BOOKS OF RtfiDIfGS - PAHO/WHO

L 

00111111 

MASTER COPY - NO TOCAR 

BOOKS OF RtfiDIfGS 

Serie Desarrollo de Recursos Humanos No. 50 

fDUCRTIOf Of OPERATIOllAL RESERRCH 

.AnD SYSTfm AnALYSIS If HERLTH 

SERVICES RDmliNlSTRRTIOI n PROGRARS 

PROGRAM OF EDUCATION IN HEALTH ADMINISTRATION 

DIVISION OF HUMAN RESOURCES & RESEARCH 

PAN AMERICAN HEALTH ORGANIZATION 

WORLD HEALTH ORGANIZATION 

J

CONTENTS 

OPERATIONAL RESEARCH & SYSTEM ANALYSIS 

Paqe No. 

INTRODUCTION .. .............................. ........ i 

FINAL REPORT ....... .................... 

List cf Articles 

Fetter, R.B., Y.Shin y cols. - Case Mix definition by 34 

diagnosis-related groups. Medical Care (l8)2 Supplement 

1-52, 1980 

Frerichs, R. y J. Prawda. - A Computer simulZation model 90 

for the control of rabies in an urban area of Colombia 

Management Science 22(4)1.41-421, 1975. 

Greenland, S., E. Watson y R. Neutra. - "The Case Control 101 

method in medical care evaluation. Med. Care 19(8),August 

1981. 

Lev, B., G. Revesz y cols. Patient flow analysis and the 108 

delivery of radiology service. Socio-Econ. Plan Sci 10. 

159-166, 1976. 

Meredith, J. Program evaluation techniques in the heaZth 116 

services. Am J. Public HeaZth 66(11): 1069-1073, 1976 

Nutting, P., G. Shorr y B. Burkhalter. Assessing the 121 

performance of medical care systems: A method and its 

application. Med. Care 19(3), March, 1981 

Reis,,,n,, A., J. Mello da Silva y J.B. Mantell. Systems and 137 

procedures of patient and information flow. Hosp Health 

Serv. Admin. Winter: 42-71, 1978 

Shuman, L.J., H Wolfe y R. Dixon Speas, Jr. "The role of 167 

operations research in regional health planning. Operations 

Research 22:234-248, 1974 

Vraciu, R. "Programming, budgeting, and control in health 181 

care organizations: the state of the art. Health Serv. Res 

14(2), Summer, 1979 

Duran, L. y A. Reisman. "Design of alternative provider 205 

team configurations: experience in both developed and 

developing countries. Technical memo 947, Department of 

Operations Research, Case Western Reserve University3 

Cleveland, Ohio, 1980

- ii - 

liancock, W., D. Magerlein y cola. - "Parametera affecting 229 

hospital occupancy and implications for facility sizing 

HeaZth Serv. Res 13(3), Pall. 1978. 

Hindl, A., N. Dierckman y cols. - "Estimating the need 243 

for additional primary care plzysicians. Health Serv. Res 

13(3), Fall, 1973 

Reisman, A., B. V. - Dean y cols. '"Physician supply and 255 

aurgical demand forecasting: a regional manpower study. 

Management Science 19(12): 1345-1354, 1973. 

Abernathy, W.J. y J. C. Herahey. "A Spatial allocation 268 

modeZ for regional health services planning. Operations 

Research 20(3):629-642, 1972 

Goldmanm J. y H.A. Knappenberger. 'THow to determine the 282 

optimum number of operating rooms. Modern Hospital 111: 

114-116, 1968 

Revelle, C., D. Bigman y cola. "Facility location: A 290 

Review of context-free and EMS models. Health Serv. 

Res. 12(22):129-146, 1977 

Brodheim, E. y G. P. Prastacos. "The Long Island blood 308 

distribution system as a prototype for regional blood 

management. Interfaces 9(5):3-20, 1979 

Duraiswamy, N.,, R. Wetton y A. Reisman. "Using computed 324 

simulation to predict ICU staffing needs. J. Nur Admin. 

Febrmary, 1981, pp. 39-44 

Harrington, M. b. "Forecasting area-wide demand for health 330 

care services: A critical review of major techniques and 

their application. Inquiry 14-254-268, 1977 

James B., Henry and Rodney, L. Roenfeldt "Cost analysis 345 

of Leasing Hospital Equipment, Inquiry, 15 No.l,pp 33-37, 

1978 

Klarman, H.E. "Application of cost-benefit analysis to the 350 

health services and the special case of technologic 

innovation. Int. J. Health Serv. 4(2):325-352, 1974 

Warner, K.E. y R. C. Hutton. "Cost benefit and coat 378 

effectiveness analysis in health care. Med. Care 18(11) 

1069-1084, 1980.

- iii - 

Paqe No. 

Jackson, M.N., J. P. LoGerfo y cole. "Elective hysterectomy 394 

a coet benefit analysis. Inquiry 15(3):275-280, 1978 ......... 

FINAL REPORT OF THE MEETING (In Spanish) ......................... 400 

List of Patiipant ........ 

.. ............................ 419

FOREWORD 

These last years, PAHO with the support of Kellogg Foundation, 

has developed a series of workshops In Health Administration for 

Latin America and the Caribbean, such as Organizational Behaviour in 

Costa Rica, Evaluation and Health Planning in Barbados, Health Economics 

and Cost Control in Brazil, Epidemiological Approach to Decision Making 

in Chile, Management and Nurse Education in Panama, Strategic Management 

(or Planning) In Brazil, and as the latest Operational Research and 

System Analysis in Venezuela. 

The materials prepared for these workshops, all in Spanish, 

have been periodically sent to all existing Administration Programs in 

the Region, including those in the Caribbean. Clearly, this Subregion 

has not benefited from those documents, and this fact has been stated 

by our colleagues in the English-speaking countries of the Caribbean. 

Final reports from these workshops and other meetings were 

published also only in Spanish in PAHO's "Revista Educaci6n Médica y 

Salud." 

During these meetings, new approaches to curriculum development, 

exchange of research information, selection of best bibliographic references, 

and the ,.nuice of articles that should be disseminated to our Region were 

discussed. This was accomplished through the input.of meeting participants. 

I have no doubt that these articles discipline oriented, are 

the tip of an iceberg of hundreds of publications which are periodically 

published all over the world. These reprints are being published in 

Spanish as books of readings for Latin American countries to support 

continuing education of faculties and post-graduate students.

In discussions about these publications with professionals 

in charge of the Administration Programs in the Caribbean, they 

expressed their frustration due to the lack of information on PAHO's 

different progr-ms in English. Bearing this in mind, the Health Care 

Administration Education Program is reproducing some original articles 

In Engllish on Operational Research and System Analysis, selected 

during its Workshop in Venezuela, 

Humberto de Moraes Novaes, M.D. 

Regional Advisor in Health Care Administration Education

Workshop .- Education in Operational Research and System Analysis 

Caracas, Venezuela 

March 8-12,1982 

I. Introduction and Background 

A fundamental component of the World Health Organization's 

goal of health for all by the year 2000 is the development and 

support o£ .ealth administration programs. To Further this ob- 

Jective in Latin America and The Caribbean, a series of workshops 

has been sponsored by the Pan American Health Organization (PARO) 

and the W.K. Kellogg Foundation. These workshops have been con- 

cerned with organizational behavior, economics and financing, 

evaluation and planning, epidemiology, and administration. This 

particular workshop has dealt with the incorporation of Opera- 

tions Research and System Analysisltechniques in training and 

continuing education programs for healtb systems and hospital 

administrators. 

This workshop has been motivated by the fact that the health 

systems decision maker in Latin America is typically a physician. 

At present decisions are generally being made based on subjective 

grounds, without the use and, in many cases, the knowledge of 

the technic,l tools of operations research and system analysis. 

It is only recently that this situation has begun to change as 

quantitative methods are being included in both public health 

and continuing education curricula. 

As late as 1970, health operations research was non-existent 

in Latin America. In 1971, PAHO sponsored a symposium on system 

Systems analysis and operations research should be broadly 

defined as encompassing those quantitative techniques of management 

science which can be used to solve real world problems. 

These techniques may range from classical industrial engineering 

to mathematical programming.

-2- 

anc.> ,s which recommended that operations research at the hospital 

level become a priority. 

Also at this time, a summer program at the Instituto Tecnoló 

gico de Estudios Superiores de Monterrey, México was initiated, in 

which Industrial Engineers students and professors worked together 

with health managers in the definition and solution of management 

hospital problems. Since that time diferents modalities of this type 

of program have been carried out in other universities of colombia, 

Costa Rica, Perú, M¿xico and Brazil. 

At present thre are about 11 programs of industrial and 

systems engineering In which students and professors have-participated 

In health services research studies and 54 academic programs that 

after some form of training in health administration. 

The specific objectives of this particular workshop were to: 

1. Exchange experiences among participants relative to the 

teaching of operations research and system analysis applied to health 

care problems. 

2. Develop a basic plan for the teaching of operations 

-.search and system analysis. 

3. Select a basic bibliography for use in updating instructors 

and health services administrators and for use in the educational process. 

4. Design a basic program for continuing education in operations 

research r-d systems analysis for updating professors, researchers and 

administrators in the health sector. 

The 19 participants at the conference represented slx Latin 

American countries, the United States and PAHO. (see Appendix 1.).

-3. 

The format of the workshop was organized to allow partici- 

pants to expound on their perceptions of the problems involved 

in the training of Latin American health administrators in 

management science techniques. This enabled the group to form 

a concensus viewpoint from which specific problems could then 

be addressed, and recommendations formulated. These concerns 

can be summarized as follows: 2 

A. Health Concerns 

1. Physicians, wbo are the health decision makers, 

do not have knowledge of the technical tools of systems analysis 

and operations research. As a result, decisions are made on 

subjective and political bases without benefit of quantitative 

methodologies and information. 

2. There is a parallel need for the development and 

utilization of information systems in approaching health systems' 

problems. At present decisions are constantly being made without 

information at both the national and local levels. 

3. Middle managers require a knowledge of descriptive 

statistics, systems analysis and evaluation techniques to im- 

prove tLcir effectiveness. 

4. Public health administrators bave viewed those 

students being trained in management science techniques as a 

threat. This has led to the deterioration of training eiforta 

in at least one country (Colombia.) 

5. Effective teaching must be geared to demonstrating 

to health administrators how to apply methodologies to their 

2 For a specific example see: Carlos Perez, Improving the Managerial 

Capabilit'y of the Colombian Health System, May, 1981, 

included in the Appendix.

-4- 

prob.ems rather than straight lecturing, However, this requires 

a l,-ge amount of preparation time and such teaching aides as 

case studies and decision making laboratories. It is further 

constrained by the problem of a lack of proficient teachers who 

can communicate effectively with students. (See number B-1) 

6. In educating administrators, particularly physicians, 

there is also the problem that, given a little knowledge obtained 

via a short course, they may assume they know a lot; that is, 

"a little knowledge is dangerous." 

7. Health institutions have not created positions for 

teams who would apply management science techniques to their 

problems. Those that do, typically have only one position which 

is not satisfactory. Consequently, health institutions have not 

yet created a demand for operations research and system analysis 

techniques. 

8. There has been a tendency by operations researchers 

to focus on the micro level problems, neglecting the more signi- 

ficant problems which must be addressed on a national scale. 

For example, .the impact of improving the efficiency of a hospital 

is insigni-,cant compared to that of improving the quality of 

water or nutritional level in the region whicb that hospital 

serves. This is particularly true in rural areas. 

B. Educational Concerns 

1. Most university professors do not have sufficient 

background in applying systems analysis and operations research 

in the health field to effectivoly mutivate and educate health 

administrative students.

-5- 

2. Programs for bringing together schools of engineer- 

-,g, business and medicine are not, a priori, guaranteed to 

succeed. 

(Conversely, at Sao Paulo, Brazil, 

a Joint program between a business school and private hospital 

has proven to be successful.) 

3. An underlying problem in the educational process 

is that schools of administration are primarily oriented to the 

private sector. They have had little experience (and often 

little interest) in the public sector. 

4. Most Latin American professors of public health 

are physicians. They have typically shown a resistance to 

the participation of other disciplines in the education process. 

This has been true in the overwhelming majority of programs in 

Latin America. 

5. The Latin American universities typically don't 

have the financial resources to keep quality faculty. The 

result is that most leave for industry, consulting or jobs in the 

USA. Thus, obtaining and maintaining human.resources may be 

the most pressing problem facing educational institutions. 

6. There are both institutional and political barriers 

which prevent a public health school from developing a formal 

relationship witb a management school. Without these formal 

relations, it is very difficult to develop any type of educa- 

tional program. This type of intransiance may not be resolved 

until those who are reiponsible for budgotary decisions are 

cognizant of the situation and the necessity for its resolution.

-6- 

As a means of addressing these concerns, the workshop 

p4rticipants formulated a series of recommendations. These 

recommendations are oriented towards the development of both 

continuing education as well as more formal education programs 

for health administrators in which the techniques of systems 

analysis and operations research are presented. In addition, 

since it is also necessary to train practitioners of those 

metbodologies who will work in the health field, these types 

of programs have also been addressed. A fundamental belief 

of the participants is that these techniques, to be successfully 

applied in solving health problems, require an interdisciplinary 

team approach. They cannot be solved by individuals, neither 

administrators nor analysts, working independently. Conse- 

quently, the education of both administrators and practitioners 

must be addressed. Further, to be most effective, these educa- 

tion programs should overlap to as great as an extent as possible. 

Thus, as students, there is an opportunity for both future 

administrators and technicians to take courses together, work 

on group projects together, and develop the appreciation of each 

other's .iscipline and capabilities that will facilitate their 

working together in the future. 

The next section of this report contains the recommendations 

of the workshop. It is followed by a section giving more speci- 

fic details of the three levels of courses (short course, one 

year certification program, two year masters) that have been 

proposed.

II. R.commendations 

-7- 

A series of 22 recommendations bas resulted from the work- 

shop. Tbese have been categorized into five subsets: 1) educa- 

tional programs and additional workshops; 2) resources and 

teaching aides; 3) information exchange programs; 4) institutional 

curricula; and, 5) general recommendations. For each recom- 

mendation those individuals and institutions who should play 

the primary role in that recommendation's implementation bave 

been noted using the following codes: 

1. PAHO (the Pan American Health Organization.) Although 

PAHO has limited financial resources, it can continue, to assume 

the important roles of facilitator and knowledge base. Speci- 

fically, it is a function of PAHO to bring together the appro- 

priate índividuals and institutions in order that the recom- 

mendations might be implemented, Part of this facilatator role 

would involve the continued development of financial resources 

from both foundations and the involved governments as required 

for program implementation. 

2. EI (Educational Institutions.) This refers to those 

educational institutions in the different participating countries 

that would either provide the proposed programs directly or 

contract with other institutions for curricula and course devel- 

opment and offering. 

3. Agencies refers to federal, regional and community 

level agencies that would provide students and funding for the 

proposed activities; erve a¡t S)ll urco or data; und provide 

specific problem areas for study.

-8- 

OR/SA (Operations Research and Systems Analysis.) 

This is directed at the participants of the workshop as well 

as selected key individuals in the various Latin American 

countries who would be involved with the implementation of 

the recommended courses and program options. 

The specific recommendations are as follows. The re- 

commended, responsible parties are indicated by ( -) after each 

recommendation. 

A. Educational Programs 

1. Three types of courses should be developed for 

teaching operations research and system analyses to Latin 

American Health Administration students and practitioners. 

These courses could be given as part of a specific program or 

in a continuing education mode. The three formats are summar- 

ized below; specific details are left for the next section. 

a. Short courses geared primarily for adminis- 

trators and other practitioners. The typical short course would 

be from 16 to 40 hours in duration and wouldV be oriented towards 

showing the jtudent how the particular methodology would be used 

in solving realistic problems of concern.: A series of 

courses should be developed since it is unrealistic to expect 

that one course would be sufficient. The courses would have no 

methematical prerequisites. That would motivate the student 

towards an understanding of the system's perspective, an appre- 

ciation of information for making decisions and recognition that

-9- 

there are technically trained individuals who could apply man- 

,ament science techniques to the solution of health systems 

problems. Similar short courses would be geared to operations 

researchers and system analysts to enable them to more effec- 

tively function in the health field. 

b. One-year certification program of approximately 

240 bourd oriented towards the practicing administrator. This 

program would be geared to giving the administrator a more 

indepth knowledge of management science techniques. However, 

it would require only mathematical proficiency in algebra as a 

prerequisite.. The course should be designed as either an even- 

ing or weekend program to enable the administrator/student to 

continue to work fulltime. 

c. A "two-year" masters program. This would be 

directed towards the administrator who wishes to acquire a 

more rigorous background in operations research and systems 

analyses. It would require a knowledge of calculus as a mathe- 

matical prerequisite. A second masters program should be 

developed to train a technically oriented individual (industrial 

engineeL, computer scientist, mathematician or physicist) to 

function as a competent health systems problem solver. Based 

on the U.S. experience, such individuals would, over a relatively 

short period (e.g., five years) assume administrative positions 

in the health field and thus become a second source of quantita- 

tively trained administrators. 

As notud, suggested detail of these programs are given 

in the next section. Specific examples of short courses are 

given in the Appendix. (PAHO,EI,OR/SA.)

-10- 

2. Operations Research and Systems Analyses studies are 

depa-ient upon data. Informed decisions cannot be made without 

valid, reliable information. However, data and information are 

typically given little attention in most programs. Workshop 

participants strongly believe that PARO should organize another 

conference on data and informatics which would complete the 

series of workshops. This latter conference would address: 

data sources, collection procedures, design and development 

of information systems, data entry procedures, editing, reduc- 

tion, data processing, quality control, collation, simple data 

analysis, networking of data systems, information exchanges aDd 

other related topics. (PAHO.) 

3. A series of directed workshops should be developed 

and held in each country. The workshops should bring together 

policy makers from government (and, where applicable, private 

industry) to interact with technical experts. These workshops 

should be aimed at addressing a specific problem of health ser- 

vices delivery. (PAHO, Agencies, El.) 

4. A follow-up workshop should be held to address the 

problem of implementation and update these recommendations in 

light of the progress made since the original workshop. Parti- 

cular emphasis should be placed upon projects instituted and 

curricula developed following the first OR/SA workshop, The 

second workshop should emphasize the problems of program/project 

implementation with particular attention paid those political, 

cultural and I'inun¡'ial constraints that effect implementation. 

Participants should include attendees from the first workshop

-11- 

and selected individuals who have been involved with the imple- 

mr' ation of operations research and systems analysis projects 

on a regional or national level. (PAHO.) 

B. Resources and Teaching Aides 

Prior to the start of the conference, two extensive 

literature searches (Spanish and English) were commissioned by 

PAHO to srve as resource material both during the workshop and 

for courses that would result from the conference's recommen- 

dations. The literature searches were primarily directed towards 

papers and documents which would provide the basis for case 

studies in the teaching of systems analysts and operations re- 

search techniques to health administrators. A secondary concern 

were papers that would acclimate technically trained individuals 

to health services delivery problems. 

The Spanish (and Portugese) literature search was conducted 

by Dr. George Kastner who located 60 relevent articles, providing 

abstracts of 43. Given the limited number of publications found, 

no screening was done; rather, all articles were included for 

use during the workshop. 

Thc 'inglish literature wasreviewed by Dr, Richard Shachtman 

with assistance from Drs. Arnold Reisman and Larry Shuman. 

Unlike the Spanish literature, computer literature searches 

identified thousands of articles concerned with health system 

analysis and operations research. 

The resultant set of 116 articles and 24 books were selected 

to encompass eight subcateguries: project evaluation, manpower 

planning and scheduling, resource allocation, demand forecasting, 

cost benefit analysis, inventory control, technology assessment

-12- 

and caLchment area analysis. Since a primary audience was to 

be adr.nistrators, the level of mathematical detail was a concern. 

Thus, articles of a purely theoretical nature were eliminated as 

were those that required a sophisticated matbematical background. 

While the more recent literature was emphasized, a number of 

earlier articles which could be considered "classics" were in- 

cluded. Also, articles that addressed issues of national priority 

were selected. These latter group included a number of articles 

concerned with the evaluation and cost benefít analysis of 

various programs. 

5. Workshop participants reviewed tbe 116 English articles 

and have selected 42 that should be translated into Spanish. 

These articles would serve as resource documents and the bases 

for case studies to be used in conjunction with the three types 

of recommended courses. (The complete list of articles and the 

courses for which they would be applicable are given in the 

Appendix, along with a classification of those articles that 

were not initially recommended for translation.) 

In addition, 14 of the 43 Spanish articles and iO documents 

previously histr'ibuted by PAHO should be included in the final 

so, of literature recommended for use by the health administra- 

tion programs. Also included should be the list of 24 English 

language books and the supplemental bibliography (articles not 

selected for translation.) This will provide the instructor witb 

a substantial set of resources. While the books and supplemental 

articles will be in English, they should be most applicable in 

the Master's programs where students will have a proficiency in 

English (at least at the reading level.) (PAHO,EI.)

-13- 

6. In order to improve the quality of instruction and 

'he relevancy of course content, PAHO should facilitate the 

development of case studies and computer software as teaching 

aides. 

a. Several case studies, ranging in complexity from 

simple versions of material presently available in the litera- 

ture to more complex cases which may require 8-16 hours of class 

and laboratory time to complete, should be developed. Problems 

chosen should be representative of those faced by Latin American 

health administrators. Source material may be provided by 

agencies within the countries themselves and/or analysts involved 

in health systems problens. The precedent for this type of case 

development has already been established in non-health related 

areas where faculty members have developed relevent cases under 

the sponsorship of private industry. Other case studies may come 

by expanding particular articles from the literature searches. 

Authors should be contacted about rewriting certain articles as 

case studies. 

b. To enhance classroom lectures (as well as certain 

case s",ldies) computer software should be developed for use in 

laboratory sessions. Sample data sets should be obtained which 

will enable students to utilize the software in order to better 

understand particular techniques and their use in problem solving. 

As part of this effort, PAHO should encourage the interchange 

of information among Latin American universities. In particular, 

abstract- or tho>st, sl.tl.wJnt prui.j.cl.s ;nd thel thl.hes< dirucQted at 

health systems problems should be circulated. Such projects 

could provide both case study and laboratory material. 

(PAHO,EI,AGENCIES,OR/SA.)

-14- 

7. An initial effort has been made during the conference 

to identify gaps in the published literature. As shown in the 

table below there are a number of areas in both the Spanish and 

English literature where participants felt deficiencies exist. 

These deficiencies should be further spelled out by representa- 

tives from Latin countries. PAHO should then commission papers 

or initiate studies which would reduce these deficiencies. 

(PAHO,OR/SA,Agencies.) 

Area 

Systems Anal. 

Informatics 

Decision Anal. 

Oper. Research 

General 

D: Deficient 

S: Satisfactory 

E: Excellent 

Status of Literature 

Type of Course 

Certificate 

Short Course Program 

Span. Eng. Span. Eng. 

D D S D 

D D D D 

D E D E 

D E D E 

D D D D 

Master's 

Program 

Span. Eng. 

D D 

D D 

D E 

E E 

D D 

8. P-r.icipants and educational institutions should consider 

collaborating in the development of applicable software for 

selected microcomputers that could'be used both in an educational 

environment as well as for real problem solving. This would 

require that participants work on compatible machines, using 

common languages. The recent advances in microprocessor tech- 

nology will enable users to obtain computers capable of solving 

complex problems at relatively low cost. The cost of software

-15- 

cnve!opment will typically be more than the machine thus under- 

,.ng the importance of collaborative efforts (EI,OR/SA.) 

9. There is considerable overlap among the various work- 

shops that have been or will be sponsored by PARO. PAHO should 

continue to provide future workshops with information from 

previous caes, particularly when such subject matter is relevent 

to more than one workshop. (PAHO.) 

10. PAHO should investigate the feasibility of obtaining 

funds for the publication of a book in Spanisb, concentrating 

on operations research applied to health services problems and 

directed at Latin American audiences. Problems of regional and 

national concerns should be emphasized. Such a book might con- 

tain some of the case studies noted above. (PAHO, OR/SA.) 

11. PAHO should facilitate the preparation of an article 

describing the workshop process to appear in the appropriate 

professional public health journal . Such an article would help 

to encourage more activities of this type, and improve the 

communication among scientists of the different countries. (PAHO.) 

12. PAHO should facilitate the dissemination of relevent 

working ,)apers and theses prepared in the U.S. to Latin American 

educators. In particular, the Catalog of Hospital Management 

Engineering Technical Papers published by the clearinghouse for 

Hospital Management Engineering, American Hospital Association; 

selected citations from Hospital Management Abstracts, and 

brief summaries or abstracts of Master's theses related to 

health services problhns (possibly o>btined thrQough the Associa- 

tion of University Programs in Hospital Administration) should

-16- 

pericaiclly ba circulated to Latin American health administration 

prog. .s. (PAHO,EI.) 

C. Information Exchange Programs 

13. PAHO should arrange for the routine surveying of 

the relevent English and Spanisb health services literature. 

It should monitor major national and international meetings of 

the operations research societies and public health associa- 

tions. This information should be distributed to Latin American 

health administration programs on a scheduled basis. Also 

included should be a listing of applicable short courses and 

professional meetings in both the U.S. and Latin American counties. 

(PAHO,OR/SA.) 

14. The information exchange function of the workshops 

could be improved by having selected participants give formal 

presentations of specific projects on research activities. 

Such presentations could be in a seminar format. Such presenta- 

tions might be scheduled prior to the actual start of the work- 

shop. (PAHO.) 

15.,. In order to most eff.iciently introduce operations 

research anr systems analyses methodologies to health adminis- 

trators while developing a cadre of sophisticated technical 

analysts, primary emphasis should be placed on developing the 

short courses and the two year master's program. The develop- 

ment of the certification program, while important, should re- 

ceive less priority. As noted, a number of short courses should 

be developed with a varJity of: foriits and o>ntent to atiisfy 

identified target populations. (EI,OR/SA.)

-17- 

16. All participants sbould promote the development 

of -dalth operations research and systems analyses programs 

within industrial engineering, medicine, management and health 

administration departments/schools. An emphasis should be 

placed on instituting joint programs among the different dis- 

ciplines, p&rticularly when only one university is involved. 

(EI,OR/SA.) 

17. An important component of both the certificate and 

masters prograns should be the offering of at least one course 

in health operations research jointly with engineering, public 

health and/or the medical school. Such a course should involve 

at least one project in which students must function as an inter- 

disciplinary team. (EI.) 

18. While Latin American universities will become the 

predominant trainers of master'sstudents, U.S. and Canadian 

universities will remain the primary training facilities for 

Ph.D.s. Consequently, it is in the best interest of Latin Ameri- 

can universities to develop more formalized institutional arrange- 

ments with the appropriate U.S. schools and programs. Latin 

American ,liiversities should strive to assure that such U.S. 

programs remain relevent to and considerate of the educational 

needs of the Latin students. If possible, Ph.D. dissertations 

should be directed at problem areas that are also of importance 

in Latin America. (EI.) 

19. As health administration programs continue to 

develop in Latin America, it is important that strong, formal 

ties are developed with both industrial engineering/operations 

research and management programs. If such ties are not developed,

-18- 

manatement science techniques will not be effectively integrated 

int> che health administration curricula. (EI,Agencies.) 

E. General Recommendations 

20. PAHO can further facilitate the development of 

research projects and collaborative programs between U.S. and 

Latin America by providing information to participants and 

other key individuals on priority problem areas, countries with 

favorable political climate and potential funding sources. (PAHO, 

EI,Agencies.) 

21. It cannot be overemphasized that operations research 

and system analyses require interdisciplinary teams to suc- 

cessfully solve health services delivery problems. All programs 

and resultant curricula should carefully reflect this fact in 

their development and delivery. (PAHO,EI,Agencies,OR/SA.) 

22. In order for operations research studies to lead to 

successful implementations, it is important that the decision 

maker be integrated into the planning process. This is why 

health administrators, at all levels, must be cognizant of the 

value and use of management science techniques. (PAHO,EI,Agencies, 

OR/SA.) 

III. Specific Examples of Courses 

Three types of courses were considered by the conference 

participants: short courses, certificate programs and full 

masters' degree programs, It was felt that the initial emphasis 

should be placed upon the development of short courses, fol- 

lowed by the masters' degree prugrums und then the certificate 

programs.

A. Short Courses 

-19- 

Short courses, by their nature, will typically be 16-40 

hours in duration and either given intensively during a two 

to five day period, or spread over several weeks or one semester. 

Since it is impossible to cover much more than introductory 

material in one short course, the offering institution should 

design a series of related courses which may build upon one 

another and provide students with a more complete educational 

experience. In all instances, an, attempt should be made to 

include problems for students to solve, both individually and 

in groups. These "laboratory" sessions may also enable students 

to have some computer interaction with either a game; e.g., 

simulation of a hospital or regional health system, or a simple 

resource allocation or forecasting model. 

Examples of short courses include: 

1. Stand alone course for hospital and health systems 

administrators. Typically, no mathematics background required, 

although some knowledge would be helpful. Course would present 

an introduction to the techniques of operations research and 

systems analysis with particular emphasis on understanding what 

these techniques can and cannot do. Students will gain a know- 

ledge of the OR approach, cost/benefit analysis, information 

systems, data sources, interdisciplinary teams.* Students will 

learn how to recognize problems that could be solved by these 

methodologies, how to estimate the approximate time and effort 

*These are suggestad topics to be included in the optimal course. 

It is recognized that no one course may sufficiently cover all 

of them.

-20- 

required for solution, how to recognize potential political 

barriers to implementation, and where to seek help in conducting 

such a study. Specific topics included in the course may be: 

"toy" or simplified decision problems, resource allocation 

problems, planning (PERT/CPM) and discounting. A team project 

in which an unstructured problem is formulated should also be 

included. 

2. A stand alone course for industrial engineers, 

operations research, or other technical disciplines who will 

become involved in solving health delivery system problems. 

Students will become familiar with health systems and the types 

of problems typically encountered. Solution procedures for these 

problems will be presented and discussed. 

Prototype courses: 

a. 40 hour course - Introduction to Operations Research 

and System Analysis for health Professionals. 

1. Introduction to the concepts of system analysis, 

planning, evaluation and decision making. This should include 

an overview, conceptual model of the systems approach as applied 

to health aelivery problems. This framework should then be used 

,.l referred back to as the topics are presented in more depth. 

2. Definition of a decision system, types of deci- 

sions that are made, types of models. Discussion of the levels 

of decision making. 

1. Introduction to modeling; the art of modeling; 

examplle m.,debl s ol' rt.le(v¢.nt heu llt.h syt.il.s pr

-21- 

4. Presentation of a case to be developed during 

the course or a mini-project to be carried out by participants. 

Emphasis should be on problem formulation. Solution of part 

of the problem may be achieved by "hands on" use of canned 

computer programs during laboratory session. 

5. Introduction to planning and evaluation approaches 

including flow charting, PERT (Program Evaluation and Review 

Technique), CPM (Critical Path Method), Gantt charts. Include 

actual examples with computer use optional. Also include dis- 

cussion of process and outcome evaluations with appropriate 

examples. 

6. Introduction to resource allocation models. 

Presentation of simple linear programming case study. Illus- 

tration of graphical solution technique. Discussion of assump- 

tions in problem formulation. Computer usage desirable. 

7. Introduction to data sources and data collection. 

Should review data management and processing concerns, data 

reliability issues, information systems and potential problem 

areas. 

8. Preparation for cost-benefit analysis and capital 

budgeting. Discounting (net present value), project costing, 

simple cost-benefit analysis example, definition and development 

of internal rate of return, and benefit/cost ratio should all 

be covered. 

9. Cost-Benefit Analysis. Definition of incremental 

benefits and costs; imlpact - incidence matrix; complete cost- 

benefit example, using hand calculations. Computer example 

desirable.

-22- 

10. Marketing health services. Basic definitions, 

distilction between selling and marketing; strategic planning; 

specific health applications. 

11. Introduction to uncertainty. Basic introduction 

to probability concepts, events, probability assessment and 

estimation. Concepts underlying Bayes rule, distributions, 

expectations. 

12. Decision Analysis including decision trees, 

alternatives, outcomes, experiments, values and their construc- 

tion. Analysis and interpretation of decision problems, simpli- 

fied case studies. Laboratory session should include a full 

decision analysis case. 

13. Forecasting and prediction techniques. Concepts 

of moving average, exponentialsmoothing, simple least squares. 

Interpretation of simple regression. Case study example; use 

of canned computer program desirable. 

14. Inventory and material management, Relevent 

costs in inventory system. Basic inventory models -- lot size 

and order le"ol. 

15. Implementation of operations research/systems 

anaiysis studies. Interdisciplinary teams; working with 

decision makers; validation of study results; roadblocks to 

implementation. 

b. 8-16 hour courses for health professionals* 

*These courses utilize the following text: Arnold Reisman, 

Systems Analysis in Health Care Delivery; Lexington, Massachusetts; 

Lexington Books, 1979.

Topic 

Introduction to 

Systems Analysis 

Systems Concepts 

Graphical Techniques 

Algebraic Techniques 

Data Collection 

Formal Aides to 

Creativity 

Art of Systems 

Analysis Strategies 

for Implementation 

One Day 

Course 

1 hr. 

1 

2 

1 

.5 

.5 

-23- 

Two Day 

Course 

1 hr. 

1 

5 

3 

1 

1.5 

Two Day 

Course 

With Math 

1 hr. 

c. Quantitative Methods Applied to the Administration 

of health care. (45 hours) 

This course is designed for health administrators, 

typically physicians and nurses, who bave limited mathematical 

and statistical background. The course, which is given at 

the University of Sao Paulo, Brazil, has been revised several 

times. 

1. Elements of Mathematics - cartesian coordinates, 

linear equations, elementary graphical solution procedures. 

2. Elements of Statistical Inference - set tbeory, 

probabilities, Bayesian probability, discrete distributions, 

continuous distributions, simple linear regression. 

3. Elements of Operations Research - critical path 

method, linear programming, queueing. 

1 

5 

8 

3 

1 

1.5

-24- 

d. Methods Improvements Techniques in Hospitals. 

This is designed as a stand alone course for bospital 

department heads and administrative staffs. Students must define 

and solve a real problem effecting their department. (40 hours) 

1. Introduction to Methods Improvement - scientific 

management; systematic approach to problem solving. 

2. Philosophy of Work Simplification - extension 

to hospital setting; problem identification; use of personnel 

in work simplification; PERT. 

3. Work Session in Problem Definition 

4. Process Analysis - definitions and symbolic nota- 

tion; flow diagrams; evaluation of flow diagrams and subsequent 

action. 

5. Operations Analysis - operation charting; motion 

economy; micro-motion-video motion techniques. 

6. Work Simplification - systematic examination of 

problems; what to look for; what to do; layout; automated 

systems. 

7. Work Session on Problem Analysis 

b. Work Sampling - sample size determination; proce- 

d,,re; interpretation of results. 

9. Other Data Gathering Techniques - time study; 

link analysis; questionnaires. 

10. Work Session on Measurement 

11. Cost Analysis - cost measures; cost estimates; 

non-monetary criteria; comparing methods.

-25- 

12. Elementary Statistics ior Analyzing Differences - 

testing results; interpretation of data. 

Studies 

13. Work Session on Evaluating Results 

14. Implementing the Results of Methods'Improvement 

15. Work Session: Reports and Critiques of Projects 

e. Health Systems Research Seminar 

This is a 45 hour course designed for both health 

administrative students and industrial engineering/operations 

research students who want to learn about the actual application 

of quantitative methods to health services problems. In addi- 

tion to classroom discussions, students form small, interdisci- 

plinary teams and solve an actual hospital or health system 

problem. Discussion topics have included: 

Research in Health 

microcosting. 

approach. 

Control 

1. Introduction to Industrial Engineering/Operations 

2. Evaluation of a Unit Dose Drug Distribution System 

3. Hospital Cost Models - step down cost allocation; 

4. Incentive Reimbursement - an industrial engineering 

5. 

6. 

7. 

8. 

Utilization Review Models 

Nurse Staffing Methodologies 

Predictive Reimbursement Models 

Grouping Hospitals for Reimbursement and Cost 

9. Operating Room Scheduling and Admissions System 

10. Linear Programming Applications: menu planning; 

theraputic radiology

C' ... cs. 

-26- 

11. Regional Planning - location of a network of 

12. Emergency Medical Services - information systems; 

simulation of EMS systems; evaluation of paramedic performance. 

13. Health Economic Analyses and Models 

14. Measuring the Effectiveness of Care 

f. Information Systems and the Evaluation of Quality of 

Health 

This 30 hour course is directed towards health services 

administrators and planners. It presents an overview of the use 

of information and data processing technology for control and 

utilization for administration and planning of health systems. 

1. Origin, nature and utilization of information in 

health systems; medical documentation; administrative records. 

2. Information for decision processes; importance 

and objectives of educating health personnel in informatics. 

3. Objectives of the users of information; functional 

levels; levels of centralization or distribution of information 

systems; impact of information on organizations. 

tration and planning. 

4.'Needs of administratorsJ information for adminis- 

5. Systems analysis; project, role of systems analysis; 

planning and control of projects; collection and distribution of 

data; process for implementation of information systems. 

6. Computers and automation: basic concepts, electronic 

data pisc:essing; files and logical structures;, data banks; com- 

puters and their components; data processing centers. 

7. Options in data processing: centralized, decen- 

tralized, and distributed systems.

-27- 

8. Applications of data processing to health systems; 

advantages and disadvantages for the use of computers in medicine. 

9. Information systems for evaluation and control; 

methods for evaluating the utilization and quality of services; 

medical auditing. 

g. Quantitative and Analytic Methods ior Health 

Administration* 

This 40 hour course seeks to: 1) create an apprecia- 

tion for, and an understanding of, the basic processes of sys- 

tems analysis modeling, decision making and control; 2) identify 

the intimate roles these processes play in the activities and 

responsibilities of the health administrator. Some of the more 

fundamental decision/control models and methodologies of systems 

analysis, management science, and economics are discussed. 

Strengths and limitations of these models for more effective 

health decision making are Stressed. Prerequisites for this 

course include college algebra, basic linear algebra and ele- 

mentary probability and statistics. Topics include: 

1. Overview of the Systems Analysis and Operations 

Researc.,/"!anagement Science Health-Related Literature 

2. Systems Approach and Systems Analysis - A Pro- 

logue for Health Care Decision Making 

3. Quantitative Models and Their Role in Health Care 

Decision Making and Control 

4. Simple Deterministic Decision Models: Inventory 

and Project Management 

*Course developed by Barnett R. Parker, Ph.D., School of Public 

Health, University of North Carolina, Chapel Hill, N.C.

cal ~Pogramming 

Sensitivity Analysis 

-28- 

5. Complex Deterministic Decision Models: Mathemati- 

6. Solving Linear Programming Problems by Computer; 

7. Generation of Data Appropriate to the Bealth Care 

Decision Making Task: Techniques of Measurement 

8. Goal Programming and Integer Programming Applica- 

tions to the Health Sector 

9. Cost-Benefit/Cost-Effectiveness Analysis: A fun- 

damental Mode of Thinking in Health Care Decision Making 

10. Feedback, Control and Program Evaluation for More 

Effective Health Care Management 

11. Guest Speakers - Special Topics 

B. Certificate Programs 

This program would be directed towards the working 

health administrator who is seeking more indepth knowledge than 

is provided through a series of short courses. It is also for 

the individual who is seeking a career change. Middle level 

managers would take this program as one step in advancing to a 

higher management position. The program would typically be 240 

hours. It would require a mathematical background through algebra. 

The graduate of the program should be able to formulate and 

solve simple quantitative problems (linear programming and queu- 

ing); perform an elementary cost-benefit:analysis; collect and 

reduce daita (descriplive statitl.i's); use "canned" computer 

programs for analysis; formulate (but not solve) more complex

-29- 

problems. The graduate should have a basic understanding of the 

couts involved in providing health services. The interdiscipline 

team concept should be stressed. A team proJect of 2-3 month's 

duration should be included. 

Courses in the program should include an introduction to 

quantitative methods (e.g., short course c.); a more rigorous, 

course in applying quantitative metbods (e.g., short course e. 

or g.); an introductory course in bealth economics; an introduc- 

tory course in biostatistics; and a course in information systems. 

Optional courses to be considered are: theory of health planning 

and evaluation; quantitative methods of planning and evaluation; 

and health organization behavior. 

C. Masters' Program 

Two forms of Masters' programs were discussed. The 

first would be directed towards hospital and health systems 

administrators who required more extensive educational training 

than offered under a certificate program. While students might 

be fulltime administrators, they would be expected to spend the 

necessary time in residence at an academic institution. The 

mathematical background should include college mathematics through 

calculus. This program would include all the courses listed 

under the certificate program, Also included would be an advanced 

course in statistics, a second level course in operations research 

methodologies (probabiliitic models), and an introductory course 

in econometrics. Students would carry out two or three team 

projects during the program. Thebe projects could be done as part 

of an interdisciplinary course (e.g., short course e.).

-30- 

The second type of Masters' program would be for engineer- 

in 6 andergraduates who wanted to solve health and hospital sys- 

tems problems. The program would provide a strong emphasis in 

operations research methodologies. In addition, students sbould 

become familiar with cost accounting concepts, economic prin- 

ciples and information systems. Graduates of this program would 

be expected to perform a number of the following functions: 

1. Solve linear programming and inventory problems; 

2. forumulate and analyze stochastic models; 

3. perform indepth statistical analyses (ANOVA and mul- 

tiple regression); 

4. design data collection studies; 

5. program in a high level computer language (FORTRAN, 

Pascal or PL-1); 

6. design program evaluation; 

7. conduct computer simulations; 

8. write technical reports; 

9. interact with decision makers and work as part of an 

interdisciplinary team; 

10. take unstructured problem, abstract it, formulate a 

mathematical model, obtain solutions and implement the best, 

acceptable solution; 

11. conduct long range planning studies. 

IV. Articles to Translate Into Spanish 

As b¡oted, an extensive literature search was prepared prior 

to the start of the conference. From this list, a total of 41 

articles were recommended for translation. These are summarized

-31- 

Lblo-w. Also given is that type of program (short course, cer- 

-ficate, masters') for which the article would be most bene- 

ficial. 

Senior Author Year Published Applicable Program 

Berkson 

Escudero 

Fetter 

Frerichs 

Greenland 

Hartunian 

Lev 

Meredith 

Nutting 

O'Connor 

Reisman 

Shoenbaum 

Shuman 

Vracin 

Duran 

Hancock 

Hindle 

Reisman 

Abernathy 

Goldman 

ReVelle 

Brodheim 

Duraiswany 

Harrington 

Centerwall 

Couch 

Eisenbe.¿ 

Henry 

Klarman 

McGregor 

Muller 

Schwartz 

Willems 

Warner 

Kendall 

Jackson 

Evans 

McNeJ1( 2 ) 

Nemhauser 

Parker 

1. Optional 

2. Series of three 

1979 

1980 

1980 

1975 

1981 

1980 

1976 

1976 

1981 

1972 

1978 

1976 

1974 

1979 

1980 

1978 

1978 

1973 

1972 

1968 

1977 

1979 

1981 

1977 

1978 

1981 

1978 

1978 

1974 

1978 

1980 

1979 

1980 

1980 

1980 

1978 

1981 

1975 

1975 

1975 

Masters 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X X 

X 

X 

X 

X 

X 

X 

xX 

X· 

X 

x 

xX xxxxx 

X 

X 

X 

X 

X 

xX xX 

xX 

xX 

xX xX 

xX 

articles in NEMJ 293:211-226 

Certificaté 

x 

X 

X(1) 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X, 

X 

X 

X 

X 

X 

X 

X 

X 

8hort 

Course 

X 

X 

X 

X 

X 

X 

xX 

X 

X 

X 

X 

x( 1 ) 

x 

X 

X 

X 

x() 

X 

x(1) 

X 

X 

X 

X 

X 

x'

-32- 

!n addition to these works in English, it was recommended 

that eAHO also distribute a number of the works already in 

Spanish. Included in this second list are several papers that 

have been published by PAHO. It was recommended that all of these 

PAIO publications be included in any distribution of papers. 

Recommended Spanish articles are: 

Senior Author 

Year Published 

Ackoff 1974 

Barrenechea 1978 

Dunia 1976 

Miembros Facultad 

de Ingenieria 1974 

Grundy,Reinke 1974 

Marin 1978 

Novaro 1973 

Rodrigues 1978 

Schmidt (p.47) 1980 

Schmidt 1981 

Schmidt (p.50) '1981 

Instituto Monterrey 1976 

Applicable Program 

SIhort 

Masters Certificate Ciourse 

X 

X 

X 

X 

X 

X 

X 

t 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X 

X

- 33 -' 

-Case Mix Definition by Diagnosis-Related Groups 

ROBERT B. FETTER, YOUNGSOO SHIN, JEAN L. FREEMIAN, 

RICHARD F. AVERILL AND JOHN D. THOMPSON

- 34 - 

Case Mix Definition by Diagnosis-Related Groups 

ROBERT B. FETTER,* YOUNGSOO SHIN,. JEAN L. FREEMAN,t 

RICHARD F. AVERILL,§ AND JOHN D. THOMPSON I 

Froml tite Centerfor for Healtl Stulies, 

lnstitlio, jbr Social ail Iu lie! .Stdies. 

Yale Untivcrsitu 

, Professor. School of Organization and Management, atad Institution for Social and Policv Studies. 

f Assistant Professor, Department of Preventioe Medicine, College of Medicine, Seoul National University, 

$eoul, Korea. 

Assistant in Research, Centerfor Healthl Studies, Institution for Social ard Polcy .Studies. 

A4 Asociate Directorfor Research, Health Management Program, School of Organization and Managemnent. 

! Profeasor of Public dealth, School of Medicine, and Institution for Social and Policy Studies. 

The research reported herein was performed pursuant to Contracta No. HCFA 600-75,0180 and No. HCFA 

77-0043.from the Health Care Financing Administration, Department of Health, Education and Welfare. 

~Iblcation of this monograph was supported by Contract No. 600-7540180, from the Health Care Financing 

;dmintstration, Department of Health, Education and Welfare. The opinions and conclusions expressed hefan 

are souely those of the authors and should not be construed ao representing the opinions or policy of any 

ncg of the United States government. Copies of a more detailed description of this research are available by 

tíilng to Al Esposito, Director, Division of Hospital Experimentation. Office of Demonstrations and Evaluan. 

ffe of Researcl. Demostrations. 

a;ud Staristics. It¿clrt Car;' Fiarswinag Adtlrnisltratio,, Rolmoai E16. 

Meadows Buildhig,. 6401 SecuritSV Blvd., Bualtimore. MD 21235. Requests should be made for AUTOGRP 

,latient Classfication Scheme atad Diagnosis Related Grmlps (DRGs). Health Care Financing Grants and 

V ontracts Report Series, HEW Publication No. (HCFA) 03011 9-79. Inquiries regarding the software aeailable 

íor analyses with AUTOGRP and DRCs should be made in writing to Ma. Marla DeMusis, SOM, 12 Prospect 

'tce, New Haoen, CT 06520.

- 35 - 

Acknowledgment 

The authors wish to acknowledge the support, insights, and cooperation during this 

research of Al Esposito, Project Olficer, Health Care Financing Administration, Department 

of Health, Education and Welfare. The successful completion of this research 

relied heavily on the programming skills and patience of Enes Elia, Senior Systems 

Prograromer, and the excellent administrative support of Ann Palmeri and Maria De- 

Musis. The assistance of the New Jersey Department of Health, especially Michael J. 

Kalibon in preparing the description of the New Jersey prospective reimbursement 

system is greatly appreciated. The material presented on the case mix approach to 

hospi:1I regional planning is based on the current doctoral research of Robert Chemow. 

The authors are indebted to the many physicians who, dunng the years of development 

of the DRCs, participated in their formuluation and evaluation. 

The research reported here represents the current stage of development of the 

Diagnosis.Related Groups. The initial work on the approach and associated technology 

for patient classification began nearly a decade ago. Significant contributions were made 

during the course of this research by Donald C. Riedel, Ronald E. Milis, Lesley Mills 

and Phyllis Pallett. 

Supplement to Medical Care Volume 18, Number 2 

ilEDICAL CAPE (ISSN 00257079) IspublishedmonthiybyJ. B.LippincottCompny. East Washington Square. Philadelphia. 

A 19105. Subsciption address: Joumai Fulfillment Deamnent. UippincontHap . 2350 Virginia Avenue. Hagersntwn. MD 

1740. Subcriptiion ates: United Stater and U.S. Possessions. one yer, $36.00; special price US. students. 20.00; all other 

untriesexceptJapanone year, 42..00. Single copies $5.00. Subscriptionorders and correspondencecncerning subscriptions in 

JabPon should be ent to Woodbell Scope. 1 1-1 Shoto 2-chome. Shibuya-ku. Tokyo 150. Japan. Printed in U.S.A. Copyright() 

!980by J. 8. LLppinco« Company. Second clas postage paid at Hagerstown. Maryland, and additional mailing ofrices.

- 36 - 

Prsfact 

DURING THE PAST DECADE, health care financing researchers have sought to 

develop equitable methods to constrain the rate of increase in health care 

expenditures. Tle ced l for a stccessfidt national Iiospital cost containment 

program has been highlighted as a major step toward overall health care cost 

containment and a comprehensive national health plan. 

In early efforts to compare and control hospital costs, researchers calculated 

product costs by unit of service, such as lab tests, rad;ology tests, or days of 

routine heoel service. Several incentive schemes were devised to encourage 

hospital effic iency and low unit cost. Experience has shown that focusing on unit 

cost alone encourages increased length of stay and ancillary utilization and 

argues that attenttioi shloutld ie li>ciisetd oni medlical-practice I);itterns. 

Recently, much attention has been ftcused on the cost per patient stay or per 

case treated. Shifting to a single hospital product required that techniques be 

developed to adjuist for variations in hospitals' patient or case mix. 

To group hospitals with a similar case mix, researchers hlave generally concentrated 

on proxy measures, such as the mix of services or t:fcilities available in each 

hospital. While these approaches are relatively easy to calculate, thféy normally 

do not address either the extent or the use of a hospital's available services or 

facilities. 

What is required is a classification scheme that is both manageable in terms of 

the number of case types defined and reasonable in terms of the variation in 

resources needed to treat each case type. This should permit direct measurement 

of a hospital's case mix. 

The development of the Diagnosis-Related Groups (DRGs) represents a significant 

step in case mix measurement and application for reimbursement purposes. 

DRGs classitfy 383 types of cases encountered in the hospital acute-care 

setting. Each DRG represents a class of patients requiring similar hospital 

services. Since DRGs are medically meaningful, they help provide a common 

basis for comparing cost effectiveness and quality of care delivered. DRGs also 

have the potential to assist the ho.¿,ital administrator as he manages his institution 

and communicates with the medical staff. 

Many PSROs have already explored the use of DRGs to review length-of-stay 

and treatment patterns. New York and Maryland have incorporated the DRG 

concept into their hospital cost containment programs and both New Jersey and 

Georgia ,tili soon be incorporating DRG methodologies in new hospital cost 

containment programs. At the federal level, DRGs are being considered for 

incorporation in new reimbursement procedures for acute care hospitals. Given 

the current emphasis on hospital cost containment, the development of DRGs is 

both an important and timely advancement of the health Ce,'e financing field. 

JAMES M. KAPLE 

Acting Director 

Office of Research, Denmonstrations aund Statistics 

Health Care Financing Administration 

Department of Health, Education, and Welfare

ML ICAL CAE1 

Februurl 1980, Vol. XVIII, No. 2, Supplement 

THE MAJOR FUNCTION of a hospital as an 

acute-care :npatient facility is to provide 

the diag.lostic and therapeutic services required 

by physicians in the clinical 

management of t'neir patients. In addition, 

the hospital also makes available certain 

hotel and social services such as meals, 

laundry and counseling, to patients while 

they are residents at the institution. Considered 

as an economic enti.y, the l: spital's 

outputs are the specific services it provides 

in temis of hours of nursing care, 

medications and laboratory tests. Its inputs 

are the labor, material and equipment used 

in the provision of these services, Each 

patient receives a specific set of outputs or 

services which is referred to here as a 

product of the hospital. 

For example, consider a patient under 

treatment for uncomplicated pneumonia. 

His hospital stay may last 6 days. Depending 

on his treatment process, he coulld receive 

several chest roentgenograms, a 

sputum Gram stain and culture, a blood 

culture, several complete blood counts, a 

urinalysis, routine blood chemistries, 2 

days of oxygen therapy, 2 days of IV fluids, 

2 days of IV penicillin followed by 4 days of 

IM penicillin and a supply of oral penicillin 

to b, .. !-"n at home. During his stay, he 

will a:so require various amounts and 

levels of nursing care. This set of outputs 

constitutes a product ordered by the physician 

and provided by the hospital to the 

patient in addition to meals and other hotel 

services. Each department within the facility 

consumes a certain amount of inputs in 

terms of standard labor, material and overhead, 

depending on the output produced. 

Por the radiology department, this would 

include the films, the technician time, and 

a portion of the departmental overhead ex- 

- 37 - 

1. Introduction 

penses for the equipment and space 

utilized in the production of the various 

types of roentgenograms. 

Since individual patients receive different 

amounts and types of services, the hospital 

may be viewed as a multiproduct firm 

with a product line that in theory is as extensive 

as the number of patients it serves. 

The particular product provided each patient 

is depeúdent upon his condition as 

we 11 as the treatment process he undergoes 

during his stay. For example, in most instances 

an ulcer patient with a major surgical 

procedure such as an exploratory 

laparotomy or gastric resection requíies 

more units of physician and nursing time, 

more medication, more ancillary services 

and remains hospitalized at least 10 days 

longer than a woman admitted for a normal 

delivery. The formér case is generally referred 

to as more complex since it rece;ves 

mnore of the institution's outputs in ternis of 

nursing time, meals, laboratory tests, and 

special services than the latter case. During 

any specified time period, an institution 

admits and discharges a variety of 

cases representing different levels of complexity. 

The relative proportions of the different 

types of cases the hospital treats are 

collectively referred to as its case mix. 

Historically, patient-days of care and the 

number of admissions or discharges have 

been used to describe hospital output 

while cost per patient-day, per cent occupancy 

and mortality rate have been used to 

evaluate hospital performance. Such 

aggregate measures presented out of context 

of the types of cases treated by an institution 

and their relative complexity is 

not particularly useful infonnation to hospital 

management for internal assessment 

of efficiency and effectiveness nor to reg-

CASF MIX DEFINITION BY DRG 

- 38 - 

ulatory agencies for inter-institutional 

comparisons. In particular, the ovutcome 

ancd cost of the individual patient care 

processes that give rise to the hospital 

products cannot be adequately determined. 

As a cose(ltence, admiiiiiistrators and 

clinicians caninot ascertain the quality aud 

cost iimplications of the treatment plans 

practiced withi.. the insitution nor can regulators 

assess the impact of alternative 

mietlhods aross institutions. 

In order to evaluate, conmpare, and provide 

relevant feedback regarding hospital 

perl>onnaiice, it is necessary to idcntity the 

specific products that institutions -,rovide. 

As defiled above, a hospital product is a set 

of services provided to a patient as part of 

the treatment process coiltrolled by his 

clinician. While each individual patient 

adlmitted to an institultion is unique, he has 

certain demographic, diagnostic and 

therapeutic attributes in common with 

other patients that determine the type and 

level of services he receives. If these classes 

of patients with the same clinical attributes 

and similar processes of care can be 

identified, then the framework within 

which to aggregate patients into case types 

is established. Moreover, if these classes 

cover the entire range of inpatients in the 

acutte-care setting, then collectively they 

constitute a classification scheme that provides 

a means for examining the products 

of the hospital, since patients within each. 

class are exr-cted to receive a similar 

product. 

The means of defining hospital case mix 

for this purpose is the construction and ap- 

Flication of a classification scheme comprised 

of subgroups of patients possessing 

similar clinical attributes and output utilization 

patternls. This itnvolves relating the 

demographic, diagnostic and therapeutic 

characteristics of patients to the output 

they are provided so that cases are c"rerentiated 

by only those variables related to 

lihe c

X;' VIII, No. 2. Suippllce,t 

- 39 - 

,'lirst and second sets of patients incurred 

Ily 13 and 30 per cent, respectively, of the 

'tosts incurred by the latter. Thus, the 

:'type of surgery performed and the pres- 

,ence or type Jfsecondary diagnosis make a 

.considerable difference in the hospital 

!!product provided to the patient. Foi .ais 

reason, patient case types determined 

solely by primary diagnosis is not acceptable 

for use in defining hospital case mix, 

since it is not precise enough to account for 

either the condition of the patient or the 

complexity of the treatment protocol. 

Another patient classification scheme in 

'wide use is that developed by the Professional 

Activity Study (PAS) of the Comnmission 

on Professional and Hospital Activities 

(CPHA), which publishes tables of 

length-of-stay statistics based on data from 

'jarticipating hospitals.' These tables present 

summaries on the basis of primary 

diagnosis, the presence of additional diag- 

An>ses, the presence of any surgeries, and 

'.age. Specifically, the PAS classification di- 

.vides all possible prinimary diagnoses into 

.349 mutually exclusive major diagnostic 

categories. Each of these major diagnostic 

categories is then divided on the presence 

: or absence of a secondary diagnosis, pres- 

; ence or absence of any surgery and 5 age 

.categories (0-19, 20-34, 35-49, 50-64, 

! 65+).tf This results in 20 subcategories for 

;; each of L ' 49 major diagnostic categories 

V~for a to.al of nearly 7,000 patient classes. 

iThe information is currently used exteni' 

sively by Professional Standards Review 

'Organizations (PSROs) in setting lengthi.;of-stay 

checkpoints as part of their concur- 

.'rent review process. 

5 Although the PAS scheme takes into ac- 

',count additional patient attributes such as 

,secondary diagnosis, age and information 

:related to the treatment process, such as 

operation status, these attributes still may 

f For major categories 342-349, pertaining exclu- 

'sively to newboms, birthweight is used in place of 

age. 

INTRODUCTION 

be insufficient in certain instances to describe 

adequately a case in terms of output 

utilization. Refer to the example described 

above were the specific type of operation 

differentiated one set of surgical ulcer patients 

from another. Thus, because of its 

uniform structure throughout all 349 diagnosis 

groups, the PAS scheme tends to 

overspecify in some diagnostic categories 

where the extra variables may not be partictilarly 

relevant as regards uitilization of 

t:acilities and underspecify in others where 

more precise information is required. 

As an alternative patient classification 

scheme, the Diagnosis-Related Groups 

(DRGs) have been constructed, based on a 

procedure referred to as the significant attribute 

method. The fuindamental piurpose 

of the DRG approach is to identifv in the 

hospital acute-care setting a set of case 

types, each representing a class of patients 

with similar processes of care and a predictable 

package of services (or product) 

from an institution. Using this Ipproach, 

the entire range of diagnostic codes was 

initially divided into broad disease areas, 

such as Diseases of the Eye, Diseases of 

the Ear, Cerebrovascular Diseases, and Infectious 

Diseases. Each of these categories 

was further subdivided into groups based 

on values for those variables that demonstrated 

an effect in predicting output as 

measured by length of hospital stay. It was 

decided to use length of stay (LOS) as opposed 

to some other output measure since 

it is an important indicator of utilization, as 

well as being easily available, well standardized, 

and reliable. The use of LOS and 

its relation to other measures is discussed 

further in the next section. The 383 

DRGs that resulted from this process are 

interpretable fromi a medical perspective 

as well as similar with respect to their patterns 

of length of stay. 

This supplemnent presents a description 

of the approach implemented to construct 

the DRGs and a discussion oftheir application 

in a number of health care settings. 

Section 2 provides information regarding

CAsE MIX DEFINITION BY DRG 

- 40 - 

the: data base, the statistical methodology, 

and the general process by which the 

methodology was applied to the data in the -, 

formation of the groups. At the end of Section 

2 is an example illustrating the procedure, 

to which readers may initially refer 

1 : 

f 

i 

1 

.Í 

M EDICAL CA}t 

for a general understanding of DRG construction. 

A discussion of the interpretation 

ana application of the DRGs in utilization 

review, budgeting, cost control, prospec. 

tive reimbursement and regional planning 

is found in Sections 4 through 6.

- 41 - 

dMLDICAL C.ARE 

'.Fcbrtanj 1980, Vol. XVIII. No. 2. Supplemnent 

£^ ^ .. .-. . 1. 1 

i. C;onstructlon ot u 

,IT PRIMARY OBJECTIVE in the construc- 

.tion of the DRGs was a definition of case 

types, eas"h of which could he expected to 

receive similar outputs or services from a 

hospital. In order that the set of d

CAME MIX DEFINITION BY DRG 

- 42 - 

'I~e use of LOS as a imeasure of case complexity 

has been studied by other researchers. 

Luke in his work on case mix measurement"l 

established the high degree ok 

correlation between LOS and total charges 

rendered the patient. Lave and Leinhardt 

found significant correlation between LOS 

and measures of case mnix complexity.'? 

The process sI lorming the DRCGs was 

beguin bIy partitioning the data )base into 

mutntally exclusive and exhaustive primary 

diagnostic areas, called Major Diaglnostic 

Categories. Each nmijor c;ategory was then 

examined separately and filrther subdivided 

into groups based on values of 

variables suggested by the statistical algorithmn. 

Physician review of these recommended 

subdivisions often led to modification. 

Thus, at each stage of the process 

the suhgroups were hased both on statistical 

criteria as well as physician judgment. 

The precise variables that were included 

in class definitions varied across the major 

categories. For example, age was determined 

to be important in explaining utilization 

for hernia patients, but not an important 

factor for gastric ulcer patients. From 

each Major Diagnostic Category, a number 

of final patient classes was formed. These 

final patient classes are the DRGs. A more 

extensive discussion of the data base, the 

statistical algorithm, and the general 

strategy used in constructing the DRGs is 

presented in the following subsect:ons. 

2.1 Data Base 

The data base used to construct the 

scheme contained approximately 500,000 

hospital records from 118 institutions in 

New Jersey, 150,000 records from 1 Connecticut 

hospital and 52,000 records of federally 

funded patients from 50 instittutions 

in a PSRO region. These records contained 

demnoglrphic inlfornation al>out each patient 

(e.g. sex, age) as well as clinical and 

diagnostic information related to his hospital 

stay (e.g. pro)lemns/(liagnoses, surgical 

procedures, special services used). 

MEDICAL CARE 

Diagnostic information in the data base 

was coded with both classification systems, 

ICDA8 and HICDA2. Since there is not a 

tlFect inatch between the 2 schelimes, data 

from all hospitals could not be combined in 

a unified data base. Thus, it was decided to 

construct the classification scheme using 

the miore prevalent ICDA8 codes as the 

standard. The ICDA8 version was then 

translated to HICDA2. This translation 

was evaluated with hospital dato. and 

itecessary modlitficatioiis were miade to insure 

the consistency of the classification 

across the 2 coding schemes. Both ICDA8 

and HICDA2 record surgical procedures 

tsiíng 3-digit codes. These procedure 

codes cover not only operations performed 

but also some therapies and minor diagnostic 

procediures. For ICDA8, the ranges of 

codes that were considered to reflect actual 

operations were 010-999 and A10-A59. 

Likewise, for HICDA2, the actual code 

range for operations is consjdered to he 

010-920 and 933-936. In constructing the 

patient classification scheine, only codes 

within these ranges were considered as 

surgical procedures. 

2.2 Statistical Methodology 

The particular statistical methodology 

employed is a variation of the Automated 

Interaction Detector (AID) method of 

Sonquist and Morgan, which has previously 

heen applied in the analysis of conmplex 

saniple survey data at the UniversitY 

of Michigan Survey Research Center.2? 

The objective of this approach is to 

examine the interrelationships of the variables 

in the data base and to determine, in 

particular, which ones are related to sonme 

specified measure of interest, referred to as 

the dependernt variable. This is accomplished 

by recursively subdividing the 

observations, through binary splits. inito 

subgroups based on values of variables that 

maximize variance reduction or minimize 

the predictive error of tlie depeindent vairiable. 

Subgroups aire designated termiiiiiinal 

'4

\'nKV 111,I No. 2, IIIJ·,ri 

- 43 - 

Lroups when tlhey caliirot le pautitionedil 

further either because the samnple sizes are . 

too sinall or the remaiiiuiig variation is 

either tio low to Ibe redticedl ltirther or iunexplainable 

iii teris of the variables in the 

data base. Each observation is contained in 

one and only one of these terminal groups. 

with a predict.ld value eqtual to thie mean of 

the group. That is, ifykj is the value of 'the 

dependent variable for thejth observation 

within the kth group, then 

Yk =- k * ek (2.1) 

whereYk is the mean tbr all menihers in the 

kth grolip und ue is thte error it: Usillg ' lo 

predict or estimnate Yk,. This proceldure 

minimizes the sulin of the (ek,)- over all 

observations. Thlis, individial oblservations 

tend to have Iv¿alules close to thle incai¡ 

valite of tlie terminal grotup to whichl they 

belong. 

It was decited that the approach had he 

be implemented on an inutiractive basis to 

accommodate a hligh level of physician iiitervention. 

This was an important considleration 

since group formation using this 

algorithm is biasically iterative in niature. 

Since no computer system existed that 

could handie large data bases efficiently in 

the interactive mode, a new technology 

was developed called AUTOGRP. 20 AU- 

TOCRP supports a facility allowing onne to 

·invoke a;n algorithm that determines pirtitions 

hbased on the variance redluction 

criterion c' tée? AID algorithm. This coinmand, 

o¡ capability, of the system is referred 

to as the CLASSIFY tacility. 

' Mathematically, the algorithm can be 

'described as follows20 : Each observation in' 

Ia.data set has a value of the independent 

yariable X and a value of the dependent 

variable Y. lf there ;iare N possible distinict 

values of the independent variable, then 

'ihe subset of observations, or records, that 

'has each value X, (1 . i - N) is called a 

'category. If there are Mi observations in the 

Ith actcgryv (1 ~ i ', N), ithe toital sitan tl' 

squares (TSSQ) of the data with respect to 

DI)C (:O()NS' UCI'I()N 

the dependent variable is denlled as 

N m 1 

T;SSQ = (Y -Y) 

i=l j=l 

(2.2) 

where Yu is the value of the dependent 

variable for thejth ol)servation in the ith 

category of independent variable, and 

N 

Y = 1 

i . 

i = 1 j=1 

y', 

N 

i=l 

M, IE (2.3) 

or the meanis valtiue of the dependent variable 

in the entire data set. The data set can 

be partitioned on the basis of the independent 

variable into G groups, where 

each group is the uniúín of speci'ied 

categories. That is, we can define the mapping 

of categories to groups with sets Rk (1 

< k 1 G), such that 

Rk n Rk = 0,k# k' 

G 

U R = {i,2,3 .... N}> 

k=l 

The "within group sium of squares" 

(WGSSQ) is the total of the sqluared deviations 

(diffe'trences) of each group's observations 

from the group mean with respect to 

the dependent variable anad cain )e expressed 

as 

.WGSSQ (k) = 

where 

i R 

i o Ry 

Mj 

j l 

(Yjj - Y,, . 1 :r k 4; C. (2.4) 

WGSSQ (k) = within group sum of 

squares for the kth group 

Ilk = set of a;ill ategories of the indtcpendent 

variable in the kth group and

" E MIX DEFINITION BY DRG 

Mi 

js1 

- 44 - 

Y ) 

is the mean value of the dependent variable 

in the kth group. The total within 

group sunm of squares (TWGSSQ) for the G 

groups is the sum of the total squared deviations 

of each grotip's observations froin 

the respective groupl meian and is given by 

TWGSSQ (G) = 

G 

Y,,= i R 

i¡ Rt 

M1, 

' (Y" -- Yk,) 

j=! 

For a given independent variable, the 

CLASSIFY algorithm partitions observations 

into the particular set of groups that 

results in the uminimization ofTWCSSQ for 

a specific dependent variable. Since 

TWGSSQ is proportional to the variance 

left unexplained by the independent variable, 

minimization of TWGSSQ results in 

the miniiiiization of the unexplained variance 

of the data. 

2.3 DRG Formation from Major 

Diagnostic Categories 

To facilitate the analysis over the wide 

range of disease conditions in the acutecare 

setting all diagnoses were initially 

divided int.i 83 mutually exclusive and 

exhaustive Major Diagnostic Categories. 

Their formation was also motivated to insure 

diagnostic hornogeneity. Thus, the 

final clusters do not contain patients that 

transcended these categories. For example, 

from the point of view of output utilization, 

it may be appropriate to form a patient 

class with hemorrhoíds, hypertrophy of 

tonsils, and normal delivery. The ou" ut 

utilization of these patients is very similar, 

often ret,s.`ring a relatively minor surgic.l 

procedure with a very short preoperative 

stay and a total hospitalization period of 2 

1. 

iE Rk 

(2.5) 

MiEDICAL CAIIE 

or 3 days. However, the physicians who 

would treat these patients as well as the 

treatment processes of the problems they! 

are presenting are quite diflerent. Therefore, 

it was felt that including such pati.iats 

ill the samine class would not define a nedlically> 

meaningthl category. 

The specification of the Major Diagnos. 

tic Categories wvas peirforíied by al coiniittee 

of clinicians, tollowing 3 general 

principles: 

1. Major Diagnostic Categories miust 

have consisteiicy ii ternus oftleira iilt(llie. 

phvsiop.atho>logic classificaitiotn. or in tlht. 

maiutier in which tlihey are cliniícally ma.iaged. 

2. iMaijor Diagnostic Categories ¡must 

have a siulicient iinuiber of patients. 

3. Major Diagnostic Categories mniust 

cover the complete range of codes withouit 

overlap. 

A list ofthese categoríes as de'ined by their 

ICDA8 and HICDA2 codes appears in 

Table 1. There is also an indication of the 

corresponding Professiohal Activity Study 

(PAS) diagnosis groups that correspondl to 

each category. Note that the categories are 

very broad, such as Diseases of the Eye. 

Diseases of the Cardiovascular System aud 

Infectious Diseases.. 

A consistent process was followed in partitioning 

each Major Diagnostic CategorY 

into DRGs. First of all, each category was 

refined by eliminating certain unwanted 

observations. Cases with dead patients or 

bad records, and those that were particiularly 

deviant, were excitided ironi tiirIIILr' 

analysis. Cases with dead patients were 

removed trom consideration since thleir 

lengths of stay were probably atypical of 

the disease or problemni under consi(lerai 

tion. Records with olbviois coding errors or 

missing data were also eliminated because 

their information could be misleading. Ob-

Vol. XVIII, No. 2, Supplenent 

- 45 - 

TABLE 1. Major Diagnostic Categories 

DRC CONSTRUCTION 

Mijor PAS lIDAS HICDA2 

Category Initial Grotup Nainies G Crollp No. Codes (odes 

1 Infectious Diseases 

1-8, 10-17 000-0689, 

071-1360 

2 

3 

4 

5 

6 

Malignant Neoilasmi of Digestive System 

Mlalignlnt NS'eolasmi of Respirato

- 46 - 

CASE MIX DEFINITION BY DRG MEDICAL CARE 

T ILE 1. Continued. 

Major PAS ICDA8 tlICDA2 

Category Initial Group Names Croup No. Codes Codes 

33 Pulmonary Embolism 

34 Phlebitis and Trombophlebitis 

35 Hemorrh.dls 

36 Hyperér

'Vol. XVIII. No. 2, Supplemient 

- 47 - 

TABLE 1. Continued. 

DRG CONSTRUCTION 

Major PAS ICDA8 HICDA2 

. Category Initial Group Namies Group No. Codes Codes 

_: . 

65 

66 

67 

68 

69 

Dis,ases of Skin and< Suhcutanetous Tissue 

Arthritis 

Derangetnent and Displacemnent of 

Intervertehbnl Disc 

l)iseaser s of Bole anid Cartilege 

O)tler Diseases of Musculo-Sikeletal Systemni 

70 AniL:ci;i ioixiailies 

71 Normal iMature Born 

72 Certain Diseases and Conditions Peculiar 

to NewbSrl Il[ats 

73 Syniptoixis anid Signs Relterable to Nervous, 

lRespiratory, aa.(d Circulatory Systelns 

74 Symptonis and Signs Referable to GI 

and Urinary System 

75 Miscellaneous Signs, Symptoms. and 

Ill-defined Conditions 

76 Fractures 

77 Dislocation and Other Musculo-Skeletal 

Injury 

78 Internal Injiury of Craniumn, Chest, 

and Other Organs 

79 Open Wound and Superficial Injury 

80 Burn 

81 Conmplication of Surgical anid Medical Care 

82 Adverse Effects of a Certain Substance 

83 Sls>.-eia a l A i¡ issi>l. s :iredl IE.x;uiuliinatiolis 

Widiout reported Diagnosis 

251. 252, 

254-255 

256-258 

265, 266 

261-264 

259. 260. 267 

268-271 

272-282 

342. 343 

283, 344-349 

680-6840, 

686-7099 

710-7150 

725-7259 

7171-7180. 

720-7249 

716-7170, 

726-7389 

7.10-7433. 

7%138-7570, 

7572-7599 

Y20-Y209, 

Y22-Y239, 

Y26-Y279 

Y21-Y219. 

Y24-Y259, 

Y28-Y299. 

YO05, 780-7799 

lo 

284-287 4432, 780-7808, 

7814-7815. 

7817-7823, 

7825-7834. 

7836-7837 

288-290 784-7865 

291-295 606-6060, 

628-6280, 

6293. 6295, 

7810-7813, 

7816, 7835, 

7866-7889, 

790-7910, 

793-7969 

296-310 800-8299 

311-316 830-8480 

317-320, 326 850-8699, 

332 950-9599, 

9953, 9954 

321-325, 870-9390 

327-330 996-9969 

331 940-9499 

338, 339 997-9991. 

9993-9999 

333-337 960-9952, 

9955-9959 

340, 341 YOO-Y004. 

Y006-Y 159, 

379-3793, 

388-3899, 

789-7899 

--- 

680-6840, 

686-7099 

710-7150 

725-7259 

7171-7180. 

720-7249 

716-7169. 

719-7199, 

726-7390 

740-7.599 

Y20-Y209. 

Y22-Y229 

Y24-Y249 

Y21-Y219, 

Y23-Y239 

Y25-Y299, 

Y40-Y489. 

760-7689 

770-7709, 

773-7769, 

778-7799 

780-7839 

771-7728, 

777-7779, 

784-7969 

800-8299 

830-8489 

850-8699, 

900-9049, 

950-9599 

870-8979, 

910-9391 

940-9498 

996-9999 

960-9959 

Y00-Y 1'.J 

Y50-Y896

CAo.. MIX DEFINITION BY DRC 

- 48 - 

servations with disproportionately high 

values of length of stay were excluded 

since a few deviant records could have a 

marked effect on the stability of a group's 

distribution. - 

The screened set of records in each 

category was then used as input to the seconcd 

stage, in which the CLASSIFY algorithm 

was applied to suggest groutps of 

observationls, on the basis of prespecitied 

incdependent variables, that may he different 

with regardl to length of stay. The set of 

independent variables selected as input to 

the alh~orithmil was intentionally limited to 

those variables descriptive of the patient, 

his disease condition and his treatment 

process that would be readily accessible on 

iimost discharge abstracts, specifically diagnoses, 

sturgical procedures, age, sex and 

clinical service. This constraint was 

applied for several reasons. First, mean 

lengths of stay are oblserved to vary across 

levels of these variables in descriptive 

statistical suminaries of hospital discharge 

data. 23 Second, these variables are always 

recorded and entered in almost all hospital 

inftrmation systems. This increases the 

classification's potential for implementation 

in most research and applied healthcare 

settings. Including other items of information 

such Ias ancillary services used 

would limit its applicability to those systems 

that collect such data, which would 

exclude, tor example, the PSRO PHDDS 

data base. Fin' ly, restricting the variables: 

to this set also simplifies the class definitions 

and controls to some extent the final 

.11imber of categories. A classification with 

numerous groups and a complex definitional 

structure is unmanageable. , 

For each independent variable. output 

from the CLASSIFY procedure inltluded 

the total numberofobservations in the data 

set, the total number of different values 

the variable assumed (i.e., number of 

categorie.; ,- c:ells), the number of groups 

formed, the total suni of squares (TSSQ) 

and the per cent reduction in total sum of 

squares attained by such a grouping 

SIEDICAL CA:RF. 

((TSSQ-TWGSSQ) / TSSQ)* 100. Once 

these results were obtained, a cliniciSi,n 

selected the inost appropriate variablle 

for division. 

--The interpretation of the partitioning 

suggested by the algorithin was a comiplex 

task, with many fiactórs examined and 

weighed simnultaneouLsly. The decision to 

accept, to reject or possibly to revise the 

recomniended partitioning was based (on 

bothl the statistical evidelnce aind the cliíi. 

cian's imedical knowleclge. The statistic,,i 

results were examined in light of certain 

criteria. Variables yielding the highll(st 

percentage reduction in variance werc 

prime candidates tbr dividing the data set. 

However, the number ofcells or values fir 

those variables and the nutmber of groups 

tobmed were also considered. It is an artifact 

of the algorithmi that many partitions 

can be created when the independent variable 

has a wide range of v ilues. Moreover, 

too many grotups forined at the first split 

become difticult to manage and are of 

(luestionable significance. For example, 

secondary diagnosis often had many diifterent 

values and thus often produced significant 

variance reduction by forming 

many subgroups. However, such grontps 

were diffticult to interpret as a prinmary partition 

and tended to be of limited value. In 

all cases, groups were further examined in 

a more descriptive framework to determine 

if the statistical significance was supported 

by medical interpretability. 

The Lict that a variable that appeared to 

be powerful in explaining variance was not 

selected at a particutlar stage does not mean 

that the variable was ignored. Two possibilities 

still existed. If that variable were 

independent of other variables in explaini 

ng variance, it would appear in subl>sequent 

stages with the saime power as at 

the earlier stage. If, howéever, it were correlated 

with a variable 'chosen at an ealier 

stage. then its explainatory power at sul>selqent 

stages woulldl be lesscled accordling 

to the strength of the correlation. For 

example, if seconclary diagnosis WaS

?Vol. XVIII. No. 2. Sttilluleareiltte 

- 49 - 

strongly related to age and Lage was 

selected ats a partitioniiiig variable, thenl 

secon(idary diagnosis would not aippear 

powerful in subsequent clustering. 

Groups were then generated bas i on 

the most appropriate variable, that is, the 

one that .et as nainy of the criteria 

specified above as possible. In particular, it 

1) exhibited a significant reduction in variance 

relative to iiost ot the other variablles, 

2) created a manageablie ni tilher of>groil>ps 

based on the relatively smaill iinlmer ol 

values of the independeit variable, and 3) 

createtl groups whiose ueaiits were signilicantly 

dlitierent. Also. grotups Ibruied were 

homogeneoxns troni a clinical perspective. 

Once each Mlajor Diagnostic Category 

was initially pairtitioled inito suilgrotips 

based on the valiues of an independent varíable, 

a decision was nmade whether or not 

to further subdivide each saubgroup based 

on any of the other available variables or to 

end the partitioning process by treating 

thenm as terminal groups. The statistical 

basis for this decision was deterniined by a 

set of stopping r ules. For any given gronup, 

the partitioning ceased when etiher one of 

the following conditions was mnet: 

1. The grrotip was not large enoughl to 

Warrunt anIother classificatiom, tlhat is, whllvl 

the ilnliilber of oihserfiltion>is in the groiup 

was less thani 100. 

2. None of the varialbles relducedl uniexplainil 

variation I)v at Ieast 1%, or 

((TSS -'I WSSQ)/TSSQ)*100 < 1 per 

cent.* 

Otherwise, the group was fuirther subdivided 

according to the criteria discuissed 

previously tor generating new, sulbgroups. 

In some cases, however, the process was 

terminated tor nonstatistical reasons regarding 

overall inanageability (e.g. inaintaining 

a low number of total groups) or 

medical interpretability. 

* This I per cent I>ouid was increased ii certain 

Major Diagnostic Categories. 

DC; :()NST'I'(:'i'ION 

Tlhis grotu),ping process resulted i¡ tihe 

lormatio, oí' :383 finai grolups or I)CG.s 

each defutied by somie set of the tollowing 

patient attributes: primary diagnosis, see- 

' ondiny diagnosis, primary surgical procedure. 

secontlary surgical procedure, age, 

and in one case, clinical service aiea. While 

other variables such as sex, tertiary diagnosis 

or surgical procediure were 

examtined, they were not imnild to be significiant 

iii explaining ouitput utilization. A list 

of these grotups with a brieft narrative descriptioei 

of their contents appears in the 

Ap)l)elctlix. A more ctiip)lete spetcification 

can be obtained fromn tlie Health Care 

Financing Administration.' 

The DRGs vary considerably in their 

structure across the iMajor Diagnostic 

Categories. Some Major Diagnostic 

Categories are not further subdivided, 

such as Category 35, Hemorrhoids, in which 

no variable demonistraited a sufficient efttct 

in further explaining outpitt uti lization. On 

the other hand, Appendicitis, Category 46, 

is ftirther subdivided on, the basis of 

specific primary diagnosis and the presence 

ot a secondary diagnosis. This restilts 

in 4 DRGs: appendivitis (without 

peritontitis) and without a secondary diagnosis, 

appendicitis (without peritonitis) 

with a secoudary diagnosis, appeidicitis 

(with peritonitis) without a secondary 

diagnosis, and appendicitis (with 

peritonitis) with a secondary diagnosis. 

This symmetric breakdown suggests that 

the effects of primary. diagnosis and the 

presence of a seconldary diagnosis are additive 

in nature. Major Diagnostic Category 

76, Fractures, has the nmost complex structure, 

resulting in 13 DRGs, indicating both 

the importance and interaction of 4 variables: 

primary diagnosis, secondary diagnosis, 

prinmary surgical procedure and age. 

It should also be noted that when variables 

are h.ghly correlated, very often only 

one applears in the classification for a 

specilic mnajor category. Asn extrieme exampie 

of this is Major Diagnostic Category 36, 

Hypertrophy of Tonsils aud AdeMoid,

- 50 - 

( ASE MIX DEFINITION BY DRC NIEDICAL Ckax 

Fic. 1. Tree diagram i!lustrating partitioning of urinary calculus patients. 

where almost everyone has a tonsillectomy 

anid/or an adenoidectomy. The high correlation 

of primary surgical procedure with 

primnary diagnosis results in no further variance 

reduction in the category that can be 

attributed to surgical procedures. Thus, although 

stirgery is aiinost always used in the 

treatmneit e.' i.,';esits with Hypertrophy of 

Tonsils anacl Adenoíds, since virtually 

every patient in this category had the same 

surgical procedure, the surgical information 

did not differentiate the utilization of 

these patients and was therefore not used 

in the formation of the DRGs for this 

category. 

2.4 An Example 

The iterative partitioning process used 

in form;,-. the DRCs can hest be illustratedl 

ti¡ tic co.ltext ol aln xamill)lc-tile 

classitication of lMajor Diagnostic Category 

55: Urinary Calculus. This category con- 

tains patients with a primary diagnosis 

(ICDA8 codes) of either 592, calculus of 

kidney and ureter, or 594, calculus of other 

parts of the urinary system. The formation 

of the DRGs from this Major Diagnostic 

Category is summarized in the tree diagram 

presented in Figure 1. First, this 

category is partitioned into 3 groups based 

on the variable primary surgical procedure. 

The first group contains; nonsurgical patients, 

which are those with either no operation 

or with a procedure code (ICDA8) 

outside the range 010-999, A10-A59,t The 

second and third groupsare fonned on the 

basis of the specific procedure perfonrned. 

In particular, the imore complicated procedures 

perfonned on patients with a urinary 

calculus-ne.phrotomy, ureterotomy, 

cystotomy-are in the third group, while 

f Operatitións oed outside these ranges ae not 

considered actual surgicil procedures since they 

represent minor procedures or therapies.

.'yo. XVIII. No. 2, Supplentcsit 

Size = 1125 

- 51 - 

DRG CONSTRRUCTION 

TAtiLE 2. Descriptive Stastistics tor the Partitioning of 

MIajor I)iaganostic Category 55 (Urinary Calculis). 

Mean '-= 6.93 S.P. - 6.44 

Partial M .Mean 

:' Independent Variance Length 

Variabhles Explainiiedl DRG No. Size lf Stay S.D. 

Priinary surgery 

. Minor 

Major 

Seconidary tLi.glo

- 52 - 

CAS¿ MIX DEFINITION BY DRG MEDICAL CkAE 

TABLE 4. Suggested Partitioning (3 groups) of Urinary Calculus 

Paticnits on tie Basis of Typ)e ofI Priiialry Surgery 

Independent 

Group Size Mean ' Vtiable Description 

I 1 2.00 749 Other antepartmn procedulres to 

ltrminate pregnancy 

1 2.00 571 Mleatotomy 

1 

1 

1 

1 

2.00 

2.t00 

3.00 

3.(X) 

277 

249 

430 

862 

Venous anastoinosis, intra-abdominal 

Other opr:ationis on peril)heri vessels 

Incision of bile (hepatic) ducts 

Arthrocentesis 

1 :3.00 601 V;asectomy 

1 3.00 921 Local excision of lesion of skin 

adii siilcittaiiteons tissie 

2:) 3.28 (000 No code 

688 

1 

4.08 

5.00 551 

No code 

Ureterectomny 

2 2 5.50 574 Repair and plastic operations 

on nrethra 

7 5.71 559 Other operationls on ureter 

218 6.25 A46 Cystoscopy and irethroscopy without 

effect upon tissue 

5 6.40 A45 Endloscopy of colon and rectum without 

eifelct upon tissue 

la 6.40 568 emioval of calculs aLnd drainage of 

bladder withoot inicision 

2 6.50 57'2 Excision or destruction of lesion 

of nrethra 

146 '6.59 557 Passage of catheter to kidliey 

21 7.14 575 Iilation of urethra 

1 9.00 A16 Biopsy of thorax 

3 2 

2 

1 

1 

10.00 

11.50 

12.00 

13.00 

A21 

566 

A44 

549 

Biopsy (contined) oF turinary tract 

Repair an.d other plastic operations 

on hladder 

Esophlagoscopy and gastroscopy without 

el'fect ipon tissue 

Other operations on kidney 

1 

, 3 

13.00 

13.67 

14.00 

556 

561 

562 

... 

Repair and plastic operations on ureter 

Local excisiot antd'estruction lesion ' 

of bladder, transurethrol 

Local excision and destruction of 

lesion of bladder. otihbr 

3 14.00 582 Prostatectomy, transurethral 

1 14.00 583 ~ 

Prostatectonmy, other 

8 14.13 545 ,Nephrectomy, complete 

72 14.46 541 Pyelotomy 

40 14.47 560 Cystototomy 

101 14.63 550 Ureterotomy 

19 15.89 540 -Nephmtomy 

1 16.00 513 : : eiieorrhoidectoiny 

11 16.82 544 Nephrectomy, partial 

1 17.)00 546 Repalir anid plastic operations 

on kidney 

570 Urethrotomy, external 

1 21.00 A27 Biiopsy of hone 

I.(1. 5

.V'oL XVIII. No. 2. Stpphlidary diagnosis 

(tLx), age, auid sex. 

The number of groups formed by the 

algorithm and the corresponding per cent 

reduction i¡n unexplained variation tbr each 

of the ,ai-'lttles are shown iii Tablle 3. 

Sinc- the greatest reduction in unexplained 

*ariation was achieved with operl, 

and a linited mnhuher *of gr'oups (3) this 

variaulle wvas considered thle prilne *- "ilidate 

for initial subdivision of the caitegory. 

The algorithli suggested 3 groups whose 

vontetlns are described inl Tabl 'l e . This 

table presetits the diillerent :;urgical procedures 

contained in each groutp, the 

corresponding number of observations 

(SIZE), and the meait lengthi of stay 

(MEAN). Note that Inorkt thal .98 per cent 

of the obseraitions il tlihe tirst rtrollp have 

no surgical procedure listed. The second 

grotlp primarily contails ol>servationis with 

relative!y mitior procedtres such as cystoscopy 

a;i¡

CASE MIX DEFINITION BY DIIC 

'ABLE 7. 

No. 

Group Observed Mlean S.D. 

No secondary 449 3.28 2.88 

Secon

Vol..'XVlII, No. 2, Supplement 

- 55 - 

All uWlW C~oi/i 

· 1425 

0 4 9 Iz b 20 24 26 2 

*imgfh ot #tey l{dr y 

lAqjlí of soy {6ays) 

No Surwqry Mtno Suqi.es to uol Su;oo.s 

n ?.71 · 

A 429 A ' 286 

m.oen 403 mon 

4 · 33 

4 430 * . 31 

e 636 mon * i499 

4Vf'( 

DRC CONSTRUCTION 

o a te iC 20 24 2? {2 

iingth of s1ay (dura) 

04 a 2 6 s 2G 24 e# 32 

loeth of * ¡oy dop.) 

0 4 a 12 4 20 24 2e 32 

tength of fey (O40 

DOR 241 

C.»u , f.. C$oqIC. t 

DO0,or i20 00ot41w 01 COAu ¡Ow lo loW0 h)0. 

1 

. 

0*G 242 

¡loddO CO4....u s tl 

n~ "t-~, CAt100. ¡oy,00*2. 

O]~,o 

8¡8 beOn 

o1 o~4·1K~ _ o4 4t 

P dM SGo avy 

0 4 t2 4 2024 2 32 

IwgiN of etoy id~r) 

Wdns o l Pe____,__t I ____oD P _oera 

W*lhu!n S.o4dr Da Wth S~odwr Oi 

| r * 449 r * 262 

_-I I * · 329~ meao ·632 

04 e I* 20 2429 232 4 6 12 4 20 24 2 32 

lengAQ of oM (¡do1o) eng *o~y of l4(dOyl 

006 230 11 I :0¡ 

wemi Co¡sw, u e a.., *.tq¡ eOoo 

.4 '..OO Su m uan r 

.i Con .o tee , tOlgl , 

y o s~orib o oup i 

FIc. 2. Sunimary of lerngth-of-stay distributions for groups fortmed in partitioning provess.

C- ' MIX DEFINITION BY DRG 

TABLE 11. 

- 56 - 

No. of Per Cent 

Variable Groups Reductiba 

age 3 1 ..05 

oper2 1 0.0 

dxl 1 0.0 

%ex I 0.0 

down ofthe surgical groups on the basis of 

secondary diagnosis. Moreover, in light of 

one of the major objectives of' keeping the 

total number of classes low, additional 

groups formed at this stage of the partitioning 

of Urinary Calcultus patients would be 

of questionable value. Therefore, the 2 

surgical grotups were not divided further 

but were considered terminal groups. 

Step 5 

The 2 subgroups formed from the nonsurgical 

cases on the basis of presence or 

absence ofother díagnoses were evaluated 

to determine if they should be partitioned 

further or left intact as terminal groups. 

The algorithm was applied and produced 

the results shown for the nonsurgical cases 

without multiple diagnoses. The algorithm 

produced the results listed in Table 11 for 

the nonsurgical cases with multiple 

diagnoses. 

With respect to the nonsurgical cases 

without multiple diagnoses, both sets of 

groups font - - ,n the basis of age and secondary 

sa ,ncal procedure, respectively, 

MEDICAL CARE 

were determined unacceptable. In each 

instance, more than 95 per cent of the ob. 

pervations fell into the first group, leaving 

the second group with fewer than 25 cases. 

For the nonsurgical cases with multiple 

diagnoses, the 3 groups formed using age 

levels were considered as potential sub.. 

groups. The age levels detining the bound. 

aries of the groups were 66 and 70. This 

partition was rejected for reasons similar to 

those above, namely, the lopsided distribution 

ofl cases in thle groups. Almost 90 per 

cent of the observations had an age under 

66. 

Thus, the nonsurgical groulps with aind 

without multiple diagnoses were consíidered 

terminal groups. 

We conclude, then, that specific surgical 

procedures and the presence of multiple 

diagnoses were ímportant variables in 

predictir.g length of stay for urinary calculus 

patients. The 4 DRGs formed were 

significantly different (a = 0.01) with respect 

to their average lengths of stay and 

are clinically interpretable. To be sure, by 

overruling some ofthe partitions suggested 

by the algorithin, a certain anmoulnt of 

explanatory power was sacrificed. But, the 

trade-off was generating a reasonable 

number of subgroups or DRGs which 

could be interpreted from a medical 

perspective. Figure 2 presents a descriptive 

summary ofthe length-of-stay distributions 

for the groups formed as part of the 

partitioning process in this example.

- 57 - 

MEDICAL CARE 

Febiuary 1980, Vol. XVIII, No. 2. Supplemteilnt 

3. Interpretation of the DRGs in Health-Care Studies 

THE DRGs AS CONSTRUCTED met the original 

rese:trch objectives; that is, they provide 

a detinition of case types by relative 

output ::tilization, where the definition of 

case type ¿has the fillowing properties: 

1. The varialbles iste in ti the deinition ol 

a case are li¡¡¡ited( t tthose reflirring to the 

patient's ;at ilre 'tiillilllll

CA .IIX DEFINITION BY DRG 

- 58 - 

differences among providers still remain 

with respect to length of stay and other 

output mneasures which cannot be accounted 

for by patient characteristics ora. 

surgical procedures. 

The development of the patient classes 

is not intended to inhibit in any way the 

practice of medicine but to offer one the 

capability of examining reasons for variations 

in service utilization, treatinent provess, 

and outcomie. In this cont..dt, the 

groups can provide a iramnework tor the 

initiation of an ongoing process of comparative 

anialysis of hlicatlh caire with the 

long-rim goal of determining both the cost 

and value of any kind of care that might be 

delivered. With such information, meaningful 

dialogue among clinicians, administrators, 

planners and regulators canl proceed 

in rationalizing of observed difterences. 

Only in this way can strategy, policy and 

ME :DICAL CAR: 

politics interact to the benefit of the con,. 

munities served by each institution. 

The classification of patient records into 

DRGs is a constantly evolving process. In 

fact, the group structure described here 

represents the third classification schelie 

developed using the methodology pre. 

sented in the previous section. As coding 

schemes change and data are collected thiat 

are more current and represelntativei o' 

acute-care institutions in the United States. 

these grotps will be re-examined anud revised 

accordingly. 

Clinently the DRCs arc being appllied in 

a number of different practical and research 

settings in the health-care field. 

While these applications are discussed extensively 

elsewhere, a brief overview is 

presented in the following sections to give 

the reader a better uinderstanding of the 

DRGs' potential utility in a variety of areas.

- 59 - 

.SlDiCL ( :AL.. 

FcbnartlrJ pI

CASE ' : DEFINITION BY DRG 

. Gcc 

tx. 

Lz . 

x 

14WW 

oEc.v 

w cn 

U-. 

< 

-Isa 

W Z 

X : 

, a 

o 

x 

- 

u ' 

-aM. 

< 

1 1 

¡ 1 

cn -1 r- 01 

- ~ .' 

- i--- - 

04 

~ r-cae-u(E ~ 

r- a a -c c c ' 

'-04et .qIl ca -ca 

- 60 - 

ci 

a 

o 

1 

oe 

9- 

u 

MIEDICAL CARe 

number of bed days consumed per dis. 

charge. A case mix-adjusted length of stay 

(CASE MIX ADJUSTED LOS) appears íu 

.colunin 4 and is the mean length of stay the 

hospital would have experienced with the 

region's case mix. For the ith hospital, it is 

computed as 

j 

J PJ j= 1,.... -383 (DRG) (4.1) 

where au is the average stay in thejth DRC 

in hospital i, and Pj is the proportion of the 

region's cases i¡i )RG j. As such, it stamil. 

ardizes or holds constantcase mix across all 

institutions and results in a measure that 

one can use to compare hospital utilization 

based oa length of stay. 

As a complement to the case mixadjusted 

length of stay, column 5 contains 

the average length of stay the hospital 

would have experiencgld with its owtn catse 

mix, but the region's average length of 

stay within each of the DRGs (LOS 

WEIGHTED CASE MIX). That is, for the 

ith hospital it is computed as 

A p, j = ....,383 (DRG) (4.2) 

where AN is the region's average stay in the 

jth DRG, and pu is the proportion of the íth 

hospita's cases in thejth DRG. Since it uses 

a standard set of relative weights (Aj) ftr the 

DRGs, it can be interpreted as a relative 

measure of case mix complexity, with 

higher values indicating a more complex 

case mix, that is, a greater proportion of 

cases receiving a high amount ofthe hospital 

outputs. In interpreting this measure, it 

should be noted that complexity is a function 

of both the condition of patients as 

well as the treatment process selected by 

physicians. Columns 6 through 8 are indices 

computed by dividing the measures in 

columns 3 through 5, respectively, by tlhe 

region's average length of stay. 

Columns 9 through 12 present the observed 

difference between the institution's

Vol. XVIII, No. 2, Suppletiment 

- 61 - 

and region's average length of stay (AV. 

ERBAGE LOS DIFFER) and indlicate to 

what extent this dit'ference can be attributed 

to DRG-specific average length of 

stay or the ai (DIFFER DUE TO LOS), the 

case mix of its patients or the PU (DIFFER 

DUE TC CASE MIX), and the combination 

of the 2 or their interaction (INTER- 

ACT DIFFER). 

The separation ot' the diltereiice intO 

these comiponents is hased ou the 

technique described by Kitagawa,'4 brietly 

dlescribed here fi)r the reader's convenlience. 

It' ; auld A reprepresent tie averalige 

length of stay for the ¿th iistitution and the 

region, respectively, then 

a, = Pj aJ;, = 1 .... 33 (DRGC) (4.3) 

A = PJAJ. j = ..... 383 (DRG) (4.4) 

The difference between a; and A can be 

expressed as 

DRGC APPLICATIONS 

hospital and the region not accounted for 

hy the LOS difference (column 10) or the 

case mix difference (columnn 11). 

The utilization of the report lies in interpreting 

the differences observed between 

an indivtdual institution and the set of institutions 

or region as a whole. For example, 

consider Hospital 4, the largest with 

41,643 cases (column 2). Its average length 

of stay was one of the longest, 7.479 days 

(column 3) or 0.708 days longer than the 

region (column 9). This diftterence can be 

attribhltedl in varying degrees to the following 

componients: 1) dileretices iii case 

composition or case mix; 2) differences in 

case-specific average lengths of stay; and 3) 

differences attributed to the interaction of 

case mix and case-specific average lengths 

of stay. With respect to the first component, 

if the hospital's case mix had been treated 

at the region's average length of stay per 

DRG, Hospitail 4's average length of stay 

would have been 6.96 days (column 5), for 

a difference of 0. 188 days over the region 

(column 11). On the other hand, if the case 

a - A = P (a, -A,)+ Aj (p - P) + (a - A (p - P 

AVERAGE 

LOS 

DIFFER 

,,., j , . / . - 

DIFFER DUE 

TO LOS 

for i = 1 .. , 383 (DRG). 

I'This .may be rewritten as 

DIFFER DUE 

TO CASE 

MIX 

k-A=( A Pa",,- PA,)+ (1 Ap p,,- A, P) + C (aU-) (pi,-P,) (4.6) 

.'i j j j J 

for j = ,...,383 (DRG), 

.from which we note that the first 3 differ- 

1:ences are computed by subtracting the 

region's average length of stay from the 

m,,,ores in columnis 3 through 5. The 

interaction component is the residual or 

the amount of the difference between the 

qr~ 

INTERACT 

DIFFER 

(4.5) 

mix were standardized or all institutions 

were treating the same types of patients, 

Hospital 4 would have had an average 

lengtlh of' stay of 7.320 days (coluann 4), 

about a half day (column 10) longer than 

the region. Moreover, relative to the

CASE MIX DEFINITION BY DRG 

- 62 - 

others, it would have had the longest 

length of stay. Thus, institution 4 has an 

average stay that is .708 days longer than 

the region; .188 days is attributed to a more 

complex case mix and .548 days to longer 

DRG-specific average lengths of stay. 

The sign and magniitude of the interaction 

ditterence (column 12) is ani indication 

of the extent to ,-which the observed difference 

between an institution's and the region's 

average length of stay is accouinted 

tor b)y I)oth case mix and DRC-speciftic 

length of stay patterns jointly. It is that 

amount of the overall difference between 

tlhe iinstitutio>iial aud regional averaige.s tlhat 

cannot be allocated independently to case 

mix or DRG-specific utilization. A large 

positive interaction might arise, for example, 

if the hospital has ani average length of 

stay higher than the region in those DRGs 

where it treats proportionately more patients 

than the region or a lower-thanaverage 

length of stay in those DRGs with 

proportionately less discliarges. In this instance, 

the particular patterns of the hospital's 

deviation from the region in both case 

mix and utilization hby DRG contibuted to it 

having a higher average length of stay. 

The interaction difference is especially 

imniportant to consider if its absolute value is 

large relative to the other differences 

(coluirins 10 and 11). Under these circumstances, 

one should not use the measures 

in the report for that hospitalin comparison 

to the'others. A large interaction 

indicates that '>osital utilization patterns 

vary by cse type and standardization 

would be misleading. Referring to the in- 

.N.mriation in Table 12 Hospital 4 has a 

small-magnitude negative interaction 

component, -. 029. On the other hand, 

consider Hospital 8. with an average 

length of stay of 4.943 days or 1.828 days 

shorter than the region. The bulk of this 

difference is attributed to lower DRGspecific 

utilization (- 1.315 cdays), but also 

to -' less .. ""'niex ca;e mix (-.986). '4owever, 

¡note tlihat these legative diti2rences 

MEDICAL CARE 

are partially offset by a half day (.473 days) 

attributed to the interaction of case mix and 

.pertrnnance. In this instance, because of 

the Ml1atively large interaction term, one 

should exercise caution in comparing this 

institution's standardized utilization 

measures with the other hospitals in the 

report. 

Sinice discharge information on all patients 

is routinely collectecl by institutions 

fir admlinistrative puriposes, ceisus data of 

complete coui1ts rather than sanmples are 

generally available for the report described 

above. However, hecause of its size or limitations 

ina thlie imbiler of services aicl stall' 

available, an individual institution may not 

treat patients representing a large range of 

cases or DRGs. In this instance or in situations 

where there is known to be lar..e interaction, 

as discussed in the prec'-ding 

paragraph, an alternative st¡ategy is to restrictone's 

comparison to a limited number 

of institutions with uniformly high nunm- 

!lers of cases or to perfornm comliparisonis on 

a limited set of DRGs, for example, those 

contained in a few Major Diagnostic 

Categories of particular interest. 

4.3 Institutional and Practitioner 

Profiles in PSRO Evaluations 

Protessional Standards Review Organizations 

(PSROs) were ruandated under an 

amendment to the Social Security Act of 

1972 to assure the quality and appropriateness 

of health-care services delivered to 

federally funded patients (Medicare, 

Medicaid, Title V). To this end, one 

$pecific responsibility of each organization 

iis "profile analysis"-that is, the construction 

and review of relevant sunmmuries of 

aggregated data pertaining to the care and 

servwves received hy patients and provided 

by practitioners and institutions in the 

PSRO area. It is basicálly a descriptive 

ainalysis of health care aind service patternis 

uIsiug mIeastires aud lorniats which serve to

Vol. XVIII, No. 2, Supplenwt 

- 63 - 

DRG APPLICATIONS 

1) facilitate the identification of excep- periniented with the uise of institutional 

tionill areas of p)erlorinilnce and 2) Irovide length-of-stay profiles based on DRGs for 

a muechanismi of the ougoing miuonitoring of 

the overall systenm, with particuilar atten- 

the retrospective inonitoring oft' utilization 

patterns in their area. 

tion to previously defined prol)len areas. 

While ,'onsiderablle latittlde is given the 

PSRO ftir the tormant ainid content of'a profile, 

th,:re are 3 basic conmponents to its 

design: 1) the patient groulp, which defines 

the sope of tihe proille i terins o" tihe 

popuilattiolu to I)e aialyzed; 2) the profile 

subjects, which detine the subset of the 

patient gmroip whose data is to be profiled; 

that is, aggregated a,'(,l :;linallilrizc, the standard 

deviation (S.D.) and the median (ME- 

DIAN) length of stay, or 50th perc'mntile. In 

this example, the median length of stay for 

the region as a whole was 2 weeks. However, 

there was considerable. variation 

amog hospitalospitals. Hosital 8 had a m¡edian 

length oft'stay of 10 days, while Hospitals 4, 

2 antd 3 had medians of 15 days. Further 

investigatipn in this area through medical 

care evaluatian studies may be warranted 

to determinine if there are quality-oftcare 

problems in the short-length-of.stays hospital 

or inappropriate utilization i¡n the 

other 3 institutions. 

As illustrated i¡t the albove exaumple, 

these profiles are potentially useful in 

identifvilng areas of' inail)pro)priate utilization, 

which can lead to addtlitioimal issues to 

adl

C... L MIX DEFINITION BY DRG 

a 

w 

-), 

M. 

a 

9 

CL u,,, 

£e. 

ui. 

g. 

a 

ir 

;t 

g- 

«.;........ 

wJÂjijijj 

r__l.Ômq^Wfw 

l* 

. _ 

4* * 

I 0 

* e 

5 o 

1 e 

o e 

e e 

e ¡ 

Ot tO 

e e 

t e 

· ( 

um i 

ow 

, vX 

ee o · III I ; 

e:;;1-: 

ce1 1 1 

@ I *- @ I * 

I1* 111 I 

. 

I I I 

· rII111 

o e 

*e 

e., 

I Ie 

aer 

: 

0 

* - 

-a 

·~ a 

a ^ a "@@wj* 

,,,~,,,i:? } 

tli · 

*,' ,ri iIl 

I·~~~~~~~ 

e e 

e 

e* - ~-e P~P,- 

--- o____ 

__ _____ 

--- a----- 

L&L O o Po 

rSci « .X;a;lll) 

o 

-1u ewa 

c 

Qu 

t 

e. 

' 

eV. 

u - 

z 

a 

r 

1c 

w 

#m 

u, 

u 

. u 

zu,. 

a 

g 

u' 10 

so' 

- 

zaM 

o" ,. 

C-9 . 

4a 

oz5Call 

1.4 

za 

- 64 - 

M1EDICAL CAIE 

sons. Moreover, each group can be revieweld 

by PSRO personiiel anid a sit 

selected for follow-up analyses. To tacili-. 

tate further the identification of aberrant 

utijization behavior, automnated screens 

based on statistical algorithmns have also 

been implemented that first flag those 

DRGs with relatively high inter-losipital 

v;riation, then determine for each hospital 

the groups in which its mean length ofastav 

was significantly higher than the area aidl( 

those in which it was significantly lower. 

4.4 A PSRO Monitoring Systecn 

The utility of the DRGs in reviewing 

PSRO data has suggested their potential 

future application as a control mnechanismii 

to replace the present costly concurrent review 

process with a timely, cost-effective 

retrospective monitoring systein. In this 

context, at the ti mne of patient dlischarge, ani 

abstract is produced l,and ia DRG nulnber 

assigned based on data in the record. When 

a small sample of records has accumulated 

in a DRG for an institution or set of i nstitutions 

under examinatiot., information froin 

the sample can be used to estimate the 

parameters ofthe distributions of 1 or morc 

dependent variables related to utilization 

or quality of care (e.g., length of stay, death 

rate, postoperative wound intection rate). 

Once acceptable patterns of practice have 

Ibeen established, thie dionitoring systenm 

can use standard statistical control procedures 

to detect changes in the process. 

iWhen: changes are detected, special 

studies may be justifie, to determine the 

cause for the change, which may be attributed 

to errors in the record, 1 or more 

particularly deviant cases, or an overall 

change in the treatment process that has 

affected all the cases. Work is currently 

under way to develop computer-based systems 

which could support such a control 

mechanism in a timely manner on either a 

hospital or PSRO level. 

.

- 65 - 

£.¡EDCAL CARE 

ebrNuarV 1;980. Vol. .YVIII. No. 2, Supplement 

5. Case Mix Accounting in Hospital Budgeting, 

Cost Control and Prospective Reimbursement 

5.1 Case Mix Accounting 

.;N NItMPt .TANT OBJECTIVE of hospital costing 

and budgetary systems is the uinder- 

·standing anct control of hospital costs. In 

traditional organizational setti ngs, cost 

control is miost su¡cccssllI ill those sitiations 

where well-e a 

precise definition of the services provided 

:by the institution. In a general sense, hospitais 

provide "patient care, .lbut more 

'specifically, they proviíde patient care of' 

.various kinds and intensities over various 

"durations based on the needs of the pa- 

.tients treated. 

Since the DRGs form a classification of 

;the patient population into classes with 

*!similar expected output utilization, they 

can provide a definíition of the services 

Yiprov. 

'w:k )y a hospital. As such, they allow 

t.he .esources consumed and costs incurred 

;to be related directly to the types of patients 

or case mix that the hospital 

t'eats. 2 .8.B This is important in a hospital 

setting, where it is not manageinent (i.e., 

administrators) but rather individual 

physicians who are reasponsible tor allocating 

resources through various services 

mnd departments in order to provide effec- 

:tive patient care. To a large extent, physi- 

.*cians act independently of each other and 

Yj'Ure not generally aware of tihe overall lib'nancial 

implicatioxis oi their individual de- 

ìsio~t. If hospital cost control is to be attaineld, 

effective communication between 

the financial systems of the hospital and its 

phys-cians must be achieved. By tobnulatitg 

the hospital biudget in terms of patient 

classes with similar piatterns of care. a direct 

linkage hetweenx the practices ot' inividual 

physiciians and the finaincial conseqcuences 

for the hospital can be realized. 

'rie goal, theni, of a caise mix amc:ounting 

systemi is to provide a complete tfinancial 

picture of the costs of treating specific 

types of patients, whose care is the basic 

service ofa hospital. Under the traditional 

organizational structure of a hospital, there 

is no department whose responsibility is to 

insure that individuial patients are financially 

well managed. Typically, the hospital's 

2 accounting systemns-linancial and 

managerial- deal with patients in the 

aggregate and not on an individual basis. 

The financial system provides the lbasic financial 

description of the hospital in ternms 

of the balance sheet, income statement and 

ftinds flow, while the managerial accounting 

system provides the financial information 

oriented at the department level (e.g. 

nursing, laboratory, medical records) for 

internal nmnagement purposes. Thus, 

hospitil accounting systems have not provided 

the integrated picture of the financial 

consequences of the care delivered to 

individual patients that case mix accounting 

is designed to produce. 

5.2 DRG Cost Model 

The process of determining the cost of 

treating patients in each of the DRGs for an 

individual hospital or collection of hospitals 

is decribed elsewhere.? In summary, 

the types of accounts in a hospital chart of 

accounts can be categorized into 6 distinct 

service areas: 1) outpatient accounts; 2)

CASE .MIX DEFINITION BY DRG 

- 66 - 

overhead accoulnts lnot related to piatienit 

care; 3) overhead accoumits related to patient 

care; 4) hotel and other general services 

accounts; 5) nursing accounts; and 6) 

ancillary services accounts. 

The DRGs currendy encompass otly tbhinpatient 

population; hospital outpatient 

costs are not included in the DRGs' costs. 

Overhead accounits are costs incurred I)y 

the hospital iii its general operation but are 

either not related or only indirectly related 

to the provision of patient care. Depreciation 

and interest cliarges are examiiiples of 

overhead costs that are not related to patient 

care and therefore are not normally 

included in the DRG costs. Other overhead 

accounts such as hotisekeel>ing or 

laundry are indirectly related to the provision 

of patient care and are included in the 

DRG cost. The definition of the overhead 

accolnts that are considered as patientcaire-related 

versus non-patient-carerelated 

can vary, depending on the goal of 

the case mix accontiting system. For 

strictly internal mianageinent purposes it is 

reasonable to include as patient-carerelated 

the various administrative services. 

However, ifthe case costs of a collection of 

hospitais are to be compared, then the admninistrative 

costs shoild not he included, 

since administrative costs can vary greatly 

across hospitais for reasons other than case 

mix. The remaining 3 types of accounts are 

all directly relatt:d t.. patient care and with 

the addition of the outpatient account, are 

referred to as the final cost centers. The 

-'vices associated with these accounts can 

he directly related to individual patients, 

allowing the costs to be apportioned to 

each patient. 

The direct costs of each final cost center 

and the portions of the cost of patientcare-related 

overhead accounts allocated 

to each final cost center (as determined by 

a special alg!)rithiii) represent the total cost 

of providing the services associated with 

each final cost center. An allocation statistic 

specific to each final cost center is used 

as the basis ofapportioning the costs to the 

MEDICAL C4#t: 

patienits in each ofthe l)RCs. Forexm;niple , 

thile cost of nursing is allocated to patientt. 

based on a DRC-specific per dieni nursing 

weight which was derived through a study 

of the amount of nursing time spent witih 

patients in each DRG. Wl ile all of the alto. 

cation statistics possess some defects, they 

are designed to reflect more equitably the 

(lauiitity of aut iínstitittitoni's resources coi. 

sumed by the patients in each DRG. rTh, 

end result of the DRG cost miodel is the 

determination of the unit cost of treating 

patients in each lI)IG. 

5.3 Hospital Budgeting 

The full case mix cost accouniting approach 

has been applied to the budgetaryl 

process in 2 test hospitais. In the initial 

year, the unit costs (i.e., average cost per 

patient) in each DRCG were determined. In 

order to establish the following year's 

budget, it was only necessary to project the 

hospital's case mix and apply the appropriate 

inflation factors. Deviations from the 

budget due to case mix were immediately 

detected and the diagnostic and service 

areas experiencing significant deviations 

fromn established unit costs were isolated. 

The resulting unit costs in the test hospitais 

typically varied across DR9s by more 

than a hundredfold. The following DRGs 

illustrate this cost variation. 

DRG 

127-Ischemi c heart 

disease except 

acute myocardial 

intaretion with 

shunt or other 

major operation 

187-Gastric and peptic 

ulcer with 

gastric reseetion 

or other 

major operation 

with a secondary 

diagnosis present 

Tyupicl 1976i 

Unit Cost 

$9.934 

7,362 

F

Vol. XVIII, No. 2, Sn",,phl- steal 

112-Otitis miedia, 

c.lroulic .'il.sto)iditis 

or otosclerosis 

withotut any 

operation 

273-False labor 

.withoutt aliy op(eration 

- 67 - 

264 

Even within a specitic diagnostic area 

the DRGs provide a higlh degree of cost 

discrimi ination. For exampi)le, paitietits with 

a prinmiary diagnosis of urinary calcullus encompass 

4 DRGs with the tollowing typical 

1976 uinmit costs: 

Urinanj Calculus DRGs 

239-Without in oper- 

2- a0 d y i agnou sis 

240-IVitlhoutt an operatioin 

witli a 

s;ec( ¢{i;ir y 

dia;lí)nsis 

241-With miinor operation 

sIuch ;is 

cystoscopy or 

cathettcr to 

kiitdney 

242-With maijor operaition 

schlil as 

nephrotoiny, 

cystotoity or 

r'reterotoiiy 

Tilpical 1976 

L!ni .:ost 

$ 394 

774 

1,032 

2,2)93 

.ThL, , "en within this narrow diagnostic 

; artea the unit costs across DRGs varied by a 

factor of nearly 6. 

: An exaniple of a munit cost report 

' for DRG 121-Actute Myocardial lntiLrc- 

;/tion-appears in Figure 4. It compares 

the cost of treating AMI patients in 

; the same hospital across 2 differenít years. 

`; The box at the top of the report summainrizes 

-: the length of stay, charges, antd costs ex- 

· perienced by AMI patients in the 2-year 

period. The bottom portion of the report 

breaks down the costs experienced in 

! terms of the final cost centers of that hospi- 

. tal. For each item in the report both the 

absolute and per cent change across the 2 

CASE MIX ACCOUNTINC 

years a.re indicated. Such a report allows a 

hiospitail aininiistrator to isolate hoth the 

diagn(ostic and service areas wilere there 

are significant differences across years or 

relative to other hospitals if comparable 

data from other hospitals is available. Once 

the potential problem areas have been 

identified, the administrator can begin 

a more ineaningftul dialogute with the 

pIl)hysicians reslonsihle f)r the idenatified 

patients and services. 

5.4 Prospective Reimbursement 

Traditionally, most health insurers have 

reimbursed hospitals retrospectively on 

the basis of reasonable and allowable costs. 

While this model of reimbursement 

guarantees coverage for nmost hospital expenditures, 

it provides little economnic incentive 

to hospitals to control costs. Hospital 

prospeé.tivc-rei nlwrsement systems 

establish the rate of hospital reimbursemenit 

before the period over which the 

rate is to apply. The rewards and penalties 

inherent in a prospective systemn can 

potentially provide the inotivation for 

hospitals to become more cost effective 

without sacrificing the quality of medical 

cure. Under contract No. 600-77-022 from 

the lHealth Care FinancingAtAdininistration, 

the State of New Jersey is in the process 

of inmoving from a per dicm reasonable- 

'cost-based reimbursement system to a 

cost-per-case incentive-based system. 2 ' 

The Standard Hospital Accounting and 

Rate Evaluation (SHARE) system is the 

per diemn cost-based reimburseiment systein 

currently in use in New Jersey. Under 

the SHARE system, costs are grouped into 

31 cost centers according to uníi form definitions 

of futictional centers such as laboratory, 

rautiology adl the like. The iinpatient 

costs are then regrouped within each cost 

center into 3 basic categories: 1) nonplhysician-controlllale 

costs; 2) physician 

costs (e.g., physician and resident salaries 

and ftes); ard 3) other costs which are 

either not controllable by the hospital, or

CA' .IX DEFINITIO)N BY DRG 

- 68 - 

DRG 121 ACUTE MYOCARDIAL INFARCTION 

,MEDICAL CAKE 

1975 1976 CHANGE % CHANGE 

PATIENTS 267 ' , 315 48 17.98 

TOTAL COST 736811.12 882896.69 146085.56 19.83 

UNIT COST 2759.59 2802.85 43.25 1.57 

DAILY COST 199.24 201.26 2.02 1.02 

BED-DAYS 3698 4387 689 18.62 

AVERAGE LOS 13.85 13.93 .08 .55 

TOTAL CHARGES 820269.81 1042593.69 222323.87 27.10 

AVERAGE CHARGE 3072.17 3309.82 237.65 7.74 

COST/CHARGES . .89825 .84683 

FINAL 1975 1976 CHANGE % CHANGE 

COST CENTER UNIT COST UNIT COST UNIT COST UNIT COST 

OIETARY 127.46 131.44 3.98 3.12 

ADMITTING 20.48 22.45 1.97 9.61 

BILLING 90.40 97.27 6.87 7.60 

HOTEL 159.28 180.74 21.45 13.47 

NURSING 702.26 681.49 -20.77. -2.96 

INTERNS & RESIDENTS 72.23 71.51 -.73 -1.00 

MEDICAL RECOROS 22.01 26.39 4.38 19.87 

SOCIAL SERVICE 16.36 17.50 1.14 , 6.95 

NEW 80RN CARE UNIT .O0 .00 .00 .00 

INTENSIVE CARE UNIT 60.34 69.37 9.03 14.96 

CORONARY CARE UNIT 636.73 509.58 -127.15. -19.97 

OPERATING ROOM .00 .00 .00 .00 

RECOVERY ROOM .00 .00 .00 .00 

ANESTHESIOLOGY .00 .00 .00 .00 

DELIVERY ROOM .00 .00 .00 .00 

OIAGNOSTIC RADIOLOGY 70.77 86.87 16:10 22.75 

RADIOISOT0?rS 3.68 4.55 .87 23.55 

RADIATIC:; THERAPY .00 .00 .00 .00 

LABORA.TORY 276.29 330.85 54.56 19.75 

EEG .00 .00 .00 00 

EKG 58.22 55.07 -3.15 -5.41 

MED-SURG SUPPLIES 101.24 136.18 34.94 34.51 

PHYSICAL MEDICINE 68.74 84.43 15.69 22.83 

RESPIRATORY THERAPY 77.25 82.49 5.24 6.79 

IV THERAPY 39.82 49.30 9.47 23.79 

PHARMACY 156.03 165.38 9.35 5.99 

OIALYSIS .00 .00 .00 .00 

UROLOGY .00 .00 .00 .00 

KIDNEY TRANSPLANT .00 .00 .00 .00 

Fic. 4. Sample management reprt foracute myo(ardial infartion patients treated duringa 2-year period.

VoL XVIII. No. 2, Supplemnlent 

- 69 - 

which are not to be included in the deterinaintion 

of the SHARE reimbursemient 

rate, After salary rates have l)een 

"equalized" in order to avoid distortions 

due to geographic wage differentials, controllable 

c,sts are grouped into 3 clusters of 

cost centers to allow for trade-ot'ts in treatment 

modes. Hlospital peer groups are 

foried and cost screens are set tbr each 

cluster of ecost celiters iu ecaih peer , mnp. 

When a hospital's cliuster of cost centers 

itils these screeuis, each cost ceniter within 

each cluster is screened again within peer 

groups. Foriiulaute then ;arc appliet to determine 

any uinreasonalble costs in the base 

period that should be disaillowed. Atter a 

hospital's reasonable costs have bheen determined, 

a preliminary per diemi reinibursement 

rate is cletermitned for the hospital. 

The proposed SHARE per diemi is 

re-examined in an informal review, often 

followed by appeals to the Departiiient, 

and in some cases to the coutrts. Once the 

SHARE per dieli is finalized, it provides 

the basis of reimbursement except for 

certain year-end adjustments made for volume 

variance, actual inflation, and cost increases 

because of legal and approved 

management changes, such as Certificates 

of Need and Blue Cross contractuial 

adjustments. 

The essential feature of the SHARE 

methodology is "peer groutping." This regulator, 

ut Ihnique was developed in the 

utility tield, wherein like institutions are 

classified according to comparable attributes 

in order to apply "tair" standards and 

develop reasonable rates. Since hospitais 

:display inore complexity than single- 

!output industries such as electric, gas or 

'telephone companies, the technisque has 

been applied not at the aggregate level but 

at the cost center level. It should I>e noted 

that the forin or unit of payment under the 

¿:SHARE system is the per diem and that the 

basis of paymient is cost. 

Altlhoulgh the SIlAIBE reirilursemnent 

i,methodology in New Jersey attained sig- 

CASE MIX ACCOUNTING 

nificant accomplishments in bringing a 

certain amount of order to cost reporting by 

hospitals in the state, there are basic elements 

to the structure of hospital reinimbursement 

which remain problems and 

whikh SHARE is not equipped to address. 

The t'uñdamental problem in cost contaíinment 

is defining the appropriate taols 

tor measuiring reasonable efficiency and eftectiveness 

in the hospital setting. Accurate 

instruiments are needed to ineasture the 

level of productivity and efftctiveiess iii 

teris of both outputs aud inputs and to 

respond with the applropriite t'inancial incentives 

or disinceintives. The attempt to 

measuire and compare hospital efficiency 

at the cost-center level fails to recognize 

the role of case mtix iii determining hospital 

costs. Ditferences or lack of differences in 

hospital costs at the cost-ceniter level can 

ihe the result of different case mix compositions 

anid mnay not reflect dit'le elices in 

hospital productivity. Further, 'te quality 

or etfectiveness oft' care must hb; pn)perly 

identified and measured in order to 

eval uate accurately hospital productivity. 

Anialyses by cost center inmay tend to 

obscure rather than claritfy these problenis. 

The second limnitation of the SHARE approach 

is in the existence of financial disincentives 

to the delivery of high-quiality, 

appropriate and efficient care. It is no secret 

that the per diem fonn of payment 

tends to encourage longer lengths of stay. 

Furthermore, the length-of-stay incentive 

problem is compounded by the use of cost 

as a basis of reinibursement. The reasonalble 

mieasure of a hospital financial manager's 

effectiveness is his success in 

maximizing reimbursement. The more 

cost is increased, the greater the revenues. 

Costs can be skillt'ully redistrihuted over 

ditferent cost centers with the el.ct of escaping 

peer cost screens. Another problem 

for the regulator, and opportunity for the 

amnbitious financial manager, results from a 

systel¡ il¡ whicli somte paticnts L)pay costs 

and others pay charges, compoaunded be-

CASE V' X DEFINITION BY DRG 

- 70 - 

cause the ratio of costs to charges can be 

manipulated. Soine opportunities t'or business 

gamesinauship will doubtless be 

found in any reimbursement inethodology. 

The reimbursement system must identify 

appropriate costs and design a financial irnf 

centive to monitor aud minimnize such 

practices. 

Finally, the need f'or effective communication 

is perhaps the most serious problem 

ficing regulation of the health-care inidustry. 

To be effective, regulators and managers 

must build a language to the physiciansi. 

It is a reflectioni of the state ofthe art 

that appeals in the hospital industry are 

managed by lawyers and accountants and 

therefore examine problems of allocation 

and tinance. If the financial and medical 

information were merged, it would become 

possible to trace the relationship betNveen 

the physicians' decisions and their 

effects on costs. From this base, the proper 

questions can be-franied to deal equitably 

with issues of effectiveness, quality and 

efficiency in patient care. 

Thus, the SHARE methodology was not 

equipped to address the problems of 1) 

defining and measuring hospital case mix, 

productivity and effectiveness; 2) providing 

incentives for better management; 3) 

avoiding business gamesmenship; and 4) 

fióstering communications between the 

hospital financia.l systems and physicians. 

In order to deal with these problenms, New 

Jersey is developing a prospective casecost 

incentive--.. in.uursement system. The 

DRGs will provide the basis for the definition 

of case types and differential reimv'nrsement 

rates will be established for' 

each DRG. In order to develop the DRG 

reimbursement rates, it is necessary to ob-' 

tain the patient abstract, billing an'd financi.al 

information for the case mix cost accounting 

model described in the previous 

section. 

Since January i976, New Jersey has 

been recei-',,g medical discharge abstracts 

from all 104 acute-care hospitals in the 

state. In addition, 21 self-selected hospitals 

MEDICAL CARr. 

are also submitting their billing infonna. 

tion. Hospital cost information is obtained¡ 

through SHARE and is grouped inito 2 

categories: costs directly related to patient 

care, and institutional and other costs. 

Costs that were judged to be relatively 

fixed in the short run, and not directly related 

to patient care, were categorized as 

institutional costs. institutional cost centers 

include managerial services, facilities 

titainteutance, anid allied hiealth, inursing 

and graduate medical education costs. A 

separate methodology was developed to 

determine reasonable institutional costs 

comprised of a fixed sum and variable. 

amounts that relate to teaching comminitment 

and to the total amount of patientcare 

dollars. The patient-care costs were 

allocated to the patients in each DRG using 

the DRG-cost accounting methodology. 

Tie case mix cost accounting approach 

per-,lits the development of the institutional 

productivitygoals desired by thet 

regulators and the proper recognition of 

case inix complexity, differences across 

hospitals. Once the DRG costs have been 

determined, reimbursement standards !y 

DRG provide an appropriate set o'f goals, 

which are balanced to reflect the interests 

of both the public and the industly. The 

preliminary rate design has set a rate pler 

DRG for each hospital, composed of ¡proportions 

of the hospital mean case cost per 

DRG and the state standard (the mean caset 

cost per DRG across the sample of 21 hospitals>, 

The hospital and state proportion 

would always sum to unity. Thnus, for 

example, in the initial years of the systenm 

the rate per DRG might consist of 75 per 

cent hospital cost and 25 per cent state 

standard. Although the proportion of state 

standard would begin to approach 100 per 

cent over time, the early emphasis on tllhe 

hospital actual cost will próvide a reasoiable 

opportunity for institutions themiselves 

to make use of management information 

by DRC. Detailed managenment information 

by DRG will be provided to eaclh 

hospital and will be organized to help the 

i 

r5 

.J

'XVII. No. 2. Supplement 

- 71 - 

P)itals etfectively focus on areas of con- 

', in order to deal with problem areas. 

X:n order to realize fully the potential ofthe 

item, the process of implementation curt¡l>y 

is planned for 2 distinct stages. Stage 

¡l]ie initif' development, gives to each 

¡spital both the opportunity and responsibility 

for reacting to the management infrmation 

hy imnplementing steps Lo 

t:n dy inetficiency, expanding eiliciency, 

and opening an effective dialogue with 

tbeir physicianis. The motivation for dischauging 

thiis >butrdl is silpplied Iy prospective. 

incentive-based reiinbursemient. 

Ihus, in any given year, a hospital would 

retain the savings achieved by bringing in 

itscost per case uinder the prospective rate. 

.Sinm the fonn of payiment is per case rather 

than per diem, the unit of reimbursenent 

ao longer poses an incentive to increased 

:, 

:.e 

i 

CASE MIX ACCOUNTING 

lengths of stay. Recalculating the next year's 

prospective rate hased on the previous 

year's actual achievement serves the public 

interest by embedding the results in an improved-standard. 

The second stage in the process of implementation 

will involve properly crystallizing 

questions of inter-hospital significance. 

Health-care issnes will be placed in 

thieir DRG-specific and medical context, 

rather than the current collection of financially 

oriented appeals, thus permitting an 

cexamiminition of efficiency,

- 72 - 

MF:r :AL CAitE 

' .tr/ I!)N. Vol. XVIII, No. 2, Suiple>>Ieu, t 

6. A Case Mix Approach to Regional Planning for 

Acute Care Hospitals 

WHILE THERE ARE ONGOING USES of the 

DRGs in the are_;- of utilization review and 

hospital budgeting, cost control, and reimbursement, 

the application of the DRGs in 

the arue of hospital regional planning is only 

in its early stages ofdevelopmiiient. 5 Regional 

planning refers to the activity of organizing 

health-care resoutrces in a defined geograplhic 

regioui to aclhieve a desired state of 

aflairs in terms of the availability of health 

care of acceptable quality and cost. The 

primary thrust of the hospital planning activity 

has traditionally tocused on hospital 

facilities, primarily beds. Through legislation 

such as the Hill-Burton Act, much ofthe 

planning activity prior to the 1970s emphasized 

the adequiacy and distribuition of 

hospital beds to meet the needs of the population. 

However, the rapid increase in 

sophisticated'medical technology has resulted 

in a need to plan on a regional basis not 

only for hospital beds but also tor specific 

hospital services and equipment. Thus, the 

planning for the quantity and distribution of 

major new equipment such as computed 

tomography scanners or specific hospital 

services such as open heart surgery has become 

an integral part of the planning activity. 

Since certain types of services and 

equipment 4. ¡-,cessary to treat specific 

patient types, planning decisions will affect 

the case mix a hospital can treat. 

The modification by the planning process 

ofthe case mix that an individual hospital 

can treat will inevitably affect the case 

mix of the other hospitals in the region. For 

example, if a new service is added in one 

hospital, then that hospital will begin to 

treat additional types of patients. This will 

likely result in a decrease in the number of 

those types of patients treated at the other 

hospitals in the region. Further, the new 

service may cause the other capacities of 

the hospital (e.g., beds or operating rooi 

time) to be exceeded, requiring that the 

hospital cease to treat patients to whom it 

previously provided ¿are. The excess patients 

will have to be treated in tihe other 

hospitals in the region. Thus, the implica. 

tions of a planning decision can be co-n. 

plex and difficult to predict. A case mix 

approach to regional planning would have 

as its central focus the patients being 

treated and the demands they place on 

hospital resources. The role of each hospital 

in the region would be defined in terms 

of the case mix it treats. 

The basic hospital regional planníing 

prob)lem must consider a number of ftctors 

siniultanieously: 

1. All patients must'have access to the 

necessary hospital services. Access wvoull 

normally be defined in terms of travel time 

by ground transportation. An aceepltahle 

travel time coul(l vary by case type. depending 

on the prevelance of the patient 

type and nature of the condition (i.e., 

chronic, emergency, etc.). 

2. The bed and other capacities of the 

hospitails in the region must be sufficient to 

ineet the demand tor care. 

3. Patients must only be treated in hospitals 

with adequate senvices, equipmnent andl 

specialists. The type of hospital facilities 

and specialists necessary for proper care 

will vary with patient types, For exarn ple, it 

might be determined that patients younger 

than 14 years of age should only be treated 

at hospitals with pediatric units. 

4. Minimal quality-of-care standards for 

each patient type must be met. The 

qia&lity.l:carc staihldirdls cnuld l e ais basic 

as a minimal mortality rate for each patient 

type. For example, if the mortality rate tor 

acute myocardial infarction patients in a 

hospital exceeded soine percentage, such 

as 20 per cent, then the planning process 

muist consider requiring that hospital to 

cease providing care to those patients. The 

qtiality-of-care standard could become

;.'. 

'Vol. XVIII. No. 2. Siupplement 

- 73 - 

¡mlore sopltisticated and include process 

and ontcoanit nita miires. 

5. The cost of providing care to the different 

types of patients must be reasonable. 

The i..formation fromi projets such as the 

Ntew Jersey prospective reiml)ursemnent 

experinment can establish the cost of providing 

care to eacli patient type iu eavh hospital. 

If a hospital proves extremely cost ineffecvtie 

in providing care to vertain types of 

patibts. llthe the pliannúing process sii.st 

consir er : ther a iore c(ust-efll'ctive alternative 

exists. 

TlIiIs, the basict hospital rcgional planning 

problem can be describedl as a system 

of hospitais with the capacity to treat 

specitied nunibers and types of patients. In 

addition, there aure a set of basic cquality and 

access constraints that must be met. However, 

given that the capacity boiunds and 

access and quality constraints can be met 

by a number of alternative cotigurattiotns 

of case mix, the alternative requiring the 

least cx)st is likely to be the preferred alternative. 

Such a probleni can be analyzed 

nsing linear programming techniques. The 

DRC can provide the basis for establishing 

hospital case mix. A case mix-based linear 

progranming model of the regional plan- 

.ning probleim should prove to have many 

CASE MIX APPROACH TO PLANNING 

iminmediate practical applications. For 

example, suptpose a proposal to close a particular 

hospital had been made. The regional 

planning model could be used to 

suygest the most cost-effective means of 

¡tistributing the closed hospital's case mix 

to the other hospitais in the region while 

still maintaining adequate access and quality 

andl not exceeding the capacities of any 

hlospital. Alternatively, if there were an 

existing recommendation for the distribution 

of the closed hospital's case mix, then 

the model could he used simply to siminlate 

the impact oft'the case mix changes un 

the other hospitals and describe that impact 

in terms of various hospital parameters, 

such as occiupancy rates and reinmbursement 

levels. There are many other 

examples of the use of such a m¡odel. If 

there were significant changes in the composition 

of case types projected becvaise of 

a new cure or an outbreak of an inflctious 

disease, then the modeliould be usedl to 

evaluate the impact on the hospitais in the 

region. Thus, while a case mix regional 

planning model is only in it's early stagesxf 

development, it does hold the potential to 

be a powerful tool for the hospital planning 

process.

- 74 - 

M i.]DICAL C(ARIF: 

Fe,-n#arl lv0. VSm. VIIll No. 2, Supplement 

PATIENTS treated in an acute-care facility 

can Yvary considerably in both the duration 

and intensity of services required to provicie 

appr(>priate patient care. The relative 

amounts and types of hospital outputs 

utilized >y individttal patients are 

depelident on both the condition oí the 

patient and the treatnment process 

employed. By relating the demographic, 

diagilostic. aud therapeitic clharaiteristics 

of patients to the hospital outputs they 

utilize, a patient classification scheme can 

be developed wvhich provides the 

tramework for hoth the specification of 

hospital case mix and the measurement of 

the impact of case mix on hospital utilization 

and performance. The Diagnosis- 

Related Groups represent an attempt to 

provide such a patient classification 

scheme. As currently defined, the DRGs 

provide a manageable number of patient 

c;asses (383) that are exhaustive and minutually 

exclusive with respect to the types of 

patients seen in an acute-care setting. 

Further, the DRGs provide patient classes 

that are clinicially consistent and that have 

similar patterns of output utilization as 

measured by length of stay. 

The comparison of patient data across 

institutions or Droyviders will invariably reveal 

the exi-t nC:e of differential levels of 

utilizatior and performance. A comparative 

analysis by average length of stay, cost, 

or any other aggregate measure is not 

meaningful unless the impact of different 

case mix compositions can be determined. 

The DRGs can provide a framework for 

establishing the effects of case mix tas well 

as for identifying diagnostic areas with potential 

problems. The goal of most comparative 

analyses is to isolate problem 

areas s(; 4h.t corrective measure, ann be 

initiated. If prograims aimed ait itiaproviuig 

the performance ofthe hospital health-care 

system are to be successful, managers and 

7. Summary and Conclusions 

regulators must establish an effective 

dialogue with those responsible for the delivery 

of services, the physician comimunity. 

The DRGs provide the first step in suclh 

a dialogue since problems defined in the 

contextof DRGs are understandable fronm a 

clinical perspective. 

The various actual and potential applications 

ofthe DRGs in the areas of utilization 

review, hospital bucdgeting and cost coatrol, 

prospective reimbursement and regional 

planning emphasize the central role 

of the patient. By focusing on the types of 

patients being treated, programs responsihle 

for these activities will share a common 

conceptual basis even though they are 

concerned with different aspects of the 

health care system. While the applications 

to date llave been implemented to meet the 

immediate needs of the individual programs, 

future work will be directed toward 

exploring the potential of the DRGs in 

achieving better integration and.coordination 

of the different program goals and. 

activities. 

The current set of 383 DRGs were developed 

in light ofthe available data and its 

limitations at the time of their construction. 

As such, they representjust one imple mentation 

of an evolving series of patient 

classification schemes. As more comprehensive 

and reliable patient data become 

available and the practice of medicine 

changes, the DRGs must adapt to reflect 

these changes. To this end, it is felt that the 

technology and strategy used in forming 

the DRGs can be applied in the developnment 

of future generations of classification 

systems. Indeed, a major revision and 

evaluation ofthe DRGs will be undertaken 

as soon as ICD-9-CM data are available in 

sníflicient quantities. Further, work has 

begun in extending the approach into otlier 

areas of health-care delivery, in particular, 

ambulatory care.2' 

.1 1

:`". XVIII, No. 2. Stupplenent 

- 75 - 

1. Departsment of Heilth, Education, anid Welfire. 

A.UfOGRP patient classification .scheme and diagssois 

related groups (DRGs). Health care financing 

lntsn a;'nd contracts report series. Washington, D. C.: 

Departnhilc' nr I healthi Edueationt. uani Welfre. 1978. 

tDHEWV publication no. (HCFA)03011 979.) 

2. Avenil RF. Atpplication of case lnix cost acecountiLg 

to hospital heulgeting and prospective reimbursement 

pre*sented at tlhie 38t1h mneeting of the 

Acadenmy oF .\laiilgeii, eit. San Franicisco, August 

3. Bays CVy. Casc-miix tlilli.retnces l)thvween ¡ionpmiit 

and fibr protit hospitais. Inquinr 1977; 14:17. 

4. Ceite-r Gibr Ileahlthl Stildies. D)i:gnosis relited 

group relorts fir l>SHO diatil)iasts. New lILven: Yale 

University Center Hlealth Studies. 1978. 

5. Chemow R. A casemix approach to regional 

planning: an alpplication ol linear programniing. 

Ph.D. dissertatioín. New vlfave'n: Yale Unitersity. i¡ 

progress. 

·6. Coniínission oit Pnrf issional; a;d H lIosptital Ac'tivities. 

IEospitail adaptation ot> ICDA. 2ncd ed. Aun 

Arbor: Conuiuission tn Prolssisumal anel Hospital Activities. 

I968. 

7. Commiission onl Prt)fssional and Hospital Activities. 

Liengthl of stay inl PAS hospitalis. hy diagnosis. 

Ann Arbor: Cnmtiiissioín on Professional a;ud Hospital 

Activities. l976. 

8. Fetter RB. Mills R. Riedel DC. Thoinpsoi JD. 

IThe application of diagnostic spevcific cost ipnuiles to 

,.tost and reimhnursement control in hospitals. J S :d 

; Systems 1977;1:137. 

.; 9. Fetter RB. Thomipilson JDI, Milis RE. A system fir 

,'ost arid reinmblursement conitrl iiin hioslitais. Yale Biol 

;J 1976;49:123. 

10. Freenman JL. Applications of the AUTOGRP 

eomanputer system as a patient care monitoring mechanism 

fbr PSRO's. Presented at the 38th meeting of the 

i:Acaden» oi' Management, San Francisco, August 

¡1978. 

? 11. Hill BA. Principles of medical statistics. New 

&Yorlk: Oxford University Press. 1971. 

e'i: 

References 

REFERENCES 

12. Health Care Financing Administration. Implementation. 

evaluation and extension of a patient 

iJme mnitoring nechanisnim. Final report Conbtact 

240-75.0051. Baltimore: Health Care Financing Administration. 

October 1977. 

13. Public Health Service. International classification 

of diseases. eighth revision. Washinglon D. C.: 

U.S. Covernnient Printing Olfice. 1968. (Publication 

no. 1693.) 

14. Ktitagit.va EIM. Comipotlests o(f dlillrfenrte Ihetween 

two rates. J Am Stabsticd Assoc 13955;50:296. 

15. Lave IR. Lave LB. The extent oirole diltbrentiation 

among hospitals. lealth Serv Res 1971;6:15. 

16. l ave JR. Leinlnuarlt S. The.- ost and lngth of£a 

hospital stay. Ihi(litiry 1976;13:327. 

17. Lee ML. Wallace RL. Classilication ofdiseu-ies 

lor hospital cost analysis. Inquiry 1972;9:69. 

18. Lee ML. Wallace RL. Problems in estinmating 

llaltil)rKlit cost hietiions: ai llplicLatioii to hos¡itals. 

West Ecrm J 1973;11:350. 

19. Luke RD. Dimnensions in hospital ea;se mix 

meitsurement. Ilquiry 1979;16:38. 

20. Mills RE. Fetter RB. Riedel DC. Averill RF. 

AUTOCGHP: an interactie corlotitetr, ystain fior the 

analysis of health care data. Mled Care 1976;14:603. 

21. New Jersey State Department ofl tealth. A pnríspective 

reimbursenment system hase4 oln patiernt 

case-mix for New Jersey hospitals. 1976-1978. Annual 

Report. November 20, 1977. 

22. Ratlirty J. Pattms ofhospital use: an iatlysis of 

short-nin variations. J PolitiCal Econ 1971;79: 154. 

23. Ranoftky. AL. Utilizatiotn ofshort stay hIospitais: 

annual sumniary firrthe United Staltes, 1976. Rc'cksille. 

MD: National Center ftr Health Statistics. 1978. (Vital 

and heitlth statistics. Series 13. tno. 26) (DHE\W publieation 

no. (PtlS)78-1788.) 

24. Social Sevurity Administration. Progress report: 

developinent. testing aind evaluation ofa proslpective 

c¿tse-paymetnt reimhurseinent system. WVashington. 

D. C., SLocial Security Administration, 1978. 

25. Sonquist JA, Morgan JN. The detection of interaction 

effeets. Ann Arbor: Institute for Social Research. 

University of Michigan, 1964.

- 76 - 

> .CAL CARE 

t brNOar 1980, Vol. XVIII, No. 2. Supplement 

Appendix 

Diagnosis-Related Group Descríptions 

Major Diagnosis Category Diagnosis-Related Croups 

01: Infectious Diseases 

02: Malignant Neop4!isnm 

lf the Digestive 

Systemn 

03: Maliignant Neoplasm 

of the Respiratory 

System 

04: Magnlignat Neoplasm 

of the Skin 

05: MaUiignant Neoplasm 

of the Breast 

06: Malignant Neoplasm 

of the Female 

Reproductive System 

001 

002 

003 

004 

005 

006 

007 

00i 

009 

010 

011 

Infectious Disease (Enteritis, Diarrhea) with Age less than )16 

lnfectious Diseiase (Enteritis. Diarrhea) with Age greater than IS 

Infectious Disease (Viral Disease, VD, Meningitis) withoiut Sec 

ondary Diagnosis 

Infectious Disease (Viral Disease, VD. Meningitis) with Sec. 

Ondary Diagnosis 

Infectiutis Discase (Blood Infectioi,. TB, Salmonenllu) w.ithouit 

Surgery 

Infectiois Disease (Blood Infection. TB. Salmonell;a) wrth 

Surgery 

Cancer of tik? Mouth, Totigue. Large Intestinc. Liver. C.allIl.~I. 

der without Siurgery 

Cancer of the G1 System (Esophagus. Stomach, Pancreas, Sm.ll 

Intestine, Recétum) without Surgery 

Cancerofthe CI System with Surgical Procedure (Biopsy, Endnscopy, 

Local Excision, Centesis) without Secondary Diagnosis 

Cancerofthe Cl System with Surgical Procedure (Biopsy. Endous 

copy, Local Excision, Draining) with Secondary Diagnosis 

Cancer of the Cl System with Surgery (Gastric Resection. Colon 

Resection, Esophagus Resection) 

012 Cancer of the Rlespiratory System (Trachea. Lung, Larynx, Thorax. 

Mediastinum) without Surgery without Secondary Iiagnosis 

013 Cancer of the Respiratory System (Trachea, Lung, Larynx. 

Thorax, Mediastinum) without Surgery with Secondary Diagnosis 

014 Cancer of the Respiratory System with Surgical Procedure (Biopsy, 

Endoscopy, Excision of Lesion) without Secondary 

Diagnosis 

015 Cancer of the Respiratory System with Surgical Procedure (Biopsy. 

Endoscopy, Excision of Lesion) with Secondary Diagnosis 

016 Cancer of Respiratory System with Surgery (Lobectomy, Laryngectomy, 

Radical Resection) 

017 

018 

019 

020 

Cancer of the Skin except Malignant Melanoma without Seeondary 

Diagnosis 

Cancer of the Skin except Malignant Melanoma with Secondary 

Diagnosis 

Cancer of the Skin-Malignant Melanoma with Surgical Proce- 

'dure without Secondary Diagnosis 

Cancer of the Skin-Malignant Melanoma with Surgical Procedure 

with Secondary Diagnosis 

021 Cancer of the Breast without Surgery with Age less than 63 

022 Cancer of the Breast without Surgery with Age greater than 62 

023 Cancer of the Breast with Surgery without Secondary Diagnosis 

024 Cancer of the Breast with Surgery with Secondary Diagnosis 

025 Cancer of the Female Reproductive System (Uterus, Cervix, Vagina, 

Ovary, Fallopian Tuahe) without Surgery without Secondary 

Diagnosis 

026 Cancer of the Female Reproductive system (Uterus, Cervix, Vagina, 

Ovary, Fallopian 'rube) without Surgery with Secondar) 

Diagnosis 

027 Cancer of the Female Reproductive System with Surgical Procedure 

(D&C, Biopsy, Excision of Lesion) withomit Secondary 

Diagnosis

,VoL XVIII, No. 2, Supplement 

Major Diaglosis Categtory 

07: Mlalignant Neoplasm 

;il the Male 

Rcrodhaehctive Systemi 

08: Malignant Neoplalissl 

o. tthe Urinary System 

"09: Malignant Neoplasm 

of O

- 78 - 

CASE b'IX DF.INITION BY DiRG MEDICAL. CAR.: 

Mliajor L)iaglno>sis Category 

1I: -BiUnig Neiol)hilson or 

the Feinale Repro. 

ductive System 

12: Benign Neoplains of 

O)ther Sites 

13: Diseases of Thyroid 

and Other Endocrine 

Clands 

14: Diabetes" 

15: Nutritional and Other 

Metabolic Diseases 

16: Diseases of the Blo>d 

and Blhod Forming 

¢ )rgai., 

(153 

054 

055 

056 

057 

058 

í)59 

Blen¿iKg, 'Frnur (P:pqilhluna. Polyp) of tihe Uteruns. Vagina;, Vu'lva 

x\ hlmut St%;vindary Diagnosis 

Benign Tumor (Pípilloma. Polyp) of the Uterus. Vagina. Vulva 


Benign Tumor (Fibroma) of the Uterus. Ovary without Surgery 

Benign Tunmor

.. 

Vol. XVIII, No. 2. Supl|lemcent 

.Major Dinagnasis Category )Diignosis-Beláted Groups 

17: Psychoses Not 

Attributed to Physical 

Conditions 

18: Neuroses 

19: Alcoholic Mental 

Disorder and Addiction 

20: Other MentaiiJ Disorders 

21: Diseases of the 

Central Nervous 

System 

Diseases of the 

Pe-'rh-ral Nervous 

S?-:em 

Discases of the 

Eye 

083 &lediterancan Anemia, ilcanpholilia witliout Stargery withlout 

Secondary Diagnosis or with Mlinor Secondary Diagnosis with 

Age greater th;rn 10 _ 

084 Disease of Blood Hemoglobin without Surgery without Secondary 

Diagnosis or with Minor Secondary Diagnosis 

085 Disease of the Blood (Anemias). Blood Forming Organs (Spleen) 

without Surgery with Major Secondary Diagnosis 

086 

087 

Disease of the Blood (Anemias). Blood Forming Organs with Surgery 

with Age 2-52 

Disease of the Blond (Anemias). Blood Forming Organs with Surgery 

with A&ge less tiian 1 or greater thli. .53 

I88 Schizophirenia (Pamnoid, Catatonic, Unspecified) Invohltional 

Meldancholia with Psychiatric Sertice 

089 Schizophrenia (Parainoid, Catatonie. Unspecified) Involutional 

SI.ialacrholiil without Psychiatric Servive 

090 SSchizophrenia (Alfi.ective, Acute Episotde), Mlaitic-Depressive 

Psychosis 

091 Neurosis (Anxiety, Hysterical, Phobic, Hypochandriacal UnslXcified) 

092 Neurosis (()Ohsessiv-Conlpulsive. Depressive), Persoiality- 

Disorders 

093 Alcoholisnm without Secondary Diagnosis or with Minor Secondauy 

Diagnosis 

094 .Alcoholismn with Major Secondary Diagnosis (Liver Cirrhosis. 

I)elirium Tremens, Other) 

095 Drug Dependence. Physical Disorder (Probably Psychiatric 

Origin), Cephalgia 

096 Psychosis, Non-Psychosis Related Brain Condition 

097 

098 

099 

100 

101 

102 

103 

104 

105 

106 

107 

108 

109 

110 

111 

- 79 - 

APPENDIX 

Epilepsy, Migraine. Brain Disurder (Unspecified) withoit Surgery 

without Second:ary Diagnosis 

Epilepsy, Migraine, Brain Disorder (Unspecified) without Surgery 


Multiple Sclerosis, Paralysis Agitaas, .leningitis, Hemiplegia 

withont Surgery 

Disease of the Central Nervous Sysfeln with Surgical Procedure 

(Nerve Block, Other) 

Disease of the Central Nervous System with Surgery (Laminectony. 

Spinal Fusion Ventricular Shunt) 

Facial Paralysis, Neuralgia (Trigeminal, Other Unspecified) withotit 

Surgery 

Sciatica, Polyneuritis without Surgery 

Disease of the Median Nerve with Surgery 

Disease of dithe Peripheral Nerves except Sedian with Surgical 

Procedure (Nerve BIlck, Other Unspecified) 

Disease of the Peripheral Nerves except Median with Surgery 

(Spinal Cord, Nerve Roots) 

Cross Eyedness, Cataract, Cyst of the Eyelid withoit Surgery 

Glaucima, Corneal [imtlanimation/Ulveration, Disease of the Iris, 

Retina witliout Smtrgery 

Disease of the Eye with Surgical Procedure (Muscle Repair of 

Eyelid, Other) 

Disease of the Eye with Surgical Procedure (Removal of Lens, 

IIL! .on into Sclera 

Dinsi.eS of the Eye witIh Surgicail Procetdurte (Reattaehment of 

Hetina. ielpair of Cornea)

CAbS . MIX DEFINITION BY DRG 

Major Diagnosis Category 

24: Disease of the 

Ear and Mastoid 

Process 

f25: Hypertensive Heart 

Diseases 

26: Acute Myocardial 

I nfrction 

27: Ischemie He.art Diseaises 

Except A1MI 

28: Arrhvthnmia and Slowed 

Conduction 

29: Heart i adlure 

30: Carditis, Valvular 

and Other Diseases 

112 

113 

114 

115 

116 

117 

118 

119 

120 

121 

- 80 - 

Diagnosis-Related Groups 

MEDICAL CARE 

Disease of the Middic Ear (Iiulanimation, Chronic Mastoid Bone 

Inflammation) without Surgery 

Disease of the Inner Ear (Infiammation, Menieres Disease) without 

Surgery. _. 

Disease of the Ear with Surgical Procedure (Incision of Membrane, 

Removal of Adenoids, Other) 

Disease ofthe Middle Earwith Surgery (Removal of Bone, Repair 

of Membrane) 

Disease of the Ear with Surgery (Removal of Mastoid Bone, Excision 

of Middle Ear, Other) 

Hypertensive HeartiDisease without Surgery without Secondary 


Hypertensive Heart Disease without Surgery with Major Secondary 

Diagnosis 

llypertensive Heart Disease (Fatal) with Kidlney Involvemient 


Hypertensive Heart Disease with Surgery 

Disease of the Heart-Acute Myocardial Infaretion 

122 Disease of the Hieart, Ischemia (Blood Deficiency) except AMI 

witlhout Surgery withouit Secondary Diagnosis 

123 Discase of the Heart, Ischeinia (Blood Deficiency) except AMI 

without Surgery with Minor Secondary :)iagnosis 

124 .sease of the Heart, Ischemia (Blood 'eficiency) except AMI 

witlmit Surgery with Major Secondar Diagnosis 

125 Disease of'tlie Heart, Ischemia (Blood Dei.ciewcy) except AMI 

with Cardiae Catheterization 

126 Disease of the Heart, Ischemia (Blood Deficiency) except AMI 

with Surgical Procedure (Endoscopy, Insertion of Electronic 

Devíice) 

127 Disease of the lieart, Ischemia (Blood Deficiency) except AMI 

with Surgery (Shunt, Otlier Major) 

128 Disease of the Heart, Irregular Heart Rhythm, Slowed Conduction 

without Surgery without Secondary Diagnosis or with N inor 

Secondary Diagnosis 

129 Disease of the Heart, Irregular Heart Rhythm, Slowed Conductio.n 

without Surgery with Major Secoandary Diagnosis 

130 Disease of the Heart, Irregular Heart Rhythm, Slowed Conduction 

with Replacement of. Heart Device or Cardiac Catheterizatiun 

131 Disease of the Heart, Irregular Heart RhytIm, Slowed Conductión 

with Insertion of Electronic Heart Device 

132 Disease of the Heart, Failure (Poor Function) without Surgery 

133. Disease of the Heurt, Failure (Poor Function) with Surgery 

134 Disease of the Heart, Inflammation, Valve Problem without Surgery 

without Secondary Diagnosis or with Minor Secondary 

Diagnosis 

135 Disease of the Heart, Inflammation, Valve Problem without Surgery 

with Major Secondary Diagnosis 

136 Disease of the Heart, Inflammation, Valve Problem with Cardiac 

Catheterization without Secondary Diagnosis or with Minor Secondary 

Diagnosis 

137 Disease of the Heart, Inflammation, Valve Problem with Cardiac 

Catheterization with Major Secondary Diagnosis 

t

i,..X VIII. No. 2. Supplemeiint 

- 81 - 

Major Diagrioosis Category Diagnosis-Ielated Grotups 

:31:- Cerebrovascular 

:: Diseases 


Vascular System 

33: Pulmonaary Emobolismn 

34: Phlebitis aud 

'1 tlro.ial>phlebitis 

35: 

Hemorrhoids 

l{ypertruphy uf Tonsil 

and Adenoid 

*37: Acute Ulpex'r Respiratory 

T'ract Infection and 

* Influenza 

.38. Other Discas-tir of the 

Uppe~r Respitatory Tract 

APPENDIX 

138l Disease of the Heart, Inilamniation. Valve Probleni with Surgery 

(Valve Replacement. Other Major) 

139 Circulatory Disorder bi the gíin. Occasional Blood Deficiency 

without Surgery without Secondary Diagnosis or with Minor Seondary 

Diagnosis 

140 Circulatory Disorder of the Brain. Occasional Blood Deficiency 


141 Blood Clot in Brain Obstructing Circulation without Surgery 

withiint Secondary Diagnosis or with Miinor Secondary Diagnosis 

142 Blotd Clot iio Brain Ol>structing Circulation witfhout Surgery 

with Major Secondary Diagnosis 

143 Brain Hemnorrhage (Stroke) without Surgery without Secondary 


144 Brlin llenmorrhage (Stroke) without Sorgery with Mijor Secundary 

Diagnosis 

145 Circulatory Dysfunction in Brain with Surgery 

146 

147 

148 

149 

150 

151 

Disease of the Circulatory System. Inflamnimation of the Lymph 

Glands. Varicose Veins (Legs). Raynauds Disease without- 

Surgery 

Discase ,of the Circuilatory Systeni (Hardening of Arterial WValls. 

Arterial Blood Clot) withont Surgery without Secondary Diagnosis 

.or with Minor Eecondary Diagnosis 

Disease otf the Cirt -datory System (Hardening of Arterial Wialis,. 

Arteri¿al Bi>oul C: t) witho(it Surgery with Muajor Se'ondary- 

Diagnosis 

Diseiase of the Circulatory System with Surgical Proced*re (Excision 

of Varicose Veins, Other) with Age less than 51 

Disease of Vascular System with Surgery (Excision of Varicose 

Veins, Other) with Age greater than .50 

Disease of Vascular Sytem with Surgery (Excision of Nterve, 

Vessel) without Secondary Diagnosis 

152 

153 

Disease of Vascular System with Surgery (Excision of Ni-&ve. 

Vessel) with Secondary Diagniosis 

Disease of`Vascular System with Surgery (Arterial Recunstruction, 

Ampitation of Extremity) 

154 

155 

156 

157 

158 

159 

160 

161 

Blood Clot ., tlhe Lung without Secondary liagnosis or with 

Minor Secondary Diagnosis 

D. .ad Clot of the Lung with Major Seconrdary Diagnosis 

IBl"uinimation of the Veins, Blood Clot without Secondary Diagnosis 

or with Minor Secondary Diagnosis 

Inflammation of the Veins. Blood Clot with Major Secondary 

Diagnosis 

Hemorrhoids 

Enlargement of the Tolnsils/Adenoids 

Acute Uppl)r Blespiratory Tract Inletio. m Ilnfluenza with Age less 

thano 45 

Acute Upper Respiratory Tract Infection Ilnfluenza with Age 

greater than 44 

162 Disease of the Upper Respiratory Tract except Acute Upper 

Respiratory lnfection and Influenza without Surgery

CAS E MIX DEFINITION BY DRG 

.Major Dininosis Category 

39: Piineinofnia 

40: Bronchitis 

41: .Asthma 

42: ()tlher Luing and 

Plecirl I)isea.es 

43: Diseases of the Oral 

,:ait'. Salivary Glandtis 

a,.1tl Jaiw 

44;: ;astrie aud Peptie 

U leer 

45: Upper Castn- Intestinal 

Diseases except 

Gasutrie and Peptie 

Ulcer 

46: AI)Ilpxtlivitis 

- 82 - 

Diagniosis-Related Cronips 

SIEDICAL CARE 

163 Disease of the Upper Respiratory Tract with Surgical Procedure 

(Biopsy, Visualizations of the Nasal Septuni) 

164 Disease of thc Upper Respiratory Tract with Surgery (Nose Re. 

construction, Inciniun andPraiinaige of Sinuis) 

165 PPneunionia with Age less than 31 

166 Pneiuanoniia without Surgery without Secondary Diagnosis wvith 

Age greater than 30 

167 Pneumnonia witlihout Surgery with Secondary Diagnosis with Age 

greater th el¡i 30 

168 Pineunionia with Surgery 

169 Bronchitis with Age less than 46 

170 Bronchitis without Seclidary Diagnosis or with Minor Second. 

alry i)iaglosi.s with Age gr.-ater thai .i5 

171 o"nchitis with Major Secondary Disgnosis withl Agt greater 

thall 45 

172 

173 

174 

175 

176 

177 

L78 

179 

180 

Asthiia with Age less tlhan 31 

Asthina without Seconildary Diagiíosis with Age greater thal :(1 

.Astlthma wxith Sevouidary liaginosis with Age greater thlia, 30 

Lung Collapse. Pleurisyg, Punlioniary congestion withsoit Surgoery 

Emtphysema, Eímpyema, Abscess. Acute Swelling without Stirgery 

witholit Secondary Diagnosis or with Minor Secondar- 

C .lgsosiss 

.kll>Wsciii. ELi)l>yetni;, Al>bsess, Acute Swevlli g witout Sil- 

^,ery with NlaUjr Secondary Viagnosis 

Dlisease of the Lutng and Pleaura with Surgical Proced¢iu (Bronchoscopy, 

Cliest acisio, Other) withiout Secoindary Diagnosis 

D)isea s the e Lung and Pletura with Surgical Procedure (Bronchooscopy, 

Chest litcision, Other) with Secondalry Di)i;uosis 

Disease orf the Lunlg adil Pleuira with Surgery (Removial of Lole: 

Other Major) 

181 .liinor Protileis of thle Teeth 

182 Mliaor Prolalemiís of the Teeth (Jaw, Salivary Clanis, Other Oral 

Solt 'risstue) 

183 

184 

"45 

186 

187 

188 

189 

190 

191 

192 

Stomach Ulcer without Surgery without Seconlary DiagnPosis 

Stomach Ulcer without Surgery with Secondary Diagnosis 

Stom;ich Ulcer with Surgical Proceduire (iop>sy, Visutlizatiou. 

Other) 

Stomltach Ulcer with Surgery (Remnoval of Portion of Stomach., 

Other tMajor) without Secondary Diagnosis 

Stomach Ulcer with Surgery (Remnoval of Portion of Stoinmach, 

OCther Slajor) with Secumnlary Diaguosis 

Upper Gl Disease Except Stomach Ulcer without Surgery without 

Seco)ndary Diagnosis 

Upper Cl Disease Except Stomach Ulcer without surgery with 

Secondary' Diagnosis 

Upper CI Disease Except Stuiomch Ulver with Surgical Procedure 

(Visualization, Other Minor) without Secondary Diagnosis 

Upper GI Disease Except Stomnach Ulcer with Surgical Procedure 

(Visualization, Other Minor) with Secondary Diagnosis 

Upper GI Disease Except Stomach Ulcer with Surgery 

¡1$1 

f Appendicitis (without Peritonitis) without Secondary Diagnosis 

1IJ4 A¡ppldivitis (witlh>iSt l'erito>itnil) withi Sccumsiary Diagnuosis 

195 Appendicitis (with Peritonitis, Other) without Secondary 

Diagnosis s

T 

XVIII. No. 2. S.pphlts"t 

.i¿jor Diagnosis Citegory 

Hernia of tlhe 

-Axidouiínal Cavity 

48: Esteritis. Divertievula, 

a¿iad Faniction[al Disorlers 

olfthe Intestineic 

*.49: D iseaises of thie Antis 

50: SlisteeIlah.ols Disea.ses 

ot' the litestine auli 

I>eritona.llie 

51: Disea.ses ol' the 

Liver 


't.all¿«hlhhr austl Bilh, 

Dlat't 

1t> A>pil)edicitis (witla Peritonitis. Otiher) with Secundary 

Diiagnusw. 

197 

198 

199 

200 

201 

203 

204 

205 

206 

207 

208 

2(09 

210 

- 83 - 

Grutips 

AbIdomainsal Hernia with Age less than 15 

lnginua!ul fHieriia (without Obstruiction) with Age greater than 14 

a.ld less thalan 65 withoiut Secondiry. Diaginosis 

liguina;l Hernia (withoiat Oblstrtuction) with Age greater than 14 

antd less tihan 65 with Secondary Diagnosis 

Alblonminal Hernia Except Simple Inguinal with Age greater than 

14 atud less thin 6.5 withosit Snrgery 

AlIlaminal lernia Except Simuple Iagaiuaal witlh Age greater than 

14 aud less tlansal 65 with Minor Surgen. 

Aldlomiituial I lernia Excel)t Sinple Inguinal with Age ireater thtan 

14 asad less than 65 with Mlajor Surgery 

Ahdsomi.al llernia with Age grecater thlan 64 wvithoitat Sargery 

AldlomialI llernia witih Age greater thllan 64 witil Ninor SuIrgery 

Ahblomiial liernia wviti Age greater than 64 with Mlajor Surgery 

Faiunctional Disorder of the Intestine without Surgery 

Intestiiial Ponuching, Regioial Enteritis. Ulcerative Colitis withiait 

Saurgery 

lntestinaJl PocilnvIiag (Functionlal Disorier) wvith Minor Surgery 

witliolnt Sevoiuldarv D)iagnosis 

lntestisial Pouching (Functional Disorder) with Mlinor Surgery 

with Secolidary Diiagiosis 

Iiltestieail P¡ witi Mlajor Surgery 

(lieê lichi, )thleOr) 

211 Disease ol'the Anos without Secondary Diagnosis 

212 Disease ofl'the Aruis with Secondairy Diagh)osi6 

A.PP.EN DIX 

213 

214 

Misceellaneo.us Disease of the Intestine and Abdominal Lining 

withl Age less th.ali 56 withotut Snrgery 

Dliscell.eons 

Disease fIt' thc lntestiLne and Abdominal Living 

with Age greater tiian 55 withont Surgery without Secondary 

Diagnosis 

215 

216 

Mliseelliaeoa.~ Disease of tile Intestine aud Abdominal Lining 

withl Age greater than 55 withouit Snrgery with Secondary 

Di¿agnlosis 

Ilisell&iai.e'ois i)isea.se *. thei Intestiice iantd Allndominlal Lining 

witl Snargical Procedure (Local Incision, Excision) 

217 

218 

.liscellaneous Disease of the Intestine and Abdominal Lining 

wilth Visinalization, of the Intestlie withont Secondlary Diagnosis 

.Mliscellaneous Disease of the Intestine and Abdominal Lining 

with Vistialization of the Intestine with Secondary Diagnosis 

219 Miscellaneous Disease of the Intestine and Abdominal Lining 

with Major Surgery withoit Secondary Diagnosis 

220 Miscellahieous Disease of the Intestinle and Abdominal Lining 

with Major Sturgery with Secondary Diagnosis 

221 Ilepatitis. (l..lcetius. Sermni) Sualacute Necrosis of the Liver 

witil Age Icss thlsan 41 

222: lelpaltitis (Ihal¿etionss Seruis) Sblacuate Nierrosis ol'f the Liver 

witil Age grea.ter than *10 

223 I.iver Cirlhlosis withoílt Secoi;d.iry Diagainsis or with Slinor Secni)dar>' 

Di.agnosis 

224 Liver Cirrhosis with Major Srcondiary Diagnosis 

225 o:a..' of the Giallli;.Ider .atl btil Dc thi Sargry with 

Agre kh-, Ilsial 51 

226 1Ii:e,

C' .., MIX DEFINITION BY DRG 

Major Diagnosis Category 

53: Diseases of ~he 

Pancreas 

54: DcDiseases of thec 

Kidney muid Ureter 

55: Urinary Calculns 

56: Cystitis and Other 

Urinary Diseases 

57: Disease of the Prostate 

227 

228 

229 

- 84 - 

Diagnosis-Related Croiups 

Disease of the Gallbladder and Bile Duct with Surgery witlho1, 

Secondary Diagnosis 

Disease of the Callbladder and Bile Duct with Surgery with 

SecoMdary Diagnosis with Age less than 65 

Disease orfihe Callbladder and Bile Duct with Surgery with 

Secondary Diagnosis with Age greater than 64 

230 Disease of the Pancreas without Surgery 

231 Disease of the Pancreas with Surgery 

232 

233 

234 

235 

236 

237 

238 

239 

2.11 

241 

242 

¿MEDICAL C,,,M 

Disease of' the Kitdley aud Bladider witlhitit Surgery withi.ti Se

·. ¥XVIII. No. 2 Supplement 

)i, "'" 

, Major Diagnosis Category 

58: Disease of the Male 

" Reproductive System 

59: Disease of th.: Femiale 

Reproductive Systenm 

60: Diseases of 

the Breast 

61: Abortion 

62: Obstetricai Diseases 

of the Antpartilm and 

Puerperiuna 

63: Normal Delivery 

64: Delivery with 

Complications 


Skini anid Subel.lasiv ems 

'Tissue 

2,56 I)is¢i-se ol tih 

Prostate) 

Prostate with Surge

- 86 - 

CASE MIX DEFINITION BY DRG NMEDICAL CARE 

MIajor Diagnosis Category 

66: Arthritis 

67: Derangenment and 

Displacement of 

Intervertebral Disc 


Bone and Cartilage 

£9: Other Disease of the 

Mlusculo-Skeletal System 

70: Congenital Anonmalies 

287 

288 

289 

290 

291 

292 

293 

294 

Skin Inflammation, Abscess, Eczema. Reddened Skin with Sur. 

gery without Secondary Diagnosis 

Skin Inflammation. Abscess, Eczema. Reddened Skin with Suir. 

gery with Secondauy Diagnosis 

Psoriasis Eruptive Skin Lesions, Chronic Skin Ulcer 

Arthritis without Surgery with Age less than 65 

Arthritis without Surgery with Age greater than 64 

Arthritis with Surgery (Excision of Bone, Joint, Mlembrane Sur. 

gical Joint Fixation) 

Artha itis with Surgery (Juint Incision. Spinal Fusions, Excision of 

Tissue Between Vertebrae) 

Arthritis with Surgery (Repair and Restoration of Joint, Removal 

ol' Mitil>nllbm, Ixtwee.ii Vertelbrae) 

295 Disorder and Displacement of disc Between Vertebrae withouit 

Surgery 

296 Di:uorder and Displacement of Disc Between Vertebrae with 

Surgery 

297 

298 

299 

300 

301 

302 

303 

304 

305 

306 

307 

308 

I)iiuioisl~cl;I~tl Groitpn: 

Rheuimatism and Inflammaition Tissue Covering Bone, Other 

Minor Bone Disease without Surgery 

Disease of the Bone. Inflammation of Mlarrow (.Acite, Chronic). 

,'ungy Bone, Un.tided Fracture without Surgery 

Disease of the Bore. and Boine Tissue Lining. with Surgery (Excisioin 

Bolie Linitng, Repair ol' Other Joiunt) 

Disease of the Boae and Bone Tissue Lining witli Surgery)' (Jint 

Ihcision, Bone Excision, Bone Fusion) 

Disease of the Bone and Bone Tissue Lining with Surgery 

(Amputation, Hip Restoration. Other Major) ' 

Inflammiaiition of the Component Parts of the Joints, Curvature of 

the Spine, Deformed Foot without Surgery 

Backache. Difluse Disease of Connective Tissue. Inflarnmation 

of Musele without Surgery without Secondary Diagnosis 

Backache. Difiuse Disease of Connective Tissue. Inflammation 

of Mlusele without Surgery with Secondary Diagnosis 

Inflanimnation of the Comipornent Parts of Joints with Detrmnity 

(Palm, Finger. 'oe) with Surgery 

Other Disease of the Muscle and Bone (Major) with Surgical 

Procedure 

()ther Disease of the' Muscle and Bone (lajor>:with Surgery 

(Removal, Repair of the Small Joint, Bone) 

Other Disease of the Muscle and Bone (Major) with Surger' 

(oining Vertebrae,. Other) 

309 Birth Defect (Bone. Stomnach, Testicle) without Surgery 

310 Birth Defect (Heart, Kidney. Other MSajor) without Surger>) 

311 Birth Defect (Testicle. Skin, Stomach. Other Minor) with Surgery 

312 Birth Defect (Heart Valve, Other Unspecified Heart Site) with 

Surgical Procedure (Cardiac Catheterization) 

313 Birth Defect (Palate, Lip. Hip or Other Extremity) with Surger>' 

(Repair of .Mouth. Fixatio of liip) 

314 Birth Detlct (Heart Valve, Other Unspecified Site) with Surgery 

(Heart Valve, Septal Repair) 

315 Congenital Diseases (Tetralogy of Fallot, Atrial Septal Deftect. 

Hypospadia, Other) with Surgical Procedure (Catheterization, 

Repair of Urethra) 

316 CoiCogenIitlil D)iscases (Tetralogy of Fallot, Atriil Septal Dei.ct. 

Other) with Surgery (Valve, Septum, Shunt)

- 87 - 

'i1. XVIII. No. -. Supplemnent APPENDIX 

Major Diagnosis Category Diagnosis-Related Grotips 

Normal blature 

Newbom 

1: Ccrtain Diseases aid 

Coulditiolns Peculiar 

to Nelhmxío Iniants 

13: Signis anMd SYnl)ptouns 

>Pertaiuing to the 

Nervous, Respiratory, 

and Circulatory Systemos 

74: Sign.s anud Synlptolms 

Pertainimng to the 

Castro-Intestinal 

and Urinary Systerms 

15: Miscellanelous Signs. 

Symptonms,i. und IlI- 

Defined Conditions 

?/6:' Fractures 

317 iirthil Deect (Spine. Cullet. Large Uowel) with Sturgery 

318 Normal Full Trmi Netvborn 

319 

320 

321 

322 

323 

32.1 

325 

326 

Wcll Baby Care (Preginanuy greater than 9 nmonths), Other Mlinor 

Disease or Condition nl' the Nevwhorni Inlifat 

Ihnmaturity). Hyaliit ne ite Disease. Other M;jor Disease 

or Co(dlition o the Ilnf.mt without Secondlarv Diagnosis 

Infiliatuirity. Hyaline Membrane Disease, Othler Major Disease 

or Conmdition of the Inilint witi Secondary Diagnosis 

mlicvatimit ifl' s Nervotis. Respiratory, Circulatory System Disease 

w itlihut Surgery witlihot Seconidary Diagnosis 

Convulsions. Fainting. Nosebleed, Chest Pain withoutt Surgery 

withl Secinidary Diagnosis 

lra iu I)iso rdlr ol' I)ixzin.ss. Shorti.ss (l' Brrea lth. (:ougiling u¡> 

Bllasid \wtiout Sourgery switlh Secoindar. Diagnosis 

Indications of Nervous. Respiratory, Circulatorory Systen Disease 

with Surgical Procedure 

Indlications of Nervous, Respiratory. Circ'ulatory System Disease 

with lMajor Suirgery 

327 niudications ofCGastro-lntestinal. Urinary Systein Disease withouit 

Surgery withoit Seconuedary Diagnosis 

328 ludicíations of Gastro-lntestinal, Urinaiury System Disease withoiut 

Surgery with Seconclary Diagnosis 

329 Indications of Gastro-lisl.thnal. Urinary Systeia Dis.e¿ice wit1h 

Surgical Procedure (Visual laispection. Other) 

330 IIndications of C.astro-Intestinal. Urinary System Disease with 

Surgery (Abdonminal, Other. Major) 

331 

332 

333 

334 

335 

Sterility (Male, Female), Admission for Ohservation without 

Surgery 

Chemical Imnalance, Headache. Fever. Other ¡ll-Delined Indication 

of Disease without Surgery with Age less than 15 

Chemical Imbalance, Headache, Fever, Other Ill-Deflined Indication 

of Disease without Surgery with Age greater than 14 

Miscellaneous Indication of Disease with Surgical Procedure 

(Visual Inspection, Other) 

Miscellaneous Indication ol' Disease with Surgery (Abdominal 

Surgory, Removal of Uterus, Other Major) 

336 Fracture (Skull, Face, Forearm, Leg, Foot. Hand) witihout Surgery 

with Age less than 30 

337 Fracture (Skull, Face, Forearm, Leg. Foot, Hand) without Surgery 

with Age greater than 29 

338 Fracture (Spine. Rihbs. Bone of the Upper Arm) without Surgery 

with Age less than 65 

3.39 Fracture (Spine, Ribs, Bone of the Upper Arm) without Surgery 

with Age greater than 64 

340 Fracture (Thigh Bone, Pelvis, Multiple) without Surgery 

341 Fracture (Nose. Foreannm. Hand, Lower Leg, Foot) with Surgical 

Procedure (Closed Redluction) withouit Secondary Diagnosis 

342 Fracture (Nose, Forearm, Hand, Lower Leg, Font) with Surgical 

Procedure (Closed Reduction) with Secondary Diagnosis 

343 Fracture (Lower Jaw, Upper Ann, Ankle) with Surgical Procedure 

(Closed Reduction. Open Reduction of Face) without Secondary 

Diagnosis 

344 Fracture (Lower Jaw. Upper Armn, Ankle) with Surgical Proce-

ASE: MIX DEFIl'l'lTION BY DII(; 

,?: I)iant:atiouis ailL 

O)thler MIuscuIo-Skeletal 

Injuries 

78: Internal injurie.. 

ofl thc Crailiiin. ('hest. 

aiad O()ther ()rgRan 

79: Ol)peln Vomim!.I -ind 

Supt>erlcial Injuries 

80: Bunms 

\ 1 ;it or 1 )i;tgs l »á ( :ai( -g§ 1 r 

81: Complications of 

Medicail and Surgical 

Care 

- 88 - 

I)iaR m~isis-l a·l;te¿tpe,,uI 

ReHductionl. External Fixationl. (ther) 

346 Fracture (Allkle. Leg Bones) %withi Suargery (Open Retudli.¡, 

External Fixation, Other) 

347 Fracture (Thiglih ioe. Pelvis) with Surgery> (O>pen HeduetFi.. I: E 

tenial Fixation. (Other) 

S148 'Fra.tur(, w¡ithi .\lijor Siurery (Aiiputation¡. Istoritiml ,,It 1h 

Joinit ()ther Major) 

349 

:.350 

:351 

352 

353 

3:54 

355 

357 

.358 

.359 

.360 

361 

362 

3i3 

364 

365 

366 

SF:I)L} AI. ( 'lti 

I)ishaati>íon (Shuliilder. Ell>ow. 'Wrist, Kiee). Spr.aits ,l 

Fuot. Hamnd) %w ithmt Surgery 

I)islo;atioi (Jaw. lil p). Sprain(s ( K Ceiw.. Sacrrouiiac. O)t lir t' r,;,.-, 

fitdl) \\ illthit Surii *-ry 

I)ishluÉllo u ( St itoulder. Elouw. I1hnl¡), Sl)riosi (El),w. `rti.- 

1la"ml) with Suiirger! 

Disliíhation (Ki(ve, Aukle). Sprains (Shioulder, Kinee. Anklce I Iu . 

Surgery 

I)isloeatiml (llip.. Multiplel. Spriains (liip. Saeroiliac. Othe.r 1') 

s¡isciiie

Vl. \'ll. No. 2. Sntiít'ult.uit 

M.l;ijr P)i;agntosis Caitvlgry 

'i1: Adversc EfSis et*1' 

:e.rtainí. Sl:st;]w. ' 

$10 Special Admissimi.% 

itid Ei; sillIli¢11 

\\witllhot tlorte.d 

Diasguosst s 

Diagniosis-Reitlated (.:ropil. 

371 Comillic:ttioins of' IMedlicail or Surgiical Care with Surgical 

P o'( tt ed ll--- ' 

372 (:Colpli;'t:lons 4,1' Sle i'Ial or Surgic;l (:.ir. withl Slrgtry n (Re. 

l hutu(tI t of' leart Devi¡e. Relpair of StoilaLh ) 

.373 (.' :t)Ihiltinios of IMedival or Slurgicval Care with Sargerry (Ret 

sioi onf Slitint. O)thlir S;Ijor) 

37.1 

:375 

.376 

.77 

- 89 - 

APPENDIX 

A\dv. r% . I' l;;.tt ' .f I) >rug. 'l' .sxi Ell t. .t 

o

MANACGEIttI SC'IEN.'t 

Vol. 2Z. No. 4, F,-cemnhr. 197' 

Pr,*,t in C'.S.A. 

-90 - 

A COMPUTER SIMUILATION MODEL FOR THE 

CONTROL OF RABI31S IN AN UJ13AN AREA OF 

COLOMBIA*t 

RAII'H R. FRI-tl CI¡S; AND JUAN PRAW §)A 

A siminulation itilItJl is develhq>ed Jcs.ribIing ihie itr:ai.llesiloll o ca ;iitic f;-lllsv: witvaill ;¿liJ 

between 116 spaiially distribuied barrios (ncighborhoods) in Calli Colombia. The discrete 

time. dynamic model considers both discrete randomni variables (incubation and infective 

periods. appearance and movemitent of rabid dogs through the city. ete.) and delerministic 

vatriables (Jc¡nogr:iphlic ouipuiluenis of harrio cainii¡ populalailions). V:hlties for tIh ilrnp 

variables were aequired throuigh field observalions, other Colombian sources. and a review of 

the literature. Vanriots caniine vaccinatlion slrategies were tested in the model over a Ien-vear 

planning horizon for liheir cost-cefeetiveness with regard to ihe prevention of canine rabies. 

The model is reommnended to ¡he l>an Amnerican Hlealth Organization lo he used a. an 

interactive gaminiug nilel to aid heali ll sysiClmli i;al:igers il (C'ali. C'thlolulia ¿aind inl otller l.;liiii 

American cities in seliedulilg I(he 

cost-effective mnanner. 

tine and locations of vaccination teams in a more 

1. hlilrOdtl¢mlion¡ 

The problem of rabies has bcen with man for many years. Celsus. as far back as the 

first century A.D.. described the infection in man and observed that the disease is 

transmitted by the bite of a rabid animal. Since then. it has been shown that rabies is 

a viral infection of ¡he nervous systemi primirily of warm-bloolded aninials. anid is 

invatriably fatal once ¡ihe virus has infected the body 120]. 1331. 

In the last twenty years. the United States has rid its urban dog populations of this 

disease but the problem still lingers in wildlife. namely skunks and foxes [35]. Since 

man has his closcst conilct with urban animials. the climination of rabies fronm dogs 

has reduced human rabies in the United Sidtes lo one or two cases per ycar. 

Unfortunately. South anid Central American countries have not liad simiilar success 

[251. Since other diseases compete for allocations within the health budget, rabies 

control programis have not always rceeived ¡he funds necessary to climilnate the 

problem in urban areas (2]. 

The World Health Organization feels that the vaccination of dogs is one of the most 

important methods for controlling rabies in cities and many immunization campaigns 

have proven thc efficacy of this approachi [41]. In addition. they recomnmend that stray 

dogs be euthanized in order lo decrease ihe canine density in the streets to such a 

point that there is little opportunity for rabies transmission. The collection and 

subsequent lilliidg of stray dogs. however, is highly unpopular as evidenced by the 

disdain shown throughout the world toward the neighborhood dog catcher. The 

assumption is made by the authors t. ,t in Colombia' a democracy with a history of 

internal violence, a caninc euthanasia campaign would be politically unacceptable. 

Therefore. the scope of our analytical effort'is liniited to a control program fcaturing 

different vaccination policies. 

Processed by Professor William W. Cooper, Departmental Editor for Public Aminiinstration and 

Associate Editor E. S. Savas: received Ociober 1974. 

t This research study was supported by the Tulane University Intlernational Center lor Medical Rcecarch 

and Training. Grant AI-10050 from the National Institute of Allergy and Infectious Diseases. National 

Institutes of Health, Unitei Siates Public llealth Scrvicc, Washington. D. C. 

: I.ouisiana Stale University Medical School. 

f Natiounal Company ol Basic ('omnoditicis (CONASUI>O). Mexico andt National Politemc lutie Institutc 

Mexico. 

. . Copyrtght £ 1975, The Institute of stani$;omcnt Sclencem

- 91 - 

R. FRIRICHIS AND J. PRAWDA 

When planning immunizíaiion programs for a given time period. the queslion arises 

as lo whlat percent of a population should he vaccinated and how often the immunization 

camplaiigns are to be repeated. Th'IelC rol ili of rabies transmission decreases 

when (ile number of inmmiunized dog.s in an area increases. lowever. if too few dogs 

are vaccinated. the risk of a rabies epidemlic will remnain higli. IEven afíer a successful 

vaccinalion prograi. tlile risk of a rabies epidemic will slowly starl to increase. since 

new susceptible puppies are boraI inIt> ile population and old immun e do.gs either Ldie 

or lose tieir immunity. Dogs in the poorer areas of a city have higher hirth and detahJ 

rales than dogs in wealthier sections. Consequenlly the rate of turnover fronm immuine 

lo susceptible varies fronm ncighlborlhod to neighlhorlhood, creating ovcr time a 

checkerboard erfect in wlúici 

outhreak. 

eacl arca hias a diferent probability of a rabies 

The problei facing tlhe heallh system manager is lo determine tile frequency andi 

intensity of vaccination campaignls which will maxiniize the cost-effectiveness of his 

control progr;ill in cvery section of a city throughlout hc pl;anii g hori,.ol. iII ordeil 

to do liis however, he musl hiave some idea as lo the number of rabid dogs which his 

policy prevents from occurring. The best policy is clearly tlhe one wic ni minimizes the 

cost per prevented rabid dog while mailtiining rabies helow a socially acceplable 

level. 

In the real world it is difficult and frequently impossible to estimate the exact 

outcome of a vaccination campaign. The inucidence of canine rabies is greatly 

under-reported, one reason being thail extensive laboratory tests are necessary lo 

confirnl or deny lie existence of tile discase. lI adidition. imalny people observe dogs 

suspected of having rabies but destroy them or allow them to die witlout reporting 

lie incident to the local health departmient. 

Maithematical discase modells are usually constructed in order to provide insight 

into ihle biological interactions necessary for tilr tra;nsmission of ille discase. Man) of 

the malhematical models reviewed by lhe authors (listed below) were found to be 

exceedingly coniplex witli no derived analytical solutions. Other miodels. altlhoughl 

elegan iii structure. may bc of very little use to health system managers who find 

themI diffiCulh Lto undersliand and, consequenlly. lo illolmplemeni. 

Among the literature reviewed. we list the following: historical review and state of 

lhe art [3). [41, [9], [101, {191. [29]: maihematical moudels of coinmunicable diseases [171. 

[22). [30:; recentc deterninistic disease models 161, 115], [21], 126]. [381-140]: stochastic 

disease moiodels 11 [31. [. [4. 11i11- 113: comiputer siniulation disease modells ¡61. 1I1. [121. 

[131, [211, [261, 321. [381-1401; anid disease control miodels [51. [181. 231. [271, [31[. 139j, 

140]. 

It was felt tliat a simulation mlodel of urban rabies paiterned after real world 

processes would enable the manager tlo experiment with different policies and observe 

the resuliant outcomes. The knowledge derived from this approach should greatly aid 

him in making the quantitative and qualitative decisions necessary for controlling 

urban rabies. 

The purpose of Ilhis paper is first lo describe a simiulation nmodel for lIIe control of 

urban canine rabies lthat was developed for thc city ol' Cali, Colonbia. alan. secondl to 

conmpare 1ti cost-e'fectivetness of various simniulalted vaccination campaig!ns in order to 

recom nicnd )o the healtlll systlen manaitgers thill preferred policy which is the mos I 

efficient ani cffective amnong all policies tested using (ti model. 

In adeveloping the model. we first liad tlo define lile problemi and(l tIhn convert the 

¡;. blem into mathematical fornm. Estimates of tihe parameler value,- were derived 

from a review of the literatulre or from prior canine survevs co;: -c"ed ¡in Cali. 

Colonmbia. Crtatin v;alues. Ihowever were ,c'nobltaiable froai ilhese s ;t. tC.¢ id had lo 

be derived from field epc.rirnenis in Cait or estimated from exislitir IKowledge. The 

conimuter siniulation was tested cnl ihe IBM 7044 at Tulane Unive

- 92 - 

RAllIES CONI'ROL IN JURBAN COL.OMBIA 

One tic le odel was developed. two sets of vaccination policies were tested. The 

firns set consisted o`f ;iassive vaciinaltion campaligns at selected intervals whilh 

resilíted iii thie imuniiiizalion of 70 percent of lhe dog population inii a parts of ilhe 

city. T'ih second sel. which we refer to as lhe Preferred Vaccination Policies. invohled 

a SCies of exl erimn ¡s in whiclh dogs were vaccinated only in tllose areas of the city 

.tiat had l ile potlential to contributle Ilos (o tlle incidenlce .if rablies. 

In ilhe followin seclions we will describe the states and interactions of caniine 

rabies as defined ror the simulation model. The mathematical formulaltion will not be 

plresnteild here but can le fouiníd in the principa;l autior's doctoral dissertation 1141. In 

aidditioll. we will dliscuss ile differeLnt policies an¡ld dlhefcc effe lh polic! h;as on canillne 

rabies. At ¡he end wve will briefly meiluion ilc rabiíes control recommendallions which 

rwcre made lo ilhe Colombianl hCealh officiais and some of the problemis which arose 

wxlhcn ;tinlíptlini. ¡o deterlini e ¡lch valdliity of the nimodel. 

2. I)efinition of Ile IProblemi 

('ali is located iii soulhweslelrn Colombia and has a population of over 800.0R}0 

pecople :red al'lOiín;alely 84.000 dots. A sizeable portion of ile population is 

ilpoverished. wiilh oinly rfouLr prcentli of 1he i;aimilies having a Lotl co,hbil.ed ilicoí,ie. 

based on a 1969 survey. of more than the equivalent of 500 dollars per mon¡th. The 

poorer areas have dirt streets, and children and dogs are seen everywhere. 

I:or this model. tile city was divided into a spatial grid wiilh 116 neighborhoods or 

:.irrios. each harrio being arlproxinialely 50 lheetires iin size (one hectare = 10.000 

squarre meters). 1The assumnption is miade that; inieclious rabid dogs migrate stochlasticall 

hbetween harrios ani)d set off new epidemics in the barrios to which they hia\e 

iigr:led. TIhie susceplible and inillune dog-)s, as observed iii Cali. are assuiled to 

reinl:tin wilin tl heir specific barrios duringg lhe entire time period. Onlly a certaiin 

proportion of al! inl'ecld rabid doogs actually transinil the disease to oihers. If ilhe 

infectie dog develops the "furious' symptoms, he tends to wander aimilessly and bite 

al ;i lh;iteer crosses shis path. while if ilhe infection results in tlhe "paralytic" symiptonis. 

the dog beconies paralyzcd and is ineffective a dlisease transminssion 171. 1i ;addition. 

the rabid dog must lhave virus present in its saliva before ¡tie disease can be 

tranisinitied througlh a biie (361. It thce model. the number of rabid udogs exhibiting tihe 

furious symiptolis of lhe disease with virus ini their saliva is Lrelted as a discrete 

ra;ildoln variablc reseiliing from tiic encounter between wanídering-rabid dogs and ¡ile 

su.sce:ptillc do-gs living ill any giveni b;arrio. The sysleín is ;1 discrete time ímodel w\itll 

3.10ll) daily irllC¡ali(),ls 'eprcsenting rabies during a ten-year planning hori,.on. 

I'lc epideJioloLy ot f urban can;inlie rabies is dIascril)dl liagratini atically in I:igure I. 

Dogs in the rabies cycle cani he catcgorized into one of ile following. five. mutually 

;exclusive states: (I) Susceptible. (2) Immune.: (3) Incubatin- (4) Infective, or (5) 

.Dead. In order for rabies to exist in a city, changes must occur in the various states of 

dogs over tinmc. 

1 )ogs whicil iare in ic susceptible stale. .\. are thlose which have never been 

v;tccainlled C:iaillns rabies or those whlich have been vaccinaled bul hatve subsC.etiiul¡ly 

lost their iniii il.iv. ()nce a proportion, t. of thie susceptibic dogs are vaccinailed hley 

becom

- 93 - 

414 R. FRRKICkIS AND J. PRAWDA 

Birthrote 

Birthrate 

Population c b 

SG. uscept G bil / Tmm nu 

X Loss of Immunity Rote Z 

Contact /1 proportion vaccinated 

Incuboting 

Incubotion Incub0 ao 

Period persod 

Wildlife Infective 

Inmmigrating PeriodM 

Robid 

Dogs 

Death A 

F¢iouir I1. Epidicmiolo.ical model of inieractions and states of canine rabies. 

if an infective dog has adequate contact with a susceptible dog, the susceptible 

becomes infected with the rabies virus and is moved to the incubating statc. H. 

Adequate contacl means that a rabid dog with virus in its saliva has bitten a 

susceptible dog enough times to '-ansmit the discase. Therefore. tlhe probabiliiy of a 

new rabid dog is dependent on the number in the infective state. Y. Y the nlumber in lhe 

susceptible state, X. and the rate of contact, B, between tIem. 

Once a dog is infected with rabies, it goes through a variable length incubaUion 

period. The length of stay in the incubation state is determined by a number drawn 

randomly from a log-normal frequency distribution which has a mean and variance 

characteristic of the incubation period in the real world [7]. [281. [33]. Only the 

proportion of rabid dogs which are in the furious symptomatic stage with virus in 

their saliva are consídered infective. The other rabid dogs. either tliose in the paralytic 

stage or those with negative saliva. are assunied inol to be involved in the transmission 

of the disease. Instead, they move direcily to the dead sate, R. 

All rabies infected dogs cventually die, since the recovery rate is assumed to be 

zero. Once in the infective state, Y, the rabid dog remains infective for a variable 

length of time. The length of stay in the infective state is determined by drawing 

randomly a value from a log-normal frequency distribution which has a mean and 

variance characteristic of the infective period derived from clinical observations [20]. 

1341. 1361. 

While in the infective state, Y. the rabid dogs may wandesr htiroughiout the city. The 

tendency to roain is a characteristic of the furious phasc of the disease. Everyday 

thiere is a certainl proba-hility thai rabid (logs will remain in the area where thev' are 

preselilly locatecd or Icave and imove to al.djacent barrios. Only one potential move is 

allowed per day. Figure 2 illustrates the movenicis of raibid dil ,s heweenl barrios. 

The probability. g. that infectives leave a barrio, is dependen¡ on swlhere they 

,ariíinated and whece they are going. For the simulation model. each day's movemients 

are hindled in au stochaestic m;iliner. l'i:tcl rabhid dog is given tlie (dily' oplion. if still 

alive, of remaining stationary or moving to one ol' I'ive neilhborig barhriios.,. After 

entering a r.-w barrio. ,he infective dog interacts with the susceptible dogs in thlat

- 94 - 

RABILES CONTIROL IN tURIAN COLOMBIA 

IMMIGRATING RABID DOGS 

ISNS SUS g 

sus sus RABID 

SU SUS SUS SUS--WLDLIFE 

SWILDLIFE 

sUS 

RABID . 'SUS S U s SUS 

WILDLIFE T 

4SUSSUS S US 

IMMIGRATING RABID DOGS 

FIoURE 2. Movement of inective rabid dogs and wildlife into and through a spatial grid representing 

an urban area. Abbrev.: INF - infeclive rabid dog. SUS - susceptible dog. g - probability of movemenl in a 

given direction. 

barrio and possibly, depending on the amount of contact, continues the transmission 

of the disease. 

A certain number, W, of rabid wildlife (see Figures I and 2) enter the peripheral 

barrios of the city and interact with the susceptible dog population. The wildlife are 

given the opportunity to make only one move since it appears likely that a wild 

antimal would be killed before it could roam very far in a city. Once in tihe urban area, 

the rabid wild animals are included in the infective state, Y. 

Urban canine population growth, G, in every barrio (see Figure 1) is assumed to be 

proportional to the human population growth in that barrio [37]. The assumption is 

also made that these entering dogs are not vaccinated and therefore increase the 

susceptible canine population in a specified barrio by G. per day. Some of the newly 

arriving dogs may have had contact with rabies outside the system and be currently 

incubating the disease. A certain number, M, would become part of the infective state, 

Y, during the year (see Figures I and 2). The probability that they will appear in any 

specific barrio is dependent on the proportion of the city's human population 

presently living in that barrio. The assumption is based on the idea that the more 

people there are in an area, the greater is the chance that their dogs will have had 

some interaction with other dogs in surrounding areas or cities where rabies currently 

exists. 

The model accepts as inputs the vaccination schedule. the number of vaccinating 

teamnis and trucks, and tlhe numher of rabid dogs anid rabid wildlife which enter ealch 

year fromi outside tihe system. 

3. Exsperiteicnts Usinsg itLe Model 

Since ti,e model is stochastic. numerous epidemic trails were run per experiment in 

order to determine thc mican numbcl cunmulative rabid dogs atl he end of o íevery 

year. All costs mentioned in the following sections are estimates. based on Colombian

- 95 - 

R. F'RERICItS'AND J. PRAWDA 

data, of what each vaccination campaign would cost if the experiment were done in 

the rcal world. 

Rather than presenting frequency distributions derived froni multiple epidemic 

trials per experiment, we will deal only with mean values. The policy of no vaccination 

and the policy of 70 percent initial vaccination were the only ones which 

consisited of 100 epideic trials per experiment. The results from all other tested 

policies are based on 10 epidemic trials per experiment. 

3a. A Single 70 Percent Vaccination Catmipaigtn 

Looking only at the niean nuiiber of ca;nilti rabies cases per year, Figure 3 

indicates that the policy of 70 percent initial vaccination was able to reduce the yearly 

incidence of rabies for approximately five years. Thereafter, the yearly incidence of 

rabies was greater than ii vwe had followed the policy of doing nothing at all (no 

vaccination). With rabies absent froni a barrio. the dcnsity of susceptible dogs and the 

concurrent likelihood of a rabies outbreak increased. All that was required was for an 

infectious rabid dog to wander by chance into the barrio and trigger a new epidemic. 

This same plhenomenon of a disease being kept out of an area and then suddJenly 

appearing in an epidemic fornm lhas bcen frequently tobserved iii the real world with 

rabies as well as with other infectious diseases [161, [24]. 

Comparing after ten years the cumulative mean nuniber of rabid dogs in the 

absence of any control policy and the value with an initial 70 percent vaccination 

campaign, we see in Table I that thc immunization program prevented 2.985 rabid 

dogs (21.588 minus 18.603). The theoretical vaccination campaign, however. took 85 

daily iterations to complete and required the operator to assign 48 vaccinators 

traveling in 8 trucks to all barrios of the city. During tcse three months. 59.323 dogs 

were vaccinated. Taking into accounti the cost of wages. the use of the trucks. ihe 

4000 - 70% ni/h/v. va 

40001 

/70% initol'tl iac. 

o 3000- 70% revac. 5J 

No vaccinat/on 

2000- 1 

%I0 0 ,Preferred ¡ 

vac. 

',~:~ _ .'opo/, y(V4-va-7O YJ 

I 2 3 4 5 6 7 8 9 10 

Year 

FiLGUR 3. Compuier Simulation Experiments: Comparison of the mean number of rabid dogs per yecar 

,t.ri;nc a ten-year period with policies of (I) no vaccination (mean, 100 epidemic triais). (2) initial 70 percent 

vacciration (mean, 100 epidemic trials), (3) two 70 percent vaccination campaigns. one a tihe beginning and 

the other after 5 years (mean, 10 epidemnic trials), and (4) the Preferred Vaccination Pulicy with VA sei al 70 

percent (mean, 10 epide.a.: irials).

- 96 - 

RABIES CONTROL IN URBAN COLOMBIA 

TABLE 1. 

C'onpuiter Sinudlaion Experiotnems: Comiparison of four wa¢cinaslion policies, orer a ten-.eur period. 

ita h rcgtrI tll s tllo' s¿ntil l l.¿tllt ti'¢' ,Ii lhl' r (If rlifld dogsx. th', flteel rct .luit*l' r of cci rsted dt>gs. oi¿d 

¡he cosi vj ¡he inditiduúal policies. 

Cost (dollars) 

Number oí Mcan 

Cl)idenilli toh l Iper ra.bid 

trials per rabid Total dogs per dog jdog 

Vaccinatiun policy experiment dogs vaccinated Total' vaccinated prevented 

No vaccinaitioi 

70%' initial vacciniation 

1¡00 21.588 O 0 0 0 

(entire city)" 

70% iniiíal vaccination 

70{P, ftvaccinatioin (yr. 5) 

100 18.603 59.323 22.157 0.37 7.42 

(;sir¢ ciíy)~' 

'referrcd Vaccin:aiun 

ISC 7.71 I 139,638 60.8 , 0.8.,1 .1.3.1 

Policy (VA- 70%')'** 10 3,197 100.282 64.982 0.65 3.53 

'lic os( of vacciniation is annually increased by an inflation factor of 5 percent 

·* Thie simuliated camupaign took 85 days and required thlc use of 8 trucks and 48 vacem-lt;ors. 

*'· The initial simulated campaign took 85 days and the second simulated campaign. due to 

the canine population growth, took 112 days. Both campaigns required the use of 8 trucks and 

4S vaccinators. 

'' The policy ;cquired the continuous usc of I truck and 2 vaccinalors. 

Rabies Control Center, and the unit cost of the vaccine, the 70 percent vaccination 

campaign theoretically cost the administration 22,157 dollars. The cost per prevented 

rabid dog is calculated for the len-ycar time period as 22,157 dollars divided by 2.985 

prevented rabid dogs or 7.42 dollars per prevented rabid dog. The question lhen 

arises, '"Can the health system manager do better with a different vaccination policy?" 

3b. Two 70 Percent Vaccination Camnpaigns 

It appears obvious that a second 70 percent vaccination campaign at year 5 would 

be very beneficial. Figure 3 compares a policy of one initial 70 percent campaign 

versus a policy of two 70 percent campaigns, one initial and one at year 5. The double 

campaign reduces rabies for about nine years before there is an epidemic. 

The cuos of the double camrpaign. as seen in Table 1, is almost triple the cost of the 

single campaign due to the increase in both the city's canine population and thc cost 

of labor, supplics, etc. associated with 'a five percení' annual inflation rate. The 

cost-per-rabid dog prevented. however, is greatly decreased due to the sizable reduction 

in the mean cumulative number of rabid dogs. 

3c. Preferred Vaccinaoion Policy 

In order to reduce further the incidence of rabies, we developed an alternative 

policy which continuously employs throughout the planning horizon two vaccinators 

traveling in one truck to various selected barrios. We refer to this control policy as the 

Preferred Vaccination Policy since wc felt i. was the hest among all tested policies. A 

formitiil for the Prefcrred Vaccination Policy was developed which ranks each barrio 

accorditng Lo thc potenti[il conuilibution it would miake to ¡Ihe inicidence of c;nine 

rabies. The risk of contributing to new rabies cases is calculated for each barrio as 

follows: 

j-_

- 97 - 

R. F'RRICHS AND J. PRAWDA 

where 

R,.,= thc index of rabies risk for a barrio i at time i, 

C, = hc relative proportion of barrio i dogs typically found in ihe streets. 

X , ,=the number of susceptible dogs in barrio i at time t, and 

i(j)=a subscript denoting neighboring barrios j,j= I ... 5 which surround 

barrio i. 

Since a rabid dog moves fromi barrio to barrio. the assu miiinIon is miade tlha tlhe sull 

of the values of Cq(,)X,(), in each of the five neighboring barrios is equally as 

important to the subsequent generation of rabies as the value of C,X, 1, ini barrio i 

itself. 

All values of R,., are ranked froni highest Lo lowest. A vaccinaling teaim aild truck 

are sent to thi highliest ranked barrio. II tiher are numerous trucks, each one goes to a 

separate high ranked barrio and remains there until the required proportion of 

susceplible dcogs is vaccinated. Thereafcíer the trucks return at diffrcent timc periods 

to the rabies center for reassignmieni to tle next highest raniked barrio. Tlhe most 

important factor is that the vaccinators go house-to-house in the selected barrios and 

vaccinate all susceptible dogs they encounter. Since the necessary informnation for the 

formitula can be continuiously updated hy tile vaccinators with limited data collected 

during their visit to the various barrios, we feel thlat the policy will be easy to operaite 

and will require little supervision. Further details of this barrio ranking procedure are 

published elsewhere (14]. 

Thc Preferred Vaccinaltion Policy is co*mpared in Figure 3 with a policy of no 

vaccination. The term "VA" refeis Lo the fact that 70 percent of tihe susceptible cdogs 

in the selected barrios were immunizei durIn; each visit by the vaccinators. Or 

restated, it means that 30 percent of the barrio dog owners were either not home or 

were urnwilling to cooperate with tl - rabies prevention prograni. As can be seen in 

Figure 3 the Preferred Vaccination Policy greatly reduced the incidence of canine 

rabies throughout tihe planning horizon. But al wiiat cost? 

Comparing the three vaccination policies which we have shown so far. Table 1 

in¡dicates that thc Preferred Vaccination Policy with 70 percent public cooperation 

cost more than either of the otlher two policies. However, the cuniulative meanll 

number of rabid dogs over the ten-year planning horizon is also reduced to a lower 

level than either of the other two policies. 

PREFERRED VACCINATION POLICY PROBLEM 

80- n 

70 1 - MEAN 

60- Ki .95 confidence interval 

50- 1 

4 40- 

30 

20O 

Q. 10 __....._______ _- 

2 4 6 8 10 12 

COST(USAdollars) per PREVENTED RABID DOGS 

FIGURI: .1 ('omlpuler Simulalitn Experitmcnls: Cost ¡..r prevented rabi do 

dd Juring a ten-ycar rc, 

with the '-ilerred Vaccination Policy sel at diffcrcnt i.:vels of VA (percent vaccim;ted in ach .lv, 

barrio) (m,':;;. 0.95 ?cfidence interval, 10 epidemic tr!:is per experiment).

-98 - 

RABIES CONTROL IN URBAN COLOMBIA 419 

The usual indicator of a programn\'s success is the cost-per-dog vaccinated. We feel 

that this is a poor indicator since the objective of a rabies control program is not to 

vaccinate dogs but rather to Jecrease the incidence of canine rabies. Consequently the 

cost-per-rabid dog prevented should be the correct indicator. As is seen in ]'able 1. the 

Preferred Vaccination Policy with 70 percent public cooperation cost the niost per 

vaccinated dog but it cost the least per prevented rabid dog. 

An additional feature of the Preferred Vaccination Policy is that it is fairly 

insensitive to the percentage of barrio residents being home or cooperating. We 

selected VA (the percent of the susceptible dogs in a barrio that are available for 

vaccination) to equal 70 percent based on prior knowledge of participation rates for 

immunization campaigns. However. even if the estimate is in error. when VA is 

decreased in multiple experinients from 80 to 10 percent (see Figure 4), the theoretical 

cost-per-prevented-rabid dog for the Preferred Policy does not go up dramaticallv 

until less than 30 percent of the susceptible dogs in a barrio are available for 

vaccination. 

4. Validity of tile Model 

Before concluding. a brief niention should he made concerning the valiJity of ihe 

model. The problem of determining the validity can be analzed i in to wta)s. The 

first deals with iniernal validiti. Does ihe simulation model perforn iii the manner it 

was intended? We feel that the internal validity of the model should be very high since 

the compuier program was repeatedly tested and "debugged" during the initial 

construction of the moJel. 

The problem is with the external validiity. The question. "Does Ihe modell tell what 

i. occurring in the real world?". is more lifficult to answer. What we wuuld liike to do 

is he able to predict future happenings. lowever. since tlis is a new simul:a model ion 

which has never heen tested for predictive accuracy, all we can do is subjectively 

evaluate it to see if the results appear logical. Therefore, instead of external validity,. 

we will consider only if the model is or is not reasonable. 

Based oun our knowledge of rabies in general antd canine rabies in ('ali. lhe model 

Joes appear to be reasonable. An opportunity to compare the computer model o(utput 

wilh results in the real world occurred in the summer of 1971, wlien more than 80 

percent of the known dog lopulaUion ií' Cali was immunited by ¡he l:an Anierican 

ilealth Organization. The reported monthly incidence of canine rabies quickly 

dropped from more than 20 to less than 5 cases per month and remained at this level 

'or at leasl one year thereafter in spite of the fact that there was increrased rabies 

surveillance. This finding is in accord ¡itih the observations made using the simila:;on 

mnoel. A further discussion ot this point can be found elsewhere 114). 

S. Concluniuon 

l'le resulis of the simul;ation experiments indicate that a single vaccination 

campaign which immunizes 70 percent of all dogs in each hbarrio of ('Cali will maintai 

rabie:: :, a minimal level for five years. During the ten-year planuiing horizon. a policy 

involving two 70 percent vaccination campaigns (one *every five years) is more 

effective per unit cosí in reducing the cumulalive number of simnulated rabiid dogs 

than a single campaign during the same time period. The Preferred Vaccination 

Policy. utilizing information about the dog population in the city in order to identify 

high risk barrios for selective immunization, resulis in the lowest cost--per-prevented 

rabid dog.

- 99 - 

420 R. MFRIlC}IS AND J. PRAW)A 

The comnputer silmulation nmodel is recomimended lo the Pan American Hcalihi 

Orgain/,aliomi ciilich tIo lelp formult;icte va;icci;iollimn strategies for C;ali. oi r lo 

serve ats ati illt;iactlie teaclimll or' 1.ítiili$ míodel lo be Lused dlturin (lcil antitia; l c;lass 

on rabies control. As ;L llil .of1 experimenlaion wilh tile mn'odel. hcJlh svslemn 

mna;ll'agcrs shliu:ld learn to ;tppieciate the onlplcxity í)f tle systeim and derive grealer 

im.iiIhl t111> tle ecvenllt; ctntrol of til':til rahcles. 

Ríc;creces 

1. AaI:Y. Ilil.l:N, "An Examiiianon of ihe Reed-Frost Theory of Eplidrnies," litann BiGlaog.. Vol. 3 

(1952). Pp. 201-233. 

2. Al.BA, Il. As AcI Ai. 1>., "1TheC Stalus of Rabiehs in the Amriclas." 'S¥tol Ucri¢'l/t Ihler-.Il .lteting oit 

IWcoot-gal-nMaluh D.t,

- 100 - 

26. RLVEL.LE. C. S., LY\'N. W. R. AN;I F£L.DlIAS%. F., "Mathematical Models for ihe Economic Allocation 

of Tuherculosis Control Activities i 

89>3 909. 

e)cvcloping Nations." Am. Rcy. Resp. Dis.. Vol. 96 (1967). pp. 

27. - , I. I)MAN, 1:. ANí> 1.Ys. W V,. "Al l ()ltialiioIaI. Moledl al ílmcl'c ulm mñs I.pld¢llllmol:.m" 

/%Imt¡l'114'1{ ,S('i('11' Vol. 16 (Y(1')>. pp. 190 21 1. 

28. ailll, ( C'lARI.E. (.'., "l{mle.." il ('1úilt' Al4diclu . AIe¢isc.i.T Vetcrtsiary i'uTll . ¡hilg 1 . a..1ta 

Blarhar;a. ('al.. 19(2. 

29. ......... .I(ul:K I-I ;ohS 

145-166. 

"`h rt,.cl K{ cImw mnl p'pldmm c i'm'l'hc.i y* JI laum fIh.l, .I, Vo ?4 -1 (1 '.2>. pp. 

30. - , "Methods for Current Statistical Analysis of Excess Pneumonia-Influenza Deaths," Public 

Health Reporis. Vol. 78 (1963). pp. 494-506. 

31. TAYIOR,. I-lowAII 

pp. 383-398. 

M. .. Somie fmlcdls i¡n Epidemic Control," AMtrlmenlical /iu.c¡ieie. Vol. 3 (1965). 

32. TAY.Oit. WIL.LA,' F.,. "Some Monte Carlo Metlhods Applied lo an Epidemic of Acute Respiratory 

Disease." Iltunr Bialogr. Vol. 30 (1958). pp. 185-200. 

33. TiERKiE.. ERNEST S., "R.abies," 

Springficld. II1.. 1'93. 

in Discases Tralsmitted from Animals iu Mun. Charles C. Thomas. 

34. - . "'Rabies." in (urreost Veeriurp, 'liherapl. 1966-67, W. U. iauijcrs Co.. I'lmilicl¡phai. 19d6. 

35. U.S. DEPAR'rMEN'I OF HIALT'Iit. EDUCATION, AND WEILIARLk, Morbidi, anid Muronilitr Aaittl Suple. 

ment, 1971, Center for Disease Control, Atlanta. Ga., 1971. 

36. VA'uIIN. J. B.. Gllzr^%Rnr. P. ÂN) NlaeI:I.. K. W., "Excretion of Street Rabies Virus in the Saliva of 

)Dos.". J. Am,:. MAI. Assm... Vol. 193 )(65), P).. 363 .368. 

37. VILLA)II:.Oo, FRANCIS(O. "A Two Percent -lousehold Survey Sample of Cali.," Universidad del Valle. 

38. 

Cali, Colombia. 1968-1970 (unpublished data). 

WAALER, H. T., "A Dynamic Model for the Epidemiology of Tuberculosis," Am. Rec. Resp. Dis., Vol. 

98 (1968), pp. 591-600. 

39. -- as I¡'o'r. M. A.. "Thc Use of an Epidemiological Model lor L.stimatini the EffecLivcness of 

Tuberculosis Control Mcasures," Biulletin Worhl lhealth Orgauni-:acion, Vol. 41 (1969). pp. 75-93. 

40. - - .AI' , "Use of an Epidemiological Model lor Estimating the Effectiveness of Tuberculosis 

Control Measures," Bullesin World Health Organization, Vo!. 43 (1970), pp. 1-16. 

41. WOKi.L IIFAI.TII ORGrANIZATION, I¡Y. I. OO. Expert Comnmitee on Rabies. Fifth Report. Technical Report 

Series 321, 19')66.

V 

- 101 - 

McUCi. (:Au 

AWuu IMI. VoL XIX Na. o 

The Case-Control Method in 

Medical Care Evaluation 

SA&n~ GCEeW.N.u M.S.. Da.P.H.,. EuCA, WáTmúON, M.D., M.P.H..t 

ND iLtrYOND R Naurm, M.D., DaP.H. 

Th. s~ meod esl belnu aplied euemively to he study .ocbomeic 

d~ lly be.sus éilts s¿dvm

VoL XIX. N. a 

- 102 - 

f tdie outeonie of interest delries two 

distinct categories (lfor exunple, death versus 

surival). outcome rates in different 

groups can, in theory, be compared by the 

case-ontrol ("retrospective") method." ' & ' 

This method inivolves the election of a 

series of earso (btho~ persons experiencing 

the outo«me event) and the selection of a 

reference series ofnoncases ("onrols" or 

"referents") for comiparison to the case 

series. Vasious itaitica4l techniques allow 

estimion of differences in outiom occurrence 

between groups based on the 

cae-onatwl data'-provided the casecontrol 

study satisfies certain conditions 

requised for the validity of results. A primary 

advantage of the case-control method 

is that the number of subjects necessary is 

usually much les tan that required for a 

prospective comparison.' In the monitorlng 

exnple, a cas~o ntrol study would 

require fewer than 400 deliveries (200 

cases and 200 controls) to detect a 50 per 

cent reduction in the neonatal death rute 

due to monitoring with 90 per cent probability 

(power), at a 0.05 significance level t 

(assuming a 50 per cent monitoring rto in 

the target population). 

More than 20 yean ago. Cornfield and 

Haenszel' summrized the two critical 

represmntativeness' conditions that were 

thought necessary for the validity of a 

cse-control study: 1) it should be "possible 

to enumerate all new cases of a disease. 

or a representative sample of them, without 

haviig to doserve all the isdlviduals ¡i 

the population"; and 2) "the sample of individuals 

not developing the dierse ithe 

contrulsl Ishould supplyl an unbiased estimrte 

of the prevalence of the chtraceristic 

under study among the entire nondiseased 

population of interest. The authors 

then went un tu lamuint that most retrospective 

studies of dtit day had dificulty 

assuring the saislacwtion of the seond reqOtler authors of 

that period put forth simnilar criteria.u' 

More recent writers have cxpressed related 

criteria lor the validity oícal-cmotrol 

CASEV-COXNlTOL MTHOD 

resultl in a samplinig theory fraunework,? 

while other authors have rejected representativeness 

criteria and ilnstead have 

empha>sized the importance ofcompanbility 

ofcases and controls." A conimon feature 

of all these points of view. however. is 

un emphasis on the importance of "section 

validity" in casfcontrol studies, in 

that subject selection should be unbiased 

in order to achieve valid results. 

aIn the time since the early commentaries, 

genend resevatioa s onceming the 

utility of the caa.control method have 

been put forth, largely o the grouds that 

representativeness or selection validity 

conditions ar difficult to fulfll or auure in 

ordinary practice." Thbe purpose of our 

commentary is to offer an appreciation of 

the particular suitability of the case-control 

approach for many problems in medical 

cae evaluation. We will discuss why the 

clasic problems of selection validity are 

frequently absent in medical care evaluation 

settinr. and how odither common problemns 

ofordinary case-control resarch will 

often be miimizod as well. 

Coalidera~ Favorina the 

CaM-Cotil Method in . 

Evaluadon Sem~is 

bepresenaatveanrs ad Se~oo 

Co.sldodnlooe 

A menlioned above, problem of selection 

bias and subject nonrepresentativeness 

haive loed o a curin degree of dinstrus 

ofl s-control methods among many clinical 

researchers. Yet it is precisely in the 

realm of preliminary comparisons or 

evaluations of medical care or medicld 

teehnology tht the -control metdihod is 

most often immune to such problems. 

This is because many evaluation setting 

involve amall, well-defined turget populabions, 

such as a hospital inpatient population 

or several co¡nbsied hospital pupulations. 

In such settings it is usually possible 

to enuumeraite the entire target population. 

selecL representative wmlples froinm hud 

a

GREENLAND ET AL 

- 103 - 

the case (pa>tient experiencing Olutcom)ii 

mid noncase (p;tients without oitotaíole) 

groups, and thus avoid tdihe problemis of 

nonrepresentativeness and classical selection 

bias. Often, the target population will 

be enumerated in advance. 

Aallyssi C4oUdersiion 

Because of the lrequent paucity of infonnation 

regarding the absolute disease 

rates in the target or source population, 

classical (etiologie) applicationis of the 

c;se-control method have placed strong 

emphasis on the odds ratio (or fu¡ ;ons of 

it) in the analysis of case-control data." - ' 

Many (if not most) problems in medical 

care or tecínology evaluation. however. 

require assessment of ellects in terrms of 

differences in exposure-specific rates, 

especially if risk-benefit analysis is required. 

s Furthermre, the use cf ratio 

mneasure without reoerence to uniderlying 

rates is inadequate isn most evaluaticn settings. 

For example, to say that a treatmnent 

reduces the odds of death by a factor of 20 

(an odds ratio of 1120) would be equally 

true if it reduced death risk from 90 per 

cent to 31 per cent or lrom 0.1 per cent to 

0.005 per cent; yet a high risk of senous 

side effects from the treatment would yield 

vastly different implications in each instance. 

Thus, the emp!hasis on odds-ratio 

analysis may be seen as another drawback 

of the case-control method. Fortunately, 

when enumerations of the target popula- 

',i ation and die cases are available (as is frequently 

tdie case in evaluation settilngs). the 

absolute exposure-specific rates can be estimated 

using Bayes' theorem,' and classical 

case-ontrol uanalysis methods can be 

deremphasized. Thus, the existence 'of 

soiil, well-delined target populations in 

evaullatiol settings provides analytic bene-, 

fits in additioín to validity assurancii s. 

Ajl Exiamiple: A Comparison of Neoiautl 

intestuive (Care Unilt 

Wae will ll hlustralte tile ailtve aud latL r 

p)ilmiNs withi aU exaip>le riolll Osir (OwiI ex- 

MWruCA CU 

perientce, a study recently carried out to 

4cmilpamre tle perlimniuiace of two diíilrent 

neonaatal intensive care units (NICUs). 

This study was an investigation of preliminary 

statistics indicatinLg an elevated death 

rate .amollg infant admintted to one of the 

usits (herealter designated unit A). A seeornd 

unit (lereather designamted unit B) serving 

a difierenit geographic region within 

the same health camre system was choseli as 

the comparison unit. Both units served 

low il5ciin, populaltio s ii tihe samle t¡ietropolitan 

urea, with no special referral jplan 

between the units. A randomized trial was 

clearly impractical. Ofprimary interest was 

whether the elevated death rate at unit A 

ciuld be explaisaed solely as; a restult of 

chalice, or as being due toi ana elevated proportion 

of high-risl infants being adminted 

to unit A. Ifeitherexplanation turned iat to 

be correct, a costly investigative comparison 

of quality of cure in the unit coulil be 

avoided. 

Civen various practical considerations, 

total samile size was limited to approximately 

200 intfants. Based on an initial estinate 

of the death rates in the units. only 

25-30 cases (neonatal deaiths) would have 

been expected in a prospective study 

based en a simple random sample of 200 

admissions t tthe NICUs. This was judged 

to be an insuflicient number of caseá to 

accurately assess the role of vanaous risk 

factors in die units. Furthermore, the admissions 

to the NICUs (tlte target population) 

were conmpletely and continuously 

enuminierated, anl thie status of admittees at 

28 days Rlier birth (aliveldead) was always 

determined. Given these consideratiois, it 

was decided that a case-control stuldy 

wouid be conducted íimom medical recotrds. 

(>ver a specified period of time, all NICU 

adUilissions that eunded in a neoinatal death 

woild be entered in the stiidy, along with a 

50 per cent randomn sumple of tie neoiaital 

sxlrvivors fiom the same target population 

(mliít A + asmit B). (Muiltiple births, i intsn 

uiader 800 gt mris hirtiiwt!ilght or wnill coug'alitall 

lUll()rlnath)ns illcalkinafihle with

Vol XIX. No. o 

- 104 - 

Iil'e or tntislkrred into our out of the NICUs 

were excluded in>)m the study. Very lew 

inLaits Iruom either NICU were translerred 

out.)Tlhere were 361 eligible iadniisi¡us to 

uits A and I over dithe study peridl ( IY9J (u 

unit A ¡id 162 to unit Bi). Of diese. 61 

suflered neonatal death and thus became 

the study cases, while 150 of the 300 

suarvivors were selected fir the eointrol 

(relcrenue) group. 

The initial results are sunmarized in 

Table 1. The crude ratio of the rUtes between 

the units, 1.9, was signifiant at t the 

0.(. level, faiirdy well ruling out chance 

aloune as an explanation of the diflference 

(Table I). Thle m¡ethoad uf rte computation 

is brietly described in the appendix. Using 

both univariate and multivariate statistical 

techniques, ' 4 more than 30 risk fatours 

were exwrnined as possible contributurs to 

die ditierence. These are listed in Table 2. 

Few tumed out tu be of any inmportanuce, 

mtot Lbecuse of being similarly distributed 

between the units. and sume because of 

their very weak ctontribution to risk of 

death. Ethnicity (race) was distributed 

quite differently between the unm and 

nearly all of dhe interunit differenee in 

death rates was concentriated in the premature 

infants (les dthan 1500 grnams birthweight). 

After adjusting for race, the rate in 

unit A was nmore tdiu three times die rate in 

unit B among prematures, and the absolute 

rate differences were quite high in this 

group (more than 35 p>er ceit). Table 3 

sulniaurizes thiese details. Further i nvestigatiln 

of the elev::.'ed rutes in unit A will 

benetit by this initial elimination ofseveral 

ix>ssil)le explanations (such as ethnie diflerences) 

aiad will be able to lcus ou. risk 

iLwtors and aspects of care that primllrily 

alieSt premature inlants. 

Olher Conideraiiiau 

A majior claas orflrolhletis in cas-cuaultnol 

statli

CREENLAND iT AL 

- 105 - 

TABLE 2. Fiacturs Exuri.iwed i r 

ColtlribltiHn tul h ilri teielit 

Death Raite Dillcsruic' 

Maternmal ae (yeAr1) 

Raie/ (Black; Hipanic; other) 

Cnavidity 

Numli.r of dpl ,umneous abor.ltís (20 weeLk 

Prenatal cue (yes; no) 

Hydramiios (yes; no)* 

Abnoeroll presentation (yes; no) 

Cord prolepie (yes; no) 

PlMaenta abnormalties (bruptio, previa, etc.) 

(yes; no) 

Oiyutoin (no; augimetaiuun; ieduciuil)* 

Vaginal delivery (no; ueouplica(ed; 

complicated-braech, moldfocepa. vacuu¡m, faled 

forcepa or vauuim. etc.) 

C-sectiou (no; repeat or pnriauy elective; 

nonelective) 

Abnormal dugnton ofaibor(no; 20 hn 

Seu 

Birthweight (grais) 

Cesattional age (Dubowitz) (weeks) 

Apar I iian. (by pediasticui) 

ApSar 5 inio. (by pedimcri¢ia) 

eauscattion al deolvery tendotra(chei 

intubatlon; no) 

VMor consenital abormality. ch(uroounie/genetic 

defect. etc. (no; yes, potentially viebl) 

slteliratory dintrins syndrumaneresiriatory miilure 

(uaswtlated with tmmiuunty andlur pernna) 

ap.ipyua) ('es; no) 

Meuniunm/biluduasnintic fluid spiratioia (ye; no) 

Provel iel(e.tiun (no; cougeni il or neonalal. Ro< 

horsiital.acluired) (infctiuns frli bo' .e of 

nouuuomiial oriyin-e.la. buterial sepuis with 

unire >3 dary pu)dielivery-were cuded "no"e 

ls;oi uiiiioíiiítiofsi4ther ígyoKeniral hernmlytic 

disorder req lirings exlchanl tria slusoiu)(yes; no) 

Signifiiclat iirth trautuilitjury s c; no) 

iirJtl oof hospital (home. iln raiulit. etc.) (yes; no) 

Nuiber ol' da»y in hospital (0-24. hr> . I day) 

'l'. ltac, tBiis silflred Irdsn ¥ ib levecls f r eímrdilag 

iluacuracy iii dte recirds esauIalUleLd 

Discussion 

MiDICtAL Ci 

Plrolpr applicatio of tlihe cUse.uitrol 

approach requires consideratiou of and 

concerted efforts tou meet a number of 

criteria for study (internal) validity. These 

criteria have been extensively discussed 

asid systemiiatzed li¡ tihe recent epi- 

- " 

demiologic literature.. Chief amolg 

the criteria have been selection validity 

criteria (concerning bias in selection), 

inlormation validity criteria (concerning 

dite quality of the inilfianatíio on the risk 

factora under study) and couiparnson validity 

criteria (concemrning the control of confounding). 

Efforts should be made to meet 

these criteria as far as possible in any area 

ofapplication ofthe case-control method. A 

researcher considering a case-control appnrach 

should carefully evaluate whetlser 

his study setting would allow fulfiliment of 

the various validity criteria. If selection. 

informnation arid comnparison valility 

criteria cannot be met. the study will be 

unsound. These criteria apply to recordbased 

cohort ("historical prospective") 

studies as well. 

Special caution is required in the use of 

routine medical records for case-control 

'and historical prospective studies, sinee 

biased recording of information on the 

study subjects will produce corresponding 

biases in tdie study results. Investigatnrs 

must be alert to the possibility tthat the use 

of a particular treatment regimen led to a 

more frequent recording of the outcoimte 

under study or thbat the treatement under 

study was recorded more freqmaently 

mutiong tbose experiencilig the outconle. 

These were clearly not possibilities in our 

NICU study (where the "treatment" was 

the NICU). but such problenms are frequent 

in nonexperimental drug evaluations." 

Impoortant in any mionexperinientil shady 

of mnedical treatments or technology will 

be txontrol of l'cnfounding.a Conflmntditlg 

warrants special attention in nonexperilmelntal 

ttclisuolohgy eva;ltl;itioi lIi.calibt' o 

tdie possiility ,l "cbu¡llouldinig by itídli.tioín"i"; 

this (otic'rs whenever prognostic

1 

- 106 - 

vol. xix N. 8 CA.E.CONTItOL METHOD 

TALLE 3. Finael Comptison vi' Neonatal Intensive Care Units 

¥lbior's Estiisntud 

iirtlh N>. No. EEILctulcd Eiá. Hice 

¡tlm. Weight Unit uses Controlss ea~lh ise* Test Iaio 

stlck Under 

1500 

Crmanm 

A 

B 

16 

2 

8 

7 

50^.L 

12.5 

p < 0.05 4.0 

ai"ck t)Over 

150t) 

Grams 

A 

is 

4 

2 

31 

8 

6.1 

11.1 

NS 0.5 

Oiheri Under 

1500 

CGiunw 

A t 

5 S 

2 

U 

73.3 

1.7 

p < .02 3.4 

Odbrl Over 

1500O 

Crun~ 

A 

1 

12 

9 

37 

48 

14.0 

8.6 

NS 1. 

* Comrputed using Blyes' Lhcurm (ee appendu). 

t Prinuaily tisplpaic. 

NS - n signifirúant 0.05 level. 

fators for the outcome under study served 

as indications or contraindicatf, s for the 

application of tie treatmnent under study. 

Analytic control of confounding will require 

that accurate information be available 

for all subjects on important potential 

confounders (e.g., in the NICU study. it 

was necessary tou have accurate iniornation 

on ethnicity, birthweight and other determinants 

of neonahill mortality). 

Control of coniounding can only be 

evaluated re!jtive to subject-matter 

knowledge and judgment regarding die 

study sitauation. Even when "hidden" (uncontrolled) 

confounding is believed to 

exist in the study results, however, the resuilts 

may still provide useful information. 

For example., in our NICU study, it is still 

possible that same unrecorded risk factors 

(perhaps with dillkrential implications for 

admiission to units A and B) rather than 

GREENLAND ET AL. 

3. Cornfleld J, Haenszel WH. Some aspects of retrospective 

studies. J Chron Dis 1960;11:523. 

4. Mantel N, Haenszel WH. Statistical aspects of 

the analysis of data from retrospective studies of disease. 

Natl Cancer lnst Monogr 1959;22:719. 

5. Neutra RR, Drolette ME. Estimating exposurespecific 

rates from case-control studies using Bayes' 

theorem. Am J Epidemiol 1978;108:214. 

6. Prentice RL, Pyke R. Logistie disease incidence 

models and case-control studies. Biometrika 

1979;66:403. 

7. Schlesselman JJ. Samnle size requirements in 

cohort and case-control studies of disease. Am J 

Epidemiol 1974;99:381. 

8. Dom HF. Some problems arising in prospective 

and retrospecti ve studies of the etiology of disease. N 

Engl J Med 1959;261:571. 

Appendix 

MEDIcL CARE 

9. Kleinbaum DG, Morgenstern H, Kupper LL. 

Selection bias in epidemiologic studies. Am J 

Epidemiol (in preas). 

10. Sackett DL. Bias in analytic research. J Chron 

Dis 1979;32:51. 

11. Cole P. The evolving case-control study. J 

Chron Dis 1979;32:15. 

12. Feinstein AR. Clinical biostatistics. St. Louis: 

Mosby, 1977. 

13. Greenland S, Neutra RR. Control of confound. 

ing in technology assessment. Int J Epidemiol (in 

press). 

14. Miettinen OS. Efficacy of therapeutic 

practice-will epidemiology provide the answers? 

In: Melmon KL, ed. Drug therapeutics. Elsevier: 

New York, 1980. 

The rates given in Tables 1 and 2 were computed from the casecontrol 

data using Bayes' theorem. A detailed and general discussion of 

this method is given in reference 5. Briefly, suppose we know the 

overall rate of the outcome, P(D), the probability of exposure among 

cases, P(E/D), and the probability of exposure among noncases, P(E I 

D). Letting P(D) = 1 - P(D), Bayes' theorem states that the outcome 

rate among exposed persons, P(D/E), is given by 

P(D 1 E) = P(E I D)P(D) 

P(E 1 D)P(D) + P(E D)PD) 

In our NICU study, the total size of the target population (unit A + unit 

B) was equal to the number of cases (61) plus twice the number of 

controls (2 x 150)or 361; the ital number ofcases over the study period 

was 61; and so P(D) = 61/361 = 0.169 and P(OD) = 0.831. To illustrate 

Bayes' theorem, consider computation of the rate in the first row of 

Table 1. Here the exposure is "unit A," P(E 1 D) = 43/61 = 0.705, and 

P(E [D) is estimated as 78/150 = 0.520. Thus, using Bayes' theorem 

P(D 1 E), the rate in unit A, is estimated as 

or 21.6 per cent. 

- 107 - 

P(D | E) = 0.705(0.169) = .216 

0.705 (0.169) + 0.520 (0.831)

.oó'io..o. M..:an. V1it. Vol, 0, PP. {5g-lb6. lerogsrlm Pies., 1976. Printed in GfetI Bilnain 

PATIENT _LOW ANALYSIS AND THE 

l)FlIVFI,¡Y OF RADIOLOGY SERVICEt 

BENJAMI N L¿E;, GEORGE REVESZ§, FRANCIS SHEA§ 

aind ROBERT CALTAGIRONE§ 

Temíple Universily, Plhiladelphia, PA 19122, U.S.A. 

(Received 5 February 1976) 

- 108 - 

Abstract-in recent years there has been an increased awareness regarding the cost of radiologic health care, and the 

patient delays encountered in the delivery to the consumer. The purpose of this paper is to demonstrate that, at least 

in one case in the Diagnostic Radiology Departinent al Temple University, the assumplion that better service can be 

given to patients provided more technicians and orderlies are available, is not valid. The facts tend to indicate that the 

real problem lies in scheduling techniques, and improved utilizalion of available equipment. Therefore, it is safle to 

conclude that for improved radiologic services, the emphasis should be directed iowards the design of the management 

systems and scheduling techniques, and not the staff andior facilities. 

INTRKOU;C'LION 

At a time, when there is a continued increase in demand 

for health care service,it is assumed that a reason for 

inclfrclivc scrvie lo pafients is the shortage of manpower 

and facilities. An analysis of Ihe I)i>gnostic Radiology 

Department (DRD) at Temple University Hospitai was 

organized, searching for reasons associated with the long 

waiting times of patients in the DRD and the low 

utilization of ecuipment. It has been previously assumed 

that betrer service could be presented to the consumers if 

more technicians, orderlies and examination rooms were 

available at the hospital. However, such a move would 

have a tendency to increase the costs of radiologic 

services al a time when costs are increasing at an alarming 

rate. In many instances, managers suspect reasons for 

their inefficient systems performance, but it is only after 

an in-depth study is made that the basic causes of the 

inefficiency are evident. Moreover, the suspected reasons 

appear to be of secondary importance. 

In this particular case, the long periods of patient 

waiting time cannot be decreased by adding more 

technicians or orderlies as some had surmised, but that 

patient waiting time can be reduced by improving 

scheduling techniques. Better management methods; such 

as more sophisticated scheduling algorithms, automated 

systems and computer control can be utilized with the 

present staff and equipment. The effects will contribute to 

decreased patient waiting time, and the total time spent in 

the DRD. At the same time, patient service capacity 

shouid be increased in the DRD with the present number 

of rooms, technicians, orderlies and staff. This increased 

cpacity will then enable administrators to accommodate 

the expected increase in demand for radiographic services 

as predicted in Knowles II, Morgan[2] and National 

Advisory Committee on Radiation[3]. 

Recently, there have been several reports alluding to a 

potential shortage of radiologists in thc United States in 

the near future. This potenlial shortage is predicted since 

the demand for radiologistís' time is increasing al a more 

tl'his work w;s supportedl in p;rit hy (¡;tl;' ;M 1454S8-0(. 

National Institute of General Medical Scienet,, United States 

Public Health Service. All correspondence shoul) be direcled lo 

the first author. 

tSchool of Business Administration. 

§Department of Diagnostic Radiology. 

1Table I lists and explains all symbols. 

rapid rale than is the supply of radiulogists. Any new 

national health insurance program, if enacted, would 

presumably cause an even greater inbalance between the 

demand for radiological services, and the supply of 

radiologisis. 

Improvements in DRD can be categorized in three 

areas. The first is increased utilization in personnel and 

facilities which has been reported by Lindhein[4] and 

Revesz et a1.(5,6]. The second area is improvements 

attempted by simulation technique. See studies by Covert 

et al.[7], Jean et al.[8], Kenny and Murrayl9] and 

Lodwick[10]. The third area emphasizing Computer 

Scheduling and Control is reported in Donald and 

Waxman[Ill, Hansen and Sniderll[2 or Hsish(13]. 

Computerization of manual methods has demonstrated 

that scheduling can also be applied to large departments. 

To date, however, no mention has been made of an on-line 

scheduling system, which can be dynamically updated as 

the patients are processed. In addition, limited research 

has been conducted concerning the development of 

scheduling rules which have general applicability. 

In view of the many possible policies that can be 

considered for improving the delivery of radiologic care, 

it is necessary to present the data quantitatively for 

analytical purposes. Consequently, this paper will investigate 

some of the effects of these policies on the DRD. In 

most instances DRDs at hospitals are confronted with 

similar problems. The methods of providing radiological 

services is apparently identical at all departments. The 

patient arrives at the department and immediately enters a 

sequence of service facilities, which culminates with an 

X-ray examination. It should be recognized, however, that 

departments may differ in the number of examination 

rooms, the procedure used andother non-major facilities. In 

most DRDs, however, the patient flow is basically the same. 

It appears to be more logical and convincing to analyze the 

DRD at Temple University Hospital, using its specific data 

and not discuss a hypothetical DRD. Moreover, the 

generalized model developed in this study is of sufficient 

generality to be associated with the DRDs in most hospitals 

with no more than minor modifications. 

SYSTEM DESCRIP'lION 

The system under study is the Temple University DRD 

which has an annual volume of 72,000 X-ray examinations. 

The service is provided both for inpatients (IP) and 

outpatients (OP), and the ratio of IP/OP is 55145. Figure

160 

I indicates the flow of patients through the department. 

Many of the patients are scheduled. Experience 

has shown that the scheduled paticrts will either arrive 

and enter the department too early, or will arrive too late. 

Scheduled inpalients are assisted by orderlies to arrive at 

I)RD in 20 min, depending on lile ivailability of clevators. 

Due to these random elements, it is assumcd l:al the 

patients enter the DRD with an interarrival distribution 

time of f(t); with probability P, of being an outpatient, 

and wilh prohahility I - P, of hcing an inpatient. The 

inpatients will wait for ocie of I¡le ¡enl orderlies (Ot : 1(1) 

for assistance to be taken to the control desk by one of 

three means of transportation; walking, wheelchair and 

stretcher with probabilities P2, P 1, P,, respectively. An 

oulpalient reporis to thc reception Jcsk itt which inme the 

individual is serviced for time r,. A fraction P5 of the 

outpatients will change to a hospital gown which takes time 

r2 if one of the dressing rooms (DR = 16) is ava!lNble. The 

outputient then reports lo the control desk and sNpends time 

,t. Each patient using thc dressing rooms will lock ¡he 

room for the entire period of the examination and therefore 

prevent others from utilizing the room until the individual 

is released. 

At a typical DRD, Ihere are over 200 different types of 

examinations which can be grouped into 13 major 

categories with minor variations within a category. 

For the most part, however, the demand for the 

different radiographic examination will vary over the day. 

Gl. BE and IVPt are performed during the morning hours 

before the patient has eaten; while a Myelogram is usually 

scheduled for the afternoon. 

Table 2 presents a list of the 13 examinations with the 

corresponding mean examination time 74, and the range 

for cach category. Pi(j) is the probability of having 

examination type j at period i = (8-10 a.m.), i=2 

(10 a.m.-12 p.m.), i = 3 (12-2 p.m.), i =4 (2-4.30 p.m.). 

The patient is then assigned an Examination Room 

(ER= 14), at the control desk. This assignment is an 

important one since it determines the length of stay for 

that patient in the queue. The DRD at Temple is one 

where not all examination rooms are equipped with the 

tExp'anations of category names are given in Table 2. 

R u. IrV orr 

,, nl 

Table 1. List and explanation of variables 

inpalient, l Ool...n.I 

.- 109 - 

1d,--. v 1 C-l'eI - ( ,* 

Th Wpellntait lo o r n po Dcesng it r 

Fi1. nrow. Maeporflol owharl tpatien 

Tjma 

sam facilities, and oy cenain rooms can prformrly 

specific examinations. Most of the examinations require 

one of the Technicians (TE= 10), while other rooms 

require radiologists. When the examination is completed. 

the patient waits time as for film processing with the 

probability P6 that the individual may need additional 

films to complete the study. If supplementary films are 

necessary, the patient then returns ¡o the control desk 

with high priority. If the study is adequate, the outpatient 

is released, and the inpatient will wait for an orderly to be 

returned ¡o the room which takes time Te. 

EVALiJAiON OF SUCH A SYSTEM 

It is important to recognize that in most large systems, 

there is mnore than one measure of performance. Thus, it is 

not surprising that in this case there are several, some 

Vt..' i.lMl, ' -xp i .na11It'on DI i lelstill ¡nll 

f (t) 

Inter Arrival Time letween Two Patienta 

rl Service Time At Reception Desk 

,t 2 . eDressing Time 

r3 Servicr Time At Control Desk 

T'1 Examination Time 

y5 Iíilm Process Timne 

(Walk 

T1 t ~Trau.,lorlotion 

Time To or rroma ~ ielechair 

1'.- IhgUinrlmelll r 

iStretcl.lr 

P1' 

k' S 

I''rceintoge Of Outpatients 

Ierentage Of Walk/ Wichelehair/ Stretcher 

l'ercentage Or Outputients Needing Dressing Room 

p; PGl'.ce'tag Of' I etrlntu N ti.dilg Adlitinoial tilm 

ER í ,,I:xtnll Jtoll NoOmS 

OR Orderlier 

TE Technicians 

DR Dressing Rooms 

Un 1 unm 

Uniform 

Uniform 

Constant (t1 lin.) 

See TabLc 2 

Constant (10 Hill.) 

Unfarorn 

Uniform 

Uli fornm 

45X (55% lnpatlent) 

5/20/7W% 

803 

llI 

10 

16

latient iow analysis and the delivery of radiology service 161 -110- 

Table 2. Data concerning various examinations: probability of having the examination (as function of time in the day). 

times(mean.range).roomequipped toperform theexaminations(before í I a.m. and afier 1 a.m.) 

, 1 2 3 4 S 6 7 8 9 10 LI 1: 1) 

NlAl.' f El.,,Il;il.ll... " 

SLik,. SIi'1.i" Cl..t .. n Ny..lo Abd. GI BE IvP tojo Fluor., PEt; :II 

Pi (J) 8-10 A.M. 0 0 32 0 0 0 20 20 20 0 0 0 u 

f 2 (J) 10-12 A.M. 13 17 42 20 0 0 0 0 0 2 2 2 

P 3 1 () 12-2 P.H. 13 17 42 20 4 0 0 0 0 1 1 I 

r'. (J> ¿-:30 P.M., U1 I 8l 20 0 12 0 0 0 0 0 U 0 

"an Exao. Time (HIn.) 12 t16 6 10 60 10 100 45 60 40 20 YU i5 

Rane Ex~.. TI~ (

162 

CONGESTION POINTS ANU CONTROL.ABLE 

VARIL'.LES 

It should be recognized however, that such a complex 

system has several congestion points. Specifically, some 

of the more important ones are: 

Q, In front of the reception desk; 

Q2 In front of the dressing room; 

Q, Inpatients waiting for an orderly to be assisted to 

the department; 

Q4 In front of the control dcsk: 

Q, In front of each examination room; 

Q" Inpatients waiting for an orderly to be returned lo 

the room. 

In addition to the queues, there is the problem of 

technician flow and orderly flow. In many instances. 

Iechnician and orderly flows are delayed because of the 

congestions in the system which are beyond their personal 

control. 

Before evaluating the behavior of the system as results 

from changes in the input, it is necessary lo determine 

what parameters can be changed. The capacity of the 

facility is greater than the volume of patients and, 

therefore, it does not seem reasonable to reject patíents. 

Thcrc is a limil to Ihe numbor of patients in somc 

caUegories, but, in general, more patients can be seen than 

at the present level. The scheduling procedure determines 

the time of arrivals but the number of patients is under 

limited control. 

There is some control regarding the patient's arrival 

time, but in many cases this situation is limited. As noted 

previously, some of the examinations may require an 

empty stomach, and therefore must be completed early in 

the moming. Another problematic area is that patients will 

arrive two hours before or after the scheduled time, and 

must be accepted for treatment. In other words, there is 

some control on the arrival but it is often limited by 

extraneous circumstances. 

Certainly, the parameters of the system can be 

changed; i.e. number of technicians, the number o' 

orderlies and the increase rate of service at the reception 

desk. But most important, the decision that is made at the 

control desk regarding the room assignment appears to be 

one of the key factors for the efficiency of such systems. 

In this sense, one way to improve the performance of the 

system would be through changes in the room assignmenlt 

decision. This decision is not easy since it must take into 

consideration the rooms that have the :uipment to 

perform Ihe examination, as well as the lines in front of 

each room. At present, this decision is performed by a 

clerk. An example of the decision tree for the room 

assignment of a patient requiring a chest examination is 

presented in Fig. 2. It should also be pointed out that 

Rooms 4 and 7 are available for chest X-rays before 

II a.m., and Room 4 does not have the equipment to 

handle stretcher cases. Decision trees for other examinations 

are even more complicated when requiring multiple 

phasc examinations, and more than one room as for cases 

concerning upper gastrointestinal studies (GI), and barium 

enemas (BE). Table 2 clearly indicates the rooms that can 

perform the various examinations before 11 a.m. an " - 

II a.m. (before 11 a.m. some of the rooms are used tor 

special studies). 

MTHOWDOLOGY 

The department described above is a large and complex 

facility. The behavior of one part of the system is 

dependent on the output of another. Therefore, the 

B. LEY et al. 

Fig. 2. Chcsl palient lecision trec for nrim assignment. 

- 111 - 

arrivais to one service station represents the departures 

from another service station. Consequently, variables are 

interrelated and there are numerous system parameters. 

Since the matter is complex, it will be impossible to 

analytically optimize the operation of the department, and 

to develop compact results in closed forms. Therefore, a 

decision was made to utilize simulation techniques which 

have been proven useful in similar applications. A 

simulation program was then written for the patient flow 

in DRD. More specifically, GPSS (General Purpose 

Simulation System V) is combined with Fortrar IV, and 

the program has been implemented on an IBM 0/165. 

For each experiment, the following sample siz, vas 

taken. Four vectors were generated: P(i) = (P,, P, .... 

":.; = 1, 2, 3, 4. All four vectors have a similar patient 

mix ¡( -me percentage of the 13 categories). For each 

one of t. - %k tors P(i), 10 repetitions were simulated 

with rando,,,._í. '-r the stochastic variables. In certain 

cases, a larger nu,. ' r of runs were performed in order to 

test convergence. T' - randomness was in the service 

time, method of patí..i, transportatinn (walking, wheelchair, 

stretcher) and oth,. ctochast's i,.ameters. Thus, 

cach experiment included 4 uu, x 160 aticnll/day x 10 

repetitions, which resulted in 6400 patients. To observe 

changes over vanrious number of technicians, ;..y 4, 6, 8, 

10, the sample size was 4 x 6400 = 25,600 patients. 

VAUDATION 

One of the important phases of any simulation is to 

verify its performance as to whether or not it replicates 

the actual system. In order to satisfy the requirements of 

thc study, statistical information w:as derivcd from DRD 

over a two year period, includin? Jata on performance, 

and corpp.-zd with the results from the simulation model. 

The results which include a X 2 test for goodness of fit at a 

level of significance, a = 0.05, c' ,níy indicates a good 

agreement between -tu; jRD and the simulation 

model. The res" ,.resented in Table 3. 

Tests wer .,ipleted in the simulation model to 

measure t' ..rtormances; under a typical daytime load 

of 1i · .ents, under a heavy load of patients (220 

pat:. ..s) and under a light load of patients (135 patients).

¥aticnt flow analysis and the delivery of radiology service 163 112 - 

Tablc 3. Comparisonbetweensimulationandactual syslem 

135 P a i e n ts ay b60 P a t i e t s I a y 

waiting Time Before StuJy (Min) 'leasured Sinmulation X2(x2) * 

(easuread Simulation , ( ') 

Al Done, 

NI ,:e 1 1 uncou, 

All. Patients 

Toitl 1imr in lepartnrnlt (Flii) 

Al I lli, .. 

Mlajor Fluoros 

tliscellaneous 

All Patients 

Techltni.í:w Utslizationl ; 

Orderly Utiltzatilo : 

Room Utili:ation : 

1 7 (31) 

1 2 (58) 

r,~ (.'n 

i1 (1 ,1 

,' ((35) 

.1 I .1.1 

102 (i5) 

63 (155) 

3.> 

42 

5! 

IB (35) 

Iih (G2) 

.It (2.11 

1i (II,) 

23 ( 35) 

.1 

(i.s) 

(,7 

(l' IS 

90 

70 (( 35) 

31-31 ranlgL 

I-433 rangc 

34-41 range 

\otc : 135 ,liatencs. * H II deiliciai*. 10 orderlics. 12 cxaJinJtiOfl rooms 

li(o patiOlVcts. IO 10 %cdalll, llai 10 o;'dei.l.s. . I14 ;unttiittio i l roums 

6. 78 (.93) 

7.t (.9JU) 

( · ) nae numbers in parentlhce.se aro the number of patients that lthe times are based upon 

21 

21 

25 

(bl ) 

(69) 

(18) 

1 (¡16) 

58 (61) 

121 ( 1) 

81 (18) 

69 (166) 

The measured data and thle simulated data are froa the same population at the leyel of a * .05 

The results disclosed that the model behaved as expected. 

(See Shea el ul. t14].) Having ascertained the aceuracy of 

the model with respect to the real life situation, sensitivity 

analysis would then be pcrformed with respect to the 

parameters, and the measure of performances at various 

conditions. The cxperiments concerning the simulation 

can be described as follows. 

A. The effect of changing the number of technicians 

Presently, there are 10 technicians in the core 

department consisting of 14 rooms. Several "runs" were 

made with the simulation model, varying the number of 

technicians from 4 to 10. As anticipated, the system is 

overloaded with 4 technicians, and the waiting times and 

total times are very high-50 min and 110 min-as seen in 

Fig. 3. When the number of technicians increases, the 

measure decreases but only to a certain level when the 

times reach a threshold. One implication is then clear; 

waiting time and total time cannot be decreased below a 

.1- 

E 

S. 

.r4 

1oo0 

ots 

60 

40 

20 

'I 

%x 

: 

4 6 a 

- 

19 

Number ot tectinicians 

Orderly util. 

Total time 

\ Technicion util. 

--Room u(til. 

' Dressing rm. ulil. 

Waiting time 

No. potients ct 

4:30p.m. 

Fig. 3. 'Me effect of changing the number of techiricianís. 

So 

54 

38 

17 (40) 

16 (70) 

20 (.J( 

1: (1:tl 

t*J (ctX) 

50 ( l(> 

87 (20) 

65 (t60) 

51-55 range 

50-7L rangii 

$g-44 range 

2.01 (.. 35) 

certain leve¡ by adding more techíticians. This fact can be 

explained by nu>ting thart the problem is nol with (he 

technicians, buz with a combination of the number of 

orderlies, rooms and ¡he scheduling procedure. The DRD 

operates efficiently with 8 technicians, but an additional 

technician is necessary for absenteeism. 

Similarly, the: number of patíenis in the: department al 

4:30 p.m. decreases, with the number of technicians, but 

the residual number stabilizes with 8 teáhnicians, and 

does not decrease further with additional technicians. 

Technician utílization decreases as the number of 

technicians increases due lo the faelt that the amount of 

work for afi technicians remains constant. Therefore, 

increasing the number of teclínicians results in a lower 

utilization for cach technician. 

The utilization of other service stations is stabilized and 

constant when the number of technicians rises above (I. 

Orderly utilization is about 0.70, recoption desk utifization 

is about 0.55, room utilization 0,43, and control desk 

utilization is, about 0.30. lt is important lo point out that 

none is affected by the change in the number of 

technicians. 

B. The effect of changing the number oj orderfies 

The next pararneter lo be varied is the number of 

orderlies in the system. At present, there are 10 orderties 

and Fig. 4 demonstrates the effect of changing the numbor 

of orderfies from 6 lo 12. Figure 4 clearly ¡radicales that 

the number of orderlies has ¡¡ttle affect on the plotted 

measures of performance. The most aflected is the 

utilízation of orderlies which decreases as the number of 

orderlies increases sínce the same amount of work is 

shared among more orderlies. lt also has an effect on the 

number of patients in the department at 4:30 p.m. This 

number decreases from about 25 for 6 ordlerties lo 10 

patients for 10 orderlies. Other than the two changes, the 

measures of performance are almost unaff ected and reach 

some constant value for 8 or more orderties. 

C. The effect of decreasing the nuniber of examnination 

rooms 

The succeeding experiment involves the possibility of 

reducing the number of examinatíon rooms from 14 at the 

present time to, 12 rooms. Table 4 ¡radicales the changes in

164 

c 8o 

Table 4. The cffect of chaing the number of examination rooms (at 135 patients/day, 10 technicians. 10 orderlies) 

Toltl Tiwe 

,l the S¥ysttr *(' 1 patients) 

h'a ting TiJmne 

( J.,pal jont I) 

(Ohtpnttents) 

i;. Ils. ,%y l.l.. JAl 1 . :. 

(Inp.ati:nts) 

(Outpatientc) 

Nwu.bLr or P'atient at t 4:30 

Teclmbilc:n Utiti.xaioii 

Ordlerly 

12 KRar 11i Room. (AL I-rvs.irt) 

79 Mi,,utes 58 Mlinutes 

99 

F,4 

17 

lO 

9 

.110 

·3lO 

I)rl.-Ksl- g Ro ' .58 .31 

60. - Totcl time 

ci Orderly util. 

40 - -0- - . Room util. 

20 

20 '________ __ _" _ " _-. W oitingtlm e 

No. ptients at 

4:30pm 

6 8 10 12 

Number of orderlies 

Fig. 4. Thc ellecl of changing the number of orderlics. 

the performance parameters and illustrates that total 

times and waiting times are going to increase significantly. 

As a consequence, total time for outpatients increases 

from 69 to 99 min, and waiting time increases for all 

patients from 12 to 28 min; for inpatients it increases from 

8 to 17 min, and for outpatients from 20 to 40 min. Thus, it 

is safe to generalize that system performance is obviously 

affected to varying degrees depending on which rooms are 

closed and this factor is now being investigated. The 

addition of examination rooms was not considered as a 

measure to decrease waiting time becausc of thc hb;ih cosi 

of each examination room ($100,000-$200,00,,. h 

surplus existing in the departments' capacity to handie 

more patients, the hospital administrators will not invest 

vast sums to decrease patient waiting time and total time. 

D. Changing the scheduling method 

An experiment was designed to observe the system 

performance for a change in scheduling methods. Se:v.. 

different scheduling procedures were considered: 

1. PS-Present Scheduling. At present, the room 

B. Lev et al. - 113 - 

.31 

1Z 

20 

.61 

.37 

assignment is based on the smallest number of patients in 

front of each examination room, regardless of the amount 

of work each patient requires. Once a patient is assigned 

to a room, the individual will not be changed to another 

room and no switching is permitted. 

2. MWL-Minimal Work Load. This scheduling procedure 

assigns the patient to the examination room 

according to the minimal work load in front of each 

examination room. The work load is defined as the total 

expected time that the patient will use the examination 

room, including the one in process. 

3. MMWL-Modified Minimal Work Load. This process 

is a modification of MWL which considers the time a 

patient enters an examination room and t: ime a 

decision of the present room assignment is i,. le. It is 

evident that this concept is more accurate and beu,,r than 

MWL. In this technique, an examination room is assigned 

*.- 4 on the smallest load in front of each examination 

roo-.., - .. ibtracting the process time that the patient has 

alreads .. - in the room. 

4. CQ-Cc- m' Queue. This algorithm' assigns the 

patients in one &,u.e queue. As one of the examination 

rooms is availabi -e first patient in the common queue 

from those who can ta- e the exami.i` ns in that room, is 

assigned to the room. 

5. MCQ-Modified Common Queue. This process is a 

modification of CQ. While in CQ, the natients were 

ordered according to the time available to lake the 

examination, MCQ orders the patients according to the 

arrival time at the radiology department. 

6. SPT-Shortest Process Time. As pointed out earlier, 

the decision by which patients enter the assigned room in 

PS is "First Come, First Served". In SPT, the decision is 

made according to the shortest examination of all patients 

waiting for this particular examination room. Consequently, 

patients with short oxamination times have 

priority over thosc with long examinalion times. 

7. TSPT-Truncated SPT. T'' -i is a mudification of 

SPT scheduling by acsignir - ' on priority to patients who 

have been waP:: .Éer beyond a certain time. The 

disadvantar ,..' i is that patients with long examination 

tim :., get low priority and stay in the system for 

a lonp .0e. Truncation was done lo guarantee that 

onr , patient waited over a certain amount of time, the 

.nuividual would enter the room regardless of the length 

of the examination time. 

Four different simulation days were performed consist-

ing of the same patient mix and arrival times. Each day 

was repeated 10 times for random service time, with the 

same expectations concening random personnel characteristics 

(dressing time, transportation time, etc.). Thus, 

for cach one of the seven scheduling procedures, there 

were: 

4 days x 160 dPai'y x 10 repetitions = 6400 patients. 

day 

Table 5 summarizes the data in relation to this experiment. 

It is clear to see that MMWL is superior for all 

measures of performance with the exception of two cases. 

One case is where the number of patients in the system at 

4:30p.m. is 15, while the best (MWL) is 14. The second 

case is when the total time for which 95% of patients 

complete their service is accomplished in 165 min, while 

the best (PS) is 163 min. 

Therc is a significant differcncc between CQ, MCQ and 

the uther five scheduling methods. In all five procedures, 

the examinalion room is assigned to the patient at the time 

the person leaves the control desk. The patient moves 

physically, and waits at the front of the examination 

room. Howcvcr, in rega.rd lo CQ and MCQ, this is not thc 

case. The patient has to wait in a common waiting room 

until an examination room becomes available. At that 

time, the technician returns to the control desk and 

determines which patient is next in line and can be 

examined in this room. The technician has to identify the 

patient and direct the individual to the examination room. 

In some instances, the distance may be as great as 50 

yards, and the patient may be on a stretcher. It is then safe 

to assume that extra activities may take about 5 min which 

was incorporated into the model. This fact may explain 

the reason that CQ and MCQ did not perform as weil as 

theoretical results previously indicated. Once again, this 

study demonstrates the possible conflict between queueing 

theory and scheduling theory. In scheduling theory, 

there is more control on the patients (jobs) in terms of the 

arrivais, and service times. In queueing theory, as in the 

department described above, there is more randomness. 

Waiting time before study 

OP 

IP 

Both 

Total time in thce system 

OP 

IP 

Both 

No. of patients in system 

at 4:30 

No. of patients waiting 

over 60 min for scrvice 

Patient flow analysis and the delivery of radiology service 

SPT and TSPT were clearly the most unsuccessful 

techniques among the seven procedures. SPT in theory 

should produce the best mean waiting time, however, this 

result tends to diminish as the flexibility of machine 

(rooms) selection increases (see Waysont15]). This 

appcars to be the case in the radiology department where 

examinations can be performed in several different 

examination rooms. 

When evaluating the various procedures, attention has 

to be given to the feasibility and to the cost of 

implementation. The CQ and MCQ procedures can be 

implemented manually but would require a considerable 

amount of bookkeeping, and a means of scanning the 

entire queue to determine which types of examinations 

are waiting. In addition, the SPT and TSPT procedures 

could also be implemented manually. However, it is 

anticipated that certain difficulties would arise in attempting 

to implement TSPT manually. Moreover, the MWL 

and MMWL procedures will require a computer to 

constantly update the expected luad. 

SUMMARY AND CONCLUSION 

This culmination of observations had desirable results 

and a model of the DRD has been developed. The model 

has been tested and verified for replication of the 

performance of the department at a satisfactory level of 

accuracy. In addition, any changes in the department are 

tested and verified first on the simulation model. 

The model used for the study indicates that increasing 

the number of technicians or orderlies will not decrease 

patient waiting times, and total time in the department. 

Increasing examination rooms, which may be promising, 

was considered and rejected because of high costs. It 

should be pointed out that decreasing the number of 

examination rooms is costly in terms of higher waiting time 

and total time. A recommendation was then made that 

there is no need for additional technician and orderly staff 

to improve service, and that a reduction of the number of 

technicians by one will not affect the department's 

performance. This is based on the fact that there is no 

Table 5. Comparison of 7 scheduling procedures for various measures of performances 

Waiting time within which 95% of 

palienis reccive scrvice 

Total time within which 95% of 

patients left the system 

'Best value. 

'Off by I patient. 

'Off by 2 min. 

PS MWL MMWL CQ MCO SPT TPST 

33 

18 

23 

79 

62 

68 

28 

12* 

18 

75 

57 

64 

26 

12' 

17* 

72* 

56' 

62' 

32 

19 

23 

78 

67 

71 

30 

19 

23 

77 

65 

69 

35 

19 

24 

81 

63 

70 

31 

19 

29 

77 

66 

70 

20 14* 15' 18 18 18 18 

16 12 II1' 25 22 20 20 

69 64 62* ' 99 93 101 88 

163' 172 1652 177 174 188 183 

165 -14 -

166 B. LEV etal. 

significant difference between the performance when there 

are 9or 10 technicians and 10 orderlies (see Fips. 3 and 4). 

As a result of the study, a decision was thn issued to 

change the scheduling procedure to MMWL. In order to 

implement (he procedure, a PDP-II Mini-Computer was 

purchased and the operation of the DRD is now 

computerized. In particular, the room assignments which 

were performed manually are now completed by MMWL, 

taking into consideration the anticipated load on the 

rooms, and updatcd feedback from the rooms conccrning 

the amount of remaining work. This is accomplished by 

the technicians who report to the control desk aftter 

completing the last patient, and the control desk then 

notifies the technician of the next patient. However, it is 

too early to report on the efficiency of this dynamic 

on-line scheduling system. Moreover, the staff is currently 

involved in the phase of adding options and functions. In 

any event, the first reaction indicates promising results 

which are al Icasi as good as thc previous performance. 

REFERENCI 

1. J. H. Knowles, Radiology-a case study in technology and 

manlipwcr-II. New Eilg. J. Med. 280. 1326 (1969). 

2. R. H. Morgan, The emergence of radíology as a major 

influence in Anmerican medicine, Am. J. Roentgenology 11l, 

449 (1971). 

3. National Advisory Committee on Radiation, Protecting and 

Inproving Health through the Radiological Sciences-Report 

to the Surgeon General, Public Health Service, p. 9. U.S. 

Department of Health, Education, and Welfare, Washington, 

D.C. (1966). 

4. R. Lindheim, Uncoupling the Radiology System, pp. 55, 150. 

University of California Center for Planning and Developnwnl 

Rescarch, Berkeley. Calif. (1967). 

5. G. Revesz, F. J. Shea and M. C. Ziskin, Patient flow and 

utilization of resources in a diagnostic radiology department, 

Radiology 104, 21-26 (1972). 

6. G. Revesz, F. J. Shea, 8. Lev and M. S. Lapayowker, 

Computer aided design for improving the delivery of 

radiological services, In Planining o/ Radiological Depuartents 

(Edited by M. Kormano and F. E. Stieve), pp. 10-106. George 

Thiemc, Stuttgart (1974). 

7. R. P. Covert, G. S. Lodwick and E. W. Wilkinson, Simulation 

modeling of a diagnostic radiology department, Proc. Conf. on 

t140n) 

(No, .. '\ 

McCreadie, . .' . Radiology departmenr planning. Hospi¡al 

Health and! aarai ?lenA (London) 26, 26 tJulb 1963i. 

Nuffield Provinciaí nos, ', Trust. Towards a Ctlarer Vi.'. : Ti, 

Organization of DL,. tic X-Ray Departnmitnls. Oxford 

University Press. Lond- '962). 

Schaefer, J. E.. Temple L...vtrsity Hospital Adminitration. 

Personal communication (197;' 

Shuman, L. T. and Wolfe, H., A Gen. ....i Model for Hospital 

Micro-Costing. University of Pittsburg Press (19702. 

Theusen, G. J. and Sullivan, W. G., Cost sensitivity analvsis for a 

radiology planning problem. Proc. 39th Naot'. ORSA Meeting, 

Dallas, Texas, May 1971. 

Trecate, A. and Zoboli, C., Elements in planning a radiology 

department for a general hospital, Cliniche Moderne (Milan J I. 

I17 (Jily/Aug. IX11I). 

Warner, T. N., Use of simulation to generate sheduling heuristics 

in a diagnostic radiology deplrirnent, Master's 'hcsis. MIT 

(1968). 

Welch, J. D. and Noble, D., Functional stud;ies in hospitals and 

particularly in X-ray departments, Riidi;,,kgy ILondon ) 27. 131 

(May 1961). 

l"' stm;n. C. E., Planning of new ,iiagnostic d,' irtnients, Brit. J. 

Radiol. 35, 265 (April 1962). 

Widegren, G. B., Georgia ' 'utJe ,' i'echnology, Personal 

communication (1971' 

- 115 -

1,# 

- 116 - 

Program Evaluation Techniques in the Health Services 

Abstract: This article addresses the issue of program 

evaluation in the area of heaith services; examples 

are drawn from the field of mental health. Current 

arguments concerning the goals, characterisiics. ;iad 

methodologies of prograni evaluatlion are discusscd 

Introdtc'tlio1 

Program evaluation is currently a subject of major attention 

in the health services field. Much of this attention has 

centered around the definition of "evaluation" itsell'. Simply 

stated. the purpose of evaluation is to lind ouit what worked, 

what did not work. and why.' However. the proper use of 

the evaluation restilts is frcquently a source of disagreement. 

Levey and Loomba 2 distinguish bctwcen on-going evaluation 

and rctrospectivc evaluation as follows: The purpose of 

on-going evalualion is uo iiieasttre prugre'ss lowrd prograni 

goais so as to direct control over a programn. In contras.t, relrospective 

evalutlallio is coniidteled lt detler't ille Ilc elYleccl of 

a prograni so as to facililile program IplRlnintg. 

If. as is generally accepted. prograim valuation is the determination 

of the degree of progress in achieving program 

objectives. then cvaluiation is a tool for control. However. 

there also exists the opposite view-that tihe consideration 

and evaluation of predetermined gouals is nol only unnecessary 

but also possibly contaminating." Thc concern here is 

thai the evaluator will develop tunnel vision hy accepting Ihe 

validity of tlhe prograni goals and Iiis overlook olhlir. per'haps 

more importalnti. cll'ects ol' hc pl'gl'ln. 'IThis viewpoiii 

of evaluatioun as an uinbiased (iii itermis oi Ihe progrilla's 

Address reprint requesis to Dr. Jack Meredith. Associate Professor 

of Management. Florida In.rnational University. Tamiami 

Trail. Miami. FL 33199. Tlhis paper~ias submitied lo Ihe Journal November 

17. 1975. revised. and accep!ed for putblicatlion June 1. 1976. 

AJPH November. 1976, Vol. 66, No. 11 

JACK MEREDITH, PHD 

and two generally useful quantiative evaluatlion models 

are.presented. The models are compared and their 

advantages for clinicians and administrators are detailcd. 

(Am. J. Public Health 66:1069-1073, 1976) 

goals) analysis thus appe;ars to place evaaliuaion a;is ;ilIpttin,,R 

a 

tool . 

There is no need to make a choice heíween these t%-o 

viewpoints. Evaluation can. and should. be used for both 

planning and control aund thus serve as an all-important feedback 

link bctweei tliese cruciail 'unctions. 

This paper compares two general-purpose programn evaluation 

models that can be used for both planning and control. 

The models are ilhistrated with applications in the complex. 

ill-defined area of mental health. To ililstrate the models 

i¡ is asstimed llh;i psychiatric ciaes can be specified 

objectively. outcome criteria can be designated anid agreed 

iupon. Irealment progranis remain silable. anmd patient selectiol. 

e¢tlr C lllen selccjtioll, ;antd (lll tside ilnllCll¡cce ;ar'e controlled 

or accounted for. Clearly. Ifew situations will satisty 

all these assumptions and ltius the models' outputs must hbe 

tempered with experience. judgment. intuition. and other 

quailitautivc flactors hefore a linal decision is reached. If so 

tempered. such models can provide important information 

tor the clinician or administrator. 

All Index Model 

The first model. adapted fromn Haipern ;aed B3iner.' is 

;ain indlx Inotlcl th;it cximfilles prog;lan's l cllectiveiess 

(how W,1. it tachieves its goa;s) and its eliciencs resuils per 

unit coD. Sulmmalry ralios lor these two meal ;' are developed 

!p program value and prograni co ,ta. Mental 

hcalth lograms may result in changed Jindi ftinctioning, 

in irvention. and protection of the indi 1ufI.! andior so-

MEREDITH 

- 117 - 

ciety. However. change in the functioning of the individual is 

pcrhaíps thc le osl impoirtant restilt. 

A sturroga.e ncaistl' c of thie v;alue of individu: d func'iolling 

is suggested hy Ha;lpcrn and i3inner. hbased on the individual's 

primuary nicanis ol'contíihuting io society: his economic 

productivity. Since the pitieni's niost recent economic piodutctivity 

ima;y ac¡¡irel lynol reillclct tii polcntlial duc Io his 

current mental handiciap. his ecoioinic píodiuclivily is bcased 

instead on (he average of his previotus 12 months' earnings 

and the expected annual earnings of iin average member of 

his cducaiional and *cctipatioinal peer groip. Annmilal earnings 

is used, rather thaln lotael expecel d fulur earnings. ¡o aidknowledge 

the general impermainencc of the improved level 

of functioning due to the program. A lower hound on earnings. 

the mininitim annuail wagc. is impossed to reflecl society's 

intplied mniniial worilh of ill iiitdividlis. 'I'his ohviales 

the dillictiltics of infl'cring a wage loi' housewives. students. 

the severely retarded, children. tile retired. and other 

such groups. 

'l'hc thought of usiFng ;.ll individutil's lirsonllal cuolioiice 

productivity as a surrogate for progrant value may be anathema 

to some because this suggests accepting only the rich or 

the working male into trealment programs in order to obtain 

the niost cos(-ctl'ccivc prograrms. Thius. a miore desirable allernative 

may be to utilize tile mcan national (or regional) income 

so as to equally value all patients. Another alternative 

would be to develop monetary values based on broader oncepts 

than wagcs. For simpliciíy's sake. however. we will illustrate 

the example here with individual economic productivity. 

The surrogate measure of improved functioning is then 

the product of'the patient's average annual productivity and 

an index of cltige. * TIhc index of change is a rough meastire 

of the effect of the program on the individual. Table I lis- 

TABLE 1-Specification of Indices of Change (A Model Input) 

Discharge Level of Impairment 

Admission Level 

ol Impairment None Siight Moderate Severe 

Slight +404 0 -30%/ -70% 

Moderate + 70%" +40% 0 -30% 

Severe +100'/o +70% +40% 0 

NOTE: 100 per cent or 1.0 represents progression trom severe impairment 

to no impairment. 70 per cent or .7 trom moderate to none. etc. Level of 

impairment calegorizaion is pro orabty detormined by place on a standardized 

acate whose reliability and validity have been eslablished. 

some exaniples ol' itndices olf change. n a gross scale. a;s a 

function ol inmpaimitinent ;il aldission imand disclihige. 

Table 2 presents hypothetical data t'or an Incarceration 

programni serving thrce individuals to illustrate the decter, 

tion of program value. The index of change is based on columns 

(4) and (5). The program value is then based oin this 

index of actual change in column (6) and ihe av erage productivily 

in columnn (3). 'I'e overall progr;IIt vathle index coinputed 

in (he table. 477. is conimputed on aier patienri hasis so 

as to be comparabhle with other progranms. The m;iximnunl programni 

valie possible in colmnin (9) is fotund in the sante manncir 

;is the I'ptgraii valiue ill colun>lniii () il li ( i nti¡e\ .or Io1sible 

change is used instead of the actual. Thus. the index of 

possible chalinge for Individual I whose admission level was 

moderatie is. from Table 1. 70 per cent (discharge level of 

nonc). I:or Indlividuail 2. ihe ntmtsinium is 10'1 per cent. and 

for Individual 3. it is ihe same value as before. 40 per cent. 

Program cosis are more straíightforward. The only difficulty 

here is the decision of whether ¡o charge the program 

with onily dliec¡ cosis or whelher ¡oi inchide indirecc costs a'S 

well. 'f the patient spent díill'centi ;inl1lts ol' tilme ill dil'erent 

cosl sialuses (hospital. Ifamily care. ouipalient) within a 

program. a total cost figure may be obtained by multiplying 

thie days spent in each stlilus hy lhe dalily cost of the sltitus 

and then summing over all staluses. 

Table 3 presents the hypothetical summary program 

evaluation indices for an agency with t'our drug treatment 

programs. The incarceration program's values. columns (2) & 

(3), were derived from Table 2: the values for the remaining 

programs would be derived in the same manner. These numbers 

are interpreted in relation to annual wages-the treatment 

home saves three times the per patient wages as incarceration 

(1,510 vs 477). Thie maxinmum possible value per 

program is shown in column (3 fIor comparison with coluimn 

(2). Thus, incarceration saved 477 out of a possible 4.480 

while methadone maintenance saved les 450. but out of a 

much smaller possible. 1.753. Thus. melhau 'e maintenance 

is less efl'ective than incarcerajion in ternis uf ';iblohlt vailue 

more n: effective in relation to its potential. as shc wn by Ihe 

efft ---ss index in column (6). 

C '

- 118 - 

TABLE 2-Example of the Determination of Program Value for Incarceration 

EVALUATION TECHNIQUES IN HEALTH SERVICE 

(1) (2) (3) (4) (5) ((?5 (6 

(6) (9) 

Index ol 

Poss. Change Pgm. Value 

Previous 12 Group Ave. Avg. ol Admission Discharge Index od (Discharge Pgm. Value Pos»sab 

Month's Earn. Annual Earn. (1) & (2) Level Level Change Level od None) (3) x (6> (3) x (7) 

Individual 1. 7.400 11 000 9.200 Mod. Severe - 30% a 70% -2.760 6.440 

Individual 2 4,680' 4,680 4,680 Sev. Mod. +40% + 100% 1,872 4,680 

Individual 3 5,100 6,500 5,800 Slight None +400% + 40% 2,320 2,320 

Total (T) 1,432 13,440 

Avg. = T/3 477 4,480 

*Based on a minimum wage dl $2.00 per hour. 

ioxili;calioln. ;lilhough lnot exremenly cllictivc iii tcl's oí' ils 

potential. is ani exlrmnicly elliciien prograin. Wc mighl thercfore 

mniove o cxpand t'realment hoeilos al Ihe expense of' incarceralion 

aínd try to improve the etlectiveness o ni mehadoane 

deloxilicaiion. 

1' ablc 3 c.an also givc overall ellicicnlcyaiid cll'clivenccss 

indices for the agency's four programs. These are lound by 

multiplying each program index by the number of patients in 

that program (column 1) und then dividing by the tolal group 

sizc: 

Overall 

Agency 

Elliciency 

.82x3 + 1.16x57 + I.28X81 + .76x 163 

3 + 57 + 81 + 163 

= .97 

Overall .11 x3 + .6X x57 + .34x81 + .26x 163 

Agency 3 + 57 + 81 + 163 

Effectiveness 

= .36 

The poor performance of incarceration is minimized in the 

overall indices due to the small number of patients in that 

program. In terrns of efficiency. the agency is saving almost 

as much. in terms of patient productivity. as the program 

cost. The cellectiveness ol' .36 indicales Ihail they are achieving 

about onc-third of the maximum possible achievablenot 

necessarily a poor showing. 

'I'his cxaimple has comparetd Jill'cent progrants hbu Ihe 

same model can also be applied to the same program al dillerent 

time periods or lo grouLps of patients categorized by 

means other Ihan treatment programs: e.g.. by sex. age. or 

diagnosis. 

Ha;lper'n and Binner 4 point out une of' the inevitable argu- 

AJPH November, 1976, Vol. 66, No. 11 

mcnis agauinst the index evailiaition model when they) stite 

1hil; "... .. ileao ;tlllilI liillof ¡ . a f it iieieil1 Ihe;illih proglra;il mi;ly 

not he able to mnax;inmize his 'relurn oni investment. In fiac. he 

maiy have lo follow s;ctleics thhail lower hi. reliurn. if he is lo 

serve those who need his hclp most." IThis. of course. raises 

a tlinidamclal;l quelilon: WVlo hlhold 'cl ccive tihe hem:fit of 

¡he program's limited res'ui'ccs For exaimple. bhould ¡he 

patient who can utilize them the niost receive the resources 

or the patient who needs them the most? There are no meth- 

Mlological or technical answers to such moral and ethical 

quest ions. 

Tlie M'ra'kov Ciin MoIdel 

TABLE 3-Drug Treatment Program Evaluation índices (Hypothetical) 

' 

Eyman's' 1 siatistical evaluation model. the lMarkov 

Chain. differs significantly fiom that of H:alpcrn and Binner 

in that it focuses exclusively on the functional level of ¡he 

patient. using it as the measure of chainge and ihe basis lor 

cvallation. T'he Markov model delineates. via contingeicy 

tables. the movement of the patient (or groups of patients) 

along a scale in terms of the patient's initial position on the 

scale. 

Eyman used the Marlkov model io evaluaate the efl:ctiveness 

of a school progranm antd ain inteinsive treatment program" 

in a hospital for the mentally retarded. I-owever. the 

Markov modcl is much more powerl'ul than these limited applicalions 

suggest. FoI example. as ,¡ill be shown. program 

costs can be included in the model so that it has the potential 

of being combined with the index model advanced by Halpern 

and Binner. In addition. under very general condilions 

the model has been used' s 

to predict the probable time- 

(1) (2) (3) (4) (5) (6) 

No. ol Pgm. Program Pgm. Eliciency Ellectiveness 

Program Palienis Value Value Poss. Cost Index = {2)/(4) Index = (2)/(3) 

Incarceration 3 477 4.480 583 .82 .11 

Treatment Home 57 1.510 2,205 1,300 1.16 .68 

Methadone 

Detroaification 81 975 2,870 760 1.28 .34 

Methadone 

Maintenance 163' 450 1,753 590 .76 .26

I 

MEREDITH 

- 119 - 

varying oulcome of a trealment program on a patient. Furthermorc. 

the modcl can handie more than one treatment 

progralm ll a lime so as sequential varia

TABLE 6-Expected Change In Cohort Distribution from 1970 

to 1974 with Varying Treatment Programs 

1974 Ablily Gfoup 

I 1I IMI tV V VI D 

1970 .- -- 

Athlly 1 oliil 

Group Pahlenis 89 72 95 133 1 18 .I- bO 

í 221 73 46 46 27 11 0 18 

II 112 12 16 26 27 19 3 9 

111 114 4 9 18 34 32 8 9 

IV 96 0 1 3 30 36 18 8 

V 52 0 0 2 14 18 14 4 

Vl 20 0 0 0 1 y 15 2 

O 0 0 0 0 0 0 0 0 

tion he will have in any future year he desires frtom which he 

1may dctcrminuie whal resotirces he will neeccl to service those 

patientis ofl' ourse. distaiit protjeclions haive .ess reilialbilily 

than near projections. 

Table 6 was generated. specifically for the 197(1 cohort 

shown. fronm ai fiur-yeair probability matrix sintilar- o tilhat of 

'I';ablc 5. Th'le next tIluce sets of' s .lupplcntli y inl'ornillion 

are all shown in Tahblc 7. This table lisls the nimlber ol' years 

a patient is expected to spend in each of the groips belfore 

either dying or being placed in a community loster home (limited 

to Grotip VI as per T'l'ihe 5) li'r the firsl time. 'The total 

cost the patient is expecled to incur lip to itis eveni is also 

shown and was derived hy multiplying the yeairs spent in 

each group by the Unnual cost ol' the prograin selecied lfor 

that group (from Table 5). L.;istly. the table gives the probability 

of each of the outcomnes occurriíng lirsl. 

Table 7 shows. for example. that a patient in the lowest 

ability group (I) is expected to spend 4.4 years in thai group. 

including time spent there due to regressing from higher 

groups. abhot two and one-hall' year.s in Groups II aund 111 

each, and about lfoutr and one-half yeai;s il Grotups IV ;and V 

each before he either dies or is placed in a l'oster home. for a 

total of 18.3 years spent in the hospital at a total cost of alnmost 

$18)0.Wf. 'T'he likceilhood is ;ianlosl Iwicc as gilc;al ilitla Iic 

will be placed in a foster home ralther than dying first. a very 

positive prognosis for a patient in this group. 

TABLE 7-Expected Stay Times in Years with Varying Treatment 

Programs Until Either First Placement or 

Death 

Outcome 

Abilily Group Total Probability 

1970 . .........- 

Ability Place- 

Group 11 il IV V Years Cost (S/Pat) ment Death 

I 4.4 2.2 2.8 4.6 4.5 18.3 179.500 .64 .36 

II 1.1 2.8 2.6 4.9 4.8 16.2 166,400 .68 .32 

111 0.4 0.9 3.2 5.1 5.1 14.6 155,400 .71 .29 

IV 0.1 0.2 0.6 5.6 5.4 11.9 130,900 .76 .24 

V 0.1 0.2 0.6 3.9 5.8 10.5 113,000 .79 .21 

AJPH November, 1976, Vol. 66, No. 11 

- 120 - 

EVALUATION TECHNIQUES IN HEALTH SERVICES 

Lastly. the Maitkova molel is not limit to ti ealinlgíed e'aitli;iiing only 

Ihose progranms ' ith which ain ;igenc\ h;, erperience. For 

insl;ince. il' ; ne%' moule ' opel';ran c'onditioning hecome% 

av:d"l;¿lt oruh ;t a llotili;ítiil lo ;ill e\ ninI lllg pr'gra;l i' con- 

Ceílípliatcd. stlbjcctivc etli; iate', ofl the t'rlnmsition probahililíes 

coulltd CqUlly %%t l ¡le ued' iii l tihe lI';tií,i¿ito prolablilty 

m;itrix. 'i'hc eiffect otl'tlihe neA o modiied i prog'ailln coild ihen 

I, lc .' i '.itiiii c t ill at iil tllf Nilli ;it i ¡'la';,I es I ;ia ld 7. ''hli, 

woutld also give thc eIl'eet oin the total coNt t ttihe hospilal antl 

tiltis asclrie ;an equivllentl wsorth to the progra;in. 

Disc'..ss.ion 

Although the Markov model and the Halpern and BIin- 

¡ner nilldel Irc.ilirel ;, Cnllsidetrail;ae aimnilUSt of' ell'aortl io Ise. 

evenl more el'ort is requictd Ilo provide aleculrlic. Inc;lninigltll 

iiiput daia l'for t¡le niodels. Clc'ily. the st;tirticmal ni:;miptil.'tion 

of inacctirate daita¡ is % 'orse thani useleC -il may veil he 

mi:slalding. Tlhus. to ¡ise these imotels for tile pu'pose they 

were intended ieqtii'¢cs ami ciis.II'mots O ;illl¡tiilt tf sstmk re'l'tll 

ascertaining the reliability andt validity ol'f the rawv dtta. In ¡tddition. 

an appropriate scoring aníd recoe'ding sy>stem must be 

designed. monitore(d. and properiy titilized. And finally. the 

rcstulis of ithe ilodel ;,, ;ailyacs i¡lil Isc iierprei.ted ;antd very 

carel'tilly expl;ined IeNs ineoire.1le inl'eirences he dtirir n. 

Ftl'hernmore. l*iy ev;.ll;ion model is onlly one lootl in 

the total program evaitiation process. ''hc M;uiinko mnudel 

;aiid tlíe Il;pilcpiin ;and li inlsicr modeili l ;icl'cr;l' It' e es¡pci;allv 

useli.l inii this regai'd. U nodouitedly . moex ;ilitíivíon ilmdetJls 

for the hcalilh services swill he dcv eloped and i.come ;v;,ilable 

as time progr'esses buit hliing lihc interinm the [v.t nmodels 

described here appear' to prose,,,, the moni genieral applicaibility 

Cfor qua;italivic plrogratil xi uatltioi. \\'hen ti.ed in 

conjunction with additionail quatli;iltiye intirmnaition and real 

world constraints these models should prove ecxtremely helpful 

to Ihe cliniciains and health atdlminiNitratir'n lced wJith the 

task'ol' progri'am evaltition. 

REFERENCES 

1. Weikel. K. Evatliuation of naiional hellh progranis. Am. J. Public 

Health 61:1801. 1971. 

2. Levey. S. and Loonmh;a. N. P. Healih Care Admninisilalion. p. 

421. I.ippincoll Co.. 1973. 

3. Scriven. M. Goal-iree evaluation. Evaui;atlio 1:b62. 1973. 

4. Halpern. J, and Binner. P. R. A model for a¡n ouipUt valuc anatlysis 

of inmenlul health programs. Administra;tion in Mental Health 

1:40-51. 1972. 

5. I:yniain. R. K.. T;arj;iil. G.. and McGlmnige. I). The Mitrktov 

Ch.i¡in as ;, Ilctilhod kt, eví.Islitiíg s.hotols ¡for ihe mlenal;lily re. 

atl. edAm. A . Menial I)cliciicny 72:435-144. 1967. 

6. LEyrií;ma. 1. iK.. Iaiij;an. C;.. ialnd CaIsady l. NI. N;ailurail hIisitary fl 

aícquisition of basic skills bh hospila;lize reitrided palienls. Am. 

J. Mental Deficiency 75: 120-129. 1970. 

7. Meredith. J. A M:ikwia .il5na'!is i ot ai ge;.Jitiic %ya;rd. M;anagenient 

Science 19:(1( 4-612. 1973. 

8. Meredilh. J. Prograni evalualtioun in a htopit;ll for ¡ihe ment;all relarded. 

Am. J. Menitl Deficiencv 781:471-48I1. 1974. 

9. Kemeny. J. C. and Snell. J. 1.. Finile Miarkov Chaiins. D. Van 

Noslitrnd. 1960.

MEDICAL CARE 

,arch 1981, Vol. XIX, No. 3 

- 121 - 

Assessing the Performance of Medical Care Systems: 

A Method and Its Application 

PAUL A. NUTING, M.D.,* GREGORY 1. SHORR, M.D.,t 

AND BARTON R. BURKHALTER, PH.D.S 

As health care becomes more differentiated, fewer people receive the majority 

of their care from a single source. Yet, most methods for assessing health care 

focus on the care provided by a single facility or group of practitioners. A 

method is described which tracks individuals through the diffuse medical care 

"system" and examines the process of care received for complete episodes of 

care. Through the use of tracer conditions the individual's pathwAay through the 

system is followed and the contribution of the various system components (e.g., 

facilities and providers) is assessed for various funetions of care (e.g.. sereening, 

diagnosis, treatmnent), thus pinpointing deficiencies in the process of care. The 

method is designed to sample systematically from the entire provider and consumer 

system. Use of this methodology in a variety of settings, including 

American Indian communities, has proved to be feasible and has uncovered 

deficiencies in the delivery of health services which might have been overlooked 

by other approaches. This article describes the assessment method and 

presents selected results which demrnstrate the assessment outputs. 

AS TECHNOLOGY and specialization increase, 

consumers of health services are 

faced with a bewildering array of different 

individual and institutional provide - 

with ever fewer receiving their care frorm a 

single source. Solutions to the problem are 

' Associate Director for Research, Office of Research 

and Development, Indian Health Service, 

Deparbntment of Health and Human Ss. ices. 

f Director, San Xavier Ambulatory Care Project, Office 

of Research and Development, Indian Health 

Service. 

t Senior Scientist,' Community Systems Foundation 

and Professor (Adjunct), Departnent of Family 

and Community Medicine, University of Arizona. 

All opinions expressed herein are the authors' and 

do not necessarily represent the policy of either the 

Indian Health Service or the Department of Health 

and t!uman Services. 

Address for reprints: Paul A. Nutting, M.D., Associate 

Director for Research, Office of Rescarclh and 

Development, Indian Health Service, P.O. Box 

11340, Tuicson. AZ 85734. 

0025-70791811030010281/$01.30 O J. B. Lippincott Co. 

offered from at least tv'o philosophical 

viewpoints. Some argue that tect, logy 

and specialization should be decrit ed 

and the goals of the health care system 

- 'ised to give more control to consumers 

an . -- attention to the environment and 

style: c 'iving,'-5 while others find 

evidence i the high degree of specialization 

shou!-' be maintained and better 

management auc ed to it. 6 

Methods of qua;'' -- 'essment and assurance 

may eventually contribute to better 

management of health care rystems. 

However, quality assessment and assurance 

have yet to demonstrate their worth. 

Most assessments ofthe quality of care address 

only narrow segments of the complex 

array of services. Some focus only on the 

care provided by single facilities or groups 

of providers. Others focus only on care 

proviaed to those patients who have

NUTTING ET AL. 

sought care for specific health problems. 

Most emphasize only diagnostic and 

treatment functions. 

More comprehensive methods are 

needed. This article presents an approach 

to quality assessment 'which tralcks members 

of the community into and through 

episodes of care provided by various parts 

of the medical care system. 

Requirements for Quality Assessment 

in the Indian Health Service 

The Indian Health Service (Department 

of Health and Human Services) has 

evolved a complex system of health services 

delivery. Charged by Congress to assure 

comprehensive health services to 

more than 600,000 American Indians and 

Alaskan Natives, the Indian Health Service 

(IHS) has developed and is operating 

more than 85 local comprehensive health 

care systems, called Service Units. The 

typical Service Unit serves a dispersed 

population, often scattered over severaithousand 

square miles. It usually consists 

of a 30-50 bed hospital and outpatient department 

staffed by a variety of health professionals 

and administrative staff. The 

main Service Unit facility serves as a referral 

center and administrative base for 

one or more full or part-time field clinics, 

public health nurses, environmental engineers, 

health educators, and a variety of 

problem-specific health programs (nutrition, 

mental health, maternal and child 

health, alcoholism, etc.) operated by the 

tribal government. Each Service Unit can 

also refer patients to secondary and tertiary 

care centers, either operated by the Indian 

Health Service or through contract within 

the private sector. 

As Service Units have grown more complex 

and Indian communities have become 

more mobile and more active in health delivery 

programs, the methods traditionally 

used by IHS to assess and assure the 

quality of care have become i. adequate. 

Therefore, the Indian Health Service has 

- 122 - MEDICAL CARE 

undertaken a long-term research and development 

effort in the assessment and 

assurance of quality. 

In an earlier study in one Service Unit, 

patients were tracked through episodes of 

ambulatory care for several prevalent 

health problems. 7 The study demonstrated 

that 1) explicit criteria for minimal care 

could be defined by the Service Unit 

physicians; 2) reliable data could be collected 

retrospectively from medical records 

by nonphysicians; 3) failures in the 

process of care could be identified; 4) the 

failures tended to occur at the same places 

in the process of care for different health 

problems; and 5) this information caused 

the Service Unit to take action aimed at 

correcting the problems. 

As a result of this experience, eight requirements 

were defined which, it was 

felt, would make the assessment method 

applicable throughout Indian Health 

Service. 

First, the assessment must examine the 

performnance of the total health system.§ 

All facilities, organizational subunits, and 

programs providing health services within 

the community should be included 

whether or not the individual components 

consider themselves as a part of a "system." 

The contribution of physician extenders, 

pharmacists, public health nurses and 

community health personnel as well as 

phSysicians should be included. The combined 

effect of consumer and provider on 

system performance should be measured. 

Second, the assessment must examine 

the care received by all members of the 

community, both patients and nonpa- 

4 The total health system contains the consumer 

subsystem and the provider subsystem. Much of our 

attention focuses on the medical care system, which 

we define as that part of the provider subsystem providing 

medical care. Components refer to the varinous 

parts of the two subsystems. For example, the assessment 

examines three types of components in the medical 

system. including facilities, organizational subunits 

and professional disciplines, although other 

components could also be identified.

%,I. XIX, No. 3 

iclnts. It should determine :i .Yrta±.n 

groups, especially those at high risk, ,tijv 

lower quality of care than others. 

Third, the assessment must examine, it ' 

,erformance of the'system across .. broad 

rangec of functions." In addition to diagníosis 

and treatment, system performance 

fr,. other functions such as prevention and 

u.rcening should be examined also. Altlia


- 124 - 

for the tracers as a group is similar to all 

care, and assumes that efforts that improve 

deficient care in the tracers will also improve 

care for other similar conditions, although 

these assumptions are not yet supported 

by experimental evidence. 

Consistent with Kessners use of tracers 

in sets, rather than singly, the assessment 

aggregates data by function across the tracers, 

in order to focus corrective action on 

function rather than disease.** The funetions 

we have used-prevention, screening, 

health status monitoring, diagnostic 

evaluation, treatment planning, follow-up 

and ongoing management-can be recombined 

to fit the needs of a particular 

analysis. Other functions, such as patient 

education or rehabilitation, can be defined 

if desired. 

Most functions can be separated into 

three sequential events, namely, contact 

between a consumer and a provider, recognition 

of the need for service once contact 

is made and provision of service after 

contact and recognition. Th ' particular' 

classification of functions and sequential 

events was chosen because it is generally 

familiar, because the adaptive processes 

required to correct deficiencies would appear 

to differ by funetion and event, and 

because the Indian Health Service may be 

in a position to influence these categories 

'* The data for a particular function are aggregated 

by 1) summing the number of consumers found to be 

in need of care in the study cohort of each one of the 

tracers; 2) summing the number found to be receiving 

adequate care in all tracer cohorts; and 3) dividing the 

second sum by the first to obtain the fraction of consumers 

for all tracers who are receiving adequate care 

for that function. This procedure weights the tracers in 

proportion to the size of their cohorts. Usuallyfthe size 

of the cohorts have been in proportion to the prevalence 

of the tracers, that is, the number of consumers 

in need ofcare for each function of atracer, although in 

some cases cohort sizes for all tracers have been equal. 

The aggregating procedure assumes that when the 

racers are weighted in proportion to cohort size, the 

care received for the tracers is representative of the 

care received for all health conditions. However, our 

experience indicates that this assumption is not always 

justified and that care should be taken in interpreting 

aggregated results. 

,1 3 

MEDICAL CARE 

of activities. This classification may be altered 

as more experience is gained with 

pattems of homogeneity across tracers and 

with their usefulness in improving care. 

When carrying out a particular assessment, 

we define an assessment spacet 

that is delineated by the functions, tracers, 

systemn components and populations to be 

studied. By defining the assessment space 

and measures of performance first, the assessment 

focuses only on issues of interest. 

Table 1 shows the functions and tracers 

that we used in early applications of the 

assessment method. Every function is 

examined using at least two tracers, by aggregating 

the data across the tracers for 

each function. This particular assessment 

space was constructed to examine common 

issues in ambulatory care, and did not address 

care for mental health problems, 

medical or surgical problems requiring 

specialized care, or rehabilitation. Since 

many of the functions are further broken 

down into the three sequential events of 

contact, recognition and provision of required 

service, the assessment space permits 

analysis of patient utilization, system 

outreach and problem recognition. However, 

this assessment space does not allow 

for analysis of inappropriate utilization of 

inpatient or outpatient services, surgical 

procedures or inappropriate drug therapy. 

For each tracer, minimal criteria for the 

process of care are established for each 

function which the tracer examines. In no 

case was a tracer used to examine a function 

for which "valid" criteria were unavailable 

or for which the criteria originally 

proposed were questioned by the local 

providers of care. Only criteria considered 

essential for basic health care are included. 

Criteria that may apply to a relatively small 

percentage of instances are useful in de- 

tt The assessment space for a particular assessment 

is defined by specifying 1) the functions tobe studied; 

2) a set of tracers for each function; 3) the components 

of the health system for each finmction; and 4)a population 

of consumers (from which an appropriate sample 

is selected) for each function of each tracer.

n'ol. XIX, No. 3 

- 125 - 

tailed examination of specific issues in 

health care, but are of less value in attempts 

to examine generic system performance. 

Since the assessment involves abstracting 

data from the medical record, criteria that 

are likely to be documented in the record 

are more often incorporated into the assessment 

design. In general, prescription 

data, measurements, lab results and diagnoses 

are reliably documented, while historical 

data, physical findings (especially 

ne gative findings)and education treatment 

plans are not. 

The assessment method produces three 

different types of indicators of health system 

performance: population -based indicators, 

encounter-based indicators, and 

healtlh-status indicators. Poptulation-based 

indicators$ I are computed from a sample 

of the coninmunity or a patient subset of the 

community and express the percentage of 

individuals in need of a specific health service 

who receive that service within a 

specified period of time. They track 

specific cohorts of the consumer popul. 

tion through the system of health care and 

examine the adequacy of the process of 

care and how it is distributed in the population. 

Encounter-based indicators are 

computed from consumer contacts with a 

particular component of the provider subsystem 

and express the percentage of 

consumcr encounters in which a specific 

need for service is satisfied. These indit 

I For many issues in health care it is not practical 

to employ a true population-based indicator, however 

desirable this may be. For example, an instructive 

Indicator might describe the percentage cl eople in 

the community with a urinary tract infection who received 

an appropriate aritibiotic withirnan appropriate 

time frame. However, short of a special survey to 

identify all people in the community with a urinary 

tract infection, this subset of the population i[ -arlv 

impossible to define. Consequently, the i..... 

would more likely express the.percentage of people in 

the community secreened positive for a urinary tract 

infection who received an appropriate antibiotic 

within three weeks. In t.is case the indicator more 

accurately might be called patient-based rather than 

population-based, although for simplicity we use 

population.based to refer to all indicators with 

number of individualb in the denominator. 

E 

c 

E 

 

* e 

c. 

.c 

·= c; 

:E - 

u fe 

t c 

o.c 

;> 

rD 

_@g 

-J 

_ - x X 

cí 

C0% X K 

E! 

A. 

e 

4>0 

tr 

yz 

E4>4 

4>. 

Z -8 

o > 

I 

2, L. l 1 l.¡ 4 ' .. 1 ..% .%k c. N ... . ... . . 

x 

X X x 

E> 

x 

-f e eg 

cix 

C . E 

i mZ W4 Z5 

p % 5.4> = ~ ¡S ~ - U. C 

e e 

e 

e 

e 

e 

1F 

-e 

c 

E 

e 

t 

e y 

E 

E 

e: 

-9,.


- liE6 - 

cators focus on the components of the provider 

system such as discipline or facility 

in order that their contribution to total system 

performance can be appraised. 

Finally, health status indica tors express 

the percentage of patients f:r whom a 

change in health status has been 

docunmented. -le.dlth status itidiciltors 

should not be equated with rneasures of 

incidence or prevalance since the latter requires 

a random sampling of the population. 

Health staitus indicators, on the other 

hand, often reflect change in health status 

of selected patient groups, e.g., only those 

who were followed up. Since health status 

indicators do not distinguish changes in 

health status that are the result of health 

care from changes related to behavioral or 

environmental variables, they should not 

be considered measures of the Guteomes of 

medical care unless the appropriate causal 

relationships linking process measures to 

health status have been demonstrated. 

Population-based indicators can be constructed 

in a sequence in order to examie 

the continuity of the process of care,§§ 

which reflects the extent to 'which 

consumers pass successfully through sequential 

steps in a defined process of care. 

Figure 1 illustrates a sequence of 

§§ The continuity of the process of care is defined 

as the likelihood that consumers dill receive needed 

heaith senrvices, in a proper sequence and within an 

appropriate interval of time, and is expressed as a 

sequence of conditional probabilities based on empirical 

data. This definition derives from Shortell, i " who 

conceptualized continuity as the extent to which medical 

services are received as a coordinated and uninterrupted 

succession of events consistent with the 

medical needs of the patients. The term "continuity" 

has been used differently. Some authors prefer a focus 

on continuity of process such we have done. ' .U L' 

while many others focus on continuity of provider by 

measuring the extent to which care is received from a 

single source or by referral. i " - » Although it appears 

that greater continuity of provider results in greater 

continuity of process,." Indian Health Service has 

often found it difficult to achieve continuity of provider 

and therefore has focused its attention on continuityofprocess. 

Here. the term"continuity-of-process" 

is used todistinguish our usage from other usages 

of the term "continuity." 

MtIC.d. CARE 

population-based indicators for urinary 

tract infection from which the continuity-ofprocess 

can be obtained for the treatment 

and follow-up finctions. The functions are 

divided into contact, recognition and provision 

of service, and indicators at each of 

these sequential events show the flow of 

patients throlugh the treatment and 

follow-up functions. Patient flow from one 

event to the next can be expressed as the 

transition rate Pu, where P is the proportion 

of patients at event i moving to eventj 

during a specified period of time. 

Likewise, tle transition rate of patients 

through multiple successive events in the 

process of care can be expressed as the 

product of the intervening rates, and the 

dropout rate between any two events, i and 

j, is 1 - Pu.I'I By examining care in this way, 

the assessnme nt can identify deficiencies in 

health systems performance and distinguish 

between problems related to provider 

performance, those related to patient 

utilization of service and those related to 

the system itself. 

Selection of the study cohorts and their 

health records largely determines the extent 

to which the assessment examines the 

entire community of consumers rather than 

those receiving care. As nearly as is prac 

tical, the study cohorts are generatec 

randomly from the entire community of po 

tential consumers. This is especially impor 

"" Care must be taken in multiplying together P,,' 

obtained from different tracer conditions and differet 

cohorts, because patients receiving adequate car 

early in the process of care may not be independent t 

those receiving adequate care later on. Specificall; 

patients receiving adequate care at one stage may al: 

be more likely to receive adequate care at another. I 

fact, such nonindependence is illustrated in the da 

reported by Novick, I ' although it was not explicit 

pointed out by him. We have avoided this problem I 

calculating continuity-of-process scores only for fun 

tions or sequences of functions where we were able 

observe the passage of the cohort throughout the se 

quence. Thus, our continuity-of-process scores as 

observed, not calculated. When greater knowledge 

obtained about the interdependence of functions %s 

may be able to combine continuity-of-process score 

obtained from different cohorts.

1 

Vol. XIX. No. 3 

- 127 - 

tant for prevention, scree¡inxg and health 

status monitoring. As an example, a cohort 

of women may be selected for examination 

of prenatal care by generating a master list 

from delivery room logs, birth certificates, 

operating room logs and laboratory logs 

(searching for prenatal lab work ordered). 

W)hen the redundancies are removed from 

the master list, a standard sampling 

technique is used to select the study 

.,e-,. For r any tracers a single document 

!¡zly b'e lusctd t.. gt'ler.ltte' age- andaor 

sex-specific study cohort. - g., birth certificates 

alone may suffice for the generation 

of a cohort for examination of infant care. 

It also is important to sample from the 

most basic source docunient available. For 

example, in generating a cohort of patients 

with urinary tract infections, it is better to 

sample the líaboratory log for patients with 

a positive urine culture than to generate a 

sample of medical records for patients who 

were diagnosed with a urinary tract infection. 

The latter technique biases the sam- _ 

ple in favor of patients who have made 

contact and for whom the system has recognized 

the problem. 

Patients are eliminated from the study 

cohort when they do not contribute to the 

objectives of the study. For example, when 

examining uncomplicated urinary tract in- 

.fections, it might be preferable to eliminate 

patients with chronic urinary tract infections, 

chronic renal disease, urinary tract 

anomalies, etc. These characteristics become 

apparent in the record review and 

patients thus eliminated from the study 

cohort may be replaced from the master 

list. 

Selection of the health records to be 

examined largely defines the scope of the 

medical care system to be examined. Since 

medical care systems usually are not 

clearly delineated, judgment is required to 

define the system in a way that will produce 

useful assessment results. Most assessments 

of IHS Service Units' must 

examine medical records at the main 

PERFORMA.NCE ASSESSMESNT 

FIG. 1. A sequence of population-based indicators 

for urinary aert infections constructed to examine 

the continuity of the process of care. 

hospital-outpatient facility and at one or 

more field clinics, public health nursing 

records, medical records at one or more 

referral centers and sometimes health records 

from tribal health programs (e.g., nutrition, 

mental health or alcoholism). In 

applications of the method outside the Indian 

Health Service, a similar set of record 

types have been included in the assessment 

procedure. In an assessment of one 

rural private practice, records were 

examined from two private practices, the 

community hospital, the county health 

clinic and the county public health nursing 

program. 

Audit instruments are designed for each 

tracer to extract data from each record for 

each individual in the study cohort. The


audit instrument documents each contact 

between the consumerand any component 

of the provider system and captures the 

date, location of contact, pro. ider of scrvices 

and any of a predeterrined list of 

services required to compute the indicator 

results for that tracer. Whcn completed, the 

audit instrument contains a complete profile 

for that consumer of each contact with 

every component of the provider system 

and each of the relevant services provided. 

The results for each of the indicators and 

continuity-of-process scores for each function 

or sequence of functions can be computed 

from the completed audit. 

Application of the Assessment Method 

The assessment method has been 

applied in 20 Service Units of the Indian 

Health Service, three rural private practices 

and two closed-panel health maintenance 

organizations. Selected data is 

presented from these appli itions to 

demonstrate the major characteristics 

of the method and some of the results 

achieved.' 

The first step in an analysis is to examine 

the performance of each function. In one 

- 128 - MEDICAL CARE 

Service Unit of the Indian Health Service, 

both population-based and encounterbased 

indicators were employed to 

examline the follow-up fimetion, using as 

tracers iron-deficiency anemia, urinary 

tract infection, laceration of the scalp and 

extremities and prenatal care. Table 2 

presents the follow-up criteria for each 

tracer and the results aggregated across the 

tracers. The consumer-provider contact 

rate is population based and expresses the 

percentage of patients due for follow-up 

who made contact with the medical care 

system during the time interval appropriate 

for follow-up. The encounter-based 

follow-up rate expresses the percentage of 

patient encounters due for follow-up, in 

which the follow-up criteria were. met. 

Finally, the population-based follow-up 

rate expresses the percentage of patients 

due for follow-up who contacted the system 

and received the follow-up services within 

the appropriate time. The aggregate data 

indicate that 71 per cent of patients make 

contact with some component of the medical 

care system when due, resulting in only 

44 per cent of patients receiving follow-up 

services. These results suggest that the relative 

weakness in follow--up care is the sys- 

TABLE 2. Examination of the Follow-up Function With Aggregate and 

Tracer-Specific Data 

Iron Urinary 

Aggregate Deficiency Tract Prenatal 

Follow-Up Funetion Results Anemia Infection Lacerations Care 

Contact rate 

(population-based) 71(921129) 58% (15/26) 63% (19/30) 65% (15/23) 86% (43/50f 

Provision of 

service rate 

(encounter-based) 51% (57/112) 44% ( 7/16) 52% (12/23) 80% (12/15) 45% (26/58 

Provision of 

service rate 

(population-based) 44% (57/129) 27% ( 7/26) 40% (12/30) 52% (12/23) 52% (26/50 

Follow-up critena include: 

Anemia-All patients placed on therapy should have a hematocrit or hemoglobin or reticulocyte coun 

between three and six weeks of the initiation of therapy. 

Urinary Tract Infection-All patients placed on antibiotic therapy should have a urine culture or micrc 

scopic urinalysis within four weeks after therapy is completed. 

Lacerations-All patients with a laceration requiring sutures should have an examination and documente. 

statement of wound healing between 5 and 21 days after the sutures were applied. 

Prenatal Care-All women delivering should have their blood pressiure documented and an examination o 

the uterus between 2 and 10 weeks after delivery.

Vi. .XIX, No. 3 

- 129 - 

tom' s recognition of and response to patients 

presenting when due for follow-up 

services, rather than the utilization behIavior 

of the patient population. Table 2 

illustrUtes lhow data for several tracers can 

¡)e aggregated for a single function, and for 

sequential events within that function. It 

also il>lustrates that the results are not alwuays 

similar across tracers. For cxamplc, 

the fact that lacerations, with sutures acting 

as an apparent clinical sign calling for attention, 

produced higher recognition and a 

ver)' different pattern than the other tracers, 

suggests that it may be possible to devise 

more efficient ways of aggregating 

across tracers or to categorize health conditions 

in a way that allows for more representative 

sets of tracers to be selected. 

In order to examine the performance of 

Ciflerent system components by funetion, 

the assessment often employs a 

population-based indicator in conjunetion 

with an encounter-based indicator that is 

disaggregated by relevant system components 

as well as by function. Table 3 shows 

the data examining infant immunization in 

a rural private practice setting, two Service 

Units of the IHS and a large closed-panel 

PERFOR.\MANCE ASSESSMENT 

health maintenance organization. The 

population-based immunization rate expresses 

the percentage of infants who had 

received three DPT and two polio immunizations 

by 12 months of age. The 

encounter-based immunization rate expresses 

the percentage of visits by infants 

due for an imniunization in which the immunization 

was provided. The DPT immunization 

was considered to be due at 2 

months of age and to be repeated monthly 

until three doses had been given. If, at the 

time of a visit, the infant had a rectal temperature 

greater than 100.5 ° , then an immunization 

was not considered due on that 

visit. 

Among the infant population served by 

the private practict, only 32 per cent had 

received three DPT and tho OPV' immunizations 

by age 12 months and the 

encounter-based indicator revealed that 

immunizations were provided on only 22 

per cent of the visits for which they were 

due. This private practice had assumed 

-.. hat infants were receiving their immunizations 

from the nearby county health 

clinic. But when the encounter-based indicator 

was sorted by the physician's office 

TABLE 3. Data for Infant Immunization From a Rural Private Practice, 

Two IHS Service Units, and One Health Maintenance Organization (HMO) 

llustrating the Performance Patterns Resulting From the Assessment Method 

Private 

Practice IHS-A IHS-B HMO 

Immunization rate 

(population-based) '2% (26150) 86% (43/50, 56% (28/50) 58% (29!50) 

Immunization rate 

(encounter-based) 22% (63/285) 46% (179/387) 38% (119/316) 86% (127/147) 

Sorted by facility 

Medical officer 19% (211112) 

County clinic 24% (42/173) 

MCH clinic 85% (103/121) 

General clinic 34% (70/208) 

2 fild clinics 11% (6/53) 

Inpatient service 0% (015) 

Sorted by provider 

discipline 

Physician 34% (64/189) 

Physician extender 75% (36/48) 

Clinic nurse 50% (6/12) 

Public health nurse 87% (13/15) 

Pharmacist 0% (0/52)


- 130 - 

and the county clinic, it revealed that 

neither location was taking advantage of its 

opportunities to provide inmmunizations. 

In discussing the results, both sites agreed 

that it had assumed the other was responsible 

for immunization, and both agreed to 

begin immuniizing niore vigorously. An informal 

follow-up study by one of the private 

practitioners several months later 

indicated that the encounter-based immunization 

rate had increased three fold at 

both locations. 

One Service Unit (IHS-A) had a 

population-based immunization rate of 86 

per cent and an encounter-based rate of 46 

per cent. When the encounter-based rate 

was sorted by clinic, it was noted that the 

MCH Clinic was performing well at85 per 

cent, while the general clinic (34 per cent) 

and the two field clinics (11 per cent) were 

missing many opportunities to provide 

immunization. Since most of the missed 

opportunities occurred at the general 

clinic, the Service Unit instituted a stalding 

order for immunizations in the general 

clinic. 

In the second Service Unit (IHS-B), 56 

per cent of the infant population was immunized 

by 1 year of age. When the 

encounter-based rate of 38 per cent was 

sorted by provider discipline, it was noted 

that the physicians werte providing immunizations 

only 34 per cent of the time 

and were commonly referring infants to the 

physician extender for well baby care. 

Also, 52 of the 316 infant visits made when 

an immunization was due had been to the 

pharmacist, who had recently begun a 

program providing nonprescription medication 

directly from the clinic pharmacy. 

This result and the pattern apparent for 

other functions led to the development of a 

checklist of potential service needs for 

prevention and chronic disease surveillance 

for use by the pharmacist while dispensing 

over-the-counter medications. 

A contrasting pattern wa. seen in the 

immunization indicators for the health 

maintenance organization. Although the 

providers were immunizing infants on 86 

per cent of the visits when an immuniza- 

NIEDICAL CARE 

tion was due, only 58 per cent of the population 

was being immunized. This pattern 

suggested that patient contact was the 

limiting factor in achieving higher immunization 

rates in the infant community. 

This was later confirmed by a study of the 

utilization pattenl of the infants, which indicated 

that many of the infants contacted 

the provider system only when they were 

ill. 

The above experiences with immunization 

illustrate several benefits of the assessment 

method. It can focus attention on 

a particular site or discipline; it can point to 

areas where deficiencies are greatest; and 

feedback of the results may, in some cases, 

generate action. Several limitations are 

also illustrated. The method does not identify 

the causes of the deficiencies; it does 

not uncover and analyze the many possible 

remedial actions; nor does it evaluate remedial 

actions during and after implementation 

unless a separate study is 

undertaken. 

In several applications, treatment and 

follow-up were examined by obtaining the 

study cohorts from the laboratory logs of 

each laboratory at every study site. For 

iron-deficiency anemia the study cohort 

included individuals with laboratory records 

showing a hematocrit less than 33 vol 

% and a hemoglobin less than 11 gm %. For 

urinary tract infections the study cohort included 

individuals with a urine culture resulting 

in greater than 105 colonies per milliliter. 

In order to examine the care for 

routine and uncomplicated conditions 

only, patients whose charts showed a noninutritional 

cause for their anemia, or that 

had a chronic urinary tract infection, urinary 

tract anomaly or chronic pyelonephritis, 

were eliminated from the studY 

cohort. 

Table 4 illustrates varying patterns of 

performance in treatment and follow-up in 

three IHS service units. The numbers in 

the table are population-based indicators 

that express the probability based on emipirical 

data that a patient at a given point in 

the process of care will pass successfully to 

the next. Likewise, the probability that a;

- 131 - 

A \ ,,1. XIX, No. 3 PEHFOHRMANCE ASSESS. iE.' l 

con"sumer %will pass successfully through 

multiple successive elemnents of care can 

be obtained directly from the number of 

individuals in the study cohort who suc- 

.essfully completed the sequence, but also 

can be calculated as the product of the intervening 

probabilities."a Thus the 

"continuity-of-process index" of Table 4 

expresses the probability that a paitient 

screened positive (in this case for either 

anemia or a urinary tract infection) will 

achieve contact with the system, have the 

problem recognized, receive tre' Rnent, 

make contact for follow-up, have the need 

for followv-up recognized, and receíive the 

followv-up servnice. As shown in Table 4, 

this probability ranges ftomr 0.55 in Service 

Unit C to 0.09 in Service Unit E. The 

aggregated results suggest that the three 

Sc'rvice Units have three distinct pattems 

of care. Service Unit C, with a continuityof-process 

index of 0.55 for treatmnent and 

follow-up, appears to have no particular 

step in the process of care which stands out 

as a relative deficiency. Service Unit D,vwith 

a continuity-of-process index of 0.36, 

is similar to Service Unit C except for recognition 

(indicators 2 and 5), which appears 

as the relative impediment. Service 

Unit E, with a continuity-of-process index 

of only 0.09, appears to have substantial 

deficiencies in both contact (indicators 1 

and 4) and recognition (indicators 2 and 5). 

The next step in the analysis would be to 

determine if these distinctions continue 

1when the results are disaggregated and 

examined by1 

tracer. 

" Care must be taken in multiplying together PU's 

olt.iii< .d fromn different tracer conditions and different 

cohorts, because patients recciving adequatet cwrt 

early in the process of care may not be independent of 

those receiving adequate care later on. Specifically, 

patients receiving adequate care at one stage may also 

he more likely to receive adequate care at another. In 

fact, such nonindependence is illustrated in the data 

reported by Novick,'° although it was not explicily 

pointed out by him. We have avoided this problem by 

calculating continuity-of-process scores only for functicns 

or sequences offiunctions where we were able to 

ob.serve the passage of the cohort throughout the sequence. 

Thus, our continuity-of-process scores are 

observed, not calculated. When greater knowledge is 

obttaiued about the interdependence of functions we 

may be able to combine continuity-of-process scores 

ohbtained from different cohorts. 

The assessment method can also 

examine the distribution of care among 

various types of consumers, with one ofthe 

most useful distinctions being consumers 

at different risk to particular health problems. 

In a study of prenatal care at one IHS 

service unit, 22 of 50 pregnant women in 

the study cohort were classified as high risk 

becaus(e they were under the age of 18 

years, over the age of 35 years, primigravida, 

with parity equal to or greater than 

5, or with a history of miscarriage or spontaneous 

abortion, while the other 28 

women were classified as average risk. 

Table 5 shows results for three indicators 

(of the 25 employed in the study) which 

constitute a simple se(luence to examine 

the continuity of process of gonorrhea 

screening. Respectively, the indicators 

examine the proportion of women achiesing 

contact itl the systemni by the 20th 

gestational week, the proportion of those 

with pregñiancy recognized by the 20th 

week, and the proportion of those having a 

= cervical culture by the 20th gestational 

week. 

As is apparent from the indicator results 

and continuity-of-process index of Table 5, 

system performance favors the average risk 

group at each step of the process. The reasons 

for the disparity are suggested by the 

encounter-based indicators for pregnancy 

recognition and gonorrhea screening disaggregated 

by site of contact as shown in 

Table 6. Unlike population-based indicators, 

encounter-based indicators are 

computed in units of patient encounters 

with the system. Thus, only 43 per cent of 

encounters by high-risk patients compared 

with 67 per cent of encounters by averagerisk 

patients due for recognition of pregnancy 

resulted in pregnancy recognition. 

Similarly, only 33 per cent of encounters 

by high-risk patients compared with 71 per 

cent of encounters by average-risk patients 

with pregnancy recognized and due for 

gonorrhea screening received a cervical 

culture. When disaggregated by site of contact, 

the results suggest that the prenatal 

clinic performs well in both the recognition 

and screening function, the public


TABLE 4. Results From Three Service Units of the Indicator Sequence 

Designed to Examine the Continuity of Process for Treatment and Follow-Up 

1. Contact for The proportiotn of patients with 

evaluation a positive screening result who 

made contact with the health 

cate system within an 

appropriate time frame 

(within 3 weeks for anemia and 

within 2 weeks for urinary 

tract infection). 

2. Recognition of The proportion of patients making 

problem contact for whom there was any 

evidence that the problem 

was recognized. 

3. Provision of The proportion of patients with 

treatment the problem recognized who 

received appropriate treatment 

within 2 weeks afer 

recognition. 

4. Contact for The proportion of patients 

follow-up achieving contact with the health 

care system within a 

time frame appropriate 

for follow-up (within 3-6 weeks 

afier treatment initiated for 

anemia and %w ! in 4 weeks aftcr 

~t..i;.~.q.~; :."i.s;Ne or-omp}eretefor 

unnary dtra'cC iiiLi.~c2i/,. 

5. Recognition of The proportion of patients making 

need for contact for whom there is any * 

follow-up evidence that the problem 

andlor the need for follow-up 

was recognized. 

6. Provision of The proportion of patients with 

follow-up recognition of problem 

and/or the need for follow-up, 

who received an appropriate 

follow.up (hemotocrit, hemoglobin, 

or reticulocyte count for anemia; 

urine culture or microscopic 

urinalysis for urinary 

tract infection). 

Continuity-of- The proportion of patients with 

process index a positive screening result who 

completed the entire health 

care sequence. (The continuityof-process 

index may also be 

computed as the product of the 

proportions for each of the 

indicators 1 through 6). 

Service Service Service 

Unit C Unit D Unit E 

0.92 (62/67) 0.93 (931100) 0.73 (73/100) 

0.90 (56;62) 0.76(71,93) 0.62 (45173) 

0.95 (53/56) 0.,99 (70/7, ) 0,93 (42/45) 

0.83 (441/53) 0.89 (62/70) 0.52 (22/42) 

0.69 (39'44) 0.61 ,i,. .6', 

'. 

0.95 (37/39) 0.95(36/38) 1.0 (9;9) 

0.55 (37/67) 0.36 (36/100) 0.09 (9/100) 

The results shown are an aggregate ofdata derived from nutritional anemia and urinarytractinfections.The study 

cohorts for Service Units D and E are random samples of 50 patients foreach tracer. The study cohorts from Senive 

Unit C consist of a total sample of 36 patients with anemia and 31 patients with urinarv tract infection. 

health nurses performn well in recognition, 

but the hospital outpatient department and 

the field clinics contribute substantially 

less to the perfornnance. 

- 132 - .MEDICAL CARE 

The data in Table 6 show that the 

average-risk women who were due for ca;re 

made a higher proportion of their encouln 

ters with the prenatal clinic than did the

V Xl. IX, No. 3 

- 133 - 

PERFORMANCE ASSESS.MENT 

TABLE 5. Results From> a Stervirt, Unit Showing Population-Based Indicators 

for Prenatal Care 

Total Cohort High Risk Average Risk 

(n - 50) (N' - 22) (N - 28) 

Contact The proportion of pregnant women who 0.64 (32150) 0.54 (12122) 0.71 (20!2/) 

made contact with the health care systemr 

by the 20th week of gestation. 

1H¢-c ignitionn Thl proportion of pregnant uomen making 0.69 (22/32) 0.50 (6112) 0.80 (16;20) 

contact who laJad tIeir pregnaiUncy rtcosgnized 

by the 20th week of gestation. 

Screening The proportion of pregnant women with 0.82 (18/22) 0.50 (3/6) 0.94 (15'16) 

pregnancy recognized uho had a cervical 

culture by the 20th week of gestation. 

Continuity-of- The proportion of pregnant women who 0.36 (18,'50) 0.14 (3/22) 0.54 (15:28) 

process index had a cervical culture 

by the O20th week of gestation. 

A patient asi considered high risk if she was less than 18 years ofage, over 35 years of age, primigra'ida, had 

parit. equal to or greater than 5, or liad a histuor) of miscarriage or spontaneous abortion. 

high-risk woomen. Specifically, for 

average-risk women 42 per cent (10 of 24) 

of all encounters when due for recognition 

were with the prenatal clinic and 67 per 

cent (14 of 21) of all encounters when due 

for screening were with the prenatal clinic, 

while for high-risk women only 7 per cent 

(1 of 14) and 33 per cent (3 of 9) of all 

encounters when due for recognition and 

screening, respectively, were with the 

prenatal clinic. Consequently, the superior 

perforrnance of the prenatal clinic favors 

the average-risk group by virtue of the different 

utilization pattems of the two risk 

groups. A similar pattern at another Service 

Unit has been studied in detail and is 

reported elsewhere. 2 ' 

The above examples illustrate how the 

assessment method operates, and how it 

can point to areas where there are de- 

TABLE 6. Results for Encounter-Based Indicators of Pregnancy Recognition 

and Screening for Gonorrhea, Disaggregated by System Component 

High Risk Average Risk 

Cohort Total Patients Patients 

Recognition rate: The proportion of 

visits by) the 20th w eek of gestation 

by patients not prev iously 

recognized as pregnant, in which 

recognition of pregnancy oc-urred. 

Hospital outpatient department 0.35 (6/17) 0.25 (218) 0.44 (4/9) 

Prenatal clinic 1.00(11/11) 1.00 (1/11) 1.00(10/10) 

Field clinic 0.00 (015) 0.00 (012) 0.00 (013) 

Home visit by public health nurse 1.00 (5/5) 1.00 (3/3) 1.00 (212) 

Total visits 0.58 (22/38) 0.43 (6/14) 0.67 (16/24) 

Screening rate: The proportion of 

visits by the 20th week of gestation 

by patients with pregnancy 

recognized and due for a cervical 

culture, in which a cervical culture 

was taken. 

HIospital outpatient department 0.11 (1/9) 0.00 (013) 0.17 (1/6) 

Prenatal clinic 1.00 (17/17) 1.00 (3/3) 1.00 (14114) 

Field clinic 0.00 (0/11) 0.00(1) - (0) 

Home visits by public health nurse 0.00 (013) 0.00 (012) 0.00 (0/1) 

Total visits 0.60 (18/30) 0.33 (3/9) 0.71 (15)21)


- 134 - 

ficiencies. The anecdotal data presented 

on corrective actions suggests that in some 

cases the method may enhance the assurance 

tfunction, although we do not yet Ihave 

systematic evidence. 

Discussion 

Requirements for a method to assess the 

quality of care in Indian Health Service 

service units were defined. A method was 

developed to try to fulfill these requirements 

and was applied in 25 health systems 

to test whether the requirements had 

been met. Some of the requirements were 

fulfilled while others were not. 

The first requirement calls for examining 

the performance of the entire health 

system and is based on the assumption that 

the performance of the total health system 

cannot be inferred from the performance of 

individual components, which is especially 

important in Indian Health Service, 

where consumers are often scattered geographically 

and highly mobile. Rather thia 

being limited to specific facilities or provider 

groups, the assessment begins with a 

sample of the community with specific 

needs for health service and tracks them 

through their encounters with all system 

components, including components that 

do not necessarily view themselves as part 

of the system, thus fulfilling the 

requirement. 

The second requirement calls for the 

examination of care for all members of the 

community, not just patients, and for differentiation 

among subgroups. By selecting 

study cohorts from the entire known 

community ofconsumers, as demonstrated 

in the immunization and prenatal care 

tracers, and using original source documents, 

as demonstrated in the anemia and 

urinary tract infection tracers, the second 

requirement is largely met. However, 

more work must be done to determine the 

extent to which the entire com-nunity of 

consumers has been identified in [HS service 

units. 

MEDICAL CARE 

The distribution of care by risk group 

was successfully examined, with the finding 

that two materna! and child health 

clinies provide d better care to avsrage-rinsk 

mothers than to high-risk mothers because 

ofdifferent utilization patterns. Although it 

appears that the method could also 

examine the distribution of care across 

other subgroups of the population, this has 

not yet been demonstrated. *o 

The assessment focuses on the performance 

of a broad range of functions, including 

especially those most important for 

ambulatory care, rather than on particular 

health conditions, as called for in the third 

requirement. However, the usefulness of 

the resulting information depends on the 

assumption that the care received for the 

tracers is representative of all care. The 

original application of this method found 

that the functions had similar relative performances 

for all the tracers used,' which 

was supportive of this assumption. More 

recent applications of the method also 

found similar pattems for most tracers, but 

with some notable exceptions, for example 

the high recognition rate for suture 

follow-up care compared with much lower 

recognition rates in follow-up care for other 

tracers. Such variation across tracers indicates 

that a'thorough investigation is 

needed on whether and how to aggregate 

across tracers. For example, classes of tracers 

might be found that have similar patterns 

of care within classes but different 

patterns across classes. Although current 

aggregates are possibly useful starting 

points, they are not based on sound 

theoretical or empirical foundations and 

therefore should be accompained by the 

tracer-specific results. 

*o In concept, any attribute of medical care can be 

distríbuted across different subgroups of the p opulation. 

The distribution of accessibility" and 

utilization" across various types of population sibgroups 

has been widely studied, but the process of 

care as we have defined it has not. Shortell hbis 

attempted a more precise definition of distrhibution of 

cwae by applying concepts of ecoinolics.

- 135 - 

The fourth requirement is largely met in 

that the assessment method traces the care 

received throughout an episode and relates 

it temporally to health status. However, 

causal linkages between process and 

health status are not established by the assessment 

and so outcome of care is not 

established. The number of health conditions 

available to use as tracers is expanded 

greatly by examliiiting only segments of the 

process of care rather than the entire process 

for every tracer, although this limits the 

conclusions that can be drawn about the 

continuity of the entire procer. jf care because 

the functions are not independent of 

one another. More work is needed to understand 

clearly how changes in one function 

influence the performance of others. 

The fifth requirement established 

criteria for the selection of tracers. These 

criteria have been met with the exception 

that the representativeness ofithe tracers as 

a group has not been demonstrated, as 

noted above. 

The sixth requirement establishes 

criteria for defining standards of care for'" 

the tracers. The standards established 

were both minimal and agreed upon by the 

local practitioners. The work of Brook 24 

suggests that because the assessment 

method uses explicit standards for the 

process of care rather than implicit judgments 

or indicators of outcome, the absolute 

performances will be relatively low. 

However, this is not important so long as 

the performance indicators are consistent 

over time and across comnponents, functions 

and consumer groups and their statistical 

attributes are not relatively worse than 

methods using implicit judgments or outcome 

indicators. More investigation of 

these assumptions is needed. 

Although the assessment method identifies 

areas of major deficiencies, it has not 

demonstrated that it systematically identifies 

remedial deficiencies, as called for in 

the seventh requirement. In order to determine 

if the deficiencies uncovered are 

able to be corrected, the assessment must 

be imbedded in an effective assurance 

process. Alternative possible remedies 

must be identified and evaluated and then 

implemented successfully. This is the 

point where other assessment methods 

have encountered the most difficulty, and 

it is widely recognized that assessments 

have not reliably led to improvements in 

quality.' 3 . 16.S4.25 

The eighth requirement deals with cost. 

Our experience indicates that the cust of 

employing this method in IHS service 

units is modest. Nonprofessionals were 

used successfully to collect and tabulate 

data from the nonstandardized health records 

currently found in Indian Health Service. 

In 10 service units it took 13 data 

collectors and two supervisors a total of 291 

working days to identiiy the cohorts and 

abstract data from the health records, plus 

another 69 working days to tabulate the 

data. Thus an average of 36 nonprofessional 

days was required to collect and 

tabulate the data per service unit. Each 

worker also spent eight days in training. 

On the average, data were abstracted from 

3.4 different health records for each consumer, 

with study cohorts of approximately 

50 consumers for each of nine tracers. This 

produces an average time for data collection 

and tabulation of 38.4 minutes per 

episode of care studied and 11.3 minutes 

per record abstracted. 

Others have also reported on the cost of 

data collection by nonprofessionals for 

episodes of care using explicit standards. 

Some agree with our experience that the 

cost is modest7 ' ' 0 while others have found 

it expensive."'16 Recent work by Albrecht 

and Kessner :t suggests that while the cost 

of abstracting may be modest, the' fixed 

1 Although not reflected in the above cost estimate, 

IHS is implementing a computerized health 

information system which integrates data from all the 

health records maintained for a given consumer. Once 

in operation, this system should significantly increase 

the efficiency of the assessment through its use of 

standardized encounter forms, a well-defined list of 

consumers from which to draw cohorts, and identification 

of the location of all relevent records.

Vol. XIX, No. 3 

costs for operating such an assessment 

method may be high. 

Thisi ,s's ti'.tit milethod is t t oflbw .s 

a replacement tor other methods currently 

available that provide an in-depth examination 

of specific system components. 

Rather, it is offered as a broad-brush assessment 

which provides a mechanism for 

detecting problems in system performance 

which may then be examined in more detail 

using the other methods. Use of this 

approach in a variety of health care settings 

has proved to be feasible and has uncovered 

areas ofdeficiency in total system performance. 

WVork on improving and applying 

the method continues by Indian Health 

Service, with particular effort being given 

to testing the reliability of its measurenents, 

examining how to select more 

representative sets of tracers, and in imbedding 

it in an assurance process that improves 

the quality of care. 

Acknowledgment 

The authors wish to expresstheir gratitude to Robert 

H. Brook, !M.D., for his many helpful comments and 

encouragement, and to Louise Burbank. Claudina 

Noriega and Sandra J. Woolbright fior their diligent 

efforts in the preparation of the manuscript. 

References 

- 136 - 

1. Carison R. The end ofmedicine. New York: John 

Wiley and Sons, 1975. 

2. Illich 1. Mledical nemesis: the expropriation of 

health. London: Calder and Bayars,. 1975. 

3. Navarro V. The underdevelopment of health of 

working America: causes, consequences. and possible 

solutions. Am J Public Health 1976;66:538. 

4. Fuchs V. Who shall live? New York: Basic Books, 

1974. 

5. Moss CE. Illness, immunity. and social interaction. 

New York: John Wiley and Sons, 1973. 

6. Lawrence PR, Lorsch JW. Differentiation and 

integration in complex organizations. Administrative 

Science Quarterly 1967;12:1. 

7. Shorr GI, Nutting PA. A population-based assessment 

of the continuity of ambulatory care. Med 

Care 1977;15:455. 

8. Kessner D l. Kalk CE, Singer J. Assessing 

health quality-the case for tracers, N Engl J .Med 

1973;288(i): 189. 

9. K!essner DMt. Kalk CE. Contrats in health status. 

Vol 2: a strategy fir evaiiuting health services. 

PERFORH.ANCE ASSESS.SMENT 

Washington, D.C.: National Academy of Sciences, 

1973. 

10. N'ovick LF. Dickinson K.Asnes R, May Lan SP. 

Lowsenlsteinl R. .Assessinc'it .faml atior s-are: application 

of the tracer methodology. !Med Care 1976; 14:1. 

11. Nobrega FT, 'lorrow GW Jr, Smoldt RKC Offord 

KP. Quality assessment in hypertension: analysis of 

process and outcome methods. N Engl J Med 

1977;296: 145. 

12. Gonnella JS, Daniel ZL, McCord JJ. The staging 

concept-an approach to the assessment of outcome 

of ambulatory care. Mled Care 1976;14:13. 

13. Hirschhorn NH. Lamstein 1, Klein SF, 

McCormnack J, Warner TN. Quality by objectives: a 

model of quality of care assessment and assurance for 

ambulatory health centers. Journal of Ambulatory 

Care Management 1978;L:55. 

14. Howell JR. Osterweis M.. Huntley RR. Curing 

and caring-a proposed method fur selifassessment in 

primary care organizations. J Communíty Health 

1976; 1:256. 

15. Shortell SM. Continuity of medical care: conceptualization 

and measurement. Med Care 

1976; 14:377. 

16. Donabedian A. Needed research in the assessment 

and monitoring of the quality of medical care. 

Washington, D.C.: Department of Health Education 

and Welfare, 1978 (DHEW publication no. PHS-78- 

3219.) 

17. Beck MH, Drachman RH. Kirscht JP. A field 

experiment to evaluate various outcomes of continuity 

of physician care. Am J Public Health 1974;64:1062. 

18. Starfield BH, Simborg DW, Hom SD, Yourtee 

SA. Continuity and coordination in primary care: their 

achievement and utility. Med Care 1976;14:625. 

19. Breslau N, Rceb RC. Continuity of care in a 

university-based practice. J Med Educ 1975;50:965. 

20. Steinwachs DM. Measuring provider continuity 

in ambulatory care: an assessment of alternative 

approaches. Mled Care 1979;17:551. 

21. Nutting PA, Barrick JE, Logue SC. The impact 

of a maternal and child health care program on the 

.quality of prenatal care: an analysis by risk group. J 

Community Health 1979;4:267. 

22,. Aday L, Anderson R. Development ofindices of 

access to medical care. Ann Arbor: Health Administration 

Press. 1975. 

23. Anderson R. Health service use: national trends 

.and variations. Washington, D.C.: DCpartment of 

Health Education and Welfare, 1972. 

24. Brook RH. Quality of care assessment: a comparison 

of five methods of peer review. Washington, 

D.C.: National Center for Health Services Research 

and Development, 1973. (Publication no. HRA 74- 

3100.) 

25. Williamson JW. Assessing and improving 

health care outcomes. Cambridge, Mass.: Ballinger 

Publishing Company, 1978. 

26. Morehead MA. Ambulatory care re íiew: a neglected 

priority, Bull NY Acad Med 1976,52:60. 

27. Albrecht JA, Kessner DM. Assessing ambulatory 

care: a coml)arittie anal>-sis of tiiree 

methodologies. (Unpublihed iéemno.)

- 137 - 

Systems and Procedures of Patients 

and Intormation Flow 

Arnold Reisman, Fh.D., Joao Mello da Silva, Ph.D., and Jloseph 

B. Mantell, Ph.D. 

THE TASK of properly scheduling an annual load of several 

hundred thousand patient visits to more than 100 doctors in 28 

different departments, and supporting these visits with timely, 

accurate, and complete flows of information is very cominplex. 

This task is further complicated by the fact that patients generally 

require timely sequencing of laboratory tests, X-rays, and 

consultative appoitments. From 1968 to 1973 in the clinic of one 

large health center, thc number of patient visits per year increased 

approximately 26% (Figure I) and the institution's physical 

plant was greatly expanded, tying up much of the administrative 

talent and resources. More importantly, however, the staff 

capability had become more diversified. It had disproportionately 

increased in number to serve the rapidlyexpanding 

inpatient population and to perform additional research 

and teaching activities. The systems and procedures for 

scheduling and processing the patients through the outpatient 

clinic degraded to the consternation of patients, doctors, and 

administrators alike. 

The objective of this study was to investigate the systems and 

procedures for outpatient flow and to recommend improvements. 

rhe specific objectives were to: 

1. Understand and describe, in operationally meaningful 

terms, outpatient flow systems and procedures in effect at 

the time of the study; 

HOSPITAL & HEALTH SERVICES ADMINISTRATION 1 WINTER 1978

Figure I. Total Patiet Visits 

125 

12n 

115 

l1n 

1 5 10 

ln5 

1in 

- 138 - 

1968 1969 1970 1971 1972 1973 

THE PATIENT VISITS IN 1968 ARE USEO AS THE BASE FIGURE ('100)

- 139 - 

2. ldentify problemn areas in a -, and all patient flow related 

operations; 

3. Collect and analyze meaningful hard data to test various 

hypotheses regarding the status of the patient flow systems 

and procedures and the reasons for the evolution of these 

systems to their current state; 

4. Delineate the most meaningful and reasonable goals and 

objectives for the patient flow system; 

5. Recommend both long-range and short-range improvements 

to the patient flow related operations. 

Historical background 

The "old appointment system" was romprised of a central appointment 

desk (CAD) and several parrially decentralized nonuniform 

procedures. Within the Medical Division, the CAD 

scheduled physical examinations only. All types of appointments 

of the surgical division as well as "forme'r" appointments for all 

departments were made either by the desk receptionists or by the 

medical secretaries. Routing Section personnel were responsible 

for consults, X-rays, laboratory tests, special exams, and reports. 

Each doctor kept an appointment sheet specifying the time of day 

during which different types of patients should be scheduled. 

Because of the rapid growth of the outpatient clinic and a lack 

of adequate support, the CAD became less effective in performing 

its job. As matters worsened, complaints abcut patient flow 

were expressed by physicians throughout the clinic. The major 

criticisms were incompetence in the scheduling function and 

favoritism (more patients were directed to certain physicians or 

departments than to others.) Consequently, some departments 

transferred control of their appointments (except reports and 

consults which were kept by the Routing Section) to appointment 

secretaries stationed on the department floors. Further movement 

toward decentralization followed. As each department devised 

its own rules for scheduling patients, a gradual degeneration 

of the clinic's overall systems and procedures for making 

appointments resulted. This degeneration was further aggravated 

by the rapid turnover of clerical personnel which was 

HOSPITAL & iHEALTH SERVICES ArtMINISTRA1ION 1 WINTER 1978

- 140 - 

spurred on by conflicting directives, lack of systematic. training 

and enforcement, and a system which was difficult to master or 

comprehend. Figure II depicts the dynamic feedback process 

which led to the degradation of systems and procedures. 

Productivity of the clerical staff, measured in terms of patient 

visitslclerical position ratio, is shown in Figure III. The ratio 

dropped sharply for the appointment-making personnel. In contrast, 

the ratio for the other clerical staff categories remained 

fairly constant. It is interesting to note that while thc "actual" 

system effectiveness was dropping, the perceived effectiveness, 

as seen by the physicians, had discontinued its downward trend 

and appeared to be improving because of the ability of physicians 

to exercise more control over clerical staff as the clinic'E 

systems became more decentralized. 

A study of the problem was made in early 1972 by a Subcommittee 

on Registration and Routing which generated a report. 

Another attempt to study the problem was made in late 

1973, with the establishment of a Patient Flow Committee. In 

January, 1974, this committee established the task force which 

was led by the authors of this paper. 

Method of study 

The aim of this study was to report on the systein description 

of the clinic and to provide a set of recommendations. It was 

recognized that these were preliminary recommendations and 

that more detailed investigation of the problem areas identified 

by this study should be undertaken. Specifically, it was felt that 

subsequent studies should repeat the activities of this study, 

utilize mathematical models, and develop an implementation 

scheme (Figure IV). 

As shown in Figure IV, this study phase was devoted to learning, 

describing and analyzing the clinic as a system. As part of 

this phase, a graphical description was obtained by flow chartfng 

the various patient flow related processes. Also, hard data were 

collected to test various hypotheses which emerged from discussions 

with members of the clinic's medical, administrative, 

and clerical staffs on patient flow related operations. Future work

c 

1 

'-a 

- 142 - 

Figure III. Patient Visits/Clerical Staff Ratio 

3.000 

900 

700 1 AVE. GR. TO AL 

° 6nn \ # RECEP. 

500 \ 

c- 400 7o1 

í 

x 11 

200 

800 

AVVE. T AVE.. .OTA L 

700 

2.000 

900 . 

- ~ AVE. GR. TOTAL 

Da- 1 # ROUTING ROUTI &'"P.R. P .7R. . x 3 

PERSONNEL 

fr n% 

69 70 71 72 . 73 74 

YEAR 

.

u. 

o 

a.- 

1 

W. 

LZi 

- 143 - 

c c S 

C. 

C 1 

o k 

re 

I9

- 144 - 

should be devoted to the development, testing, and evaluation of 

each of the preliminary recommendations through mathematical 

analysis andlor computer simulation. 

Patient, appointment and physician classification 

The patients at the clinic are classified into three categories: 

"new," "new-old," and "trmer." A new patient is one that is 

coming to the clinic for the first time. In most medical departments, 

a former patient is one who has been at that de-partment 

previously and has had a complete physical examination 

within a reasonable t.l:e A "new-old" patient is orne who, regardless 

of having been at that ¿»partment previously, has not 

had a complete physical examiner. ion recently. This length of 

time varies from one to two years depending on the department. 

Some medical and surgical departments do not require physical 

examinations. In these departments, a former patient is one who 

has been at that department within a specified length of time, 

while a new-old patient is one who is coming to that departmert 

for the first time or returning after the specified length of time. 

Patient appointments are classified into five categories: new, 

new-old, former, consult or referral, and report. A consult is necessary 

when the patient's doctor feels that the patient should be 

seen by specialists or subspecialists. Normallv this takes place ini 

another department. A report is an appoinUiient (which may or 

may not involve an examination) during which the doctor informs 

the patient of test results and diagnoses. 

Some of the clinic trainees, called "fellows," are allowed to 

perform work-ups of patients for the staff doctors in certain departments. 

Depending on the department, the fellows either 

form a pool and work up patients for any staff doctor or each one 

works up patients for only one staff doctor. 

Senior fellows may be given the privilege of assuming patient 

responsibility under siaff supervision. The number of appointments 

available by a department depends, therefore, not only on 

the staff's available time to see patients but also on the number of 

fellows assigned to the department and on the functions they are 

allowed to perform.

- 145 - 

Thc cu rrent appointmren t- niakinig system 

In ger .,al, there are two basic appointment-making systems. 

One is the centralized system, in which all appointments for all 

departments are made in one location. The other is the decentralized 

system. in which all the appointments of each department 

are made by a secretary or equivalent personnel 

stationed on the department floor. Every request regarding appointments 

is directed tu . central area for the centralized system 

and to the corresponding secretary of a department for the decentralized 

system. All other appointment systems are variants 

andlor combinations of these two basic systems. Both systems 

have distinct advantages. 

In the centralized system, since there is only one location responsible 

for all the scheduling: 

1. Calls for appointments will always be correctly directed or 

connected. In cont. ;t, calls to the decentralized system may 

be transferred several times before reaching the correct location. 

2. Persons making the appointmer.ts know the available times 

for all doctors. Function:, such as coordination of multiple 

appointment.s, and modifications of appointment schedules 

to keep up with the demands of the various appointment 

types can therefore be efficiently performed. 

3. Patients will have only one appointment card. In the case of 

a multiple áppointment, paper work will be kept to a 

minimum and the Record Room will receive only one request 

for the patient's chart. 

4. Economy of scale may result. 

In the decentralized system, since each department has its own 

appointment-making person(s): 

1. Appointments are made for but a few doctors in a single 

specialty. Thus it is easier for the person to know the doctors' 

characteristics (e.g., preferences for certain cases and 

50 HOSPITAL & HEALTH SERVICES ADMINISTRATION I WINTER 1978

- 146 - 

the lengths of time usually required for the varicus appointment 

types) than it is for an appointment secretary in a 

central location vbc^ must know the characteristics of all 

clinic doctors. 

2. The leaming period is shorter and so replacement personnel 

are more easily trained. 

3. Follow-up appointments are made at the floor just after the 

examinations. 

4. The doctors can easily check their schedules for 

availabilities and patient load and any changes in schedule 

can be easily communicated to the appointment secretary. 

The clinic studied had a combined centralized-decentralized 

system. The centralization was characterized by the work of the 

Routing Section. This section made most of the consult and report 

appointments, and scheduled special exams, e.g.: EKG, 

EEG, basal metabolic rate test. The decentralization .was characterized 

by the fact that the other appointment types (and also 

some reports and consults) were ¡nade in the departments by 

several clerical people. A study was conducted to determine who 

was scheduling patients for each department. 

Five types of appointment-making personnel were identified: 

appointment secretaries, receptionists, medical secretaries, Routing 

Department personnel, and personnel from the Appointment 

Service Desk. l For nearly all departments, different types of personnel 

were responsible for different appointment types and different 

appointment-making procedures were being used. :n general, 

the study clearly sho4 d the lack of uniformity of systems 

and procedures throughout the clinic. 

Current appointment practices 

Upon making an. appointrnent, a patient is assigned to a doctor 

or for a special exam in a specific department at a specified time. 

Appointment systems may be classified into three categories: 

pure block, individual, or mixed block-individual. 

'The Appointment Service Desk makes some of the new and new-old appointments. Its 

main function is to screen calis which have not specified a doctor or a department. 

REISMAN, DA SILVA, MANTELL 1 Systems

- 147 - 

A pure block appointment system assigns the same appointment 

:ine ..t the beginning of the clinic session to all patients. An 

individual system assigns a different appointment time to each 

one of the patients. A mixed block-individual assigns a group Ujt 

patients to certair. times (generally at the beginning of a clinic 

session), but also assigns individual appointment times to each 

one of the other patients. 

At the time of the stuJv only one department of the clinic used 

the pure block system. The individual appointment system was 

used by rriost doctors; the mixed block-individual appointment 

system was used by the others. 

There are two basic types of prescheduling. In "prior prescheduling" 

the patient is scheduled for certain laboratory tests, 

X-rays andlor special exams before an appointment with the primary 

doctor. In "posterior pre-scheduling," tenative appointment 

times are saved for the patient, in addition to the patient's 

primary doctor appointment. After the primary examniinat;oií, the 

doctor decides whether the tentative appointments should be 

used or cancelled. Combinations of the two types of prescheduling 

can and have been used. There are advantages and 

disadvantages to butis kinds of prescheduling. 

One advantage of prior pFc.ciheduling is that the primary doctor 

will have laboratory, X-ray and special studies results available 

at the time of the appointment. This may help in the initial 

examination. Another advantage is that when no further tests 

andlor consults are necessary for the patient, the doctor may be 

able to give a full diagnosis at the initial appointment. One disadvantage 

is that the patient may sometimes take unnecessary 

tests. Another is that as a result of the doctor's examination, the 

patient may be required to return to the laboratory 2 or to the 

Radiology Department. 

The advantage of posterior prescheduling is that the time span 

for the complete routing may be less than the time span which 

could be obtained on the day of the appointmnent. One of the 

disadvantages is that when the doctor decides to cancel some of 

the prescheduled appointments, these appointments may go un- 

'In s¢-ne cases. it nmay be possible to eliminate this inconvenwieince to the patiúLnt by 

.aving specimens and blood -aimples for a certain period of time in the lab. 


- 148 - 

filled due to the short lead-time of the cancellation. In addition, 

the advantage of this type of prescheduling may be negated 

when the doctor finds it necessary to order additional tests and/or 

consults. 

At the time the study was conducted, neither kind of prescheduling 

was extensively used at the clinic. 

Systems-related problenms 

Long lead-times for initial appointments were of great concem 

to the clinic as well as to the patients. If their appointment is too 

far into the fucure, patients may look for medical care elsewhere. 

Consequently, appointment times are not used because such patients 

often do not cancel (they become no-shows on the day of 

the appointment) or they cancel too late for the clinic to react and 

fill the appointment slot. 

A two-week appointment distribution study 3 was run to determine 

the number of no-shows (including unfilled cancellations) 

and appointments, botl-, unfilled and kept, for the different 

appointment types for each department. The study revealed that 

a significant percentage of appointment times, especially those 

reserved for consult and report appointments, were going unfilled 

in many departments. Also revealed was the fact that noshows 

are a significant problem for the clinic. For the clinic as a 

whole the percentage of no-shows for new, new/old, and forrnie; 

appointments was more than 12%. 

Proper balancing of patient appointment types on the physicians' 

day sheets was a problem. This can be seen quite vividly 

by examining the results of the appointment distribution study 

mentioned above and comparing these with the results of 

another study which examined the appointment availabilities for 

new, new-old, and consult appointment types, as they were one 

week after the beginning of the project. In this study it was found 

that in some departments the next available new, new-old, or 

consult appointment was often as long as one to two months into 

'Only patients who actually made appointments were considered. 1 hose who did not 

make appointments due to the long lead-time or any other reason were not considered. 


- 149 - 

ihe future. It was discovered that in some departments certain 

appointment times, e.g., reports, were unfilled while at the same 

tii... the demands for new, newlold and consult times exceeded 

their availability. 

Figures V and VI identify the problems: 

1. The number of ways that a patient can be tr:..sferred before 

making an appointment or cancellation. 

2. Patients wishii.; to make cancellations may not reach the 

correct locations. Consequently, if a patient has multiplc 

appointments, some appointment times will remain reserved, 

even though the patient has contacted the clinic to 

cancel them. 4 

Both problems resulted because the various types of appointments 

are made in many different locations and, iri general, the 

persons making appointments (and cancellations) do not know 

exactly who make 'ihe various types of appointments for other 

departments. 

Especially for the out-of-town patients, the time span of routing 

should be kept as short as possible. s The major obstacles to a 

short time spa.. are the difficulty in obtaining consults, time 

availability constraints (i.e., certain tests are only available at 

certain tines of the day), and procedural constraints (i.e., some 

tests which require special preparation may not be followcd by 

certain other tests.) 

It was found that missing or incomplete charts were a major 

problem for many people in the clinic. A missing or incomplete 

chart may cause delay in the patient examination, a deay ;in the 

report, and often results in a great loss of tC ie i-i locating the 

missing material. 

When an appointment is made, the person who makes it 

should generate an appointment slip and send a copy to the chart 

processing area. This is called preregistration. If a patient is not 

preregistered, the paper work is started when the patient arrives 

'A patient with multiple appointments may assume that if he cancels one of the appointments, 

the remaining ones are also cancelled. Usually this is not true. 

sit was the clinic's policy to try to schedule out-of-town patients for the shortest possible 

routing time span. 


a o 

0 

44 

u 

LI 

Xr> a 

i, 

LU 

u- 

:1 

- 150 - 

I 4 

1 

1

- 151 -

- 152 - 

at the clinic. As a consequence, the patient may be late for the 

appointment or may be missing necessary paper work. 

In the case of multiple appointments, it is possible for the 

patient to be preregAs.:.zd more than once. This occurs when all 

the appointments are not made by the same person. As a result, 

there is duplication of paper work, the chart may not be sent to 

thc place of the first appoi -.ment (it will be sent to the first place 

requesting it), and if the patient decides to cancel, appointment 

times are usually lost (Figures V and VI). 

Physical examinations are usually scheduled early for both 

morning and afternoon clinic sessions, mostly for the morning. 

Often routing, Ilboratory tests, and X-rays are required after 

these examinations. Since a physical, on the average, takes one to 

two hours, the bulk of the physicals are finished after 10 a.m. and 

before noon. As a consequence, bottleneck situations develop in 

routing, laboratory, and radiology between 10 a.m. and 1:30 p.m. 

A smaller bottleneck occurs around 3 p.m. as a result of tne conmpletion 

of the early afternoon physicals. Figure Vil shows these 

peak-load situations at thc laboratory. The situation is worsened 

by the fact that the heaviest patier;t load occurs during the lunch 

hours when these three areas (routing, radiology and laboratory) 

are often short of personnel. 

Room availability is another source of problems. These problems 

tend to be more noticeable in the medical departments, 

especially those performing complete physical examinations. 

This is less of a problem for the surgical division because surgeons, 

in general, spend more time in the hospital tha ' the medical 

division physicians. As a result of room constraints both the 

physician and the patient may be required to wait for a vacant 

examination room. 

A study of this problem was conducted. In recognition of the 

difference in needs for roóms between medical and surgical divisions, 

the data on the examination room availability were separated 

by division. The number of examination rooms in a department 

was compared with the number of staff physicians on 

duty. It was apparent from this study that some departments had 

more than enough rooms while others had too few. 

Physicians' schedules also have an impact on the problem. For 

REISMAN, DA SILVA, MANTELL I Systems

- 153 - 

Figui.'¢ VII. Patient Arrival Distribution for Blood Drawing in 

Laboratory-Day of WeeklTime of Day 

60,1 

55 

50 

45 e * 

· 

40 

THURS .-. : 

I \ 

30 35 o .. . \ 

MON. 

'Ji ';. 

,~~~~~~~~~~~~er 

· e~. ·WED. 

1 S d>\ ¡ \ 

15 I 

FRI. 

in 

5 

1

- 154 - 

example, a physician may schedule two physicals at 8 a.m. These 

patients will be worked up by fellows while the staff doctor may 

be seeing former patients. In th,.s .ituation, the staff physician's 

patients will be using three or more examination rooms at the 

same time. As was seen in the study, few departments can afford 

this many examination rooms for each staff physician. 

Personnel-related problems 

Obviously, the clinic's patient flow system is very complex. For 

the entire existing system to work properly, it was necessary for 

the many patient flow related processes to interact smoothly. 

Thus, it was imperative that all personnel understand how their 

work affected not only their particular process but also all the 

others with which it interacts. At the time of the study, there 

were no structured training prograr;is in which the new employees 

were given an overall perspective of their job as it related 

to other clinic operations. The only training which took place was 

on-the-job because a. new employee was hired only when the 

position was already vacant. Therefore, adequate training was 

very difficult because of the urgency with which the employee 

was needed. Poor job performance and lack of coordination resulted 

since a new employee did not know how his or her work 

interacted with the other employees' functions. 

In some areas there were certain inequities in job classitications. 

Some jobs, where specific skills were a definite requirement 

(such as typing), were classified on the same level as jobs 

requiring no specific skills, no ratient contact, and no machinery 

to operate. This was due, in part, to incorrect original classifications 

but more often it was the result of lack of reclassification as 

job duties changed and became more complex. There was no 

procedure apparent for regular updating of job descriptions and 

classifications. The job classification inequity seriously delayed 

recruitment of qualified personnel. To compensate for this problem 

supervisors and department heads juggled personnel within 

their departments to insure efficient operation. They, at times, 

had to shift employees from one classification to another without 

changing salary or job classification. This often resulted in per-

- 155 - 

sonnel working in positions outside their classitication and in 

dissension among employees if variations in salary scales became 

known. 

Depending on the department, the work-load varied conriderably 

for the same job classification. For example, an appointment 

secretary of a department attending many patients had a 

heavier work load than one stationed in a department attending 

fewer patients. 

A lack of upward mobility in the job often resulted iii lack of 

interest and motivation to perform well since there was no incentive 

or additional compensation for doing so. There were only 

two job levels for medical secretaries; medical secretary and 

senior medical secretary. Appointment secretariec had but one 

job category and no opportunity for advancement. 

Conflicting directives caused personnel prob!errs. In many instances, 

clerical personnel ::ceived instructions from physicians 

as well as from a supervisor. Sometimes these instnlctions conflicted 

with one anether and it therefore became difficult for the 

supervisors to enforce those systems and procedures which 

should be uniform throughout the clinic. This situation presented 

a dilemma for clerical personnel and resulted in a decline 

of morale. 

Rapid p.rsonnel turnover further aggravated the problems. 

This lack of training resulted in poor job pertormance and criticism 

from both physicians and supervisors (sometimes conflicting 

directives), which in turn resulted in low morale. Low morale 

causes personnel turnover. As one can see, the situation was a 

vicious circle in which turnover led to more turnover, to further 

degeneration of systems and procedures, which led to staff dissatisfaction, 

etc. 

Ultimate objectives 

The objectives reflect a synthesis of hard data and subjective 

inputs from medical, administrative and clerical staff regarding 

existing and desired operations. They are to: 

1. Tailor make to the needs, wishes and constraints of every

- 156 - 

doctor, the best possible methods of scheduling. 

2. Effectively, quickly and uniformly process all 

appointment-making functions, e.g., making, cancelling, 

and rescheduling appointments, in a manner which is consistent 

with good medical care, and with human and material 

resource constraints in mind. 

3. Give each patient the advantages of a lárge group practice, 

e.g., laboratory and radiology facilities, in-house expert 

consultative capabilities, attached hospital facilities, etc. 

4. Effectively triage patients. 

5. Enable each doctor and/or department to pretest any and all 

changes of scheduling procedures. 

6. Effectively requisition, deliver in a timely fashion, and 

control patient-related information flow, e.g., charts, X-ray 

and laboratory reports, and f.htancial information. 

7. Balance patient flow to avoid overloading clinic units. 

8. Have a patient flow system which can easily adapt to 

changes. 

9. Have available, in an easily accessible and updated form, 

all relevant data regarding patient flow, e.g., appointment 

availabilities. 

10. Keep both the patient waiting time and physician idle time 

low. 

11. Make appointments available in a manner which is consistent 

with the patient's needs. 

12. Efficiently handle multiple appointments and the time 

span necessary to complete a multi-apFointment diagnostic 

andlor treatment plan. 

13. Reduce the number of no-shows and of unfilled time slots. 

14. Provide for flexibility ,a doctor's schedules within a reasonable 

planning horizon so that changes in availability 

times and policies can be made easily. 

15. Have the advantages of both centralized and decentralized 

scheduling systems. 

16. Make available to physicians innovative means of scheduling 

patients and an ability to pretest such in a "laboratory" 

environment, e.g., prior and posterior prescheduling.

- 157 - 

17. Keep the time span of the patient routing as short as possible. 

18. Improve receivables handling, e.g., reaucing number of 

lost billings. 

19. Identify all -:.itient flow associated administrative costs to 

design a system which will optimize the same. 

20. Improve personnel morale and cffectiveness. 

21. Improve hiring practices. 

22. Have the administrative support necessary to implement 

the clinic's patient flow objectives. 

Personnel related recommendations 

1. Using inputs from this project, a patient flow procedures 

manual should be generated and kept elrrent. 

2. A systematic training of existing clerical staff in systems 

and procedures should be established. This trainirng 

should consist of lecture sessions on the various patient 

flow processes. r;fis could be done with the help of the 

process diagrams discussed. The essence of each job (goals 

and objectives as well as systems and procedures) and its 

role in the clinic operations must be discussed in detail 

with the employees performing the job. 

3. A systematic training of new medical and clerical personnel 

in systems and procedures should also be established. 

4. A systematic enforcement of systems and procedures is 

imperative. 

5. An improvement in hiring practices is necessary. This can 

be done by proper testing procedures and by maintaining 

a small pool of trained personnel to move into both temporary 

and permanent vacancies. 

6. A reclassification of jobs, allowing more upward mobility, 

should be made. For the same job positions in different 

departments with different work loads, some kind of incentive 

should be established. 

7. Commendations from doctors, when work is done properly, 

st. ild be encouraged and would help the morale of 

clerical personnel. 

HOSPITAL & HEALTH SERVICES ADMINISTRATION I WINTER 1978

- 158 - 

8. Consider relieving medical secretaries of appointmentmaking 

duties, where feasible. 

9. Define more clea.J'y lines of responsibility and control of 

clerical staff to avoid conflicting directives. 

Scheduling-related recommendLa'ions 

1. Consider providing, at .ach heavy consultative load desk, 

scheduled "standby" time on a rotating basis, to take up 

urgent consults without overloading any one doctor. This 

could be accomplished by having at least one physician at 

all times performing some clinic functiorns which could be 

easily interrupted. 

2. Based on historical data, consider devising a battery of lab 

andlor X-ray studies which result from some percentage 

(say 80% or 90%) of physicals for a given doctor or department. 

3. At the doctor's discretion, consider prescheduling the 

above lab and/or X ray studies prior to the patient's appointment 

for a physical with sufficient lead-time to allow 

for processing, reporting and delivery of results to the doctor. 

6 

4. When historical data indicates that certain appointment 

types with a given doctor, inevitably or to a large extent, 

are followed by some other appointment, either with the 

same doctor or elsewhere in the clinic, then consider prescheduling 

the latter as well. 

5. Consider storage of specimens for some prescribed period 

of time in the laboratory. The inconveniences caused by 

additional visits to the laboratory will be avoided. This 

would be especially useful for prescheduled patients. 

6. Consider a patient standby list. The patients in this list 

should be easily reached and would help to alleviate the 

problem of no-show and late cancellation appointment 

slots going unfilled. 

'As a first attempt to identify which further appointments result from physical appointmnents. 

studies were run in each department to determine how many of each type of 

consult, X-ray tests, laboratory tests, and special studies were ordered over two-week 

periods. The results of these studies are displayed in Table 1. 


- 159 - 

TABLE I: Pnmary DesklConsultant Desk InputlOutput 

UROLOCI 

I[ 

1 1 1 

1 1 

1 1 

* . . . 

. . 

E 

I I 1 

. . #. . 

. . 

. . 

:S

- 16u - 

7. Consider overbooking policies based on no-show and late 

cancellation statistics, by department or by physician. 

8. Consider calling patients to ascertain arrival. By doing 

this, many of the predictable no-shows could be determined. 

Personnel requirements for performing this job 

need to be considered. 

9. Consider altemative scheduling methods to optimize patient 

waiting time and physician idle time. The following 

factors are among those which should be considered: 

staff-fellow interaction, room constraint, patient load, and 

appointment sheet time increments. Computer Simulation 

can be used to pretest the efficacy of various altematives. 

10. Consider altering the availability of each appointment 

type, based or both physician determined criteria and patient 

needs. Patient needs can be determined to a certain 

extent by examining appointment availability data and 

data on unfilled appointments. 

11. Study room availability constraints and the effect of same 

on doctors' schedules, fellows' schedules, staff-fellow 

interaction, and hospital rounds. Again, Computer Simulation 

is a good study vehicle. 

12. Define a patient priority classification schc ne, based on 

patient's urgency and importance, and doctor's professional 

interest. Institute this classification scheme 

where feasible and acceptable to the individual's practice. 

13. Consider providing booking options which will have, for 

example, 100% filled appoilntment sheets for the most immediate 

time periou, 80% filled for the next time period, 

60% filled .. , etc. Such a booking procedure will allow 

more urgent patients (as determined by the patient priority 

scheme, described above in 12) to be scheduled sooner and 

without overloading the doctors. 

14. Define the optimal time periods for the above, for each 

desk or each doctor and each appointment type. As an 

example, a department could have new appointment times 

fully booked for a two week period while the consult times 

could be obtained on the day of the request.

.Recomme;tded systenz 

- 161 - 

.;any of the clinic's problems, identified earlier, are concemed 

with information flow. Problems arise when information is not 

readily available or is not transmitted properly. 

In respon - ' to, these problems, a patient flow information data 

bank was recommended. This information bank should contain 

all information relevant to the patient flow processes and should 

be easily and quickl;- ,~cessible. This information would include 

appointment availabilities for all doctors and central file of 

scheduled appointments. In addition to receiving, the data bank 

would be responsible for distribution of all information regarding 

patient flow, such as billing, scheduled appointments, and 

demographic information. 

In the recommended system all doctors' appoir.tments would 

be made at the departments, with the exception of consults, reports, 

and special exams which would be held by the routing 

area. This is, in fact, what was presently being done at the clinic. 

After an appointment is made, all informatloiU, regarding it 

should be sent to an appointment processing area. This area 

would be responsible for all paperwork associated with appointments 

and updating ¡i¡e central file of appointments as well 

as appointment availabilit es. Further, by checking the central 

file of scheduled appointments, this area would easily locate appointments, 

make cancellations, catch duplicate appoin¿n.^nts, 

and coordinate multiple appointments, multiple cancellations, 

and rescheduling. 

In the recommended system, appointment information would 

be updated both at the departments and in the appointment processing 

area. When a doctor or a department is not specified, the 

call will be directed to the Appointment Service Desk where a 

triage is made. Some patient triage situations are too corrplex to 

be properly handled by a person with non-medical training. In 

response to this problem, a Physician Triage Officer should be 

established. It could possibly be implemented by providing 

physicians' consultative services (on a rotating basis) to handle 

these triaging of non-routine patients. With such a service available, 

a non-medical - orson who does triaging of patients would 


- 162 - 

have the opportunity to conr:',: a physician about difficult situations. 

For both routine and non-routine triaging situations, a questionnaire 

could be developed to be completed by the patient, 

which would provide a means for systematic evaluation of triaging 

accuracy. On the ba;-s of this questionnaire, an appointment 

could be changed at the discretion of the Physician Triage Officer 

or any other medical staff. 

For multiple appointments, the appointment processing area 

will make a triage (with the help of the Physician Triage Officer 

when necessary). Then the area will call the proper deE!;z and 

coordinate the appointments. This coordination will be made 

easily becau,e all appointment availabilities will be known at the 

appointment processing area. In tl.e new system, the problem of 

excessive transferring of calls fromr department to departrnent 

will be avoided. 

A major problem for the clinic was that of no-shows. 7 However, 

many of these no-shows are predictable. For example, a 

patient with multiple appointments may call his primary department 

to say that he will not be able to make this appointment. 

The primary department appointment will be cancelled; 

however, often the person making the cancellation does not 

know about the other appointments for that day. These appointments 

will not be cancelled and will result in no-shows. This 

situation can bei avoided in the recommended system since, 

whenever a patient cancels an appointment, all other appointments 

can easily be checked in the scheduled appointment file. 

Part of the updating procedure in the appointment processing 

area could be to check in this file (urnder the patient's name) for 

additional appointments for the same day or associated, in some 

way, with the cancelled appointmentS. 

A predictable no-show can also occur when . a patient with an 

appointment scheduled sometime in the future has an appointment 

with the same doctor at an earlier date. This often happens 

due to a change in medical urgency for the patient. In the present 

'The billings foregone due to patient no-shows are estimated to be in the $1,000,000 to 

$2.000,000 range per year.

- 163 - 

system, there is nr'. :vstematic procedure for cancelling the original 

appointment and often the result is a no-show. This situation 

can also be avoided in the recommended system by making it 

part of the updating procedure to check for these duplicate appointments. 

Another feature of the recommended system is that data collection 

for studies such as the ones described in this paper would be 

much more easily collected since all patient flow related data 

could be collected at one location, i.e., the information data bank. 

Thus, such an appointment system would have most of the 

advantages of both centralization and decentralization. Further, 

it can be easily implemented since no major changes in clinic 

operations are necessary. 

Strategy of going towPtr: ',¡e ultinate system 

There are two ways in which improvements in systems and 

procedures c,: be implemented. One is to schedule a complete 

changeover on a certain day. The other is to do it gradually, 

step-by-step. 

The first method would be too risky and too disruptive to the 

clinic operations and not readily accepted by the clinic personnel. 

On the other hand, gradual improvement will not be so disruptive 

and will be more easily implemented. In this implemnentation 

scheme, as tangible improvements are realized, 

credibility and increased confidence and patience with the system 

will result. Consequently, further changes will be more 

readily accepted. Figure VIII shows the dynamics of this implementation 

process. 

As an initial step in the implementation process, a manual 

system could be devised to handle appointment-related information. 

As soon as an appointment is made, information regarding 

appointment availability will be immediately updated at the department 

(on the appointment sheet). At the same time, some 

mechanism for transferring this information to the appointment 

processing area would be initiated. In this manual system, the 

means of transferring this information would be by telephone, 

lift or runners. Such a manual system would not be disruptive to 

present clinic operations.

- 164 - 

oo 

a, 

r. 

3 

»e 

' 

i 

e, 

n(fD 

-_. 

la

- 165 - 

- A major problem with such a system would be the lead-time 

for trar.sferring information from the desks to the appointment 

processing. During this time, the appointment availability file 

will not be up to date. This would cause some inefficiency in the 

appointment processing area's ability to reschedule and coordinate 

multiple apl, intments. 

The ideal situation would be to have the appointment 

availability file updated ;tmmediately from the desk. This feature 

necessitates some kind oi mcchanized information transmission. 

This mechanized transmission could be accomplished by using 

teletypewriters or some other telecommunication device at the 

desk to transfer information to the appointment processing area. 

In addition to the speed with which information would be 

transmitted, this system would have the advantage of being able 

to generate typewritten copies of the appointment information at 

the appointment processing area directly from the desks immediately 

after the appointment is made. This capability would 

eliminate the need for filling out and sending a pregistration 

form, since all the necessary. information could be sent via the 

teletype. Further stiidy could be made to determine the feasibility 

and desirability _f .e;iding other types of information, such as 

requests for charts and X-ray, ~ the same manner to other clinic 

areas. 

Once a mechanized information transmission device is in operation, 

the major system bottleneck would be in the time necessary 

to manually update the files. This problem could be solved 

by mechanizing the updating procedures using some computer 

configuration. 

Once the appointment information system is mechanized, the 

next step would be to incorporate other patient-related information, 

such as billing, which could be accessed easily from the 

proper places. Thus, this appointment information system could 

be the foundation of a complete communications-based information 

system. 

An advantage of this implementation scheme is that, at each 

step, improvements are made in the patient flow system. Also, 

each of these steps ser -s as a transition and learning period for 

the next. Further, if, at any step in the implementation scheme, 


- 166 - 

the feasibility and desirability do not justify the implementation 

of the next step, all the advantages of the current stage will be 

retained. Moreover, with proper forethought and planning, each 

of the implementation steps could serve as a backup mode for the 

more advanced system sta 's. 

It is important to note that the successful implementation of 

each step is, to a large extent, dependent on personnel support. 

Therefore, improvements in the personnel situation must accompany 

any and all changes in systems and procedures. 

In future work, detailed study of each of these steps should be 

made. Cost-benefit analyses of different means by which each 

step could be implemented should be performned. Also, the 

scheduling recommendatior.s presented earlier should be extensively 

investigated. 

More than a year has passed since this study was completed. 

The clinic personnel are currently methodically implementing 

most if not all of the recommendations delineated, using the 

directed step-by-step approach of achieving the ultimate goals 

and objectives.

- 167 - 

T1c Iteole of Olieraionls Rcscarlch iii 

Rcgioulal IIeCilIh Plauniiig, 

Larry J. Shlmatgant, llarvey W1olfe 

Unive s",, g/ ittlisbrgh, P'ittisbudrgl, P'mJs!lh ni( 

It. I)ixois Sjl)'is, Jr. 

,imaiJord Ul,.. r,!/, , Slnalford, California 

(llec'eived Augist 1, 1972) 

Operations-research workers have not met with much success in being accepted 

as integral members of regional-health-planning teams, owing in part to a lack 

of understanding by health planners of the skills the operations researcher 

has to offer and in part the analyst's inability to demonstrate that he can close 

the gap between theoretical modeling and the impleulentation of his res-lts. 

This paper explores tite growth of regional health planning in the United 

States and highlights its important problem areas. The literature of operations-research 

applications to health planning is reviewed critically with respect 

to thie feasibility of miodels atid the appropriateness of assuinptious. 

Specific problems with the types of studies currently in the literature are identified 

and recommendations are made for improved coordination between operations-research 

workers and health planners. 

H 'AI,TII Pl1,ANNEli Iiave! inlriltld ' ,liv(ry sstemn tliat d(,vtlol),)d lini>ost 

randomly iii accord withi local politica. economic, and/or parochial interests. 

Cammuility3-wide planining typically pl:tyed a minor role. As a result, tlih( untvent 

distribution of healit resoaurces :ad s(rvi(ces tirotugalio t a region Ih:.s Ih(4et a per. 

tist(eit Irobhlmi coimnlic:ttinig pliininig. F'urlher, aus long as h(iolitals wvr( required 

to provide relatively simtple procedures, no great strain vwas placed on nianpow(,r 

sul)plies, nor was the ecoeonomi burdent caused by duplicatiozn of servic((s undluly 

heavy. Specialization and advanciing technology lhave elihaigd thlis situation. 

'T'oday's compl('x medical Iroccedur(es r'luirae X,.l),isiv(,, soplisticite(ld (lUiil)mei't 

facilities, as well as teams of higlhly skilled personiíel. 

ntl 

2IAYslI h1as described the history of health planning in the United States as 

highilighted by thr(e lanidmvark é(vSents: Tli( pass:ag( of tite Hill-Burton Ael (1>1. 79- 

725) in 194(i; tithe 1Xi joiit publ)lic(atio, by tie, Amnerican Hlosp)ital As.ociationm anud 

Ih(, Jublic Ilailtlh Service of tllt report, "Area-wvide lhlalinitig for Isl)pit:is andl 

lelated He(alth l Facilities;" tui and th(e pamssage of the Co)mpr-ehensive Ilealth Pllanning 

Act of 1966 (J>L 89-749). To th(lies might be added the Regional Medical 

Irograni (1>1, S!)-23:i)) eitclted i¡n 1!XiS, ai(d thi(! I',xpeiriillttl l leaultlh Servicets l)e. 

liver3y Systems programn adopted iii 1971. 

Health planinning int the United States most likely originated with a volulntary 

groul) that formed tihe Cinciiimmti Public Healtli Council iin 1917.11 Tihe growth lof 

Coinmiunity hicalth couiicils miimlSrooi(cmld; by 1(M99 tih.re ;were 1222 suieih l!o:ld anil 

st:te orgtaniz:ttiots. Ma\lanl of thi(s councis served ns the maini htulti-)lt:iiiiiiiLg

- 168 - 

Regional Health Planning 

bodies for their localities until the early 1960'.ns lin 'area-wíi

- 169 - 

Larry J. Shumc- Harvey Wolfe, ond R. Dixon Speas 

effeetively availabici< t(

- 170 - 

larry J. Shuman, Harvey Wolfe, and R. Dixon Speos 

very lcast, a qualitative inprovicni(uit. liucltud('d aiuímong tilhs. a('liVii(i s flir ¡liiihi 

planning decisions lhave resilt( d ill shar(irl einiterprises are ma:ternit y sa.¡d (nni rg.n¡cy 

care, data processing, inservice e¡du(.ition, labonrtories, a¿¡d híuris ¡ i,.is'. 

to forinl a joint veil ure for sl,.. :.g s( rvi)'es ,I i' iiot diflicult to ini)l¢,l1i1(.11, 

the obvious logistical probhleis th¡lt resul. 

::¡ith, I'r o,: 

Iinplheinntation is not so siilps>h, wIin oniie or mlor(,r institutions giví' up i ,a ) .irvi,'í. 

(ilir(ely, while s(l('ectld olsil)itllis ¡('1 att:

- 171 - 

'- ionoal Heolth Ploanning 

be vicvwed as a planning model, yet the(y arc disenchanted with the sucee.ss these 

models have had with respect to implenmentation. 

Twao in-hosl)ital studies that utilized somie ,f ti(h operations.resear(.lc techliilues 

iinow being applied to regional-l)lapu ing problenms are by WoLFE V 5iJ aid I"ETTEII ANI) 

TH~oMPsos.i'i In 1964, Wolfe developPed an inte(ger-prograniming mnodel for allo. 

cating nursingservices *within a hospital. An imxportant attribute of Wolfe's model 

wvas the adaptationl of psycholne

- 172 - 

Larry J. Shuman, Harvey Wolfe, ond R. Dixon Speos 

and Thompson to the abstract mod(l postulated by I:lagle. S.IALL00O1D, -0NIK,J, 

AND OFFENISENDI5 I 1 

Ipropose 

a-. '; ial simulation niodel thliat describes the patieniit's 

movemenit tlirough t he hc:lltlh systeiii, detaIliiig resources required for various stages 

of illness and recovery. ''The nod

- 173 - 

Regional Health Planning 

are propo.~sd in tile Revelle et al., Feldstein, and Iirystynak papers. Thie first twn 

are con(erne(d , ith tluh(reulosis a¡n( tihe last w¡ithl rheulmatie fev(er. ieveli. et al. 

lirst posthiltited a deseriptive mod(ll of tlul)erculosis el)hilniiolehgy for a ¢hvehloling 

lnation i¡t order to evaluate tier eost eftetive)e'ss of various metilods of control. 

Tihe des(ril)tive model ean ihe formulated as a linear program. The solution of the 

iin¿eaJr-l)ro>gr'i n)iig motl, yieldlM tie miiiiininl-cost metilod for attaining a given 

reduittlti-l it ltul rcutilar. c:s.s h ti iitg :1a S.t.i ilic ti t i i;ridl. ol, 

FIeldstein proposes a tube(rculo.sis plianiting model for use ini developing eountrie(s. 

This model selects thi( set of activities that woulld maximiz(! tio total econonmic 

benefit for given resource coistraints. It consilders the interrelations of public- 

Ix)lircy I>re('f,

- 174 - 

Larry J. Shunmn, Harvey Wolfe, and R. Dixon Speos 

ith( patioeit and provid'r costs attrilhutablh 

r4gion. 

to Ille( locatioi.s of lhospitals witilin a 

CSIE.NIlIACK AS\>) C IIEIL .'S. . 

t lg 

i)rl),s'Sti :a st:ttia.ll L('ea llitlUeiiq'e knowi\l :Is s(('orgranims 

for describin g a híosplital srvicce ar,;'a whi(,', n¡atuiradl arri(,rs e.xist. lii nei tlal 

applic.timi, tihe ove'rllappiig of hospitail sarvice areas mnake definitig uniiqniues 

botilldaíries for: Ii¡:li¡cell:a a 

rcígiu r ísxl 14 ni and Morrill i)rtvide :¿noihbhr applroaei tu lb(' Ioeationail analysis of 

rcgiozíal hIos¡)ital syst .ils. I']AIilKM.SONPi21 ts.a :1 1irinsl)orl:ti il li mrly served Il aiient groips is w(e11 as poorly Ioeait(l ( ealth f:iiliti ¡s. Th(.y 

I)rOlOUe a;1 i4`r(41 ('OMlii)isx MiitIhlI w\\hl¡ici, .is :t liirast (>te¡), Nsiiiiil:l 

!IEI.LEVI.aI illi(:Cl('d IluLt th( siímullatlio model was al)le to Irovide a good replic.ation 

on t1he (olic: g(o Ihitsl)i;l: sytsltii alnd 0tihal hli gravit.y m(o wtas (I a g as

- 175 - 

Regio.,oi Health Planning 

wvho wnouii' hoo.ns the prepaid group practice mode of th(e reeviving eare. Utilization 

is considered ti funection of travel time as estimated from existilng hospit:ii.uiage 

patterns. Constraints are inicluded that ¡pli c mlinimmini ciapaciti(es or euach linic, 

limit. tihe nuniber oí clinics, assig il eople to tl¡o neairest eli.ic, and restricrt exxllnditures 

to the available Iredget. 

ABEUNÁATIY AND IIEI¡u¿HEY'V 1 propose to solve the healtli-facility-loeatln probilen 

by uising thei mtoldl idll'eitol I¡eviisly dts.rilwdl. IFouir olbjectivwf rxtietions are 

postulated: (1) maixinize. utilization, (2) iniiiiiiiize distance per capita to til( ne:arest 

facility, (3) minimize distance ;. , i 't, and (4) minimize the decline in utilization. 

Population is divided into census tracts and then suhdivided into socieiononil. 

classes. Utilization is a funiction of distance, social chIaU, anid personal preferene. 

The Abernatliy/lHerslcey model consilders aill poinits in a region as possilhl sil(. x 

rMther titan evaluating pIred(et(erlined I)oltential lcations. A modii(ed s('ar(' 

algorithm is used to select the locations. lii considering all p)ossible points as locations, 

tliere is a p,:s:bilit.3 that sites that are infeasible, e.g., rivers, pIarks, or very 

'expensive prol pert!', imy 1ie cvhosen \as Ineati os. 

AlnU.i' of 1l(! iSlt:4. (Of 0loch tial iO(,h ls »roo.s'l by g>gr:n)hli(.'rs :by( o operations 

researchers have been incorporat(ed hy B]AUMi'i into a simuhltor for evaluating 

primary- and anibulatory-care delivery systems. Baum's simulator is general in 

ti it. it. cut('si hlcr aniy ronfligurati(ml of h)slíeiltal oIull.tLItinl!t altivities, .sa 1elihi, 

clinics, and neiglhborliood healthl ceniters. ln additio¡n to patient, travel tigel to 

facilities, thce model treats some behavioral aspects by considering tie prohlability 

of facility choice, keeping appointments, and remaining wvithin a particular delivery 

system. 

Loeatioai-aly::llsisu .iii.' :,, ,'vilv h v:alil:le metll(odl s for evalhlating tle( s):ltial 

relations betwecn ai health delivery sys' .. and the conmmunity. lIowever, they 

do not consider a complete spcctrum of E vices, nor do thoy treat tihe pioblem of 

distributing specific services throughout the health region nor the consequent effects 

of rcmoving or ndding servicés. The nmaasures of effTectivoene(ss used in tlhese mIodels 

¡Vive been overly simplistic and inadequate. Except for travel costs, minimal consideration 

is given to the cost compo¡íents of thc system. Tlihcre have been no invcstigations 

of thc economic costs involved in closing units, relocating physicians 

and beds, or constructing new facilities, even though in many cases these costs may 

dominate patient travel costs. 

models is presented by SCOTT. 

A compr!lensive review of location/atllocattion 

1 i'i 

planning is given by Guoss.lim 

A critique of locational models applied to health 

Thc operations-reseatrch approach has evolved from in-hospital to multihealth 

facility applicatioins. Yet, mnany peolle, includimng those actively involvedl inii ialth 

operatiois r:secuiare, liave (luestiouitld tIhe valie arid inlpuct of ntost nl udits. J3.th 

GuzEi' AND YOUNo 16' 1 have noted that very few, if any, operations-research projccts 

have had a profound effecct on the delivery of health care. Gross(""l has pointed out 

tihat, vithl few excetlions, healll operatioiw-r(esarchi studies have resulted in 

cruditc models that cither have little meaning or are given low priority by health 

professionals. He cites the serious lack of mutual respect, understanding, and 

communication between physician, medical rtsearcher, and systems analyst as 

another rcason for the lack of success in implementing quantitative studies. "Before 

healthi planning ¡moves from the rhetoric to tihe acti(on stage, physicians will 

h1ave to be conviicel, thlrough (edueatium i¡id timr)ugh diJun.inji 'tion, tha:t plinniiig

- 176 - 

isN ort lwIeI i ,... .iul h ha t e cal ti Ihlltiiii-iiig I)ro('(ss is nor(. :lt extrava'gallLt %i h¡il of 

academics, intellectuals, andl bir.auerats ) silting in splendid 

world." 

isolation frounm ith real 

,ICLAuo HLIN, 1 ',I l oun the other hand, is not as pessimistic. He notes that in the 

health ciiviriinmi it it is extreit*l altraetive to selhct first thie problens thalt artmost 

tractable, *laniing legisl:tion sotglht 

to >overcIome, aLI ilablility to look beyonxd its own traditiUnal methoids. Only a few 

comprehensive-lhealthi-planning agenciecs have lihad the courage or sophistication lo 

support meaningful operations-ret¿. .rch activity. 

Even so, a great nimany studies have been carried out, and it seems apparent that 

much of the failure of these studies to be implemented lies withi those doing tile 

quantitative work. 'Thliy llave lOt integrated tlhCiselves iito tlhe plalnning ,: - 

tem. They operate as outsiders viewing the planning system o- an abstract structure, 

ignoring the fact that it is political, social, and highly dynamic. DEVISEIIII 

has noted that a chas !; exists betwcen researclheis and lhcalth planners to the point 

that they do not speak the same language. Consetquently, he hUas recommended 

that interdisciplinary exclianges take place among concertned parties. Accuding 

to DeVise, prescriptive models must be capable of specifying alternative future 

plans in the system, predicting the consequences of choosing each alternative, and 

v:aluating th( coisexluuences accordilng to ani acelptablc mcasure of planil¡ing goals. 

HILLEBOE AND SCHAEPERI2 have stated that, in the health field more than almost 

any other area, the professionals have usually asserted that the subject matter 

is too technical anud the consumcr too inexpert to play an important role in planning.. 

Meaningful health planniiiing, they philosophisc, will be controversial if it is to meot 

society's needs. Health planners calnnot be content with attempting a better coordination 

of the existing system, but will have to changc this system. WVhile it is 

true that plannoers have not extensively embraced quantitative approaches, it is also 

true thait qiuatitillivelyc.orilted anialysts lhave slot produced anuy utIk tiha is 

meaningful in the immediate context of planners' needs. 

Operations research has not bocn useful in regional planning because no one has 

used the operations-research approach to aid in the kinds of decisions that current 

planiners are making. A large portionl of the quantitative work that has been done

- 177 - 

Regionao rfeclth Planning 

las ignored tite bIalsíe telets of oi(ratio.s resea'rch and hlis degenrated is;, aun 

exercise ini matlihena:tieal modeling: the prolleims that have beeni attenipted hayve 

iot beena \well formiliulated; tlihe ol)j(tive fum(tioiis l:ave bev(n naively 'osnslrueted 

and tll inasure, of etv yiel ulTtl otii tizatios; quanstitative variables are 

considered of prime impilortaince, but b(lihavioral variabhles are lleglected; alternat ives 

that are politically and socially infeasible are still considered ina the fe.asible set. 

The faihltre of pliianners to view operations research as a viable toul is partially 

attributable l«o : nisi.silulhrstas:ding of h t.l' inlohgV :aind I)o\w(r of tli(. op(rationisresearch 

process. lMde!l building aid thc developing mathematical techniques caninot 

be construed as s ynolymnous with operations rosearch. Yet, many people confuse 

thew t terns, expectisg model building by itself to lead to implementable solutionis. 

This conilusiomi is firthler a('ee(ituat(edl by operationis rwselrehlers whio lirst 

conceive a millthenlirial niodel anad t(heni atteiil)t to lit a health-reltt d problem to 

it usnder the guise, of an aplplicatiui. Altihough this :echisique quite ofte(n leads to 

journal articles, it rarely results i.a impleme,'t ble decisions. The plamner is lcft 

with a theoreitical cont ribution that ihe car. .er use nor appreciate, wvhile t1 hi analyst 

(lu:alilies his mod(el hIy e'lainúing thlit no di:ta were availd)hle for validatioii. 

ill effect, tile Iih:ll-lh-plisiinig problvss tliaLt I:ave been delinied aid soilve-d do nlot resemblie 

the problems for which rationual decisions must. be made. 

Develop)ing and eoin(el)taiilizisig nmathenimatical nmodels and t(ehniq

- 178 - 

Larry J. Shuman, Harvey Wolfe, and R. Dixon Speas 

I Fe IMI ENC:ES 

1. W J. APIIERNATIIY AXI) J. C. III.:U víI, "A , l:1iu1' A;IOe-ation .Mí(hel ifor I{egnhIaI l lv':lIlJl 

BServiecs I>¡anliiiig," Opns. Res. 20, 629-642(1972). 

2. A.rilSTiACl' OF IOSI'ITAI, A. Ai:MiF:NT TiUis,'l'l) e (' olwrtive hlformation ('eitcr, 

'Ihe School otf Public liealtl, IUiverS;ty of .Mi.ltigan, AiI Arlor, MIeihgalln. 

;. 8. I1.itXo.N AX.l, II. W\\I.F:, .llrogranuning Approah for Silertiang 

au Optimuni Health Progtran (aseC Mix," plpcr presented at the 40th National lMeeting 

of thc OPrI.:ATiosls R:S:AR.CII SOCIF:'eIY C.- Arp:ItICA, Analeiml, Califoritia, 1071. 

9. II 1). ('ilt:RXi l Ani J. It. SAlihiladelpilia, ]leiiisylvalnia, 1967. 

10. R. D. CouonILIs, "Hospital Complex Atnalysis: An Approach to Analysis for llnimliúng a 

Nteiropolitash System of ,Serviee Farilities," UInIpllished 1>11.1). I)issertiatii , tniviersity 

of 

I . 1>. I1):,sJ:', "Aletilods ui( 

I(1i95. 9l'eisylvai.ii:m, 

( o'idt'elts ot ll In lt(ri.a'ildilinar Regional 1lospital Stumly, 

Ilealth Services Res. 3, 166-173 (Fall, 1968). 

12, R. EARICe S0ox, "Ti)c Case for D)cccntralizing Cook County Hospital: Sornc ArPllications 

of Liniear Optiniization iln loslpital 'lanniing," Hospital Ila:iimlug ('ounceil for 

Mletropllitail ('hiago, VWorkiog >liqeml, Vol. II , No. 1, C( icago, ¡llin¡ois, :3 . 

13. "Expterimenital lcalth Scrvices l)clivery Systeins: lnformatiola 

Services and Mental lHealth Administration, 1972. 

for Appiliants," llealthh 

14. A1. S. F1ELSTEIx, "Cost lIcefit Analyísis ald lhIltil Plrogramn Pla;ini¡ng iii 1)Deelulinig 

Cotintries," palper presIclted at thc Secondl Conferencec t¡n t'le EcoIInmics of l1ealth, 

l:altimiore,: Mlaryland, 196i8. 

a 

15. R. 13. FErrlER AND J. 1). Ttfotm)dp , "iA 1)ecision Model for D)esign and Olxpratioli of a 

Progressive Patient Care Hospital," Mledical Care 7, 450-462 (1969). 

16. - ANs - , "The Simulatiotl of liospital Systemis," Opns. ReJ. 13, 689-711 

(1965). 

17. C. D. FLAGLE, "Simulation Tech!niques Appllicablc to Plublic IHealth Admiiiistrationii," 

pailer presented at thc Filst Anntual Conferncee ot Americatl Statistical As..ciation 

and Public Health A.ssociationt, PIroceedings oan Simulation il BIusiness aid Public 

liealth, New York, 1966. 

18. A- A.I> J. 1'. Youxa, "A¡pl)lie:tiioln f O¡p .io Itens ,'elaci und hl¡dustritil Enginmeering 

to 1Problems ot liealtih Services, 1Hospitalis and Public Ilealth," J. Indis¡l. 1Eng. 17, 

G09-614 (1966). 

19. R. H1. GIGLIO, F. C. KAMINSKY, ANS> J. \'A'rrs, "A Mathemnatical/Computer P>lalming 

Mlodel for Emergency Medtlical T'raisportationl Systems," pa.Pler preitcnted al tli,. 

Conference ot Areawide ('ompireenicsive 1 lealth Carme Pllaleics, Septeinber, 1i969. 

20. (;I.ASG

A^CJonoal neonah ronning 

21. 1'. F. (inn.i, Souw D(c!r "itran/is of D(tlol,¡rt and Imsple¡menltalion of ltuistl C'are 

Technolog!, and So,'e L.:,ss of thc l.S. Erptrienc, I'iiiVctrity of .iaskathliecali, 

I{egiru, Canadla, 1971. 

22. - , "Ùr.an !ecalth D)isaolerd, Spalial Aiialysis ,id the El.o¡ouiies of Iltaliah Faciity 

Location,!'paper presented at the 40th National M. eeting of the OPEmRATIONS RI:PEARCII 

SocXmr or A. -alcA, Aiahieiim, California, 1971. 

23. R. GuE, "Opcratioiiat Rescarth ¡oi 11clth a;ndl. Hlspiital Aulmimitratiol," ilospital Admtinitration 

10, 0-25 (Fall, 1965). 

24. 11. HILWLEDOZ ANDU M. 8CHAEPER, "EvalwUation in Commraity Hlealth: Relaiig Results 

to Goals,"-Bu/U. New York Acgqdemy of lledicine 41, 140-158 (¡19>. 

25. F. ICiINIOiroK, "Oieratiol lRle.warch aiwl Regiomal l Mdir ogra mS," paeinr pre, 

iikitled al tibm Naimitit aiaiiing," New Einglaid J. of .ledidic 279, 1314-1318 (19G8). 

28. JOINT C:OMuI3TE: oF TIlE AMERICAN HOSPITAL ASSOCIATION ANX PUBLIC lFALTU 

SiERVICE, "Areav;, '- >"':iiilag for Hospitals a!td llelated Health Fa:ilitiea," 1'1IS--, 

WVahington, D.C., lu61. 

29. F. D. KESNX:IIY ANI) M1%. B. WVoob. :., Developmenf of a Simnulaticn .1!.cI l a n Co!i. 

itunity Heloh Srcice Sy8lcn, 'Th Hcalth Sealth rvice Syatem--Ieccml:h T'iaOgje latititute, 

Vol. 111, 1968. 

30. W. L. KMSICK, "P>lalning l>rogramming and Budgeting in IIcalth," Mldicai Ca.-: 5S 

201-220 (1967). 

31. L. F. In'Twry',4AK, "A Afode(l for the Rlegiolual C'otroil of Rlleunatic IFever," lI:il*'r 

presenited at the 38th National Meeting of the OPERATIOXSa IiEtE.iCII SOCILTY OF 

AMERicA, October, 1970. 

32. R. 1. LEE AND L. W. JOSES, The Fundamentala of Good .ledical Care, University of 

Chicago >res, Chicago, Illinois, 1933. 

33. 1. G. Lovs, R. A. MATHIAS, AND G. TRtEoal, "l)yinaic Iilnrniming of lecaltlh Care $ystenms," 

paper presented at thc 37th National Ieecitlg of the OPEatR.TIOXs RlkEaidCiO 

OCímy or AMEalAc^, Washington, D.C., April, 1970. 

34. J. W. LUBL%, D. L. DROs.Ess, A.X L. G. WrLtE, "Highway Network 3linimul, IPath 

Selection Applied to Facility laiiiiiing," Public Health Reporla 80, 771-778 (1035). 

35. C. AIMcLuaOLn , "lHcalth Operatiomls Rescarch atl Svystems Analvysi Literature," 

Syslenm atnd lMedical Care, ALsAX SHELIDO, FRAxK B1AKER AXI) CuarrIas MCLAUGILIX 

(ede.) MIT Proes, 1Boton, IM-a_~Sciuseutts, 1971. 

36. J. J. MAY, "Health Planning: Its Past and Potential," Health Administration Perspective 

A5, Center for Health Administrationa Studies, Univercity of Chicago, Chicago, 

l¡¡iniois, 1907. 

37. R. L. MORRILL AND R. EABICKSO., "Locational Efficiency of Chicago Hospitais: An 

Experimental Model," HfIllh Semri~ Res. 4, 128-141 (1969). 

38. - N M. DI. KzeLLr, "'lhe Sinulatioln of lluspital Use atad the l'stiutio 

of Iocation Efficiency," Geographic A nal. 2, 283-300 (1970). 

39. NATIONAL COMMIRION OX COMMUXITT lHEALTH ElRVtCB#, .Aioan-l'lanning Jor Com. 

rnuniy IlcailA Secric, PIublic Afíairs I'rems, Wa.,hitgtoni, D.C., 1967. 

40. V. NAvARRO AxD R. PARKER, "A Pl>alinllg Model for l>ersomial lealth Serviees," VWorkinmg 

Iaper, ]cpartment of Medical Careo ¡aad Iloipitimi, Tlle Jolama; lolt¡kiitm U Livtcrsity, 

laltimore, Maryland, 1967. 

41. L. D. Poia, "Areawide Comprehenwivc Healtl IPlanning: The Philadelplia Story," 

American J. Public H*Jalh 59, 760-764 (1969). 

42. THE Pa~EsDEArT' COMMUIIBS10 O. HelItiI ';..fi- Wl , CASCIER AND TfRtOgiE, "RtelOrt to 

f 1

- 180 - 

Larry J. Shumon, Horvy.y Wolfe, ond R. Dixon Speos 

the lPresident-A National Program to Conquer lleart Disease, Cancer and Stroke," 

Feblruary, 1965. 

43. PuILIC IIh:.ALTII IF}:EIL.TIOX OF TllR CISCIXXATI Alt:A, 40th Annual nReport 1957, 

Cinciinati: The Fclieration, 1957. 

44. L. S. RElE ^N) 1i. IIOLLINSGSoWOTI, "IIOw MIany General llos.pital Beds are Needed? 

A lReappraisal of lesd Nee(ds ii Relationshipo to l opulatioi ," 1>11S-309, 1!I:\W, 195.3. 

45. C. RtV:VtLL:, F. FF:l,>MA. , AXs ) W. LYN.N, "An Optimuization lModel of Tuberculosis 

IEl'didi,¡iioloiy," Ml< 'iuu'únt &c1:i., ApIlilietion .,'.ri.es, 16, 11-190.11211 (1969). 

46. - ANs 

(1970). 

t. SWAIN, "Central FIacility location," G'ograplhical Anal. 2, 30-12 

47. R. SASULY ASD 1>. WAnD, "Two Approaches to llealthi 

Progmnalie," Medical Care 7, 235-241 (1969). 

Planning, The Ideal vs. the- 

48. J. ScisiX:i)ita, "A New Approach to I>l:uning thie Growti of a Metrolpolitanl lospital 

System," Regional Scienec Refearch Institute, l>hiladelpilia, IPe>i'sylvallin, lC;5. 

49. , "Aleasuring the Locatioial Efliciency of the Urban f-lospil ,!," lilrcli 

Srenvie Res. 2, i'4-169 (1967). 

50. A. J. ScoTr, "LOCawnll-AIOleatio,, Systems: A Review," Gcographical Anal. 2, 95- 

119 ¡1970). 

51. SEIoW\ICK COUNTY IIEALTli FACILITITS 1>LANh.,:.C COUSNCil SURV:V AND IR:COMMEXI¡iArioxs, 

Sedgvwick Couatmy llealth Facility Planning Couincil; Kaimswas lealth Facilities 

Informatioi Serviee, PIublication No. 494, Topcka, IKansas. 

52. L. J. SIIUMAN, '"athemaltiWal Models for Ilealtlh MIanpow>er ilamning," Department 

53. 

of Op'rationis Research, The Johlis llopkins University, Ballimiore, Maryland, 196)9. 

-- , J. P. Youso, AND E. NADnOR, "A Planning MIodel for Regional lHealth Services," 

lIcalth S~ncea Res. 6, 103-119 (1971). 

54. - , C. P. IIARDWICK, ASD G. HUBER, "A Miodel for the Allocation of Ambulatory 

lealtlh C-are Centers in a Afctropolitan Area," Icallth Scices Ruc. 8, 121-13S (1973). 

55. S. SumvEItTS, "linlluences of Arncawidc Pilaming," l/ospitalo 44, 63-65 (1970). 

56. R. D. SJUALLwooo, E. J. SOSDIX, AND F. L. OFExSS xin, "Towards Rn intcg-atcd 

57. - 

Methodology for the Analysis of Health Care Systems," Opns. Res. 19, 1300 1322 

(1971). 

, "A Quantitative Appronach to Analyzing Regional licalth Care Systems," 

paper presented at 39th National Mleeting of the OP0RATIOxS RESsEAnCII SOCIETY OF 

,AmeRicA, Dallas, Texas, 1971. 

58. J. D. SSYDER AND lM. J. ENRIOHT, "Federal Aid," Hospital '.:anagemcn 102, 37-56 

(1966). 

59. A. R. SoaEins, "Hospitais Costs and lPayment: Suggestíonsi 

Balance," Medical Care 7, 348-360 (1969). 

for Stabilizing the Uneasy 

60. D. II. STIaiso. AND R. H. STIMsoN, "Operations Research and Systems Analysis in 

61 

Hlospital Administration," Chicago: Hospital Research and Educational Trust, 1972. 

C. ToiIEGAS, R. SWAIN, AND C. ,REVELLE, "Thie Location of Emergency Service Facilities," 

Opns. Res. 19, 1363-1373 (1971). 

62. G. WV. TOIR^xACE, W. 11. THoMAS, AND 1). L. SACT-rr, "A Generalized Cost-Effective- 

:ness Model for Evaluation of Health Programs," paper presented at 39th National 

Meetingi of tile OPERATIOXS RF.8,ACH SocIrTY Or AUr.RICA, Dallas, Texas, 1971. 

63. J. vo.N NEUMANN ANO O. MOICEoS.rF:tSN, 'I'lUorg of Ga|es 

Princeton University Press, Princeton, 1947. 

und EcoCnomic chdaniior, 

64. J. 1P. Youso, "No Easy Solutions," GEonGE K. Caiícso (cd.), "The Recognition of 

Systems in Health Services," paper presented at the 36th National ;Mecting of the 

OPErIATIOXS RI:S:EARCII SOCI:'TY Or ALERICA, ArlingtoIn, Virgi;i;a, 1969. 

65. H1. I\OLFE, "A MultilolC Assignment Model for Stafling thie Nuraing Units," Dcpartment 

of Oeratiois Rllescarch, Johus llopkins University, Baltimore, Marylanid, 1964.

- .,1 - 

Programming, Budgeting, and Control 

in Health Care Organizations: 

The State of the Art 

RPc, rt A. Vraciu 

The planning, budgeting. and controlling processes (PBCP) largely subsume all of the 

planning and controlling activities of an organization. This paperdiscusses these activities 

within the context of a single management control system, focusing on three topics. 

First, a brief historical perspective of management concerns which relate to PBCP is presented 

and several important external pressures currently imposed on the health care 

industry are discussed. Second, normative models of the processes-progri'mming, budgeting, 

and controlling-are presented. The discussion focuses on the elements and relationships 

of these processes, and numerous references to the literature are provided. Third, 

several issues related to the gap between the state of the art in PBCP for hospitals and the 

current state of practice are discussed. 

Much of the management control 

process involves informal communication 

and interactions. lni,. .uí 

communication occurs by means 

of memoranda, meetings, conversations, 

and even by such signals'as 

facial expressions. Although 

these informal activities are of 

great importance, ... most companies 

also have a formal management 

control system. [1, p. 191 

I OGaamoNG, budgeting, and controlling 

processes (PBCP) represent 

the three major phases of such a formal 

management control system. A simple 

schematic diagram (Figure 1) illustrates 

the relationships between these activities 

as they occur in a regular cycle and 

the role of strategic planning as a 

starting point. 

"Programming is the process of deciding 

on the programs that the company 

will undertake and the appropri- 

ate amount of resources that are to be 

allocated to each program" [2, p. 670]. 

' is aspec! of the planning process 

lcludes the formulaticn ,f corporate 

objectives and strategies based on marketing, 

legal, regulatory, and social concerms, 

as well as on the organization's 

own perception of its role. During this 

stage, health care organizations would 

identify the types of medical services to 

be provided, the types of teaching and 

research activities to be conducted, the 

service populations to be served, and 

the general mode of operations for both 

a long-run and a short-run planning 

horizon. 

The second phase is the budgeting 

process which expresses the programming 

decisions in monetary terms, and 

covers a specific time period, generally 

one year. The budget represents the 

best plan for allocating resources to 

achieve the objectives and implement 

Address communications and requests for reprints to Robert A. Vraciu. Assistant Professor, 

Program in Hospital Administration, School of Public Health, University of Michigan, 

Ann Arbor, MI 48109. 

0017-9124/79/1402-0126/$2.50/0

- 182 - 

Programing, Budgeting, and Control 

ii i i i i i · i i~~~~~~~ 

Figure 1: 

Management Control Process* 

*Adepted from R.N. Anthony and J.S. Reece, 12, p. 670). 

the programs approved in the programming 

phase. The agreed-upon budget is 

often considered a bilateral commitment 

in which responsibility center 

managers commit themselves to produce 

the planned outpuf with the 

agreed-upon amount of resources, and 

their superiors commit themselves to a 

definition of satisfactory performance 

[3, p. 31). 

. The tbhird phase of the management 

control process represents a feedback 

mechanism in which plans and expectations 

for subsequent years can be 

modified based on experience from 

previous years. The prcess involves 

the reporting and analysis of performance. 

Such reports and analyses provide 

the basis for control over expenditures 

and useful data for identifying the 

causes of operating problems. 

This model of management control 

systems is applicable to any organization 

and is used as a framework for 

discussing .programming, budgeting, 

and control in health care institutions. 

Superirnpo.ed over the provider organizations 

are severa] layers of planners, 

regulators, etc., each trying to manage 

the system. As the focus of the discussion 

is on provider institutions, these 

interests are viewed as constraints and 

external factors which impinge upon 

the PBCP. It is recognized that these 

constraints can be influenced by providers, 

but this type of interaction is 

not discussed here. 

The remainder of this paper presents 

a discussion and evaluation of the state 

of the art of PBCP, with a distinction 

being made between state of the art, 

(i.e., that which is technically possible) 

and the state of practice (i.e., that 

which is currently being employed in

- 183 - 

the health care industí , While much 

of the discussion is focused on the 

nonprofit, nongovernment hospital industry, 

many of the process-. -nd 

issues are relevant to other health care 

provider groups. Because the topic area 

of PBCP subsumes other 0reas of financial 

mana¿ement, and because it overlaps 

with many other areas of management 

activities not typically associated 

with financial management, discussion 

here must be narrowly focused. The 

paper presents a normative model of 

the three processes (i.e., PBCP) and 

discusses several issues related to the 

gap between the state of the art and the 

state of practice. To accomplish this, 

the paper is divided into tibe¿ parts. 

The first section provides a orief his 

torical perspective to PBCP and sur. 

marizes several important external 

pressures currently imposed on the 

health care industry. The second section 

presents a generic model of each of 

the three processes (PBCP). The elements 

of the models and the relationships 

between their elements are highlighted. 

The final section discusses 

several issues associated with orientation, 

design and implementation of 

PBCP currently practiced in the hospital 

field. 

Background-Legacy and 

Inertia 

As the hospital industry developed 

from a cottage industry to being the 

nucleus of a technologically sophisticated 

health care system in the early 

1970s, it did so with few constraints. 

Each institution had cash-flow concerns 

like any other economic entity; 

however, major programming decisions 

were made largely independent of any 

external planning or regulatory bodies. 

By 1970, the hospital industry was 

undergoing rapid growth in the hope of 

fulfilling the promise that "high quality 

Health Services Research 

health care was d -:;ht." The inertia 

generated by this period is still influencing 

managerial decisions and has 

left the industry with a legacy of high 

and rapidly increasing costs. Enthoven 

[4, p. 12291 summarizes the management 

philosophies that have resulted: 

In the system of fee-for-service, 

cost reimbursement and third. 

party intermediaries that dominate 

health care fiaancini today, the 

question of efficient use of resources 

does not even arise. The 

problem of how best to spend a 

given amount of money for the 

health care of c pupulation is not 

posed. Providers are not required 

to set priorities, look at alternatives 

and make hard choices. From the 

point of view of the provider, there 

is an apparently unlinrited amount 

of money. The system rewards 

cost-increasing behavior with more 

revenue; it punishes cost-reducing 

behavior with less revenue. Such a 

system must produce inflation in 

prices and waste in the use of 

resources. 

To better understand the elements of 

this philosophy as they relate to PBCP, 

four general areas of planning concern 

are discussed: service mix, production 

efficiency, production effectiveness, 

and capacity. 

Service Mix. The miy of services 

available in a given hospital was influenced 

by the rapid technological 

change in the medical field [5], the high 

degree of medical specialization [61, 

and the ability of hospitals to rapidly 

adopt new technologies. The rapid 

technological change was stimulated 

by research funded through the federal 

government, as well as universities, 

foundations, and industry. The resultant 

new clinical technology required 

expensive and highly specialized hospital 

facilities.. The clinical efficacy of 

these technologies were often not ques-

fioned or tested 1[7 and the cost-based 

reimbursement adopted by Medicare 

and Medicaid programs," .. created 

an unprecedented opportunity for physicians 

and hospital administrators to 

do what they always wanted to d'.improve 

the quality of care as they see 

it. This means more equipment, more 

personnel, more tests, more x-rays, and 

so on" [8, p. 94]. This natural tendency 

of health care providers and administrators 

received both moral and fiscal 

support from the federally financed 

Regional Medical Program created in 

1965. As a result, no hospital wanted to 

be technologically inferior and sought 

to provide a wide range of services 

regardless of the impact (or lack 

thereof) on the health status of its 

service population or the resultant duplication 

of services. 

Production Efficiency. The hospital 

industry is often characterized as having 

few inherent incentives for efficiency. 

The profit motive is generally 

lacking, contributions have historically 

allowed many hospitals to operate at a 

loss, health insurance programs have 

insulated the consumer from the provider, 

and the guaranteed payment of 

cost-based reimbursement have created 

an environment in which managers 

have had little need or incentive to 

strive for efficient production of services 

[8]. In addition, cost-based reimbursement 

even provides disincentives 

to reduce costs through improved efficiency 

[9, pp. 165-1671. As a result, 

hospital management has generally 

failed to introduce efficiency-minded 

styles of management and similarly 

failed to develop information systems 

necessary for proper control. 

Production Effectiveness. Historically, 

hospitals have had little concern 

about providing the right number and 

right types of health services to their 

- 184 - 

service populations. Until recently, few 

people have questioned medical judgments 

regarding admcissions, length of 

stay, and surgery rates. As a result, the 

incentives inherent in the fee-for-service, 

litigation-prone health care system 

have dictated the modus operandi, 

and increasing evidence of excessive 

utilization is surfacing l10,11]. 

Capacity. While the medical profession 

operated on the "more is better" 

philosophy described above, hospital 

administrators developed a "bigger is 

better" philosophy toward facility 

planning. The availability of donations, 

the guaranteed reimbursemcnt of depreciation 

and interest expense by 

Medicare, Medicaid, and many Blue 

Cros: programs, and the availability of 

debt financing provided hospitals with 

sufficient means to add to their bed 

complement and replace existing facilities. 

Even though regional planning 

was an intended feature of the iiill- 

Burton Program, the Regional Medical 

Program, and later the Comprehensive 

Health Planning Act, large numbers of 

unused hospital beds are found in mosi 

regions of the couiry. rhe depreciation, 

interest, and ope:ating expenses 

associated with these facilities and subsequent 

replacement costs are all considered 

legitimate expenses and reimbulrsed 

ur.der cost-based reimbursement. 

As a .result, there have been few 

financial constraints or penalties 

linked to excessive capacity. 

While many other forces and activities 

were at play prior to 1970, those 

cited above were instrumental in shaping 

the hospital industry as it is currently 

operating. Programming decisions 

were made based on the "more is 

better" and "bigger is better" philosophies 

adopted by the medical staff, 

administration, and trustees. There was 

apparently little concern for the efficient 

allocation of resources since there

- 185 - 

was apparently no scarcity. The net 

result is the often cited increase in 

hospital costs. 

Given the trends toward raic,.y increasing 

costs of health care, the early 

1970s saw the federal government, state 

governments, and private industry exert 

pressure on the hospital industry. 

These major purchasers of health care 

services began to recognize that they 

would soon be unable to afford the 

health care system which was evolving. 

The attempts to control systems costs 

include rate regulation [121, controls on 

capital investments, budgeting requirements, 

and utilization review. The associated 

regulations act es constraints 

to the PBCP. 

Attempts to control capital invr - 

ments include the early comprehensi, ¡ 

health planning structure, state certificate 

of need requirements, and certificate 

of need review mandated by P.L. 

93-641. Under P.L. 93-641, states are 

required to designate an agency which 

performs a reviewing and comment 

process for capital expenditures which 

exceed $150,000, change the bed supply, 

or substantially change the services 

of the facility. P.L. 93-641 also 

authorizes health systei;.s agencies to 

recommend to this state agency 

whether new institutional health services 

should be improved or not. While 

the experience of such programs is 

mixed [13,14], the process of requiring 

hospitals to justify proposed expenditures 

does force a level of planning not 

previously observed. Moreover, the 

federal government and some states 

have made inroads in defining and 

implementing quantitative criteria 

based on conservative estimates of justifiable 

demands 115,16]. 

Annual operating budgets and capital 

budgets with a three-year planning 

horizon are mandated of all hospitals 

under Section 234 of P.L. 92-603 [17, 

p. 369]. While the law can make this 


requirement of a piucess, it cannot 

insure that the process is effective in 

the context of the management control 

process outlined in Figure 1. 

Professional standards review is 

another effort to affect costs by monitciing 

the appropriate use of hospital 

services. "But since the mechanism is 

that of peer review-one physician 

vis-h-vis another-the conflict of quality 

versus cost is immediately established, 

and with the emphasis on the maximizing 

of care, the net result may be 

increased rather than decreased services" 

[18, p. 6). 

The four categories of regulatory programs 

represent direct attempts to lower 

the rate of increase in hospital costs by 

1) providing services which are medically 

necessary rather than simply demanded, 

2) reducing the capacity of the 

system to correspond to so.-.c -,,tifiable 

demand for medical services, and 3) 

providing incentives for production efficiency 

and effectiveness. 

The models of the programming, 

budgeting, and controlling processes 

presented in the next section adopt 

these three objectives for institutional 

planning and control. (Cnst-effective 

planning is the term used by Dowling 

to refer to "determining the least costly 

arrangement of facilities and services 

consonant with community needs" [19, 

p. 26). The two definitions are seen to 

be identical in meaning.) It should be 

noted, however, that cost-effective 

planning as defined here does not 

necessarily coincide with the private 

interests of health care providers. It is 

this inherent conflict of interests which 

spawns many of the issues discussed in 

the final section of this paper. 

Programming, Budgeting, and 

Control-A Model of Processes 

This section presents an overview of 

the state of' the art of programming,

udgeting, and controlling processes 

(PBCP) for health care institutions. 

Schematic diagrams of each of the th.,:. 

processes are presented (Figures 2, 3, 

and 5) for the purposes of highlighting 

the important data elements, relationships, 

and decision points. These diagrams 

are not intended to preclude 

additional considerations or steps leading 

toward decisions, and the ensuing 

discussions must, because of space 

constraints, be restricted to general 

characteristics of the diagram elements. 

(For more detailed discussions of methodological 

options, evaluations, etc., 

see [201.) 

Programming Phase 

In the management control model 

described above, the initial planning 

step is that of deciding on the strategic 

plans for the organization. Once decided, 

the programs to be undertaken 

and the approximate amount of resources 

to be allocated to each program 

are determined. The outputs of this 

programmng process are decisions regarding 

the revenues, expenses, and 

capital requirements of patient care, 

teaching, research, and overhead programs; 

facility size; and objectives in 

the area of improving or ensuring production 

efficiency and effectiveness. 

The programming decisions are based 

on four major categories of informetion 

(see Figure 2). 

Marketing Information. The market 

served by the hospital should be understood 

well enough to permit costeffective 

planning, i.e., to determine 

the least costly arrangement of facilities 

and services, consonant with community 

needs. This should include three 

important types of information: 1) determination 

of service population of the 

hospital and its characteristics 121], 2) 

assessment of medical needs of the 

service population [221, and 3) assess- 

- 186 - 


ment of potential areas for cooperation 

with other hospitals and providers [23]. 

It is recognized that difficulties arise in 

trying to define and measure need fnr 

medical services [24]. As a result, many 

of the contemporary planning models 

and the marketing information are demand 

based. It should be noted that 

increasingly hospitals are trying to influence 

the amount and configuration 

of their demand through various marketing 

efforts. 

Organizational Goals and $.trmiegic 

Plans. The developmeni, statement, 

and updating of organizatiorlal goals 

and strategic plans is the second major 

cf-t.gory of information necessary for 

;h- programming process. These goals 

initially provide direction for the decision 

process and, later, become yardsticks 

along which various program;.3 

can be evaluated. It is important that 

the formulation of these goals reflect 

- inputs from a variety of individuals 

within the organization, e.g., trustees, 

administrators, depar,..en* heads, etc. 

[17]. Such an exchange of ideas between 

individuals at the different levels 

not only provides useful suggestions 

in the planning process, but helps 

foster a córr-itment to the organization 

arid its plans [25]. 

Previous Year's Performance. The 

feedback loop in the general management 

control process can suggest areas 

whete adjustments are necessary and 

areas where expectations should be 

changed. The primary use of this information 

is the .evaluation of ongoing 

programs [3, Chap. 91, and the data is 

generated during the Controlling Phase 

which is discussed below. 

Externa) Constraints. The hospital 

industry operates within a larger legal 

and economic system and is regulated 

by a host of statutes and agencies.

ut, 

u 

a. 

íL 

._c 

E 

0 

t42 

a .. 

0m 

u 

y 

E 

o, 

- 187 - 

Health Services Research

- 188 - 

Consequently, hospital programming 

decisions must be made with a full understanding 

of the reimbursemnt restrictions, 

legal constraints, tax culnsiderations, 

and certificate of L.eed restrictions 

associated with each program. 

Information from these four categories 

are necessary for hospital managers 

to make cost-effective programming decisions. 

A purely algorithmic approach 

to programming decisions is not available, 

and it is doubtful whether one 

could be of much benefit given the 

complex nature of organizational decision 

making. Such decisions generally 

result from consideration of a large 

number of political, competitive, social, 

economic, and persone i cencerns. 

The application of algorithms would 

undoubtedly be plagued with the same 

types of conceptual and measurement 

problems identified with cost-benefit 

analysis [261. 

Nevertheless, useful decision models 

have been developed for several types 

of programming decisions that are conceptually 

consistent with the objective 

of cost-effective decision making. 

Where they can differ, and where application 

is weakened, is ofen in the 

specification of key parameters. A 

number of these models are cited 

below. 

Facility Requirements. Dowling [271 

identifies five steps for translating demand 

forecasts into facility requirements. 

1. Identify the facility units (e.g., medical/surgical 

beds, obstetrical beds, 

operating rooms, x-ray units, etc.) 

associated with each major medical 

service (e.g., deliveries, surgical 

procedures, etc.). 

2. Determine the service producing capacity 

of a single facility unit. 

3. Determine the number of facility 

units required to accomodate the 

average demand for service. 


4. Determine the nunuer of additional 

facility units that must be available 

to accommodate emergency demands 

or peak loads based on the 

desired level of protection sought 

(the reserve or standby requirement). 

5. Calculate the total number of facility 

units required and the load factor 

(e.g., occupancy level) from the results 

of Steps 3 and 4. 

The calculations associated with each 

step are necessary to match Lise capacity 

of a facility unit to it.s expected 

demand in an efficient iwai. The demand 

estimates are derived from service 

population estima t cs 121,28). Reductions 

in use rates based on normative 

standards can be made to demand 

estimates (use rates for fixed service 4 

population) to provide activity levels 

that are closer to need than the observed 

demand [16]. The validity of 

such a reduction is based on an as- 

.sumption of excessive utilization due 

to excessive supply and inefficient 

management of medical services 16, pp. 

96-100; 10,16,29]. )nce the demand is 

characterized, four gei.aral approaches 

can be used to determine the total 

facility requirements [27]: 

1. Use o doormative occupancy levels 

(e.g., build and/or staff enough medical/surgical 

beds to provide an 85 

percent occupancy level). 

2. Analyze the past census data to 

identify the number of beds neces- 

,sary to accommodate the expected 

census, x percent of the time (e.g., 95 

percent of the time). 

3. Apply a formula based on an assumption 

of a simple (generally 

Poisson) distribution of arrivals. 

Such models range from the simple 

square root model used by the Hill- 

Burton Program [30] to more complex 

mathematical formulations (see 

analysis and references in 1311). 

4. Simulation models for more com-

plex stochastic distribution of arrivals 

and census data 132, Chap. 7]. 

These models should be quite useful 

for determining target bed sizes of 

hospitals. When applied to institutional 

planning, the need for cooperation 

between health facilities is apparent. 

The technology has not been developed 

to the same degree for sizing 

ancillary departments. (Useful studies 

include those by Thomas and Stokes 

[331], and Conrad et al. 1341.) 

Support Services. All organizations 

incur systems' costs for "overhead programs." 

The mix, scope, and level of 

such programs must be addressed in 

the programming phase. Neumann (351 

discusses why such programs often go 

unevaluated and how organizations 

have a natural tendency to foster their 

growth. Overhead value analysis [351 

and functional value analysis [361 are 

pragmatic methods for evaluating the 

worth of such overhead programs tosthe 

organization. 

For almost every overhead service, 

there is an implicit "make versus buy" 

decision. Increasingly, hospitals are 

buying many services (e.g., management 

contracts, food services, data processing, 

etc.) and increasingly, hospitals 

are looking to shared service arrangements 

with other hospitals for fiscal 

services, purchasing, laundry, etc. 

Programs to Improve Production Efficiency 

and Effectiveness. Significant 

cost savings can be achieved by implementing 

contemporary' admissions 

scheduling systems, preadmission testing, 

concurrent review, and outpatient 

surgery [371. Successful implementation 

of these methodologies requires 

considerable advanced planning/programming 

to develop the requisite data 

systems [38] and to develop the necessary 

support from critical actors in the 

system [39]. The involvement in this 

- 189 - 


planning process of all affected parties 

and the "top-down" support is essential 

to the successful implementation of 

these efficiency methodologies. 

The importance of these methodologies 

in the efficient management of 

patient services should be noted. A 

hospital which for example does not 

have a preadmission testing program 

and outpatient surgery will experience 

a higher inpatient census than if the 

hospital had the programs. Using the 

approaches described above, this 

higher demand will be translated into 

more hospital beds. Thus, for the same 

medical services, a larger inpatient facility 

would be justified. In addition, if 

'admission and operating room scheduling 

systems were not used, the variance 

of admissions would likely be higher, 

thus justifying more beds for the same 

target occupancy. In both cases, implementation 

of these efficiency technologies 

will tend to reduce the number 

of necessary beds for the service 

population. If the facility resizes accordingly, 

it will probably increase 

outpatient expenses, but there should 

be net savings. 

It is through the programming decisions 

identified above and in Figure 2 

that hospitals can plan in advance how 

to allocate resources and provide facilities 

and services which meet the needs 

of the population served, while minimizing 

capital and operating costs. 

These decisions provide the context 

within which budgeting and control 

decisions are made and, to a large 

extent, set the stage for a hospital's 

level of production efficiency and effectiveness. 

The major cost-related decisions 

are program decisions; budgeting 

and control can be viewed as fine 

tuning. 

Budgeting Phase 

The programming decisions provide 

the fiscal officer alid budget committee

with a programmatic statemer.t of organizational 

objectives. Du.iig the 

budget process, the approved programs 

are translated into a detailed statement 

of monetary requirements and financi_; 

consequences, i.e., the budget package 

is prepared. 

The process of budgeting is useful to 

management because it formalizes 

communication between the hospital's 

governance, administration, department 

heads, and others. This dialogue 

serves a useful plannng function and 

can be extremely important if the budget 

is to be used effectively in the 

control process. "Management by objectives" 

is a formal program to facilitate 

and improve the effectiveness of 

this process [40]. 

The end result of the budgeting process, 

the budget package, can serve 

three major functions. First, it describes 

the implementation and consequences 

of the programming decisions in the 

preceding step. Thus, it may be viewed 

and used as a simulator of financial 

consequences of programming decisions. 

Second, the approved budget is 

used as a standard in the control process. 

Thus, the measures of performance 

(often costs) must be stated in 

terms of responsibility (discussed below). 

The budget's third function is in 

the area of reimbursement. In some 

states, rate setting programs of third 

party participating contracts require 

hospitals to negotiate an expense budget. 

Subsequent reimbursement is then 

based on the results of this negotiation 

process. When a budget is used in a rate 

setting context, it may reflect ploys to 

increase reimbursement rather than accurately 

reflecting management intentions. 

If this is the case, te budget 

would have questionable value in the 

controlling phase. 

There are several ways of structuring 

the budget schedules depending upon 

management needs. Four dimensions 

- 190 - 

to this structure are identified and 

discussed below. 

Basis of Accumulating Costs. The 

expense budget is a key operating budget 

and the basis of accumulating costs 

represents the basic unit for planning 

and control. Management faces a 

choice of bases, and four commonly 

cited options are: 

1. Line items. A budget developed on 

this basis would show expenses 

grouped by major items of expenditures 

for the entire organization, e.g., 

salaries broken into several categories, 

medical supplies, laboratory 

supplies, etc. With expelises accumulated 

this way, management can 

control items of expenditure and 

plan for their use. There is, however, 

no clear link to output and such a 

classification fails to identify responsibility 

for expenditures. 

- :. Program. A budget prepared on this 

basis would augment the programs 

identified in the programming pr,h.sby 

showing more 'l.taile:l budgets 

for each program. Using such a base, 

organizations must decide whether 

they are planning for or intend to 

control only the direct program costs 

or the full program costs (direct plus 

indirect). in either case, this method 

of preparation .is useful in planning, 

but often fails to accumulate costs 

along lines of responsibility. Thus, 

the control function is subverted. 

3, Medical Services. A budget prepared 

using this basis identifies the costs 

associated with each category of 

medical services. Costs can be defined 

in total, on a per-unit basis, 

and by direct and indirect costs. 

Bundles of medical services associated 

with specific diagnoses are 

the base proposed by Thompson 

(411. 

4. Responsibility Center. Costs are ac-

- 191 

cumulated according to any organizational 

unit headed by a responsible 

manager. In many cases these 

responsibility centers correspond to 

departments, although not necessarily. 

Budgets prepare¿ using this 

basis are quite useful for controlling 

expenditures, but often lose their 

direct usefulness in planning based 

upon medical needs of a service 

population. In addition, the control 

function can be hampered in cases 

where areas of responsibility overlap. 

Fixed versus Flexible Budgets. A 

fixed budget is one developed using a 

single estimate of activities. The usefulness 

of such a budget is contingent 

upon the accuracy of the activity forecasts. 

A flexible budget is essentially a 

set of fixed budgets covering a specified 

range of activity levels [17,42,43]. 

Time Periods. The expense foresast 

can correspond to the budget year, in 

total, each of the four quarters, each of 

the 12 months, or 13 four-week periods. 

The smaller the unit of time, the more 

tedious the planning process. However, 

the ability to use variable budgeting 

(i.e., coordinating staffing levels with 

seasonal variations in demand) can 

only be accomplished efficiently with 

reasonably small units of time. In addition, 

the specified period of time is 

directly related to the time lag between 

the occurrence of unfavorable performance 

and its detection using budget 

reports. 

Continuous or Discrete. A budget can 

be prepared on a discrete basis, i.e., a 

fixed 12-month period, or on a continuous 

basis, i.e., the budget is updated on 

a monthly or quarterly basis, always 

extending 12 months in advance. 

Management can select any combination 

of these different types of budgets, 


each with their own strengths and 

weaknesses. For our purposes, it is 

assumed that the budget is being prepared 

for a 13-period year [44] on a 

discrete basis, accumulating costs by 

responsibility center [21 and on a fixed 

basis. (The benefits of flexible budgets 

in the planning process will depend on 

the uncertainty of activity levels. In 

cases of high uncertainty, a flexible 

budget should be used to specify the 

range of options and consequences. 

However, at some point, a "likely" 

activity level will have to be assumed. 

For control purposes, other means can 

be used to analyze performance.) Having 

made these assumptions, we can 

discuss the three principal steps in the 

budgeting phase (see Figure 3). 

Step 1 generates forecasts of the 

activity levels for each responsibility 

center, for each of the 13 periods. In 

some cases, mathematical models employing 

the historical data (time series, 

leading indicators, causal factors, and/ 

or some combination) can be quite 

useful for predicting activity levels [32. 

45-471. In other cases, historical data 

has little bearing on the future. For 

example, when a new medical service 

is implemented, there is no experience 

on which to base forecasts. In all cases, 

there will be some risk associated with 

each forecast. Any biases associated 

with the forecasts of activity levels 

should be identified to the extent possible, 

since forecasts of expenses and 

revenues will generally reflect the same 

bias. In practice, the forecasting of 

activity levels is seen as a combined 

quantitative and qualitative approach. 

That is, historical data and mathematical 

models can be useful for identifying 

trends and underlying relationships; 

however, department heads' understanding 

of exogenous and endogenous 

factors which bear upon the future 

must be subjeciively weighed in the 

forecasts.

1' 

- 193 - 

Step 2 represents the heart of an 

efficient budget preparation process. In 

this step, expected levels of output are 

translated into resource requirements, 

i.e., personnel at various levels, supplies, 

etc. The intent is to match an 

efficient level of inputs with the expected 

level of output, given a desired 

level of quality. For two major areas of. 

requirements-manpower 'and supplies-major 

elements of the process 

can be highlighted. 

Manpower Requirements. Salary expenses 

represent 55 to 70 percent of a 

hospital's total expenses. Thus manpower 

budgets represent a major area of 

budgetary concern. Warner (481 discusses 

the current research on nurse 

staffing, scheduling and reallocation 

activities. McNally 1491 evaluates six 

techniques for evaluating manpower 

levels/needs for many hospital departments. 

Kaplan [50] uses regression 

analysis to staff different nursing units. 

These formula-based approaches tó 

staffing decisions can be effective if 

there is significant department head 

input into staffing standards [51]. Lipson 

and Hensel [52] describe a manpower 

budget process which is based 

on negotiation between responsibility 

center heads and administration. Such 

participation in which department 

heads not only formulate but commit 

themselves to performance levels is 

considered important for an effective 

control system [53]. Through practice, 

careful feedback and commitment, activity 

indexes can be developed for 

specific nursing units, and production 

functions developed for ancillary departments 

which represent efficient 

and agreed upon technologies. 

Supply Requirements. Supplies vary 

from approximately 5 percent of nursing 

unit expenses to 75 percent for 

pharmacy. For many departments, the 


amount of supplies used can be correlated 

to some level of hospital output, 

e.g., patient-days, surgical procedures, 

etc. Estimated relationships (with an 

adjustment for inflation) can be used to 

forecast the amount of supplies for each 

department. An effective materials 

management program can reduce supply 

costs, including distribution and 

inventory costs [541. 

The important elements of this resource 

identification step are to 1) 

attempt to develop an input-output 

model for each department and 2) 

allow department heads to participate 

in developing the model to be used for 

their department. The models can be 

developed using historical data 1551 or 

can be constructed using subjective 

estimates of the parameters [561. Caution 

must be exercised that past inefficiencies 

are not perpetuated, but the 

actual process of trying to understand 

the use of inputs for different levels of 

output will have educational benefits 

to administrators and department 

heads. By making current patterns of 

resource use explicit, the status quo can 

be challenged and better ways sought. 

(Zero-based budgeting is a systematized 

approach to categorizing and 

judging current expenditures l571.) The 

field testing of these better ways and 

feedback through the budget reporting 

system provide the potential for period 

by period improvements in performance. 

The operating and financial budgets 

are prepared in Step 3. It is important 

that the full consequences of the hospital's 

set of plans are identified. Following 

the completion of Step 2, financial 

officers must predict or set salary levels, 

fringe benefits, supply prices, 

utility expenses, interdepartment 

charges, etc., in order to complete the 

expense budget. With the expense budget 

completed, revenues from costbased 

reimbursers can be estimated,

and charges set at a level to co. r total 

financial requirements [58]. Total financial 

requirements are the sum of 

operating expenses (patient care, teack 

ing, research, etc.) and capital needs 

(working capital, purchases of plant 

and equipment, etc.). The setting of 

charges involves decisions about crosssubsidization 

of departments and input 

from responsibility heads. Integrating 

these two budgets yields the pro forma 

income statement which can be used in 

preparing the cash budget, finalizing 

the capital budget, and preparing the 

pro forma balance sheet. 

This package of budgets should be 

viewed as a model of interrelated decisions 

where one financial decision influences 

a number of others (see Figure 

4). For example, decisions regarding 

long-term financing directly influence 

the cash budget, the expense budget, 

and the revenue budget (in particular 

the setting of charges). Understanding 

these interrelationships enables the 

budget package to be used as a simulation 

device whereby a set of programming 

decisions can be used to estimate 

activity levels, and resource requirement 

assumptions can be examined in 

terms of the financial consequences 

displayed in the budget package. Computerizing 

such a model can enhance 

its usefulness 159). 

Controlling Phase 

The concern in this section is with 

the control of performance at the responsibility 

center level rather than 

the more general management control 

process shown in Figure 1. The focus 

is on evaluation of operational performance 

where the results ,! the planning 

phases are considered "givens" 

[1, Chap. 1]. It is important that both 

financial and nonfinancial performance 

be evaluated in this retrospective 

analysis. The evaluation of financial 

performance generally begins with 

- 194 - 


an analysis of operating .rd financial 

budgets relative to actual performance. 

Typically, this so-called Budgetary 

Control process stops with the financial 

review. However, for nonprofit 

health care organizations, nonfinancial 

objectives are often key elements of 

strategic plans and programming decisions. 

Consequently, a valid analysis 

of performance miust include an evaluation 

of nonfinancial characteristics 

of operations. 

Management must develop the structure 

and processes of its control systems, 

taking into consideration the ability 

to detect deviant behavior as well as 

the systems' effect on emplcyee behavior. 

The detection concerns are disc.ssed 

below. The behavioral concerns 

inciude the following: 1) Does the 

control system induce behavioral patterns 

that are conducive to achieving 

organizational objectives? 2)What unintended 

behavioral effects does the 

control system have? and 3) What latitude 

does the system provide for 

people to implement ploys? These behavioral 

concerns have been explored 

in the literature 19, p. 75; 25,53,60]. 

These studies, and others, emphasize 

the importance of such elements as 

participaticn, Pccurate information systems, 

communication, and incentives. 

This last consideration is receiving 

considerable attention in the hospital 

industry and a number of hospitals 

have explored profit sharing or incentive 

payment programs with their employees 

161-63). I 

The major elements of the controlling 

phase are summarized in Figure 5. At 

the heart of this phase is an analysis of 

performance in which expected outcomes 

are compared with actual outcomes. 

For the operation and financial 

aspects of the organization, the operating 

and financial budgets generally provide 

the expected outcomes. They represent 

the best plans resulting from 

1

ÍA 

en .aa 

1 

co a 

cg 

.wb 

u 

iz 1 

PoIn 

00 

2;2 

E 

ce 

- 196 - 

,

- 197 - 

analysis, forecasting, and negotiation 

activities in the programming and budgeting 

phases. For the nonfinancial objectives, 

explicit statements of expectations 

should be available from the planning 

phase. 

Information about actual outcomes 

should be available from the organization's 

management information system. 

The data should be accurate, provided 

on a timely basis, and must be comparable 

with data from the budgeting 

system. This generally means that the 

hospital's financial accounting system 

must have the ability to 1) accumulate 

expense items by responsibility center, 

2) provide cost accounting systems for 

calculating per-únit costs, and 3) collect 

operating statistics as an integral 

part of the systematic management information 

system. Davis and Freeman 

(641 offer some suggestions on how to 

evaluate management information systems, 

and DATAPRO [651 specifies 

some guidelines for judging commercial 

software packages. 

The assessment of performance 

should be as timely as data permits. 

The first cut at the analysis should be 

standardized management reports of 

four types: 

1. Productivity Monitoring. Reports in 

which the productivity (output + 

input) of responsibility centers and/ 

or individual providers is measured. 

These reports compare actual measures 

with internal and/or external 

standards [17, Chap. 18; 66-681. 

2. Variance Analysis. Methodologies 

for estimating the difference between 

budgeted and actual accounts. 

Such analyses can be performed at 

the total or gross variance level for 

specific responsibility center or line 

items [691 or at the component variance 

levels [9, pp. 82-85; 701. Component 

variance analysis attempts to 

partition the total experience into 


parts which correspond to identifiable 

causes, e.g., price, volume, supply 

quantity, etc. 

3. Operational Audit. Process of examining 

the operations of departments 

and organizational control systems 

to assess their effectiveness. The 

process is concerned with problem 

detection rather than problem solving 

[711. 

4. Financial Statement Analysis. 

Methodologies for analyzing the organization's 

financial performance 

as a whole and assessing its financial 

position 172,73]. 

The purpose óf this first cut analysis is 

to systematically identify problems 

and, to the extent possible, their apparent 

causes. When problems are indicated, 

further analysis and discussions 

with responsibility managers are generally 

necessary to identify the true 

causes and the appropriate courses of 

action. 

The reporting and analysis of performance 

serve several purposes. First, the 

reports are used to keep management 

informed of what is happening in the 

different responsibility centers. Management 

can, when necessary, take appropriate 

action on the basis of these 

reports, or can initiate additional analyses. 

Second, they can be used to evaluate 

the performance of responsibility 

center heads. Third, such reports may 

indicate that the budgeting process 

should be altered. Fourth, the reports 

can be useful in educating department 

heads to the financial consequences of 

their decisions. Without such reports, 

department heads may be unable to 

make staffing and scheduling decisions 

which are consistent with the overall 

goals of efficiency and effectiveness. 

The controlling phase should be 

viewed as a feedback loop to the 

programming and budgeting phase. 

Analyzing performance will have little

- 198 - 

useful effect if it is done aF an independent 

management acu .;;y. However, 

if conducted for the purposes of 

improving the planning and budgeting 

processes, gaining a better underst ,,.'. 

ing of production processes, and improving 

management skills, performance 

evaluation can be an essential 

link in a hospital's attempt to improve 

its cost effectiveness. 

State of the Art-Issues and 

Failings 

The models presented in the preceding 

section outline the scope and elements 

of contemporary planning, budgeting, 

and controlling processes 

(PBCP). Their focus is on ccrurate 

cost-effective planning and effective 

management control. This section presents 

a number of issues associated 

with the orientation, design, and implementation 

of PBCP in the health 

field by highlighting the major reasons 

for observed differences between the 

state of practice and the state of the art. 

'Planning Attitudes 

There remains a significant gap between 

the public interest and the private 

interest of health care providers. 

The public interest was characterized 

in the first section as being served by 

cost-effective programming decisions, 

i.e., based on the health care needs of a 

service population and using criteria 

which maximize system efficiency and 

effectiveness. The private interest is, 

for the most part, still geared toward 

increasing technological capabilities, 

facility size, and the volume of patients, 

and not toward redur.ing the 

levels of expenditures. Exceptions do 

exist and, in some cases, this growth is 

consistent with development of rational 

networks and multiple hospital 

.systems. In many cases, however, this 

growth in assets and operating ex- 

penses is motivated by fear of becoming 

technologically obsoietc. fear of being 

labeled inferior to other hospitals, and 

fear of losing medical staff, patients, 

and status. 

As yet, the external constraints are 

still too loose to prevent hospitals from 

reacting to these fears out of pure 

self-interest. Enthoven 14, p. 1229] 

summarizes the attitude: · 

The problem of how tc hest spend 

a given amount of money for the 

health care of a population is not 

posed. Providers are not required 

to set priorities, look at alterriativocs 

and make hard choicec 

While the debate contir.ues on how 

best tu improve institutional planning 

dttitudes (i.e.,, by increased direct ecou.omic 

regulation or restrucr.turing the 

financial incentives), efforts are being 

made to improve the means of making 

population-based decisions 174], and 

prominent students of the industry are 

raising the visibility of problems and 

methods of increasing cost-effective decision 

making, e.g., curtailmng the "fatof-the-curve 

medicine," stimulating regionalization 

of health care, introducing 

cost considerations into the physician's 

decision making process, and 

controlling the introduction of new 

technology j4]. 

It should be noted that in the current 

environment, cost-effective planning 

decisions may be impractical and ineffective 

unless there is strong community 

planning. A hospital which is very 

discriminating in the types of medical 

equipment purchased, aggressive in reducing 

unnecessary utilization, and effective 

in managing the admissions to a 

properly sized facility runs the risk of 

losing its medical staff to competing 

hospitals that are less restrictive. Such 

a danger would, however, not be as 

great if similar efforts were being made 

by all hospitals in a community.

Budgeting Attitudes 

- 199 - 

The planning attitudes extend to the 

budgeting process; in particular, there 

is no indication that hospitals are motivated 

to develop technically efficient 

budgets. Many believe this results from 

the fee-for-service, cost-based reimbursement 

system which fails to provide 

real inceitives for efficiency. 

Under this system, there is little incentive 

to move away from peak-load 

staffing and to contend with the problems 

of part-time and temporary staff; 

there is little incentive to search for 

better modes of operation which might 

lower costs; and there is little incentive 

to say "no" to requests for additional 

staff which are intended to increase the 

quality of care without regrad to effectiveness. 

The negative effects of hospitals' 

ability to attract patients are 

greatly reduced by the widespread absence 

of these incentives and the general 

lack of competitive pressures in {ieindustry. 

The net result has been a 

widespread increase in costs, quantity 

and perhaps quality without any type 

of competitive or effective regulatory 

mechanisms for forcing the system toward 

an equilibrium' position. 

There appears to be concensus that 

this system will change through either 

more restrictive means of calculating 

reimbursement rates, or a restructuring 

of financial incentives. While this future 

scenario has not altered current 

budgeting decisions, it has increased 

hospital managers' desire for improved 

budgeting capabilities. In some cases, 

hospitals are preparing for this future 

environment by increasing their expense 

base with the hope that future 

restrictions will be less painful. 

In addition, many financial managers 

have sought to develop formulistic approaches 

to budgeting, e.g., developing 

departmental cost functions, and/or 

rigid productivity indexes. While these 


are important elements in the models 

described above, there is a danger that 

mathematical models will be used to 

develop the budget without department 

head input, review or understanding. 

Such a unilateral preparation process 

will reduce the effectiveness of the 

entire management control process and 

not be in the long-run interests of the 

hospital. 

Attitudes Toward Control 

The power structure in hospitals is 

such that management often finds its 

hands tied. Many decisions geared toward 

improving efficiency or insuring 

compliance with the budget often involve 

a perceived sacrifice by physicians 

and other professionals. Confrontations 

between administrators and the 

medical staff are often one-sided in that 

administrators are easier to replace 

than medical staffs. Since efficient behavior 

is not rewarded financially, nor 

does the system as yet demand efficient 

management, the route of choice is 

often the route of minimum resistance. 

Without clear directives and support by 

hospital boards, this cannot and will 

not change. 

Information Systems 

Accurate, valid, and timely data are 

essential for the management decisions 

described in PBCP. In an abstract sense, 

there are four key relationships which 

should be geherated from the information 

systems: costs per responsibility 

center, costs per inputs, inputs per 

.output, and cost per output. The information 

system must provide this data 

for the hospital's own operations and 

should include comparable data from 

other institutions. Drebin [75, p. 881 

criticizes the capabilities of many (not 

all) hospital information systems. 

(The hospital industry] is unable to 

define its product or quantify the 

value of its product, cannot specify

its production costs or explaibi .any 

those costs are increasing at a rate 

higher than the average rate of 

costs increase in the economy.... 

- 200 - 

This current state of practice results 

more from a historical belief that managers 

did not need good information 

systems rather than from a set of 

unsolvable problems. There has simply 

been a general avoidance of implementation 

issues, largely in two 

areas: augmenting financial accounting 

systems for management purposes, and 

developing accurate cost-accounting 

systems. 

The financial accounting system of a 

hospital is frequently used to provide 

the basic financial information regarding 

operations. Because the financial 

accounting and managerial accounting 

functions are quite different [2, Chap. 

141. two major problems can arise. 

First, the data can be invalid for management 

decisions. The financial accounting 

system is based on externally 

determined standard accounting practices 

which are geared toward disclosure 

of financial information to the 

outside world. In developing these 

standard accounting practices, criteria 

are used which may not be consistent 

with management purposes, e.g., the 

criterion of objectivity often conflicts 

with the criterion of usefulness. This is 

illustrated in the accounting principles 

relating to the measurement of fixed 

assets. Financial accounting initially 

records an asset at cost (an objectively 

determined amount) and systematically 

reduces that value using depreciation 

methods (objective estimates). For 

the purposes of planning, it has been 

argued that the replacement cost of an 

asset is the more appropriate valuation 

base. The second problem arises in the 

definition of cost centers. There is 

considerable pressure for hospitals to 

adopt a uniform accounting system 


[76], and there are fears ,.a! such a 

system will define cost centers on the 

basis of function rather than responsibility 

177]. If this occurs, using the 

financial accounting system for budgeting 

and control purposes will necessarily 

interfere with the effectiveness of 

the management control system [77]. 

Attempts are being made to develop 

systems which are compatible with 

both functional and responsibility accounting 

[78]. 

Historically, there has been little 

need to perform cost accounting beyond 

what was required bv third party 

reimbursement formulas. As a result, 

definitions of output have become synonymous 

with patient days and fee 

coúes. cost allocation is equated to the 

stea. down method, and statistical 

bases for allocation are selected if they 

maximize revenue rather than if they 

accurately reflect the use of overhead 

services. As a result, most hospitals 

findthemselves unable to measure the 

cost of programs, responsibility centers 

and specific outputs, and must make 

planning and control decisions based 

on faulty information. Berman and 

Maloney [79] describe these problems 

and their consequences for the oultpatient 

departments of teaching hospitals, 

and Thompson and Cannon [801 

discuss the general benefits and elements 

of an improved cost-accounting 

system. ? 

Conclusions 

The future fiscal environment will 

provide hospital managers i with 

stronger incentives to develop costeffective 

PBCP. The elements of such 

processes have béen outlined in this 

paper and to a large extent represent a 

straightforward application of the techniques, 

activities, and philosophies 

practiced for decades in other sectors of 

the economy. The observation that

- 201 


most hospitals need to improve signifi- emerge, the state of practice should 

cantly their management control capa- approach the current state of the art. 

bilities along these lines appears to This change in management capabilihave 

resulted from a perceived lack of ties cannot occur overnight, and pruneed, 

followed by a failure to develop dent hospital managers will recognize 

and implement-not from significant the time lags associated with such 

technical barriers. As the fiscal envi- changes and begin implementation beronment 

becomes more hostile, and fore the hea!th care crisis becomes a 

as more multiple hospital systems personal problem. 

REFERENCES 

1. Anthony, R.N. and J. Dearden. Management Control Systems: Text and Cases. Homewood, 

IL: Richard D. lrwnvin, Inc., 1976. 

2. Anthony, R.N. and J.S. Reece. Management Accounting: Text and Cases, 5th ed. Homewood, 

lL: Richard D. Irwin, Inc., 1975. 

3. Anthony. R.N. and R. Herzlinger. Management Control in Non-Profit Organizations. 

Homewood, IL: Richard D. Irwin, Inc., 1975. 

4. Enthoven, A.C. Cutting costs without cutting the quality of care. New England Journal 

of Medicine 298(22):1229, June 1, 1978. 

5. Fuchs. V.R. and M.J. Kramer. Determinants of Expenditures for Physicians' Services in 

the United States, 1948-1968. New York: National Bureau of Economic Research, 

1973. 

6. Stevens, R. American Medicine aRd the Public Interest. New Haven, CT: Yale University 

Press, 1971. 

7. Cochrane, A.L. Effectiveness and Efficiency: Random Reflections on Health Services. 

London: The Nuffield Prrcincial Hospital Trust, 1972. 

8. Fuchs, V.R. Who Shall Uve? New York: Basic Books, Inc.. 1974. 

9. Silvers, J.B. and C.K. Prahalad. Financial Management of Health Core Institutions. 

Flushing. NY: Spectrum Publications, Inc., 1974. 

10. Wennberg, J.E. Testimony and statement presented at the Hearings before the House of 

Representatives, Subcommittee on Oversight and Investigations of the Committee on 

Interstate and Foreign Commerce, 94th Congress, 1st Session, July 15, 1975. In Getting 

Ready for National Health Insurance: Unnecessary Surgery, pp. 73-85. Washington, 

DC: U.S. Government Printing Office, 1975. 

11. McCarthy, E.G. Testimony presented at the Hearings before the House of Representatives, 

Subcommittee on Oversight and Investigations of the Committee on Interstate 

and Foreign Commerce, 94th Congress, 1st Session, July 17, 1975. In Getting Ready for 

National Health Insurance: Unnecessary Surgery, pp. 104-116. Washington, DC: U.S. 

Government Printing Office, 1975. 

12. Moore, R. Payments to health care institutions: State-of-the-art. Paper presented at the 

AUPHA Faculty Institute on Financial Management, Cincinnati, OH, Nov. 7-8, 1978. 

13. Salkever, D.S. and T. Bice. The impact of certificate-of-need controls on hospital investment. 

Milbank Memorial Fund Quarterly 54(2):185, 1976. 

14. Hellinger, F.J. The effect of certificate-of-need legislation on hospital investment. 

Inquiry 13(2):187, June 1976. 

15. Department of Health, Education and Welfare. National Health Planning Guidelines. 

Hearing before the U.S. Congress, Subcommittee on Health aid the Environment of the 

Committee on Interstate and Foreign Commerce, 95th Congress, 1st Session. Washington, 

DC: U.S. Government Printing Office, 1978. 

16. Griffith, J.R. and R. Chemow. Cost effective acute care facilities planning in Michigan. 

Inquiry 14(3)]229. Sept. 1977.

- 202 - 

Programing, Budgeting. and Control 

17. Berman, lI.J. and L.E. Weeks. The Financial Management of Hospitals, 3rd ed. Ann 

- Arbor: Health Administration Press, 1976. 

18. Hanft, R., 1. Raskin and M. Zubkoff. Intr:-uction. In R. Hanft, 1. Raskin and M. Zubkoff 

(eds.), Hospital Cost Contoinment: Sele4.eu Notes for Future Policy, pp. 1-30. New 

York: Prodist, 1978. 

19. Dowling, W.L. A Procedure for Rational Planning. In J.R. Griffith, W.M. Hancock and 

F.C. Munson (eds.), Cost Control in Hospitals, pp. 25-35. Ann Arbor: Health Administration 

Press, 1976. 

20. Association of University Programs in Health Administration. Financial Management 

of Health Care Organizations: A Reference Outline and Annoteted Bibliography. 

Washington, DC: Association of University Programs in Hospital Administration, 1978. 

21. Griffith, ].R. Measuring Service Areas and Forecasting Demand. In ].R. Griffith, W.M. 

Hancock and F.C. Munson, (eds.), Cost Control in Hospitals, pp. 36-39. Ann Arbor: 

Health Administration Press, 1976. 

22. MacStravic, R.E. Determining Health Needs. Ann Arbor: Health Administration Press, 

1978. 

23. Brown, M. and M. Enright (eds.) Shared services. Topics in Health Care Financing 2(4), 

Summer 1976. 

24. Donabedian, A. Aspects of Medical C"re Administration: Specifying the Requiremenls 

for Health Care. Cambridge, MA: Harvard Univ.,rsity Press, 1973. 

25. Dunbar, R.L.M. Budgeting for control. Admims",. tive Science Quorterly 16[l):98, Mar. 

1971. 

26. Klarman, H.E. Application of cost-benefit analysis to the health services and the special 

case of technologic innovation. International Journal of Health Services 4(2):325, 

Spring 1974. 

27. Dowling. W.L. Converting Demand Forecasts Into Facility Requirements. In. J.R. 

Griffith, W.M. Hancock and F.C. Munson (eds.), Cost Control in Hospitals, pp. 70-89. 

Ann Arbor: Health Administration Press, 1976.- 

28. Boardman, J.J. Utilization Data and the Planning Process. In Ann Somers (ed.), The 

Kaiser-Permanente Medical Care Program: A Symposium, pp. 61-70. New York: The 

Commonwealth Fund, 1971. 

29. Griffith, J.R., W.M. Hancock and F.C. Munson. Practical ways to contain hospital costs. 

Harvard Business Review 51(6):131, Nov.-Dec. 173. 

30. Blue Cross Association. Comprehensive Health Planning: Analytic Techniques. 

Chicago: Blue Cross Association, 1974. 

31. McClain, J.O. Bed planning using queuing theory, models of hospital occupancy: A 

sensitivity analysis. Inquiry 13(2):167, June 1976. 

32. Griffith, J.R. Quantitative Techniques for Hospital Planning and Control. Lexington, 

MA: Lexington Books, 1972, 

33. Thomas, D. and H.H. Stokes. How many beds should a hospital department serve? 

Health Services Research 11(3):241, Fall 1976. 

34. Conrad, R.B., D.A. Knee, J.M. Meade and LM. Parrish. Utilization study saved hospitals 

from needless expansion of radiology facility. Hospital Financial Management 

27(9):40, Sept. 1973. 

35. Neumann, J.L. Make overhe.d cuts that last. Harvard Business Review 53(3):116. May- 

June 1975. 

36. Bennett, J.E. and J. Krasny. Functional value analysis: A technique for reducing hospital 

overhead costs. Topics in Health Care Financing 3(4):35, Summer 1977. 

37. Magerlein, D.B., W.M. Hancock, F.W. Butler, G.W. Mallett and D.R. Young. New systems 

can mean real savings. Parts 1 and 2. Hospital Financial Management 32(4):10, 

Apr. 1978 and 32(5):18, May 1978. 

38. Hancock, W.M. Dynamics of Hospital Operational Control Systems. In ].R. Griffith, 

W.M. Hancock and F.C. Munson (eds.), Cost Control in Hospitals, pp. 129-149. Ann 

Arbor: Health Administration Press, 1976. 

,

a. 

- 203 - 


39. Munson, F.C. and W.M. Hancock. Implementation of Hospital Control Systems. In J.R. 

Griffith, W.M. Hancock and F.C. Munson (eds.), Cost Control in Hospitals, pp. kg7- 

316. Ann Arbor: Health Administration Press, 1976. 

40. Deegan, A.X. Managemetlt By Objectives for Hospitals. Germantos;-, MD: Aspen Systems 

Corporation, 1977. 

41. Thompson, J. Planning, budgeting, and controlling-One look at the future: case-mix 

cost accounting. Health Services -._: earch 14(2):111, Summer 1979. 

42. Houser, R. How to build and use a flexible budget. Hospital Financial Management 

28(8):12, Aug. 1974. 

43. Cleverley, W.O. One step fwurther-The multi-variable flexible budget system. Hospital 

Financíal Management 30(4):34, Apr. 1976. 

44. Heda, S.S. and S.H. Lipson. Dealing with a 13-period year. Hospital Financial Management 

28(12):54, Dec. 1974. 

45. Chambers, J.C., S.K. Mullick and D.D. Smith. How to choose the right forecasting 

technique. Harvard Business Review 49(4):45. July-Aug. 1971. 

46. Donlon, V.W. Statistical methods to forecast volumes of service for the revenue budget. 

Hospital Financial Management 29(4):38, 1975. 

47. Koza, R.C. Time-series data is essential to health care planning. Hospita! Financial 

Management 28(1):32, Jan. 1974. 

48. Warner, D.M. Nurse ataffing, scheduling, and reallocation in the ho.-pital. Pospital and 

Health Services Administration 21(3):77, Summer 1976. 

49. McNally, J.K. How to find manpower levels/lneeds before budgetling. Hospital Financial 

Management 30(12):34. De.. t,76d. 

50. Kaplan, R.S. Analysis and control t.f aurse staffing. Health Services Research 

10(3):278, Fall 197.5. 

51. Flynn, J.W. Budgeting salary saving with department head Input. Hospital Financial 

Management 32(6):38, June 1978. 

52. Lipson, S.H. and M.D. Hensel. Hospital Manpower Budget Preparation Manual. Ann 

Arbor: Health'Administration Press, 1975. 

53. Bucldley, A. and E. McKenna. Budgetary control and business behavior. Accounting 

and Business Research 2(6):137, Spring 1972. 

54. Holmgren, J.H. Purchasing for the Health Care Facility. Springfield. IL: ChaA:Ls sC. 

Thomas, 1975. 

55. Cleverley, W.O. Input-output analysis and the hospital budgeting r:ocess. Health Services 

Research 10(1):36, Spring 1975. 

56. Davis, R.J. Testing the linearity assumption in hospital budgeting and control. Unpublished 

Master's thesis, Program in Hospital Administraticon, The University of Michigan, 

1977. 

57. Pyhrr, P.A. Zero-based budgeting. Harvard Business Review 48(6):111, Nov.-Dec. 

1970. 

58. Lemer. R.S. and D.E. Willman. Rate setting. Topics in Health Care Financing 1(2), 

Winter 1974. 

59. Benson, F.S. and M.A. Santullano. Modeling malces budgeting more efficient. Hospitals 

50(1):93, Jan. 1, 1976. 

60. Lawler, E.E. and J.G. Rhode. Information and Control in Organizations. Pacific Palisad«s, 

CA: Goodyear Publishing Company, Inc., 1975. 

61. Groner, P.N. Employee incentives. Topics In Health Care Financing 3(3):63, Spring 

1977. 

62. Austin. C.J. Wage incentive systems: A review. Hospital Progress 51(4):36, Apr. 1970. 

63. Williarsn, F.G. And D.C. Anderson. Cost control incentive programs: Appropriate for 

non-profit? Hospital Financial Management 32(5):14. May 1978. 

64. Davis, S. and J.R Freeman. Hospital managers need management information systems. 

Health Care Management Review 1(4):65, Fall 1976.

- 204 - 

Programing, Budgeting. and Control 

65. DATAPRO. How to buy software packages. DATAPRO. DATAPRO Research Corp., Jan. 

1972. 

66. Herkimer, A.G. Improving productivity. Topics in Health Care Financing 4(2). Winter 

1977. 

67. American Hospital Association. The Management of Hospital Employee Prouucti,'ity: 

An Introductory Handbook. Chicago: American Hospital Association, 1973. 

68. Moriuchi, M., J.D. Durham and R.D. Roberts. Productivity monitoring: Fi;st step to 

effective manpower allocation. Hospital Financial Management 32(S):30, June 1978. 

69. American Hospital Association. Dudgeting Procedures for Hospitals, Chap. 9. Chicago: 

American Hospital Association, 1¿7'1. 

70. Holder, W.W. and J. Williams.: Better cost co.:tral with flexible budgets and variance 

analysis. Hospital Financial Management 30(1;.12, Jan. 1976. 

71. Flesher, D.L. Operations Auditing in Hospitals. Lexington, MA: Lexington Books, Inc., 

1976. 

72. Choate, G.M. Financial ratio analysis. Hospital Progress 55(1):49, Jan. 1974. 

73. Bernstein, L.A. Financial Statement Analysis: Theory, Application and Interpretation. 

Homewood, IL: Richard D. Irwin, Inc., 1974. 

74. Griffith, J.R. Measuring Hospital Performonce. Chicago: Blue Cross Association. 1978. 

75. Drebin, M.E. Financial information systems: Key to hospitals' survival, Hospitala 

52(12):88, June 16, 1978. 

76. Kovener, R.R. Uniform accounting, the background. Hospital Finonciol Monogement 

31(6):20, June 1977. 

77. Ingram, ).D. Uniform accounting, how it will affect hospitals. Hospital Financial Management 

31(6):10, June 1977. 

78. American Hospital Association. Chart of Accounts for :I itals. . Chicago: American 

Hospital Association, 1976. 

79. Berman, R.A. and T.W. Maloney. Are outpatient departments responsible for the fiscal 

crisis facing teaching hospitals? The Journal of Ambulatory Care Management 1(1):37, 

Jan. 1978. 

80. Thompson, G.B. and W.G. Cannon. Hospitals, like industry, must apply cost accounting 

techniques. Hospitals 52(12):129, June 16, 1978.

- 205 - 

DESIGN OF ALTERNATIVE PROVIDER 

TEAM CONFIGURATIONS: 

EXPERIENCE IN BOTH DEVELOPED 

AND DEVELOPING COUNTRIES 

by 

Dr. Lilia Duran and Dr. Arnold Reisman 

Technical Memo 497 

ABSTRACT 

Job Evaluation, a time-honored industrial engineering 

technique developed for manual labor rate-setting was used to 

design new personnel resource teams for providing anesthesiology 

services in well-known U.S. hospitals on the one hand, and for 

delivering primary care in regions of Latin America on the other. 

This paper discusses the use of this technique in designing job 

descriptions for new allied professionals, and for designing the 

curricula necessary to train same. Experience to date in 

integrating the products of these curricula into health provider 

teams in U.S. tertiary care institutions and in primary care within 

several countries of Latin America indicates the soundness of this 

approach. 

Senior Researcher, Centro Universitario de Tecnologia 

Educacional para la Salud (CEUTES), National University of Mexico, 

Mexico City. 

** 

Professor of Operations Research, Case Western Reserve 

University, Cleveland, Ohio.

- 206 - 

BACKGROUND; THE UNITED STATES 

During the last two decades or more, the United States has 

been experiencing a shortage of physician-anesthetists relative to 

the need for their service [1,2,3]. Many anesthesiology 

residencies over the years have gone unfilled as have positions 

within hospitals across the country. As late as.1976 it was 

reported that 50% of the anesthetics in the U.S. were given by 

nurse anesthetists [4]. Parenthetically, this paper will not take 

a position as to whether administration of anesthesia ought to be 

performed by physician-anesthesiologists or nursc-anesthetists, 

both are in short supply. The American Board of Anesthesiology 

has reduced the number of i~sidency training programs 

significantly, during the last five years. The programs are now 

concentrated in major medical centers. Residencies in smaller 

peripheral hospitals are few. Yet, the available residency 

positions go unfilled. 

The shortage of anesthesiologists w.ill probably be influenced 

in the future by the following: 

These figures have been questioned because there are no data 

to establish what proport.ion of these anesthetics were given by 

nurses functioning alone as opposed to those supervised by 

anesthesiologists [5].

- 207 - 

1. A change in the immigration policy for foreign medical 

graduates has led to the reduction of the number of foreign 

medical graduates available to enter this specialty. In 1972 for 

example, 58% of anesthesiologists in formal training programs were 

graduates of foreign medical schools [6]. 

2. The percentage of U. S. medical school graduates entering 

anesthesia is decreasing. 

3. The anesthesiologists who entered the specialty during 

the immediate post-World War II period - a time of the highest 

recruitment into the specialty - are now approaching retirement. 

Much has been said about physician maldistribution. 

Anesthesiologists reflect the same maldistribution. The pronounced 

migration out of the snow belt into the southern and western 

coastal areas and into the sun belt of the population in general 

is reflected also in the movement of anesthesiologists. This 

outer-migration further reinforces the problem of anesthesia 

coverage in small peripheral hospitals of the North. The small and 

peripheral community hospitals have historically experienced the 

most difficulty in acquiring anesthesiologists. Of the 7000 

hospitals in the U.S., one half have 100 or fewer beds. Of this 

latter 3500 hospitals, only 9% have anesthesiologist (MD) coverage 

[7]. There is a maldistribution of coverage for obstetrical needs 

as opposed to surgical needs [8,9]. Lastly, in situations where 

anesthesiologists work alone and without support personnel, it is 

fair to assume that they cannot spend the time outside the

- 208 - 

operating room to devote their attention to work in the recovery 

room, intensive care unit or as- a person who responds to cardiac 

arrests or functions as consultant for pulmonary management. The 

best use of anesthesiologists would be to have them be involved in 

d¿cision-making skills that ¿ae needed from a physician and 

delegate the more technical skills to people with lesser training 

[i]. 

Time-sharing of anesthesiologists: the Solution 

The use of physician extenders is a rationale reflecting a 

national trend to making tho time of physicians available to more 

patients by extending their ability to see patients. 

Telecommunications technology in medicine or Telemedicine has been 

proposed as a solution to problems in medical care during the 

1960s [10,11]. Health care professionals in this mode _f iractice 

use the telecommunication channels to communicate with each other 

or with patients in order to improve the cost-effectiveness and/or 

cost benefit of the delivery of health care services. Systems 

analysis studies have shown that using either or both the new 

communication technologies and different ways to organize the 

practice of medicine can result in improving both the quality and 

These studies were funded by grants from the U. S. Public 

Health Service and from the National Institute of Health.

- 209 - 

accessibility of health care delivered while at the same time 

reducing the costs of same [12,13]. 

Two programs to train Physician Assistants in Anesthesia were 

developedl based on the perceived need for more non-physician 

anesthesia personnel possessing training that is different in kind 

from the Nurse Anesthetist. This training, it was thought, should 

make the graduates be on a par with a Nurse Anesthetist yet allow 

them increased upward mobility with access to professional station 

in or outside of medicine. Further, it was hoped that some of the 

inter-professional political strains which have traditionally 

existed between MD's and RN's would thus be avoided. 

It was thought the people drawn into this kind of education 

would not commonly be those from the nursing pool. Nursing at the 

time of development of the above curricula was experiencing its 

own shortage problems. These shortages persist. It was also 

thought that the use of a non-physician to extend the contact time 

physicians could have with patients would extend their 

decision-making qualities to more patients while reducing the 

costs of service. 

1 Case Western Reserve University in Cleveland, Ohio and Emory 

University in Atlanta, Georgia.

- 210 - 

Prior to developing the new curriculum, a task analysis was 

undertaken in anesthesiology (14]. The study identified all tasks 

performed and established which of them could reasonably be 

delegated to non-physicians in a manner that would conserve the 

anesthesiologist's time yet re-slt in quality patient care and 

wider coverage. The determination of which tasks can be 

delegated, under what conditions and to whom was made by a group 

of qualified anesthesiologists. 

The design of the curriculum was obviously an extension of 

the results of the task analysis and of the uob Evaluation 

Point-Rating System [15,16,17] to be discussed later in this 

paper. An assessment of t.h types of activities that 

non-physicians could reasonably engage in. resulted in the 

curriculum for Anesthesiologists Associates (18]. 

Lastly, a methodology was developed to cesign optimal 

anesthesia team involving Board-qualified MD's and the Anesthesia 

Associates. The mix of provider categories was considered in this 

methodology to be constrained by budget, personnel availabilities 

and the acceptable levels of worker "overloads". The overloads 

were used as surrogate measures of the quality of care. 

The computer simulation developed to study the efficacy of 

alternative team configurations generates a set of daily overload 

indicators for different team configurations given some patient 

load and mix distribution. The results are plotted on a two

- 211 - 

dimensional graph to illustrate the trade-offs between tolerable 

overload levels and manpower requirements. These simulation 

results provide a basis for manpower mix tradeoffs. An overload 

was defined as the unavailability of a given worker within a 

5-minute time period to respn¿d to demands for his or her services 

because of scheduled or unscheduled commitments to another case. 

The three overload categories were further defined as: 

(1) Emergency Overload: Number of periods of unsatisfied 

emergency calls per day. 

(2) Urgent Overload: Number of periods of unsatisfied 

induction, early maintenance and 

surgical procedure tasks per day. 

(3) Routine Overload: Number of periods of unsatisfied tasks 

per day. 

Experience with the curriculum in both programs at the 

Baccalaureate (CWRU) and the Masters level (Emory) indicates that 

the content is adequate to train people to function in 

anesthesiology at the level of a Nurse Anesthetist. Yet the new 

professionals bring more technical or specialized expertise, into 

the system. This additional expertise is primarily in respiratory 

monitoring and care and in the utilization of more sophisticated 

equipment. Acceptance of graduates 'from both programs has been 

quite good. Although many graduates have gone on to other kinds of 

activities and to graduate study, those who chose to remain in the 

profession and work in anesthesia are indeed gainfully employed.

-- 212 - 

BACKGROUND; LATIN AMERICA 

One of the main problems in primary health care (PHC) 

ddlivery systems in Latin-Ar,.erican countries is the lack of 

coordination between the providers of PHC services and the 

educational institutions engaged in training health personnel 

[19-26]. This lack of coordination translates itself into 

ill-defined statements about the quantity, quality and mix of 

professional resources needed to be trained. The result is an 

excess and maldistribution of health personnel in some areas and 

almost total absence of personnel in other areas. These 

distribution problems are further zomplicated by an accumulation 

of tasks at some professional levels while other levels are left 

with few tasks to perform. 

Task analysis, job and team design, much like in the 

Anesthesiology studies, were used as the methodological tools to 

generate a data base for use by planners of health services and by 

educational ihstitutions. The study developed a methodology 

which identifies the professional profiles needed to satisfy the 

special conditions of several regions of! Mex-ico. 

This study was and continues to be funded 'by the National 

University of Mexico and the National Council of Science and 

Technology (MEXICO).

- 213 - 

Primary Care Health Service System: The Solution 

The main objectives of the study were a) to develop a 

technique which will allow the definition of the number and kind 

of health personnel that ought to be trained; b) to provide a 

communication flow, between the training units and the providers 

of health care; and c) to establish a feasible way to integrate 

curricula for each of the categories of health personnel, thereby 

avoiding the proliferation of disorganized actions in providing 

health care. 

As a point of departure, the study defined a model of the 

health service system approach to providing care. It was decided 

to work under a model in which the service was more preventive 

' 

' . .. " , . ,:. . ,: , , , , . , ' i , ',~ ' :i '· , ._:;'.c'3, J _ 

than curative oriented (19,20,24,25], even though some curative 

tasks were included. The model focuses on the delegation of tasks 

to primary care and preventive medicine personnel, and on the 

design of provider team configurations. 

In ordet to do this, it was necessary to establish the 

different sequences of health care requiring different levels of 

professional skills. These sequences start with self-care, all 

tasks performed by the individual to preserve his or his family's 

health and end up with a sequence of tasks requiring the highest 

professional level. Lastly, the most significant health problems 

of Mexico were identified and prioritized. From the above a

- 214 - 

matrix (Figure 1) was constructed relating the sequences of care 

to the health problems found to be important. 

Job Design 

Following the methodology delineated in the Anesthesiology 

studies [14] modifications of job evaluation techniques were used 

to design new job descriptions for health personnel. Using the 

downward delegation of tasks principle, the needed health 

personnel profiles were thus clearly defined. All tasks that were 

actually performed by personnel were identified a.d delegated to 

the least expensive and/or least trained professional capable of 

performing the task adequately. 

According to the scheme shown in Table 1, each box described 

the functions that each level of professional has to perform with 

respect to each problem in order to maintain and/or preserve a 

health outcome. As in anesthesia, this resulted in job profiles 

not in existence at the time. This,: an additional. benefit of the 

analysis, lead to a more integrated approach to health care 

delivery. 

Ra.her than going to the field where the tasks were actually 

performed as called for by the traditional job evaluation 

technique, and as done in the Anesthesia curriculum study, panels 

from each of the professional levels in practice at

- 215 - 

different health and educational institutions in Mexico were 

identified and used in this study. Each panel had a high degree 

of expertise in each of the selected health problem areas. The 

task assigned to these panels was to fill each of the boxes in the 

table. They were asked to write the ideal professional functions 

that each level will be performing within the next 10 to 15 years, 

and to consider what would be the ideal mix of team 

configurations. 

As in Anesthesia, this study used the four basic criteria of 

job evaluation, namely 1) SKILL, 2) EFFORT, 3) RESPONSIBILITY, 4) 

JOB CONDITIONS [15,16,17]. Each of the above factors was in turn 

broken down to as many levels as necessary to rate the tasks 

necessary for handling each of the selected health problems. The 

panels were instrumental in delineating/refining all the factors 

and subfactors of this plan. Table 2 - shows the degrees assigned 

to each of the subfactors and the description given to each 

degree. The panels were able to establish a clear definition for 

each of the break points on the scale, avoiding uncertainty and 

favoring a more precise rating. The tasks thus delineated and 

their ratings with respect to malnutrition, a significant health 

problem in Mexico, are presented in Table 3. Currently work is 

proceeding in the mental health area, and the same information 

will be developed for 25 other major health problems. With all 

this information in hand, it will be possible to mix and

- 216 - 

interrelate all same-level tasks to each of the health problems. 

This procedure will further provide descriptions for new allied 

health personnel. 

This technique has prco.n useful to assign more adequate 

functions to already existing personnel. In Mexico, it has 

already defined provider functions for the family planning 

program. A continuing education workshop was organized in family 

planning for social assistants, social workers, nurse auxiliaries, 

registered nurses, general practitioners and gynecologists. The 

coverage of this workshop was nationwide, and it included all 

major Mexican health institutions. 

The first step was to establish what task cught to be 

performed by the personnel, irrespective of what they were 

actually doing in their respective institutions. Th,'s, the 

procedure described earlier was again applied. 

With the help of this panel, the tasks that each. personnel 

category should realize were defined and rated according to a 

number of criteria. Data were then gathered regarding who was 

currently performing some of the tasks. A questionnaire was 

administered to solicit the panel opinion with respect to the 

importance of each task in achieving the objectives of the family 

planning program and the frequency of task performance.

- 217 - 

The tasks that were rated "very important" and performed on a 

daily basis were thus identified and given greater priority. 

Based on this study, a course was developed and manuals for 

several health personnel categories were developed, including 

among others, an instruction manual, student manual, and 

actual-practice manual. 

The course was developed mainly to teach "multipliers" all 

around the country to handle these materials, by performing the 

double function of instruction while providing content and method 

for the training of other personnel as instructors. 

In August 1980, the first workshop including 150 employees of 

the Secretary of Health and Welfare from all over the country was 

conducted. It has been reported that this workshop had an impact 

in improving the performance of the personnel assigned to family 

planning programs. 

Using this approach, mid-career PHC workers have been trained 

for Colombia, Peru, Ecuador, Honduras, Brazil, Bolivia and other 

areas of Latin-America. Among these were PHC physicians, nurses, 

social workers, etc. Feedback indicates that these procedures are 

already being implemented in a number of agencies of family 

planning, mental health and other programs within Mexico, Colombia 

and Nicaragua. Lastly, the task analysis/job design approach 

outlined here has already been incorporated in redesigning 

curricula at several universities in Mexico. 

12

- 218 - 

SUMMARY 

This paper indicated how a technique developed for and used 

in industrial settings to evaluate manual-job compensation levels 

[16,17] was used to design new professional level job categories 

for providing more cost-effective health care. Moreover, the same 

basic approach was shown to apply in acute care U.S. settings 

[14,18] as well as in primary care within developing regions of 

Latin America [29,30,31,32]. 

13

IC1 

o 

-a 

- 

i 

- 

w 

im 

W..i C 

-J 

M. 

- 

o 

go 2 

N~ 

U- 

O 

o 

U- 

- 219 - 

-a u4 

1 

-J- _ 

EA~~~~~~~~~o v 

C O U.. 

O.- u. - H 0 

- _ 

o E 

/:Y i i .-.- r iO, 

_r 

I 

, u ! . . - . I 

= o 

E i.i 

,* E , - 

- E 

i- c_ - 

o , 

-. § _o u .- ,c o o_ 

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

I

.r 

-J 

-3 

uIt 

í 

- 220 - 

|-.. _.J Z..~ ,...e O ,e | _rVl e- e- 04 _' C _... 4I 

.- 

_- ~' 1.- O C z-i |0 w..-Qa" ,,z-3 4> . 

O0 O 

1- 

U 

. , . " 

i _u i ic Ol rA C a ° _ o cu L N - 

> 

cct- o a 

4> 

°CC_ O i - L. 

i i Ii 

-z 

LiO- ¡ 

O 4- - 

e 

_ Z *_ W x _ O - i_ eV _ L E C O- 

4>l a>Q L- 0.> U- L 

CJ C - C% L. ¡ - 

^ O _ O O .^ eOOi e . 

i ::L i 2 j li O . _ 

_ O 4>> 4> £_ 

cc o- e diO 

4 a 

!l = v I, 

c d A o 

w-- di 

*~~~~~~~~~~~ 

=0 ~.-·U- E a., ~~~~~~~~~~~ .Ein o :~n= 

0~ .~ 1n E a>.r > O a' j 

- C O 

z 

4> 

a> o

0 

z 

0 

u 

_ 

I- 

0 

Ut. 

I,, 

uj 

CD Q 

U- ,- 

,,J 

CD 

Li 

LU 

m.. _ ,J :Ez .( 

4) 0 LLI IO Q O 0 . O 

cL C _ ~ C'%_ Q 4@ r a 

_ Ll O S. ZIJ 4~~) C - Z- cr, 0 0 a) 0 C d w Cr CC x 

_I O r- · _ Q E · r _X 

C.) ~ 

"'" 0 I ~ ~3 

«3 ~ ~ 

CD~~C 

v- 

~:Z IIvCL 

c 

- _, :ZC~ 

-~~~r 

O 

-. 

C 

I , 

_ 

- 

~ 

Q 

0 

, 

,a 

0 ~ 

O 

o 

I 

E1C-.Nn= 

. r · q 

rL V 

s.N 0 

*- CiLI V V 

.O O 0 C( L C> a) I _:r r = Q · 

Z Cs _e uu E u'l ~ ~ I ~ ~ ~ :O~ 

b NN4 

LC~~~]~~~Q a 

. ., 

,:> h - i tN Q

w 

z 

c 

L 

-J 

C- 

u, 

o 

o 

C> 

C3 PI 

en 

- 222 -

TABLE 3 

TASK ANALYSIS OF MALNUTRITION 

T TASK A S K I |II | II' ! | i IVI 

! 

II__________________A _IB _ lt ID_ -a F IG HA I A I D I A 1 1 I B i ICC 

- . 1 . . . ~~~~~~~~~~~~ 

1 1 . . . . . 1 1 . 1~~~~~~~~~~~~~ 

SELF-CARE 

1. Weioht and measure the child periodically. 

2. Watch over the child in his daily activities 

and educational progress. 

_. Assist to the service of well baby care. 

4. To provide to the child a balanced diet 

prepared hysíenically according to the 

family's economic conditions. 

5. To promote that the child be engaged in 

frequent recreative activities in open 

places. 

MONITOR 

6. Transmite information to organized groups. 

7. Realize surveys anc distribute printed 

material with informration on malnutrition. 

8. Control the ralization of problem's 

detection campaigns and their prevention. 

9. To differentiate presumtive degrees of 

malnutrition. 

10. To canalize to specialized treatment the 

required cases. 

1ST. PROFESSIONAL LEVEL 

11. To carry out lab. examinations. 

12. To capacitate monitors. 

i3. To supervise monitors. 

2ND. PROFESSIONAL LEVEL 

i4. To establish clinic diagnostic of certainty 

of malnutrition and degrees of it. 

!5. To indicate and to interpret lab. exams. 

i16. Soecific trea;ment of the cases. 

i.7. Follow directly the evolution of acute 

cases. 

'S. Recister in total form the cases treated. 

19. To interput :reatment resulis. 

20. Send to subordinated levels tne cases in 

resolution pnase in order So control them. 

1 

0 

0 

1 

1 1 

612 

0 

0 

0 

1 

2 

3 

3 

3 

[3 

3 

13 

3 

o0 1 o 

01010 

0 G10 

111 

0 1o 0 

2 

2 

2 

2 

2 

2 

2 

2 

c 

1 i 

1 11 

1 

1 

1 

O 

1 

1 

1 

1 

1 

1 

1 

1 

2 12 21 2 

2 

2 

2 

2 

2 

312 

2 

2 

2 

2 

2 

1 

1 

1 

1 

i 

1 

0 

O 

O 

0 0 

2 

2 

2 

2 

2 

2 

2 

2 

2 

2 

2_ 

2 13 

o IC 

O 11 

0 1I o 

2 

1 

2 

2 

2 

2 

212 

2 

2 

2 

2 

2 

2 

312 

3 

3 

- 223 - 

2 21j 

2 S 

i2 ! 1 

i 

2 

: 9 

1 0 0 lo0 

0 10 lo 

lo 

1 

2 

2 

2 

3 

2 

S 

010 

0 oo00 

0 0 

1 

1 

o 

O 

0 

2 

1 

1 

2Z1 

012 

v 

0 

0 

0 

.I 

I 

0 

1 

o 

0 

112101 

0 

G 

3 

1 

r1 .o 

1 

3 

2 

Ol Io 

1 1 o 112 

1 lo0 1111 

1 

i 

2 

1 

2 

2 

2 

1 

3 

1 

i 

0 2 12 

0 

3 

12 :1 j2 

2 

3 

J 

.1 I- 

L ! , l . , 

r i ! 

· · ~ . . ~ --- 

I 

i 

1 

o ! 

0 

1 

G!I 

- 1 

itC~ 

Irr 

¡1 

1''

uj 

Ou> 

a-L YCI-qn 

-a 

u- 

L- 

o 

Ca =_.= 

Vmlm 

z 

W 

3- 

o- 

1< 

~ ~ ~ ~ U 

z 

1z- 

J 

L < 

~c 

a < 

9¿ 

u, 

cn 

w 

1oo 

ii 

0 

3: 

i 

- 224 - 

o --a 

a- (.I V:cnc 

$~~~~~~~~~~~~~~~~~~3 I-v I L.&. 

-L f~ -W~ =L r 

'a- 'J V))

11 

- 225 - 

TASK ANALYSIS OF MALNUTR TIOj 

TAS K J I II IIII IV 

A D)EF BC G H IG A CB JC A IB ¡A B 1 C D)E 

21. To maintain interchanges of experiences. 5 2 4 2 21 2 3 3 2 2 1 2 1 4 3 3 0 0 

of the case with health units related to 

the problem. 

22. Plan and develop training courses for 7 2 4 3 2 1 2 3 3 2 2 2 2 2 3 3 3 0 O 

auxiliar personnel. 

3RD. PROFESSIONAL LEVEL 

23. Planning of detection campaigns. 7 2 4 3 2 1 2 3 312 2 2.2 1 4 3 3 0 0 

24. To diagnose nutritional deficiences due 

to specific factors. 6 2 4 3 2 1 2 3 3 2 3 O 2 1 4 3 O O 

25. To prescribe specific treatment.. 6'2 4 3 2 1 2 3 3 2 3 0 2 1 4 3 3 O O0 

26. To dq research on malnutrition. 6 2 4 3 2 1 2 3 3 2 3 012 1 4 3 3 O O0 

27. To publish papers with respect to research 6 2 4 3 2 1 2 3 3 2 3 1 2 1 4 3 3 O 

outcomes. 

28. To disseminate research's results. 6 2 4 3 2 1 2 3 3 2 3 1 2 1 4 3 30 0 

29. E aborate eoordinate and administer 7 2 4 3 2 1 2 3 3 2 3 22 1 4 3 3 O O0 

teaching and service plans. 

30. To interchange Interinstitutional 6 2 4 3 2 1 2 3 3 2 3 1 214 3 0 

experiences. 

1 1 1 1 ~ . 

1 1 1 . 1 1 1 !- . 

-

- 226 - 

REFERENCES 

1. Gravenstein, J.S., J.E. Steinhaus, P.P. Volpitto. "Analysis 

of Manpower in Anesthesiology," Anesthesiology, 33(3), 

September 1970. 

2. Gravenstein, J.S., A.Esogbue, A. Reisman, B.V. Dean, V.V. 

Aggarwal, V. Kaujalgi, and P. Lewy. "Physician Supply and 

Surgical Demand Forecasting: A Regional Manpower Study," 

Management Science, August 1973. 

3. Reisman, A., B.V. Dean, A.O. Esogbue, V. Aggarwal, V. 

Kaujalgi, P. Lewy, C. deKluyver, J.S. Gravenstein. "Supply 

and Demand of Anesthesiologists in Cuyahoga County, Ohio," 

The Ohio State Medical Journal, 69(10), pp. 760-763, 

October 1973. 

4. Moore, Francis, "Ravenstine Lecture," Anesthesiology 

48:125-138, 1978. 

5. Ament, Richard, "Anesthesia and Surgical Care: Manpower 

Needs and Ultilization," Anesthesiology, 48:88-90, 1978. 

6. Orkin, Fred, American Society of Anesthesiologists, 1981 

annual meeting, unpublished report. 

7. Professional Activities Survey, P.A.S. Ann Arbor Michigan 

1978-1980. 

8. "National Study of Maternity Care: Survey of Obstetric 

Practice and Associated Services in Hospitals in the United 

States - 1970," Committee of Maternal Health, American 

College of Obstetricians and Gynecologists, 1970. 

9. Rasmussen, J., T. George, A. Reisman, W. Cull, J.;S. 

Gravenstein, "Obstetrical Anesthesiology Practice and 

Attitudes in Cuyahoga County, Ohio Hospitals: A Survey," 

The Ohio State Medical Journal, Vol. 73, pp. 139-144. 

10. Bashshur, R.L., Armstrong, P.A., Youssef, Z.I. Telemedicine: 

Explorations in the Use of Telecommunications in Health 

Care. Springfield, Charles C. Thomas, 1975. 

11. Gravenstein, J.S., B. Grundy, Y. Pao, A. Reisman, F. Staub, 

"The Cleveland Telemedicine Project," Telecommunications 

For Civic And Social Services, No. 20, 1976 Wescon 

Professional Program Proceedings, Los Angeles, pp. 1-2, 

September 1976.

- 227 - 

12. Reisman, A., F.Staub, B.V. Dean, J.S. Gravenstein, V. 

Aggarwal, V. Kaujalgi, C. deKluyver. "Anesthesiology 

Manpower Planning Study Phase III - Progress Report I," 

Technical Memorandum No. 280, Department of Operations 

Research, Case Western Reserve University, December 1972. 

13. Reisman, A., W. Cull, H. Emmons, B. Dean, C. Lin, J. 

Rasmussen, P. Darukhanavala, T. George, "On the Design of 

Alternative Obstetric Anesthesia Team Configurations," 

Management Science, Vol 23, No. 6, February, 1977. 

14. Reisman, A., B.V. Dean, V. Kaujalgi, V. Aggarwal,.P. Lewy, 

J.S. Gravenstein. "A Task Analysis in a Clinical Specialty 

Providing Data for a New Curriculum for Anesthesia 

Personnel," Socio-Economic Planning Sciences, Vol. 7, 

pp. 371-379, 1973. 

15. Job Evaluation Plan, American Institute of Industrial 

Management. 

16. Lanham, E., Job Evaluation. McGraw-Hill, New York, 1955. 

17. Lytle, C.W., Job Evaluation Methods. Ronald Press, New York, 

1954. 

18. Gravenstein, J.S., and M. F. Rhoton. "Teaching Anesthesia to 

Undergraduate College Students," Anesthesiology, 37(6), 

641-646, 1972. 

19. Arango, Jaime "Planeacion Curricular y Evaluacion en Funcion 

de las Necesidades de un Pais," Educacion Medica y Salud, 

6, 3 y 4, 1972. 

20. Butler, Willis, "The Undergraduate Education of Physicians in 

Cuba," Journal of Medical Education, Vol. 48, September, 

1973. 

21. Galvao Lobo, Luiz Carlos, "O Uso de Nova Tecnologia 

Educacional no Formacao de Recursos Humanos," Educacion 

Medica y Salud, 8, 2, 1974. 

22. Gutierrez, Rodrigo, "Participacion de la Universidad en la 

Planificacion de la Salud," Educacion Medica y Salud, 

7,3-4, 1978. 

23 Hutchins, Edwin B. y Hoack, Horst, "Evaluacion de los 

Programas de las Escuelas de Medicina," Educacion Medica y 

Salud, 4. 4. 1970. 

24. Duran, Lilia, "An Alternative Way of Curriculum Planning in 

the Health Sciences." Conference presented at the 

International Coloquium of Education, National University 

of Mexico, October, 1979.

- 228 - 

25. Duran, Lilia, "Desarrollo de una Metodologia para Establecer 

Prioridades de Atencion en el Sector Salud." Internal 

working paper, Centro Universitario de Tecnologia 

Educacional para la Salud (CEUTES) , June, 1981.25. 

26. LaSalle, Gerald, "Nuevos Modelos en la Formacion Profesional 

del Personal de la Salud," Universidad de Sherbrooke, 

Educacion Medica y Salud, 4, 4, 1970. 

27. Maddison, D.C. "What's Wrong with Medical Education?" Medical 

Education. 1978, 12, pp. 97-102. 

28. Rexed, Bror, "The Role of Medical Education in Planning. The 

Development of a National Health Care System," Journal of 

Medical Education, Vol. 49, January 1974. 

29. Duran, Lilia, "Una Aplicacion de la Tecnica de Analisis de 

Tareas a Partir de un Modelo de Servicio en el Area de la 

Salud y su Derivacion en Objetivos Generals de Carrera." 

Technical Report. Centro Latinoamericano de Tecnologia 

Educacional para la Salud (CLATES), 1979. 

30. Duran, Lilia y Torres, A., "Planificacion Familiar: Una 

Aplicacion de la Tecnica de Analisis de Tareas." Technical 

Report. Centro Latinoamericano de Tecnologia Educacional 

para la Salud (CLATES), 1979. 

31. Duran, Lilia, "Algunas Sugerencias de Apoyo al Catalogo 

Nacional de Ocupaciones en el Area de Planeamiento 

Curricular." Documento Presentado a la Direccion General 

de Planeacion de la S.E.P. (Secretary of Public Education) 

Marzo, 1980. 

32. Duran, Lilia, "Technology Development for a Quantitative and 

Qualitative Definition of Human Resourcres in the Health 

Field." Research proposal submitted to the National 

Council of Science and Technology, June, 1979.

- 229 - 

METHODS 

Parameters Affecting Hospital Occupancy 

and Implications for Facility Sizing 

By Walton M. Hancock, David B. Magerlein, Robert H. Siorer, and 

lames B. Martin 

Simulation is used to investigate the effects on hospital occupancy of the 

number of beds in the facility, the percentage of patients who are emergencies, 

the percentage of elective patienta who are acheduled, and the average 

lengths of stay of emergency and elective patients. A practical method is 

presented for estimating the optimum suze of a short-term hospital on the 

basis of expected demand, and use of the resulta in planning is discussed. 

To operate at minimum cost it is imperative that a hospital not 

contain more beds than necessary to meet demand. The nontrivial 

costs of building, stafflag, and maintaining unused beds are unnecessary 

and are ultimately bdrne by the health care consumer. The variable 

cost of staffing and maintaining a hospital bed was found to be 

$10,130 per year in one study [1] and $16,201 per year in another [2], 

and the fixed cost of building a bed has been quoted as $50,000 [3]. 

Thus, considerable savings are possible by eliminating beds in an 

overbedded Hospital (saving variable costs) and by preventing the construction 

of unneeded beds (saving both fixed and variable costs). 

A simulation-based analysis of the effects of several parameters 

on hospital occupancy is presented. The parameters investigated are 

number of beds in the facility, percentage of patients who are emergencies, 

percentage of elective patients whose admission date is set in 

advance (scheduled patients), and mean lengths of stay of emergency 

and elective patients. A practical method is developed for predicting 

the correct size of hospital facilities, given the expected demand. Current 

planning methods such as the Hill-Burton formula, the Poisson 

assumption, and that of Shonick [4] are shown to be inappropriate 

ThiJ study was supported by PHS grant no. HS00228 from ¿he National Center 

for HIalth Services Rescarch (DHEW) and by the W.K. Kellogg Foundation 

through the Greater Detroit Arca Hospital Council, Inc. 

Addrdeu communications and requests for reprints to Walton M. Hancock, 

HEALTH ProCessor of Induswlrial and Operations Engineering and Hospital Administration, 

SERVICES Program and Bureau of Hospital Administration, University oJ Michigan School of 

RESEARCH Public Health, 1420 Washington Heights, Ann Arbor, MI 48109. David P. Mager. 

ARC. A'I cin is a graduate student and Robert H. Storr is an undergraduate student, both 

in industrial and operations engineering; and ]ames B. Martin is assistant professor 

of hospital adminisuration. All of the authors are at the University of Michigan. 

0017-9124/78/03027614/$02.00/0 

( 1978 Hospital Research and Educational Trust

- 230 - 

since they substantially overestimate the necessary number of beds in SIMULATION 

most cases. OF HOSPITAL 

OCCUPANCY 

Simulation Model 

The Admissions Scheduling and Control System 

In order to model and simulate a hospital, specific rules, policies, 

and priorities regarding patient admissions must be defined. The 

admissions -guidelines of the Admissions Sclieduling and Control System 

(ASCS) developed by Hancock et al. [51 are used in this model. 

Simulation studies have predicted that hospitals using this system 

would operate at occupancies in excess of those found in most hospitals, 

and implementation of the ASCS in several hospitals has shown 

these predictions to be realistic [6,7]. A facility using the ASCS should 

require fewer beds to meet demand and thus should operate at a lower 

cost. A model using the ASCS should predict the optimum number 

of beds needed to operate a hospital. 

In the ASCS, admissions to the hospital are classified in three 

categories: emergency, scheduled, and call-in. An emergency patient is 

defined as one requiring immediate admission to the hospital. Thus 

an emergency admission is uncontrollable and may be considered a 

random event. Scheduled and call-in patients are elective and thus do 

not require'immediate admission; they may be put on a v.aiting list 

or scheduled for admission at some future date. The admission of a 

scheduled patient is planned for a specific time-in the future, and a 

call-in patient is called in for admission at the hospital's convenience. 

Thus, if at some time during the day the hospital has beds available, 

patients are called in. 

A turnaway is defined as an emergency patient who cannot be 

accommodated in the normal manner because all hospital beds are 

full. A cancellation is defined as a scheduled admission that must 

be cancelled in order to save room for emergency patients who may 

arrive before the next day's discharges. Thus a scheduled admission 

is cancelled in order to prevent the possibility of an emergency turnaway 

later in the day. In practice, cancelled patients are rescheduled 

at the next open date or are called in with the highest priority. In 

this study, turnaways were constrained to be between 1 and S percent 

of all emergency arrivals and cancellations were constrained to be 

between 1 and 3 percent of all scheduled admissions. These percentages 

were chosen because they appeared to be acceptable to the hospitals 

that had become aware of the ASCS. 

At some point during the day when all discharges are known, the 

hospital must decide if it is necessary to call in patients or to cancel 

any scheduled admissions. This decision is based on the census reduction 

allowance (CRA) and the cancellation allowance (CA). The CRA 

and the CA represent the upper and lower bounds on the number of 

beds left empty at the decision point and may be different for each 

day ,. e week. The number of filled beds includes those to be occupied 

by patients scheduled for admission later in the day. If the number 

of empty beds is g_ eter than the CRA, patients are called in until

- 231 - 

~HNCOCK Fig. 1. Daily census with and without call-ins. 

ET AL 

Without Coll-in$- 

45 - Census Mean , 172.63 

Census Standard Devioation · 11.459 

40 - With Co'l-ins 

Census Mean*187.85 

35 - Census Standard DeviQtion 5.956 

25 - 

z II, 

120o 

¡40 'N45 150 ¡55 160 .t65 170 175 180 185 190 195 200 

CENSUS 

the number of empty beds equals the CRA. If the number of empty 

beds is less than the CA, scheduled admissions are cancelled until the 

number of empty beds equals the CA. Thus the census at the decision 

point is always between the CRA and the CA. The overall effect 

of these allowances is to reduce significantly the variance in the hospital 

census and thus increase the attainable occupancy while maintaining 

a given turnaway level. 

Reduction of census variance through use of the call-in algorithm 

is the mechanism that allows the ASCS to achieve high average occupancies. 

This census variance reduction is illustrated in Fig. 1, which 

shows two separate simulation runs in which the numbers of scheduled 

and emergency patients admitted are equal. It is apparent that a 

facility using the call-in algorithm will operate at a higher average 

occupancy and with lower census variance than a unit without callins. 

In addition, a facility using the call-in algorithm will be at its 

bed capacity (200 beds in Fig. 1) much more often than a facility without 

call-ins. 

Probability Distributions of Emergency Arrivals 

and Patient Lengths of Stay 

In order to simulate the randomness of a hospital, it is necessary 

to assign a probability density function to emergency arrivals and to 

patient length of stay. The Poisson distribution is used here to model 

HEALTH emergency arrivals. This has been done often and has been found to 

E VEICRS be a good fit to empirical daily emergency arrival distributions [8,9] 

as well as being theoretically appealing. In the simulation model, a 

day is divided into two periods. The first period extends from the

- 232 - 

time when discharges begin to the time when all discharges are known 

(the decision point). The second period extends from the decision 

point to the time when the next day's discharges begin. For simplicity, 

it is assumed that the distributions of emergency arrivals in each period 

are identical. The assumption should have little effect on the 

results. The percentage of emergency arrivals, a variable closely allied 

to the overall mean emergency arrival rate, is one of the parameters 

whose effects are investigated in this study. 

Assigning a probability distribution to inpatient length of stay 

(LOS) is difficult. The nature of these distributions varies greatly 

among hospitals, services, and days. Some researchers have used negative 

exponential, lognormal, and gamma distributions to model LOS 

[7,8,9). Others report that no distribution fits their LOS data [7]. One 

possibility is to use an empirical distribution from a specific hospital, 

but that idea is discarded here in an effort to keep the model as generally 

applicable as possible. The use of a distribution described by 

a mathematical function makes it easy to produce probability density 

functions with given means and variances. This in turn allows LOS 

to be varied as a parameter. 

To model LOS, the lognormal distribution, which has been found 

to be a good approximation to empirical LOS distributions, is used 

[10]. To justify its use, 140 LOS distributions*-om five different hos. 

pitals were examined. The use of the lognormal distribution appeared 

to be reasonable, and it was concluded that a variance of about 

seven times the mean is a good approximation. In this study, mean 

LOS is assumed to be somewhat higher for emergency patients than 

for elective patients. 

Because of computer capacity restrictions, the LOS distributions 

were truncated at 50 days within the simulation, but this had little or 

no effect on the results because the LOS distributions had low frequencies 

at 50 days. (All means and variances refer to the truncated 

distributions.) 

Simulation Runs 

As stated previously, the effects of several parameters on the maximum 

average occupancy of a facility were examined. Maximum aver. 

age occupancy is the highest average occupancy that can be achieved 

by searching for the best choices of the CRA and CA for each day of 

the week. Initially, the following paramaters were investigated: number 

of beds, percent of emergency admissions (percent EMG), and percent 

of elective patients scheduled. 

During the first phase of the study, patients were scheduled five 

days a week (to reflect the scheduling policy of most hospitals) and 

emergency and elective LOS means were 10.38 and 8.39 days, respectively. 

These lengths of stay are typical of those found in midwestern 

general hospitals. 

The number of beds was varied over 40, 80, 120, 160, and 200. At 

each level the percent emergencies and percent scheduled were varied 

as shown in Table 1 (p. 280). 

SIMULATION 

OF HOSPITAL 

OCCUPANCY 

FALL 

1978

- 233 - 

HANCOCK Table 1. Parameters Varied for Simulation of 

ET AL 

Hospital with Five-Day Scheduling to Determine 

Maximum Average Occupancy at 40, 80, 120, 

160, and 200 Beds 

Phase 2 

Phase I 

E1%~'~ % ~ 

EMfG scheduled 

Mean 

emeremnc 

LOS 

Mean 

elecotuve 

LOS 

acheduled 

30 40, 70, 90 5.1 43 40, 70. 90 

50 40,70,90 10.4 8.4 40,70.90 

66 40,70.90 14.9 13.1 40,70.90 

95 0 

In the second phase of the study, the effects of mean LOS on maximum 

average occupancy were investigated. Percent EMG was held 

constant at 50 percent and patients were scheduled five days a week. 

Again, the number of beds was varied over the same five levels. Within 

each level, mean LOS and percent scheduled were also varied as shown 

in Table 1. 

In the third part of the study, the effect of scheduling seven days 

a week rather than five was investigated. Percent EMG was held constant 

at 50 percent and the emergency and elective LOS means were 

10.38 and 8.39 days, respectively. The number of beds was again varied 

over 40, 80, 120, 160, and 200, and percent scheduled was varied 

over 40, 70, and 90 percent. 

A Fortran simulation program developed by Hamilton, Hancock, 

and Hawley [11] was used. This program embodies the rules of the 

ASCS and allows much flexibility in modeling different hospital and 

facility settings. The simulator has been documented and validated 

[5,11], but an attempt was made here to estimate the magnitude of 

possible errors involved in the simulation process. 

The parameters percent emergencies and percent scheduled are 

not directly controllable through the input but are dependent on the 

number of call-ins that occur. Thus a search procedure is involved in 

which the inputs are varied until the proper levels of the parameters 

are approximated and the turnaway and cancellation constraints are 

satisfied. Once this is accomplished, the allowances are varied until 

the maximum average occupancy at the given parameter levels is 

determined. This procedure causes two possible sources of error. 

It is impossible to set all of the parameters exactly at the specified 

levels. Percent emergencies and percent scheduled are generally 

within -2 percent of the assigned level, and the error that this induces 

in maximum average occupancy may be estimated from the 

HEALTH results using linear interpolation. For example, Fig. 2b shows that in 

XUEACI 

K a facility with 40 beds. a change of 30 percent in percent scheduled 

results in a change of 6.5 percent in maximum average occupancy. 

Thus an error of 2 percent in percent scheduled would cause an error

- 234 - 

Fig 2. Maximum average occupancy vs. number of beds for different percentages 

of emergency arrivals and with 40, 70, and 90 percent of electives scheduled. 

4- 

c 

Q) 

o, 

8 0 

w 

(Z 

ir 

2 

2 

00ooE 

95 

90 

85, 

100 r- 

- 30%EMG 

-50% EMG 

- 66% EMG 

(a) 

i .... 1 i .... 1.,. ... i .. 1 

I. 1 * . . ... 

........... ......... . .. 

951- 

90 

30% EMG 

50% EMG 

85 -66% EMG 

A .1 

l u 

100 r 

95 .. 

90 

50% EMG 

85 -66% EMG 

30% EMG 

Oo. 

40% of the Electives are Scheduled 

I . 

(b) 

1 , .... l . ., .e i . 1 e 1 . IIa 

- - - - - - - - - - - - - - - 


cele 1 e1 IIo le! e ej le m te 

50 75 t00 125 150 

NUMBER OF BEDS 


(c) 

1 1 I le L í 

175 200 .

- 235 - 

Fig. 3. Maximum average occupancy vs number of beds for different 

percentages of scheduled arrivals and with 30, 50, and 66 percent 

emergency arrivahs. 

100 r 

95 

90 

85 

- A 

c lw 

o 

g 95 - 

uo 

o 

90 _ 

o 85 - 

, < r_ 

95 - 

40% 0SCH 

l90% SCH 

' f . lilit 1 i I I1 i" 

40%SCH 

90- 40% SCH>/ 

70%SC 

85 9 0 %SC' 

Ot '- '550 

(a) 

1 I tf I 3 I m 1I 1 ¡ t ¡ t i 

30% of the Arrivalis ore Emergencies 

50% of the Arrivais are Emergencies 

66O, 

(b) 

(c) 

1 1 1 

75 

le iI i 1 1 1 i lb i i ¡te i 

100 125 150 


i I i i i i I 

175 200 

YO of the Arrivas 'are Emergencies

- 236 - 

of (0.02 x 0.065/0.30) x 100 = 0.43% in maximum average occupancy. SIMULATION 


Similarly, for percent EMG (Fig. 3b, 40 beds), an error of 2 percent in OCCUPANCY 

percent emergency would cause the following error in maximum 

average occupancy for the 30 and 50 percent emergency levels: 

(0.02 x (0.937 - 0.897)/0.20) x 100 = 0.40% 

These figures represent the maximum errors in occupancy. In general 

(at different parameter levels) the error will be much smaller because 

the differences in the numerator are smaller. The possible errors are 

easily calculated from the graphs for an individual case. (It is interesting 

that the error increases as the number of beds decreases.) 

As with any simulation, there is a certain variability in results 

due to pseudorandom generation of numbers. In order to estimate 

the variance in occupancy within the simulation, 20 100-week runs 

were performed with identical inputs but with different seeds for the 

random number generators. The sample standard deviation in mean 

percent occupancy between runs was found to be 0.033 percent. 

It is possible to estimate the total error in predicted maximum 

average occupancy due to the simulation. Assuming that the error 

due to pseudorandom number generation is normally distributed 

and using a 99-percent confidence interval, the maximum error in 

occupancy (percent) is 

2.58 x (o-EM 2 + arEL 2 + 'RNp2) X 100 

where oE, = the standard deviation of the error due to variation 

in percent emergency. The maximum error is already 

computed as 0.43 percent. Assuming thie error to be 

normally distributed and using 99-percent confidence 

intervals 

YO -EM 0.43/(2.58 x 100) = 0.00167 

oEL = the standard deviation due to variation in percent electives. 

The maximum error is computed as 0.40 electives. 

Using the same assumptions'as for o'Em 

EZL f 0.40/(2.58 X 100) --'0.00155 

CraN = the standard deviation due to pseudorandom number 

generation. This has previously been found to be 

0.00033. Therefore, maximum error in occupancy = 

±2.58 x (0.001672 + 0.001552 + 0.000332)J x 100 = +0.594 

percent occupancy. This figure, ±0.594 percent occupancy, 

is an estimate of the maximum error due to the 

sources discussed above. 

Restu. . FALL. 

1978 

The results of the procedures outlined are shown in Figs. 2-5, 

and the effects of the different parameters on occupancy are discussed

- 237 - 

HANCOCK 

ET AL Fig. 4. Maximum average occupancy vs. number 

of beds with 95 percent emergency arrivals and 

all electives called in. 

o95 

100oo 

90 

85 

0% CH 

0 50 75 o00 125 150 175 200 


below. Perhaps it would be desirable to use Fig. 4. as a benchmark 

when examining the remaining figures since it is closest to the historical 

Poisson arrival process. 

In all cases, maximum average occupancy increases with increasing 

facility capacity. The higher percent occupancy in larger facilities 

results from a decreasing coefficient of variation of census (standard 

deviation/mean) as the number of beds increases. When the coefficient 

of variation is small, the maximum average mean census may be 

doser to the facility-capacity while still maintaining the turnaway 

and cancellation constraints. 

For a given number of beds maximum average occupancy is observed 

to increase with decreasing percent scheduled. This is a result 

of the fact that as percent scheduled decreases, the number of call-ins 

is increased. Thus the ability of call-ins to reduce census variance is 

improved as percent scheduled decreases. 

In this study when patients are scheduled five days a week, a lack 

of call-ins is felt most heavily on weekends. With no patients scheduled 

for Friday and Saturday, it is necessary to call patients in to 

maintain a high maximum average occupancy. If the call-ins are 

unavailable due to a high percent scheduled, a weekend drop-off in 

average occupancy occurs. (Remember that the ratio of call-ins to 

scheduled patients is constant.) 

In most cases, maximum average occupancy is observed to increase 

as percent EMG decreases. This is because, in general, emergency 

arrivals introduce a greater variance in the census than elective arrivals. 

In certain cases, however, this is not true. In this study, when 

percent scheduled is 90 percent, the variance in census caused by elective 

admissions is larger due to the drastic weekend drop-off in averlHELTH 

age census. Thus, when percent scheduled is high and patients are 

ReIEARCH scheduled five days a week, maximum average occupancy will decrease 

as percent EMG decreases and the percent of elective patients 

increases.

- 238 - 

Fig. 5. Maximum average occupancy vs. number of beds with 50 percent emergency 

arrivais and 40, 70, and 90 percent of electives scheduled. Results of model are 

compared with Hill-Burton formula and Poisson model in (c). 

c 

Q>

- 239 - 

AECOCL. Table 2. Eflect of Changc in Schedulcd 

Caincellation (SC) and Emergency Turnaway (ET) 

Rates on Maximum Average Occupancy 

Maximum average Average no. 

No. of o/< 

bds 

beds EMG 

% 

sched.uled 

occupancy SC + ET for 

1-3% con 

SC and SC and 

ET: 1-3% ET: 2/mo tra 

320 30 30 99.3 98.7 8.3 

320 30 90 94.6 90.6 25.6 

320 90 90 96.1 90. 21.9 

80 30 30 97.6 96.5 3.2 

80 30 90 90.7 87.3 5.4 

80 90 30 90.8 85.0 4.7 

*Fewer than two per month but as close to tw as posible. 

In general, if percent scheduled is below a certain level (approximately 

80-90 percent) or if patients are scheduled seven days a week, 

maximum average occupancy will increase as percent EMG dqcreases. 

It is advantageous (from a maximum average occupancy standpoint) 

to schedule seven days a week because no weekend drop-off 

then occurs. Figure 5a shows that when percent scheduled is high, the 

use of seven-day scheduling allows a higher maximum average occupancy 

than does the use of five-day scheduling. This is to be expected 

since the'weekend drop.off for five-day scheduling has a greater effect 

when percent scheduled is high. Thus, when sizing a facility it is necessary 

to be aware of how patients are to be scheduled. 

The results of varying mean LOS are shown in Fig. 5b, where it 

can be seen that mean LOS has a much smaller effect on maximum 

average occupancy than the other parameters do. In sizing larger 

facilities, the mean LOS will have little effect on the results, but the 

same cannot be said for small facilities (fewer than 40 beds). 

As mentioned previously, the maximum average occupancy is affected 

by percent cancellations and turnaways. In Table 2 the maximum 

average occupancy with the 1-3 percent constraint on cancellations 

and turnaways is compared with the maximum average occupancy 

with the sum of the cancellations and turnaways being less than 

but as close as possible to 2 per month. For 320 and 80 beds, the constraint 

change results in a decrease in the cancellations and turnaways 

and the maximum average occupancy also decreases. For comparison, 

the sum of cancellations and turnaways for the 1-3 percent constraint 

is also given in average occurrences per month. For purposes of 

planning or determining the number of beds needed, the scheduledcancellation 

and emergency.-turnaway rates should be viewed as a 

HEALTH matter of policy. Once agreement is reached on their acceptable level, 

LSEAR' CH lthen maximum average occupancies can be computed using the simulator. 

It should be noted here that the occupancy data for 320 beds 

are not strictly part of this research and are furnished only for their

- 240 - 

vallie in thc sensitivity analysis of the cancellation and turnaway SIMULATION 

constraint. OF HOSPITAL 

OCCUPANCY 

Application of Results in Facility Sizing 

-The results of this study may be applied in determining the correct 

size of a hospital. The method is as follows. 

A. Obtain an estimate of the average daily census (ADC) of the 

facility. For planning purposes, this is frequently obtained by 

taking the present ADC and adjusting for demographic factors 

over thc planning horizon. 

B. Estimate the parameters for the facility (percent EMG, percent 

scheduled, mean LOS, and scheduling pattern). The usual 

assumption is that the percentages will not change over the planning 

horizon. 

C. Find the graph in Figs. 2-5 that is closest to the parameters 

estimated in B above. 

D. Determine the necessary number of beds (NB) using successive 

approximations as follows. 

1. Use ADC as an initial estimate of the number of beds, and 

determine a percent occupancy from the appropriate graph. 

2. Determine the number of beds needed using NB = ADC/ 

(% occupancy/100). 

3. Use the number of beds found in step 2 to determine a 

revised percent occupancy. 

4. Return to step 2 and compute a new NB using the revised 

occupancy of step 3. 

5. Repeat steps 2, S, and 4 until the bed-number estimates 

converge. 

Use of the algorithm described above assumes, of course, that the 

ASCS system will be used to admit patients to the facility. The specific 

schedules, CRA and CA, which are specific for day of the week, 

can be quickly obtained using the admissions simulator. These values, 

of course, will vary for any point on the particular curve used. 

As an example, consider a facility with average daily census = 

180.0, percent EMG = 66, and percent scheduled = 70. To determine 

the optimal number of beds, find the occupancy estimate Of 96.3 percent 

from Fig. 2b using number of beds = 180. Then, 

NB = 180.0/(96.3/100) = 186.9 

This rounds to 187 beds. Figure 2b gives the occupancy estimate of 

96.5 percent for 187 beds. Thus the second estimate of beds is 

NB = 180.0/(96.5/100) = 186.5 

This again rounds upward to 187 beds, the sequence has converged, 

and the optimal number of beds of such a facility is 187. 

The results of this study should not be extrapolated to small 

(fewer than 40 beds) facilities, which should be simulated individually 

sinc. percent occupancy is extremely sensitive to the number of beds 

in the facility.

- 241 - 

HANCOCK Discussion 

ET AL These results ma> be used to determine occupancy factors for 

bed-planning methods. In comparison, other current planning methods, 

such as that of Shonick [4], the Hill.-Burton formula, and the 

Poisson sizing assumption, overestimate the number of beds needed in 

most cases and are thus inappropriate. Shonick's methodology is the 

same as the one in this study, but his model is different, and his results 

cannot be put in the format used here. The lack of a call-in algorithm 

as a census restorer will cause Shonick's method to have substantially 

lower maximum average occupancies under identical cancellation and 

turnaway constraints. 

In Fig. 5c the results of the present study are compared with the 

Hill-Burton formula, the Poisson assumptions, and the results obtained 

by Hancock, Martin, and Storer [12], who used a similar but 

less extensive approach. It is apparent that facilities can operate at 

occupancies much higher than those predicted by Shonick, the Poisson 

asssumptions, and the Hill-Burton formula. The exceptions occur 

in cases where percent scheduled is high and the number of beds is 

less than 75. In these cases the Hill-Burton formula predicts a somewhat 

higher occupancy than is possible. Both the Hill-Burton and 

Poisson modeis ignore important facility parameters that determine 

maximum average occupancy. 

Figure 5c shows that the occupancy curve derived by Hancock 

falls in the same range as the results of this study, but its different 

shape is attributable to the fact that Hancock used different turnaway 

and cancellation constraints. His constraints were set at two cancellations 

and two turnaways per month. Thus, in small facilities, two 

per month represents a large percentage of arrivals, whereas in larger 

faciliiies the percentage becomes smaller. This explains the "flatness" 

of the curve derived by Hancock and also serves to point up the 

sensitivity of the turnaway and cancellation constraints mentioned 

earlier. 

When sizing hospital facilities, all important parameters and characteristics 

of these facilities must be evaluated. All parameters must 

be consiaered collectively since their effects on occupancy are not independent. 

The results of this study may be applied to the sizing of 

individual facilities although it is important that factors not dealt 

with in this paper also be taken into account. Specifically, if the turnaway 

and cancellation constraints differ from those used here, one 

must expect the occupancy to differ as well. Other factors such as 

scheduling pattern, seasonal variations that cannot be smoothed by 

admission scheduling, and (to a lesser extent) mean lengths of stay 

should also be considered. 

REFERENCES 

1. Magerlein. D.B., W.M. Hancock, F.W. Butier, G.M. Mallett. and D.R. Young. 

HEALTH New systems an mean real savings. Hosp Financ Manage 32:10 Apr 1978 and 

SERVICES 32:18 May 1978. 

RESER.CH 2. Magerlein. D.B.., R.J. Davis, and W.M. Hancock. The Prediction of Departmental 

Activity and Itr Use in the Budgeting Proceus. Report No. 76-1. Bureau of 

Hospital Administration. University of Michigan, Dec. 1976.

- 242 - 

S. Facilities Engincering and Construction Agency. DHEW. Representative Construction 

Costs o1 Hojpitals and Related Health Facilities, p. iv. DHEW Pub. 

No. (OS) 73-6. Washington, DC: US. Government Printing Office, 1972. 

4. Shonick, W. Understanding the nature of random fluctuations of the hospital 

daily census: An important health planning Lool. Mcd Care 10:118 Mar.-Apr. 

1972. 

5. Hancock, W.M., D.M. Warner, S. Heda, and P. Fuhs. Admissions Scheduling 

and Control Systems. In J.R. Griffith, W.M. Hancock, and F.C. Munson (eds.), 

Cost Control in Hospitals, pp. 150-185. Ann Arbor, MI: Health Administration 

Press, 1976. 

6. Johnston. C.M., W.M. Hancock, and D. Steiger. The First Three Months of Im. 

plementation of ¡he Admissions Scheduling and Control System at Delta Hospital. 

Report No. 75-9. Bureau of Hospital Administration, University of 

Michigan, 1975. 

7. Yannitelli, P.F. and W.M. Hancock. Implementation of ¡he Admissions Scheduling 

and Control System into Charles Hospital. Report No. 75-3. Bureau of 

Hospital Administration, University of Michigan, 1975. 

8. Handyside, A.J. and D. Morris. Simulation of emergency b. d occupancy. Health 

Serv Res 2:287 Fall-Winter 1967. 

9. Young, J.P. Stabilization of inpatent bed occupancy through control of admissions. 

HospitaLs 39:41 Oct. 1, 1965. 

10. Balintfy, J.L. Mathenatical Models and Analysis of Certain Stochastic Processes 

in General Hospiltai. Doctoral dissertation, Department of Engineering, Johns 

Hopkins University, 1962. 

11. Hamilton, R.A., W.M. Hancock, and K.J. Hawley. The ¿Admissioun Schcduling 

and Control Systen: The Admissions Simulator. Report No. 75-1, Bureau of 

Hospital Administration, Univernity of Michigan, 1974. 

12. Hancock, W.M., J.B. Martin, and R.H. Storer. Simulation-based occupancy recommendations 

for adult medical/surgical units usiiñg admisions scheduling 

systems. Inquiry 15:25 Mar. 1978. 

SIMULATION 


OCCUPANCY

#12 

Estinating the Need for Additional 

Primary Care Physicians 

By Arntbmy Hinddk, Nicholas Dierckman, Charles R. Slandridge, 

Harry DVlchcr, Raymond Murray, and A. Alan B. Prisker 

A /uems approach i¡ used to ea^ the primary health care delivery 

system in Indiana. The outpui (ofce viit) of prinmary care physidcana i 

estimated and compared with the demand for her r vicea. Indexes of de. 

mand, aupply. cou, and need are derived and used to determine the addi. 

ton u aumbr of priry care phycanada n d in ac urea. The raesul 

ol thb study are belh used to encorage aduating medical tudents to 

pr~ in ua in naed of additiomnl prima medial care. 

PRimuy medical catu has becon an are~ of- oonen because of 

the ppa deere in the umber of primary medical care proid. 

e m specdly g8eral and ahmily pantido;ne inD tbe United Sancs 

and spdcklly in tdie ut of lndiana [1-S]. A tchnique was dcvcl. 

oped la thib remebc to ~umr the primary health car delivery sy~ 

in Indima and to provide inmaIua regaprding the ned for addi. 

tiaN pd~ care phystam nla drcat areu of tbe sate. 

Dibcuuom bet~w qms anm mullns t Purdue Universiy and 

phbiu am at the R]qenrieef Instute for Health Care revled that 

li ord or r the a modl be ushlf, it parame~t would haye to be 

dma b Ol actua data. Thus de modeling proca wu con. 

Uda~ed by tbe aivblM ky of dnu prduding for xample, considaea 

rb d lt e fie of Shnncim ma as Blue Cros/Blue 

Shied, M icare, d Medicaid. A discuion of thes modeling i¡nes 

can be formd n hnddg, Pri , and Delche [4]. 

Meha.d¡o y for Aein Primay Medical Care Servie 

hPby M01a Cmar avk A s 

ac h in ll n [S,6] have dscribed the distribution of 

pTimay care enics by using rmii of population per primary carc 

TkL rs--- r& sJ mppaorad by had Rngguari~ IgJfutaf, 4 divion of thc 

Rrsr,. .uJtUo~. c,. IlpdU.~ 

dd _amuum n~ r rad f sa er r~rInu lo Charlu R. Stdndri4e. 

imt. Pons, C e.l ,M~Lw. Ssteu D~vls¡i. Unrsitb ofl 1o°, 

¡m* Cal, 1. ,I 1 HI i udb U· akclmr in m* d.~mmta of opverU~om 

ALTaU m *,Sl V.ItIU > e rLmmau~; N~k Dkr~ .b k da mlesm ~aIonm e-. 

u av M Ld by T.e*sk. 

8|l 

MNd~icl lJ lorvmin 

7 Dur . de a lnur 

SJstrs CorporAlJio; 

eJl ume,,u, nm. Umwnivn; R~ymnd MUvU 

p1refullr a4d ec mmi . t d rsn0 of mId~nMe, MicAhigin SSi.e Uni. 

nwry; ed U. 1 1. pJ r. r proao~r in hA S~hool olnduurial Engincrnng. 

SM l~?74WO~r//(1oc/;T. oo/o 

O 197) H eial !facr uF d E.dumaloal Tnuri

-- 2ti4 -- 

physician in a county. However, a political unit like a county may 

not accurately represent a medical service area since it is possible for 

cities of subustantial population to be located relatively doe to a 

county line, thus providing medical care to residents of contiguous 

counties,. or lor physiciana to be located within casy acces of people 

lrom another county. 

In this study, pimary medical care service arcas are defined as 

population cetera and their environs, specifically, towns or cities with 

population of at least 2,000 (the approximate number of people per 

primary care physician in the United States [7]) plus thc people in the 

immediate environs of these towns or citie. The environs are defined 

as the area within a five-mile radius of the town or city since 

that is a reasonable distane o travel to a physician. If the boundary 

drawn around the city overlaps that of another city, a combined service 

area re uluL Acording to this definltion, Indiana contains 79 

primay care atate's rc. The counties are grouped into 12 districts. 

each of which u divided into primuary care center and the residual 

rural area surrounding he centers (lee accompanying figure, p. 292). 

F himain cthe Capadty of the FulLTime Primury Care Pbysdcin 

A f~ e priamuy are r P) phylan is defined as a general 

or fmmly practition the age group S3S9, wh!-h he, the highest 

utput in term O visitu per year fOr this speciaty 8]. (Tables 1 and 

2, p. 29.) Prmary care phyiciam are condred to comprise nonfed. 

aeralo a~bed .hyidam, boCd medical anad ostophic, who are 

gen ralmd * aMaily pationer, pediariciba, internis, obstetriciangYo{o{ 

or ger~ l faoo The output of one ~FPC physician 

a d~mdas he nusber o prma are visits per year for a general 

« fnilyS prncdt r in the a ge group 5-9. The output of phypiciam 

wt oth r ad spedlty charalt b expresed as a percmte~ 

d tbe output of tde IMPC pbydian. Vitsa per week and 

we~s praciem pr yar by an FFPC phyuidan were computed from 

Rc

Pri c= arvicc a ia Indina. 

- 245 -

- 246 - 

Table 1. Relative Numb of Primary Care ViLt iuFOR u 

ft, Di~rent Speciahi~ FHYSICIANS 

SViddty Weeks/ Rclative vliitf/ 

day yedar year 

Genral and hmlly .. 3931 

Internal medLcine .... 2095 

Ped latrtic ............ 3.1 

47.73 

47.20 

47.77 

1.000 

0.24 

0~25 

Obbanrñ.gy ldog . 2714 

CaGe n l urgry ..... 1873 

47.70 

47.0 

0.i2 

0.4M7 

E' l lted Ir.. Ddkber. Ra>tckh, and Murray 19. 

Et tlmated (rom Rrfernce Dato o, the Profil of Medical 

Prt"rc [-lll. 

incom of $15,000 and over since it is amumed that this income group 

has all of iu expreed demand satisfied and would be able to bear the 

full monetary cos of receiving care. In 1971 people in this income 

group visited all phycians at the rate of 5.1 viits per year, and, of 

thee, 3.0 viiut were to primary care physlcians [15]. 

For he purpoS of this study, the ideal population i3 also defined 

as having damoep phic characriuica similar to the population of 

Indiana and it la sumed tht geogrphic acccaibility to .. m-a care 

(pentage of the potentlal demand that is expreued ae allowing 

for die comt &njd iíuavenlnce of traveling to a phypician) for this 

group l tihe ame au the average geopaphik accessbility for Indiana, 

whkch rus esablUb u a 0.9373. Thc demand rate for primary care 

ot be ideal population n thbn be calculted u 3.0/0.9373 or 3.20 

visiu Thaefore dth yearly capacity of an FTPC physician i 8,762/ 

320 or 2700 individuals of the ideal population. With this informa. 

tion, an algorithn has been developed for obtaining indexes of pri. 

ary care delivery for the previoudly defned service areas. 

Tbl . ui~ve Nu~be d ViI Atene~ by 

Giernal Und Faiy PCra of D ut 

Ae G~ 

(Sore: 5l~dr~ et aL [U8 

yer 

Unde 16 0.12 

vI39 UA24 

45-49 0191 

55-69 9.10 

¡(1~0-em6( amasI~ 

Owd~~~-BO~~.~~ 0127dP ~ ~ 

~FALL 

~~1978 

Ore.. _ . .~~~~~~~~~~~~~~~~~~~

. 1' - - 

IIINDLE Algorithmn for Determiuiang Pámu-y Care Service Index 

£T1 AL 1. Calculalc hce Equnwlnit Number of FTPC Physicians in Each 

Service Area. The visits attlilAde by each physician can be expressed 

relative to those attended by tLe FTPC physician as a function cf 

age and specialty (8-12,16-18]. For comparison among specialties 

(Table 1), the number of visits attended per day is estimated from the 

Delcher, Raykovich, and hMurra.y rtudy of Indiana physicians [19]. 

The weeks worked per year are estimated from Reference Data on the 

Profie of Medical Praclice [9-111]. The days worked per week are assumed 

not to vary by specialty. 

Similarly, the output of physicianu can be compared among age 

groupa Standridge et al. [8 report relative values for general and 

family practitionen in Indiana (Table 2). Age and output are assumed 

to be independent of spccialty because the available data do 

not allow estimation of the relationship between age and output as a 

function ofl pecialty. 

The output of cach physician in a service ares, expressed as a 

percentage of the outpu of the FTPC physician, is computed as the 

product of the relative number of vuiiu of tlose in the specialty and 

the relative number of visita of those in the age group of that physi. 

cian (see Tables 1 and 2). Then. the number of FTPC physicians in 

a mervice area is the sum of these producu for all the physicians in 

that ure 

Physician extendern. that is, physician asistants and nunre practiuoa, 

acr not included in the model. Standridge [1] estimated that 

thee providen supplied leu than 0.S percent of the primary arre visit 

in Indina in 1975. However, it would be easy to extend the model 

to include thee províders of primary care. 

2. Compute a CGographic Acce#ibility Measure. Geographic accembility 

is defined as the percentage of the potential demand that 

becomes expr~eed demand after allowing for the cost and inconvenenc 

of traveling to a phyician. The accessibility measure is derived 

by auín g that &lU the people in a Mevice area are saved by the 

phyic amn located in that area The simplest case is the mall, appraitnmely 

circular, single.phyican area. If the population is randomly 

located within the area, thbe minimum expected distance to the 

pbh ui given by locauing the phydiclan in the center of the area. 

The appropriate model for this situation was developed by Eilon, 

Wauon ydy, and Chriitod [20]. 

The problan becores more difficult ii the area contains more 

than ae pbys~ n loca~ion, but, lor most areas, use of a nearestlocation 

algorithm produces a set of polygons with the physician cenrally 

located in each polygon. The area of the individual polygon is 

estimated a the arca of the wrvice area divided by the number of 

physidan locationa. Although a precie result for expected distance 

HALST to the phyaician it not pouiLlc, a lower bound is given by assuming 

«zMUN dcircular subareas. Experimenul procedures and a sensitivity analysis 

lCd to the selection of this pro"dure [21]. Thus the expectdl Cartesian 

distance to the phytician in miles is given by

O.667(A/(wL))I 

where A is the area of the servce area in square miles and L is the 

number of physician locatio;s in ihe service auea. Expected distaníce 

was etimated for each revice area allowing for the different types of 

practice--olo, partnerhip, aid group [22]. A convernion factor, estimated 

to be 13 [22]. was employed o convert expected distance into 

expected road distance. 

Lacking informatin specific to Indiana, we anumed a simple 

linear relationship betnen distance and demand rate. Using data 

obtained from a sudy by Kane [2]J, we estimated the sope of this 

line aud ausumed that the rclatiozhip observed for rural Kentucky 

would also hold for rural Indiana becae of similar topography. This 

set of assumptiocs led to the following equation for ,eographic accessibility 

in a service ara: 

4CCESS, = 100 - 2.5D 

where ACCESS i the geopphic acceibility in tervice area j, expresed 

as pe ntge, and DJ is tbe pectd~ road distance to physidcum 

in lerv ~ j. The lMerpreta~on of the ucembility calculatian 

b ti 100-ACCESI b tbe percentage of potential demand 

tha doaes n bcoe exprced demand becaue of distance to a 

physidan. 

3. Com>ap #e ],ct ofl c omic St tusw on Expresucd tcmannd. 

A ponaran :J Ph. potental demand kfr primary medical care is de- 

Bflead by hc moaat ourn of obulning r The ecoaomic fecai. 

blit7 of ob~ning cue b defined oa the br of restricted activity 

dap par ymr, mui tr daa uad p~ccdurcs of Weds and Greenlick 

[24]. As down la Table 8, peraina in low.inmme famillie ned more 

mM~i l am thih pan In amll~ with bigher jcmnas It b as. 

anme d (mnd tbis maumptou ba borne out by the data in Tabl 3) that 

the hgber a fuity LaoCe. the arg the percentage of needa (po. 

tental d and) ezxpea ua visita to physician. It is further assumed 

TaS 8 Ceompma f _ dmaic Pembili Index 

ldana iicv 

o / , P restzLCted ~rlbilty 

NED FOIL 

PRaIMLRY CARE 

PHYSICIANS 

- 249 - 

HINDLE that all tlie primary medical care needs of those perons whose family 

tT AL incomes are $15,000 or greater arc expresed as demands for physician 

servics-that i¡a, economic feasibility is indexed as a functijn of their 

satisfied demand for ervices (»e Table for a]culationf). 

The following logarithmic function is used to calculate the eco. 

nomic feasibility index for each servic area: 

FEAS, =oax In(l)+ b 

wlhere FEAS i¡ the index of economic leasibility (expreued as a percentage) 

for service area j, I is average family income in service arca j, 

and a and b are cosants to be atimated by regression analysis. The 

values of 1 used wer tbe midpoints of the income categories except 

for the highest category, for which $15,000 was used. Uaing the data 

in Table 5, regression analyas yielded values of 0.274 and -1.68 for 

a and b, rapecaively, with an lS of 0.955. 

The interpretaion of these calculation ia that 100-FEASj is 

the porio of tb potential den d in a service are that does not 

bcoome expreued demand because of the monetary cost of primary 

hcalth are. 

4. ComputM Ihe Potentlh Demand. Potential demand ui the ned 

for primary halth care as expericed by the population. disregarding 

factor that could interfere with satifaction of that need [112,1]. 

This tbhe demand the population would place on the provider of 

primay cdre the cost of obaining such care, both monetary and 

no me cty , ee aro and, of course, auming that the population 

l uniformly able to reco~el conditioa for which medical care ia 

Potenial demand bi conderd bere as a lunction of two charactaik 

of he populatio: age aMd need for car (15, 25-29]. The 

idml populuoua aiunaed to have the ame age distribution as the 

mute of Indiana. The average utilization rate for the ideal popula. 

tion of ldiana (

250 - 

Table 4. Primary Cre Ulization Rates by 

Patient Age Group 

Visu/penon/ 

Age grup r« 

(4) ((.) 

Under 18 229 

O18r 2 

Over 64 4D0 

· ii i Bi i~~~~~~~~~~~~~~~~~~~ 

The mean number of restricted activity or bcd days for the popu. 

lation of the service area lb ic dewmined as the population.weighbted 

average of restricted activity daya over the income groups. The mean 

number of retriced activity days by income group is estimated from 

data ia ref. $1 and 32 snd prented in Table 5. 

i T eThee calculatioau for age d family income are used to deter 

mine the poateial demanud in numbers of the ideal population in a 

given service ar by the folowing equation: 

ve, = N, X (1/o) x (x/Bo) 

where Yot is tb potencial demand in unitu of the ideal population for 

rvce area j and NI b the ictual population in wervice area j. 

5. Dc:..íin: he; Exprcued Dcmund. Al of the potential demaid 

bowever, will not be expre~ed as visiu to physiciau. Facton 

that may impede some people frn making visita to physicians must 

be tuar fito -aaat [18). Two such factor georaphic acceibility 

and econoaic kfalbilty, are ~ite into accIunt in tdh model and are 

auumed to act depently of one another. Thub the expresed 

deAnad iD a serviC are, VY, is obtained by multiplying the potential 

dena i that & ares, Y, by tbe a:eibiliUty and economic feasibility 

factor. tltarea a com caputed in seps 2 and 3: 

Y, = ej, x FEAS, X ACCESS, 

Tibe 5. admad ~ c d ky Dhys 

of Pathemtn la DI t ame Cm p 

Eniad a~ed 

Incrue aouy mvty.dap 

Under powny kwlt a0. 

Owr poway idl aid 

¡e a $15000D 14. 

15G000 ad over llJ! 

* rjasdd frk de b a Ddm.b* D~ I31) ud C.û o1 

to U.u: ¡1f0 p1. 

t *u. Cn_ .f oL.U~: 19f0 (o . 

NEFD FOR 

PRIMARY CARE 

PHYSICIANS 

FALL 

197'

- 251 - 

HINDLE 6. Computc Mhe Availability Index. Availability is defined as the 

rT AL percentage of the expres~ed demand that is satisfied by the suppliers 

of primary care, thit is, that becomes satisfied demand. The latter is 

defined as the amount of medical services actually given by the providers 

of primary bealth care in responwe to population demand [15, 

33]. An availability index is costrued based on a model of the 

services that flow arou the service area boundariea. The basic assumptions 

of this model are: 

1. If the phyiciam have spare capacity after all the expred 

demands of thdi individuals in thc ame service area are satisfied, 

this spare capacity is made avalable to the adjacent rural arca. 

2. If people in a defined service arca can receive bectt service in 

the rural area, they will make demanda on the primary care physi. 

ciana located there. 

3. Persons do not coua district boundaries in order to obtain 

primary medical care services. 

Tbe availability index is defined as the ratio of the supply of visita 

available in the sevice ara to the expreud demand. The supply 

available Ia ic initial supply plus (minus) the amount of supply 

transferrd into (out of) the rvice area. The availability index for 

a service area (AVAILj) is computed by the following formula: 

A VrAIL, = (MDi x P)/V, 

whele MD a tbc number of FTPC phyidan in area j, P is the capac 

ity of owne FTPC physidan in teru of numbera of the ideal popula. 

tios aid Yt ex pr ed deanandin service aea j (the total number 

of phyiia n visits or the yer). An it¡raive procedure was developed 

o Mind the miber of FPC phy~dm in service ea j udng the 

tbree auumpiuas dacri be awa aud applied to each of the 

db.ua Im dpcadzaalty. 

7. Comnpulk tkl Primn, Care Sericc Index. The primary care 

cerl index lfr a evice ae (PCSI) is defined as te ratio of thc 

sttald Ad to bc poteCtial demand-tha t ¡s, is it the product 

of tbc exa deand demmmand the availability index, divided by the 

potential denad expreed a a percentage: 

PCSI, = (Y x AVAILj)/V,, 

This index i¡ thus the percentage of the potential demand that is 

atidiied. 

8. Comte khe Phtician Requirement and Need Factors. The 

pbhykin ruequirment fMctor, XMD, is the number of additional 

FTPC phy~idam that would be required in order to aatisfy the expeed 

demand t i not 

m ncucrendy being met: 

IHALTH 

UVA xZ 

XMD,=(v,/P)-MDj, 

where V ¡ the expr~ed demand in savice arca j (the total number 

of pbhyicin visiu pr year for the aa,. in numbenr of the ideal popu.

- 252 - 

lation); MDJ is the total number of FTPC physicians in service area NF- FOR 

j; and P is the patient capacity of one FTPC physician in numbers PMARY CARE 

of the ideal population. The physician requirement factor for each 

district is the sum of the requirement factors for the service areas in 

tlae district. 

The need factor is the estimated average number of member of 

the ideal population who are beyond the capacity of each FTPC physician 

to serve but who would have to be served in order to satisfy that 

portion of the ex~presed demand that is not currently being satisfied. 

The need factor, XP/. is expressed as follows: 

XP, = ',/MD,-P 

The need factor for each district, XPd, is 

XP, = (XMD, x P)/MD, 

where XMD, i, the number of additional TPC physicians required 

to atiify that portion of epressed demand that is not cumnrently being 

met in district d and MD 1 is the total number of FTPC phyicians in 

disrit d 

Meru~ df Eecue of tbe Prmary Medical Came System 

ElEctiv«ee indexes provide a mesure of the capability of the 

primary a e povida e in ach servtce are to met the demand of the 

popultion of that rca. To aras the performance of the primary 

medial care divs sd>em of a state, a policymaker need satewide 

mnaMures of effectiveane Several maurcs of performance that can 

be deived fron the inkd t aure diumed below. 

Iudtxa Raekd to Minimum LmbL of Serice or Availability. 

Fomulon of 

tto n that tbe g 

tme mem of effecuiveai i¡ based on the a*ump. 

o oí he poficy~ that some minimum level of 

eitbe the ervi ~ de a va;iabity idex is obtainrL The per. 

cetage of arvie u that have index values below a defined minimum 

leved 1 one uh meaure. Alarnatively, a population.weighted 

average of index levd below tbe minimum level ould be used. 

Popuwlio.Seighek d ldex por tl Stc. This measure combine 

the indexes lfr the ~rvc ar by multiplying either the primary 

care ice ind or the avaiblty inde for a particular service 

ares by the fraction of the Mate*' populaion in that arca and sum. 

ming the produc of ll aras. 

Pkhyucin R gquiremnt Factor for thc Stal. This measure is the 

number of additonal FIPC phucians needr in the state to atíify 

all of the exprsed demand 

of the serve area 

and is the aum of thd requirement factors 

The Analy for Indiana 

The procedure developed in thi reecarch is being used to evaluate 

the primary health care system of Indiana. It was fint uced in F 

1976 by the Board of Trusaes of the Indiana Medical Diustribution 

Loan Fund to identify areas in need of primary care physicians.

- 253 - 

HINDLA Tablek Meamm of Zficvemn of the Prmary 

Medal Care Sm: lndin, 1975 

Ma~wr Value 

ALAITH 

s=vK:M 

Pe~rct merice amas wih 

srvke lnd bdow 70 43.9 

Prca ewvice aras with 

avnbllWty bebow 85% 0.8 

Populatobn-wdghted 

mkr ha (%) 69.6 

Popula~i.wdlghtd 

avaiblUlty ~da (%) 87.1 

sauwi. pbhdd 

requrrn factor 251 

Data Soura for Physicain and Population 

Data conning primary care pbyaician in Indiana and their 

cbahnaairs wcrc obuined rom Lhe Indiana Physicimn Proile [7]. 

Dma com.ning the dic and age distribution of Indimaa'. population 

were hb d oD tdc Indi~n Coun"y Popuhbtion Projections [301. The 

kvel of family incme in each crvce ara was derived from data in 

tie 1970 United Suite Ceuus [32]. 

Rwail for tbc Btac d Indiana 

Tabe 6 summri the meaures of effectivenes of the primary 

medical arm sym of Indiana. It hows that 45.9 percent of the 

avice area bad a service inde of ls than 70 percent or that 43.9 

p~ t oe thc svice arca* were meeting lesa than 70 percent of the 

poentdal demamd F\uermore, 80.8 percent of the srvice areas hsd 

phul dan avalability ~ndexa that were lower than 85 percent. 

The popultionweighted wrvice index shows that le¡s than 70 

Table 7. lh~y an ll equiteme and Need 

Fpaca for Ind DhIm 

Phda Need factor: 

Di~ requl~t Idal pop/ 

bctwor phydan 

I. Gary ..................... 

2. Stih ida ............... 

8. Fort Wa .... 

4. Laayet ................. 

5. Aam ................. 

6. Te ame .............. 

7. diampol .. 

ILicuo m t. ~ ............... 

g. I doid~ .............. 

0o. Lawmm~bua g ............. 

¡i. vanl ................. 

12. N Alb.oy ......... 

0 

45 

35.4 

8.7 

2.7 

21.4 

4............ .1 

6.6 

s.s 

.419 

28.9 

143 

348 

59 

62i 

261 

415 

748 

291 

220 

97 

610 

521 

i i ,

- 254 - 

Table 8. Service Arcas with Need Factorn Greater than 900 

Accfc Economlc Avail- Prim"ry Physician 

Service Aclt Need 

arta ~Ser*iac~ ribility ublmt fueasiiú rbilt ice e air. c ec-. factor 

*rre tndex inde index indx O 

Lawrence. 

burg-nrl ...... 77 79 37 22 11.8 4 418 

Fort Waynerurail 

........... 83 U 44 31 17» 8302 

Muton .......... 91 84 44 33 2.9 03( 

Angola ........... 92 79 46 34 2.3 506o 

New Albanyrurai 

........... 79 78 49 30 10.1 2 753 

Terre Iauterual 

........... 8D 80 52 33 11.3 2428 

l0oo0ote ......... 89 79 54 58 1.0 2291 

Zvamvlukrural 

........... 77 80 54 3 12.7 2291 

Hmuinuro ....... 94 84 62 49 3. 1 595 

ML Vcmon ....... 93 79 63 46 1. 1527 

Bol~ ........ 9 80 52 3S 113 I1475 

G~ry-t i ........ 84 80 66 45 9.2 139 

Ktn t~bnm ...... 89 81 67 48 0. 1 20 

iatal ........... 81 81 67 44 .820 

uval ........... 3 U8 8 47 9 1 215 

S F-o- - ........... ®e 77 71 so 0.7 099 

Lsil ............. 93 78 72 52 1. 1 m2 

percen of the ponti al demand can be satisld. The population. 

m~tged avalabilty indae ahows that 87 percet of expred demn~d 

¡ uatid. The value of the utaewde pbycian requirement 

factor ¡i 251, which shows that an additional 251 FTPC phyicians are 

needed in oard to meet bhe 18 prent of the expr~ued demand that 

i not a~ed Sincte fullm~ physician by definition provides thce 

mont viui per year of any phy~ln, the requirement factor may be 

viewd u the minimum number of additional physician required. 

ul fo Iai Dh ta 

The requirenmt factor and need factor for cach of the 12 Indiana 

district are presented in Tabk 7. The need factor reflects both 

pimuy care availability ad the extra work each phycian has to 

perL>rm if he o he ab to suisfy al of the capresed demand. Table 

7 shows tha 8 of the 12 ditu have need facton of 10 prcent or 

more of the capaclty of an F phyiscian (2,700) and four disricts 

have oecd facto d 20 paret ~ mare of thb value Thee result 

indicate that he demad for care placed on primar care phyicin 

in di.:' -o in aioU put of Indoiana snifsamtly exceed their capacity 

for providing such care. 

NEED FOR 

PRiMARY CARE 

PHYSICIANS 

FALL 

1978 

e

HINDLE 

Tr AL 

HIALTH 

falvCU 

- 255 - 

Table 9. Cumulative Distribution of Service Area 

Need Factors by Prcent of the Equivalent 

FTPC Capacity 

Lmee Number of Percent of 

pe t 1service 

a reas cervice arcas 

All ...................... 

O ....................... 

91 

40 

100.0 

44.0 

0-10 .................... 1 165 

10-20 ................... 

2-30 .................... 

40-60 ................... 

Ii 

8 

9 

12.1 

8.8 

9.9 

70 ................... -O 

90-100 .................. 

0 

S 

0.0 

SJ 

100-170 ................. 5 55 

ielu for Sevia Aras 

Values fo the gographic accessibility index, the economic feasibility 

index, the availability indcx, the primary care service index, the 

physician requiremnt factor, and the need factor were compiled for 

cach ervice area. Table 8 (p. 501) shows those 17 areas with need facton 

greater than 900, which is one-third of the capacity of one FTPC 

phidian. 

Finally the distribution of the arvice area need factorn in term of 

perct nuga of the iTPC phyician capacity (2,700) is thown in Table 

9. The tabla that ian ader to met all of the cpreued demand 

in the tate of Indiana, each physidan in 36 of the service arcas would 

hae to arry a paient lo bd at let 10 percent higher than hib or her 

esm ed maximam. In 17 of the service areas, ach phypician would 

haye to are for at lest 40 pernt nore patienta than his or her 

maximumw. Table 8 sbows that a a dditional number of phyidamn 

equivalnt to 11 FWC I phyaicuana required to meet the exprsued 

demad ln these 17 service are On dhe other hand, 40 service areas 

wer found to have no nced of additional primary care phyiciaan. 

Applk dof aria Idiana 

BeCginnig in Apil 1976, thc ramurch has ben used to usist the 

manben of thbe Bard of Truac~s o #he Indiana Medical Diutrbutba 

Loan Fund to determine the aru in Indiana in greatest need 

of addtblmal primary are phyici . The fund is a loan-forgivene 

pom that rwppar medical students who agree, in exchn~ge, to 

practic primary mdical care in an aea of m ed. Studenu funded by 

this pogram recive a list of the ar in Indiana in greatest need of 

additol primary care pbyucia, and medical graduaa select practice 

lacaom jintld with the loan fund board of truAte lndiana 

has about S00 m gaduates per year, and 95 studenu have par. 

tidpated in bis prnuo~ to date. Sixtyfour of these are in postgradute 

tralin and ave a prici n phypia The remainig 26 

are itl i medial c oL It b hoped tha this research will continue

in helping to rectify the maldistribution of primary care physicians NIED FOR 

;! ~~~~~Indiana;~ . MPRIMARY CARE 

PHYSICIANS 

REFERENCES 

1. Standridgc. Prima., Jielthi Care Man power Projeclions and Polic AuLcu 

men,. Matr' tha4 Shool of Induul ZngJEcring, Purdue Unlerity, 

Aug. 1977. 

2. 1970 Americn Mcdil Directory. Ch~lop: Acn Medical AoclatIoa. 1970. 

3. 1973 Ameri¡cm Medll thcctory. Ch~ko: Acra~ Mcdial Amaton, 1973. 

4. Standridg, -R., A.A.L. PRiL~r. and H. Ddch«r. Ier in the ddop t of 

a health mapower pnlnng ~mod. Simhdtion 31:9 july 1i7a 

5. Axdroo. S. &nd . Murray. lndiarn Pthyiu aud Rural Healh. Rq lf 

Instutute Rtpon. Rtgnx Innit.e, Indanpolls, IN. Oct. 1974. 

6. Boruff, J. Dldbuat¡in of phyl~dus ISBH Bul 76:3 July 1972. 

7. IndiaMn Phtkien Proile. Seven Area-Wlde Comprehbivc Health Ptmnning 

Ag~es o Indiana and Ocalth Servccs Manlag~gt. Inc., Indianapolls, IN, 

Dec. 1975. 

8. Suandrldge. C.R, C. Macal. A... Prkáer, H. Dclha, and . Murny. A Sim. 

atoo Mlodol 1 te Pra~ y Hk~l Cre S o Indian. In IIJ. H4hland, 

.GIL: iJe. ad J.W. Shmldt (edL). ProcdQ of tha Wfinr SimuhaUp 

CCrm.~, P. -359. GaltUsbus MD. Dec. 1977. 

9. CCLntor I hmloa Saian m dmrch ad b.Developme. Refer~a Do os fa 

Prkb of AMd~MI rtrfk, iJ7. Cha . Amian Mcdlcal AodItbno, 9P7. 

10. Cmnau m h svias~ IRe b and b RafStrK~n Daot o fthe 

uwg Mia Pre tica, 1973. cao: u n Mcdicl Amodaon, 1974. 

II. Care b§r lcalh crvioa iraach a sd DCevlopmL Rorf*nce Data on the 

P~ et A.'-c Pra, 1974. ica: m Mdal AoLbn 1975. 

¡L D_ d Ja, A. >e/ . Mé~cl Cm AdmhsUtá ¿ro: speif$ng Requrcn~ 

or H~ CavL C&bd MA : P~d Unvi.a:y Per. 1973. 

¡i Wa T. and 5. a~ml Sr . A 

ana o th e b derlnanu oC phydan 

utúb~mda. a.r? p~.'nted at the lOs anual meeticng o the Am~ean 

bic HL A c ~ m a sa., San F~l. CA. Nov. 1975. 

14 Ay. L Iami:e d amu barrri ta e u of neted ditnkl r. 

lar MYd Co, m 15:44 J~ _1. 

U5 N# fw b hl&yb ~ada a. P , Voluma, ud Intamil 

&Um Lat VlOd, U~Jbd SM~di-i1. Vital Id Haltb Sicstiat Pubeia-as 

S«l 10, No. 97. DUW ib. No. (HIA) 7f152L Wahizgo. DC. UV5. 

G- t~mri d 08 , Mar. 1975. 

10. A Pdl Puml~ Cutan Phls ln Natoa~ Dlos~o and Thapeuftc In. 

da. Vda. 4 No. Al, PlA: PAI AmArla LUd. Jume 1975. 

17. VaMbda, LG. Unbtdka M Uplda kS lSe« In S.G. Vahovic anud P. 

Ahbv (eda.). Rfenr cait D o r Ah M ot MJdlal Prcie, 197. pp. 47o0. 

~ p Am~eria M~eda AIodaex~ 1974. 

I t L ¡~ J. phidmi A . la ~o se rc~ Qanarllica id P er. 

tra. la §.. LJ W w . Ab. A . d G.A. Ityui (ed&). RlJrrnc Di oen 

th& i~ Af ~7 I073. 3.s~, pp W=M. a po: Am~eri n Mial 

19 Daa . , %DdId , d .L amay. A ur t ludIn Ph!r~u' 

iJ~k CGnJ,~i &%b Tea kka~n. uui litute pot.TL 

Rmk~r l c. lUiaa g, Nl. M. 1!r. l 

0 San, S. CN. Wa*Osmdo. G md í. s~Tdes DltrbtU ManaKuerntl: 

ilath~.imUrl odWUi amd P lyulJs Nie Yak: Hatnmr. 1971. 

21. ldk. A. mod N. Dlitua- M>owir fr Pimuryi Madl~ Car in In l o. 

Repmtld laUnh lekpart. PImr Insgthe. lndnapola IN, July 197S. 

22. Dklremaa. NJ. P r JediCl Cm i ld. Matl s thaci, Schod d 

,Idud l Rhcarcek~ Purde Um~dVsly. Aug 19¡97 

23. &aum. RL Dacbtlb d hbl ai r pioritm and acpcuationa~ ,m 

ruIa mm HML Jnm k R 4:142 Sumin, 1¡09. 

24. Wdm . CawnLkh Deisnaa o medial lar utlllation: The d. 

len do l dn ~d dia mc a acu t h e ik l cure si Med 

C4m 8i:4M N v.-D 19. FALL 

23. Aday. L mud La lUdahom. ThA UUtl¿tlobn of Siuth ~es: lndiu d 1978 

Co~rasL DHLW Pub. No. (aM) 7ISM0. Wad~l. DC: U.S. Goern 

-t iU Or~. Dec. 1973.

- 257 - 

HIiDLZ 26. Ado, An . ad L. A adIca Patrn d Use ol Health Serics. In HzL 

IT AL F eman. . Levine, ad 1LG. IRMe d(). Hidbod o1 Mcdl S.oloO , 

pp. 353.60. tngleeod (Wt NJ: h~mkes-Hal 12. 

27. Asdeno . O. ud J. Ne~ Sodal ad bdMduml denimu¡sau of medal 

oase utlLoai 1tihe Unlied Sate. YMUlabm Atn Pid Q 51:95 Winaer 

mPV~s 

HtlTl~l mVM 

RURAIM 

S. adas, L uad O. Ando. A Dmd o/ 1m Shiv as. Chlago: Unlier. 

dq dof Cbop P, 9in. 

29. li, T. .and L Wbhc. Yaoa re bid to te bw os bohald sias: An iater. 

dm aprl mve dy. oMR d Cwr 7:1U Mar-Ap. IN. 

»o. udIu~ Ce.um7 P~Ibam Pvo d, 19o1-20. s replmd for the adia 

Siute auid d H b ihe DhMl d a~d, d~ l ud B ,~ la.dl 

- Ualverity,. booion. IN. Jime ^1I. 

1l. NMJaml Ct~ br Habf Sa~t DblU Deys. Vital mad Hasth Sits 

da Pubaddb Sedrs 10. No. 9. WaMbla,. DC: UL Govemaem dat. 

ins O~ 974. 

S32. U.& U of the Cesmm Cme~ o PoptdWo: q7O. VoL 1, C/mmet/lsla 

?l «p ~bu. hn 1 (ldl~). 

33. NiadJ C~ctur Hoe Ilib hadbl Ndia~d Ambulory MYdial Cre dur. 

W7: ]JB a YsMeddole. Vitl uad Hatbh Saatdi PubUclaos 

er kr No. ci. DIEW b. No. (HRA) -IIS, . Wmhbiga DC: US. Coy. 

ezurt 0dadng OEc, 1978

MANAGEMENT SCIENCE 

Vol. 19, No. 12, August, 1973 

Prinled in U.S.A. 

- 258 - 

#13 

PHYSICIAN SUPPLY AND SURGICAL DEMAND 

FORECASTING: A REGIONAL MANPOWER STUDY* 

A. RE[SMAN'*, B. V. DEAN**, A. O. ESOGB1UEt, V. A(OGARWAL**, V. KAUJALGI'*, 

P. LEWY: AND J. S. GRAVENSTEIN** 

Thie paper discuases the methoda and the results of a forecast for the demand 

for operative and obstetrical procedures anid the supply of aniesthesiologista in 

Cuyahoga County, Ohio. The techniques anid resulta of a ten-year forecast for the 

deniand for and supply of aneathesiologists and auxiliary personnel in greater Cleveland 

are dlicusaed. Several regression modele were used to forecast asupply based on 

popuilation, number of physicians, and the income per capita. The demand models 

wcre based onr population, age and sex distribution projections, and historical data 

regarding operative and obstetrical procedures. The results of these "objective" 

models were then compared to forecasts tunder tincertainty geznerated by a panel of 

experts tU-ing the Delphi Method. Alternative states of healti care delivery were 

investigarted and irnplications for future aimiastlhesiologist maanapower requiremenlts 

detailed. 

1. Introduction 

It is gcnerally believed tlhat medical manlpower availability is a kecy element in the 

never ending struggle to solve the nationi's healtl problems 113]. Matry regions are 

facing severe shortages of health care professionals. There is general agreement that 

persomnnl qualified to administer anesthesia are badly rnaldistributed; somce believe 

tbat cven today many thousand anesthesia specialists are lacling. W\itlh the average 

agc of anesthesiologists in thc nation bcing forty-five years and an attrition rate of 

30% over thc next 10 years, a definition of what constitutes a properly balanced 

supply and demand becomes important. In Cleveland n group of coneerned anesth(!siologists' 

addressed itself to the problem of current shlortages in anesthesia manpower. 

Under a grant from the U. S. Public Health Servicc, 2 the present authors in 

cooperation with the Cleveland Committee on Anesthesia i\[anpower critically iavestigated 

the current and future status of the anesthesia matnpo\we(r situation in the 

county. Specifically, the problems of supply and demand of annestlcesiologists ia 

Cuyalioga County, Olhio up to 19SO wcre investigated and models were developed:to 

analyze these problems. 

2. Supply Model 

The number of anesthesiologists in Cuyahoga Courity in the year 190S can be forecasst 

by considering different supply mnodels. These models are esseritially recgressiona 

niodels. PaLranieters for tíhese models were, estima:ted usilg data fromt 1963(-1969. Ia 

the study by Dougharty [4], a set of regression models for the supply of plhysicians in 

the state of Arkansas was developed. The models proposed lhere correlate thle number 

of phl sicir:is to the population and per capita income, wvith the data obtalined fronm the 

* leceived Junre 1972; revised October 1972. 

** Case WVestern Reberve University, Cleveland, Ohio 14106. 

t Georgia Instittute of Technology, Atlanta, Georgia 30332. 

lt alliol College, Oxford, I:nglarid. 

' !)irect .,". Anaesslhesia D)epartmentsl in ninre major Cleveland lIospital:i. 

sCuntra('t Nos. NIII 70-4033 a:id NIII 71-4022 (Anesthesi¢logy N:tlilpnwrei 

Period Jai'uaary 1, 1971 to Marc'i 31, 1972. 

Problerias) for thit 

('opyriglt ) 1973. Tlhe InatitutL' 1ft M iarneuent heieuces

- 259 - 

REISMIAN, DEAN, ESOOBUE, AGOARWAL, KAUJALGI, LEWY & ORAVENSTEN ;._ 

nation and the state. Since our work was confined to supply predictions for Cuyahoga 

County, two translateral prediction approaches were possible. One could either relate 

national figures first to the state and then to the county or relate the state figures 

directly to the county. A basic hypothesis in this study was that a relationship exists 

between the numbers of anesthesiologists, surgeons, and physicians in general. 

A. Model I 

This model considered population and per capita income as the most important 

factors which affect the number of physicians in a given region. Iistorical data on 

the number of anesthesiologists in Cuyahoga County wvere not available. However, 

the ratio of the number of anesthesiologists to the number of physicians was approximately 

constant lor the state of Ohio frorn 1963 to 1069. 'The average ratio was 0.038, 

with a range from 0.036 to 0.040. We therefore accepted 0.038 as a reasoxiable ratio 

for our projections and treated Cuyalioga County as a microcosm of the state. In 

Nlhdel I the nutnber of phlysciauiis ini an. year was assumed to be related to per capita 

inicome and population exponentially. Specifically, if P(t) = number of physicians 

in year t, S(t) = population of the county in year l and 1(t) = per capita income of 

the county in year 1 then P (t) = AS (t)"'I(t)cL, where t = X-1900 (X = 1960, 1961, 

1962, ... ) and Al, B", C, are equation constants. Note that both population and 

per capita income were assumed to be related to year 1. Choice of base year as 1960 

is irrelevant and does not change the results. S(t) = aIbl and l() = ab' where 

al, b1, a2, b2.are equation constants. 

D. AModel I (A ) 

This model predicts the supply of physicians; assuming an exponexntial relation between 

the ratio of per capita income of the county anud the per capita income of the 

nation. If 

P(t) = number of physicians/1000 population, 

I(t) = (t)/(), 

I.(t) = per capita income of the county in year t, 

1. (t) = per capita income of the nation in year t, 

then P(t) = Aj2l(¿)B, A 2, B2 are equation constants. It was assumcd that I (t) is 

exponentially related to ycar t, I(t) = A 4tb ' . 

Now if D (t) = number of physicians iii the county in year t, and A (1) = (population 

of the county) X 10 a then D () = P(t).'A(t). 

The population of the county vwas assumed to be linearly related to year t, as 

A (l) = a 3t + b3, where a3, b 3 are equation constants. 

C. AfMoel II(B) 

A similar model can be built wherc the estimation of the supply of physicians was 

from the state to the county which is called Model II (B). If 

P (t) = number of physicians/1000 population, 

I(I) = I()/,(), 

I1 (1) = per capita income of the county in year 1, 

I.(t) = per capita income of the state in year 1, 

thien, P (t) = AI (l)''"' :Ld D (t) = 1, ( (). l).11 

The remaining symbols are similar to those used in Model II (A). 

D. Model III 

In this model the supply of pliysiciaiis was assumed to grow expollc,,tially with 

year t. If P(t) = number of physicians in the county in year t aud 1 = X-1900 (where

ANESTHESIOLOGISTS 

PHYSICIANS 

- 260 - 

PHYSICIAN SUPPLY AND SURGICAL DEMAND FORECASTING 

·-"- DATA =j FORECAST - 

192 5000 MODEL 

MODEL n 

í84 4800 4184S00 - MODEL m 

176 4600 40 - MODEL 

68 s 4400 

160 4200. 

152 4000 - 

144 3800 - 

136 3600 

128 3400 

1963 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 

CALENDAR YEAR 

Fiouni. 1. Supply of AnIesthesiologists/PhyYicians in Cuyahoga County up to 19SO.* 

X = 1963, 1964, 1965, - ) then P(t) = Al4e" where A4 and 1¡4 are equntion constants. 

The detailed analysis of the models is reported by the present authors in 161. The 

results of thlese models are shown in Figure 1. 

It can be seen from the graph that Model III forecasts 182 anesthesiologists, tho 

highest, and MIodel II (A) 164 anesthesiologists, the lowest estimate. The predictions 

of Mlodels II(B) and I with 170 and 177 anesthesiologists fall within these values. 

The estimation based on time gives the highest, those based on per capita income, thc 

lowest value. 

Accordinig to these models, in 1980 there will be a 12% to 24 % increase in the number 

of anesthesiologists since wec counted 147 in 1969 in Cuyahoga County. Thce average 

amnnual rate of incrense in the number of anesthesiologists is 2 to 3 ancstlesiologists 

per year. 

3. Age-Sex Specific Demand Model 

In order to leffectively plan for the supply of manpower aLLd other resources it any 

professional discipline, it is necessary to have a means of assessing the future demand 

of the variables of concern. Manpower demand in any discipline over a period of 

time depends on demography, industrial mix, manpower utilization, and cconomic as 

well as social factors operative in a given comnmunity. Since these factors are subsumed 

in a model predicting incidence of operative procedures in any arca, any 

meaningful projections of the demand for anesthesiology services in Cuyalhoga County 

would normally include elaborate consideration of the demography of the area. This 

part diseusses the development and application of a deremand model for anetsthe.sia 

Personlrel bas :d on the demAnd for surgical procedures. In particular, an age-sex 

Specific m odel is considered. 

* Th app:trent discoittiruity at 1969 occurs hecause diffcrent regression mnodels werc uaed on 

the }listorical data from 1963-1969.

- 261 - 

REISMAN, DEAN, ESOGBUE, AGGARWAL, KAUJALGI, LEWY & ORAVENSTEIN l 

Let the populatioól be grouped by sex and age distribution, where the age groupl 

0-4, 5-9, ... , 80-84, S5 years and over are represented by j = 1, 2, -, J - 1; 

respectively. A total of 18 age groups were considered. Tlie surgical procedures weri 

similarly classified using the standard 16 international codifications (i.e., Neurosur. 

gery, Opthalmology, etc.). The following matrix of demand model was then devel. 

oped. '' 

Let A' be the matrix of surgical procedures for males and A' be the corresponding. 

matrix of surgical procedures for females. Tlifs distinetion is necessary since past stud. 

ies, (1] for example, indicated differences in demand indices betiween the sexes. The, 

cells of the matrix [alj] then give the coefficient of demand and represent the number 

of procedlures of type i performed on males. i¡ age groulp j in Cuyathoga County.in 

1970. Similarly, cells [a{jl represent the number of procedures of type i performed on 

females in age group j in Cuyahoga County in 1970. In performing the calculations 

we must consider all values of i and j where i - 1, 2, ... , 16, j 5 1, 2, ... , 18.-. . 

Let b" be a vector giving the distribution of males in Cuyahoga Counity in 1970 and 

hJ be a vector giving the distribution of females in Cuyahoga County in 1970. 

Thus, b" number of males in age group j in Cuyahoga County in 1970, and 

bJ = number of females in age group j in Cuyahoga County in 1970. 

Similarly let d", d t denote the vectors of projected distributions for males and 

females respectively in Cuyahoga County in 1980. 

Define a vector pw as the ratio of the projected male population by age group in 

1980 to that in 1970 in Cuyahoga County. Similarly define pl for females. Thus, 

(1) pj' = d="/bf' 

and 

(2) p/= d //bl, j = 1, 2, ... 1S. 

We must then compute a vector (X) = (X'4, X') defined as 

(3) (X')= =[A(pj"), 

(4) (X) = [i,' (p/), 

wvhich was obtained by direct matrix multiplication. 

It is clear that (Xj) gives the projected demands for surgical procedures by males in 

Cuyahoga Countty in 1)SO, with (X/) giving the correspondirng Iprojectioiis for females. 

Thus, X"' gives tlie projected number of surgical procedures of typc i to be performed 

on males in Cuyahoga County in 1980, and X/ the similar projections for females. 

If we define N" as the total number of surgical procedures required by males in 

1980 and N' thie corresponding figure for fcmales, then N"' = 'I[ X"'" auid N' = 

-!, X,/, hence NV = N' + N / is the projected total number of surgical procedures 

required in Cuyahoga County in 1980, if it is assumed that the current demand patterns 

continue till that time. 

The data for matrices A"' and A l (follo win¡g pages) wcre collected from tlhe QUEST 

division of Blue Cross of Northeast Ohio. The data cover 79% (estimated) of all 

surgical and obstetrical procedures performed in 1970.3 This factor was used througlout 

the initial phases of the calculations and then adjusted accordingly to accoftut for 

its partial coverage. The matrices A' and A' are shown in Tables 1 and 2 respectively. 

The data for vectors b' and b t were obtained from [10]. Thie vectors d ' and d t giving 

a In Nurtiheat Ohio.

- 262 - 

PHYSICIAN SUPPLY AND SURGICAL DEMAND FORECASTING 

:he age-distribution of the population in Cuyahoga County in 1980 were not readily 

vailable from any known source. Consequeutly, estimates were obtained by comparing 

tlie population distributions of Cuyahoga County and the whole U.S.A. Data 

;rere collected from [8]. 

Twvo types of projections from this report, namely Series B and Series E, were 

rniployed. The basic difference between these two projections is due to the assumption 

under Series B projections of an average of 3.1 children per female iin the popuistion 

and for Series E, an average of 2.11 children per female. Employing these two 

projections leads to two types of results for the expected number of surgical procedures 

for botlh the male and female segments of the county. 

Statistical tests were first applied to the null hypothesis: "Thic distribution of the 

population of Cuyalioga County, by age and sex in 1970 was the same as for the U.S. 

in 1970." The null hypothesis was accepted on the basis of Kolmogorov-Smirnov 

Nonparametric test and Chi-Square Goodness of Fit test. 

'lhe vectLmo p'" and p/ were calculated using relations (1) and (2), and lthey are 

:hown in Table 3. 

The total iiumber of surgical procedures in 1970 in Cuyahoga County for males Nias 

51,199 and for females (including obstetrics) 96,229. Hence the total number of procedures 

for both sexes was 147,428. The estimated demand for surgical procedures for 

males in 1980 is 57,666 by Series B and 55,941 by Series E projections. Similarly, the 

,stimated demand for surgical procedures for females in 1980 is 116,041 by Series B 

and 114,606 by Series E. This estimation predicts an increase of from 9 to 13 percent 

in the surgical procedures for males and an increase of froin 19 to 21 percent in surgical 

aud obstetrical procedures required by females ia the county. hlie results of this 

analysis also present data on the projected number of procedures for each surgical 

category up to the year 19S0. These values are shown in Table 4. 

Th'le models thus provide us with the demand figures for Surgical Proectlures for 

cach category of surgery. While these figures are interesting in their own right, the 

greater aim of this study wvas the derivation of the number of anesthetics rcquired by 

the population in thc county in 1980. Consequently, to obtain this number, we proceed 

as follow-s: 

Let a¡ = the conversion factor giving the number of anesthetics required by the ith 

category of surgery. 

Rational estimates for the values of ai, i = 1, ... , 16, vwere obtained froni [21 and 

are shown in Table 4. The total number of 132,575 anesthetics in 1970 was then computed. 

According to [71, there were 149 anesthesiologists in Cuyahoga County in 

1970. This implies that S90 anesthetics werc performed on the average by an anestliesiologist. 

Using the above conversion factors and the precomputed projected number of surgical 

procedures, the number of anesthetics estimated by Series B forecasts in 1980 

is 156,227. Under these assuniptions and assuming identical workload in 1980 as in 

1970, the estimated number of anesthesiologists NA demanded in 1980 i¡S given by 

SA -Number of anesthetics in 19S0/ 

Number of anes thetics per anesthesiologist = 176. 

Similarly the number of anesthetics estimated by Series E in 1980 is 154,824. -lence 

the nu,L_ -:- f anesthesiologists demanded in 1980 is 174. Thus tlie assumlption of 3.1 

children per female in 1980 as opposed to a figure of 2.1 accounts for a diffcrence of 

two aunesthesiologists dem',rnled ini 1980 as provided by thcse niodels.

- 263 - 

1350 REISMiAN, DEAN, ESOOBUE, AGOOARWAL, KAUJALI, LEYV & OGIAVENSTEIN 

O 

Z» 

o04 

O 

.a 00,1 

- 

. 

c G 

A 

*0 

O 

le 

u0 

: 

C 

u4. o 

elo 

-U 

6> 

-e 

Ci. 3..:. 

e; 

E:a 

a. 

al 5.Z 

_ 

es 

u. 

s a. 

o1 

o 

l 

*0g so o! ,> -lo 

lo _O o e>-4 ç v O 

LO bo 00 .4 w el CI 

lo0~ =00c~e§ c4 C§ C 

_ -I - X _~- 

s ~ N4 ~ X ~ N 0 &4 1_0d oOXt ew 

a co ma 

| R: tO; N* °>2- °°4 

|.< °O, z- O> E - O C> - 

Cl c Ni 1 cI. C 

N 

'~c~ R^: o oo O" 

Z> ¢ O O u 4> P 

w 

_ o c oN * &.>. O _ cac r_ . 4>Oo 

I> c·O ,~ ~E c ~ 

b-> G G 4 0._CO i 

o 7UC? 

- - 

JI 

F - 

fi 

.> es 

w 

b 

O 

3. 

4> 

B 

.0 

f 

3.. 

u 

e- 

CO 

.0 os 

g 

n 

O 

O 

ue 

- 

V 

O 

"c 

3 , 

o 

* 1 

6> 

u. 

o 

a 

3 

o 

O 

O 

C h, 

lO 

f~2i 

vl c 

>4 - 

CE.d 

n>4. 

.= .]? 

E.gg 

_ _ 

O - 

o O , 4'O

01 

O 

·L 

C 

O 

'3 

a- 

E 

LS O 

LS 

5> 

*LS 

6 

·- 

LS 

C.s 

r: 5> 

u-. 

5. 

a. 

-u- 

u. 

s 

2 

o 

1 

- 264 - 

PHYSICIAN SUPPLY AND SURGICAL DEXIAND FORECASTING 

01 -> 

0s c" 

Otg3 

t_ aC m LO

- 265 - 

REISMAN, DEAN, ESOOBUE, AGOARWVAL, KAUJALGI, LE\WY & GIIAVENSTEISi 

TABLE 3 

Ratios of Prcjected ¿Male and Female Populalion by Age Group in ': 

1980 to that in 1970 in Cuyahoga County 

Ma-l . Ytm.ls '. 

Age rup * 1

- 266 

PHYSICIAN SUPPLY AND SUIRGICAL DE.MAND FORECASTING 1353 

4. Conclusions 

Both the supply and demand models presented above are primarily projection 

,dclIs and the results can thus be compared. The demand for anesthesiologists under 

rreat practices has been predicted to be from 174 to 176 by 1980. The range for the 

.,pply of anesthesiologists in 1980 has been estimated to be from 164 to 182. Tihe 

;iults for the supply level are somewvhat less than the demand that is expected to 

1jst for anesthesiologists in 19S0. 

The main limitation of the above models is that they neglect the etíects of techno- 

,qical, legal, ethical and economic changes. It is, however, conceivable that these 

etors would influence the supply of anesthesiologists and the demand for surgical 

:;*ccdures ald hlence the demand for anesthesiologists ini 1980. 

To incorporate various states of nature that cuuld conceivably occur ten years 

:.nce into the prediction for the desired level of anestlhesia personnel, another model 

i iorecasting demand under uncertainty was constructed, which is reported in detail 

,1)i}. In that model, the subjective opinions of a panel of cxperts were used. The 

n;nel included individuals affiliated with a representative group of iistitutions or 

.ganizations, and having expertise in all areas that have a bearing on the problem 

n.der consideration. The panel was provided with factual data and information ob- 

,ined from the supply models. Tlrough several successive rounds of a DI:LIHI 

*.ercise, consensus of the experts in the health care area was reached on a set of 

.iestions concerning the future. The questions were coacerned with the following: 

:.e attrition rate among anriesthesiologists; the probabilities of occurrence of various 

aures over the next decade iii regard to tihe quality of aeasthesia cure and the num- 

:.r of anesthetics per capita; thec perceived requirements of anesthesia personnel 

*ith the change in the quantity of anesthesia care assuming no change in the popula- 

'.ju and the number of anesthetics per capita; and the possible uses of the new .class 

anesthesia persorinnel currently in training at Case West&en Reserve University 

::d elsewlhere. Based on the results of thce DELPHI certain predictions were made. 

was observed that in 1980 there will be an estiniated need for about 50 per rent 

:7,re anesthesiologists, about 50 per cent more anesthesia residents in training, and a 

ficienit number of trained nonphysician personncl in anesthíesia comprising a work 

't' of about 70 in 1980. Through the DELPHI exercise, the constraints thlat con- 

.ibute to thegap between current and "ideal" nianpower levels in anesthesiology were 

., identified and ranked in order of their importance. 

The DELIPHI method predicted a demand for 19S0 of 35 anesthesiologists more than 

' mraathematical projections. A high demand would exist for anesthesiologists if ali 

mproved level of health care were provided by 1980. Thc DELPIII model considered 

':i aspect explicitly anld in fact reaclied a consensus on the figure of 0.57 as the 

:robability for improved quality of health care with 10 per cenit inere:ase i the 

-umber of anesthetics per capita by 1980. The projectioin mnodel, on the otlher hand, 

=ade the preilictions under the current state of health care delivery, hence its predie- 

'ions are lower than those obtained through the' DELPHI. 

Acxxs>, .. l;.%t:xNT. .. ThI e present work wasl stipported under cointract a gr.iit, nid iide.lo .vailavbI? 

.:ughi N'iI 70-4033 :rild NIli 70-4002. The Ruthiors aire greatly indebted tu IDrs. J. P'otter, N. 

-Piero, S. Katz, S. Ko'v:ac, II. Iretlinhmer, E. Malia, 13. S:nlkcy, J. Smart, J. Viljiern; all Inelil- 

.t u lihe CleveImld C,ímilxittee on A':o^theslilogy .Manpowcer, aln to I)r. M. lRhteCn, frr thelir 

lklatnee, cooperation, nnd nssistance i¡n c',ndcting this atudy.

- 267 - 

1354 REISMAN, IEAN, ESOGCUE, ACGGAIIWAL, KAUJALOI, LIlWY & OIIAVEtNS1E¡ú1 

Ilarefertnce 

1. AuL:iR, (1. It., "Ana Analysis of Patient Load in a Medium Sizo Community Ilospital, M.LH 

Thesis, Techinical ltoport 1.1028.02, DIepartment of Industrial lnginmeoring, Syracuse ITi. 

vorsity, Junte 1070. 

2. Communlications with l)r. J. Potter, Chairniiasn, DIepartment of Anestlhesiology, Clevelah 

Clinic, Cleveland, Ohio. 

3. I)>:PUt:no, N. G. :S. AL.,"The Clevelaid Pilot Program ili Aneathesia Education," Marymouat 

llopital, Garfield Heights, Ohio 44125. 

4. I)ovHAHRTY, L. A., "The Supply of Physiciana in the State of Arkansas," Rand, l1Mq3i4 , 

APC, August 1970. 

5. (ClAV,:NSew:um, J. S. et al., "Analysis of Manpower in Anesthesiology," Anealheciology, Vol. 

33, No. 3, (September 1970). 

(. (iiAV:NaST:lN, J. S. et al. "Analyais amad Forecasting of Anestiestia Matapower in Cuyahoga 

County, Ohio," Techinical Memorandum No. 259, I)epartment of Operationa ltcéeareh, Car 

Western lteserve University. Mareh, 1072. 

7. " lealth Itelated I)ata," Nortilheat Ohio Regional Medical Program, Part 11, Sections.V, VI 

antl VII. 

n 

4. "Popllati o 

Is-timnates and Projectimns: Projectionri of Lilo Popildationn of the United States 

by Age and Sex, 1970 to 2020," U. S. I)epartment of Commerce Service, P-25, No. 470 (l#-. 

vcmber 1071). " ' 

9. It*:l*M*.\N, A. et al. " Analyai anId Codification of Pro- Lird Prwt-Operativo Anetlhesia Related 

Tasku," Tochniical Memorandum No. 229, Department of Operationis Itcsearch, Cae 

Western lReserve University (June, 1971). 

10. " Recport of Anesthesia fManpower Coinmittee," New York State Society of Anesthesiologist' 

Comrmitteo on Total Anesthesia Care, 1970. -: 

11. "ttatistical Abstracts of Ohio-1909," Econoinic tResearch I)ivision, I)cvelopment Depart-. 

menot, State of Ohio, 1970. 

12. "Statistieal Abstracta of the U.S., "U. S. Buroeau of Census, 90th Edition, 1909. 

13. ST:WAUT, W. H., "Manpower for Dotter Hiealtl Services," Public Heallh Ieport, Vol. 81, No.S 

(May 1966), pp. 393-396.

- 268 - 

A Spatial-Allocation Model bfor Regional 

Health-Services Planning 

,Villiam J. AL-:rnathy and John C. llertshey 

Stanford Unwersi*y, Staniord, California 

(letccivc~l A\utgust 12, 1'971) 

In planning for health services, the need arises to determine the location, 

capacity. and number of health centers for a geographically defined region. 

The present paper formulates this problem in a form convenient for solution 

and presents results from the model to clarify some important aspects of this 

allocation decision. The planning region is assumed to consist of geographically 

defined subareas or census blocks, of known location. The population is 

stratified in such a way that each stratum exhibits relatively homogeneous 

patteras of health-care utilization. The model characterizes the effects of 

center locations upon aggregate utilization and utilization of individual centers, 

sad gives optimal locations of centers with respect to several alternative criteru. 

An example illustrating computational feasibility and the implications 

of various criteria for the location decision is presented. 

HE PROBLENl, OF' determining ipatial locatioxns of primary health-care :enters 

for a dispersed population is an important issue in primary health-care planning. 

This paper forniulates the problem iii terms oi four different criteria. 'The importance 

of accepting census statistics as .source data is recognized in the formulation 

·and in the choice of optimization procedure. 

The distance from individuals in a community to sources of care is a recognized 

' ' t 

barrier to utilization." "1 Several types of decay functions have been proposed 

to describe the inverse relation between distance to medical facilities and utilization 

·M these facilities.l " '"- Recent research by the authors analyzes choice behavior 

among facilities when several facilities at varying distances are available. z'l 

It ¡e 

demonstrated that the influencc of distance upon choice behavior iN much stronger 

tairn ik the influence of distaince ulpn overall utilization. 

In addition, it is well known that demographic, socioeconomic, and health- 

*-tatus factors all influence individuals' medical-facility utilization. Strata that are 

?differentiated by these factors show different utilization patterns as a funetion of 

distane .12.'," That ,''I , is, distancc is a larger barrier to utilization for ~some 

strata than for others. Substantial progress has been mnade in understanding the 

e oomplex interactions of distance and socioeconomic variables. However, a method 

for using these interactions to suggest alternativc locations of medical facilities has 

not heretofore been developed. 

A frequently stated goal of regional primary-ihealth-care planners i¡ to allocate 

facilitie to locations in sueih a manner os to provide as much of the primary care 

Idemanded by the population as possible. Y'et this may conflict with other goals. 

For example, if the health planner maximizet total utilization, centers may be 

' placed closer to areas that contain large numbers of people for whom distance is a

- 269 - 

William J. Abernothy and John C. Hershey 

stronger barrier to utilization thau to areas with populations that are more mobile. 

Such an allocation of centers may be far from optimal in terms of thc goal of mini. 

mizing the average distance patienxts must travel for cach vis4it. 

The point is that alternative objectives i¡n thc spatial-allocatiosi problem will 

often eofliict. Any proposued louationie of health centera mnust be atudied i¡s Il ,. 

light of cach objective. 

This paper pres*enta a model that can a.rist a hcalth-.yístemn planncr in dlett.'r. 

mining the location and capacities of a specified number of health-ciLre (.'iuirs 

within a region in a'cord:uiec witi ritated criteria. Thc formulation reli(e4 ui,t, 

0c,',wuS df nmakilg tnruleTffs among multiple objective.s is almo explol*rld. 

Ans e xamphl olTerinig ,mnutirical experienice with the model ia givwn that

U 

u 

- 270 - 

Regional Health-Services Planning 

.: 

t: 

.zW 

9 

.i 

ti: 

¿C

- 271 - 

William J. Abernathy and John C. Hershey 

SUlppl>se ; prilmiary care acilities (or ueniters) r'e [O ebc located in the area. WVe 

dein,. dist:aince betiwveen two points it tihe ar'a tis the rertangular .separltion It ll,. 

iumni of the vertit.al arid hlorizonital distanees) betwee¡n the tswo points. Thir is 

,'nlr. Iiit inorf rl valid l1h:.i Ihle dirret, distani e I)\ !we', 1lw twvo pointis. siart, i¡ 

¡livitidualsl -will tr:ivel iiglh\\tways i tihe tegelmr l per value of the power hlas iol Iue)ct empirically establisheid ior 

the general eaise, and various valued are reported.1"`1 The present formulation assumes 

a power of 1 for simplicity. The numerical method used for assignting fatilities 

to locations places no constraints on the use of different values, should they 1íe 

appropriate for a particular application. 

Wlhen more tllha osne renter is included withit a planieitg region, tlie ¡ormula tion 

inust be extended to recognizc the effect of distane on bioth the user's choice of a 

parti'ular ceiter among those available and the chalinge iii overal3 utilization thiatl 

r.sults wh(.n nu,'iío crínte í lrs ar: availal>l. 'l'lu .ecffet ofl distac, oa"c it" iliza.ioa i. bIt t 

meswured iii term.t of the proximity to thie clo.sest ceniter. ''lhe nminimunim dist:ilíec 

from a cen sus block i to a center is defined as bi niln,-i, ... .. b,,. 

A us.r's choice of a particular cente.r from among tliose availaltle is stron gly i¡iflueticed 

by tlhe proximnity of the cent.er to the use'r i rlatioí to til, proximit y ofC 

* tilher :',.te.rs. Rtece!int re.s.arcrh by the presi.nt tiittlLíori sh,o' tsli.it a regressjim liiO

- 272 - 

Regional Heclth-Services Planning 

distance can explain over one-half of the variance in the percent of all families that 

choose a particular center. tl t In the present nodel the choice of a particular center 

k by individuals in census block i is represented by the probability Pn.. Thiis probability 

is conditional upon the specific spatial relations among centers. Thus, if 

uij is defined as the total number of visits to center k by individuais in stratum j 

in census block i, then 

uiik Pin ije//( 1 + dib ). (:3) 

The particular probability distribution function tihat is ut¡sed in the numerical 

examples discussed later is as follows: 

Pa=- 1/[ _,'k' (bi/b,,) ]. j (b/l>bk). (4) 

This .tari tliat r the pirol)bility of c hoosiing a I)articular center k is a funetion of ithe 

distance to center k relativc to the distance to the nearcst center, or bi/bk. Theo 

probability function is formed by normalizing thiis ratio by the term k a" ( biJb), 

to satisfy the constraint .,- _ Pk, = 1. 

The specifii functional relations that will bhest represent patient behavior in¡ the 

choice of a facility j

- 273 - 

William 1. Abernothy ond John C. Hershey 

maximize or minimize a specified criterion. 

tives. for a regional planuer are apparent: 

Several reanonable cla~.,es of objev. 

1. tMa.o.niúc utilizatiopn. Mlaximize the total utilization of prinary health 

services i¡u the region, i.e., m:L(xU. 

2. JlinimUize dlisiace per capita. Tlhiu criterion mighlt be represmalted i¡ two 

differenit -wa:ys: ta) .\Minimize the average distaxnce that cucrlh individual in the, 

comminunity mu.st travel to reachi the enter of choice. (b) .\linimize the' averag,, 

dista.l'e per :apita t ttlihe (lo.s-t te iter. Tlhe latter alternative is choiiine for su.l 

~.lK'uent an-:lysis. .\[athenumti('ally, tlis t.ales the form minl( E.': :,' )./b"n 

:. .l litimiz,. i slitlce per c/ i. ,tinmizi the average ix.r-visit travel di.t:mn", 

to tlic n(arest eenrter, where all visits as computed in equation (2) are inchlulld. 

i.e.. miáiZ:-L £ - . m i,b/l'). i An alternative statemenit of thi.i ,riterion is t(i 

minimiz, thi- tvf.r:ige li.st.l:,. trav.hvdil IK r visit to thel u.ier's e, t(,Lr of choice. TI,. 

lirst vertsion i.s 'omsidre(d preferable for nmost applications, howvever, imice tui¡ 

minnimum reqluired distanve exlusçive of c*hoiec better represont.i the effeet of di.tante 

ea a barrier. 

4. .lin&iilize percent deleyralalion in iutilizaíion. Thisi eriterion minimize.s ¡hei 

averag

- 274 - 

Regional Heolth-Services Planning 635 

requiren:mitas, a in e.:haaliasm for exanliniing tlic' in:renmental gaitis to be expected fro)m 

adding ad(litio(:: l e(enter,. alid a ii'i:1s to e)xplore th( 

tívc cntriteria upn solutions. 

it(ierentia:l (ffetets of alterna- 

Suppose thatl the hcalth planner can) select a criterion, atnd wishis to determine 

the optimal loc1ationri of G primiary caree ciitters. wliere O· cea:u r:tg. frLrom on, upa to 

some maximum fiasible ¡number. (If somne trieuters :air(,ady exist iti thc region, lle 

may want to hiolhd these illociations constant anil conisider only the optimal locationis 

of additiot.id centers.) By using the model, lie eaní compare the expected incremental 

gaiiis in thie c-riterion valuti for ea:h ah:itiotiual center. If tlie lhoicje :umong 

sevcral olthjttives is unia::tr, lihe prorlss of allor:nting Celiters can l)e relp.:ated to 

maximize eacll objective. for vach value of G. This facilita:tes a compari4on of 

different locations in terms of their relative erfec:ts on muiltiple objectivec. 

Suchla :a us( of thi( mod(el is illustr:at(d by 1lW( h xa tinl)el it thie followiltg ;se(liot. 

The ,i))ln'ai':atl restilts iar. o|r t:'hi):l for a conill)letc, ont-lille, ilnl era:ctive, FUILTUAN IV 

version of the mode-l, programmein to be suitable for several hundred census blockli4, 

and up to 20 centers and 8 substrata. The program determines the values for all 

TAISI.u.: 1 

UTiI.i.%iTioN C'iA.ll.\ ¢'TIr:tlIl'-T.^ 1O:A\II STIIA.\rUM 

Utilization rate at immediate proximity 

|litgh Low 

BIarrier effect 

tanee 

f dis-. Stl. ratum l j 

v, -5.2 

I¡ .Slratium j-4 

& ¿=l1.9 

d -. O.0005 d4-0.0008 

lligh Sratrtitmjj=3 St rat tiau j -2 

; =eli.2 t a2 .1 

: · ,Z,,o d.i lai 

115 

four criteria described above although only oune is optimized at a tinme. wse of the 

model provides understanding of the problem and insights into the impact of various 

cnrteria on the solution in a hypothetical example. 

AN EXA.IPLE 

AMsuYE TIIAT health care centera are to be located withii the regiom d.,cribed itn Fig. 1. 

The regiotl has 4 strata, with vi and di a.s given itn Tanhble 1. Strata 1 and 3 have manrkedly 

higher utilizatiotn rates al imtme

- 275 - 

William J. Abrnarthy and John C. Herhshy 

It is uieful to explore tile ¡inplications of thc variouu criteria for a given number of he:i(1h. 

care. centeor and then to examine the ellect of increa.sing the number of centers. Thle IfTawt 

of alternative criteria is the most obvious when one center is i>ositioned to satidfy the ent ire 

region'l; requirement. Figure 2 liows tlhe three differeltt locationa for onlul center that :ure 

T.,BLI:, I1 

l.,ATrlON. I.t)uXU, .xi)> POPUL.ATION uO C';:.SU 1B.ucLrs IY ¿VT.\lAT'M 

No. Coordinates Radius 

-1U.0 1 1.2 1.0 

2 t - I Iu.0o 12.5 1.0 

: -S.-8.7 1I2.(l i.0 

¡ -'. 4',j 1 1O.O 0.7 

, -11.5 10.4 1.0 

Ei -12.0 14.8 2.0 

7 -7.5 14.0 2.0 

t -6_. 3 i 10.4 2.0 

9 -10.2 7.3 1.8 

10 -13.7 1 9.1 2.0 

1 -1.2 11.5 2.2 

12 -3.2 j 3 5.0 

13 -12.5 4.2 1.8 

14 -12.0 -0.3 4.U 

15 4. ij t9.0 4.0 

i -8.3 -5.2 3.5 

17 i 2.7 -2.0 1.7 

I18 3.0 1.2 i 

2.0 

19 5. -3.7 2.0 

20 1 0. -4.8 2.0 

21 It.4 11.3 2.5 

22 ' 20.3 5.7 3.5 

!3 

2>4 j 

2 11.5 

10.9 

3. 

1--5.8 

1 5.5 

! 4.0 

25 j l.O -0.85 3.0 

21 ! 2.0 0.3 1.8 

27 24.8 --6.3 1.5 

't8 ' 18.0 -6.0 2.0 

29 21.8 - -3.3 0.8 

30 k2.4 -1 1.0 

31 22.8 -3.7 1.0 

32 20.6 -4.3 1.0 

StratunI 1 Straum 2 Suatum Stratum Slrau 

1O(Xi . i. 4;5 i 1;) 

634 1617 IX) ;t; 

j w33 

2 ;O 55 

1 7 

175 

! ¡? 14;x; 

¡0 llt ¡iuu 

i633 16i i (X}) .5;. 

1100 - ~0 130 454 

700 400 : 200 : ») 

tax,) 200 120 150 

ti}5 '.H1O 120 5lZ) 

700 T Ld0 1'25 ( ti) 

850 i J0 12,5 700 

i 512 1to o 338 j 5) 

44i 2i00 37 350 

34i2 200 1 38 250 

5)00 

iO0 

750 

112 

112 

112 

113 

113 

113 

2'25 

250 

167 

167 

166 

575 

375 

375 

450 

350 

700 

350 

450 

450 

450 

lU0 

534 

533 

533 

4;0 

275 S 

275 

250 

137 

288 

28 8 

787 

438 

537 

475 

850 

600O 

60O0 

6OO 

i 

.00 

500 

lictated by eriteria 1, 2, aid 3 with repeet to the three cities. The three dots indicate center 

ioatims; assoe~ited triangles identify respective criteria. The precise locations with resleet 

to the coordinates in Fig. I are given ii Table II. I)ata offeriig a comparison of each 

location with respect to all criteria, including criterion 4, are pIrovideti i¡u Table IV. 

Tlle tiree criteria yield eritirely different locations. Vhen criterion I (maximize utiliza- 

140 

150 

150 

150 

150 

350 

217 

216 

216

-¡' -l _ *-s 0 

- 276 - 

Regional Health-Services Plnning 

-50 

5 

. Center 

_/~ Criterion govcrning 

center locations 

O o 15 20 25 

gltJ1) ^^ B "'->)t 

\sQ~ty 

Location of I Center. for 3 Difzerent Criteria 

Fig. 2. Tlio e*ffe t úf eaeIh criterion iUpon the locatioii Of olle heaIlth *(entter. 

Center number 

I Wiil 1 I e'lter 

11 Wiith 2 eo.lotrs 

2 

1 

1ll With 3 cmleri. 

1 

3 

JV Wilh t ee-lhr, q 

4 

" ° 

V Witb 5 velder 

4 

TABS L! lll 

.%1) SIIEUI*'fIr:IJ Mn..MEII F* ( (':XTEI:. 

O. c Fig. I f.,r I,,e ,.,rs ..f ('lmrdil:Ie,) 

Criterion 

(1) Maximizc (2) Maximize distance (3) .Minimizedistance 

utilization per capita 1 per encounter 

y x y x 

21. j11 - 3.1X) U 1.$4 1 .2 - .71) 10.1)0 

21.-1 -3.7 I l?. I -3.;1M) -3.3 1i .)! U) 

- i .U4 .i 

.SI) 10.4 -9.U I .10..U 

'2.40 -3 .1 21.52i -2.78 22.30 -3.20) 

-10. tIl 10.40 -10.¿¡ 10.40 -10. 0 1) 10.40 

3.-63 - .75 3.iO -2.80 3.i0 ! -2.80 

2 .» 402 .) 3.r14j t 22.00 -:J.50 1 21.23 -3.04 

-- IU.;dU LU .4U - IU.IU ¡u.5O 

IU.4U 

3.59 

11.32 

-2.7& 

- 2.25 

2.0YJ 

3.76 

-4 

3.04 

3.1il 

- 11.3.5 

-2.70 

3.00 

22.40 

3. 24 

11.04 l l .li 

ae Delermir:d ,eo.ly ftr Ihe tiran eriterico. 

-3.10 

lO.lil 

-3.19 

3.21 

-L ) 

.1~

- 277 - 

William J. Abernathy and John C. Hersihy 

tiona) i , u.le, the ceniter is located near the center of C(ity C. 'G'lis is due lo the fItct thai. 

thi4s anra tcontains : large intuber of individuals for wlhom distutcec is a strong barrier. Aí. 

the satne time this choice of a locationi tends to inerea.se overall distance per capita vi-a-vis 

other criteria greatly, as can be seeni from Table IV. One might argue that this criterion is 

most equit:able, siice it favors tho*e wvhose utilization is mnust curtailed by larme travel di.. 

ta:'nes. lowever, a cotiter :argumneit i.s thlt the resliltinig location shlift* the cost nf trans. 

portationi to those who are less .cansitive to lLsitance barriers, irresipective of their ability to 

accept it. 

'ihe tlird criterion (minimnize distance per enicounter) lends to a location near thi * r'net r 

of Ihe latriist *ity (.I). This loeat ion i¡. traduliionial hut not Ipartii.ularly equitable. ()vt.r:tll 

TABLE IV 

CH.AtA.CTV:HISTICS OF P)TIMULM ' LOCtATI.O 

Criterion 

Maximize Minimize distance Minimnze distance 

utilizatiuon per capita per encounter 

Will) I t;ml. Ier 

Utilizalti(,m. U i 12,29 11,07 119,572 

)istance/capita ¡ 25.82 18.00 1 * 1.18 

I)ist riceencouiter 24. 8 16.5 1 14 .110 

!lldex-.'- deerease ill -utiliza- 33.2% 35.4% f :..';;. 

Wilh 2 ('elnters 

Utilization, U 146,891 143,237 144,195 

b)istlace,ecapitu 8.53 8.14 8.17 

Distane/eneounter 7.39 7.41 7.39 

Index-%c decreae.¡ in utiliza. 21.7T% 23.4% q 22.0: 

tiotw 

With 3 Ceiters 

Utilizatiowl. U 155,086 154,575 155,051 

Distanee/capitt 5.32 5.31 5.32 

Distaniee enemxinter 4.00 4.94 i 4.90 

Index-%; decrease in utiliza. 16.9% 17.15% 1 16.915% 

tint 

With 4 C:enters 

Utilizatioas, U 156,932 156,000 155,514 

I)iataneei~'r:ipita," 4.87 4.77 4.79 

Diaitance/encounter 4.58 4.43 4.39 

Index-% decrease ini utiliza- 15.84% 1i.26% 16.567O 

tion 

tutilizaticin is the lowest with this eriterion and the pereent reduction in averuge utilizttiol, is 

highest (criterion 4). This third criterion is appropriately desecribwd tas a local one. That i., 

it does not consi.tently represent needs throughout the region. In the present example, the 

optimium results from a simultaneous decrenae in utilisation in some arens and an increa.ce in 

uothers. .\s the center is moved away from distaince-.iensitive remote locI:tiont;, distance p er 

encounter decrea.es over the rasige considered. This is because utilization varies invers. -ly 

with, and at a higher marginal rate than, distance. At the same time, thbe center is moved 

toward large population unatenr so as to inereas~ utilisation at elo*e proximity. The criterion 

i;s analogous to the action that well-intentioned medical practitioners might take in 

independently locating their individual facilities-minimize the distance that their paticntl

- 278 - 

Regional Halth-Services Planning 

mwus travel. The e.'ssemtial pIoiit is thlat sirh a'l rule leads to nmuch diffcrcut re.sults tlhal are 

obtnine,íl with more global criteria. 

The .eeond criterion yield, a location tihat might be anticipated. PIer eaplitas distanee 

is miimiztired lien tlie center is located iii City B, betwen the two larger citieis. It ir interestinl 

to note that this mile, like the otlher two, tendx to locate ¡le ce eter in un area of high 

population ratliher tha: iii a s.par.sly Ilopilated 'elsus block. 

There is little difference in the effects of tihe three criteria uion locatio.as when tvwo or 

three eenters are allo:cated. As. lhow ti T:able i 1 for tlhe twao-reiter case, locatitu.s are near 

the cemiern of the two I'Lrage cities (.4 adml () for all itriteria. Siil'nil:rly witll tlinre eiEters, 

the! a e:msi-r.l hlly lo'ald .lr hel etr e f th rlenters l e e riles. 

The four-eenter ca.us, like the tione-center ease, aigain shows a distiset differece animong all 

tiree crinteria. This i9 depicted in Fig. 3. Three cetr are e oiiteit ly located iin the 

* Center 

/ó\15 \ ~/X Criterion governing 

-i C 10 

( \ t ) "/Ce Ce ntr O 

lo 15 2 

t tions 

Locotion o; 4 Centers -or 3 Di-s rennt Criteria 

Fig. 3. The effect of each criterion ulpol the location of four health centers. 

eities buit the location of the fourth one delcmlds r'ritieally on the criterion. The effect of 

each criterionI upon the location of thibs fourth econer is similar to the effeet of eacti criterion 

upon the locatiou of onne center. 

The diffelu.l ial effects of different criteria for hetalth -ervices tendl to dimil liu.-l the 

number of ceniters increa.-es. The ímpact of different criteri in a muel le&s whemí there are 

four centers than when there is only one. Figure 4 providei a direct comparisota of tihe 

criteria ai the number of centers is inraceased from one to four. 

The three separate graphl shown iii Fig. 4 pre.~ent the valuc of each criterionm wielm it 

merves uai the basis for optimizing the locatioii of frona one to four renters. Eaíh gmrapl li l.so 

provides a comparisaoí of this with the dlegradation in the same criteriomn measure wlen each 

of the other two erintearisa optimized. The eommol pauttern with each criterion ia a sigííificarlt 

improvement a* the ¡unmber of centers i inereaased above 1 but a marked decrease iii the rate 

of improvemerit a. more venterJ are added- a ntro:zg tetsdeic.y towvard dimiiishing rettirlat. to

- 279 - 

William J. Aberncthy and John C. Hershey 

Increase in distance per encounter when 

criteria lond 2 ore optimized (miles) 

o 4r 0 o 

(sgall.w) Jaunooua Jad aouos.!- 

Increase in distcance per copita when 

críterio lond 3 ore optimized (miles) 

t 2E 

u t 

eln o 

I 1 . 

~ I . Z 

Decreo'e in utilizotion when criterio 

Z ond 3 ore optimized (totol visits>x i000) 

P O o + , r - 

(o000oo1 u te.! ^ 1oo40) UO!,OZ!ll.n

- 280 - 

Regional Health.Services Planning 641 

scale. When niore fhau fotr i.e¡ter.s :te :all¿xated, thlere is little reltive inmproveimelit i 

any of the imtea.urcs, 

:uid liflerene.es among tilhe regíion-widc imeasires are negligible. 

The thrce graphs incliuded iii Fig. 4 deincm.trate the use of Ihe model in the inlportant 

aranlytic nrle of deternlining the incremental Ienefits from inrea:sin lthe nítiber of eenlers ii 

a region. The siecifics ouf :ai'dy.is ean Iwx varied to aicomiodaitc t he planning ie problemi at 

ligd. One or more centers cln be fixed in location or capacity to represFent existinig farilities, 

different criteria can be studlied, and miniimum or maxiiiiiini ecnter sizes can be included iii 

the formuílamtion. The moidel also gives vohlines bhy !.troat:s for eachl renter so thnt i¡.m'os related 

to liiirmln size iiol :ei a ofie ,.rf rai e:il. e:I,1w dealt wii .hit Wlilh sp'ilie refinlements 

can xie infrlimiruued to accolllxoditte a pIatticular plaiíning probleni, Fig. 4 illastrates the etylx 

of analysis that can be performed to leterniine increniental beuefits from a.dding nmore eenters. 

A health pl:iaitter can comniare tlie rost impllications of sugg.lted inimbers, sizer, and 

ocaLatioi, of center.s wvithl thl( mriiír.iiic:l Ibeefits ofof l ervie, wlhere IM-nefit, :Mr, 

inme.sllreu4 ii t1'm111ís (l . .l cMh 1 ritrio L :s ttilizaitiol tr',l¡'el di.,.lnse,, or per alfilat p ereent:íge 

decrease i¡n utilization oun a region wide hAt.-is. 

In the tinal aínalysis, the decision maker must decide how to weigh different 

objeetivw,$. No olnl( curiterion is best. Thc ultimate decision as to tlw number. 

capacity, aind location of health vueter moinst be hoised upoIn a broad range uf coínsiderations. 

However, formal analysis can be extremely useful in reducin¡g ihe 

number of variables that must be considered subjectively. Furthermilore, by 

forcing the mniodel to evailate decisions that are proposed on other hbaes, the model 

can serve' to qu 1itify th e'Oll(seluent s of alterliative choices for comparison with 

the optimum solutioins derived from explicit criteria. 

C(ONCLUS .ON 

Tg REOIOXA4L allocation problem is a central issue in research on health-eare 

delivery systems. This paper has undertaken to develop a method that can be 

applied praectically by a liealth plasittcr in bridging tlihe gap betwveen re.eareh resulta 

anid implementation. Emphazis is, on providing a formulatiotL tihat usefully 

eharaeterizes the problem within a region rather thaun on the elegance of optimization 

procedure!s. (As a benchnmark for eomputatioínal feasibility. cach optimization 

for the above example reqccuired about 20 seconds of CPU time onu an IB.M 

360-67.) The formulation distinguishes the separate impact of distance on various 

.ocioeconomie strata with respect to their choice of a particular center as well as 

overall utilization. Thc model can be extended to acecommodate considerations of 

inlxOrtaunc! ina partituil:r aPpplieatioris by eh:alngiuíg tlíi cirteria: uisl for oIplinliza:tioi 

autd the onllstraintls imposed ;11 the formnulationi. 

ACKN.OWLED(;GM'ENT 

RESEARCH REPOITED i11 this palper was 

supported in part by thé National Center 

for Health Services Rewtarch anid Dcvelopment, Departmeiit of Health, Education, 

and Welfare.

- 281 - 

642 William J. Abernathy and John C. Hershey 

REFKERENCES 

1. WV. J. ADEM.NATYI AND J. R. MoOORE, "Regional Pltulling of Primnary Healthl 

vices," .IUdicol Care, forthCloming (1972). 

Care S,'v-. 

2. - - AX) E. L.. SCIIii;M.s, ")i>tiSaIa alid llcalthl Service.: Is.U¡es of I.lilizaii¡O :andi 

Facility Choice for l)emographic Strata." Re.wearch I>aper l 19, Stanlford I'niver.it y. 

Graduate &hool of Busincs, July, 1971. 

3. E. S. BurVA A.> W. V t1. TA.UERT, "EvaluatioII of D)irect Conlí>uter S.arch .MellkI i";. - 

th( .\n.g:r i1at P'lmingiPK I 'ul)hll," m 1l ¿du. .I[Iu,('IIti»i nl¡¢., 193"?(6 1. lll, , 67. 

*1. A. ('IoCCo ANi 1. ALmA.N, -1eical Areas aSrvice únad 1 )istaluCes *Iravehed for l'bhyi.i;:l 

Cate iii WVestewn Pe.i .>ylva uia," Iublic Hcalth *iottogra:ll 

Service, 1959. 

No. 19, Ul.,S. I'ulliv 1 h:ill;, 

5. J. 1). FonllEs AND W. T. ZI:mImA, "EstimationI of Supernmtrket D)rawiig ')ow\r: Al. 1:\tensjion 

onf l.Oc:at; 'I Therry :ud I'nrii:tice," .Annanls Rl¿iontl Súi. (forthlIcling). 

6. W. L. G.AR-ISOX, 1). F. AI.nRLE, J. D. N¥sTUEN, AsN) R. L. MIORRILL, Sl.udies oi ¡n:l/f 

wag Deelopment and Geographic Change, University of Washington -'ress, 

Wa.hington, 1959. 

Seia:ttl. 

7. P. F. GROSH, '"rbanU Health Disorders, Spatial Analysis auid thc Ecounomics >f o[ h.ahl 

Facility Locatiut," Iaper presented at the 40th National Confesrreniee of the í ¡,,t; V 

TIOss RtUM.EAR.C SOCIET:Y oF AM.I;RICA, Anaheinim, California (Octoxber 28, 1971). 

S. 1'. llaggett, Locario.ial .4naludis in Human Geogragphy, St. Martin's Pre., 

1966. 

New York. 

9. R. HooKE .AND T. A. JEEVtS, "Direct Search WSolution of Nutmerical and Statistical Pruo 

lemrs," J. .ALoc. Computing .1achinery 8, 212-229 (1961). 

10O. P. J. .IHLII *a.oD R. L. IC.NA..MARA, "Tlle Relationi of D)i.:tance to the l)ifferential; l ', 

of Certain Health Pers>onel and Facilities and to the Extent of l3ed lllnes.," Rurai 

Soc. 17, 261-26.5 (1952). 

11. R. L. KANE, 'lI)etermination of Health-Care Priorities anid Expectations amunong 1tua:1l 

Consumers," Hcalth SCervice8 Res. 4, 142-151 (1969). 

12 . . Siu. X N. i, R. L. BA3aSHua, AND C. A. NIEZNER, "The Concept ouf Disuance U:t a 

Factor in Acesaibility and Utilization of Health Care," .lledical Care Rev. 26, 143-1Al 

(1969). 

13. J. A. SONQUIST AND J. N. MIORG.A, The Detection of iteractiun Ejfectsa Silurvc! Re.:Lrcl11 

Center, University of Michigan, Monograph 35, 1964. 

14. W. H. TAtEr, "A Search Decision Rule for the Aggregate Scheduling Problem," 

15. - 

Management &i. 16, 343-359 (1968). 

, The .Search-Detiion-Rule Approaeh lo Operalions P>lanning; Unpulblishliedl )octr:dl 

l)i.-,.rtation, U.C.L.A., 1968. 

16. J. E. WEiss AND .%i. R. GREEXLICK, "I)etermninuLts of Medical Care Utilization: Th' 

Effects of Social Class aud Distance on Contacts with thc lMedical Cnre System

How to determine - 282 - 

the optimum numbie ofoperating 

rooms 

Jay (;olhilmm. D).S,'. t(od II. A. K,,l)l)jhr' rh. r. I). 

11owv 'lmainy ol)icratimg§ rooiims do 

y)ol rcally ííe((tl? Is ii necessari y to 

acq

Decmand in Mintu és/Day 

Numnlber oi Olenj.tillg 1t l1lí1s 

Ca¡pacily ilt ¡ lllu.ie,' ]).i)'! 

Lua(di I"a' t.' 

A~···8 'tl-cluge aUviltf ).t \s W.,ilvd 2. 

:) 

- 283 - 

I. ,s(i, -/l ..... .... O.)(i )!X --- til ll>t ivel tile .i'c . '.,c ,r(,lliit (Lily d'e- 

:. iil .2 73.ii i:li 771-. * . tl (i,. - :;lilf :, 1 Ii t y.i itn7.. 

thiIh ¡xd \ o hl ¡t " lul 1 r. lOl)')ratil"rZ " 

TJ-i; I.I' 2. J"lí s,. 1,,l , ,iil ,"it 1,'Ii '1, s "Id,¡r 

u I.. l ,iír 

displays el..í'.c " t', 'icb, ' ¡or ou . ") fm'5 (

0r 

1.' 

* SI 

' 

S5 

II MAND 

it'Al 1. D'IAYI 

lOAD FA':IOK 0.S'4 

FOR 5 ROOM5 

- 284 - 

?240 

').7,s C,,il9.1 ! .%917 0.?3R U.lJ'b 0./7 1.0) 

I 'l(:(- 1 í;t11111 ai d ,, . ' i. ,',pl u 

a.s a tIIW'ti>ií orf . 0' 

for a demand of 2,175.1. minites 

per day, 

Ci : C, -- 2:!Sn (0.R5 - 044) 

365 (7 03 - 0 .) - 

for a dmernad of 2,27l4 miimntes pc' 

day, almd 

Ci = r, -- 250C: 0 .91 - 0.481 

365 (23.09 -- 0.621 

for a demanaid of 2,372.9 minutes 

per day. 

(Conftined on Page 1.98) 

Disposal System 326 

IIstlic I,',i,í l s,'i X\i,. il' iii, , i" S ;,,,ii; l 

\ll l ~l;ll,-r ('(Uxlllilt. ;I : 1", u,1 .. ;"1,mf 

t1r:l(.1h3 . c;(.d,1h "1 .(i ip. i. "I' Id, 

3 i "i ',,l,, "1 ~ ,If,' I"''' ll(1 ,) ;'1' 

¡ l'ti\ i:llh' li'r l! '{} %IN ('i tit i"i.:i 

i''iii i;lth Iil. slihiiil. ,i¡l, Ii "i"l 

r,. i, l r ih S' \'¡l.t ' 

ilftt ll¡i (h1'. ,111, 

!.,,, 1]',II ) .:!i *ij sl ¡t it ililh: I:, i l-;i " \;! 

-'-',' 'h.. t'llt;. .Am('ri( :11 ( ::, ( 'o. 

Respirator 327 

ilir. ;i v.hll,,( li7>i(: l. I!,u l it íl.ísh "' 

O"V·\ 1 

1- - . ' 

.,0. :La 

- ZO5 - 

''" '' l &l 'v ic .' :l3si *iiIl('(1,- i 'i,'\\l , l.io. 

- (';iltii. vh.,(i"i". uIhr ;,tiiio i;iI. h;i'.,1 

;l. i I r li l h ';l, sisllt Illaxil)';i 

;dal , ', v 11 t l I ; a . .l , 1 .1 i 1h, \h m j; l(,- 

: 111(' ;i ;11li.¡ 1I ,li'; t' i¡,\·. ( i -';l1;s. 

I;lIm,'llS, ILKi', 

Multiple-Seating 328 

'I'iu ii¡' \\i í.i lí,l - -' ,ili:'. . s t!'."i 

S i-th r.',írí; -I.'' - Sh hu. í 

wa -dl.' 

s('atin,« ;IrrI'ill"('l*l('llt. ,~ .t('m (xOi 

:i1.i lia)h. n i( 'i'ly ti ,t' I'a.tri' 

.\Nl i.i.h.s iilh t. 3 str .. rici l' 

.Xladlis.sl I"tr.cihntX I.dulusrie,. 

Blood Warmer 329 

Dll, iii il. til t u,,¡t',, l(

! 

21 i 

Monitoring Syste:m 330 

\ Ii' '\ i, i telielíl;.l11. %!'i', e i"""'"'l"i:: 

" ¡""I n \glil " 1"X 

Die11 111 il l 11s 1 i' ti11.1' 

i*Itijee . llleI iliít I li ' ;e ';lll 

,l ""il-. I th llee. .' i,' :li!lb ; 

i!'!:: 1l1.':li; l i1:' ;. li\ i!, ::,:e : -';l 

Urinal & Bed Pan 331 

/ ee,11',lelell'h'e te l'( .iil. Ii !l.le e l,. li.: i(. 

bm ('.5 - cd: ',j,:,I' e le l líle. lUei";;i. l (';e 

- ZOO - 

tlr-iilgl~ :a \l itle llal. lbamt. plrtevt lids 

.,ilí ,í.h , .c : i .. .1l rel lldh lie 

hr,.bh ,d "" dlt.1m, d 'r ¡l 4. ,," % ilt 

tille ilil, pliie'fd,. pl, íh';% (:oi. 

Recordinfg Titrator 332 

t 

\ lll :e , l' le rIle. (cc ll(lic,'4 litl;hte 

e' 1 l,,' i11,-1, il .1 ;ilet , e teece -. 

elli! e :lle . ele,*:e eti:, c 'l li ill e i ;l, 

: '! 'Ii .lll\ ::iti : "i , 1, .{ ,Ji' , : ,! .i: 

e(e l - 1 el leí . íee l * 11 1e i\ le-. % ei ! cl(í 

'l.l, hi'.1 ;,.c lll' ,cIf c i ''.' 1lA(i11 ll' ills 

tl"el ",el 1 

,i l, el le., Nlll. \11 n;Ih.ítli;'ls 

ill t'l iia 1 o I('" li- 

:1Si; . lut¢ini:cleil li.esIlu liii.lefil. eli-,. 

t $l.¡LI li 

Calculator 333 

.l : ,e:ie. ,s ,;" e ,. le 'le le, .eei l e .1. 

ililll: m, Ill .:Jill' ,u,(./,,~ ' ! , fl,, 

r",', ".,le S ¡jel l : 1- ee "; . . "1i. 

h1r .i I j|I :|| i i';I !l . ', , | j 

I 

;,feí í,-i,'e' ltli.,l e!S: h, :h'.i. .ll, hq, ,í 

Iltle ll ell , hlelS e jt e j e'. l-' l;ll:e e e l'i :- 

Jil",« 'i"",!, l'l'.]'.J: : i)J)'llli~llt \: llJl 

i 1 ' l !ll, %le'" .. :i "'t li.ll í . 1a 1 _ ' Ie 

i " l} y." 1 «' 1'ti a, 't ,X 1 . , ' 

.irilt tl e,(-ie l (.;t : e :ii e le :e l. -! e ¡ : i _ 

I¡.,lirablioriei. s it-. 

This new AAF horizontal Roll-O-Matic 

air filter is ideal for central station 

air-handling units or ventilating systems 

where headroom is limited. 

I _ 

11

i 

Trays 334 

Colorful pattenis anid obllontg, rowl(l 

;i.!i 

:, tJl(r-olustati' mixiny, \.iv' [)prcsct 

.. . 

JI., 1 

for 105' 1'. .edíl a :;vitch lthat penurits 

,n,:tuu|il adjí¡stin(,lt. (3) Tinencd hydrotilerapy, 

with sliut-olf ut any time 

pl tl :i() 30 ,imtcs. ({4) 'Iit(.l (leplty-. 

iníg an1 auto¡ilnatie: S)lt-Mlf of water 

pIunip. Ille Electric Corp. 

Ice Dispenser 337 

Nhdvl XID)(;-29() t:.)il>ttlil, i í(e 

dlislpcscr relcas4's culbes ¡isi, ice 

Im 'k 'ts. pitht.l irs tlr ti!er conldtairnr. 

:, itliout c( ntamniril tion 1):. i, 

;ila Ila;iduls. ite unl it. Initlco'; up 

t, 3 10 pn,,,ilms of i t o cu ), ts (t: aly, 

;aitd ,as ;I ii isitxllil'1 iseiSl)lsihig aiud 

.ti>ruug (.;il;cityt tL4 275. I>nllrd>. 

\l:itXo(~: J, (lUil)ine-lt \!orks 

New AAF horizontal «-o0MATIC 

o0 

air filter. . . Comnp!etely automatic opteration 

- Entire unit located within plenurn 

i Arailílable mn 6-foot-hiqh unit 

The new Tvpe H lModel G Roll-O-Mlatic 

seits new indt.,!:y slandard for horizonial 

roll-tyipe filers with many advantages rolt 

availabli in any other unit. 

One of thc utniue features al the enew 

l ype H is the ;ibilitv ¡o change mediaii rom 

oulside the plenmiil tiespile ihe fact Ihe culire 

unit is hotised within the pleaiiii. Drive and 

conirols are Iocaled outside lhe plenunm aiind 

are readily 'ac'csible. 4ll electcical equipimein 

is otaiside Ihe air strea'tn. 

A patenied ainti-sag device provides posiise 

imedia seal. aik idier rolls al each end 

I I i- 

plrc idJc m,,re unifrim tracking olf me, it. 

l'c-vrii. ;il. id .lid e rasCtbl' Mitly .;¡l;iiI¡i. 

1 i0-'oial C(jl'!GCij{ki1 ¡tl C boiit'Ol lbOX i.s i(l 

licld wi-inoC, icqul:'ed. 

The ;new Typle 11 Rill-()-Nlilic. aivailaIll 

i .t sv¡ide .rane. of sizo .ind flutir litlercuit 

.ha io!itrS, . ol'cf. s lt.'ler applic.tlij 

I!exibiln, th.iii ne ,r lefoe. 

l"or clllj'1lct in,.I; l;ioII, call ytllr llei'r- 

)byA.,i I: i elpcrti ive; o, tite ile tr r Bullhiilm 

No. 24,', Robert Sloore. Anierican Air 

F ill er Cmpnisi. ll l c., i Central Avectie, 

Louisville, Kentuckv 4-0208. 

American Air [ilter 

BETTER AIR IS OUR BUSINESS 

I - -- - -Y- 

F,,r umm, , 1.,?.Y -ii -r..li .- /1 

t 

i

LITERATURE & SERVICES 

*ii . . i 

Partitions 338 

1,aríw1íí3il3.q the l'AI' 1;e), Jtii I)hn1. 

htr(rhsi!l

tiie, gency L:ghting 347 

,i',., .:. !l1i.' Ii 'lll)3 H'r333 345 L: )13t3';li 

, .. .,..." 33ll... ,{ . "im i.: !.'xi §., '. . , (. . .i, 3i. P 1.1 :'3 .1.i !. .! 

SUPPLIERS NEWS 

_ _ ag____ 

1.d1.lr 1i.,,;()v1r::h \llulti :,; , ( :(:.. , 

n--('-eh l:. . lo . 'hi: ., . 

4' 5t:tn('el: ,I lit':: 's Il -\ 

::! í 5I; 

\. 

; .c .' I'i4 ' 

3Ih"'.i . ":3 'l,, r (',,rl3., ..l

,' i6 

- 290 - 

Facility Location: A Review of 

Context.free and EMS Models 

By Charles Re Fllc, Dalid Bigmian, Da :id Schilling, 

Jared Coho, annd Richard Church 

EM.S hl(ation mriodels arc tlae Iorriiilald to address s5pc(ilc probllris 

of cmergercy mnedical services systemrvs; (ontecxt-free lonation niodels arc liaose 

developed *wilhout referencec o partirtular applications. The literature onx 

dI,cee t,'o iypcs, of pullic Iacility Iioc;tinu mrindeli is revise'cli. and Ic devel. 

opmnent of the maximnal .overiíng¡ nio(lel [rons several cma lier context-free 

models is described. witis cmphasis on poblen statcments anid articulatious 

ol service objectives. An applicationl of tlhe naitnial coverUsg mnodel lo fire 

truck location points up the ability of tlis rnodcl to handc nmultiple objerc 

tises: its ability to tomlpare alhicia solutions gives it great utiility for 

i>lanini>; and cvashllting EiMS systcus o(l a wide raut;c of comiplexity. I'o 

tential applications of the maximal covering inodel arc discussed rcgarding 

EtlS problenms involving multiplc timic satandards and service objectives, 

location of special equipment, aid siting oI fixed facilities. 

'i lda(cment of facilities un a nciwork to rcspond to demands 

or tu alract consumers has blccome a subjcc lfor both study and 

teaching. The cxtensive bibliography by Lea [I] indicates how thc 

literaturc in this arca has burgconed in recent ycars. IThe literaturc 

includes investigations of the location of both public and private 

fa;...ies; much of the literature concerning both the public and private 

sectors throughl about 1970 was brought together in an articlc 

by ReVelle. Marks. and Liebman [2]. Private sector facilities arc 

often located to fulfill precisely stated objectives. suclh as minimnum 

cost or maximum profit. in conitast, ithe goals and objectives of public 

facility location are more difficult to capture and translate into 

quantifiable teras. 

The gradually evolving interpretation ol tlhe goal of maximum 

public welfare has led to a number of possible problemni statements in 

thc sphre of public facility location. Each of these diffcerent statemelits 

has the potential lo fit tle perccptions of some decision niaker. 

RKcogniring the vatiklity of nmany ot>jectives has bro¡xght us to wha;t 

Addres communicatiolis and iXqlu.sss for ip:lilu t(o Charlic ReVrlle. Pro. 

ipam in Syuemu Analyais a*nd Ecoomnis íor Publir Deisieons Makisg. Johss Hopkins 

University. Bahimore. MD 21218. 

David Bigman is an «'onomlr with thre Iniernational Monetarn Fun(I; Dasid 

kShiUing ia asistant profesuo at the Centcr tor Technulory and Adrainistration. SUMMER 

American, Univel>ity; lad Coalol is ,u ite prnlcaor in tIre loIhns Ilopkins Pro 1977 

gram in Syerms Analyris and Eonounics lor Public )elision Making: and Richard 

Churth i aistiíant profesor of

- 296 - 

REVEI.LE we believe to be the most uselul oí probleni stal;tmcnts for the pub: 

E:' AL. lic f.cility prohlem. It i¡ a statementc thal adtnis tic multiplicity oí 

locational objcctives and sccks those spatial pattirns that are superior 

under a number of ,hese objectives. 

'IThe purpose ol tiis article is to rcview thic location analysis literaturc 

of the last fecw ycars that deals with public eiergency lfacilities. 

Sorne of thisi literature represents studies directed toward specific 

problems ol emcrgcncy medical services (EMS) systeras. primarily 

location of ambulance dispatching points. Much of the lictrature. 

lhowvecr. reports contex-free mode dels developed o atidlrcss location 

probleiis imn general; these mnodels are not ticd to specilic applications. 

Nevertheless, cach oí IIhe context-free miodels we shall discuss has 

been applied in planning some forni of cmergency facilities. Wc 

sh;lil trace iie develoipmient of thc coilteI .frcc moidcis inod .discuss 

Iheir pieniaiil for furthier application iii emergency medical recovery 

systciis. T'Ihere is an ad:d-tionial body oí location literature thait trals 

locations on a plane, using ithe Euclidean metric; these works will not 

he rcvicwed liere becalisc we feel that thc aippro:cil lacks rcalism fro 

thie ypes of problems we wish o ; !dress. 

Location-Allocation Models for EMS Systems 

Ihe location-oriented literaturc Lth:t has ben developed in the 

specific context of planmling emergency sysiems may e dlivided into 

twu catcgories: thc first is coniLrtic n mainily with locationial and 

sr!aial rcnsidcralions and occasionally inclides notions of randonm 

.v,::? thiat are explored in an iecrative fa.slion throtugil simulation or 

IqueCeing. Severnii o thlese works draw on the context-free location 

literature. Thie secontd ci.tegory addresses thli event-service sequence, 

with eniphasis on the random components o tlhe events. Location 

anl. ', when it appears. is generally an ad Ihoc rather 

integral leature oí thlese investigations. 

ithan an 

Spatial Analysia Applied to EIMS 

Focusing on the average response timne. Vol¿ 3] applied a "steep. 

est descent" procedure to eici,iiine locatiuois of ambulances. The 

procedure, unlortunately, is subject to entrapinmet at local optinia. 

Alhhouglt best soliations are possible, there is no way shiort of enu. 

meration to verify whetiher a solution is inferior to the true optimum. 

The model's suitability for planning ptirposes is hindered by its requiremnents 

for data and computer time. Its complex structure makes 

it very dillfcult to introduce other considerations and additional objectives. 

The assumptiíon that emcrgenicy calls arrive as a Bernoulli 

procesU (i. , with stationary independelnt inrcreients) may be unrealistic 

in certain contexts. Volz added other assumptions thiat reduce 

thie mathematical coniplexihy but impair themoldel's useuilness. NeviL^.LTkH 

ertheles, thlis model was the fi st to trcal the interactiiou I>ctwcen 

SlRVICUS 

RESEARCH lo

- 292 - 

from tile demandl sites. He used both a simulation an< a ql

- 293 

R£VELLE is apparent in light of accusations about wastelul diuplications and 

eT AL underutilization of critical care units. Willemain's model illustrates 

the inierrelationships among'important planning variables bl>t does 

not lend itself direcily to empirical use. 

Event-Service Analyses 

A multiple-server queiucing model was uscd by licl¡ and Allen 

(l]j to deternmine tie riiniber of amblulances neicdcd to :achievc specified 

response times. They assumed that tihe temporal distribution ol 

calls .ould be describcld by, and meet tihe assumnptions of, a Poisson 

distribution; tihat is, that each event occurs independently of tihe 

preceding one. Unfortunately. tihe extension to multiple servers distributed 

across the network posed severe complications beyondl tihe 

mnathematicil coniplexities because of the need to define rules of 

prccedence and responsibilitics for each arca. 

Savas [121 analyzed ehe cost.cffectiveness ol pioposedl challges in 

New York's ambulaince service with a descriptivc sinimulation model. 

'I'he simulation was of a standard sort. developed witli I¡BM; ahlerna 

tive locations were investigated ,I trial. 'ie costcffectiveness was 

dlisplayedl as a benefit/cost ratio o. zminlites saved to additioínal (ol. 

lars expended per mionlil. 'Th'lis was a stlamigitlorward bti pioicering 

eclort to use quantitative methods on anr emcrgency service system. 

NMeasuring benefit in lives saved, Smith [13] developed a benefit/ 

ost analysis ol a coronary emergency rescte service. He used botit a 

quetieing atI a simulation model, with no appreciable difference in 

·. h. C.aly aggregatc data werc av;ailable. whichí reduced t¡he reli. 

ability o tlihe an.;lysis. 

Stevenson [j1 used queucing theory to estimate dispatch delay in 

ambulance service operations. For a given number of ambulances, 

he provided order.of[magnitude estimates of the minimum expected 

response time, using aggregate data antl a simple queueing model. 

The data reflected general regional characteristics such as average 

speed. area, and total number of calls. The results were expected to 

help the decision maker determine equipment allocation levels. Ex. 

plicit locational considerations wvould probably have provided more 

uselul result. 

Chailken and l.arson (15] explored the characteristics of urban 

emergency scrvices and reviewed tlie operational problemns of these 

systems. Their concern was with vehitle deployment and application 

ol queueing (chiefly) and simulation approaches to these problems. 

They considered allocation policies relatecd to the number of units, 

location and relocation of units,. response areas, and patrol patterns. 

Acton [16) studied tihe cost-effectiveness of five alternative heart 

attack programa. Many ol the uncertain parameters in the analysis 

were bhaed un estimates hy a panel of experts. He usedl a Mlonte Carlo 

IEIALTH simulation to ciaTulate expccted lives saved and assoiated costs of 

RESEARCH cach program. The benefits duc , , savel lives were ectimated by 

livelihood saved and by willingness-o.tpay measures; the latter were 

obtained from questionnaires and require further study to gauge

- 294 - 

thcir validity. Cretin [!7] analyzcd ilhe process of dcathi followinii FACILTY 

nmyocardial infarction and developed a model of risk of death fronm LOCATION 

inriarction. ¡ler model combined features of both the disease process 

a;i.d tie msiedical iolpunse and was used to evaluate policy alternatives. 

Thec disease was nimodeled as a discrete Mlarkov process, withi transition 

probabilities and other paranimeer: obtaincd froin statistical regression. 

I'lhi is an interesting approach, whlici coulhl form 3n integral 

l:art ou broa;der rm ,dels. 

l1cvcs,,olS aind Willeniain [18,19] vicwed thie scrccnitig process as 

a triage system. in which calis were alloated to prinmary service if 

Ihey wiece ciecrgcncies and to secondary service if they werc not. 

Cadlls wicrle aso routed to tile ecoundary service il all priuniary service 

vehicilhs were buisy. lhe analysis consisted of comibining assiarimed decision 

and evesst probabilities. Although the approaci was straightl. 

forward. tile assurmed prohabilities were not justified by the autihors. 

Ti'ese models and ,,ncceptualizations representl tle state of the 

;ar as relletied by tihe EMS-oriented literature iup to about oIIc ycar 

ago. 1i is clear tisat there is great potential íor lurther application of 

locatiota niodels to ENIS problems. We believe such applications will 

draw heavily on the cointext.free location litera;iav specifically on 

tihe Imaxinmal covering model and its refisenicenis, whllli atre distIissce 

below. 

Context-free Location Analysis 

An interlocking set ol tihree network loiation mnodels will be 

described here. A!' :aideb hta;.. have been applied. at least conceptually, 

lo emergency seivices. Tite third model (tile ttmaxiinal covering 

model), whose lurdier developmient and refincient we sihall discusIs 

has recently bcen applied to the location o[ fire stationis and 

equipment in Baltimore,'MD. I'lis i¡, to our knowlelge. tíhe first 

applications of this model to such a seiting, although it is chiaracterized 

by the moht versatile and adaptable objective of all thrce niodels. 

The p¡Median Problem 

The location model that underlies, tieoretic!"l)y and historically, 

the others we sliall discuss is the p-median problemn. whicil was pro- 

Iseld by Hakimi [20,21] in the context of locating switchiing centers 

in a communications network. It was recognized. lhowever, that the 

probiem statement could be extended to many otiher fiames of reference. 

We will state it in the setting in which people come to facilities 

lor service, but it can easily be turned around to pertlin to the situation 

of serven going from facilities to points of demand. 

Locate p faciliies on a network ol denreds so tha tihe average travel 

lime of all uen is a minimnum. Since fhe Ifailities ate not dstiinguishetl 

Lt sire or specialiaiion. it is asumed tihat every user will travel lo hisa 

uearest facility. ,tJ MM ER 

Numerous researchers iave propxned sulution , .lihods tor thc 

p-median prublenm. and solution procedcsílt' are imporirant: a well-

- 295 - 

REVEt.EI. 

ET AL. 

poSJcd probleni stalticlcit withiout at adel a icatc solitioún |proedluic is 

of litle use. (Necveitilclss, olr primary concerl ha;is bcen withl prob- 

¡lem statementes. Capturin§g tic Io

- 296 - 

The Location Set Covering Model FACILITY 

A possiblc otitcome oí gradually tiilictiilng ihe miaxini¡ill tithme ! .)CTION 

constriintis is tit;li, at some point. tile p Itailities will hecome insufficient 

in nuirilbe lo t oer r all points of- dcialudl withiiil tlhc timec (onstraint. 

'I'he logical questiou t lh:t follows is, given tile time standlaríJd, 

wliat is tihe leais niumnber of facilities required? This question lcd to 

ite aot"".uIl;tisim 

and solution of thle loatiorn set covelilg probllnim by 

1'l'cls ct ¡,1 . 1:121. A stateimelt o1 tev plr)blcm is: 

Fiidl tihe iniaiiurnuin t¡iiiinLber of fatliieis ¡ aiad lihcir lou.lio.nas suct la : 

eJt I i. i lt ul dIci¡i1anld Ia;is 'a .ciliti) witilsis / tsllC u Isits. 

iterestiniigly. several specialistsL ins ciiie:rtiiy hicailil services have 

also posed this prob¡em, aliloghtil tlhey were unaw:;rc of ihe tractablle 

niathiematical strtucture uitndcrlyiuig it. liitmley [34] ga;vec i(t lolow. 

i"g verb.,Il It'illltCW)a 

' : 

Hlow nmasiy amlbulaictes? Arc li'tre riotiligh? 'l;lii is Jal .ICeCil ¡ehI: zC. 

ispons¢ timie? Usually iti a inctiopolilan arel. i¡ is 15 nmiúitsl .... II 

a 15 misnute responxs tinei il deiimallded. how mi;otv a:inmulances are rc. 

quirrd; wlrrc muull rley bc Iosi(ioiuilt ¡1 ptu iu ;,;.l( alsla Ct 

thal tihis criterion is nicie 

The saiime undamental questionis were asked by (:assid)y and Wilson 

in thlicis stmldy of thie l.oinlot Anmbilalile Se:vitc [3.iJ: 

llow maity velCiclc; 1.aulId .It Seivi£c hlavc asid luow siOUhild tIíh I.ll 

llcet be divid; *..'Cca .irceclicr auibula.igcs amad ithe vaJ¡iou s¡izs of 

sittiitig

- 297 - 

RM.VELLE 

EL AL. 

of facilities and thieir locations. A facility sitc tilat is eligible to cover 

a number of denlan¡d poinis is generally a miore ellicieni choite tIhan 

a site with only a Iew appearanices in tile nmany denianid sets. 

Wlien tile location set coverinig problemn was first stated anid 

solved. its precision ol problemn statenent. its strong relation to stated 

necls. and its straighlitforward solutíion metilod quiickly established it 

as a widely recognizc( nimodel for thei location of facilities. Thle suc. 

ccsslul ellorts of Public 'i'ccliology. Ial'. ini promintig thiis model floi 

locatitng fire stations lostered this recognition. 'I'lThe model has been 

rcquested for ause iti njearly 100 cities. 

The localioJt set covering niodel as tile first to use sl e ,nl:xiiiiii 

uiadi;;nce (uo tiime) ;is a deter-iiiiii;:l ol tic spati;il cotiligulatioit 

of facilities. As such, it was tile first to neet thie need for distante or 

tinie stantdal;r thal t liad beet stateid iitd he ic clirs ;saiiih!iii·c f alid tr;ivcl tlie lii .a git'si ssiiiiitle /, pI 'i f;il¡it. líe lh.a. 

IILALt I 11 tion Ket covertig niollvl igt:,res pIo¡ lationí, ;iiI ltil tile tI litliillííí 

SERVICl" tiunitber ouf Ifilities ietieledl so c 'rs' .1 set uo tlel. ml.l i,luiit>s withiiil ;1 

iaslputle-tiinte stlandlarl. TI'hc .Ixialt.iil overitg mitlutc 1 .ihil.iit 1the 

ilsipurttilie uf IxJuiiatiul n stul drel.,il .í CSleiOt'.SiíS tili' .t.tl' 1: ti(

- 298 - 

Fig. 1. Maximal covering model results: Percentage of 

population covered within 2.0 time units by various numb .rJ 

of optinmally placed facilities. Point B is a solution tou 

the location set covering problem. 

c 

o 

100 

o 

v . 

oa. 

o 

e 

- 50 

o 

o 

c 

e 

. _ 

00 ,E 

o 

o 

4 8 

Number o1 flciliHties ued 

Ia¡ ilitie are to be placed so a> to cover as much of the pIoptulatiolu (or 

as tsallny dmi;atld Ipints) as Possible. Wilen p, the liuint un I:t ilities, 

is niacreilsd to tihe !)oint iiat 1(10 percent of hi1 tdel:iítantl is covered 

by tihe a axiiiial bovering iiodel, its va;luei is sle aic a- tilte unii.i 

iitiii numinbCI indicated by the Ioc:alion set (overilag nmodel; tile latility 

placemenits indicated by tite two incdci, el howevari. 

sarily be tihe same. 

will oltu Icls. 

TIlia. maximal covering model fill ata ilii m,,.tant need in locatioii 

analyis. It

REVELLE 

ET AL. 

HALTOI 

SERVICUS 

RESEARCH 

- 299 - 

Fig. 2. A solution to the locatino set covering 

problem. Facilitiea (circled dots) are placed so that 

all demand points and 100 percent of the 

populatiun are witiin 15-unit limits (jolid lines\ 

from some facility. Dashed linres are 10-unit limits 

from cach facility; they enclose only :31 ¡perccnt 

ol tihe populalion. 

placed facilities will cover 90 pcrcent ofl tle dcimamidi. elic (litcstioi 

is imrnediately apparent: are t!he four aduidiioinal la ilitics wortlh 

their cst. wben thiey cover only the last 10 pertent of the denm;md? 

The location et covering muodcl would Iave yielded only the inforination 

underlying point 13 of thc grapih. 

Solutions available from thie locatíio set coceting miodel are 

rigid. in that all points must be coveedsl; ini additioun, the sohltions 

are not "tight." Extensive computatioinal experience indicates tihat 

the mininimuni nunmber ol lacilities to provitde :over mu;y br ;a¡r;aiged 

in many ways without significaultly violatinig the futd.ilidimt:il uovirage 

standard. Applictaion of the mi.íxiíiil. *osc iti¡g niKlCl *a0t oftCi 

improve on solutions oblainable Irom thle lc;ilio e' t i>is¡ls. liiws ii ¡: g I. I .íi¡

- 300 - 

Fig. 3. A maximal cuvering solution to the location 

problem of Fig. 2. Facilities (circled dots) are placed 

lo that all demand points aml 100 percent ol 

the population are withian 1i-unít limits (solid lines) 

from some facility. Dashed linc, are 10.unit limits 

from each facility; they enciose 55 percent 

ol tihe population. 

- - 

is one of tlhe many possiblc solutions to the prolblem of minirnizing 

the number of facilities (five) needed to cover all points within 15 

timenc (ur distanice) units. The tdlusters surrounted by solid lines in. 

clude points within 15 units of thec Iacility to whicl tibat cluster of 

uiints is assignietd. lhiec dashed ¡lisn suiround puints tat aerc withini 

10 units of eacai cluster's facility; these points represrint oily 31 pertent 

oL the area's populatiosi. The reinai:iring G9 percent o thic populatium. 

althoughl it falsa withioi the 15-unit limit. is mirce than 10 

units away fromi alny lacility. 

Figure 3 representa anothier solution to tlie probtlen. obtained 

with the maximal covcring model. Th*is solution was Loomputed with 1977 

¡he ,ame number ol facilities. and all pointis stil Iie within 15 units 

ol thleir lusest latility. T'le omullutiaioiil. however. sought to mllaxi- 

FACIIITY 

L)CATION

REVELLE 

ET Al.. 

- 301 - 

Fig. 4. Another maximal covering approach to the 

location problem of Fig. 2, with 15-minute constraint 

relaxed. Facilities (circled dots) are placed so that 

95 percent of the population is within 10-unit limit 

(dashed lines) from tome facility; 98 percent is 

atill within 15-unit limits (solid lines). 

mize the population within 10 units, given tice countra;int of keeping 

the entire population within 15 unitu. In the new configuration 55 

percent of the population is within 10 uniht of a facility. Although 

the location set model nuight have foutd thiis solution by chance, tihe 

probability of its doing so is low; in general. onily explicit optitnization 

using the maximal tovering model cain hc expected to yield sudc 

solutions. 

If the rigid constraint of complete cever-gec witil¡n as: ouiter liiit 

is reLaxed, alternative configuratiois witl puosilly molu e sic'l'l charirIAMit 

acieristiis may be derived. Figure 4 depicts a solutioii ohtaiiicd witi 

SRaLaaiUR tbhe standard of 15 units discarded. Fiv'. facilities can cover '.Y5 percent 

ol the population witlin 10 units wilce leavislg ulry '2 l.'cint 

o the popuilation otusieir a 15.unit p

- 302 

Multi-objccuve Maximal Covering and Fire Station Locatimo FACILITY 

Iri thc past year we have bcen assisting the fire department of the LOCATION 

city of Baltimore in planning its future system configuration. Thc 

National Board of Fire Underwiiters (now tihe Insurance Services Organization) 

has stated guidelines lor fire protection (391 that arc similar 

to those given by the EMIS regulations except that they are given 

in teirms of distance. The fire siandards are absoluitc: they do not 

include thie 95-percent specification thlat gives the EMS standards 

their realistic flcxibility. In practicc, however, it is accepted that 

absolute attainment of the standards is not within reach; thus the 

maximal covering model is appropriate and useful in this setting. 

Althiough thc fire standards are expressed for coverage of property. 

coverage of the population is anothler obvious objectivc. The 

fire hazard to population and thie fire hazard to property. determined 

as the product of population (ur property valuc) anid fie fre frtlcecy 

for subarcas of the city, constitutic still other mncasures againllst whi: 

to gauge coverage. Coverage ol fires, as reflected in fire frcquncicy, is 

also an important ofjective. These issues wcre worked out in consultation 

with the firc

- 303 - 

REVet. 

¡T AL 

Act of 1973. The menaning of an "adequlate nul>cier of emergency 

vehicles, called for in the legislatioan. is interpreted in the regulations 

as a number suifficient tlhat 95 percent of relquests for assistance can be 

met within 30 minutes in rural areas and 10 minutcr. in urban areas. 

To the location analyst. tihe framers of the regulations lor emergency 

services appear ao he among the most articiulate of decision makers. 

for their specificatiolis of elfectiveness mnicsures are cear, concise 

mathematical entities. Tihe analyst is seldom fortunate enough to 

encounter such precise problem statemcnts. 

The coincidence of viewpoints of tie frainers of t¡he rcgulations 

arid the autihors of the naxinial covcring model suggests, correctly. 

that this model was structured to meet real needs. In fact, the origin 

of thle model lay in a 1970 report of the Carnegie Comnission on 

Higher Education (42]. That report called for establishment ol 55 

t 

lieaith i are teiters acroís the natiol, loatced iii stlch a way that 95 

percent of the populatio.! was within :an hoiir', drive of a faciliy. 

IlThe numinbers aside. this concise stat,:imtciit sparked the lcvelopinicivt 

of the maximal covering problem by Chur(hl and ReVelle [41] out ol 

lthe location set cove'ring problem. 

The maximal covering inoxlel ullers adtlitiona.sl informination beyond 

that requested by the EMS regulations. As discussed previously 

in connection wiíth Fig. 1 (jp. 137), tihe maximal covcring model can 

indicate noi only the n numbe r and locations of vchicles requuired for 

95 percent of responses to o(xcur within 3S minites. but also the nunmhkr 

arul !ocatiois needed for any coverag;e level one wislies. allowing 

..

- 304 - 

population center; the use of secondary outer standards could ame. FACILTY 

liorate the situation o thiose who represent the 5 percent of demand LOCATION 

that is ex mpted Iromn tlie regulation standard. 

Still other uses u thie maximal coverirng model are possible: alter 

maximizing coverage within a stated time, one may seek tiue solution 

tihat minimizes the average service time for tiose outside the stand;trd. 

Tite economic and geograpliic options available to planning 

regiouis difler cnoughl that each region miglit benefit. both in service 

and in cost. from "fine-tuning" its EMS plan by examining many 

such secondary solutions withisn the regulation standards. 

Sonme work has already beien done oin ;aplying time other two 

toutexxit.ee models, tihe p-median andl tihe location set covering models. 

to emnergenry ambulance services. The work of ReVelle, Tore. 

gas, and Falkson [38) cited carlier extends the location set covering 

model t to twn nore detai:ld tspects o! tie EMIS locatiosn problemn. 

()nc ol limes, .Isp:Cts concerns travel stieC oli tihe link froni demant d 

¡point to care facility alter time patient is pickcd t up whlidi needs to bc 

considered in determining ambulance locations. The other aspect is 

that demands in the real world tend to occur along continuouis roads 

rather tihan at discrete o Berins. erln. l. 'elle. anit Elzinga [43) used 

time p.mediant framework to investigate tie riisinimum average time 

when the travel from demand to point of care is included. We 

have not yet applied the maximal covering nmodel to these more detailed 

expressions ol the problem, but it will be usefucl to discuss here 

omrne possibilities inherent in sulch an application. These possibili. 

ties si: :.d ir,- of interest to EMS policy makers, since they allow the 

evalative standarda diseussed in the regulations to be extended to 

include the link from demand point to point of care. 

When an ambulance arrives at the scene o a call. certain limited 

care may be given the patient, depending on time training and equipment 

of the arriving personnel. The time standards for response to 

calls apparently apply only to the arrival of the ambulance at the 

scene. Thle purpose of tire ambulance is twofold. however. and the 

second function is transporting the patient to a staffed antd equipped 

emergency care center. 

Without meaning to suBgest dilterent values lor the time standards. 

it would seem useful to apply a time standard to the entire 

time between dispatch of the vehicle and the patient's arrival at the 

emergency care facility. 11 thie 30.minute response standard were extended 

to cover both functions. more vehicles and locations would 

be required to achieve the stated coverage. Two standards might be 

formulated. one for time time from ambulance dispatch to the emergency 

scene, tihe other for the entire time from dispatch to the appropriate 

site of care. Time addition of this new standard provides 

another yardstick widi whicl lto nmeasure the effectivenes of an emergency 

recovery system. Modeling with this added standard does noU 

seem to present any insurmountable challenge. UMNER 

Still another standard could be applit 1 lor the time from dernand 

cene to hospital. Application of such a stalidard, however, is not

- 305 - 

REVELLE possible within ilhe location concepts discussed tliur, lar. whici focus 

ET Al.. on tile dispatchl location of emergency vehicles. Achievement of a 

standard lor tile demandl-point.to-hospital Ieg of the transportation 

triangle (depot lo demand to hospital to depot) can ciily be influenced 

by the locations of emerg;ency care hospitals. 

The locations of aiuch hospitais in turn affect the vehicle loca. 

tions needed to achieve standards on the entire depot.to-point.ol-care 

time. The twin problems are interlocking and complex, but potentially 

solvable. Together the two problems of ambulance location 

and location of emergency care facilities comprise almost all of tihe 

transport segment of emergency care. Seen as two interlocking problenms. 

they give rise io a more basic question: 

With limited resources lor purchase and opcration of vehicles and for 

establishment and opcratio.m of emergency care facilities. dctennine the 

nunmber and localionís of boill ¡Ihat will maximize the proportion of resloScs 

occurring withi.i the lime sandard. 

A complementarity of achievement exists between emergency vehirles 

and emergcncy care facilities. An additional emcrgency care 

facility miglht oblvi;:te the need for I ,r (to choose a number) emergenty 

vehi(lies by dliiiiisling thc total time from dispatch to final 

care site. The maximal covering model will permit these trade-offs 

to be examined. The relative costs of mobile and fixed facilities 

would indicate whiclyhuoicses were appropriate and open the possi. 

bility nl cost.eflcctilness comparisons that could Icad to EMS syst(.. 

,. wuld be'1.both functionally and economically optimal. 

Summary and Conclusions 

We have discused the remarkable hainiony between the structure 

of the maximal covering model and ihe goals of policy nmakera. 

The ability of the model to deal with multiple objectives and to opti. 

mally asaign special equipment to optimally-located depoits, as in our 

fire department study, indicates the model' general utility in a wide 

range of location problema. It should be particularly useful in plan. 

ning and evaluating emergency nicdical systems. 

In location research (perhaps in other types of resc'rchl as well) a 

aymbiosis exists between the setting of a particular application asid 

the development of models. Lcation models aid in conceptualizing 

a problem in a particular setting and the special characteristica of 

the setting feed back to enrich the location model. We expect such 

aymbiosis from the EMS setting. We believe that location analysis 

can make a significant contribution to the evaluative structure oí 

such services and that tlis setting will add new and important di. 

mensions to location ¡clcarch. 

l1tALTH.' REFERENCES 

silavíC I. Lea. A. An annouted bibliogiaphy olf , lei io.aUocaIoU I Wolking pH'ry. I>. 

lESAtmti of GCeogaphy, Uniacnisiy olf uunto. 1973. 

ReVelke. C. D. Marks, and J.C. l.iáeman. An analysti ol privale and publitc wc 

lor loation modrit. Mondge Sti 16:62 July 197'0.

- 306 - 

3. Voli. R.A. Optimum ambulance location in acmi-rural arcas. Trans, .Sri 5:198 FACILITY 

May 1971. 

4. Fitsuimmons. J. A methodology for emelgency ambulance deploy)rrtit. Manage 

Sei 19:627 Fcb. 1973. 

LOCATION 

5. lerlin, G. and J.C. Lltbiran. Matilrmnatical anal)yis of 

location. Socio.Economic Plann Sri 8:323 Dcc. 1914. 

tnc-lgciucy amnibula cc 

6. Stidl'lim. S.. C. Toregas. and G. Iisher. Operatiolial analy.si of dcmand-.es-poi. 

isvc a>stema. Department oí Envlmnrnmental Systemins Eumgincaing. Cnimell Uni- 

imbity. · Report to the New York Siate Scicnce amd T'ed.lillogy F'oundation. 

Jan. 1971. 

7. Stidham. S. Stochasiic dcsign modcls for location ant allocatiorn of public scrvice 

lacilities. Depanment of Environmental Systems Engineering. Cornell 

Univsenity. Reron co the New York State Science and Technology Foundation. 

May 1971. 

8.

- 307 - 

RLVELI.E 29- Khumawala. a. An cffricie algorithm tor h. 1973. 

30. Swoveland. C.. D. Uyeno. 1. Vertinsky. and R. Vickson. Anilmtlance locallon,: 

A probabilistic enunication apprlahla. Alanoge Sci 20:616 DI)c. 1973. 

31. Carbone. R. Public facilities location unelcr stochastic dcmand. INFORl 12:201 

Oa. 1974. 

32. Toregas, C., R. Swain. C. Re'.Vclb.. and I. Ihkm.nlnti. 'rhuc lkliiin tío a Lti( al Iti

.'17 

- 308 - 

INITER.'ACES Copyright (e 1979, The Institutc of Management Sciences 

Vol. 9. No. S. Novcmiber 1979 0o92-2102/79/090510003501.25 

THE LONG ISLAND BLOOD DISTRIBUTION SYSTEM 

AS A PROTOTYPE FOR REGIONAL BLOODMANAGEMENT 

Eric Brodheim 

Lindsley F. Kimball Research Institute, The New York Blood Center, New York, New York 10021 

and 

Gregory P. Prastacos 

Department of Decision Science, The Wharton School. University of Pennsylvania. 

Philadelphia, Pennsylvania 19104 

AISTRACT. Eacti year ovcr two million htospitalizcd Americaiis dci.n1tl 

pon Ilhc ttilm ely av;iilaililiiy ofl Ihle riglht yp.c of hinod i)rojucls . t 6.O(XH) 

ioslpitlal hlhto banks (HUUB's) in thie United States. If the right bIood products 

are not available al thc HBB whcn requircel. thcn medical coinplicalions 

or postponements of clectivc surgery can result which translates to extra 

days of hospitalization and expense. On the other hand, since most blood 

prodtlucls may only Ihe adllinislerLcd io)t ia pliCllh of tic saillc bild typlc withiln 

21 days of collectiot., overstocking at HBB's leads to low utilization. which 

increases costs and is wasteful of the scarce blood resource. 

The Long Island blood distribution system was set up as a prototype of a 

regional blood ccnter and thc hospital blood banks that it services collaborating 

lo preplan regional blood tlow. It maximizes blood availability and 

utilization accordling to a Prograinmcd Blood Distribution Systenm (PIDS) 

model and strategy that has been shown to be generally applicable. PBDS 

schedules blood deliveries according lo statistical estimates of the needs of 

each HBB and monitors actual requirements lo adjust deliverics whcn indicated 

by control chart techniques. In addition, it provides a daily forecasting 

of shorn-erm shortages and surpluses for the next several days that results in 

controlled movement of blood to and from adjoining regions. Finally, the 

system is able to adjust the regional strategy so that availability is reduced 

uniformly at all HBB's during periods of seasonal. regional shortages. 

PBDS has drastically improved utilization and availability of blood on 

Long Island: wastage has been reduced by 80%, and delivery costs by 64%.; 

This prototype is acting as a model for other regional blood centers in the 

United States and for other national blood services as a basis for planning 

and controlling blood flow in a geographic area. It usually replaces preexisting 

procedures where a regional blood center collects blood based upon 

gross estimates and reacts to requests for blood by individual HBB's on the 

basis of experience and on the currently prevailing inventory situation. 

Introduction 

The Operations Research Laboratory of the New York Blood Center, in collaboration 

with the Long Island Blood Services (LIBS) division oí the Greater New 

York Blood Program (GNYBP), has been studying the probleni of providing a high 

availability of perishable blood products at each of the hospitais in a region, while 

HEALTH CARE-BLOOD BANK 

INTERFACES November 1979 

Z.; ~~ ~~. 

; y.,,p `, ~~~~. , ¿;~~. 11- 1.

- 309 - 

assuring thal the maxinmum nuniber of these products are utilized during their 21-day 

lifetime. This problem requires thai the dual and seemingly conflicting concerns 

about availability and utilization be reconciled. It also rcquires a radical change in 

management decisicn concepts so that the regional blood center (RBC) that is responsible 

for collecting and distributing blood and the hospital blood banks (HBB's) 

that stock it for possible use recognize common objectives and collaborate in implementing 

a strategy that optimizes the regional utilization and availability of these 

scarce products. 

The first section of this paper describes the national problem. The Management 

Science techniques lthat were utilized to create a transferable mnodel for a distribution 

plan with near-optimial characteristics are dcscribed in the second section. This is 

followed by a description of the implementation of this model by the Programmed 

Blood Distribution System (PBDS) and of its testing al LIBS as a regional prototype 

for the Cvolutioni of a national blood distribution netwourk and operation. The managerial 

and economic impact and the implications and extensions of this prototype model 

are then discussed in the concluding sections. 

Background 

The national problem 

Mosl blood products in the United States are derived from whole hlood thlat is 

collected in units of one pint from volunteer donors by approximately 200 RBC's. 

After laboratory processing and testing, blood products derived from the whole blood 

are distributed to the HBB's in the region where they are stored to be available for 

transfusion when requested. This paper is restricted to whole blood and red blood 

cells (i.c.. whole blood from which plasma and other components have been separated), 

which together account for over 95% of all transfusions. Both have a lifetime 

of only 21 days during which they can be transfused to a patient of the same type and 

after which they have to be discarded. 

Historically, HBB's have maintained high inventories of most of ihe eight 

different types of each of these products in order to provide high availability to 

satisfy patient needs'and have accepted the low utilization resulting from spoilage. 

Consequently, the national utilization rate of whole blood and red blood cells prior to 

expiration was estimated to be only 80% in 1974. At that time, the federal government 

adopted a national blood policy that called for an all volunteer blood supply to 

be accessible to all segments of the public. The blood supply was to be efficiently 

administered through the formation of regional associations of blood service units in 

each of which a RBC and the HBB's that it serves would collaborate to achieve these 

objectives. 

The role of Management Science 

It had long been recognized that Management Science techniques were required 

to improve blood utilization and availability. Strategies with desirable characteristics 

were formulated for individual hospital blood banks and pragmatic regional approaches 

that improved performance in a certain region had been implemented and 

INTERFACES November 1979

- 310 - 

reponed. Thec Nutional Heart, Lung and Blood Institute Resources Panel in its 1973 

recommendations, which were a major factor in formulating the National Blood 

Policy [4], stated that: 

One of ihe strongect arguliculs I'ivoring tihe introduction ol s)'slcins management lo 

blood service is that it will result in improved service. in more economical utilization of ihe 

blood resources, and most important. in iniproved effectiveness and efficiency. For these 

reasons ii is recommended ihat emphiasis be placed on developing micasures of activity und 

relalting thcse io objectives. Only ii ihis wa;y will it bc Ipossiblc to oill;lio secire antl credillc 

cvaluatioin of the chaigcs lo ble mi'dc. 

This study addressed these issues by: 

(1) creating a model for relating measures of blood banking activity to 

availability and utilization; 

(2) developing a management tool called the Programmed Blood Distribution 

System (PBDS) to implement a regional blood distribution strategy that is based upon 

this model; 

(3) evalluating the effectiveness of PBDS in a prototype region (LIBS). 

Complexity of the problemn 

The complexity of the blood distribution problenm is primarily due lo the perish.able 

characteristics of blood, to the uncertainties involved in its availability to the 

RBC, and in the variable daily demand and usage for it at each of the HBB's. Thc 

situation is complicated by the large variation in the size of the HBB's supplied, the 

incidence of the different blood groups, and the rcquircments for wholc blood and red 

blood celis at individull hospitals. 

Since it is a national policy for blood to be derived from volunteer donors, its 

availability is uncertain and is a function of a number of factors thait canniot be 

controlled by the RBC. The demand and usage of blood at HBB's arc also unicerlain 

and vary from day to day and between hospital facilities. The HBB's within a region 

may range from those transfusing a few hundred units per year to those transfusing 

tens of thousands of units per year. The most frequently occurring blood type (0 

positive) occurs in approximately 39% of the population, while the least frequently 

occurring blood type (AB negative) occurs'in only about 0.5% of the population. 

While most medical authorities agree that at least 90% of all blood trans.fusions could 

be in the form of red blood cellis, sonme hospitals transfuse almost cntirely red blood 

cells while others transfuse entirely whole blood with the ratio of whole blood to red 

blood cells frequently changing with time as transfusion practices improve. 

Approach 

The transfusion services throughout the nation are characterized by diversity. 

Each RBC has independently evolved its own philosophies and tcchniques for blood 

distribution. Each region strives for "self-sufficiency" iii supplying the blood nceds 

of the hospitais in its region from donors who also reside in approximately the same 

area. Because of these factors, it is essential that any strategy devised be defensible 

from the point of view of both the RBC and caclh of the wide range of HiB13's that it 

serves. Furthermore, any strategy that involves interactions between RBC's niust 

provide for clearly defined benefits for all participants. In addition, it is desirable that 


-311 - 

the implemented strategy he c hitaraileri/cd h ixo imauaqgeii ment cn tcepis: rotation o l 

blood producis between HBB¾.. aínd irc.ilhCeduLId deliveries to ¡he HBB's. 

Any strategy Iha¡t aill>oeaes hltoitl Iprnnd itct, tio he retained t until ransfused or 

outdated will result in low uili.zation. cspciiaill in the case of tiie small usage HBB's 

which, in aggregate. aecoun

0 6 

- 312 - 

FIGURE 1. Availability Rate Model. 

le · · · ·- · · A nvAiLABiiTY 

1 

> I IJl/ I I J 

I90 

>- 1MEAN DAILY DEMAND, D 

MEAN DA11Y DEMAINlD, D 

It was similarly established that the daily usage co'uld be modeled as an exponential-type 

distribution whose paraimeter is relcted to tie mician dily us;agec (U). 

The dcinUand-to-uslge ratio (D/U) was found 1o vary betwceni 1.5 and 4.0 Ior mllost 

HBB's with an averagc valuc of about 2.5 iii most regionsl' Thcse unalyses shliowed 

that the parameters for the models of demand and usage coi Id be readily estimated 

from records maintained by HBB's and ftirther tlihat aviabilily ratc eCoi(d bc reliahly 

estimated by thec mIodel. 

Utiliza Iion /1 

The utilization rate (i.e., fraction of the periodic supply'iwhich is transfused) al 

an HBB depencis on the size as well as the age mix of its blopd supplies. In order to 

derive a model for utilization rate, ihe following basic distribution strategy was 

developed, based on the premise that periodic shipments are rade to each HBB. At 

the begirining of every "period" the RBC ships a fixed number of fresh (1-2 days 

old) rotation units and a fixed number of older (6-9 days old) retention units to each 

HBB. The retention units are permanently retained by the HBB.until transfused or 

discarded at the end of their useful:ife. The rotation units that are in excess of u fixed 

"desired inventory level" at the nd of the period are'returied to the RBC for 

redistribution as retention units. Since the rotation units a're fno subject lo spoilage, 

the utilization rate is determined by the behavior of the retention inventory.**' 

The number of retention unitsin inventory at a HBB immediately after each 

delivery can be represented by a hfinite-state Markov chain, whose trUnsition probabilities 

are a function of the fixed periodic input (i.e., che fised retention shipments), 

and of the variable demandland usage. Under the Ussu.Imp-ion that Ihe oldest 

unit in inventory is transfused firc, the steady state solution of thc system can be 

computed and related to the utilization rate. This model was examined in [3], where 

**Ulilized Units = Total Supply - Units Spoilad. 


- 313 - 

analytical approximalions were derived for Poisson usage, relating the number of 

retention units in each shipment to the resultant utilization rates with the desired 

inventory level as a parameter. 

This rclationship is illustrated for a fixed utilization rate of 98% by the family of 

broken lines in Figure 2 where the scheduling factor P is the fraction of mean daily 

usage that is replaced by retention shipments. As an example, if a HBB's mean daily 

usage for a given blood product is 1.5, thecn the HBB can achicve a utilization rate of 

98% by any of the lollowing combinations: desired inventory 1= 1 and P =0.89, or 

1=3 and P =0.82, or 1=5 and P =0.70. 

0.9 

a: 0 7 

FIGURE 2. Utilization Rale Model. 

/ 

- / CALCULATED 

L0. / I ¡ 3 /OPTIMAL 

I 0.ó r I / ,--- I=S 

FEASIBLE 

.(/) Y I / SOLUTIONS 

: 1INVENTORY 

LEVEL 

0.5 I I I 

0.5 1.0 1.5 2.0 2.5 

MEAN DAILY USAGE, U 

It was shown thatthis stocking procedure maintains the mean inveqtory close to 

this desired inventory level most of the time. It was also shown that adding additional 

stages of returns and redistribution would make only slight improvements in the 

availability rate and utilization rate achieved. Since multiple redistributions introduce 

severe logistical problems and significant transportation costs, distribution 

strategies involving more than two stages of distribution were not investigated. 

Properties of desirable regional allocation strategies 

Having derived models enabling us to predict the availability and utilization 

rates of a HBB fc- any rotation/retention policy implemented by the RBC, the 

regional allocation problem was examined. It was assumed that some fixed penalty 

costs were associated with every nonavailable unit and every nonutilized unit, and 

the objective was to determine the distribution policy parameters so as to minimize 

the total expected regional cost. 


First, the policy that minimizes the total expected one-period cost was derived 

[5]. It was shown that this policy involves the following operaiions: 

(I) first allocate all available retention units so as to equalize the utilization 

rates at all HBB's; 

(2) then allocate all available rotation units (which are not subject to spoilage) 

so as to equalize the availability rates at ai1 HBB's. 

It was shown thait this pucicy is independent of the unit penalty. costs, and that is 

maximizes both the availability and the utilization of blood in the region, simultaneously. 

That is, any deviation froni the policy that would reduce utilization would also 

result in reduced availability for the next period, and vicc versa. It was next shown 

[6] that this policy was not only myopically optimal but also approximately optimal 

in the long run. Further, in a large number of cases that were tested by computer, the 

utilization and availability rates computed froni the myopic results also corresponded 

to the absoluite optimal values comnputed. 

This result established the principle that a distribution policy should seek to 

equalizc utilization rates and availability rates. This is also a policy that has thc 

essential elerents of "fairness' in equally spr!ading the nonavailability and 

nonutilization risks among hospitais regardless of their relative size and is consequently 

a highly defensible policy. 

Finally, it was shown that the highest possiblc regionil availability and utilization 

rates are achieved whcin the desired inventory level for each blood type in each 

HBB is at the value that minimizes the total number of rotational units that are 

required to achieve these availability and utilization rates. 

It is a straightforward effort by computer to calculate the combination of inventory 

level and schcduling factor that requires the minimum number of rotational 

units. The minimum number of rotational units required to achieve a fixed utilization 

rate of 98% and an availability rate of 95% are indicated by the points connected by 

the straight line segments in Figure 2. Thc irregular behavior of this solution is due to 

the fact that inventory levels must be integer values and rounding occurs on very 

small absolute values. As an example, the minimum rotational shipments required to 

an HBB of mean usage of 1.5 units daily to obtain the target goals above occur when 

the desired inventory is 5 units, and the scheduling factor is set to 0.67. The trend 

line which is drawn in the heavy line in Figure 2 is meant to indicate simultaneously 

the optimal values in inventory level and scheduling factor for given values of mean 

usage. 

Adding operational constrainis 

The above distribution model of equalizing availability rates and utilization 

rates among the HBB's is illustrated by the two curved lines in Figure 3. The upper 

curved line shows the minimum total shipments required to achieve a fixed 

availability rate at a HBB of a given mean usage. The lower curved line shows the 

maximum retention shipments to achieve a fixed utilization rate. The area between 

the curves would have to be met by rotational shipments. As can be seen from the 

right end of the curves where the tails meet, this results in a situation where the larger 

usage HBB's receive almost all of their shipments in older retention units, while the 

smaller usage HBB's receive almost all of their shipments in fresh rotation units. 

UITERFACES November 1979 

- 314 -

- 315 - 

rcquired Ior shipping and oilher piurposes, is iiscd to creale data files. Fronim hese data 

files, and using lihe model described above, "policy selection tables" are generated. 

These tables indicate the minimum total fresh supply needed to be distributed on rotation 

in the region over a two-week period, in order io achieve certain *acceptable' values for 

availability and utilization. On the basis olf hese tables, of the amount and the stability of 

the collections, and of the reserve to be kept at the RBC, the attainable values are 

determined. and the "target" values are selected for these performance measures. 

'-TO INITIALIZE SYSTEM 

Q 

I:l¢itt:l-i 4. Pnogrnlíinicd B l oold Distribution Flowchart. 

CREATE DaT* FILES | 

_J~~~PREPA 

SELE( 

IN STEADY - STATE 


- 316 - 

Once thc planning phase is completed, hospitais are assigned to delivery routes. 

These delivery routes are constrained lo provide for deliveries lo aill hospitals at a fixed 

time each delivery day. Delivery day intevals are cither one, two, or four days, 

dclpendiíng on ihe size of lhe hospital alid special requirements or strong preferences 

expressed. Froni these assigniments "regional distribution sumniaries" per delivcry day 

are prepare(l for eval:uation. Thlcsc ae revised as needed to equalize the amnount of blood 

distributed cach day as far as possible. Individual hospital "sumnmaries of delivery 

schcdules" anid of 'desired inventory levels" are then prepared and sent lo all HBB's. 

After discussions during which the above distribution schedules are confirmed or 

modified as required andtl extensive educational sessions with the HBB's' management 

and operational personnel take place, thle operation is ready to start. The final step is the 

preparation of packing documents which are prepared in the order in which deliveries are 

to be made. 

Once operational, da¡ I'files or dislribution schedules are modified by one of two 

meanis. As new hospitais are added, as hospitals are removed, as changes in usage occur, 

such as those from incrcased bed capacity, or as chlilages in us;age are deleccd by hle 

control procedutire, i¡e tl;al; I'iles ae muodilied. 'I'he olicy selectiomn tables are revised 

eacl tiIme tilere ale revisiois ii usage estimates. The revised tables are then nmanually 

evaluated lo determine if the changes are substantial enough to require a change in 

targets. A change in targets may also be required if a substantial increaise or decrease iii 

the blood supply is anticipated froni othler iniformation. If such changes are required, then 

the regional distribution forecasting procedure is performed again as described above. 

Scheduling deliíveries 

A major advaintage of PBDS, both to the RBC and to the HBB's, is the ability to 

preschedule most deliveries. Prior to PBDS being implemented, a number of delivery 

vehicles were dispatched as orders came in. For urgent orders, vehicles were dispatched 

immediately, while for more routine orders an attenipt wa.vs made to hold vehicles back 

until several deliveries in the samne geographical area could be combined. This procedure 

was expensive and, perhaps more importantly, resulted in situations where even urgent 

orders were delayed, since delivery vehicles were not always available.during peak 

delivery hours. 

With the PBDS inost deliveries are prescheduled, and take advantage of known 

traffic patterns in order to mininiize delivery time. An interactive, comiputer-aided 

procedure was dcvised which assigns HBB's to delivery routes so as to meet their time 

and frequency of delivery requirenients. The twelve delivery day planning cycle is split 

into three groups of four delivery days, after which the delivery cycle repeats. In each 

four-delivery-day cycle each HBB receives either one, two, or four deliveries. Thíc 

procedure tries lo satisfy the delivery requircmenis without leaving gaps in consecutive 

time siots, since an emnpty time sloi indicates idle time. 

An opportunity to test the flexibility of this delivery schene occurred recently when 

the LIBS Blood Center was mnoved from one location to ainoiher several muiles away. 11 

was found that the delivcry roules couid be adjusted rapidly and the required reassign. 

ment of the HBB's was determined conveniently. 


- 317 - 

Controlling ¡he system 

The shortage deliveries and the rotational returns for cach Iospital arc nionitored in 

order to dltect chainges iii hospilal requircilecils. Evcy Iwo wccks thc "discrepaincy" 

hctween tIhe hospilal's cstinialtdt and expected usagc during these two wceks is computed. 

This discrcpancy is added to the cuniulative discrepancy frorn prior weeks to form 

an updated cumulative discrepancy. which, togctiher with thic nutiiber ol wccks included 

in it. ald ihe vyaluc ol thic lhospitall's expected usage. are used to compute the "normalized 

cumulative discrepancy" of this week. This last value is compared with a statistically 

establishcd "limit;" if it excecds the lirnit, it is concluded that a shift in usagc level has 

occurred. New usage cstimaUes arc conmputcd. and new distribution schedules are prepared. 

Otlherwise, no action is takcn. 

Since the mix of the eight blood types is a function of the ethnic mix of the 

population, which is in a state of transition for some of the hospitais tested, the above 

control procedure was established for cach of the blood types as wcll as for all types of 

whole hlood and all types of red blood cells. Adjustmenis were made either by blood type 

or for all blood types. It was found that this had the unfortunatc characteristic that if 

overall usage was increasing or decrcaising, individual blood types tended to go out of 

control in consecutive evaluaítioni cycles belore the total usage chart went out of control. 

liis causcd an excessive number of distribution changes. For this reason, the concept of 

"waming limits" and "action limits' was set up. A changc is only milde il onc of the 

blood types exceeded Ihe action limnits. However, at that time if any other blood type also 

excecded tlhe warning limits, then that distribution would also be changed at the same 

time. 

Daily inve'ntory ailjustment 

The resulting regional blood flow is illustrated in Figure 5. In this figure the 

aging of the RBC inventory is indicated down the center of the figure with the 

scheduled movement of blood to HBB's indicated to the left of the figure and the 

nonscheduled movenment to the right. The long-dated, stock-dated, and short-dated 

RBC inventories refer to blood units that are suitable respectively for rotation shipments, 

for retention shipments, and solely for supplemental shipments - which are 

filled by the oldest available units. The arrows indicate the blood flow that is normalized 

to 1,000 units collected. 

On the basis of this anticipated regional blood flow, the RBC's inventory is 

evaluated and adjusted daily. Stock-dated inventory balancing is performed late each 

afternoon after all rotational returns have been received. It involves the part of the 

flow circled towards the bottom of Figure 5. The available stock-dated inventory is 

compared to the retention shipments that are scheduled, the anticipated supplemental 

shipments plus a small reserve for unusual circumstances which is shown as becoming 

short-dated inventory. When the inventory for any product exceeds these requirements, 

the excess units are designated as surplus, and transshipped to the New 

York Blood Services (NYBS) division of GNYBP. When stock-dated inventory is 

tThe usage is determined by combining the known weekly scheduled distributions with the recorded 

supplemental deliveries and rolational relums to form an estimate of actual usage. 

INTERFACES Noveniber 1979

- 318 - 

I'Fu Itii 5. Regional B3lood Flow Baliacing. 

SCHEDULED DISTRIBUTION RBC NON-SCHEDULED DISTRIBUTION 

TO HBs i TO HBBs 

I NVENTORY \Ni / 

STOCK- DATED_ ~ INVENTORY 

STxo OO LONG-DATED 

V \ / ~~ I~NVENTORY 

LONG DATED ALANCING 

RBC 

RBC 

O T DAT 0ED 

OUTDATED 

DESIGN\TED 

below requirements then either surplus long-dated units (if available) will be retaíined, 

or the shortage will be made up from the other divisions of GNYBP if 

p~ssihle. 

Long-dated inventory balancing is performed each morning after the bulk of the 

bloods collected the previous day have been typed. It involves the part of tile flow 

circled at the top of Figure 5. The long-dated regional inventory which is expected to 

bccome available during that day is compared Lo the commitmcnt ol' units lor 

scheduled rotation shipments plus units required to meet open heart surgery needs 

(which is a specialized procedure where only fresh blood units are suitable). Any 

units in excess of these requirements are either retained to make up for shortagcs in 

stock-dated inventory as discussed above, or are made available for transshipment to 

other divisions of GNYBP. Since LIBS collects in excess of its needs', there is 

generally a surplus of rotation units especially in the more common blood types. 

Computer operation 

The effectiveness of PBDS depends upon accurate and timely data on operations, 

which is achieved in part by running the system on a niinicomputer, and by 

utilizing machine-readable bar codes on blood products and on test samples [ ]. The 

machine-readable codes on test samples are used in conjunction with automated 

equipment which performs blood type determinations and links the daita concernilng 

thc unit and its test results directly in the computer. This provides the earliest 

possible indication of what products will be reaching inventory during that day. The 

bar codes, which indicate product, blood type, identification numbers, etc., are also 

scanned as blood units arc shipped out and rctuirned, Lo maiilntainl perpttlul inventory. 

In this manner, the total inventory picture al the RBC is accurately maintained in real 

time. 

INTERFACES Novenlber 1979 

l~ ~~~~~-

- 319 - 

The light pen techniques are also utilized to control computer operations and lo 

identify the locations to which blood products are shipped or from wherc they were 

received. This is done by scanning bar codes on menu sheets that list all shipping 

locations, all types of transactions, and all available types of computer operations. 

This way, the computer is effectively used by nontechnical personnel, and any errors 

in entering data are minimized. 

In order to maintain a modular approach that can be utilized by all RBC's 

regardless of size, a network of miniconlpiuters is hcing used to h:ndic aHl the RBC 

blood processing needs. In a larger operation, such as LI3BS, thc operations are split 

functionally (i.e., separatc miniconiputers in the laboratory and disiribution arcas), 

while in the smaller centers a single minicomputer would handle all functions. 

Impact, Implications, and extensions 

Scientific in¡puct 

The successful operation of PBDS at LIBS has demonstrated the ability to set 

performance measures on the basis of regional planning. It establishes the first 

quantitative management guide for the selection of fcasiblc targets and slrategics. for 

the evaluation of options within a class of strategies, and for establishing the best 

possible performance within that class as a refercnce for the stratcgy selccted. It 

further demonstrates the ability to identify 4eviations from anticipated performance 

and, consequently, the ability to manage by exception. 

An example of this is the analysis of reported utilization by 28 HBB's that was 

performed recently. Statistical techniques indicated that the utilization rate pattern 

for 18 of the 28 HBB's were statistically not differentiable and showed an average 

utilization rate of 96%, which is close to the 98% value predicted; 6 HBB utilization 

rates were indicated as significantly below this norm and 4 utilization rates werc 

indicated as significantly above the norm. Since the same utilization rate is achievable 

for all HBB's under PBDS, deviations from the statistical norm can be confidently 

attributed to assignable causes. The utilization rates above the norm are 

attributed to sophisticatd techniques such as holding the same blood unit for several 

possible recipients concurrently. The poorer utilization performance of the 6 HBB's 

is attributed to poor blood banking practices such as failure to retum untransfused 

units to inventory promptly, or to special hospital practices. 

Management decision implications 

PBDS implies major changes in management decision `models both for the RBC 

and the HBB's. Most regions follow a procedure characterized by a decentralized and 

reactive distribution. In this mode of operation, the HBB checks its inventory status 

one or more times per day and, if it deems it to be low, places an order for additional 

blood. The RBC makes a decision whether its inventory is sufficient to fill the order. 

If it is, it delivers the requested quantity. If it is not, then it seeks to modify the order 

to a lesser amount, to substitute red blood celis for whole blood, etc. This "discussion" 

results in a modified order which is actually delivered. Both the HBB and the 


- 320 - 

RBC are making short-term decisions regarding each delivery, sometimes with conflicting 

objectives and a,-e rcacting to a situation more than anything else. 

PBDS allows the system ¡o operate in a predictive distribution mode. In this 

mode the RBC assumes responsibility for the long-term scheduling decisions, while 

each HBB assumes responsibility for ihe daily finc adjustments. Thc RBC starts with 

a distribution stralegy as described abovc. A copy of this, in the form of a shipping 

schedule and recommended inventory levels to be maintained, is furnished to each 

HBB. The HBB inventory checking is then reduced to two steps. First, bcforc each 

scheduled shipment is due, any inventory a;bove the desired levels is returned and 

second, whcnever the inventory is insufficient to meet demand, a supplemental order 

is placed which is automatically filled under normal situalions since it is assumed that 

it is heeded immediately. 

The supplemental deliveries and thc returns are monitored over a two-week 

period, at the end of which period a decision is made based on the control statistics on 

whether or not to revise the distribution. This is a further example of how PBDS has 

achieved a management by exception principle. 

Economic inipact 

* The economic impact of PBDS can be most directly measured in terms of 

improvement in blood utilization. Prior to thc implemnentation of PBDS, thc utilization 

rate in the LIBS region was 80%, which was also then the national average. 

Since the implementation of PBDS, the utilization rate for LIBS has improved by 

16%, while the national average has improved litile if at all during tihe same interval. 

The improvecient in utilization at LIBS translates to 80% reduction in wastage, and 

therefore, to annual savings of $500,000 per year. 

Of lesser economic impact is the reduction in the number of deliveries. Before 

PBDS was implemented, an average of 7.8 weekly deliveries were made to each 

hospital, all of which were unscheduled; after PBDS was implemented, the number 

of deliveries dropped to 4.2, but of which only 1.4 are unscheduled. By associating a 

$10 cost to an unscheduled delivery to an HBB and a $5 cost to a scheduled delivery 

(which is part of a route), PBDS has achieved a 64% reduction in dclivEry costs. Thlis 

translales to annual cost savings of $100,000. Additional important, though less 

tangible, costs savings are achieved by the implementation of sounder blood banking 

practices to reduce discrepancies between actual and achievable performance for 

individual HBB's. 

Probably the most important savings in the national health care bill brought 

about by PBDS is realized by improved blood availability to patients. Since deliveries 

to the hospitals are mostly prescheduled, elective surgeries can be prescheduled 

also, so as to minimize the number of surgeries postponed because of lack 

of the right blood products. However, savings from this improved availability are 

extremely difficult to estimate and quantify. 

National implications 

It can be seen from the model of Figure 3 that, in order to achieve high 

availability and utilization rates for the rarer blood types (small usage), most of the 


- 321 - 

shipnients have to be on rotation. In practice, however, RBC's can only schedule 

between 70% and 85% of the average blood collections for rotation shipments because 

of uncertainties in blood availability, and the need to keep an RBC reserve. 

Therefore, if thc amnount of rotational blood available for distribution is approximnately 

equal to that which will be utilized within the region (i.e., if u strict selfsufficiency 

criterion is applied), then for these blood groups either very low 

availability or alternatively a very significant rcduction iti utilization must he accepled. 

Since neithecr of these alterniatives is desirable, un alternative is to rotate nmore 

blood units in this blood groupl thal will be cventually transfused in the region and 

redistribute some portion of the returned blood units lo another region utilizing the 

inventory balancing techniqLues that wcrc described earlier. This is a feasible approach 

in LIBS since, like most regions thal are primarily suburban and rural in 

nature, it has the capacity for collecting blood significantly in excess of the overall 

regional needs. 

LIBS collccts approximaitely 20%/1 more blood than will be ultimately transfuised 

in the region. Virtually all of thc rarer bloods are first rotated to HBB's iii the region, 

while the excess collection in the more common bloods are mnostly immediately sent 

to the New York Blood Services (NYBS) division of GNYBP. This division encoiipasses 

the metropolitan New York arca and, like n1ost large urban arcas, has difficulty 

in collecting enough blood to nieet the needs of the imajor medical centers anid 

other hospitals in this area. The influx from LIBS is of significant help and the small 

fraction of older, rarer bloods thilt are reccived as stock-dated units can rcadily he 

absorbed. Thus the units collected in LIBS in excess of its own transfusion necds 

improves the quality of blood services in LIBS as well as in NYBS. On the other 

hand, the major medical centers in the New York Division of GNYBP also serve the 

needs of much of the population residing iii the arca covered by LIBS. 

It should be noted that not all major regions such as LIBS require interaction 

with larger regions. A planning exercise performed for the regional blood center in 

Richmond, Virginia, which services only seven HBB's, indicated that this region 

was remarkably self-sufficient. This is primarily because it includes a major medical 

center which accounts for a significant part of the regional usage. However, for the 

most part there is a.mutually beneficial interaction feasible between the smaller 

RBC's serving mainly suburban and rural arcas and the larger ones which will mostly 

service larger urban areas. 

These conclusions suggest that the concept of total regional self-sufficiency 

which has long prevailed in blood banking is in conflict with the goals of the Nutional 

Blood Policy that calis for high availability with efficiency. Rather, a modest level of 

resource sharing between regions as outlined above will work to the benefit of all 

participants. 

Extensions 

The success of PBDS has fostered a favorable climate for the further application 

of Management Science techniques both on a national and international basis. Thc 

International Society of Blood Transfusion Expert Committee on Automation has 


- 322 - 

reccnily endorsed the concepts inherent in PBDS and the Swiss Red Cross Transfusion 

Service is studying it as a basis for the creation of a national blood distribution 

system for that country. 

On the national level, as the PBDS strategy is expanded to the other three 

divisions of the GNYBP, it will provide an opportunity to more thoroughly investigate 

a broader regional network. Concurrently, other regional blood programs are 

investigating the use of PBDS. The American Red Cross Blood Services - Northcasi 

Region has already implemented a distribution plan for its six RBC's based upon 

the PBDS concept. These implementations represent broader geographic groupings 

of RBC's which, together with other such groupings, could eventually interact 

through a hypothetical national blood clearinghousc. 

Such developments would contribute significantly towards meeting the objectives 

of the National Blood Policy. It would act to alleviate the "national blood 

shortage" which, in part, at least is thought to be not so much a national blood 

shortage but rather the lack of a national logistical system for moving blood from 

where it is available to where it is needed. They would thus contribute to a reduction 

in the consumer's hospitalization costs and the delivery of higher level health care. 

Acknowledgements 

We are indebted to Aaron Kellner, M.D., President of the New York Blood 

Center, Inc.; Johanna Pindyck, M.D., Director; Robert Hirsch, M.D., Medical Director 

of the Greater New York Blood Program; and Thcodorc Robcrtson, M.D., 

Director of Long Island Blood Services, for their wholehearted collaboration and 

encouragement. We are also grateful for contributions made by Professors Cyrus 

Derman, Ed Ignall, Peter Kolesar, and Donald Smith of Columbia University, at 

various phases of this work, and for helpful comments and encour4gement provided 

by Professors Howard Kunreuther, William Pierskalla, and Morris Cohen of the 

University of Pennsylvania. 

This research was supported in part by the National Institutes of Health, National 

Heart, Lung and Blood Institute Grant No. HL 09011-16 at The New York 

Blood Center. 

REFERENCES 

t11 Brodheim. E., "Regional Blood Center Automation." Transfusion, Vol. 18, 1978. pp. 298-303. 

[2] Btodheim, E., Hirsch, R., and Prastacos, G. P., "Setting Inventory Levels for Hospital Blood 

Banks." Transfusion, Vol. 16. 1976. pp. 63-70. 

1[3 Brodheim. E.. Derman, C., and Prastacos, G. P., "On the Evaluation of a Class of Inventory 

Policies for Perishable Producis such as Blood," Management Science. Vol. 21, 1975, pp. 1320- 

1326. 

[4] National Blood Policy, Department of Health. Education and Welfare, Vol. 39. 1974. 176. 

[51 Prastacos, G. P., "Optimal Myopic Allocation of a Product with Fixed Lifetime," Journal of the 

Operational Research Sociery. Vol. 29. 1978. pp. 905---913. 

[6] Prastacos, G. P., "Allocation of Perishable Inventory," forthcoming in Operations Research. Also 

available as Technical Repon No. 70, Operations Research Group. Columbia University. 1977. 

[7] Prastacos, G. i. and Brodheim, E., "A Mathematical Model for Regional Blood Distribution." 

Working Paper 79.02-91. Depanment of Dccision Sciences, The Wharton School. University of 

Pennsylvania. 1979. 


- 323 

THE CREATER NEIW' YORK BLOOD PROCRAM ¡ 

The New York Blood Center ' American Red Cross 

Dr. RaJ Nigam 

RCA 

David Sarnoff Research Center 

Princeton, New Jersey 08540 

Dear Dr. Nigeam, 

310 Esst 67 Street. New York. N Y. 10021. 794-3000. 

August 3, 1979 

The Greater Newv York Blood Program,the largest blood program in the world, 

has assumed the responsibility for the blood needsof 18 million people 

residing in New York City and in much of its surrounding area. We recruit 

people to donate their blood, collect this living tissue from them, process 

it in our laboratories and distribute it to the 262 hospitals in our community 

vhere it must be available to care for the people in need. Our aim is to have 

enough blood of the right types on hand to meet the usual hospital demands and 

the emergencies vhich may arise, but not so much that it vill be vasted since 

blood can only be stored for 21 days outside the body. We face the major 

problem of hov to maximize the availability of blood to each of the 262 

hospitals that ve service vhile effrectively discharging our implicit covenant 

to our donors to see that their gift is efficiently utilized. 

We turned to management science to help us achieve this goal and provided 

the Long Island Blood Services Division of our program as a test site. The 

system that vas developed has been in operation for two years. It integrates 

the blood supply of a region by developing a complex pre-planned rotational 

system so that the availability of blood of all types becomes the ssae in 

hospitals regardless of size and usage, and utilization of the blood also 

becomes the same. 

Patients in the community have equivalent protection for blood services at all 

the hospitals vithout burdening some of them vith excessive vaste. This system 

has given us a mcechanism to do exactly vhat ve urgently needed - Takc the Crisis 

Out of Blood Banking. We have offered to our hospitals a rational, effective, 

functioning system - vhich vill soon be extended both throughout our service 

area and elsewvhere i the United States and abroad. It is not unfair to state 

that this program can and vill contribute to an improvement in health care 

throughout our nation, and therefore touch each and every one of us at some time 

in our life. We are proud of our achievement and appreciate this recognition. 

JP:ck 

I - 

Sincerely, 

Pindyck, N.D. 

ie President snd Director 

The Greater New York Blood Progrsa 

I ' ls - -II - IP B - - - -· ~·s~WI 


"'1 .'. 

- 324 - 

Using Computer Simulation 

to Predict ICU Staffing Needs 

by N. Duraiswamy, R. Welton, and A. Reisman 

Tis aricl dacritbe a compurt imubdon study to deiwufs 

¡he ambr oM!f fuff-¡u ;te:t d nmra sequdt 

to ddv~r afe nur h a 20.had medical 

_Lawev mr usat n Le asheonr' compua dmul~ion 

madd p d/ffwu, stafrdu r i* /for two sLr- 

MIh* lo np a e aborot~l he.ao tMhe w 

ddi wn am omplhd ad ndisuahe dost omd 

/fffO, f/ ,, o, t dmt loa ohr e t. 

ami~p. 

The rfi nutning cervi.e stadrd of the Joint Commisaioa 

cm Acdliuati of HoSptir (JCAH) cabll for "a 

fflca mnumber of duly iUcnsed rqlered nures on 

dry all en a ... to live patient the nurnin car 

timt mq~ir the judgmant ud pcilkted sklil of a 

_ ~tr au m"ll¡1. Many rect muanmcnt dudk-1 

4e a mued tn to define how many nurnes consImuem 

a ufllciem ?umbur. 

m qh aay bpita, nurdn rnaffins is done -lmost by 

tidon;a~ maa addas, wbrmt. and r aem.,la 

taff on the baais of a ubjective prception of what 

msems o~siary. Nurs are often reaacned from noto~uy 

uare to cover busy ars$l5. However, in thc 

mitlve cue unlt (ICU). whau ou must be knowlodgab 

a caxtive, observant, diQnostic, and therapWdc 

daona of altial care(6), reasniln nunrcs 

from other au~ can jeopardlte c unit's standard of 

are. Ths intuiti approah to rtafMng al o difficult 

N. O. D , PaD., ik iata porI or Of bulesa m ladr 

Uam V, Ko. .k ~recdd u MS. md Ph.D. la opea 

da_ mrerb kr~ Cau Wemua Ium Ui vduy. 

IL Wll, LNR., IM. LN., bk a m,~t dese ~uind e lba e n Prono 

P~ lid Sodd do Nl~ . Can Wqm a~maere U-ailty. 

A. l5m , P.D., P., kL pmomw o 

Wmm Iammw Umial~y. 

o opa~i rtuar., Cmu 

*19SI The Jounml ot Nunda Admitnmiwn 

Thb Joural of Nwiqa Admimmu / Feb~ruy I1N 

ID translp i to quantifiabl terms that are uaderstandable 

to othrs, including hospital administrators, boards 

of trustees, and other nurses. 

This article describes a system-analytic study of the 

nurse staffing needs of the Medical Intensive Care Unit 

(MICU) at the University Hospitais of Cleveland. la 

consdering how to determine the efficacy of various 

staffmrn alternatives, we decided that a computer simulaion 

woud be the best method. Computer simulation 

acraes a computer version of the real system and provides 

laboratory conditions in which decision makers 

can tet lternative policies. Simulation allows the testing 

of muffina choices in the light of the conflcti~g 

objective of cost continment and quality of care. 

Morcover, it allows consideration of variations in patient 

admission and discharge ratcs, lesnth of stay, 

leves of acuity, and levels of nursing care requirements. 

Traditional assessments for staffing 

Syutematic mcthods for staffin8 have taken several approche. 

One approach takes such dements as diet, 

toileting, vital sign measurement, respiratory aids, hyiene. 

suctionn,. and turning and/or positioning(7-9) 

and applie a wcigbted point *ystem to these uarcas of 

care. Ihe point total is then converted to some unit 

equlvaat o a specific amount of nursing care time. 

Other approaches have lumped all care categories tother 

and subjctively established required levels of 

care ranfin8 from hih to minimal(0l] without def'ning 

the criteia to be used. An "hours per patient-day" formula 

multiplies the number of patients by standard 

houn of nunin care, such as 5.0 per patient, and then 

multiplies this total by 7 to obtain the required hours of 

nutinl care per week. Another approach uses patient 

care categoric such as minimal care, partial care, modrate 

care, complet care, and intensive carell 1-131 but 

includc criteria for indirect ad supportive elements of

_,~ .- 

- 3. 2- - 

TABLE 1. PATIENT CLASSIFICATION 

Ca Re Aummt : : Wmd Poes .o4rdilnO to LCr 

Cl'"'y ' ", ' 

.ol tCue 

Intse HIlgh Moderaot e Minimal 

Physlcal Dp-idan~ 2.4 1t48 11. 4 

Obwvakn .8 19.6 OA 

8octEmot,~loal Ce 2a0 18.0 10.8 4. 

RshFIIUorTeaCNno 22.0 1L2 924 4.4 

the nursing process, thus allowing a comprehensive asscssment 

of actual care required. Such a comprehensive 

approach forma the basis for the patient clasification 

ysiyem developed a University Hospitals of Clveland, 

and this approach was used in the study described here. 

Baekgrlound 

¡be University Hospitals of Cleveland include a 

1000-bed ;ace c medicl center located on the easi 

aíde of the city of Cleveland. lts aix specialty hospitals, 

il umd om cnra maamet, are closy afrilated 

with theb profesional and graduate schools of the healih 

adc disciplnes of Case Wern Reserve Univerity. 

Tle Medicael Itensive Care Unil (MICU) of Univerity 

Hospital conlir0 of two patent care arueu: thc intmelve 

cm unit (ICU) where crically ill patienut are 

biitally admitte and cared for during the acute tage of 

their illns, and the progrceive care unit (PCU) where 

p~rtls ar e cared for lfter the acue stae of Iheir illnm. 

Both units were desigpd to accommodate 10 patins 

ecah with 10 single roomr in the ICU and 5 double 

roomu in the PCU, yielding a total posible census of 20 

patient for the MICU. 

After the firt year of operation at 100 per cent occu- 

TABLE 2 PATIENT CARE CATORIE8 

ACCORDIN TO TOTAL POC 8CORE 

To~ PO 8Moe CdaMV 

80or dox, 

m0a40 

o.~ 40 

Mt~- 

.~Modmmm 

mlml 

pancy, the 44 budgeted full-tinic nursing positions 

proved to be inadequate to provide safe nursing care in 

both Ihe ICU and PCU when both were at 100 per cent 

occupancy. For this reason, 4 beds in the MICU were 

closed to patients, reducing the total possible census 

from 20 to 16 patients for both units. We undertook our 

study to determine how the original 20-bed unit could be 

restored without resorting to overstaffing or understaffing. 

Data collection 

We used the department of nursing's Patient Care Categorization 

(PCC) assessment tool, with the charge 

nurse as the source of data. This classification system 

uses four categories of care requirements: physical dependency, 

observation required, social/emotional requirements, 

and rehabilitation and/or teaching requirement. 

Within each of these categories, there are four 

levels of required care. The charge nurse chooses which 

level of care each patient requires within each of the 

four categories. 

A total acore, called a PCC score, is then obtained by 

adding weighted points for the level of care chosen in 

each category. (See Table 1). Based on the ¡otal PCC 

acore, a patient is assigned to one of four care categories. 

Table 2 lists total PCC scores and their corresponding 

care categories. 

Using thesc four categories, the nursing department 

developed the policy for nurse-patient assignments 

(Table 3). This policy provides adequate patient-nurse 

rations to ensure safe care and is currently used for 

nurslng assignments in the MICU. 

The PCC tool had been in use for two years prior to 

this study. During that time, ihe PCC assessment was 

done one week of each month on the 7:00 A.M. to 3:00 

P.M. shlft by the MICU charge nurse; this sample was 

The Journl of Nun'in Admitbttion / February 1981

_;-inarP - - i - i ,_I 

found to be representative of data obtained from initial 

daily assessments using the PCC tool. Wc computed the 

probability of a 'atient's being in. a specific care category 

and the nim ber of patients in a specific care category 

per day. 1. ing these data, which spanned two calendar 

years, w, identified seasonal variations, such as a 

greater numbez of admissions of patients with diagnosis 

of drug overd -se or myocardial infarction during the 

spring and sur. mer months. 

The Daily E atient Census Forms for the ICU and 

PCU were usec on a sampling basis to obtain admission 

The Jomeal d NWnY4 Mmisuun.i IA Fe~uay I11I 

- 326 - 

FIGURE 1. FLOW CHART OF THE SIMULATION MODEL 

and transfer or discharge data. The sampling included 

al patients admitted on 112 randomly sdelcted days over 

a onc-year period (1977-1978) to determine actual 

Iength of stay in both units. It was feldt that this approach 

would provide a statistically significant sample 

yet minimize data collection time and effort. We then 

used this data base to determine the length of stay distribution 

in our simulation model. 

There is a significant fluctuation in the daily patient 

load in the MICU. Consequently, one would exppet a 

similar fluctuation in :"s daily nurse staffrang nceds. The 

| LENGTH OF STAY 

-- - DISTRIBUTION

TABLE 3. 

MEDICAL INTENSIVE CARE 

UNIT (MICU) POUCY 

FOR NURSE-PATIENT ASSIGNMENT 

- 327 - 

Nue.patlent aslgnment wlll be made In temis of 

pallent PCC ecore and wlll comply wllth ihe core and 

nuepatient ratlos alletd below. 

Patent Catm 

Catogory Score 

Intense 

HNh 

ModSate 

Minimal 

NurcePatlent Ratlo 

1:1 

1:1 

1:2 

'3 

The above &aalgnment policy la applicable lr all ahilis 

in bolh hb ICU aud PCU. 

patient mix in addition to patient load determines :he 

staffing requirement, but the random arrival of patients 

and the random length of stay suggest that the patient 

mix probabilities may not vary every day. 

The simulatien model 

Fgure I ilUustrates the simulation model. Patients are 

admitted to the ICU for a certain length of time, as determined 

from the length of stay distribution for ICU. 

At the end of this period, they are edher transferred lo 

PCU or dichuared. The length of mty of the patient in 

PCU is aiain determined, using the length of stay distrib¡tion 

for PCU. 

Since no additio of beds is planed for either the 

ICU or PCU, capacity constaint at the current ten-bed 

lecL for each unit have been imposed in the model. Accordil4 

to current practice, when all beds are occupied, 

thbe let sick patient is transferred from the ICU to the 

PCU in order to admit a newly arrived patient. When a 

siair situation arises in the PCU, the least sick patient 

la transferred out of the PCU. The simulation model 

coalam tbe blsic to hade tbee úguations. 

The nursing cae requirement for each patient i¡ expred 

in terms lin and of ou ente the simulation 

model to determine the total nursing care hours needed 

in a shift. The model asmumes that patients in an intensive 

care setting wih require the mame level of care during 

the otber two shifts of the day. 

The simulation covera a full year. We designed the 

imulation modal to print out the number of nures 

neaeded on each simulated day during the one-year 

paiod. lao ddition, the model generaed descriptive statistical 

information about the average number of nurses 

required u well u the standard deviation for any 

chosen period of time (Figure 2a-d). 

Validity of the model 

The simulation model was validated using the historical 

data base from the MICU. The simulation results are 

therefore valid for prescriptive purposes if the utilization 

patterns do not change. However, if the utilization 

patterns are projected to increase or decrease, the simulation 

can be used as a vehicle to develop new plans for 

staffing. 

Analysis 

The total number of patient days for both l..e ICU and 

PCU is 308 as determined by thc analysis of 112 randomly 

selected days during a one-year period. The patient 

mix probabilíties used in this study are shown in 

Table 4. hbe values for patient mix probabilities were 

obtained. trom the data collected by the department of 

nursing over one full year during 1977. The patient mix 

probabilities for the winter, spring, summcr. and fall 

seasons wcre obíained separately to capture seasonal 

variations in the staffing needs. During the course of the 

study, data on the patient mix probabilities were collected 

individually for both ICU and PCU through direct 

observation. However, these individual data werc 

not used for the simulation because the nursing care required 

for a patient is the same in each unit for each patient 

category. 

Staffing by seasonal demand 

As evident in Figure 2, the variation in staffing requirements 

between fall and winter months is not significant. 

The average requirements for spring and summer were 

identical with only a small difference in the standard deviation. 

However, there is a notable variation in staffing 

requirements between the combined fall/winter and 

combined spring/summer periods. These results indicated 

a need for two staffing levels, each for a sixmonth 

period. Since there is no specific staffing level 

which meets the staffing requirenent optimally on each 

day of the six-month period, we offered the data in Tables 

5 and 6 to show the percentage of days of overstaff- 

TABLE 4. PATIENT MIX PROBABILITIES 

.... ,, ' '.'' . POAB ' 

etiCos 

kI~ 

wihn Who, 

0.31 

. .. Sum;i, . 

0. 0 ' ' 

a.:: 

;' '0 04.3 ' 

_ ;ode 033 0.12 0.18 0i0 

;Minknoa 0.13 0.03 0.02 0.13 

I hc Journal uo SNui\ln F AdJmnmlnmtr¿atn Ihraiy / 191

ing and understaffing associated with each staffing 

level. This information allows management to weigh the 

impact of alternative staffing levels against cosís and at 

the same time to comply with HSA and JCAH standards. 

Scheduling complications 

The average staffing levels of 30 nurses per day for fall 

and winter and 36 nurses per day for spring and summer 

appear most desirable in terms of overstaffing and 

understaffing. The large turnover of nursing staff during 

the spring and summer months (approximately 15 

terminations or transfers out of the MICU between 

March and September of 1977)1161 may have have implications 

for scheduling vacations for experien.. 

I 

lo 

00 

c. F mr (Augul) 

a 10 lo la 25 

DOY 

Ibe Journal oí unran Admenut l Fwbrlwuar) #lvTI 

- 328 - 

- I -·V - i-- 

FIGURE 2. SEASONAL M.CU 

NURSING STAFF REQUIREMENTS 

- 

* ^4W : 38 

810. Der. ' 2 

nursos. Ncw graduate orientation to the MICU and education 

in critical care, which also occur during spring 

and summer can increase the average number of nurses 

needed during these months. 

While the simulation resultod in recommended staffing 

of 30 or 36 nurses per day, levels which are lower 

than the actual budgeted staffing (44 full-time nursing 

positions per day) for the MICU, several considerations 

must be noted. First, the staffing levels suggested by the 

model do not include provisions for vacations, days off, 

holidays, illness, absenteeism, and staff development 

commitments. Second, the staffing recommendations 

apply to situations in which every nurse is experier, -d in 

critical care, has successfully completed the entire ICU 

orientation, and has demonstrated safe, indc-endent 

¡ 

1 O. 

80 

4O 

20 

t0 

'llo~~~ ~Std. 

AWae : 3JO 

DeW. : .S 

MtulnYm: 42 

5 

Fell (NOvmb 

0 1S 

%Y 

20 25 30 

W_

Nursing Stalf Level 

(i of nurss/day) 

26 

28 

30 

32 

34 

Nursing Stafl Level 

(I of nurseacday) 

30 

32 

34 

38 

38 

.40 

- 329 - 

... , I I I I 

I _i 

TABLE 5. ANALYSIS OF STAFFING LEVELS (FALL AND WINTER) 

% of Days of 

Overstatffing 

18.8 

37.2 

39.4 

58.8 

72.2 

% of Days of 

Understaflfing 

81.2 

62.8 

41.2 

41.2 

27.8 

TABLE 6. ANALYSIS OF STAFFING LEVELS (SPRING AND SUMMER) 

% of Daya of 

Overtaffling 

5.6 

15.6 

27.8 

28.9 

52.2 

71.7 

practice without close supervision. Finally, the recommended 

stuaffing levels are for fully qualified nurses for 

dircm patient care; thcy do not include provisions for 

ICU nurses away from the bcdsidc involved in administrative, 

educational. o* tsther indirect patient care activities. 

Costa of tLe atudy 

Since this study was conducted to fulfill academic requiinmets 

to study real-life problems, there were no 

out-of-pocket couts for development of the model and 

its ooinmputer implementation. Data collection, analysis, 

cmiputer programming, and computer tíme expenses 

would occur in other settings. Nursing time used for 

collection of data used in this study was approximately 

rive minutes per day for seven consecutive days each 

month. However, this responsibility has bcen a regular 

pat of the MICU charge nurse role for several years. 

Data analysis and programming of the simulation 

model required less than one person month. 

Applying ihe computer simulation 

la other sietdis 

Although this study was carried out in a large urban 

medical center, Ihe same approach can be used in small 

community hospitais. A staff of inhouse cngineers anJ 

prorammers i not n:csury. Small community hos. 

pitais with access lo local universities can use universitybased 

expertie to perform such studies. Moreover, 

many public or private universities are looking for reallie 

experiences for both faculty and students and will 

often perform sludies on a gratis basis. 

The seasonal variations in ihe dita collected from a 

% of Days of 

Understafflng 

84.4 

84.4 

72.2 

4".8 

47.8 

2S>3 

single institution should be considered illustrativc only. 

The approach, however, is transferrable to other instituiions, 

other services, and other seltings. 

References 

S of Days of 

Optimum Staff 

o 

O 

19.4 

o 

o 

% of Days of 

Optimum 5.1ff 

10.0 

o 

o 

23.3 

O 

o 

1. Accredrod ion Moanualfor lIospitols. Joint Commission on Accreditation 

of Hospitais Chicago. 1973 (updated). 

2. Ryan. T.. Barker, 8. L.. and Marcianto, F. A. A system for determing 

appropriate nurse starfing. J. Nurs. Admin.. 5(S):30-38. 

1975. 

3. Norby, R. B., Freund. L. E.. and Wagner. B. A nurse staffing 

system based upon assignment difficulty. J. Nurs. Admin.. 

7(9):2-24. 1977. 

4. Clark. E. L. A model of nurse staffing for effective patient care. 

J. Nurs. Admin.. 7(2):22-27. 1977. 

5. Norby. R. 8. ec al. 1977. pp. 2-24. 

6. Burrell. A. L.. and Burrell, L. O. Criricol Core. Si. Louis: Tlhe 

C. V. Mosby Company, 1977. 

7. Clark. E. L., and Diggs. W. W. Quantifying patient care needs. 

Haspiras. J.A.H.A.,. 9:96-100, 1971. 

8. Poland. M., el al. PETO: A systein or assessing and meeting pa. 

tient care needs. Am. J. Nurs., 7:479-1482. 1970. 

9. Willilms, M. A. Quantification of direct nursing care activities. 

J. Nus. Admin., 7(8):15-18. 1977. 

10. Heara. C. How many high care patients: Deployment of nursing 

utaff. Nursing Times, 17:504. 1972. 

II. Norby. R. 8. a al. 1977, pp. 2-24. 

12. Phipps. W. J., and Philps, C. Definition of Terms Used lo Classify 

Patiens According to Their Nursing Care Requirements. Departnent 

of Nursing, University Hospitais of Cleveland. 1977. 

13. Pardee, G. Classifying patients to predica staff requirements. 

Am. J. Nurn.. 3:517-520, 1968. 

14. Zleers. L. J. Calculating a nurse staffing budget or a 20-bed unit 

at 100%e Occupancy. J. Nuns. Admin.. 7(2):11-14. 1977. 

15. Newman. M. A.. and O'Brien, R. A. Experiencing the research 

process via computer simulation. Image. 10:5-9. 1978. 

16. Alekn, M. Annual Rportn. MICU. Unpublished paper. Dcpart. 

ment of Nursing. Univerity Hospitals of Cleveland: 1-6. 1977. 

I hc Journal of Nurnms Adn:initration; February 1981

1 9 

Michael B. Harrington Forecasting Areawide 

Demand For Health Care 

Services: A Critical 

Review of Major 

TeChniques And 

Their Application 

Since well before Cassandra, people havc 

sought to forecast the future. In modem times, 

forecasting methods have adopted fairly scientific 

guises. But it is not clear that their predictive 

accuracy has improved appreciably over 

.hat of their predecessor. This is especJally 

utre of methods applied lo large-scale public 

problems, such as areawidec health planning. 

Improved methods have been offset substantially 

by the greater complcxity, more rapid 

change, and more pervasivc interdependence 

characceristic of modern society. Forecasting 

skill have grown, but so has the difficulty 

of the ¡;.',. 

The impetub for developing improved methods 

for use by areawide health plannera, begun 

nder tbc Comprehensive Health Planning 

lesilation of 1966,' no doubt will be reintorced 

by the Health Planning and Resources 

Developmct Act of 1974 (P.L. 93-641).2 

M1mut . Hia po~ i D, bs Senior Auociat ltot 

the Sysclm Evaluallon Departnen, CACI. Inc.- 

Federal (Arlaliton, VA 22209), and Professor of 

ialth Economcs, Souhetemrn Unlveity. 

fai marerch or thbis piapr w completed under 

Contract liA 230-76.00~ aeored by the Bureau 

Healthb Planaing aMd Rmurces DevelopmcnM. 

Health Resources AdminiArítion, DHEW. In de. 

veop~n ibs par, Ub aubhor benited grealy 

(mam prticipatíAon ia eries o( three ympoi~ Wpon. 

ored by thibe Bureau of Health Planning d Re. 

bouets Developenl, DHEW. durinl the fali of 

1974 und prling of 1975. Of coun. neither BHPRD 

anr tib other rympoml paricipans are responlible 

for shorioominp thir paper may cntain. In additbn. 

aluable crituncsm was rendered by Sarh P. 

Far of the WAublni Office of Arthur Young 

& Comniay; Wiliam Orfes. Dcprnment ot Medci.ne 

and SurUlry. VA(Mas Admnlnt(mbul . WuashJion. 

D.C.; and Dr. Chester McCall. Manager. Syucms 

Evaluation Dcpanmen, CACI. Inc.-Federnl Arlingl 

to. VA. 

- 3 30 - 

__ ______________ ______ 

Somewhere betwcen perfect forecasts of the 

health care futurc and none at all lies a "zonc 

of feasible forccasting" (i.c., forccasts both useful 

and within the capabilities and budgcts of 

health planncrs). This paper explores the zone 

of feasible forecasting, by considering the following 

points: 

O The objectives of forecasting in areawide 

health planning; 

0 Specific factors contributing to thc demand 

for health services that enter into any serious 

forecasting effort; 

O Major sources of change and continuity in 

the areawide health care system that are 

identified in conjunction with demand forecasts; 

O Six generic kinds of forecasting techniques 

in the health planning context, including 

an evaluation of the strengths and weaknesses 

of each; 

C A simple framework through which each 

forecasting method might be brought to 

bear most effectively; 

o Some apparent dimensions of the "zone of 

feasible forecasting," given the current state 

of the art. 

Objectlves of Effective Forecasting 

At least three major objectives for forecasting 

areawidc health service requirements recur in 

public discussion:' 

1 Improving the foresight and skill with which 

public and private health care resources are 

allocated. Improved forecasts would help 

achieve lhis objective primarily by providing 

an arcawide "picture" of future necds 

1

Table 1. Factors ¡i denad for medical care* 

- 331 - 

Foecuing Health Core Servnic 

Fadcon afieelng a Faton affecting Healt h vkl 

pfhedm'd'da.and for pbyddau' prel~ ualy demade ed 

berth ernks d albetb tneaices m · resul 

1. Incideace of ilne or injury . Patient chamctcrisia. including 1. Physician cae in an of fice or 

2. Severity of iliness or injury relative cost to the patient of other outpatient setting 

3. Cultural values and atitudes using different health services 2. Emergency care 

4. Demographic factors 2. Physician's institutional 3. Inpatient hospital care 

5 Costa (tactorn affliatiluons 4. Referrals to apecialized ervices 

J. Cast fab~cío ~ 3. Physciar's knowledge and long-trm care facilities. or 

6. Perceived accessibility. actitudes other care-giving institutiona 

availability aad convicnicncc 4. Relative costs to physician of S. Home health services 

of service provider prescribing alternative health 

7. Family. peer group. or other services 

authoritative pressure 5. Ava·ilbility of prescribed or 

8. Life stylc facions affecting desired services 

health 

These factorn are not necessarily ranked in order of imponrtance. 

an, demanda into which public and private 

objc.ives can be harmonized. 

2 A'pneating both quality and cat control 

in medical care. Improved forecasts would 

do thib by helping to avoid both alltion 

of "excesaive" community recourcer to some 

halth servic in the future, and the allocatio 

of insufficient" resources to others. 

3 Anticipatin Wnd planning for the impact of 

deveopiag tchnology and chag·ln method 

of organizatioa upon the health care 

spatea' 

These are ambitious objectives. Many argue 

tha bey are won uabitious for exi~iag forecastia· 

techniquea, that ourt u skilis are 

too limitd to achieve them. Thi paper explores 

this question in greter depth. 

C Pf Cfodb so b Dead 

fo Ham~ SeIrve 

"Need" and "demad" have recived considerable 

ateatioa in the litertum. TbI paper 

conccntrta upon the concept of demaud, the 

more tracable ol the two no frm fra a forecauing 

viewpoint.* Demad for health servia 

-g erally defind u the aount o serviesa 

actually used by indiidu in the commuaity 

-- a tuneioa d a oumber of economic , socil, 

pbycal and envioramml froa (cc 

Tale 1). Tbe. fason ae simgficat in any 

comprehensivc forcas o requirema for 

particular health services. But taken together, 

they presenat formidable problems ¡o the forecaster. 

Frt, the nature of each and the 

interaction among them are difficult to determine 

with pr~eion, especially ona an areawide 

basis.' Moreover, the influence of les 

readily measured variables ii Table 1 (eg.. 

peer group pressure, perc:ived accessibility) 

is very poo~ly understood from a predictive 

standpoint. Ssond, the importance of tbe 

two cdlumm varis in term of r Jainflu- ve 

ence upon service demand. Pacton in the finth 

columa tend to be more influential in thd demand 

for mergncy ad ambulatory services, 

tor example, while facton in the second column 

asumc greater importance when forecasting 

demand for inpatient and long-tenn carn. 

Third, the rlaive contribution madc by eotora 

wilhin each column is different from service 

lo srvic. Pamily or peer group pressure, 

for exmple, tends to be very ignifiant 

in the demand for long-term care, but less 

important in cmergency care situationa. 

To be fully useful, bowever, forecatiog 

techuiques should be able to deal with these 

demd-oriented complexities in a reasonably 

effective way. 

Ma Suoas dl Cofa Choa~e ad 

Cad k A cdqt Daem d for Seuice 

Miky fmaon Etmin demud ur 

e fidy m«ble, 

p _ticuilly in tbe bort nm. OtbeaM m ma

Inquiry/Volume XIV., September 1977 

- 332 

TabMe 2. Possible sowues of change and continuity in the communily health care system 

Soures of change 

1. Overriding problem(s); 

examples include: 

* Improving quality of care for aekcted 

target populationa 

* Gaining control over runaway costs 

* Redefining the balance of interestr between 

providern and consumenr of health care 

servics 

* Broadening access to care for selected 

target poputationa 

* Ovcrcoming fragmentation of service 

delivery palterns 

2. Prime Movers; 


* Medicare. Medicaid 

* Blue Cross. Blue Shield 

NHI, it enacted 

* Associationa of health care professionala 

(eg.. AMA, AHA, ANHA, etc.) 

· PirIO* 

volatile. When forecasting, it may be useful 

to etimate commuahty forces promoting change 

afiJ encouraging preservation of the nsatus quo 

i thdese factorn. Some commonly advocated 

stracgies include:' 

Forecetnlt Major Sowuce ol Chanse: 

A least w readily identified hitiaon d rof atifcamt 

chanp in the health ca system are 

sucepqible lo amalysis using one or mae of 

ti. lorecaf _l techaiques to be evaluated in 

Ihe exit sciohan: 

O T/1 OverrldMin Problem(s):. ldentify and 

evaluate thbe problm(s) tadng the com 

munity heath care yrtem thim re widAly 

thought to require solution ad thit, i 

"solved," wil result ia signif~nt chlagls. 

Por a sample of conmmoa probrs, mee 

Table 2. 

O The Pri~ Movrs: Identity and evaluate 

thase inmtituioas ad proceaes that strongly 

and indepemdtndy affect the way health 

car: is delivered (see Table 2). 

Forecasfta Malor Srces o0 CotLnulty 

Whil re me aspecs 'of the blth care system 

will chane within the plbaaal dme frame, 

perhaps dranatically so, most wlU notL It often 

Sources of continuity 

1. Organizational and demographic superstructure; 


* Instiiutional and professional licensure 

requirements 

* The complex of other federal, state. and 

local regulationa 

* Longestablished federal, state, and local 

financing programs 

* The demographic mix of the community 

population 

* The geographic dislribution of the 

population 

2. Kind and capacity of health care facilities; 


* Shonr-term, general hospital 

* Specialty hospitala and institutions 

* Nursing, personal, or domiciliary care 

establishments 

· Custodial care homes 

is helpful to identify asnd evaluate these aspects 

explicily.' ° Several generic sources of continuity 

include the so-called organizational and 

demographic superstructure and the kind and 

capacity of health service facilities in the arca. 

Both sets of factors change relatively slowly 

and both strongly affect demand for services. 

Table 2 includes examples of each. 

Of the problems in weighing these factors, 

perhaps the most difficult arise when appraising 

sources of change. There is a large amount 

of folk wisdom available on how to judge the 

streDgth of political and social forces. But successful 

measurement and precise forecasting of 

such forces is rarely (if ever) achieved. It is 

feasible to determine the general direction in 

wbich "trends" are going. It is quite difficult, 

however, to be correct about their magnitude 

and timing. This is complicated by the large 

number of decision-makers and their correspondingly 

largo measure of autonomy and 

discretion. 

Six Siglkt Forecasting Techniques 

So far, factor that contribute to demand for 

health ervica, ad the broader set of factors 

that contribute to both change and continuity 

in the community health care picture, have been 

described without reference lo how cach might

e forecast. This section descijbcs su ;aeti. 

techniques for dealing with each sct more systematically, 

and suggests the rJjour sti

Inquiry,' Voluine XIV. Septeimber 1977 

- 334 - 

E] There is very littlc thcory available about 

how one should forecast using historical 

analyses; liitlc or no formal methodology 

exists and few criteria for judging "good" 

forecasts are available. 

Delphiec Tecimniques 

Delphi is a method by which "expert opinion" 

is employed in systematic fashion. A wcllstructured 

forecasting problem is presented to 

a group of peoplc who are well grounded in the 

field or who are otherwisc "expert." Each 

one makes forecasts concerning specified probleats 

by scparately and anonymously employing 

specially prepared questionnaires. Results 

are summarizcd in a way that displays the 

range of responscs (usually in some quantitativc 

w..); these then are circulated among 

forecasters. The process is repeated for as 

many cycles as desirable in order to focus the 

anay of forecasts. All significant interactions 

~a.oel participante are in writing rather than 

face-to-face." 

Coskider a hypothetical example: An HSA 

situated in a rarpe. metropolitan area is concerned 

with p:a,.ning tt meet future demand 

for emergncy medical services. It has reason 

to think that computer-assisted diagnosis technique, 

currently beig discussed for emergency 

room use, wll strongly affect the way 

pl~ainb should be done. To explbr these 

iau, EMS experts employed Delphi at a 

workshop rsion of a joint physican, computer-industry 

symposium. They predicted 

tht by 1985 approximately 25 pacent of local 

ermen~cy room phycians will have acceas 

to conpute auisted diagnstic service. They 

employed expert opinion and data on licreas 

in local experimentation with the se of automated 

techiques in ithe medicl proeuion, 

toether with data on related EMS trndr. They 

arrived at thei forecas over a three-day per 

ad, employiag Delphi techdiques in tree 

iterations. 

Delphi has enjoyed caniderabe populait y 

in recent y~n, beinlg myd in a wide varity 

d f ortrctasi probln . Sone of th 

major reasons are sted bdow: 

Strenglihs Claimned: 

O Use of more than one "expert" (and more 

than one kind of :xpert) is possiblc in a 

well-structured w;,y. 

O The use of exrcrts is systematic; all confront 

essentiallv the same forecasting problem 

in the sa.ne context. 

O There is specific, purposeful intcraction 

among participants on the same problem. 

This tends to sharpen judgment and ideas 

and tends to reducc idle, unfocused speculation. 

O Disturbances or biases, owing to group dynamics 

or dominant personalitices, can be 

minimized. 

O Individual responsibility for the accuracy 

of forecasts (and hence possiblc inhibitions) 

is avoided. 

O The process itself can provide considerable 

educational benefit for those who participate 

through clarification of problems, issues, 

and goals. In addition, it can open 

communication lines between representatives 

of groups that ordinarily do not maintain 

contact with one another, or that have a 

history of inadequate communication. 

O The process can be especially effective in 

dealing with prediction problems that require 

intuitive judgment, and the marshaling 

of subconscious processes; problems, that 

is, which do not centrally depend upon cmpirical 

data and inductive processes. 

O The process can serve as both preparation 

and' follow-up for face-to-face meetings devoted 

to forecasting matters. 

O The technique can be applied in any forecasting 

situation wherein "expert" or otherwise 

desirable opinion exists, and can be 

reached by mailed or other questionnaires. 

Possible Weaknesses: 

O It is diffScult to design questionnaires for 

Delphic application that meet acceptable 

scientific standards for validity, appropriate 

selection and use of "expert opinion," reliability, 

mad the avoidance of error due to 

maUl or biasd samples of respondents. In 

practc, littie attention seems to have

ei, paid to these standards in conducting 

Delphi.' 3 

n '/he iterative cípnvcrgc,:: : ,i cxput upiimon 

may reflect increasing c :iiiotmity insicad 

of increasing accuracy. 

O "Expert" forecasts may bc more susceptible 

to errors arising from "scnools of thought," 

conflicts, or from unanticipated developmcnis 

in ficlds outside the expertise in the 

group. The mcthod's usual total reliance 

on expert opinion, unleavened by other 

methodological checks, requires spccial 

caution." 

O The strength and accuracy of piedictions 

are heavily dependent upon the initial conditions 

specified and upon thc qualily and 

appropriateness of the information supplied 

to the participants during the iterative process. 

Poor information tends to produce 

polr results, in Delphi as elsewhere. 

Trend Extrapolation 

- 3;- -- 

A wide range of methods fall under this heading. 

Each involves a review of the historical 

data (primarily quantitative) pertaining to 

sone problen: or issue. Using these data, trends 

are "jdermined," assumptions made, intervening 

vuariables are identified where pouaible, and 

the interaction of some combination of these 

facton is projected into the future as a forecasi. 

Regresion or time series analysis are 

among the mit frequently cmployed statislical 

Uend exrapdlation techniques.' m 

Two major variations of this method are 

common. The first projecus data that cxpress 

directly the phenomena being forecasted. Most 

bed-necd formulas used in health service forecasting 

are a variation on this theme. Por example, 

the well-known Hill-Burton formula 

projects bed requirements on the basis of trends 

in population, use-rates, and average daily 

censaus.' 

The socond common varicty is based upon 

data that are thought or known to be correatled 

with thc phenomenon being projected. 

This method ofien is employed when little or 

no dta exist for thc phenomenon one wishes to 

forecast. Use of demographic data as a surrogate 

for ith factors in column one of Table 

1 is a comnmon example. Or-citing the basic 

Forecasting Hiealh Care Services 

hypothetical ample used throughout this 

paper-an inao,, th study of EMS facilities that 

already have adopted computer diagnostic consulttion 

services might show that their usage 

is related in a complex way to some ten diffcrent 

variables. These might include such factors 

as physician work load, degree of medical 

specialization, access to and use of other consultative 

services, cost of the computer service, 

and others. Previously completed studies make 

it possible to predict the growth rate through 

1985 for these ten variables. Using these 

growth rates and the historical correlation between 

these factors and the use of computers 

by EMS personnel, it might be possiblc to predict 

that 28 percent of local EMS pcrsonnel 

will employ computer diagnostic consultation 

in 1985. 

Trend extrapolation methods have a number 

of strengths to commend them to areawidc 

planners. Here are some of the more important 

ones: 

Streng/lhs Claired: 

O Future developments often may be straightforward, 

predictable continuations of the 

present and immediate past, at least within 

"acceptable" error limits (e.g., population 

growth and mix, incidence of disease, level 

and distributions of income). Where this 

is true, trend extrapolations can be very 

effective at minimal cost, particularly over 

the short run. 

O These methods encourage the search for 

trends and continuities and for factors that 

cause, augment, or inhibit such trends. 

O Since these metheds deal with "objective" 

data and with matlhematical techniques, they 

tend to reduce (though, of course, not eliminate) 

the effects of analyst prejudice or 

bias. 

U The average behavior of large numbers of 

people is much more stable and "predictable" 

than that of individuals or small 

groups. The capacity of quantitative (particularly 

statistical) methods for dealing with 

a large volume of data is very helpful on issues 

involving the "trend" behavior of large 

populations. 

O Most statistical methods are based upon 

1¿

Inquiry,/Volume XIV, September 1977 

well-established statistical thceory that specifies 

the confidence one can have in their 

rcsults (c.g., the various methods for appraising 

the dispersion ot data about their 

central tendency). This tlieoretical supersltructure 

is very useful for drawing atten,tion 

to potential sources of imprecision or 

diffusion of focus in particular kinds of 

trend cxtrapolations. 


O Apparent trends in the data have no life 

of their own. Their continuity over time 

may be very tenuous. Though the temptation 

exista, such trends cannot be taken 

for granted. 

D No fomrn meas exists for taking "new" 

trends or !.-tor cmerging in thc future 

into account ila advance. Inded, the ch 

may be inhibited by the perceived "objectivity" 

of extrapolad techniques. 

O Most quantitative techniquaes require data 

that are quite precise and reliable if their 

major strengths are to be capitalized upon. 

Data most frequeantly available to areawide 

plannrs tend 1, be ol uncertain reliability 

and valdity. 

O IU thc uawc~ of the lorocuatig problcm 

canot be expreUed in quantitative terms 

Md in termo flr which the data re available 

on a year-to-year bast, most available 

techniquaes are only partially uble. 

Morpl o Ancyus id Rd e Trem 

1e ov l insat af metioda f~lli under 

thiese two nbica s ato mae a foroaati ng prob. 

hm more man= ubl by bre it into subproblem. 

One seeks to discen the relative 

imporence of each of the subprobkm and to 

improve prodicion accuracy by working with 

more cfcúively dched subproblems Specifc 

attention is paid to enumerating the set 

d *il itgificn t (fcusble) facon (alterativa) 

for explaining some outaome (echieving 

aome obioctive). Thae lctan eo altea=tives 

may be defined b~ally or throqb ime of inductive 

mathematical techniques, such multiple 

regreio or tbc compul#~id Aulmatic 

Interaction Detector (AID, pragram. The 

_ '?4 _ 

ovcrall process is called morphological analysis; 

the graphic display of subproblems showing 

their interrelationships is called a relevance 

tree 7. 

Consider the following hypothetical example: 

An HSA was concerned with the future 

effect on total demand for services that will 

result from a rapid expansion of ambulatory 

care facilities. A large amount of data on both 

the working behavior of physicians and on 

factors contributing to demand for local ambulatory 

facilities was analyzed using the AID 

program. This program worked through the 

data, establishing "factors" through a series of 

dichotomous splits. These factors were chosen 

by the program because the dichotomy splits 

that were selected maximized the betweengroup 

variance. The graphic display of the 

series of these dichotomized factors represented 

a "picture" of all factors that would contribute 

to changes in demand for ambulatory care, expreased 

in terma of relative importance. Tlcse 

factors then were appraised by "experts" .ho 

rendered judgments about the extent to which 

each factor would change over the next few 

year. Taken together, the graphic display and 

expert judgment resulted in the forecast that 

demand for outpatient services would increase 

directly with an expeanion in such facilities, 

Lcading to a relative decline in use of key inpatient 

servicea. 

Whether graphic or quantitative, morphological 

analyis has not yet seen many applicatiom 

in the health care context. Its strengths 

Lrgely remain to be tested, but might include 

the following: 

Sreunguhs Claiumed: 

O Brosd, large-scale forecasting problems are 

much easiar to handie in "pieces." 

O Relevance trees readily show what is 

(thought to be) contingent upon what, thus 

chains of causality, potential arenas for 

conflicts of interes, and areas where organizational 

change are necessary (or are likely) 

can be shown more clearly. 

O Recombiastion of subproblems may show 

a varicty of ways in which changes might 

occur and might thus dispel possible tendencis 

to fall into "single possibility" ruts.

O Emphasis u.on the explicit enumeration of 

many possible alternative factors or solutions 

to each subproblem has significant 

heuristic value, encouraging the development 

of fresh perspectives. 

C3 Feelings for specific policies, organibational 

developments, or social factors that effectively 

enhance or slow progress toward areawide 

health goals, often emerge from relevance 

tree analysis. 

PoJsible Weaknesses: 

- 337 - 

O Where quantitative techniques arc no¡ used, 

successful analysis is dependent upon asking 

the "right" questions (i.e., decomposing the 

problem properly). Few effective guidelines 

cxist, particularly if the problem is ¡ldefiprd 

or poorly understood, as are most 

aceawUke health forecasting problems. 

O Where quantitative techniques are used, 

moast of the posible disadvantages described 

in connection with quantitative trend cxtrapolation 

metho~ds pply. 

O Few a prior quantitalive or qualitative 

guides cxist folr chbosing the factors that are 

relevant to a piven forecasting problem. 

O "A11 psbiltics" for a given problem may 

be extrmely large in number, particularly 

in auiwlde planninig-- many that the 

capacity of HISA paning staffb to evaluate 

them ¡ swamped. 

Compusrr Simulaons 

A conoeptual model s constructed of the major 

cadition and procesas of the situation to 

be preodicted. Thee lements are related to 

one anoother nla loicl and/or mathematica 

rlatioahips that expres, es accurately u posdbie, 

their "real word" interactioa within spci¡fed 

constraint. Employing the model and 

tbe ~cmputer, one can nerate time series (ie., 

dchae in the value o( key varuiables from time 

perod to time period) that expresa the interaction 

of the model' elements. Assumptions, 

variables, processes ad the intecton among 

them, can be variad to simulate various hypoetial 

conditios or the adoption ol proposeg 

pdicia in to "rel" word." 

Consid«r the fcllowlS example: An HSA 

Forecasting Health Care Services 

collaborated with a local university over a 

two-year period to develop a model that expresses 

the interaction of demand for various 

levels of health care, including ambulatory, 

inpatient, emergency and long-term. Confronting 

numerous pressures to restrict new construction 

of inpatient beds, the HSA sought to simulate 

future demand for all services, assuming 

a policy that held the number of inpatient beds 

constant. The results shed light on the yearby-year 

increased ambulatory and long-term 

care capacity that would be required by this 

policy: 

Strengths Claimed: 

O The continuing, simultaneous interaction of 

an indefinitely large number of quantitative 

elements can be considered. This interaction 

can be simulated throughout an indefinitely 

long planning period. The only limitations 

are data availability, cost, and analyst 

imagination. 

O Complex interactions over substantial periods 

of time frequently produce results that 

are counter-intuitive. Given the limited 

ability of the human mind to trace out such 

interactions unaided, simulations may be 

the only fully effective means of projecting 

complex interactive situations. 

O The modl's conditions can be varied at 

will in order to test the effect that varlous 

hypothesized changes in the real world will 

have upon future developments. 

O Unlke neariy all oprations research and 

decison theory models, there is no requiremen. 

to limit analyiss to the achievement od 

a single objective. Progrs toward multiple 

objoctives can be projected unce the model 

need not be predicated upon nodons of 

optimality. 

O Both the construction of the model and its 

actual use are processes with imprasive 

heuristic value. Issues and problemns frequcndy 

are exposed and clarified that would 

not have emerged otherwise. 

O Tbe "tracing out" by the computer of inplicationsnialspacifid 

in the model i¡ ncay mimune fram 

analyst prejdi~e and bias.

InquiryiVolurne XIV. September 1977 


O Thc situalion to be simulated must be wcll 

understood (i.e., ne models elements, thcir 

rclevancc, and thcir interaction must correspond 

well with "reality") in order to 

develop a model whosc results arc not misleading 

or simply erroneous. This possible 

weakness is cspecially salient in areawide 

planning where the range of considerations 

is extrcmely wide and the amount of verified 

knowledge is typically rather narrow. 

O The variables employed in the model must 

bc measurable with considerable validity, 

recliability, and precision. since the mathematical 

manipulation involved in even 

fairly simple models can cpuse the error 

rate lo "ecxplode" in unpredictable ways.'9 

O Modeling a situation in a way that captures 

the essen- o" some dynamic situalion, 

without mislcadmng oversimplification or misrepresentation, 

is much more art than science. 

Few guides for constructing "good" 

models exist, particularly those of complex 

social situations. 

O Time seria data, upon which mist simulation 

modeL are based, must be treated 

carefully. The .nd oi summary indices 

that usually must be employed (e.g., "visits" 

or beda per unit of population) can be mis- 

Ieading. Even where such indiccs do a 

creditable job o capturing the rence of 

the p enon _hem they reprsent, such statistic 

may not readily reflect changes in 

qua~ly s coatralted with chan~ in numerical 

value. n Improvematen (or retogrtenuu 

) in efficiency or efectivaoes are 

difficult to take inato acont, a* well. 

O The deveupmeatl and oting cwt o 

a computer simulation o any real significance 

bi quite high. Por thbi roaen lone, 

simulations may be beyond the reach of 

any but the largest HSs for tbe immediate 

future. 

Docisioq Theory 

- 13 7, - 

This approach i repetativ d many found 

in the field o( operta rtesucb. It is useful 

in making what might be cad caditiona 

teoan a (i.e., predaiOm thwt i tb strategies 

prescribed by the techiques are adopd, the 

decision-maker can expect the most favorable 

future outcome under the conditions assumed). 

The basic method is lo define a series of alternatives 

among which a choice is lo be made, 

and to appraise each alternative in terms of 

its performance during a specified, comparable 

future time period. Thc consequences of adopting 

a given alternative frequently are traced 

out explicitly in the form of a decision tree, 

whose cnd points are expressed in numerical 

value payoff(s). Thc futurc environment may 

be expressed cither in terms of risk (i.e., the 

future assumcs one of several possible known 

states, each with a known probability of occurrence), 

or it can be dcalt with in terms of 

uncertainty (i.e., the future assumes onec of 

several known states, each with an unknown 

probability of occurrence). Once the alternatives 

are defined, and the range of possible 

interactions with the future environment is 

expressed in terms of possible payoffs, decision 

theory offers a range of criteria for 

choosing the best alternative.P' 

Suppose, for example, the HSA is considering 

four proposals lo expand long-term care 

(LTC) capacity in thc area. Each proposal 

differn in bed size. HSA staff studies indicate 

that future demand for LTC is somewhat unpredictable. 

Three posibilities are projected: 

low, medium, and high demand-each with an 

estimated probability of occurrence. A matrix 

is constructed that relates the additional capacity 

of each proposal to the three possible 

future demand levels. The HSA then makes 

ita decision using a variant of the maximum 

expected value criterion: 

Stengths Claimed: 

OGiven its emphasis upon explicitly tracing 

out as many possibilities as feasible, thc 

method encourages one to consider forecasting 

problems from as broad a perspective 

u possible, viewing the future as having 

multiple possibilities. 

3 It forces analysat and decision-maker alike 

to be as explicit as possible about the process 

by which alternatives are defined, examined, 

and evaluated. 

O It facilitatsc use and communication of expert 

opinion by structuring forecasting

*1 

1 21 

1 

«i 

1 

1 

1 1 

- 331, -- 

Forecasting Health Care Services

Inqliry/Voluite XIV. September 1977 

- 340 - 

problems so as to take advantage of specific 

expertise. In particular, it helps distinguish 

and sort out issues of fact, conjecture, preference, 

strategy, probability of outcome, and 

value of outcome. This is very helpful for 

gearing available expertise lo those aspects 

for which it is uniquely suited (e.g., questions 

of fact vs. issues of stralegy vs. malters 

of preference). 

O It aids decision-makers in sorting out their 

own altitudes and preferences with respect 

to situations of risk, uncertainty, and indeterminancy 

(as these terms are defined by 

decision theorists). 

O it helps keep debate within decision-making 

groups focused upon important, objective 

aspects of the problem at hand by providing 

a 'picture" (e.g., the decision tree 

or table) from which to work. 


O The outcome or payoff corresponding to 

a givea alternative oftea is a complex bundle 

of conaquences not readily summarized by 

avaiable quantitative m ures, sch as 

dolian or patients treated. Where such 

eaures are uses, ,he risk of ovenimplificiation 

is substantial. 

O Oucome asociated s with pouibe alternativs 

often are highly uncenain entitea for 

wbhkl aoly rough estimates can be made. 

Tie p t prc~hn of a single-valued 

payoff may be higy mbkding. 

O Tbe dicrete, maunldpnAít alternatives 

Mequired in deciba thbeory dtr. may not 

be approprite, tbive tauaal situatioa. 

This a esrpecily true whbe te trnge of 

poaible ernati ve i actuly continuous 

(as contrasted wth dicrete), whre no 

man flor meaurina the reuts of a given 

alternative eicats, or where the environment 

a highly unctan or rapidly changing. 

O Moai decisiona must be made ina an organizatial 

contLt Such eavirames feature 

inter- and i-ntarltoa barpining 

ova obocv~ and th sbultneous pursuit 

of multiple, ac pme y oals.n 

'Tluse atn tend to pcddo aplbo 

of smne of t major advanta that decioa 

theory has to offer. Pua~duly vulnerable 

ure uch atrenl u t be cocept 

of utility maximization, uniform notions of 

optimality, well-defined criteria for choice, 

and the like. 

O The technique is not concerned with forecasting 

per se as much as it is with guiding 

decision making when the main dimensions 

of the future environment are known, or 

can be assumed with reasonable confidence. 

A Paradig for Applying Forecasting 

Techiqes to Factrs Influendng Future 

Dnemnd for Heiltb Services 

At this point in the discussion, it is necessary 

lo mention a framework within which existing 

forecasting techniques can be applied to advantage. 

Table 3 provides a first approximation 

of this framcwork. 

Each of aix major forecasting techniques 

discused in the preceding section is represented 

in the columnas of TabIe 3. Each of the major 

factors influencing demand is represented in 

the rows. The cell at the intersection of a given 

row and column represents the application of 

the column technique to the, row factor. Table 

3 representa a single technique applied to 

a single factor. In actual practice, of course, 

no such limitation would exist. In many cases, 

best results would be achieved employing two 

(or more) methods jointly. 

A representative sample of cells have been 

selected because of the attention that the topics 

they rprepsent receive in the literature. Blank 

cells indicate. areas where little or no research 

activity has taken place. As Table 3 indicates, 

these unexplored areas predominate over arcas 

where work has been done. Each cell is evaluated 

in summary fashion in terms of two 

criteria. Tiee criteria are intended to express 

the overall usefulness of a given technique, 

as applied to a given factor. The criteria: 

O Validity-to what extent can the technique 

be applied to the "essence" of the forecasting 

problem (as distinguished from its 

peripheral, ifdirectly measurable or readily 

undent~od, aspects)? 

O Precision-to what extent is the output 

of (he technique expressed in precise terms? 

Each of these ia expressed along a single 

dimeanion-low, moderalte, high.

Therc are ihreec basic prcessecs fin gauginig 

the validity and precision of a itchnique for 

forecasting where the technique is bised upon 

formal theory. 

: ' First, nne r:ar vcrify the log 

ical reasoning that led from the thci., to the 

prediction in question. Secoid, une can evaluate 

the "reasonableness" of the assumptions 

that underlie the porUayal of the real world 

situation being forecasted. And third, one can 

verify some of the key predictions, comparing 

thcm against actual events as thcy unfold. 

These processes can be used either singly or 

in combination to appraise the qualaty of the 

output produced by the various forecasting 

methods.' 4 

The resulis shown in Table 3 were nos arrived 

at through rigorous application of these 

proceses. Such an investigation is well beyond 

".. scope of this paper, requiring examination 

of specific applications of each technique. Inste.d, 

Table 3 is intended to convey an overall 

impreuion based upon discuwsions noted carlier. 

With this caveat in mind, several interim 

conclusions emerge: 

First, where possible, two or more forecastling 

methods should be employed in mutually 

rcintf:unr v:ava. In particular, historical analogie 

and Delphi are almost always more cffective 

as exploratory techniques, and/or in 

conjunction with one or more of the remaining 

t~acImqu«. No ásige method should be relied 

upon ecluwively. 

Scod dven the largo amount of unavoidable 

uaocrtainty in forecasting, it i important 

lo dc~lop atregic a*od anaytical techniques 

for estimating and hedging against posible 

error. FPoreca that provide an "optimsiicpessimistic" 

range rather than a point estimate, 

or that include discuison of pouible mitituing 

Iacton, serve this end. 

Third, as Table 3 dcates, much o the 

fwld o demand forecasting remaina largely 

unexpbrod at present. BEpecally deserving of 

ttention appear to be those factors that afect 

the phyidan's impact on demand. 

Fourth, dhc more sophitsicated the forecastlag 

tednique, the better the thuation to be 

forcasted mi be undersood to employ it, 

particularly on an ·reawide basIs. Relatively 

simple mb , ocused on the right ismus, 

can be quite effective. Sophisticaed methods 

Forecasting Health Care Servces 

vidl.,it oaiw oaf sufficient quality and quantity, 

or vit).,.ut the proper issue focus, may be of 

liírile ue. 

SUou. Apparent Dimetnsions of the 

"Zone of Feasible Forecasting" 

An effective way to chart some of the major 

'dimensions of this zone is to compare the 

state of the art against the objectives of areawide 

forecasting cited in the first section of 

this paper. 

To Provide a Picture of the Future Context 

Helplul to All Planners, Public and Private 

The ncw liSAs will be required to produce 

forecasts (perhaps updated annually) of the 

future areawide health care picture. These 

will be useful in carrying out their mandated 

functions and also can be used by all parties, 

public and private. The state of the art is 

perhaps strongest with respect to this objective. 

Fairly long-term forecasts, composed partly 

of quantitative and qualitative information, are 

well within the capability of existing methods. 

These probably can be made with "enough" 

assurance to provide a useful planning context 

for Dal interested parties. 

To Improve Quality and Cost Control 

The picture is less promising with respect to 

the capacity of current methods to meet this 

objective. Available methods can forecast dcmand 

for individual services with reasonable 

accuracy. But the state of the art reflects two 

major deficiencies: First, most methods to 

date have not been geared to forecasting demand 

for rwo or more services simultaneously 

or to forecasting demand for subspecialities 

within a given type of care (computer simulation 

being the major exception). This means 

the danger of ignoring the impact of changes 

in one service (or one subspecialty) upon the 

other is quite real. Second, most methods do 

not focus upon the appropriateness o utilization, 

hence the ráik of forecasting patiers of 

demnad that are inefficient, ineffective, or extrencly 

costly (and making poor docisions 

based upon the results) is quite high. It should

Inquiry/Volume XIV. September 1977 

be noted that failure to gauge "appropriateness 

of utilization" is not an intrinsic failure of forecasting 

methods; skillful and imaginative use 

of existing methods could overcome this deficiency, 

especially if supported by appropriate 

utilization studies. 

To Anticipate and Plan aor New Technologies 

and Methods ol Organir.tion 

Available forccasting methods, based upon current 

demand data, are poorly suited to achieving 

this objective. Demand is not a direct 

function of either technology or methods of 

organization. Hence, the use of current demand 

data (influenced to whatever degree by 

historical technology or methods of organization) 

is a poor way to achieve the objective. 

RIaercs aud Nots 

I US. Conlre, The Comprehensive Healtil Planning 

od Publk Hfm:ls Jer.vkts Amendtnnmu ol 1966, S 

OM. 9h Co.1grm. Nov. 3, 1966. 

2 US. Coneta. TIh Nalonal Healrh Planning and 

Pres.esj Derviopmwn Act ol 19?. S-2994. 93rd 

Can , Ia. 4, 1975. 

3 Se dlcuíuo of ihes objhva in¡ th Ibolowinl: 

M d, M. F.: "alic Comcepla nd Crucial luus 

in Health Plaini Ametrkn Journol of Publk 

HelWt. 59: 1, pp. 1697,. 1969. Colt, A. M.: 

*EkwIum o« Comprehenaive Health Plaing." 

A~rlnr louenul of Publkc Heldh. 60: 7. pp. 1194- 

1204. 1970. Dcky. W. 1. t al.: 'Comnpreheadve 

Halth PlanninaLFedcn a. Sate. Ltoal: Concepts 

Id imitll", W~sn Lw Revkl,. pp. 339-78, 

1970. Fhalb. nto t

- 343 - 

Forecasting Hcalth Care Servlce 

Sciences," 96: 639-998, Sumnaer 1967. 11 reports Zamowiz. Victor: "On Tbc Accuracy And Proper- 

the procedings of a symposium devoted to issues in tiae Of Short-Term Economic Foreca " The TrSAs 

large-scale social forecasting. 

Of Economies: Tire Forty-Filth Annual Report OJ 

9 For a discussion of how this concept has bceen de- Thie'Natioal Bureau Of Economic Restarch, 1965; 

veloped and used in a variety of forecasting con- also Hersch. Charles. The Discontent Explosion in 

text. see: Bell, D.: '"Twelve Modea of Prediction- Mental Health," American Psychologist, Vold. 23. 

A Preliminary Sonring. Daedalus: The Proceedings 1968. pp. 497-506. While exclusive use of ecpern 

of Tlhe Anmerican Academy of Arts and Sciences. 9j: opinion applied to situations in the presenr bas becn 

845-73. Summer 1964. 

largely discredited on acientific grounds. it could be 

10 See the entire issue of: Tite Annals of The Amrri- argued that future situation (iLe.. foecasting) are a 

can Academy of Poliical and Social Sclences. 399: different matter. This may be true, but the burden 

1-160. January 1973. devoted to the geeral topic ot proof cmns to lic with those who advocate this 

of: "ihe Nationsi Health: Some luues." edited by view. al present. 

Sylvester E. Berki and Alan W. Weston. I5 

For a oad discuasion of the major quantitaivd ex- 

II This brief description, of course, severely underatates trapolation methode in mn econometric context. see 

the tcope of a major methodological ficld-historiog- Fox, Karl A.. Intermediate Economic Statlstics, 

raphy. While not within the acope of this paper, a (New York: Wiley & Sons) 1968, especially Chap- 

discussion of the methodological problema in reconten 7, 8., 9 and 11. 

aructingl recent h;story ia worth reviewing A con- 16 

Report of The Committee of Consultanta, Evaluation 

venient survey of philodophies of history. which 01 Trhe Hill. Burton Bed Need Formula: Short-Term 

clearly indicates thc importance of the "t'eory of General Hospital reds, Health Care Facilitics Ser- 

history" employed with resard to selectivity of evi. vice Evaluation Project 71-4. Augua 1972. 

dence and the variability with which it is weighted in 17 

For an example of morphological analysis applied in 

evalualing · liven historical situtlion, can be found a noahealth contexi, ce Shillif. 1. E.. and Smith, 

in Patrick eGailiner (ed). Theorlis ol Hlloory. R. D.: "A Forecasting Method for Setting Short- 

(N.Y.: Free Prer), 1959, and in Dray, William H.. Rango Rescarch Objectives." Researchi Management. 

PhUloophy ol llitory., (Englcwood Cliffs. NJ.). 12: 24-34. March 1972. For a description of the 

1964. Chapter J. Sc aio Mazlihb. Bruce (cd.): AID program, see Anderon, R., et al: Automatic 

Tbe Rallrod and The Space Program-An4 Erplora- Inter·ction Detector Program For Analyzing 

blna In Nlnorkial Analogy (Cmbridge, MA: MIT Heath Survey Data." Healsit Services Research. 6: 

Pra)., 1965. 

165-U3, Sumnmer 1971. For a comparison amunong 

2 Por om of the best d the carly treateinta of this multi.variate stalcul tchniqua (including AID) 

topie, c: Helmer. Olaf, Soelal Tclsnololy. (N.Y.: from a methodoloical viewpoint, see Reinke. Wil- 

Bisc BDoo&). 1966. Aleo me: Vickern, Sir Oeot* liam A.: "Analysi of Multiple Sources of Varia- 

try. The .4r of Ietden.c. .t: A Study of Po!tcy Ma.tin: Comparion of Three Techniques." Holih 

ing. (L~ada: C. 1 nml ilUI), M& 1965. For a more Services Reearch. 9: 311-21, Winter 1973. For a 

m~m evaluatib o Delphi in the conalt of the technical discuasion of relevance trees see Fiacher, 

idM o wsubjective scallng, PUil, Jura: 'The Del- Manfred, Technologlcal Forecasting and Social 

pi Mtbod: Subiance. Contex., A Critiquc uad an Chanle 1: 4 (1970), 381-89. 

AaeM ?d_ Biogrpby." Soclo-conome Plinnlng 18 For a primer o the basic methods of computer 

S~luese. Vd. J. No. 1, Pp. 57-71. 1971. Pcr~ap simulation, ee Fomrrter, 1. W.: Principles of Sys- 

6 mnu ctical review of Delphi i cona te ha ¡ tem., Cambridaf, MA: MIT Pres 1967. For an 

Harold Son rmas' Delphl Cerlfe: Erpert Opinlon. application of simuladon in a ficid related to area 

FPoceuWl,. odW Grop Procet,. (LetngLaon MA: health plunnin. s Forreter. 1. W.: Urban Dy- 

Lamao. Boale ) 1975. An aU~ of Delphi namins (Caunbrldle, MA: MIT Pres) 1969. F:or 

in the iong-term care context my be found in: spcific auawide hbalt plannin application me 

Clark. L H.. aod Coahran. W.: "Needr o Oider YVt D. e, Dabek. L. and Intrilliaptor. M. D.: 

America Amnemd by Delphi Tcbnique" lJoural 'A Micro-Eccoa t[ric Model for Rcllal Health 

of Gerou~ogy 27(2): 275-7, Apri 1972. 

Plh~'" Buniu d Economi Bulletin, 24(1): 

13 Tbhi pouani w nca~ Delpbi b¡ parap mat 1-21, PFall 1971. 

frequettly overlookld in practic. A ri~ro as- 19 Thls probleam discuted effectively by Alono., W.: 

resmnat of Delphi agpun sadad or quetilea "Prdiica Bi With Imperfect Dau." iournal 01 

mire design. au embodied in the Stdard# Por Edu. The Americ~n Intliute 0f Planners. 34: 248--5, 

catbnal and Pjychlogkal Tesr uld Mnu U~ July 1968. 

pubiad by tb Amriarn Psco Arocs ti 20 Bauer, L A. (cd.). SocIal Indictors. (Cambridg, 

(1966), b conductd by Sackman in Chapter 3 of MA: MIT Prma) 1966. 

hia Delphi Crilwue: E&perl Opinlio. Feoreclng 21 Por a coin arv~y of the mior coaceptual sa 

And Group Procnes, (Lxinton. MA: Lcxhtalo in tbe idd o formal decliion theory. see: Ward 

Books) 1975. pp. 11-27. 

Eduarda and Amos Tvenky (ed.), Decl¡on MakitnS: 

14 E.uasive uwe d capef opinion to omaue uament Seltcln~ Readiint. (Middlexs. E id: P ua 

and to validate ites has beca valued in a number BooL), 1967. or Raifa, H.: DecLon Analydls: io- 

of casteMz. mli~ hbv tendtd to bow tiau, rod~ory L. csrer O Choies Under Uncerrulny 

even in ituatioa wberm ti problm aen farlry el (Rdend MA: Addlon Wetley) 1968. An inter- 

structured and thc informafion av-llaMc a sncas e~Lng appUaa~ of deciiu thcory in re oal 

ably complt,. aepen opiaion ia not speally d. hab Iplal appears Grinmes. d a.: Urn od 

lective or reliable as compered with plauidble ater. D~c Toy~ in Reioral PhnaLn" HNdalih Set. 

ntive mc~odr Por dudie in vrand coMtai nec vrtes RMaku,. 7: 73-78, 1974. Of particular Ibar.

Inquiry/Volumne XIV. September 1977 

- 344 - 

cst lo health planners in the project review funclion solved. For example, see Helmer, O. and Rescher, 

is Reutlinter..Schlomo: Techniques For Projecí Ap- N.: "On the Epistemology of the Inexact Sciences" 

praisl Under Uncerairnty. World Bank Staff Occa- Managenernr Sciencr, 6: 1. ¡959 pp. 25-52. A more 

sional Papera Number Ten (~London Johns Hopkins general vicew of the major issues is contained in 

Press). 1970. For a discussion ol the context into Frank. P.: Plillosophy of Science (Englewood 

which decision theory lits. ecc Chapter X of Kahn. Cliffs: Prentice-Hall), 1957, especially Chapters II 

H. and Wiener. A. J1.: The Year 2000: A Frmre. and 12. and in Ikle, F. C.: "Can Social Predictions 

work lor Speculation. (New York: Macmillan Co.), Be Evaluated?" Daedalus: Journal ol tie American 

1967. 

Academy of Aris and Sciences. 96(3): 733-58. 

22 For one of the most succinct discusiona of this or* Summer 1967. 

pnizationl tendency. ee Cyert. R. N., and March, 

J. G.: 

24 For 

A 

a rendering 

Beharioral 

of 

Tleor)o 

"the basic philosophical 

Of The Firm, 

problem 

(Englewood 

Cliffs: Prentice Hall), 1963, Chapter 

facing 

3. Also 

the forecaster'" rom the perspective of each 

valuable b: Thompson, J. D.. and McEwen. W. J.: 

of five philosophically plausible points of view, ee 

Orpnizational Goals Mnd Environment: Goal Set- Mitrofl. W. 1.. and Turoff. M.: *Technological 

ting As An Interactive Procas." Amerlcan Sociologl- Forecasting and Asaum ent-Science and/or Myral 

Rerlew. 23: 23-31. 1958. 

thology" Teclhnologkal Forecasting and Social 

23 The epistemology of this matter a by no meana re- Cliange. Vol. 5, 1973.

#20 

Jaime 1B. Henrs 

Rtxlnc. I.. Ro:nl'IJI 

- 345 - 

l:o man1, I¢are th..%pitll% thlo4tughttltl the 

Unictd Si;11e¢ have ueJ fund fíini' m gíni1N1. 

dtJnaltion. and othcr free monc) %i.ti,'c% I' 

underrtrie capilail cx:penditur¢ % foi equipmcnt. 

Unforiunal¢ely. ihcec fi¢l fundl. h.a.. 

nol kepl pac*e a lh ith r.spad incrie..ec in dem;and 

[íor caipital epe:ndilure- As aun fllernalite 

t¢o ihe osuright purchaing of equipmncnl. 

.omeC ho,.pit;al. have itír.ted tio le .ing. Although 

no c.cil . ¡r.ti. .ue d.s ilahlc. ile i cs'. 

tinilllc'd tha;l o .w':" i:l;8llil vo rlh of1' ;1s.%' 

.ar Ied.cJ lo h osp'.it. in the ':nit1el Si;itel 

e.uch >ca¿Ir. 

In .in a.ttempl tos dJclcrmninc u hoilhtI' 

cIiI>% 1. Io Icel¢e. ilmdlinl lilttl , ;ld cL)nil oller', 

of .2 non-pirofit hi.pit.il. in Stluth ('arfinm.a 

v.e.: inter.icus. Sd. icv .fal re'.otsn% v crc o'fkied 

for ihcm ¢hooi-nig le¢a.ing i% ;i Ifina nc.ing 

.:hilí. Six of the rca;,>. nl¡hel mlt 'luqiCtl) 

%uggtCd . tre li.ceJ hd helow: 

I I h1 .a...' r 1 , e:wur d Io ha. '. v high de...e of1 

teChnuoloíigi c, l íh%,b leccnLe. lhco el.'t t the 

dildol I ti, i .e in 'ql iplni:ll Is dc.'llcd 

.2 .S'f hl'. 1% K,'cl oi n I .'.led cquipmin l'nt

I1':,1":" ,o1l.,' A`V. %h' h i, 19I7N 

I.m.ini ¡%. .'i iiI i¡nie'iJed I, 1 'pl.ill. i. li- 

.:1i ¡h. lI;imge of Ih, ,le 1C¢. 2 

True I.eaw %'erwus ('ondiional Sale 

Nllhi :'h · s.m.,l yI)pc. of aí;r-atngtl.nl· .,ic 

liriequicnll) I lr"cd lo u, Ics;c f in.aning h! 

l'e,,o .·;ind Ie'nc¢·J . h,,pil;al% %hutuld he a;e..eic 

,,f P' JiffJecni'c., in flinancing lai*raingcmcnl,. 

%·iie' senl ,%¢ e .¡eno oinlr;e¢t, are aclu;all ce.ndilitn.eil 

, Wilh : ,;lle. Iru ;tlea%e. rental pa; nl'rl , 

a;re expeii'\ d fiur r¢cinhíimu-cllinl duiing th. 

t¢,ei mi1 shich Ihct) uoccur; on ihe 0lh¢e h;nd. 

msilh .i Ic.·%e.purc.ha%·c (ctinJititin;il .ile fin..c,lJ 

h%.. Ic.íving i

- 347 - 

I v,'N ti.K 1tHm.pi.l !, t1t4tlM,¡1w,; 

'3.4 

II 

31 Z 

3 4 

SlIMN.43iOll & .'%e' 

32 6 

I(i I'A. 

' .|I H) nlntlh ,,nll.,tk, 

k.%timnted %lm; «ar valucr tPercrniage id t I4I llh4. .J JOi.iI ! 1 . '1:. ., 

,'4 "i44' 4t, :

. . - , . . 

hlIs0, . '%'o,.t. ). .I k 197" .t , t vt AE 

- 348 - - / 

Talr 2. Beore-tax kss rilts for hospilsl equipmeit', August 197.\-.'ebruar% 1975. Primr e.,: 

eh -_ _3 _ __ 

_Eqwl~ ta 

11.3' 11.9 

67.1 

I9l 

17: 

21.. 

brumaetd rIhtae ,alue IPerceiuLare d roh 

4 1~ lase 25"&r % 

14 13; 

14 7 

3 .2 

61«l 

32 y 

19. 1 

191. 

25 4 

l i

Ra.fo.renem.. aid %.hs '. 

- 31!9 - 

1 ¡l¡

Mcthodologic Articles 

- 350 - 

APPLICATION OF COST-BENEFIT ANALYSIS TO THE 

HEALTH SERVICES AND THE SPECIAL CASE OF 

TECHNOLOGIC INNOVATION 

Herbert E. Klannan 

As an economic technique for evaluating apecific projecta or programa it the public 

ctor, cost-benefit analysis is elatively new. In this paper, the theory and pra~ico of 

costbenefit aalysis in general ue discussed as a bass for conssdcring ita role Jn ua~e . . 

~tedoigy in the health §eicea A review of the literature on applications of 

costbenefit or cost-effectiveness analysis to the health field reveals that few complete 

stud~s have bcen conducted to date. It is suggestcd that an adequate analysis rtquiraes m 

empirical approach in which costs and benefits ae juxtaposed, and in which presuaed 

benefit reflect an ascertined relationship between inputs and outputs. A threefold 

claufition of benefita is commonly employed: direct, indiect, and intangible. Since 

the latter pose difficulty, cost-effectiveness analysis is often the more practicab 

procedur. After summanzing some problems in predicting how technologic developments 

me likely to affect costa and benefits, the method of cost-benefit analysu is 

applied to developments of hcalth systems technology in two sttings-the hospital aud 

automasted multiphasic screening These examples underscore the importance of solving 

problems of measurement and valuation of a project or program in its concrete setting. 

Finally, bain s to the performance of sound and systematic analysi are Usted, and the 

political context of decision making in the public sector is emphasized. 

The purpose of this paper is to discuss the application of cost-benefit analysis to the 

assessment of technology in the health services. With the few exceptions that are noted, 

the focus of this paper is on services, not research. 

In carrying out this task, there is no substantial body of empirical research literature to 

draw upon, analyze, and synthesize. Accordingly, the task will be approached in three 

distinct steps. First, the theory and practice Of cost-benefit analysis in general will be 

reviewed. Second, applications to the health field will be discussed. Third, the 

potentialities and limitations of cost-benefit analysis for the assessment of health systems 

technology will be suggested, using concrete illfstrations. 

THEORY AND PRACTICE OF COST-BENEFIT ANALYSIS 

As a formal and systematic approach to choosing among investments in public 

projects, cost-benefit analysis is only a generation old. It derives from the marriage of 

theoretical advances in the new welfare economics and the previously undernourished 

public expenditures branch of public finance (1). In reviews of the cost-benefit literature, 

few references are encountered that antedate 1958 (2-4). Most of the theoretical au well 

This auie ia revisin of a paper presented at the Confer~nce on T~cinology and Hhadth COa 

Systems in the 1980s it San Pranco in January 1972, aponsored by the Health Care System Study 

Sction, Natial Center for lt Svis Rerh anud Development. 

Inmtrntional Journl of Health Sevic~, Volunw 4, Number 2. 1974 

zi

; / larman 

u empirical research has been carned out in connection with water resources 

projects (5-10). 

Aúm and Critea of Choice 

- '@jl 

Cost-benefit analysis aima to do in the public sector what the better known 

supply-demand analysis does in the competitive, private sector of the economy. When 

market failure occurs-whethor through the absence of a market or through the existing 

market's behaving in undesirable ways-public intervention comes under consideration 

(11). Cost-benefit analysis is helpful in determining the nature and scope of such 

intervention. 

The most egregious example of lack of a market is given by the case of the pure public 

good. Such a good is collective, usually entails governmental action, and is characterized 

by a particular feature: when more of it is consumed by A, B need not consume less (12). 

National defense is one example frequently encountered in the literature, and the 

lighthouse on the shore is another. Certain aspects of basic research and the dissemination 

of research findings share this feature, since the acquisition of new knowledge by D does 

not diminish its value for C, the original investigator who developed it. 

In the context of cost-benefit analysis the most important cause of market failure is 

the presence of substantial extemal effects. Such effects are called economies if positive, 

and diseconomies if negative. Vaccination against a communicable disease is perhaps the 

most commonly cited example of benefits accruing to a third party or to the community, 

in addition to the benefits received by the patient and health worker, who are directly 

involved in the transaction'(13, p. 18). Still another example from the health field is the 

protection accorded to the community by hospitalizing persons with severe mental 

illness(14, p. 12). 

The goal of public policy is to adopt those projects or programs of service that yield 

the greatest surplus of benefits over costs. Evaluation of projects is prospective, oriented 

toward the future. The criterion of choice, analogous to that of maximizing profits in the 

market economy, is to maximize present value. Stated differently, but meaning the same, 

the criterion is to equalize marginal benefit and marginal cost. Strictly speaking, as 

Stigler (15) notes, maximizing present value is also the criterion for optimum behavior in 

the private sector. As Fuchs (16) pointed out, this criterion is quite different from that of 

attaining the maximum amount of a particular indicator of benefit. 

Of course, the notion of balancing benefits and costs is by no means alien to medicine. 

Lasagna (17) states, "Since no drug-free of toxicity-has ever been introduced that is 

effective for anything, those of us who are pharmacologists have learned to live 

reasonably comfortably with the notion of paying some sort of toxicological price for 

welfare." 

A possible source of misunderstanding about cost-benefit analysis is that benefits are 

usually costs presently borne that would be averted if the program in question proved to 

be effective. It is essential to distinguish these present and potentially avertible costs 

from the resource costs required to conduct that program. Since the two types of cost are 

not always juxtaposed, this distinction is not obvious, and the failure to draw it is not 

evident. 

There are two essential characteristics of cost-benefit analysis: breadth of scope, and 

length of time horizon. The objective is to include al costs and all benefits of a program,

- 352 - 

Application of Cost-Benefit Analysis to Health Servic~s 1 

no matter "to whonmsoever they accrue," over as long a period u is pertinent and 

practicable (3). 

Countir, Meaa~, g, and Valuhg Benefits and Cost 

When an agency wishes to undertake a project or program, it may be tempted to go far 

afield in counting benefits and to neglect some costs. For example, in water resources 

projects, certain secondary benefits may be included improperly. In vocational 

rehabilitation programs both savings in public assistance grants and income taxes on 

subsequent earnings are sometimes counted as benefits, even though neither item entails a 

saving in the use of resources. Both grants and taxes are transfer payments, that ai, they 

represent a transfer in command over resources (18, p. 47). 

The distinction between costs and transfers is not meant to suggest that the sole 

justification of public projects or programs is an increase in output or gross national 

product (GNP). On the contrary, there is increasing recognition that public projects or 

programs may carry multiple objectives, including income distribution, more jobs, and 

regional growth (19-22). There is no reason why, for purposes of cost-benefit analysis, 

earnings on a job cannot be assigned greater weight than the same amount of money 

received in public assistance grants (18, p. 167). What must be recognized is that such 

weights are judgmental, are likely to be arbitrary (at least initially), should be derived in 

the public arena, and, above all, must be clearly stated. 

Similarly, as shown by the progressive individual income tax, we seem to act on the 

belief in the United States that an extra dollar accruing to a low income person is worth 

more than an extra dollar accruing to a high income person (19). Again, assigning relative 

weights may help to improve analysis for public policy. There is no reason to believe, 

however, and no intention to claim, that agreement on such weights is imminent. 

On the cost side a good example of the tendency toward understatement is the neglect 

of compliance costs imposed on individuals and firms in calculating such costs as those of 

administering the individual income tax, and of Medicare for the aged. Once the 

installation of seat belts in automobiles is made mandatory, the temptation arises to 

disregard the cost of seat belts to car owners (23, 24). 

Counting benefits and costs involves deciding what to include and what to exclude. 

When an item that may be properly included can be measured, the next problem is that 

of valuation. The ease of valuation, indeed its possibility, depends largely on whether the 

item in question is traded in the market and bears a price. In that case there are many 

good reasons for simply adopting that price (25). When market price is deemed to be a 

defective measure of value, however, an attempt is made to estimate an imputed or 

shadow price (26). One modification of market price that is widely accepted is to set a 

lower value on unemployed resources; the size of this adjustment may vary not only with 

the state of the economy at large, but also by geographic region and by occupation (20). 

When an item, even an important one, lacks a market price, the tendency is to omit it 

from calculations. If total benefits are thereby understated, a program may be 

erroneously deleted. More important, perhaps, programs with a sizeable proportion of 

unvalued to total benefits stand to lose in competition for funda with progrnam thai have 

few, if any, unvalued benefits (27). Among the items most llcely to be omitted are 

so-called intangible benefits; such benefits are especially prominent in the health field. It 

is not that they can never be valued (28). Rather, one may distinguish between intangible

I Klarman 

- >3 - 

benefits that for the time being remain difficult to value and pure public goods, which are 

not traded on the market and therefore cannot be valued. 

Tbhe dilemnmas of valuation can be escaped by retreating from cost-benefit analysis to 

cost-effectiveness analysis. The latter i the less demanding approach, since it does not 

require the valuation of all benefits in terms of a common numéraire. Cost-effectiveneu 

analysis requires only that benefits be measured in physical terms. Once an objective or 

output is specified, the aim is to minimiz e the cost of attaining it. The cost data required 

for cost-effectiveness analysis are, however, the same as for cost-benefit analysis (29). 

Retreating from the valuation of benefits to measurement alone entails a substantial 

loss: analysis can no longer assist in setting priorities among several fields of public 

activity. The reason is simple. While cost-benefit analysis cuts across diverse objects of 

public expenditure, cost-effectiveness analysis can only help in choosing among alternate 

means of achieving a given, presumably desired, outcome (30). It is cost-effectiveness 

analysis that was incorporated as a major element in the planning, program, and 

budgeting (PPB) systems of the federal government. After its initial development by the 

Rand Corporation, PPB was introduced by the Department of Defense in 1961 and 

extended to other departments and agencies by Executive Order in 1965 (31, 32). 

Both cost-benefit analysis and cost-effectiveness analysis imply the measurement of 

outcomes that are associated with particular projects or programs of service. Presumably 

there is a link between inputs and outputs that is measurable and known. It does not 

matter whether bc.,avior follows a deterministic or probabilistic pattern. In the 

development of water resources the design of a particular project almost guarantees the 

emergence of certain physical outcomes-so much land will be lost to flooding, so much 

more land will be protected from flooding, so much land will be irrigated, etc. In national 

defense the outcome of a proposed course of action is much more uncertain, since other 

countries can take evasive and retaliatory action. In the health field, as will be shown, the 

presumed link between inputs and outputs is sometimes tenuous (33). Too often, the task 

of measurement, which necessarily precedes valuation, has been neglected. 

The Rate of Discount 

There is wide consensus in economics that a dollar is worth more today than it will be 

worth a year or two from now; this holds true when overall price level remains constant. 

As long as assets are safe, consumers are believed to have a positive time preference, that 

is, prefer to consume now rather than later. For producers investment may be productive 

through either the lapse of time, as in wine making, or the adoption of more round-about 

methods of production. Borrowers are therefore willing to pay interest for the use of 

capital, and lenders, in a capitalist economy, expect to receive interest. In a socialist 

economy an accounting or imputed rate of interest is employed to help allocate resources 

over time. The interest rate that calculates the present values of future streams of benefits 

and future streams of costs for public projects or programs is the well-known discount 

rate of cost-benefit analysis (18, p. 165). 

Although economists agree that a discount rate is necessary for rendering 

commensurate benefits accruing and costs incurred at different times, they do not agree 

on the size of the discount rate. Marked differences of opinion prevail for a number of 

reasons. One is that a diversity of interest rate structures exists in the real world, owing to 

capital market imperfections, differences in risk, and governmental monetary

- 354 - 

Application of Cost-Benefit Analysis to Health Services / 

pd~cies (2, 3, 34). There is controversy regarding which imperfectiona to allow for and 

how to allow for themn Another reason for differences of opinion, as Musgrave (35) 

o~ua cles, is the saitce of financing-private consumption or investment. Still another 

reason for differences of opinion is a value judgment: whether the proper measure of tho 

dLcount rate for public projects is the opportunity cost of capital in the private sector, 

or, to the contrary, whether it is the social rate of time preference. The private rate may 

be high, well above 10 per cent, particulady when allowance is made for the corporation 

income tax of 50 per cent (36). The social rate of time preference is usually much lower, 

based on a longer time horizon or greater readiness by the community acting together 

than by individuals to postpone gratification in favor of future generations. The social 

rate, which has been justified in terms of future risk and uncertainty, probability of 

personal survival, and the diminishing marginal utility of additional income or 

consumption as per capita income grows over time, is not a number that we know how to 

ascertain empirically (34). Accordingly, still another procedure, which combines private 

opportunity cost and social time preference, is also not measurable. 

In practice, the agencies of the federal government have employed a wide range of 

discount rates, usually without giving a reason (37). Nevertheless, the consequences of 

choosing a high or a low discount rate are clear. A low discount rate favors projects or 

programs with benefits accruing in the distant future. In effect, as Boulding (38) has 

suggested, a high interest rate favors the aged and a low one favors the middle aged. When 

a project or program is short-lived, with both benefits and costs concentrated in the near 

future, the choice of discount rate is of minor or no consequence; indeed, for a 

short-lived program discounting may be dispensed with. Some economists are averse to 

selecting a particular discount rate, on the ground that they are in no position to choose 

between generations (13, p. 57). The tendency is to display calculations of the present 

values of benefits and costs under two or more discount rates. It seems to me that such 

alternative calculations do not afford helpful guidance to the policy maker, unless he is 

advised when to employ one or the other. 

Even in the present state of the controversy there may be some merit to employing a 

single number for all public projects or for all public human investment projects. The 

combined method, recommended by a panel of consultants to the Bureau of the Budget 

in 1961 (39), and subsequently developed by Feldstein (40), can furnish an adequate 

rationale even though it does not yet yield a specific number. Such a number admittedly 

would be arbitrary, a reflection of value judgment (2). Henderson (34) reports that the 

French have adopted a centrally determined rate of discount of 7 per cent, to be applied 

to all public enterprises. This rate is higher than that encountered in many Ame~ican 

cost-benefit studies. 

APPLICATIONS OF COST-BENEFIT ANALYSIS TO THE HEALTH FIELD 

The health services literature contains many affirmations of the importance of 

cost-benefit analysis for improving the allocation of resources to and within the health 

field. It may prove to be a source of astonishment that relatively few complete 

cost-benefit studies of health programs have been carried out. Perhaps it is appropriate 

that fewer cost-benefit studies have been performed than advocated. Where the am is to 

minimize the cost of producing a given good or service, or even of constructing a hospital 

of specified size and with suitable appurtenances, the apparatus of cost-benefit analysis is 

superfluous (33). It then suffices to compare unit costs.

I Klaman 

Ctitena for Inchmion 

- 355 - 

The major reason for the shortness of the list of complete cost-beneflt studies is that 

most studies conducted to date are limited in one or more respects. First, in the 1970s. 

there seems to be little point to considering nonempirical analyses. Thus, today Mushkin's 

seminal work (41) in conceptualizing the application of cost-benefit analysis to the health 

feld must be excluded from consideration. 

A second, perhaps more critical, requirement for including a study is that both the 

benefits and costs of specified programs be measured and valued simultaneously, with 

their respective present values juxtaposed and compared. By this criterion, the majority 

of empirical studies so far performed in the health field are excluded, including that by 

Fein (42) on mental illness, by Rice (43, 44) on a number of diagnostic categories, and 

my own on syphilis (45) and on heart disease (46). While all of these studies attempt to 

measure and value the cost of a disease, thereby, in effect, measuring and valuing the total 

benefits of eradicating that disease, none attempts to estimate the costs of conducting 

programs with specified contents and aims. Although each study has made a contribution 

to the counting, measurement, and valuation of direct and indirect tangible benefits, and 

two have explored the valuation of intangible benefits, none has presented a comparison 

of costs and benefits under a specified set of conditions. 

The above two requirements-quantification and juxtaposition of costs and 

benefits-impress me as being incontestable. A third requirement can be defended as 

equally necessary: that benerits and costs reflect a known link, alluded to above, between 

program and outcome, i.e. between inputs and outputs. Such a link should be empirically 

based. Today, speculative or hypothetical relationships do not suffice (47). To apply 

economic valuation to hypothetical relationships between programs and outcomes is to 

indulge in an academic exercise, since the results of such valuation cannot transcend the 

quality of the underlying measurements. Such an exercise is not only idie, in that it can 

make no contribution to policy formulation, but it may be counterproductive if it 

obscures the fact that the relationships between inputs and outputs are not yet known 

and remain to be ascertained (14, p. 29). 

In an article discussing the contribution of health services to the U.S. economy, 

Fuchs (48) has demonstrated the importance of information concerning the efficacy of 

health services. The economist can indicate the types of data he requires, but he is seldom 

in a position to procure them by himself; he must rely on other investigators in health 

services research to help him to obtain them. 

This third requirement implies an important corollary. The size of a problem, as 

measured by the total costs of a disease, is not a reliable guide for policy (49, 50). Even in 

communicable diseases, less than eradication may be an acceptable goal. For most 

diagnostic conditions it is essential to know the extent to which a given program is likely 

to reduce the size of the problem. This point is often overlooked. It lends itself to 

oversight particularly when benefits and costs are not juxtaposed. In the early 

cost-benefit studies in the health field there may have been a further tendency for 

economists to attribute greater efficacy to medical care than was perhaps warranted (51). 

Weisbrod (13) performed the earliest of a small number of such studies and his remains 

one of the most systematic. He compared the benefits and costs of intervening in three 

diseases-cancer, polio, and tuberculosis. Drawing in a creative way on Bowen's work in 

deriving the demand curve for a public good (52), Weisbrod was frequently reduced to

- 356 - 

Application of Cost-Benefit Analysis to Health Serices / 

obtainng cost data and some notion of the link between inputs and outcomes from 

pe~onal communications with clinicians and administrators. His threefold classification 

of benefits-direct, indirect, and intangible-followed Mushkin and has beconm the 

convention. 

How such benefits are measured and valued, as well as an assessment of the current 

state of the arts, will be given below. Both accomplishments to date, and possible 

shortcomings in the accepted procedures will be presented. 

Diúect Benefits 

Direct benefits are that portion of averted costs currently borne which are associated 

with spending for health services. They represent potential tangible savings in the use of 

health resources. Certainly in the long run manpower not required to diagnose and treat 

disease and injury does become available for other uses It is reasonable to suppose that 

our economy, like others, has a vast variety of wants in the face of a totality of relatively 

scarce resources, so that freeing resources for other, desired, objectives represents a 

contribution to economic welfare. 

In the absence of a specific program of services to be evaluated, the measure of direct 

benefits is usually taken to be total resource costs currently incurred. The appropriateness 

of this measure as a basis for policy is questionable, as indicated above. Nor is it helpful 

to take some fraction of the total. In terms of resource use, diminishing marginal 

productivity is likely to set in as a program expands beyond a certain point. In terms of 

valuation of benefits, diminishing marginal utility is often a plausible assumption. 

While it is usually taken for granted that direct benefits, or the current costs of care 

that will be averted, can be measured with precision, this is true only when a firm 

produces a single good or service, such as maternity care in a special hospital. In most 

instances several goods or services are produced jointly. Under conditions of joint 

production it is possible to calculate the extra or marginal cost for each product. but not 

its average unit cost (10, pp. 44-45). When average unit cost figures are presented, they 

reflect an allocation of overhead and joint costs; and such allocation is necessarily an 

arbitrary accounting procedure, even where it is systematic and replicable. An alternative 

procedure, which is no less arbitrary, is to assign to a diagnostic category its proportion of 

total costs, with the proportion taken from the percentage distribution of patients or 

services. In the absence of facilities that produce only a single product, it might be helpful 

to analyze cost data for facilities with varying diagnostic compositions of patient load. 

However, other factors are also at play, and there is no logical solution to the problem of 

determining average cost under conditions of joint production of multiple outputs (18, p. 

166). 

Another complication, which affects the calculation of direct benefits and also of 

indirect benefits, is the simultaneous presence of two or more diseases in a patient. The 

presence of disease B when intervention is attempted in disease A serves to raise or lower 

the costs of intervention and therefore the corresponding benefits (45). The reason that 

indirect benefits, which represent gains in future earnings, are also affected is that the 

presence of diseases A and B in a patient may reduce the probability of successful 

outcome from the treatment of either. The effect is to overstate the benefits expected 

from reducing the incidence of one or the other disease (51). The magnitude of this effect 

is not known.

IClarman 

- 37 - 

The prevailing tendency is to take direct benefits from a single-year estimat of 

costs (44). Since surviours will also experience morbidity in the futuro, some medical caro 

costa are being neglected. Initially this procedure may have been associated with ma 

emphasis on single-year estimates, to the exclusion of present value estimates (50). Once 

the necessity of present value estimates is recognized, other explanations must be sought 

for this shortcut. A possible explanation is that survivors will experience only average 

morbidity in the future; when extra morbidity is absent, there is perhaps no need to deal 

with morbidity. A more plausible explanation lies in the lack of longitudinal data on the 

morbidity experience of defined population cohorts. 

The fact is that a single-year estimate reflecta the prevalence of a disease, not its 

incidence. It may be that the prevalence figure is sufficiently greater than the incidence 

figure for chronic conditions, so that it makes ample allowance for future events. Indeed, 

the prevalence figure in the base year is the same as the sum of the incidence figures for 

all survivors to this year, if certain factors remain constant, such as the size of population, 

death rates for the particular diagnostic group, and the incidence rate. When any of these 

factors follows a rising trend, however, the prevalence figure exceeds the cumulative sum 

of the past and present incidence figures and falls short of the sum of incidence figures 

expected in the future. 

To the extent that unit costs or prices tend to increase faster in the health services 

sector than in the economy at large, the value of direct benefits will also increase. In my 

own work 1 have incorporated an adjustment for this factor into the discount rate, 

deriving thereby a net discount rate (45, 46). If econoinic growth were to slow down in 

this country, the lag in productivity gains of the health services sector behind the 

economy at large would be reduced, and so would the size of this adjustment. 

Transportation expenses for medical care are a resource cost which is disregarded in 

cost-benefit analysis, although they are allowed as deductions under the individual 

income tax. When the physician made home calls, his travel expenses were automatically 

included in health service expenditures. The foremost reason for neglecting them today is, 

most likely, lack of reliable data. There may be the further, implicit assumption that 

patients' transportation costs are of a small order of magnitude. 

Indiect Benefits 

Earnings lost due to premature death or disability, which will be averted, are indirect 

benefits. Debility as an impairin, factor in production has not attained the prominence in 

empirical studies that Mushkin (41, 51) attached to it from a conceptual standpoint. 

Since the publication of Rice's studies (53) it is no longer necessary to estimate loss of 

earnings on the back of an envelope. Drawing fully on the data resources of the federal 

government and using unpublished tabulations almost as much as published ones, 

Rice (43-44) prepared her estimates in systematic fashion. She applied labor force 

participation rates, employment rates, and mean earnings, inclusive of fringe benefits, to 

the population cohort in question. For men and women separately, she derived estimates 

of the present values of lost earnings due to mortality under alternative discount rates and 

a one-year estimate of lost earnings due to disability or morbidity. 

Several elements of the benefit calculation that were still at issue a decade or so ago 

appear to be more or less settled now, some perhaps prematurely. These can be 

summarized as follows:

- 358 

Application of Cost-Benefit Analysis to Health Sevmic / 

1. Our ordinary concern is with loss in earnings, not income. The latter includue 

income from property. 

2. Consumption by survivors is no longer subtracted from gross earnings in order to 

artive at net earnings. Viewed prospectively, everybody is a member of society, including 

the patient (54). 

3. The value of housewives' services is recognized, despite the fact that such services 

are not traded in the market and are omitted from the GNP. Weisbrod (13, pp. 114-119) 

developed and applied a complex method for measuring the cost of a substitute 

housekeeper, but subsequent writers have followed Kuznets (55) in employing a simpler 

approach, putting the value of the services of a housewife at the level of earnings of a 

full-time domestic servant. To employ a single number is the more practical procedure by 

far. The magnitude of that number is a separate question, however. It is increasingly 

evident that the value given by the earnings of a domestic servant is not adequate (56). 

Thus, the accepted value of the housewife's contribution would increase substantially if 

day care centers for working women were expanded at public cost. 

An alternative approach is to value the housewife's contribution at the opportunity 

cost of her staying out of the labor force (45). Implementation of this approach is 

impeded by two considerations (57). First, the method is complicated, since values would 

vary with the individual housewife's educational attainments, type of occupation, amount 

of job experience, full- or part-time employment status, etc. Second, nonpecuniary 

factors, which certainly influence the labor force participation rates of women, are 

difficult to measure and may behave erratically. When total family income permits, the 

pecuniary opportunity cost of the wife's staying home has been known to be as low as 

zero or even negative. 

4. The employment rate has been typically taken at 96 per cent, or an overall level of 

4 per cent unemployment at the level of "full" employment (44). In the 1970s the 

magnitude of this rate is at issue. Whatever the magnitude, Mushkin's argument is 

accepted that the health services system should not be charged with failures by the 

economy to provide jobs to all who seek them (41, 58). 

What is often not taken into account is the tendency for persons rehabilitated after 

serious illness or injury to find fewer job opportunities than persons who have remained 

healthy and on the job. In my study of syphilis (45), 1 recognized the loss of earnings due 

to the "stigma" attached to this and similar diseases. When prevention is feasible, it seems 

appropriate to assign to it an extra weight or bonus for this reason. 

5. Calculations of indirect benefits rest on the implicit assumption that the life 

expectancies of cohorts of potential survivors are known. Usually standard life tables are 

employed, separately for men and women. For diseases of low frequency it seems 

reasonable to disregard any effect on the total death rate occasioned by the deletion of a 

particular cause of death. For major diseases the problem is important, although simple 

deletion may be incorrect. As Weisbrod (13, pp. 34-35) recognized more than a decade 

ago, survivors who have avoided a particular cause of death may have a higher or lower 

susceptibility to other, competing causes of death. I compared the effects of simply 

deieting heart disease as a cause of death on life expectancy and on work-life expectancy. 

The former was large-1 1 to 12 years-and the latter was small-less than a year (46). For 

a disease with heavier impact at the younger ages, the effect on work-life expectancy 

would be relatively larger, and correspondingly greater attention would have to be paid to 

the effect of competing causes of death. 

J

1 Klarman 

Intargibl Benefits 

- .5 - 

Pain, discomfort, and grief are among the costs of illness currently borne, which 

constitute the intangible benefits of a program of health services that averts them. The 

benefits accrue partly to the patients and partly to their friends, relatives, and society at 

large, to the extent that we. take pleasure in the happiness of others. That positive 

externai effects in consumption exist is indicated by personal and philanthropic gifts. to 

the extent that they are not subsidized by the deductibility provisions of the income 

tax (59). Looming even larger is the averted premature loss of human life. Since none of 

these effects is traded on the market, none carries a price tag. In attempting to put a value 

on averting them the question arises: what would one be willing to pay to avoid them? 

In my paper on syphilis (45), 1 estimated willingness to pay for escaping the early and 

late manifestations of the disease by examining expenditures incurred in connection with 

other diseases that met certain conditions. After consultation with clinicians 1 adopted 

psoriasis as the analogue for early syphilis, and terminal cancer as the analogue for its late 

stage. The conditions specified were that the expenditures for medical care represented 

principally a willingness to pay for freedom from the particular disease, since in neither 

case could direct or indirect tangible benefits, as defined above, be realized. To the extent 

that payments were made only by the patient (directly or through health insurance), 

willingness to pay by others was neglected and total willingness to pay was understated. 

Neenan (60) has estimated the consumer benefit of a community chest x-ray program 

for tuberculosis. With the help of some fee data indicating willingness to pay, he obtained 

very high estimates of value. 

Several years have elapsed since intangible benefits were valued. The analogous diseases 

approach has not been repeated; this suggests that neither the estimates themselves nor 

the procedures for obtaining them have been found useful. One reason is obvious: the 

approach is specific, calling for the development of estimates, disease by disease. 

A larger body of literature is devoted to the value of human itfe than to the other 

types of intangible health benefit. Life insurance holdings are clearly not applicable to 

bachelors and jury verdicts are inconsistent (13, p. 37). The implications of public policy 

decisions or govemmental spending are difficult to elicit in the absence of information on 

the alternatives that faced the decision makers (19). Moreover, such valuation may lack 

stability and consistency (24, pp. 133-134). 

Schelling(61) has proposed a different approach. He would measure the value of 

human life, as distinguished from livelihood, by the arnount people are willing to spend to 

buy a specified reduction in the statistical probability of death. Acton (24, p. 258) 

applied this approach, and derived an estimate of the value of human life at $28,000. This 

amount serves as a substitute for the net value of lost earnings and is not an additional 

sum. 

I am not sanguine about the applicability of Acton's numerical estimate to the 

evaluation of program alternatives. Acton was the first to criticize the defects in his 

estimate, including the small size of his sample, and its apparent biases. While these 

defects can be remedied in the future, what troubles me is the likelihood that respondents 

to this type of question may not grasp its meaning. Do respondents know the actual 

probabilities of their dying in the coming year? How is a small-e.g. 1 per cent-reduction 

in statistical probability perceived? How much more is a 10 per cent reduction worth 

than a 1 per cent reduction? Is it plausible to postulate a strictly linear relationship

- 360 - 

Application of Cost-Benefit Analysis to Health Services 1 

between increase in risk and willingness to pay to cover it? (54). Moreover, does not the 

value of a gain depeno somewhat on the starting point? (62). If all payments come from 

tbl consumer, the distribution of income must exert a sizeable influence; by how much 

would will igneas to pay change if the task of reducing the death rato were viewed es a 

collective responsibility that is fully financed from public funds? 

Titmusa (63) regards the value of human life as priceless and beyond valuation. Yet 

implicit values are being placed on human life whenever public policy decisions are made 

on highway design, auto safety, airport landing devices and traffic control measures. 

mining hazards, factory safeguards, etc. In emphasizing voluntary giving, the sense of 

community that the gift relationship in blood both reflects and promotes, Titmuss seems 

to be pointing to a large external benefits component that is neglected when life-time 

earnings are taken as the proxy for the value of human life. Although the concern for the 

altruistic motive is salubrious and appropriate, the conclusion does not follow that human 

life is priceless. 

As Mishan (54) observes, a rough measure of a precise concept is superior to a precise 

measure of an erroneous concept. It is agreed that the notion of the value of human life, 

apart from livelihood, is sound. A numerical estimate of this value would be useful in 

comparing the worthwhileness of alternative programs. Comparisons of programs would 

gain in relevance and aptness if all benefits were counted, including the saving of human 

life or improvements in life expectancy. This potential gain is much more likely to be 

realized if all benefits are entered into the model, rather than if some appear only in 

footnotes. 

I am unable to say at this time how such a number or set of numbers for the several 

age groups can best be derived. Certainly Schelling's questionnaire method (61) can be 

improved. Perhaps the implications of past or existing public policies will yield a narrower 

range than one expects. It is conceivable that a committee can do a better job in the 

realm of values than in the realm of fact. In any event, the value of human life is probably 

higher for identified and known individuals than for members of statistical populations. If 

so, incurring extraordinarily large expenditures in behalf of the former is far from 

conclusive evidence of irrational behavior. 

Weisbrod (13, p. 96) avoided dealing with the problem of valuing intangible benefits 

by assuming proportionality to tangible benefits. This is an unsatisfactory solution, given 

the'differential impacts of various diseases on life expectancy, disability, and morbidity. 

However, a solution to this problem was not needed when the emphasis of public 

expenditures analysis shifted from cost-benefit to cost-effectiveness. To repeat, in 

cost-effectiveness analysis outcome is expressed in physical terms, e.g. iife years gained, 

and the task of analysis is to discover the program that will yield the desired outcome at 

the lowest unit cost. In the health services it goes without saying that desired outcome 

incorporates a constant level of quality of care, or at least an acceptable level. 

Cost of Progrum 

The estimate of the cost of a proposed program, with which benefits are compared, 

poses no special difficulties. A budget is prepared in terms of the market prices of inputs, 

which may be adjusted by shadow prices when warranted. 

If programs vary in size, it is appropriate to examine the possibility that economies of 

scale exist (14, pp. 82-83). However, since health services are rendered in the local area,

Kllarman 

- 361 - 

thi prospects of realizing such economn ¡ies are much more limited than in the manufacture 

of goods. Moreover, when the size of a program increases, factor costs may rise. Finally, 

ua the scopo of a program approaches the size of the total population ?t risk, the extra 

coast of additional units of output increases when increasingly resitant grouap are 

encountered. Conversely, it has been suggested that in the early phases of a progran unit 

cast is likely to be higher than later on, since administratora learn by doing (14, p. 24) 

Cost-Benefit Verau Cost-Effectivenss Analys 

Although it is not so difficult to estimate the costs of programs, it is quite difficult to 

formulate the contents and expected outcomes of programs. In my judgment this has 

been the chief obstacle to the useful application of cost-benefit or cost-effectiveness 

analysis in the health field. 

Elsewhere 1 have listed the data required by the economist for valuing outcomes (46). 

A clear statement of each type of outcome is necessary. Certain events, such as death, 

disability, extra unemployment, and the use of health services must be entered on a 

calendar, beginning with the base year, and assigned a duration. The data should extend 

for a period as close to a person's lifetime as possible, with particular attention to the 

possible recurrence of illness and its exacerbation. 

This list of data requirements implies a degree of knowledge about the effects of 

health services on the health of a population that is often lacking. The obstacles to the 

attainment of such knowledge are many. Medicine is not an exact science, and physicians 

may disagree among themselves and the same physician may disagree with his own past 

findings. Field studies are complicated by what Morris (64) calls the iceberg 

phenomenon: members of the designated control group, who are presumably normal, 

may in fact have the disease under investigation in asymptomatic form. The possibility of 

inducing iatrogenic disease means that only studies performed on normal populations in 

the community, which are far more costly than studies of captive clinical populations, 

can yield valid results (65). 

A serious gap in existing data arises from the lack of longitudinal studies of 

populations. Few investigators possess the requisite patience and dedication, or 

experience the necessary career stability. The funding agencies, under conditions of 

budgetary stringency, have even shorter time horizons. Although statistical manipulation 

of existing cross-section and time-series data is a much cheaper and almost alwaya 

available approach, it may not afford an adequate substitute in many instances, especially 

when a high degree of correlation exists among the independent variables under scrutiny. 

In 1965 I reported that only one study met the longitudinal data requirementa listed 

above-Saslaw's study (66) on rheumatic fever. Unfortunately, the report on this study 

was truncated in publication. Neenan's study of chest x-rays for tuberculosis (60), 

conducted in 1964, concentrated on the short term, on the ground that a recovered 

patient suffers no impairment of earnings while early detection alone does not alter the 

long-term outlook. No evidence is adduced for these assumptions. 

Acton (24) has recently completed a cost-benefit analysis of alternative programn for 

reducing deaths from heart attacks. He considered five programa: an ambulance with 

specially trained nonphysician personnel; a mobile coronary care unit with a physician; a 

community triage center; a triage center combined with the ambulance; and a program to 

screen, monitor, and pretreat the population. The largest net benefit, whether measured

- 362 - 

Application of Cost-Beneflt Analysis to Health Services / 

by the number of live saved, or valued by the criteron of eaninm or that of wm _l~gms 

to pay, is given by the screening, monitoring, and pretreatment progrmnL However, tie 

v~lu of personal time lost in screening is neglected, and the screeing prosgrn seema to 

dplay grat variability in outcome. Acton's cost-benefit analysis follows his costeffectivenes 

analysis, in which iUves saved are not asagned a value. The screenin, 

monitoring and pretreatment program yields the largest number of lives saved but thi 

verage cost per life saved is second from the highest and the marginal cost of saving two 

dditional lives is 524,000 each, compared with the estimated average and marginal cost 

of $3,200 for saving the first 11 lives under the ambulance program (24, p. 117). Acton's 

work is noteworthy for the wealth of detail on the epidemiology of heart attacks. 

physiology, treatment, and delivery systems for treatment or prevention. This economist 

drew extensively on the expertise of health services speciallsts and investigaton. 

The report of the Gottschalk Cornmmittee to the Bureau of the Budget (BOB) (67) 

contains a cost-effectiveness analysis of altemative modalities for treating chronic 

end-stage kidney disease. The problem facing the BOB, and posed to the Committee, was 

to define the appropriate role of the federal government in this field. The conclusion that 

a substantially expanded federal role was warranted was reached on other grounds, which 

did not entail an economic analysis. These included: some veterans were already receiving 

free care in veterans administration hospitals; several foreign countries, each poorer than 

the United States, were committed to delivering this service to all patients; voluntary 

insurance plans in the U.S. were not paying for the cost of prolonged hemodialysis and 

their leaders saw no prospect of doing so; the patients requiring treatment largely 

comprised middle-aged adults; and what was still a unique life-saving measure was 

available for application to known individuals, persons who would otherwise die in short 

order. Once the recommendation was made in favor of an expanded role for the federal 

government and a feasible mechanism was designed to finance the care rendered to 

individuals, the problem that remained for economic analysis was how best to discharge 

this responsibility-through hemodialysis in an institution, hemodialysis in the patient's 

home, kidney transplantation, or some mixture of these modalities. 

The cost-effectiveness analysis clearly pointed to the superiority of the transplantation 

route, which incorporates hemodialysis both for initial and back-up support. When 

hemodialysis is necessary, doing it at horne is much cheaper (29). These findings 

influenced the Gottschalk's recommendations to the BOB that kidney transplantation be 

expanded as far as possible. (The best mix of modalities was not solved for because the 

implied assumptions of constant unit cost and constant utility of a life-year made it 

unnecessary.) Again, as noted with respect to Acton's study, the economic analysis drew 

heavily on the underlying epidemiologic, physiologic, and clinical data developed by or in 

behalf of the Committee. 

Had the Gottschalk Committee performed a cost-benefit analysis, it seems plausible to 

postulate that a shortage of kidneys for transplantation and the relatively greater ease 

with which hemodialysis facilities can expand might have yielded a higher net benefit 

value for dialysis, at least in the near future. However, allowance for the superior quaiity 

of life under transplantation would constitute a partial offset (29). 

Contrary to some impressions(68), the Gottschalk Committee did inquire into 

available prevention programs and determined that, for the foreseeable future, the 

number of eligible patienta with end-stage kidney disease would not change. The 

Committee did not inquire intra the dispersion of the distribution of life years gained.

- 363 - 

Thua, it did not consider whether an average gain of 10 yeara is worth the nme wben it ia 

the product of 10 years each gained by 100 per cent of the population at ditk, or of 20 

yean each gained by 50 per cent of the population, or of 40 years each gained by 25 por 

cont of the population. Can it be said that the marginal utility of an additional year is 

constant or does the principlo of diminishing manuginal utility govern? 

The Committee did not have to deal with two problems that might arise undor 

different circumstances. One is that even cost-effectiveness analysis is not so simple as it 

appears to be when two or more types of outcome are sought as goals. If only one 

outcome, such as life years gained, is preeminent, other outcomes may be neglected. 

Where all outcomes are important-reduced mortality, lower morbidity, and less 

disability-it becomes necessary once again, as under cost-benefit analysis, to arrive at 

common or weighted measures of outcome for alternative programs (69). Only the 

problem of valuing intangible benefits is escaped. However, in cost-effectiveness analysis 

the focus is confined to outcomes common to health services programs, and the weighting 

problem is serious only when the several types of outcome do not occur in the same 

proportions for every program. The second problem not faced by the Gottschalk 

Committee is the appropriate role for government to assume if expensive life-saving 

measures became practicable for other organs of the body. Nor did the Gottschalk 

Committee attempt to deal with the question of increases in patient load if the very 

success of the program it sponsored led to the relaxation of criteria governing patient 

eligibility for treatment. The quantitative effect of such relaxation may be appreciable. 

In the years 1966-1967, during the early spread of PPB in the federal government, a 

number of cost-effectiveness studies were carried out in the Department of Health, 

Education, .and Welfare (23, 49, 68). Although costs and benefits were calculated 

simultaneously, the link between the inputs and outputs of programs was measured too 

often by means of hypothetical numbers. Once the relationships were postulated, no 

effort was made to pursue the measurement problem through empirical inquiry in 

subsequent budgetary periods. In certain instances only expenditures chargeable to the 

federal budget were counted as costs, neglecting expenditures incurred by individuals and 

by other levels of government (23, 24). 

PROBLEMS IN ASSESSING HEALTH SYSTEMS TECHNOLOGY 

This discussion of the potentialities and hitherto modest achievements of cost-benefit 

analysis and cost-effectiveness analysis in the health field bears directly on the analysis of 

the development and spread of health systems technology. However, changes in 

technology bring to the fore an additional factor: a heightened degree of uncertainty 

concerning future benefits and costs. According to systems analysis, one appropriate 

response to the prospect of uncertainty is to perform a sensitivity analysis, concentrating 

on a few key factors or assumptions to which the measure of costs or benefits appears to 

be especially sensitive (34, 70). This proposition strikes me to be a formal one, awaiting 

empirical content. 

Nevertheless, within the time frame of a decade, any allowance for uncertainty due to 

developments in technology may be excessive. It has been suggested that the technology 

that will be applied in the next ten years is already known, and that the pattern of 

technologic diffusion is discernible (71). This view may be too sanguine, but it is not 

contradicted by the record of the Gottschalk Committee. By wisdom or good hluck, the

364 - 

Application of Cost-Benefit Analysis to Health Services ¡ 

Comnittee's projections of survivorship of patients with transplanted kidneys and the 

cot of hemodialysis at home, both of which were originally supported by scanty data, 

hane been borne out (72). 

If technologic developments over the next decade are, in effect, already known to 

those gifted with early recognition, what can be said about prospective benefits and 

costa? In a plea at a health services research seminar in New York City for more research 

and development funds, Bennett (73) argued that the half.finished invention is the most 

costly product, so that technologic progress is bound to bring a lower unit cost of service, 

a well as improved performance. 

In those cases where straightforward development takes place and serious adverse 

ide.effects are not encountered, Bennett's view of the cost-reducing and benefitenhancing 

effects of technologic progress is undoubtedly correct. However, in many 

respects the future is shrouded in uncertainties. Such factors as the size and geographic 

distribution of population, value structures, and political decisions are uncertain for the 

future, even if technologic developments are not. Public policies are also known to create 

unintended and unanticipated consequences. An accepted way to deal with uncertainty is 

to provide for flexible operation, that is, to avoid a finely tuned operation which yields a 

minimum cost only for a particular scale of output. Similarly, if manpower is to be used 

flexibly in the future, it must be endowed with a more general education than otherwise. 

Thus, flexibility, whatever its cause or source, imposes a modest extra cost over a 

moderate range of outputs (18, pp. 105, 123-124). 

The Histoncal Record 

Rather than pursue this argument of pros and cons, I propose to examine the historical 

record. What have been the effects of past changes in health systems technology on costs 

and on benefits? A review of the literature on this subject reveals sharp differences of 

opinion. 

In a monograph on hospital expenditures sponsored by the National Center for Health 

Services Research and Development, Feldstein (74) attributes most of the postwar 

increase in hospital cost to an increase in demand, or, more precisely, to an upward shift 

in the demand curve. To paraphrase his argument, technical change in the absence of 

scientific progress may occur for two different reasons. Economic analysis has 

emphasized technical change in response to a shift in the relative prices of inputs (75). If 

wages rise faster than the prices of other inputs, for example, hospitals will economize on 

labor by using more disposable items, by automating laboratory procedures, etc. The 

effect of such substitution is to prevent costs from rising as fast as they otherwise would 

have. 

The second reason for technical change without scientific progress, which Feldstein 

emphasizes, is a shift in the demand for hospital care. This type of change generally yields 

a new product. The spreading of high-cost techniques is primarily due to rising income 

and increased health insurance coverage. As income increases, patients tend to raise the 

valuation of more costly care by relatively more than the valuation of less costly care. An 

increase in the proportion of the hospital bill paid by insurance will shift hospitala to 

more expensive technology, as the out-of-pocket price per unit of benefit is lowered. 

Gains in scientific knowledge, including managerial innovations, that have the 

potential of lowering the cost of care may actually have the opposite effect. This happens

I Klannan 

- 365 - 

again if the new scientific knowledge raises the benefits of expensive care by relativly 

more than the benefits of inexpensive care. In addition, if patients' real preferencs do 

not prevail but hospitals persist in producing services with the most expensdv techniques 

for which benefits are not less than cost, scientific progres cannot lower cot per patient 

day. 

In a monograph on physician expenditures, Fuchs and Kramer (76) draw a dhrp 

distinction between the effects of demand factors and those of technology. Their 

arguments concerning technology reflect an historical perspective, and may be 

paraphrased as follows. The late 1940s and early 1950s were marked by the introduction 

and widespread diffusion of many new drugs, particularly the antibiotica, which had a 

pronounced effect on the length and severity of infectious diseases. Since the mid-1950s, 

advances in medical technology have not brought about a similar improvement in the 

ability of physicians to improve health. Renal dialysis, cancer chemotherapy, and open 

heart surgery may achieve dramatic effects in particular cases, but bring about only 

marginal improvement in general indexes of health. Moreover, the early advances tended 

to be physician-saving, while the later ones were characteristically physician-using. The 

improvement in health resulting from the early advances was so great, that it turned the 

anticipated slight rise in demand for physician services into a slight decline. The reason is, 

according to Grossman (77), that healthier people have less objective need for physicians' 

services. By contrast, Fuchs and Kramer conclude that changes in demand factors had 

little effect on expenditures for physician services before the advent of Medicare and 

Medicaid in the mid- 1960s. 

In effect, whereas Fuchs and Kramer view technology and the conventional demand 

forces as being independent of one another, Feldstein holds that the effects of technology 

may also be exerted through a shift in demand. Both positions are stated ably and 

forcefully. As often happens, each raises more questions than it can answer. It would be 

premature, therefore, to attempt to pass judgment on the validity of the respective 

findings concerning the effects of technology in the postwar era. 

In a study focusing on the marked acceleration in the upward trends of costs and 

expenditures for hospital and physician services in 1966 (78), I have argued, though by 

no means-conclusively, that the large expansion in cost reimbursement to hospitals and 

the adoption of a new, previously untried method of paying physicians at reasonable and 

customary fees, subject to the prevailing distribution of fees in a local area, must have 

exerted strong effects of their own. In the case of hospitals, cost reimbursement for most 

patients leads to an impairment of financial self-discipline, since a dollar need only be 

spent in order to be gotten back. In my judgment, this proposition holds true for any 

institution, whether it be under voluntary nonprofit, governmental, or proprietary 

auspices. So far I am not persuaded by the empirical studies that have reached 

conclusions to the contrary (79, 80). 

A number of works have appeared that attempt to explain the behavior of the 

nonprofit hospital (81-85). They are, for the most part, far.-ranging and enlightening. One 

is also entertaining, positing a theory of conspicuous production, with the hospital's 

objective taken to be the closing of a status gap (85). None really attempts to deal with 

the sharp discontinuity in hospital cost and price behavior beginning in 1966. 

A rise in personal income may lead to greater reliance on technology for still another 

reason. For example, many persons are unable to stop smoking. A higher income enables 

them to pay more for cigarettes with a filter and reduced tar and nicotine contents.

- 566 - 

Application of Cost-Benefit Analysis to Health Senvice. 

Sbmilay, a higher income permits people to spend more on automobiles with safety 

dppsU, reducing the noed to exert influence on the behavior of drives. It may be mort 

effct~vh to operate on impersonal environmental forces than to try to chingo the 

bair of individuals (86). 

At this tim no general answer is discernible to the question of how change in halth 

uystan technology affect costa and benefita. It happens only once in a generation. 

perdapa even less frequently, that an idea such as early ambulation after surgery is born of 

ecessity in wartime, effects huge savings in tha use of health resoutrcs, and also exerta a 

positive effect on health. In most cases, the effects of technology will be mixed. Often 

the product is new, in the sense that a treatment is created that was not available 

previously and therefore could not have been demanded. The decision of whether or not 

to adopt a piece of technology, and the extent of its spread once adopted, depend on a 

number of factors, including the values of consumers, the motivations of providers, tha 

availability of funds, methods of provider remuneration, as weUl as the cost and efficacy 

of the service in question. 

Such a general formulation of the problem of assessing health systems technology, as 

provided above, affords practically no guidance to decision making. Only the concrete 

circumstances surrounding a project or program can indicate the special problems of 

measurement and valuation and the unique opportunities for solving them, what is to be 

emphasized in the analysis, and what may be neglected with only a moderate degree of 

trepidation. Accordingly, I will examine two examples in detail: hospitals and automated 

multiphasic screening (58). 

The Hospital 

Economists have offered essentially three views concerning capital investment in the 

hospital. First, hospitals invest too little capital, hence their productivity gains lag behind 

those of the economy at large (87). Second, hospitals invest too much, because grants and 

bequests accrue to them at zero price (88). Third, there is no optimum amount of 

investment in hospital beds, since there is no standard of appropriate hospital use (89). 

Conceivably, each position may have some merit to the extent that it reflects the 

situation in different sectors of the hospital. 

For simplicity I shall employ a threefold classification of hospital capital 

investment-patient beds, supporting housekeeping serviCes, and ancillary medical 

services (82). The unique problems of measurement and valuation facing the application 

of cost-benefit or cost-effectiveness analysis will be explored for each sector. 

Patient Beds The heart of the exercise in evaluating a project to expand hospital bed 

capacity, in my judgment, lies in one's explanation of the phenomenon of hospital use. 

At one pole, if the primary deterrninants of use are biologic in nature, an increase in bed 

supply beyond a certain point must result in additional empty beds. If hospitals are paid 

at stated charges, empty beds inflict a heavy financial burden on each institution (79). 

The reason is that fixed costs constitute two-thirds to three-quarters of total operating 

costsa (90). Each institution would therefore be subject to financial self-discipline in 

building beds, and there would be little occasion for outside intervention beyond the 

provision of information on the plans of other hospitals. The effect of introducing more 

technology might well be to increase the proportion of fixed costa to total operating 

costs, thereby reinforcing the efficacy of financial self-discipline.

Klarman 

- 367 - 

At the other pole, if all beds built tend to be used under conditions of prepayment, u 

Roemer(91,92) first suggested, there is no automatic criterion for an optinum bed 

supply. In the absence of evidence that low hospital use has an unfavorable effect oa 

health status, the appropriate public policy is to dcamp a tight lid on bed supply (79). The 

applUcation of more or lens technology ia the hospital is beside the point, although it does 

mem preferable to operate any extra beds as cheaply as possible. 

Patient census a a function of bed supply in the long run; combined with patient mix, 

it seta the requirement for nursing personnel, which may be viewed largdy au a 

requirement for personal services, with little or no substitution of equipment permitted. 

However, substitution is possible among levels of nursing personnel. The extent of actual 

substitution of low-paid for high-paid staff is perhaps overstated by the failure of hospital 

budgets to incorporate expenditures for special duty nurses. 

Housekeeping Services. 1 do not see any problems of sophisticated analysis in the area 

of supporting housekeeping services. Here the appropriate criterion for decision making is 

that of cost minimization. Bed sheets and towels are to be washed as cheaply as possible, 

for a given specification of whiteness. Patients' rooms and corridors are to be kept clean 

as cheaply as possible. Meals of a given quality-nutrition, calories, hot or cold-are to 

cost as little as possible. 

Once it is recognized that certain products or services need not be produced by the 

hospital but can be purchased from the outside, the problem is that of developing valid 

comparisons of unit cost. In addition, some administrators may wish to allow for certain 

risk factors. In the absence of competition among suppliers, the sales price may be quoted 

artificially low at the outset, only to be raised later. Also, in the absence of competition, 

purchases from the outside may increase the risk of running out of inventory. 

Apart from an allowance for lower risk associated with production within the hospital, 

estimates of internal cost of production should include only differential cost. No portion 

of overhead cost should be attributed, because this would continue in entirety after 

internal production ceased. Moreover, top management will perform the same role as 

coordinator, whether some goods and services are produced inside the hospital or 

acquired by purchase. 

In fact, the rise in hospital wages and gains in productivity attainable in large-scale 

manufacturing have led hospitals to increase the purchase and use of disposable items and 

ready packaged supplies. As Flagle (93) reports, gains in productivity from investment in 

large-scale plant have been achieved outside the health care system, which shares in them 

through purchase. 

If the objective of cost minimization is for a given level of cleanliness or nutrition, how 

this level is to be deterrnmined must be established. I doubt whether much would be 

accomplished by searching for effects on the health of patients. Rather, the criteria must 

be either patients' satisfaction or acceptability to management. Expressions of 

satisfaction are somewhat suspect, since patients are likely to be impressed by any display 

of interest in their opinions. A more practicable approach would be to compare 

alternative standards of service, none of them falling below adequacy, with the additional 

cost of attaining successively higher leveis. 

In some respects the computer partakes of a supporting housekeeping service and in 

other respects, when participating in diagnosis, it is akin to an ancillary medical 

service (94). The computer is a housekeeping service when it processes the payroll and 

issues bills to patients and insurance plans. As a substitute for older ways of bookkeeping

- 368 - 

Application of Cost-Benefit Analysis to Health Se~vlc / 

sud billlnS, the evaluation of compuler performance is straightforward. Docs it r~edu 

costs? If o, by how much? 

Medical Sa~ice Even when the computer helps in diagnosis the test is still coet 

reduction, if an older way of performing the same task is being replaced. There may be a 

complication, however. The cost of operating the computer falls on the hospital, whil 

savings in physician time accrue to the attending physician. The presence of distributional 

considerations suggests that the decision reached is not independent of who the decision 

maker is, or who exerts predominant influence on hia. 

Apart from the distributional considerations of who pays and who saves, evaluation of 

the worthwhileness of the computer in assisting in diagnosis is no different from the way 

another ancillary medical service, the laboratory, is evaluated. With respect to services 

that were rendered in the past, the test is simple. Does the new equipment save money or 

does it expand services for the same amount of money? In the laboratory additional and 

more costly equipment does replace technical personnel. A possible offset is the tendency 

to prescribe more services (95), although within the limits of existing capacity of 

equipment and staff the marginal cost of additional units of service is low. What is not 

known is how much good is accomplished, particularly in the absence of information on 

the timeliness of delivery of the reports on these services. 

Flagle (93) has reported economies achieved in patient surveillance due to continuity 

of use of the monitoring system in infusing blood. This finding strikes me as analogous to 

the finding in his early work (96) that a single channel is more efficient than two channels 

when the demand for services vanes stochastically. 

The intensive care unit is a more complex operation to evaluate. To the extent that it 

substitutes equipment for nurses it should cost less. However, the unit is also intended to 

save lives. The yield in life-years gained is properly subject to more sophisticated analysis. 

From this discussion it appears that cost-benefit or cost-effectiveness analysis is a 

plausible approach only if the service rendered is a new one or if the old product has 

changed appreciably, gaining new dimensions. When all benefits take the form of savings 

in health resources, that is, are direct and tangible benefits, the appropriate form of 

analysis is cost-benefit. When the preponderant benefits are intangible or life-saving, the 

dilemma is to choose between cost-benefit and cost-effectiveness analysis. On the one 

hand, cost-effectiveness analysis is easier to perform, since intangible benefits need only 

be measured but not valued. Indeed, according to Feldstein (97), even the problem of 

choice of discount rate is simpler in the case of cost-effectiveness analysis, with only the 

social time preference rate being relevant. On the other hand, to resort to 

cost-effectiveness analysis is to give up in advance whatever help analysis can offer in 

choosing among several objectives or program areas. It then becomes necessary to malke 

the choice among programs on other grounds, as the Gottschalk Committee did. 

1 am unable to discern a general resolution to this dilemma. It is certainly not evident 

how to establish priorities in a systematic way when cost-benefit analysis is abandoned. 

Perhaps the choice can still be made in a practicable way, with reasons explicitly stated, 

when remarkable benefits are under consideration, as in the treatment of end-stage 

kidney disease. When the benefits in question are modest but difficult to value, how is 

one to decide whether or not to adopt a particular piece of technology? To follow the 

lead of pace-setting organizations is almost always to say yes. Perhaps we should put trust 

in our ability to continue to improve the valuation of intangible bnefits in the 

future (28). Setting standard values on gains in life expectancy at various ages would

- 369 - 

seem worth exploring. However, I c-in also see increasing difficulty in the future in 

valuing direct tangible benefits, if fewer market prices become available for health services 

'n the event that provider reimbursernent shifts away from fee-for-service toward 

capitation and salay methods. 

Autonmted Multhipic Screening 

Often cited and discussed as an example of technologic development in the health fleld 

is autonmated multiphaasic health screening (98). The reports issued from the Kaiser- 

Permanente laboratories in Oakland and San Francisco reveal a good deal about the 

organization and staffing of such a service and present data on unit costs (99-103). No 

evaluation akin to cost-benefi¡ or cost-effectiveness analysis was attempted prior to 1973, 

when a preliminary cost-benefit analysis for middle-aged men was issued (104). 

Collen and associates(101) report that total costs for screening an individual are 

$21.32; which, they note, is only one-fourth or one-fifth of the cost of a periodic health 

examination employing more conventional modalities. The position of the authors is that 

this comparison will serve for the time being, pending determination of the efficacy of 

multiphasic health screening. The fact is that some people do.undergo a periodic health 

examination, whatever its efficacy may be. 

Garfield's position (105, 106) differs from that of Collen, in that the effectiveness of 

screening in arresting or curing prevíously unknown disease is beside the point. For 

Garfield, automated multiphasic screening has assumed a useful social function, serving as 

a sorting mechanism tor patlents witlh ,,epayment who would otherwise flood the health 

services system. 

I have difficulty with both positions. Collen's comparison of cost with that of the 

periodic health examination reminds one that the latter procedure is notoriously 

controversial, with the central issue revolving precisely about its effectiveness. Among 

physicians there appear to be true believers, persistent skeptics, and ambivalent 

prescribers (107-109). Furthermore, as emphastzed in the Nuffield report (110), screening 

implies an invitation to the patient to come and see the doctor who pronmises him a 

'favorable outcome. This is in contrast to the more usual visit initiated by the patient who 

has symptoms and seeks relief. 

My criticismns of Garfield's position are more serious, for his view that automated 

multiphasic screening shouJd be regarded as a sorting mechanism, a substitute for the 

rationing of services by pricc, raises a host of questions. Apparently, judging from a more 

recent presentation of his pGsition (1 1), much of Garfield's argument is based on an 

interpretation of what happened under Medicare and Medicaid. To my knowledge, the 

Medicare program experienced only a modest increase in the use of services and a huge, 

unexpected, increase in unit cost. There is no way to interpret the unanticipated rise in 

expenditures under Medicaid in the absence of data on trends in size of the eligible 

population, per capita use, and unit cost. My own view is that the increase in eligible 

population may have been the major factor.' 

Garfield (111) has hypothesized a difference in price elasticity of demand between the 

sick and his other three categories of patient-the well, the worried well, and the early 

sick. However, there have been no empirical studies of the demand for physician servica 

i See Klanan, H. E. Ma~j public initiative in health care. PubUc Intret 34: 106-123, 1974.

"~ ·I~t;~3?r·,,' ~ I o,,;xrvWj ', x· .. i li, 1 ' 

- 370 - 

Application of Cost-Benefit Analysis to Health Setvices / 

in which people are so classified. From other studies it would appear that a host f( 

factor such as health insurance, earnings as an expression of the value of time, ae, arad 

the supply of providers, ae important determinants of the demand for physici a 

ei ices (1 2). 

The assertion that the supply of services for sick care is inelastic is not unique to 

Garfield. In the area of trends in the education of physicians, which ltakes longr and 

therefore responds more slowly than any other health occupation, my own readin 

indicates that even this system has been somewhat responsive, even while insiting that 

clas size in medical school must be kept small (18, p. 101), and still moro responsive 

after the policy decision to expand enrollment was made and implemented by funding. 

Whether the supply response has been sufficient to meet rising demand is, of course, a 

different issue. 

The most serious reservation I have about Garfield's position touches closely on the 

nature and function of cost-benefit analysis. If complete prepayment serves to create a 

condition of perpetual excessive demand, then some rationing or control measures are 

dearly indicated. Why assume, without c¿mparing alternatives, that automated 

multiphasic screening is the most appropriate instrumentality? It seems to me that wher 

the stated purposes of a program change, so should the menu of alternatives to be 

considered. 

Two reports by Collen (101, 102) on the cost of screening fil a real need. Two 

measures are presented-cost per test and cost per screening. Cost per test reflects only 

direct departmental costs, whde cost per screening incorporates an allocation of overhead 

expense. The article published in 1970 (102) offers a costing rule: in order to allow for all 

costs incurred, double the reported cost per test. The earlier article (101), which appears 

to present essentially the same data, suggests a blow-up of 50 per cent; I am unable to 

account for the difference. 

Since the screening process is automatic, the capital equipment is indivisible, and all 

procedures are schedulable, economies of scale are to be expected. The larger the scale of 

operation, the lower is the average unit cost. However, to achieve the lower cost, full 

utilization of existing facilities is essential. Accordingly, it is said to be advantageous to 

have available a source of stand-by patients, such as those awaiting admission to the 

hospital (103). 

Collen's second article (102) goes beyond cost per test or per screening, and reports 

cost per positive case. For manmmography a prevalence rate of 1.2 per cent converts the 

unit cost of S4.90 into a cost per positive case of $408. Since one-fifth of the women 

with positive mammograms have cancer of the breast, the screening cost per true positive 

case is 52,000. His doubling rule would raise the cost to S4,000. The cost of diagnosis for 

all five women and of treatment for one is still excluded. 

The proportion of false positives is a function not only of the accuracy of the 

screening test but also of the prevalence rate (113, 114). There are two reasons for aiming 

to keep down the number of false positives: to avoid neediess anxiety, and to prevent 

iatrogenic disease associated with the diagnostic process itself. 

The data reported to date from the Kaiser-Permanente laboratories indicate that 

automated multiphasic screening is both feasible and affordable. The quaetion is whether 

it if; worthwhile. One answer is in terms of its effects on health. The Advisory Committee 

on Automated Multiphasic Health Testing and Services (AMHTS) (1 15) states that much 

of disease uncovered by testing will be chronic or not reversible; it will not yield a saving

iKlarman 

- 371 - 

in the use of services or an improvernent in health. There seems to be Little point to using 

multiphasic screening if this is the case. 

A second answer is that of Garfield(lll), which I have criticized at length. He 

provides no persuasive reason for choosing this instrumentality to control the use of 

physician services. 

A third answer is possible: that automated multiphasic screening is an integral part of a 

package of comprehensive health services to which everybody has a right. Usually a 

service is aspired to by the poor because the middle and upper classes are already getting 

it. This is not yet the case regarding automated multiphasic screening. 

Clearly, a reasonable answer can only be provided through an evaluation of automated 

health screening for its worthwhileness. The report by the Advisory Committee(115) 

states, "There are elements of AMHTS that defy cost-effectiveness analysis, but which 

depend primarily on medical, social, and scientific objectives." If I understand the 

statement, I disagree with it. It may be, however, that I do not understand it. What are 

the medical, or social, or scientific objectives that defy measurement? 

Following the formulation of data requirements given in the preceding section, 1 

propose that data be compiled to evaluate automated multiphasic screening as follows: 

the volume of disease detected that was not previously known;-what could be and in fact 

was done about all this disease; what the outcomes in terms of health status and 

subsequent utilization of services were; and at how much cost, inclusive of diagnosis and 

treatment, the outcomes were attained (116, 117). It must be added that, as indicated by 

a recent paper(ll8) which compares study and control groups for such measures of 

outcome as work and health services utilization, CoUen's group is steadily compiling more 

and more of the requisite data. Still lacking is information on costs that correspond to the 

specified benefits. 

Barriers to Systematic Analysis 

To bring some focus to a discussion of the necessary steps ahead, I have prepared a list 

of barriers to the systematic and rational analysis of expenditures for health systems 

technology. At the same time 1 shall assess the prospects for lowering or overcoming each 

barrier. 

1. When the costs of operation mount beyond all projections, the tendency is to argue 

that the computer or automated laboratory, as the case may be, is not merely providing 

services but is performing a research function. Yet doing things we know little about does 

not define research. Certain features of research, such as formulation of hypotheses, 

design of study, and capability for statistical analysis of data, are not necessarily available 

wherever services are rendered. Although some replication of research is desirable, it 

should be intentional and need not be universal (119). It follows that sources of research 

funds should exercise discrimination in allocating them. If the absorption of so-called 

research costs by patients is precluded, this tendency to encourage pseudoresearch will be 

minimizae 

2. A tendency exists to expand the range of functions said to be performed by new 

equipment. Surely, data on payroll could assist management in controlling cost by 

department; data on billings could provide a proxy for cost data by diagnosis. The first of 

these applications can be evaluated according to a strict criterion: is potential cost control

- 32 - 

Application of Cost-Benefit Analysis to Health Services / 

achev~d, so that savinga are realized? The second application can be judged on it owM 

merits as an intermediate good: of what value is such information and to whom? 

3. la the health field there ia a tendency to adopt the best available and atost 

teohnology in every institution. This drive is promoted by the medical ethic of dolng tbhe 

utmost for the individual patient and reinforced by current methods of paying providera 

by third parties. The voluntary nonprofit form of organizing hospitals is frequently 

mentioned *s a factor. Still another factor is usually neglected, namely, the nature of tho 

physician-hospital relationship in this country. Physicians who specialize in treating 

patients with a given disease will not accede to its exclusion from hospital A, where thoy 

hold a staff appointment, unless they are granted staff privileges in hospital B, where the 

planning agency would like to concentrate all facilities for diagnosis and treatment. Only 

in part are financial interests involved; equally, or even more important, is the 

preservation and application of professional skills 

4. Economic valuation has no meaning without a firm basis in the underlying data on 

the link between the inputs and outputs of specific programs. It is not often that 

economists can develop such data. Other investigators must be persuaded and enabled to 

do this by investing their time and energies in longitudinal studies. 

5. It is discouraging to perform technical analysis, to persuade the decision makers of 

its usefulness, to have it adopted, and then to discover that funds for health services are 

cut off because total government spending is being curtailed. Adjusting aggregate demand 

in the economy through changes in total expenditures is bound to result in the 

stop-and-go operation of individual programs. This is both wasteful and frustrating, and 

poses a substantial threat to continuity in the provision of health services through public 

financing. 

6. Since cost-benefit or cost-effectiveness analysis is economic evaluation of public 

projects or programs, it must inevitably take place in a political climate. While the 

economic tool of cost-benefit analysis implies a delineation of goals and an articulation of 

values, the imperatives of the political process may call for a blurring of differences and 

potential conflicts, in order to facilitate the building of coalitions aimed at the 

accomplishment of particular ends. Schultze(31) has observed this paradox: PPB has 

been applied most in an area, national defense, where future uncertainty is greatest but 

value differences among citizens have been traditionally least; PPB is not applied much in 

the human resources area, where the problem of uncertainty is not so serious, but 

differences in values among citizens prevail, as well as a great many vested interests. 

Some political scientists, such as Wildavsky (120, 121), would agree with the above 

description and conclude that such are the facts of life. Most changes in governmental 

budgets are incremental anyway and do not-indeed cannot-derive from base zero (122). 

Within the boundaries set by defined political understandings, there are ample 

opportunities to improve decision making through systematic analysis. There is no reason 

to believe that politicians prefer to make poor decisions over good ones. In cases that are 

of vital importance to the body politic, many politicians, when persuaded of the right 

thing to do, would be willing to use up some of the credit they have accumulated and 

make the tough, though unpopular, choice. They cannot take such a stand on every issue, 

however. Therefore, the exceptionally capable practitioner of economic cost-benefit 

analysis must know how and when to make an allowance for the existence of a political 

cost-benefit calculus (120).

Kmnnan 

- 373 - 

REFERENCES 

1. Haveman, R. H. Publc expendituras and poicy ailysis: An oervibw. In Publkc A'xpmnftu 

ard Poey Anaysyds, editd by R. L Haveman and 1. MargL pp. 1-18. Martha PGuiddn 

Company, Chcago, 1970. 

2. Eckssin, O. A suvey of the theory of public expenditure criterda. In Publ Fbics~ Neme, 

Sourees, and Utilizaton, edited by J. M. Buchanan, pp. 439-494. Prineton Univity re, 

Princton, NJ., 1961. 

3. Pr, A. R., ;nd Turvey, R. Cost-~ fit analys;: A survey. Economic Jowni 75(300): 

683-735, 1965. 

4. Stuseer, P. O. Public Expenditure Budcing. The Brookingo Istitution, Washington, D.C., 1969. 

S. Ecktein, O. Water Rcsources Development. Harvard University Press, Caunbridge, 1958. 

-6. Hirahleifer, 1., DeHaven, J. C., and Milliman, J. W. Water Supply: Econemi, Technology md 

Policy. University of Chicago Presa Chicago, 1960. 

7. Knese, A. V., and Smith, S. C. Water Research. Johns Hopkins Presa, Baltimore, 1966. 

8. Krutilla, J. V., and Eckstein, O. Multiple Purpate River Development. Johns Hopias Press, 

Baltimore, 1958. 

9. Mas, A., Hufschmidt, M. M., Dorfman, R., Thomas, H. A., Mrgiin, S. A., and Fair, G. M. 

Design of Water.Resource Systems. Harvard University Presa, Cambridge, 1962. 

10. McKean, R. N. Efficiency in Government Through Systems Analysir. John Wiley & Sons, Inc., 

New York, 1958. 

11. Arrow, K. 1. The organization of economic activity: lasues pertinent to the choice of mrnket 

venus nonmarket allocation. In Public Expenditures and Policy Anaiysis, edited by R H. 

Haveman and J. Margoli, pp. 59-73. Markham Publishing Company, Chicago, 1970. 

12. Samuelson, P. A. The pure theory of public expenditure. Review of Economics and Stastiria 

36(4): 387-389, 1954. 

13. Weisbrod, B. A. Economics of Public Health. University of Pennsylvania Pres, Philadelphia, 

1961. 

14. Sloan, F. A. Planning Public Expenditure on Mental Health Service Deli¿ery. The New York City 

Rand Institute, New York, 1971. 

15. Stigler, G. 1. The Theory of Price, rev. ed., p. 150. Macmillan and Company, New York, 1952. 

16. Fuchs, V. R. Health care and the U.S. economic system: An essay in abnormal physiology. la 

Technology and Health Care Systems in the 1980's, edited by M. F. CoUen, pp. 169-175. 

National Center for Health Services Research and Development, Washington, D.C., 1973. 

17. Lasagna, L. Quoted in Decision-Maaking on the Efficacy and Safery of Drugs, edited by J. D. 

Cooper, p. 2. Interdisciplinary Communication Associates, Washington, D.C., 1971. 

18. Klarmnnan, H. E The Economics of Health. Columbia University Presa, New York, 1965. 

19. Freeman, A. M. Project design and evaluation with multiple objectives. In Public Expenditures 

and Policy Analysis, edited by R. H. Haveman and J. Margolis, pp. 347-363. Markham Publishing 

Company, Chicago, 1970. 

20. Haveman, R. H., and Krutilla, J. V. Unemployment, Idlie Capacity and the Evaluation of Public 

Expendinres. Johns Hopkins Press, Baltimore, 1968. 

21. McGuire, M. C., and Garn, H. A. The integration of equity and efficiency critería in public 

project selection. Economic Journal 79(316): 882-893, 1969. 

22. Weisbrod, B. A. Income redistribution effects and benefit-cost analysis. In Probiems in Publc 

Expenditure Analysis, edited by S. B. Chase, pp. 177-209. The Brookings Institution, 

Washington, D.C., 1968. 

23. United States Department of Health, Education, and Welfare. Motor Vehicle Injury Preventon 

Progrwm,' Cancer Control; Arthrnit Conntro; Selected Discase Control Programs. Office of 

Assistant Secretary for Program Coordination, Washington, D.C., 1961. 

24. Acton, J. P. Evaluation of a Life-Saving Program: The Case of Heart Attacka. Doctoral 

Dliertation, Harvard University, Cambridp, 1970. 

25. McKcan, R. N. The use of shadow prices. In Problem in Public Expenditure Anaiyas, edited by 

S. B. Chase, pp. 33-65. The Brookings Institution, Washington, D.C., 1968. 

26. Marolis, J. Shadow-pricas for incorrect or nonexistent market values. In Puble Expenditura 

and Policy Analysu, edited by R. H. Haveman and J. Margolis, pp. 314-329. Markhun Publihl~ 

Company, Chicago, 1970. 

27. Fein, R. Defminition and acope of the problem: Economic aspecta. InAsorean the Elffecltienen 

of Child Health Servnces, editad by A. B. Bergman, pp. 44-SO. Rous Labotories, Columbus, 

Ohio, 1967. 

28. Weuabrod, B. A. Concepta of costa and benefita. ln Problems bh Publc Expendltuve An*b, 

edited by S. B. Chase, pp. 257-262. The Brookinp Institution, Washington D.C., 1968.

..· 74 - 

Application of Cost-Benefit Analysis to Health Snices / 

29. KIhn0m, H. E., Francia, J. O'S., and Rosenal, G. D. Cost effve~v anu appl~d to be 

tríatment of cboc renaul disoue. Med Care 6(1): 48-54, 1968. 

30. Gothabm, W. Allocatin federal tources amog competing socdl need HEW Ind m pp. 

1-13, August 1966. 

31. Schultz, C. L T7he Ptlds ad Economic ofPublic Spend~, pp. 1-34. 77-102. Ibs B1mdoo 

Institution, Washingto, D.C., 1968. 

32. United States Bur dof the BudLet Bulleti 66-3, Supplement, and Bultín 68-2 an 

Planninsftogra Budgetipn October 12, 1965, February 21, 1966, md July 18, 1967. 

Repnted in PinnM& Progmmnf Budgetint: A Systemr A ppoa to Mdmig~.et, ed by 

F. J. Lyden and E. G. Miller, pp. 405-443. Markham Publih~ng Company, Chico, 1968 

33. Klarman, H. E. Presnt status of coat-benefit nalyas in the health field. in J. Puble Hcdrl 

57(11): 1948-1953, 1967. 

34. Henderson, P. D. Investment criteria for public enterprie . Iln Pubc Ente se, edited by R. 

Tutvey, pp. 86-169. Penguin Modem Economies Readinga, Penguin Books, Baltimne, 1968. 

35. Musgrave, R. A. Cost-benefit analysis and the theory of public finance. Jounai of eo~ c 

Litemature 7(3): 797-806. 1969. 

36. Baunol, W. 1. On the discount rato for public project. In Public Expenditres aíd Pdlcy 

Analysis, edited by R. H. Haveman and J. Margolis, pp. 273-290. Markh Publshing Company, 

Chicago, 1970. 

37. Interest Rate Guidelines for Fedeml Decision.Mabing. Joint Economic Commnittee, Congrua of 

thi United Stats. U.S. Governnment Printing Offico, Washington, D.C., 1968 

38. Boulding, K. E. Notes on a theory of philanthropy. In Phnthropy and Pube Polly, edited by 

F. G. Dickinson, pp. 57-71. National Bureau of Economic Research, New York, 1962. 

39. Hufschmidt, M. M., Krutilla, J., and Margolis, J. Standards and criteria for formulating and 

evaluating federal water resources developments. In Heanngs on Guidelines for Estnri~w the 

Benefins of Public Expenditures, pp. 135-212. Joint Economic Comnmitte, Congre of the 

United States. U.S. Government Printing Office, Washington, D.C., 1969. 

40. Feldstein, M. S. The social time preference discount rate in cost benefit analysis. Economtc 

Journal 74(294): 360-379, 1964. 

41. Mushkin, S. 1., and Collings, F. d'A. Economic costa of disease and injury. Publc Health Rep. 

74(9): 795-809, 1959. 

42. Fein, R. Economics of Mentalllness. Basic Books, New York, 1958. 

43. Rice. D. P. Economic Costs of Cardiovascular Diseaes and Cancer. Health Economics Series No. 

5. U.S. Government Printing Office, Washington, D.C., 1965. 

44. Rice, D. P. Estimating the Cost of Iliness. Health Economics Series No. 6. U.S. Governmsnt 

Printing Office, Washington, D.C., 1966. 

45. Klarman, H. E. Syphilis control progpams. In Measuring Benefits of Government Investment, 

edited by R. Dorfman, pp. 367-410. The Brookings Institution, Washington, D.C., 1965. 

46. Klaxman, H. E. Socioeconomic impact of heart disease. In The Heart and Ciculation, VoL 2, pp. 

693-707. Federation of American Societies for Experimental Biology, Washington, D.C., 1965. 

47. Menz, F. C. Economicsa of disease prevention: Infectious kidney diseas. Inquiry 8(4): 3-18, 

1971. 

48. Fuchs, V. R. The contribution of health services to the American economy. Mllbani t Mcm Fd 

Q. 44(4, part 2): 65-101, 1966. 

49. Grosse, R. N. Cost-benefit analysis of health service. Annait 399: 89-99, January 1972. 

50. Klarnan, H. E. Conference on the economics of medical reaerch.L In Report of thd PReidenYs 

Commission on Heart Disease, Cancer, and Stroke, VoL 2, pp. 631-644. U.S. Govera^nt 

Printing Office, Washington, D.C., 1965. 

51. Mushkin, S. J. Health auan investment Joumal ofPolitical Economy 70(5): 129-157, 1962. 

52. Bowen, H. R. Towrd Social Economy. Rinehart and Company, Inc., New York, 1948. 

53. Rice, D. P., and Cooper, B. S. The economic value of human life. Am 1. Pub. Hralth 57(11): 

1954-1966, 1967. 

54. Mishan, E. J. Evaluation of life aid limb. Journal ofPolitlcalEconomy 79(4): 687-705, 1971. 

55. Kuzmet S. National Incom and In Composioias 919.1938, pp. 22-23. National Bur~a of 

Economic Research, New York, 1947. 

56. Walker, K. E., and Gauger, W. H. The Dollar Value of Housah~d WodL New Yok Stite CoD. 

of Human Ecology, ComeUl Univerty, Ithaca, New York, 1973 (procssed). 

57. Siramldin, 1. A-H. Non-Market Componento of Nariona lncome. Institute fr Sod E rch, 

Univerity of Michign, Ann Arbor, 1969. 

58. Gdman, A. C. Murltph~ic Healfh Testig Systems: Rcweu adAu notdotn . U.S. Gotnm 

Pinting Office, Wahiton, D.C., 1971.

- 375 - 

59. Valr y, W. S. One eonoist's vie w of phbllathmpy. In Phianthropy and Pu bhkl Poy, , edtd 

by F. G. Dickinaoa pp. 31-56. National Bureau of Economic Resarch, New York, 1962. 

60. Nmnan, W. B. Nomietive Evaluodon of a Pblkc HelIth Prom. Instituta of IPblUc 

Adminiratioa, Universty of Micigan, Ana Arbor, 1967. 

61. Sdlin T. C. Tibhe if you a~e may be yonr own. la Problemn Pftc iun Expma~ Anly, 

editd by 8 S. . Che, pp. 127-162. The Brookings lustitutb a, Was~itoa, D.C.. 1968. 

62.ó uow, L. Invesmn lon evnn Capittl, p 134. Wadsworth Publuh~ Compmr bic, , 

b bnoat, California, 1970. 

63. ltmu~ M . . 77T GIftRclationhip, p. 198. Pantheon Book , New York, 1971. 

64. Mona, J. N. The U# m of Epidemiy, p. 45. E. A S. Uvingsto , Edlbugh. 1957. 

65. Sdceff, T. J. Prefaerd e~ in diagnosi. Med C 2(3): 166-172, 1964. 

66. Sasuaw, LM., Vieta, A., and Myerburg R. Cost of rheumatic fe~r and itia prnt. AnM J. 

PuMic Health 55(3): 429-434, 1965. 

67. Report of the Committee on Chronic Kidney Dieas (C. W. Gottsc~L, chairman). Burenu of 

the Budget, Washington, D.C., 1967. 

68. Grosse, R. N. Problems of resource allocation in health. In Plubl Expendiur and Polkcy 

Andlysy, edited by R. H. Havenman and J. Margolí, pp. 5 18-548. Markham Publshing Company, 

Chicogo, 1970. 

69. Feldtein, M. S. Health sector planning in dewloping countriu. Economica, New Seaies 37(146): 

139-163, 1970. 

70. Quade, E. S. Systems analysi techniques for plannng-progrnming-budgetig. In PknniqL 

Programming, Budgeting: A Systerm Approeh to Meanaement, edited by F. J. Lyden and E. G. 

Miller, pp. 292-312. Mukham Publshing Company, Chicago, 1968. 

71. Sturmn H. M. Technology and Manpower in the Healtkh Servc Industry, 1965-1975. U.S. 

Government Printing Office, Washington, D.C., 1968. 

72. Kiarmn, H. E. Economic Aspects of Chronic Kidney Disease Revisited. Paper presented at 

annual meeting of American Public Health Association, November 11, 1969 (processed). 

73. Bennett, 1. L. The Scientific and Educational Basis for Medical Care. Paper presented at New 

York Health Services Research and Policy Seminar, December 1, 1970 (processed). 

74. Feldstein, M. S. The Rising Cost of Hospital Care. Information Resources Presa, Washington, 

D.C., 1971. 

75. Blaug, M. A survey of the theory of process innovation. In The Economics of Techndogcal 

Change, edited by N. Rosenberg, pp. 86-113. Penguin Moderm Economics Readings, Pengun~ 

Books, Baltimore, 1971. 

76. Fuchs, V. R., and Kramer, M. J. Expenditures for Physicians' Services in the United States, 

1948-1968, pp. 35-42. National Center for Health Services Research and Development, 


77. Grossman, M. The Demand for Health: A Theoretical and Empiricl Investigation. National 

Bureau of Economic Research, New York, 1972. 

78. Klarman, H. E. Increase in the cost of physician and hospital services. Inquiry 7(1): 22-36, 1970. 

79. Klarman, H. E. Approaches to moderating the increases in medical cate costa. Med Cere 7(3): 

175-190, 1969. 

80. Pauly, M. V., and Drake, D. F. Effect of thitd-party methods of reimbursement on hospital 

performance. In Empirical Studia, in Health Economics, edited by H. E. Klarman, pp. 297-314. 

lohns Hopkins Press, Baltimore, 1970. 

81. Davis, K. A. Theory of Economic Behavior in Non-Profit Hospital Doctoral Disertiaon, Rics 

University, Houston, 1969. 

82. Ginaburg, P. B. Capital in Non-Profit Hospitala. Doctorn Dis~ttion, Harvard Universty, 

Cambridge, 1970. 

83. Newhouse, J. P. Toward a theory of nonprofit institutions: An economic model of a hospital. 

Amerian Economic Review 60(1): 64-74, 1970. 

84. Reder, M. W. Some problems in the economics of hospitala. American Economic Revew (Papen 

and Proceedings) 55(2): 472-480, 1965. 

85. Lee. M. L. A conspicuous production theory of hospital behavior. Southern Economic Jaournal 

38(1): 48-58, 1971. 

86. Rosnstock, 1. M. Why peopie use health sevices. MEibank MemL Fund Q. 44(3, part 2): 94-124, 

1966. 

87. Ginzberg, E. Men, Money. and Medicine, p. 55. Columbia University Presa, New York, 1969. 

88. Long, M. F. Efficient use of hospital. la The Economics of Health and Me~drl Cat, editd by 

S. 1. Mushkin, pp. 211-226. Bureau of Public Health Economica, Univrsity of l~chin, Ana 

Arb~r, 1964. 

89. Anderm, O. W. Trenda in hospital usea and their policy implicationa. i Ftih Annual

376 - 

Application of Cost-Benefit Analysis to Health Servic / 

Sympoomnn on Hospital Affins, Where is Hospital Use Hcoded? pp. 2-5. Gradute Sehod of 

Bus, Univerity of Chicao, Chicago, 1964. 

90. Feldtin, P. J. An Empia~l Investigation of the Margil Cost of HospiIt S~rais C.Grduat 

Prognr in Hospital Admistration, University of Chicgo, Chicao. 1961. 

91. Roerer, M. 1. Bed supply and hospital utbization: A national experiment Holpilrs 35(21): 

36-42, 1961. 

92. Shain, M., and Roemrer, iL L Hospital costs relate to the supply of beda. Mo Hosp. 92(4): 

71-73, 168, 1959. 

93. Flale, C. D. Evaluation and control of technology in health senvices In Tecdnoky andHlcath 

Care SyJtems in the 1980'r, edited by M. F. Collen, pp. 213-224. National Conter for Health 

Services Research and Development, Washington, D.C., 1973. 

94. Shegog, R. F. A. Reviewing some applications of computers to medicine. n Problems and 

Pgress in Medical Cre, Third Series, edited by G. McLachlan, pp. 145-170. Oxford University 

Press, London, 1968. 

95. Ahlvin, R. C. Biochenical screening-A critique. New EngL J. Med 283(20): 1084-1086, 1970. 

96. Fla~e, C. D., Gabrielson, 1. W., Soriano, A., and Taylor, M. M. Analysis of Congestion in an 

Out-patient Clinic. Operations Research Division, The Johns Hopkins Hospital, Baltimore, 1959. 

97. Feldstein, M. S. Choice of technique in the public sector: A simplification. Economic Journal 

80(320): 985-990, 1970. 

98. National Academy of Engineering. A Study of Technology Assessment, pp. 96-104. Committee 

on Science and Astronautics, U.S. House of Representatives. U.S. Government Pinting Office, 


99. Collen, M. F. Statement in Subcommittee on Health of the Elderly, Special Committee a.. 

Aging, U. S. Senate. Hearings on Detection and Prevention of Chroníc Disease Utilizing 

Multiphasic Health Screening Techniques, pp. 214-222. U.S. Government Printing Office, 


100. Collen, M. F. Periodic health examinations using an automated multitest laboratory. J.A.M.A. 

195(10): 830-833, 1966. 

101. Collen, M. F., Kidd, P. H., Feldman, R., and Cutler, J. L. Cost analysis of a multiphasic screening 

programu NewEngL J. Med. 280: 1043-1045, 1969. 

102. Collen, M. F., Feldman, R., Siegelaub, A. B., and Crawford, D. Dollar cost per positive test for 

automated multiphasic screening. New Engl J. Mcd 283: 459463, 1970. 

103. Collen, M. F. Automated multiphasic health testing: Implementation of a system. Hospitals 

45(5): 49-50, 56-58, 1971. 

104. Collen, M. F., Dales, L. G., Friedman, G. D., Flagle, C. D., Feldman, R., and Siegelaub, A.B. 

Multiphasic Checkup Evaluation Study: 4. Preliminary Cost-Benefit Analysis for Middle Aged 

Men. Medical Methods Research, Oakland, California, 1973 (processed). 

105. Garfield, S. R. The delivery of medical care. Sci. Am- 222(4): 15-23, 1970. 

106. Garfield, S. R. Multiphasic health testing and medical care as a right. New EngL J. Med 283(20): 

1087-1089, 1970. 

107. IngalUs, T. H., and Gordon, J. E. Periodic health' eamination, 1900-1965. Am J. Med Sci 

251(3): 123-140, 1966. 

108. Siegel, G. S. An American dilemma-The periodic .health examination. Arec. Environ. Health 

13(3): 292-295, 1966. 

109. Wade, L., Thorpe, J., Elias, T., and Bock, G. Ae periodic health exauinationa worthwhile? Ann 

Intern Med 56(1): 81-93, 1962. 

110. McKeown, T. Validation of screening proceduxea In Screening in Medical Car: Revivcing the 

Evidence, pp. 1-13. Oxford University Presa, London, 1968. 

111. Garfield, S. R. A look at the economics of medical cart. In Technology and Health Care Systems 

in ¡he 1980's, edited by M. F. Collen, pp. 169-175. National Center for Health Services Researi 

and Development, Washington, D.C., 1973. 

112. Feldatein, P. J. Research on the demand for health services. Milbank Mem Fund Q. 44 (3, part 

2): 128-162, 1966. 

113. Blumberg, M. S. Evaluating health screening procedurea. Operations Researc 5(3): 351-360, 

1957. i 

114. Thornet, R. R., and Remein, Q. I. Principles andProcedures in the Evaluaon ofScreeningfr 

Dialuer. U.S. Govermnent Printing Office, Wasuington, D.C., 1971. 

115. Report of the AMHTS Advisory Committee to the National Center for Health Services Reteuh 

and Development, VoL 1, pp. 29, 31. U.S. Government Printing Office, Washington, D.C., 1970. 

116. Pole, 1. D. Economic aspects of screening foz disea~ In Screening in Medical Care: Revicwa~ 

the Evidence, pp. 141-157. Oxford University Pres, London, 1968. 

117. Suchman, E. A. Evaluative Research, p. 65. Ruseln Sage Foundation, New York, 1967.

- 377 - 

i aSll a _ . b, CuaMu L. L, PFda. R., Sb,_m A. L3., a Caunp IL. Fddmm, aG D., 

Dae, L. md Clbn, L F. KYF .Pemmamtu MYdtphM Sde ¡l h O * up Ei En uabi h jct: 

2. D~y uad COm Dioe~ lt. Snm Yo d uipe th ~ ~ P~ 

op wmd at dt Eplmidelm Scdml A a P H h Amo~d,, Octbas 11, 

1971 (po~sud). 

119. EaLa, A., ad Re~ R TRhnolalcal *tiharu to ioci cdum Sdmc. 175: 31-38, 

Jum 7, 1972. 

120 Wiav~y. A. Ibo pdiicd ecomo~y of Bcy: CoobenfIt mul]M, md proman 

buptin& Púbh Adimhkbwld 2 Rep~ 26(4): 292310, 1966. 

12L WUdavky, A., ad Ham~P A. Com~p e sn inaental budgtlst in tho 

depatmut agiculmm. Admmilbtw Scda~c Q~urter 10(3): 321-346, 1965. 

122. ndblo, C. E De d=oin* u in¡a btxaotn mud expndituf la Puabc Fa~es: Na~e, 

Saomce md Utatb, editad by J. .L Bucdiman pp. 295-329. Pdn~cton Univmity Pre 

Pdnton, N.J., 1961. 

Diect reprint requerts to: 

Dr. Herbert E. Klanmu 

Professor of Economica 

New York University 

Graduate School of Public Administration 

4 Washington Square North 

New York, New York 10003 

MaJ serpt mAbndttd for pubilction, AprT 12, 1973

#22 

MAED&L CÂE 

Nowaenrber 1960. Vol. XVIII. No. 1 

- 378- 

Original Articles 

Cost-Benefit and Cost-Effectiveness Analysis 

in Health Care 

Growth and Coinpositiont of the Literature 

KENNETH E. WARNER, PH.D..,' AND REBFCCA C. HUTTrON, M.H.S.A., M.A.A.E.f 

Concern ab the etaclaing cost d health services is refitted in the rapid 

growth othe litehure on cod-benefit snd eod-effcctivenb M nailysis (CBA find 

CEA, resctfvely) in heialthd ce. A earch d dui litcrure Fr 1966-7811 produed 

a bibliogphy d more ithi 5l releva reerenees, gowing fom halfa 

dozen per year ti die beginaing d the period to dowe to 100 cach a thd mot 

recent 2 years. Th litture ~owth has been more rapid in mcedil than 

nonmedical journals and a prcference lur CEA over CBA appearn to be emrgin& 

Studies related to da uIs mid tre menl have uined in popularity, while 

the euly pominence d udies with a substantive prevention dtheme has diminirshd 

Consistent with the nre~asin medical focu of the literature, num- 

.. I L' aticles aoented towwd individual pr~tioner decision maing have 

apr».. mare nipidly than tho e oried towd o ni arl or soeieta decision 

maidns. In addition to docurrenting dthese nd this atide identifies 

published reviews of health care CIA/CEA ad boois and utides oteeti~nti 

to cenvey the prineiples r CBA/CEA to the health care community. nIe liele 

ncludes with specul·tion oe likeliy ner4ure trend in the iteatiure nd 

consideration o the quality implictions o the rpid rowth. 

CAÁNCERN alhimt time high and risitlg e)stN .rf 

health care has pronipted a wide varneil ,f, 

eost--ot 1 l ¡o 10 Ilt,t , ~ 1I l * J H l.tImtll,,f.t :.. 

efl.letive and under what crecumstances, 

arid tíhei to *c mpare alternative interventiuns 

iii terrms of their ostliness. to assess 

their relati e c(st f-feutiiveness. The task is 

a monunimierital one; determiinng effectiveshess 

is itself immensely difflieult.' 

To identibf and convey the meaning of 

os

WAHRNER ANI) I|UTTr)N 

The puroses In of this article are lto descriltw safety, eínvironmental protection and cer- 

thi growthl and i)mpositio uíof thie health tain aspecits of individual health behavior 

care CBA/CEA literature, to slpl)¡a~ia¢s tli. t 

C(ost-lietlefit alid k íst-elecctienless 

.i.. sis hiave 'omlae t, r.efer to fiormal 

ali.l tl.iiatixe t(.se of resoed iza aniy 

.irnim;tlaer ¡' ;im >l)l¡.atiol.s. th e ,,il.x (ll i, ;!iir dI 

ii(tll(n lalt.die' al p ml)li.ati(ial s(.la i i.hs. au 

HlltJa ol thi't.i t

"'ol. XVIII, Nu. II1 

I. S ,Si 4.. l .. I.ib.111 1. l t . ."i ,.Isii.§a !l g .t . 

..Ts ,re' -',t 5 IIt.1LA 

- 380 - 

t II ¡Ii: l1li¡ ,illltlin, '. li chl -" ' 1 *, s et J,}. ".f' . 

or .*5 .d .ah- fI..ie sl. 4oJ t.. r1¡l . ",".íilh e ". ,. 

(:CST.BENEFIT AND) COST-EFFECTIVENESS 

cess of benefits over costs? Because ,il 

costs and benefits are measured in the 

same (inonetary) unit, CBA can lie 

employed to compare similar or widely divergent 

types of programs. Thus. in theory 

at least, CBA might be used to decide 

whether publie resourees should be allocated 

to construction of a dam or constniction 

of a hospital. 

In CEA, desirable program consequen¢es 

are not valued in monetary tenni, 

but rather arte measured in some other unit; 

in health care CEAs. cominon nmeasures 

include years of life saved and days ol' inorbidity 

or disahility avoided. The reasoní lor 

a iaoninoetariy mcasure of l prgranil tieLctiveuness 

is .. ither the in)p.ossil)ilit or .uudesirability 

of valuing imnportaiit onitel mei. e 

in dollars. Thus. thte " lottoiI h¡m." a 

CEA is not. lilke a C(BA. a nel nlt i.la'Lry 

value; rather. it is expretssed iii Uiaits liike 

"dollars per year of lilf saved." CEA ipermits 

eo'niparison of e.ost per unil (itf eot.leetivriness 

anolng conilx'ting progralla alte'rnat1; 

.. .1, -. 'ifd to serne the salie I)asi(' 

purpose, biut unlikel CBA it des inot allou 

em)lllaristil fe I)r prgrasail, ¡lailag dliltere(tlt 

oDieotives. I)ca(.ise thle etli etlr'.lv', , , r 

eouitenilue uaí.'eieres ditl'r. Nor dI.'s it I¡. riit 

assessmiient of t Il. inhil, renit .rhl nl .ifa 

progranli: is a c .si ol'S($.5O().0r(J >t'r or lih-. 

sas;ved auteltal)le? ()lvioisly. tlis ilue · - 

tion reqjlires a sowial iind piolitivall jmdg- 

Ilmena:; it i ¡¡iot sililyll .h lic.il ..ner. 

(:BA andl (:iEA,:I lJa I 'v'tl sti.iedl ¡lr 

dlev-ad.(%. b)y cntrib)utions 

appeared in print, with the exeeption 

of a chapter by Grosse4 in a 1970 

book n.;iented toward stiidents of 

econoumiies and policy analysis. This chapter 

is parti':ularly noteworthy Icr its reiew 

ofí C:A/CEA applications in tll,. Departmeni 

of l.ahlth, Education, antd 'elll'are 

(HEW) during the author's tc, uure as all 

HEW fllicial. Grosse cun)eyed miiIlh of 

thi' samle lmaterial in ani article publishedl 2 

years later/ though agairi tie audience was 

o¡it heallth care p)rnt'ssioaiais. That same 

year. houweer. witnessed p>ulblicatioi of a 

h),>,k which I )h'anie one of the health care 

c milsi¡"íity's mniost widely read asd fre- 

(lueIntl cited contributions, Coxhrane's 

Effe¿tlrm ene*s url EI:ffirideny: Randonr Reflie 

icS,'i * o> Ilulthl Ser.tit'(e." lil o)lur e(StirIllntlll 

this short IJxx)k hadl a i)roil(aimd inpa¡lt 

ii tlanling the' thouuglhts olf hiealtia care 

prol.s.aicinalis toward issties of resource 

scareity and thie link ewtweeln eflicieley 

anlid vq'ily. It is at least possilleh that (:Kehti.me'.s 

I>,c>k pliared a sigilgniftiat rle iii til' 

ralpid grousth irn healti cvare (:BA/(:IEA 

wh lvla I('ganm tile -fcitoewiig yvar. 

I¡i !971J. Klarmlan pul)lished twu- articlehs 

.ia CHA . iethodlioh,el. . and al)pllicationis. 

()1 ' 1,s fil-' Iist a tti.n cit(led re- it.' al.Ld l ;s- 

( ,%-i>ii el th'athh -. care (:A/(:I'EA'. li.he 

,pll¢'r ;,vs tli'. firt se,1 ii .prtpll ' let al)i)¢e-11 il 

.1 Ilmedlí..l lOlill.icl I 'lI' h teolissin ye'air. 

1),IhI. ", I¡ hhsh' .,I t

WARNER AND HUTfON 

- 381 - 

observers. this Journal issue stands as a 

landmark in the evaluation of medical 

practice. 

Two years later, another issut ,,f the New 

EnglandJournal of /Medicine offered readers 

a discussiop of CEA methodology." a 

, iphisticated application of it,' and an i.n- 

portant, thoughtful treatbnent of the limitations 

ol formal analysis.?' Many health services 

researchers adjudge this package. 

1omnhined with Weinstein and Stason's 

b>ook on hypertension policy,? to be a 

smilestone in health care CBAICEA. At the 

end of that same year. in Deemilber 1977. 

the Arthur D. Littie Company euinpleted 

an "Iiatroduction to Cost-Bene'`t Analysis 

Applied ta, New Healthl TechnMlogies."`' 

This duocumiment, intended to assist health 

plaanners and otliers, was prepared under a 

tvimtract with the 3Bureau of Hiealth Planitling 

tand Heoiíir.t D'v'elO>hpilctnt ¡in tihe 

llealth Restaurces AdministrUtion. 1fiEW. 

In December 1978at the Url>ian Ilnititute 

(:onflfren'er .)!-te t.lllrm. 

F!.i i l ,. ir . m I..eml, \V.mirnll .ímd, 

i.,., 

' 

)l(' i '.11 il( 1 I lJli)(Jg' ,1 (:iv.¡lk ' 

(:1i: -\. a. i .me .mmel ld II, ;.1 .li-cl )% i(ln iii 

I)t,. le llti l , Im

Vol. XVll. No. 11 

- 382 - 

(:()ST-BE:NEFIT AND COST-EFFECIIVENESS 

Development of Health Care CBA/CEA 

Biblioglphy and Clasification 

o Refereces 

The sucmeeding two seetion.s present an 

empirical andalysis of the growth and content 

of the health care CBA/CEA literature. 

The analysis derives from counts and celassiRication 

of more than 500 referentes in a 

bibliogra'., 't uovering the period 1966 

through 1197. atnd including CBAs and 

CEAs on persoraial health sernies topíies, 

reviews and conimnents on such literature, 

asd dis.cussions ofCBA/CEA iinethodology 

direted spl ,'ifically to health care professionals. 

Excluded fromn tle t il)liography 

arte s'ores of CBA/CEAs on Iialth-rn-le- 

.'ant It ,nomiuedical suibjtcts (..g., Irafíic 

·%alt'D' anad toitroll o>f t'n viilta uirit';t.íI l, 

hlltiak,.% ald ari- 

*l'hs 0it (:BAi(:t:A Il

.AHNF:R AND IIl'TIoN 

- 383 - 

rare I)rolssioBml:d' introdmuction tt (CBH.: 

CEA, w. lilt it ai)i)>npri;ate tl, inicudr. i. 

articles whichl purport to relate to thi¡ 

thelme, and not only those whitl. ChB?/ 

(:EA exlxurtb. wultld clracterize a.. teciiirally 

soíund.§ 

1',ll.hw u)ilag g comptiom ol' ihe Iiliiiogr.qalhy, 

w(.i lassified e achi relerenm.e aceording 

ti thec fi)lohing (lini, ,L~is.!: 

tI) Yt-ar ( 1966-1¡978) 

i2 T)IWm (C(:A.(:EA. g sr 'ral lnikiim,)iwl,) 

; l>' Il,hli s.ti( , :t'llitl.< (iiclt;r.dt juinm.ul. 

plirlhil. ilitenscl'.l ' ri a 14.r ¡ril oMM- 

itr.lthr%, *Ar Ihalth >en ic. r, n.rt s lI(.I'. 

*ssIsIlí*¡lli:líl j(luníl. sítl,' r> 

'i i' li.iíl I. iHtII'tJiílj .l jrr.sílUll rl h. 11h1..l- 

Vo'1a. XViii., ,. 11 

F'i;.. I (;r,'stl ,1 lih-Jtl 

I. 

r.'.*- (:IHA'!(:I l t MrJars. 

ir 

a. 

tiA 

o 

z 

0 

1002 

90. so' 

80 

70' 

60' 

50. 

40' 

30, 

20' 

1o. 

total la.% alu.ys 'xv('dc eJ 88i. (See' als. 

Tal)h. 1.)o 

,aaa1iat 'l',. isa ¡s *'aXaa'a si- íl. t I tv .l . l tu all.tít.I Ia'saalt . a a sl. . tla ra ! st. al Th, W . ', 

msa ssa.aaaí«4.. I ,saaaaal 

lit'd arJ*luah. , i.l.i~,i 

.ll tall 1.a,sas mí. 

Trd 

M. al' adll.. I i ..,'t ía 11 4< la a a t'ra lst ls ta a' l ,l'l' NL't- 'la, 

taslal, asasa., t l.ia SII.ss u.Atlaat rsí,alat as fas. 

¡,I aasrl s. z l . w _ "E' lsl. t ali'$ f il* *I it'afais -fh as' g 'ta a .a Iu al 

"l.. I. ,,ith (al,' ta E'F t4 < ¡A lit¢-fdniaa' y. 1(fh tid t ith l ' 

A,. ,tl ,,aa,,l., 4.- ldta , ,s,~ Ih a, a'l, u, EhJ.l 

la gt 't í- ,sai- st'la.'l .tl, ttl.t ls.g¡ - I' ,u a.. t. 

, 9* ,l 1ala (a t:t, dlfh :at1., , 1 . ., a m , § .a..,:,. ;,t a sta. , , a aa ,, tu -le 

d , ;... ,a -is 3>* ltal is ,> . ,. ,, 

...... a»esas.¡ Ih4. tía.' - E 'hb . a < es - *. líE. .11 F... 2t.atua Ul' h'f .r tKrtu w , gi... las A I.1 l^ti. 4,,m l .§ , st 

:, ." tlt.. í1 F'% lstat aasa s t . t- tsa .- ,t'.t aaat laat li. 

I.dl:.'1 a.ía. , ."'ta'aa''.a.aial ,o,,~,,,'f .h ,,1111i.i. a*talí... ~.~,-th as% ia..a ,,( a.- l,. ti (la. ,, l ,'s, 1 ~Jl ' ailí t.1 ",. .. .1. 

* al- § | .s a.! a. , , í . t . E" ¡ ' . E j t - a-1 

.. .|. .§ ..,1tl.l.*.. .. l.. tís H 1 4:. t .: . -, 

- 384 - 

CO~ST-BENE:FIT AND :)r-)ÈTI'S6 

CBAs, 

CEAs 

-66 F7 68 69 70 7172 73 74 75 76 77 78 

YEAR

WARNER AND HUTTrU., 

TABLE 1. Trenda i¡m Health Care 

CBA/CEA, 1966-1973 and 1974-19786 

66~- 1974- 

1973 1978 

I Avcge~ annual number of 

publictions 17.0 73.2 

2. Publictiorns in meditcI joumrii 

as of total joumal publications 40.2 62.7 

3 (:EA, as e odf (:EA + CBAs 42.1 53.2 

I. if art;icle> o0,. 

Preventuion 44.7 22.0 

D)iig"".,ul> 18.8 30.9 

TI'u'tl*lllt 36 5 47.2 

5. 'c f aritl..s with OrwellmtiiOll ol. 

Individual .3 I - 

(Ogaiiu.luizaiuu 21.3 IO.8 

Not'Jet) (1.04 7 3.4 

· All dill.. .... h rI33li('.3itit ap ' 0.05 

- 385 - 

IWO)s seei liikely to !raitotc rII,.0IiI? \ I " .' l 

'..ll Al/l'll It 'l i '1 " 1 . Il( r '. ""l t.,' si. l I , .% 

ll i,..' 's.l..,- í i . . f.f 5 .f ,'uit I, .1....,* ,,l, it,. 

114 Jt'IlINII~.i~ IllJ j J~ 14- . 'Il I* J,. ~t-I. t u' .'I 

&HIJ&llllllJ.hlr ristmJ~l .J... .JJ 

I *,' l .till 1 .' [l ..... , " it l' I % I.. t ' 4. . it x1 ..i. 

Publication Vehicek 

MEDICAL CARE 

Figure 2 plots the annual numbers of 

articles in medical and nonmedical journals. 

The purpose is to examine the proportion 

of the literature which has been intended 

primarily for a physician audience 

anid how this propoirtion has changed over 

time. While the time paths follow each 

other *losely, the graph shows a shift from a 

rough parity prior to 1973 to a clear majority 

of medical journal articles after 1973. In 

other words, the rate of growth of the iedical 

literature 1 -s exceeded that of tite 

nonmedical journal literature. particularly 

in receait years. This shift is statistically 

significant (p = 0.05). (See Table 1.) 

In c-ategorizing references by publication 

vehicle, we kept track of a subset of 

medical articles, nantely those published 

in the New Eingla.id Journwl of .Medicirie. 

\\'e isolated theste articles becatuse of tihe 

Jluiri;Il'.s ge'neral )lsitioín of leadershilp in 

thie ieadical literature anad because several 

uf' the est and uiost influential health care 

(:CA/C:EAs have heenu pullished in the 

Jtur«ul. t i is iiitresting to obserne that 

htbId)re 1975, tice miiii)er of CBA/CEArehlvant 

contrihitions in tle Journal excee 

(ed One only oc. t (in 1968, when two 

c)untril>)itionis were idenltified). TheJ,)ulrual 

pi>u)lishled sere.í relevalit articles ini 

1975 antd f;oiur or muior edth ticI. y e.a r 

MNii of (:BAs and CEAs 

Prior to the inost recelit years. the amiuital 

>umibMe.r ,f (:BAs genernilly exceede(d the 

iiuiilher ofl' CEAs. Since the tnid- 1970s th,. 

r(.vírst- lh.as tn'( tn . Tie ei. mix ii th.u, 

*."tt. i!Sc '1'.ille I.)'lis pro rLs i s., tl)pprt 

¡,r Wrm3tNteti,'I

Vol. XVIII. No. II 

FIC. 2. Dilfuslun of 

(:BA/CEAs by type uf 

>,umaJ 

w 

u. 

o 

z 

- 386 - 

s'aluing Ixneftits in CBAs. u 

,maniplemeuntary 

or altemative explanations relate to tihe 

appaLrent relative tonceptuai simplicity ou 

CEA: analysts use CEA blcauobe it is ersier 

for the econormic layperson-c.g., the 

physit'ian--to understand; also. the receilt 

relative growth in dtie literature in mnedi¡:al 

journals appears.to incliude relatively niore 

contributions by physicilaos who. as 

ectonomie Iayperso

W'AHNER AND IHUTTON 

- 387 - 

treatment, compared wíth prevention, than 

are nonphysician health professionals (including 

xboth providers and health servites 

researchers). Also, consistent uith ihe 

principal early non-health care applications 

of CBAICEA, early healtih care 

CBA/CEAs toncentrated relatively rnore 

oi, health care "public goods," including 

especially commininicable disease .ontrol., 

rather than individual patient care, a gromwin! 

t nce ner, today. Several excellent cornmiiiiinicable 

disease prevenitioo studies are 

iliid iti thle recent medical literature, 4 , " ' 

l,>it this ii ne of the fw . substantive areas 

ifl wli,). , the )(llll'¡ .of ¡>re-1974 Ilnpers 

aIL' all y ec

Vol. XVIII, No. 11 

analyses whose decision-asissting pers 

pective is narower dthan that of"society." 

Our data suggeCt iat the social perspective 

has dominated the literature over the 

entire period studied, accounting for 

roughly 70 per cent of all publications in 

both the early and more recent years. 

However, articles oriented toward individual 

(e.g., practitioner) decision making 

have increased most rapidly in recent 

years. Comparng the pre- 1974 period with 

the years 1974 through 1978. one observes 

a near doubling of the share of papers 

oriented toward the individual perspet'tive. 

This growth has come at the expense 

of papers with an organizational orientation. 

While the two categories together accnunt 

for fewer than 30 per cent of the 

literature contribuitions. the shiift is statisticaily 

significant. (See Table 1.) 

ii,. 1i t, " t f 1. '... t. h's . 'i. Iho .. 

i id íl , í,l . l, l,',*s 

- 388 - 

COST-BENEFIT AND COST-EFFECTIVENESS 

have addressed mental illness probhlnms 

and programs. with the same number of 

citations relating to dental care. Together. 

drug abuse and alcoholism ac.ount for a 

similar nUmhber of references. Renal discase 

has received an amount of attention 

(almost 20 papers) dispmportionate to its 

health importance but reflective of the 

political and economic importance associated 

with public funding of dialysis. 

The federal government's mid-1960s 

interest in disease conitrol programs. and in 

kidney discase in particular, uande this the 

only disease problem to liave more than 

one citation before 1969. 

Two general classes of healh problesIas 

have captured con)siderable attention. A 

ariety of commnunicable diseases have 

>.é-en the subject of more than two dozen 

papers. Since cmrmmunical)le disease and 

Its treatirient have distinet "public goods-" 

Substantive Topics and Areas of characteristics," this is a logical subject of 

Interest in Health Care CBA/CEA§§ CBA/CEA. and it is not twoo surprising to 

find that ialf of all th«ie cnimmlníicalble dis- 

The he-: t. r.- C(BA/CEA literatuire 

eaM,% Iai.¡ rs date from ti Mfire 1974. By ion- 

icovers a ví'as array of.liseas, irol>l)his atad. 

trast. tlit' Set:iiid class 

as 

of 

indicated 

pIrolblens-the 

iii the Ireviou. ert.ioit. all iot 

I)rt'si tilima ofi hrtil, tfeerts.--hias 

the 

I>)een 

major mendical finiatiois. Yel despit(: 

st.iditl nmaiach imotrt ill reetmllt 

the diversity. 

)'yars. 

a 

w¡ith 

fe.w sUiilecas aind tolteer.as 

otnl 2 of 15 ipai'rs predlatinlg 1975. 

acoaunt 

Se. e.ral 

for a large share f the liiteratare. 

hirthi deliet disease prohienis 

The 

liave 

single 

redisease 

e(lass whi.h has cap> 

*cci

AAHNER AND HU'TON 

- 389 - 

Several disease problenis emferge in the 

guise of surgeriec intended to treat them. 

Each of the following surgeries is the focal 

point of at least one reference in the bibliography: 

radical cystectonmy, tonsilleetomy, 

choleeystectomy, herniorrhaphy, 

appendectomy,, synovectom. joint replacemnent 

and hysterectomny. In addition, 

there is an equal number of paper, relevant 

to surgery and CBAICEA, but not identifiable 

with a specific surgery. Many of the 

surgery papers were contributions to a recent 

hook on the subject." 

We classified some two dozen papers as 

nonspecifi screening and prevention. 

Sorne of these related to particular acti 

ities (e g., multiphasic screening). while 

others discussed CBA/CEA issues rmutnr 

generally. A few represented attenmpis to 

cover several separatt activities. 

In recent year. a grt-at ded;l ih>lsliq d 

a(l¡liitinaol ftíum I)apl'rs oun i.obitutiotinal der- 

SIIl hlillm - (cart., with th(- Ip.itient typl( niot 

i¡, 1977 aind 1978. 

Inid('amtool (' ;i -li c'eui'nt )rohlh'lems .u1(1 .1id - 

\'Will otl.r cfitiril)u(. t (,,,l...,ld 1..('hn 

,Ilgi,.' 'mn.rge ans tihe 'ulit

Vol., XVIII. No. 11 

lustrate another instance in which technical 

innovations-often, in this case, substitution 

of one type of personnel for 

another-produce outcomes which are difficult 

to quantify usefully. Nevertheless, 

analysts have made a dozen contributions 

on this subject. 

We noted previously a couple oi conspicuous 

absences from the literature: certain 

major disease problems have received 

very limited CBA/CEA attention, diabetes 

and cancer therapies serving as good 

examples; and CBA/CEAs specific to 

sophisticated capital equipment are less 

common than policy interest might lead 

one to suispect, though, again. their frequency 

may increase due to both growth in 

interest and supp-.rt of analysis. 

Related to dearth of equipinmetsplcific 

stuiies, apart from w.retning procediures 

relatively few diagnostic procedures 

hav¿e been the subject of CBA/CEA 

attention. A few pro'edures have received 

isolated dis'a.si ,- hut onily radiology lhas 

received :; *;. : .- alentiou. Wein.stein" 

has iientified the evaluati.ni of diagnostic 

proeedures as deserving of CBA/CEA el- 

- 390 - 

COST-BENEFIT AND COST-EFFECTIVENESS 

and, within the social sciences, numí.rous 

analyses of niedical technical change. Yet 

"side from implicit and tangential interest 

in them (e.g., as a component of hypertension 

management), drugs have not often 

concemed CBA/CEA analysts. 

Finally, the literature reveals very little 

evidence of attempts to compare the costs 

and benefits (effectiveness) of specific 

medical and nonmnedical interventions to 

deal with health problems. While our bialiography 

search focused on medical ap 

proaches, one might have 2nticipated identification 

of a few studies which cross the 

nmedical-nonmedical border. Yet with the 

exception of the early HEW efforts- ' 7 tlhe 

principal mechanismn for crossing that borler 

is the reader's location and comparison 

.,separate. independent studies. For botd 

conceptual and empirical reasons, this is 

mnt a highly rewarding analytical strategy. 

Conceivably, heightened awareness of 

prevention altematives will motivate formral 

efforts to grapple with medicali¡ionaiedical 

coiilaparisofis iii the filture. 

fLrts. ¡is¡ plel i¡s supported 1»y the growing 

Conclusion 

boJdy of literature that indiíts the inrceasing 

use of diagnostic tests as a Inajor source 

of medical (osnt inflatioi,. T'he evidente 

suggests that everydayt . mnaiindane ttsts are 

at least as significanit cwotril>utors to that, 

inflation as the more sophistiuated aid exspensihe 

tec.hnbI>golo es.u yet the fomner 

tiave recreived í rry littlh CBA/CEA attention.. 

Agaira. i)rol)>lhis f mita.tingiK asid 

vahlmiusg tilh (1t(teollt-% i >diaglultstit Ip)r< ('dures 

st.ied ¡i¡ the w ía.' r)fr.ath app)lication 

' 

of1 (BA./(:EA. 4," 

liti (,1>llBgtlias ec(tioian. A í (. .)t* t ) otfi er 

ar,.. 1s wJ m.h %í.í¡ f tnle. rrE1í.I)t ,'#te- ¡ . th, e 

Ite rat.írta . I r it 14- l-st .\' ral ih'.,.it.-N. 

Reilevtking ec>ncern with the high ard rising 

costs (of personal health senrices. the 

health care CBA/CEA literature has grown 

* rapidlly in recent years. The social miliieu 

seems likely to encourage sustained interest 

in controlling the costs ofclare; with that 

ninterest should c)iyie continuing growth in 

analysts' efforts to gripple with issues of 

List effc'tiveness. 

lThe relative growth in medical journals' 

share of the literatura reflects increasing 

professioiUal concetrn abut the spiraling 

.ustNs of health snrvices: medical professioniAs 

are exprt u.iiig the.ir interest in the 

IS3m-; alld nlnonmedical analysts are exhihit- 

, illIL g y.O , .ll( si'lt'lltifhI 

fr, l tlh l i1 iii l t iit'l' ll'- .a11i¡ (h.lí.íhí .!it .alt're.(d 

. ,1. s )r.u'tlt' . d.i ()11tt'l)111-%. fi ,iJ lh-alth 

(ir.i- I)rttgs liit ( Ix.es" thd. oh i"Is . (lt,,- l- t 

dr.,;., otii I)i,'l;rsm l. .i. íaád 1 , da, . l tI¡i u.'. 

ng a d eu'il' tI, wo(rk with physiciaris to ai;- 

'% Irt ( gi(. vttrtit l i health care 

i.AI/(:EA% iii ti>.* medical literature has 

}l.sl a litIllilxi'I of railnficathinas: stuidies with 

.i (hicl>d1E)%1 I. tr' 'at>ljnIit f(.'us--the princi-

WARNER AND HUTTON 

pal orientations of physicians-have increased 

much more rapidly than those with 

a prevention theme; consistent with the 

increasing emphasis on personal health 

services. analyses oriented toward individual 

practitioner decision making appear 

lo be growing more rapidly tha. those with 

an orientation toward social or organizational 

decision malking; CEA, a more readily 

comprehensible technique, has come to 

dominate CBA. 

Some of the recent trends seem likely to 

ixprsist in the near fuiture, while others are 

more prohlematic We expect continued 

growth in the medical joumals' share of 

the literature. through increasing interest in 

CBA/CEA among nonphysician health 

professionals might restriet such relative 

growth in the physician-orienited literature. 

In any case. we would anticipate intensification 

of the trend tuward concern 

with individual practitioner decision making 

and increasing use of CEA relative to 

CBA. The r" .'1 n diagiosi s ¡;id treatmi('nt 

orien.it.is. . ;h..y' #)r IllI¡¡¡V nit ik' St1statined. 

Interest in thie .t Il spl)e ifice 

calpital-inteiisive inedical t'ehtu)logies 

shotild prinmiote a diagRosis n¿d treatiment 

ioctus, I)l!t the eIjerging, conven. COir healthl 

proaoftimn atndl disease IprT.iviint.>in cill 

re-verse the relative decvline iii ClBA/CEA 

interest in prevention. 

(;ro , ial'. .. 

h.s ' 1I;'ll gr.,sl)Jglill; ser.cyitss. witil (hi- 

¡,. 1 It (: IA/(:I :A ¡ruíilch.ils ii ¡t tie 'uh.t .bi 

¡im..r I'i In, lt}l í.. rí ..¡q¡lií .tlhm.. A. .. tila- 

, I 1,, l(,. ,.iitl, lotl.iit..I IbrI.tslz,' le. d1 

1 iil. 

- 391 - 

t l . tu li.Is t i ¡oqt , 11¡il e1¡1.1u . 1 1 . ti. 1 

n,.lIti-..ttrih,mmha misil>le tir delivering 

health senrvices. By eonitrast. sul:perior 

anjlyt'.s nmay achieve little impact .ss a const.llation 

of tiactors-et)nomnie incentivei.s 

politicatl concernas, etc.-itiay domiiiriate 

('Inhi(lh'r¿ationls ,)f sl('itl ('(ost-efe('z'tix 

('tl('N', ' 

Actknowledrgment 

l I.I>. .h. u,di.,tt " '1'"- *,, .i.,ieg, d*I.. . *..! t Lh . thlm. 1.. 

I .fl1. ll t . *il *" J'li,'l' * ' rt,dil ,iZ: r l II ta l.tll 

'., 1 " i I,, 1 I"tlh' ., , d%'.I '1,.. - , - I.,¡r .-- l.¡l' - .re dl 

I. -l. r1. Iv .ld 

MWciCAL. CARE 

I \ \..... , i i i . i t * st l 5. ", i< 

1t - y k .- I-- 11 "1 --l ,,, I." ,! " , ;. ' ..

SVol XVIII. l '>. . II 

:;Ei ¡ y7i 

1. 1 \\ 1,',i.¡ cc, \. h. Fh,-. .,Ie Ic,,'h . .. P...sl.,-, I.l- 

e, . ... h ... . ,e- I I .I1eie l . i.l'*. 1 , .1 - * ' . 

*'1 , I lIt 2,!111 

- 392 - 

C(:ST-BENEFI r AND C()ST-EFFECTIVENESS 

tiU. nd pietentd. Ann Agbtr: H«Iilth 4dn#m.í,atnduei 

Ptess (uidIiming). 

lr 

2. Congrtas tof thhe United States. iNlSI.F% 1 ¡cec ( í) l .ce (te c ,1ren> iill 

(1321ii¡. 

1 ' 4 7 eIIee11 c': le -1e1 ,aIe) I9els, ss eled) 3.3'Jlth ( : e'. 

%1.,l .1 hce l.I. . ,,-'. e .. lee Ilse It.áltl caf.' '..-- '!" e 

N%\ .¡ec.i? .e i alhl.s, c, 1 i'' i l- ce ..1 t -S) ."' 19 t9 

h e ,h e , I · %1, ,he , e [,, h, e.c¡ .

w'ARNER AND UL'TTOit'u ' . N Engl I .MIl- 176f. 

2L"a:,0I6. 

42. .htuw-nl.nam, S. Mc.Pil B. KaNP t J. iThe s-io ¡¡unJllenzd 

det ision. N i:ngl J Med 1976i.25:759 

43. Ilealhhy peouple:: h' S.irgeml (¢.i.ral', nr.i)rt 

i llhealthi h pru)iolils 1 i, ld dismdaít Ipr.'eiulionu. 

%\%,.lígt(m I) (':.: t1 .. (u>%-n.mniei Pn¡ltislg ()lt' e-. 

1979. (IIIEW $piallihiatulm eit. (PUiN 7W.55071 

44. Buiuike-r J. iar.ec D. .t(iiít-lvr) F. ' l*d. (:isit% 

riíLs .uid loi-it .. (ll't. tal ug'.r N ,,rk .t ()Oid,,r lu',i- 

4s. .%it. Pne 1977. ¥ A. 

4ti SIu'AL

# 23 

Morgan N. Jickson 

Jarnie, P. .>oGierfo 

Paula i)ichr 

Carolyn A. Watit 

Williamni Richardson 

- 394 - 

Hylv,,erecinomn t% one of ihe mol I'requentll 

performed major surgical procedurec in ¡he 

United State,, tloda)' l he facl thal incidence 

rate% var, t.onidcrahly heit4een l*ngland and 

the Uniled Stales." and vary even wilhin the 

iamne geographic area of the U .S .:' ha caused 

1i Io be ciled a, an example of a procedure 

which is ofien performed unneee.e,.armly. Although 

the surgery obviale,, . woatiitn'\ concern 

abhult he risks of an unplanned pregnanyv 

or cancer of ihe cervix or uterus and freec 

her from nienstruaion. it may expoNe her unneceNrily 

Iu i .¡.o:. , ,r nl..idlil) annd morhidity 

from the %u,.gery or it, complication,. 

Thi, study wUs undertaken ¡o an;^d/e the direil 

coti and direel hbnefti of electlive sirmple 

hytercciomy. Cosl of aliernaltive melhod, of 

contraception and their compliciilotn,, v cre 

not included in the analysi,. 

Rice4- ha, invetigated ¡he determination ol' 

indirect io.st of ilne,%. However. an ana.l) >1 

of elective hyntercctomy uNing direct co,,t. 

i.e.. charge, paid for inpatient and amhulaltor> 

ecrvices. hae not been reportedJ 11,, ui,,iumed 

thai .uch :a report 'would provide informatiton 

uNelful in making the deiiimon au %u twhether ur, 

nol the proccdltire should he d.loul.ged. 

Anal> /n onlg direct i cor i ,¡void' the conirooer>) 

%urrounding an appropriaite .uiIluti(n of 

1. ,*hr." lfi.1 f, S, holio .0. ,hc ?!i r%1 s '3 4 .¡*hllí,lq. 

,il Il| P I ., :. I .I mc.. ln ( Ih.t'11o .% | i ' t# J.nN.e 

P. I.o4.erli. 1.)., M P.H.. 1s .e-sIJn I 't.f,:,'O I 

Hct.lih '\1ca:i , .&aid A. .! It ,*.(% .1 i.i 

Pha thlr. Pbil). i, A,..ss.fnl I>'"i*. ) sM,1 l .t ~ l.sll 

(.' rdI Ak. '1, 1 . II.I>., i' A'ssíl.,nl P,.WI.'., 

~"rl,)l,:~ ..f',. k', s1.dA\ 1 I'ihP>, ', ,,I ' 3 ,, ,,I,,, 

,. 1 ¡1. , ,lth 

4 %illiam 

.,"c ,-.h , h. . % 1,,.c1 .1 i%1lsh. tic..llh .red1 ,(.mmíH, ,l. 

( 

'd,-lh , 5(r.li~ , ··

I,,qr *ir 'ilu.,i XV. Seplteri¡br 1978 

- 395 - 

year-old woman is justified on the basis of 

known risks and benefits." Deane and Ulene' 

compared hystereclomy and tubal ligalion as 

alternative sterilizaltion strategies in )oung 

women who desired permanent contraception. 

Using decision analysis and cost-benefit analyi,. 

they found thal al interest rates atove 

5/r. tubal ligation would be the preferred stralegy 

for all women over 20 years old. since il 

would be less cosily in both mortality and dollar 

dimensions. Cole and Berlin" examined the 

benefis and cosIs of elective hysterectomy for 

women at age 35 years. Assuming a 0.06r'/ 

operative mortality rate and using rlhorecical 

cosi projections. the) concluded thal the opcration 

would increase life expectancy by 0.2 

years and Ihal ihe cosl of performing ihe operalion 

on one million women would exceed 

the benefil by $570 million when discounted 

to present value. They concluded that can. 

prophylaxis was not adequate justificaiion f. 

elective hysterectomy. 

Olher siudies have documented the adverse 

psychological sequelae' ° " ' thal may befall 

women undert .-,-,. .- sl:rectomy;: arker':' 

iuggested that the-;e were more likely lo be 

fomund in a woman in whom ihe procedure had 

been eleclive. The consensus of current literature 

concerning intangible considcralions favorn 

he notion that the cosis lo a woman of 

having an eleclive hysiereciomy outweigh thc 

benefils. 

Marib and Methods 

Data for the study were obtained from the Seattle 

Prepaid Health Care Project"' a fouryear 

wcial experiment designed lo evaluate 

the effect of provider financing and organization 

on patienl health status and utilization of 

medical services. Near-pIoor families residing 

in the Model Cilies larget arca of Seattle who 

met pecific income requircmcnt% were Clig,. 

ble lor the Project. The charge data (amount% 

paid io providers) were collected on paticnt% 

enrolled in a very comprehensive prcpaymcnt 

plan thai full) covered all medict:l servicc% 

provided by membcrs if a Physicians' Service 

Bureau (Kng C'ounty Med":aL'Blue Shieid) 

and community hospita.is The providers in. 

Cludedl alI couínunity hospital., pharmace%. 

and virtuall) all non-federal. n)n-.goup health 

coopcralive pihysiicians in King ('ouniv. 

WashingIon. There were no deductibles or 

other copayments. For ¡he services considered 

in this paper. and within the time period 

examined. information on amounts paid lo 

providers for inpatient and ambulatory medical 

services is essentially complete. 

The data base consisted of all paid charges 

thal occurred due lo the specified uterine-related 

gynecologic diseases listed in Table 1. 

These data were collecied by selecting information 

using three-digit diagnosis codes from 

ihe EiKghh Revi.sion. Internarional Clussificarion 

of Dise.ses. ' Some overlap occurred in 

the diagnoses between conditions which could 

be attribuied lo an intact uterus and ihose 

which would nol occur afler hysterectonmy. 

The categories included in the list are those 

w here vinually all of the cases could hbe ashted 

lo be attributable lo the presence of an 

nltact uterus. If most of the diagnoses in a 

d .gory could occur in a woman posthyster- 

.omy. this calegory was excluded. In this 

--. ¡. the charge data served as an estimate of 

ihe to,.' cosl of gynecologic services consumed 

because of uterine-relaled disease. In 

women who have had a hysterectorny. 

charges for none of ihese diagnoses should be 

incurred. and those foregone charges therefore 

would be considered tangible benefits,, of 

surgery. 

Table 1. Uterine-related diagoses used for 

dala s~let'ion 

tH I('I)A 

MalqnJan neop1am, uihc r ervxu U4t) 

Maiarnaln neoplasms ol ihe uiteus I.2 

Utenine fibrona t I« 

¢(her beni#n ncuplam% fd ihc uierus '19 

Ncu0plkm% uniptc'ised >< the uicru, 214 

Irun d

Tabi 2. Outpatlknl cose data 

- 396 - 

¡'.'.lt,ic Ilv hl'r'c lhl!' . A (C,.¡-B'. li, "c Anl'/sii 

Asc 

Pcnili I coshs 

AdjuIedJ io 

1972 dojlar, 1974 doII.tr. 

PcnxJl 2 o% 

'I ot.l 

uJuipJiien 

Paiient 

y )cars of 

c¢p>vurc 

Average 

annual 

cos per 

corl per 

woman (in 

rounded 

dollars! 

3-3 .4 . S7.363 S.20 .49th $12.Xh 349 537 

35-39 4.091 4.6Il '.27h 9.57 252 .39 

40-441 22.276 2..72 4.807 7.379 201) 37 

45-49 6.194 h .W9 3.1? 7 llIU.ihi 230 44 

50-.4 4.345 4.9110 2.6h1 7.

In.lumin- i',)i.eIu XV. Seph.irhb'r 1978 

- 397 - 

Table 4. Exposure snd expenditure data from Ihe Prepaid Health Care Project 

Calulated 

tota 

wolnan 

yc¿lr% of Oulpatlenl Inp a.cni 

Age e s rDoure co.:% Co , 

36--4 349 S37 $44 

35- 9 252 39 59 

40-44 2tu1 37 44 

45-49 ?.10 44 mn 

-.4 191 39 25 

55-49 139 22 ] 

4tL IMi 9 2" 

1Ir .I prtjected cXp 'Ihe wotien 

Annualied .ivcragc cxpccted expenJturc per uoman 

tin ro)unded Jollar,,) or uterinec-relatcd didgnosc%. 

PV )ol PV of 

' orltl am.oun totail cot total .om) 

of projertecd for cach age for cach age 

Ih%s per wom;an tor ¡he spesified 

ulerine-rel.ilcd diiagnote%. This table also 

ahows the nel direct henefil as a result of a 

hylrcr¡lomy lo he $2.735. undiscounied. 

When discounied ait a 3.'# interet ratle. the 

value of the henefil is reduced ¡o $1.822. By 

discounting al a ¡ia te of 6.. 5'. the pteent valtc 

hecomea $1.240. 

und.'rgoin¡g l C.'r omy have ne: uterine paihology 

anu *:xpracnce no muiohidily rel'er;ble 

Io the uterut. T'hi% ib Ihe exlrcme cajc 

where there are no medical indicaiion% for the 

operation. If a henetit were found in analyzing 

an operation pertormed under Ihee condilions. 

then le%% rigid crileria would Nho. 

greater benefil. 4) The palient. are healilhy 

and experience minimum operatilv mortalily. 

5) 'I'herc arc no %ignificani complicalion, *)i 

the %urgcr>. excluisive of fatality. l'hi% .,%- 

:.umpnion %ould Icad lo a conerv¿alive etimale 

of the co,%¡% that might he realized h> the 

operation. 

Once the annualized cot projecllon% were 

hobtained. the .%ream fl' epcetcd henefiti werc 

dJiícounied to preenl value u-,ing a rate of re- 

Itull oíf t.; . a- ucll a. a `'S; lisCitinl tate tlt 

ici the %en%iisvlyt of the hnlingn 1ti Ih¡t ple.ent 

vahlue calcula¿lion%. The rallonaile for u,,ing 

¡hese Ito hglrc% i., dlicust,,ed heilo 

The mean ct),t of a hytercctoniy in ¡he Prepuid 

iHealth C'ae Projel sa is 51.h67 (s.d. J. 

.36h. N 22. Although both presenl value, 

1I the heneftil,, ¡e ilhin (ne standard deviation 

of Ihc co(P%. the imporiance of the discounl 

r.ate can he sen h! the faic. thalí. at a rate of 

1',. I¡lc ie ¡ .a -s,.d nl t heIllelil of .S18.- 

Rt'~ 

I he calculaíion o., >' .gc-.pe k:i ,i:.-:, h,r ,iat 

wh:te.s '. italh a. dlsnini tale of h .';. there is 

. nc(l :o,( of $197 KRc.liS:? Ill I¡he stlialiegy 

,1 tihe ,snal lsl U; .s chosecn Io mt;lxiillmue the 

palaeni .mnd inpalient Itcalmenl o uerlcHlle lelalcd 

dJeaJae are iel forth mn I ahlk 2 und .l 

I'ahle 2 ,h( thatl the average annu¿al per :arpila 

c%.% rlo uherine-relalc(l anthul.fi(, utac 

eparciled dtrel htncarili. relaailtrn of an) of 

¡he icstli liti:. ;surnmplmri.r osf he hanal¿. .i% 

,ouIhl lireascc e the nel henelil (1o incrcei,, ihe 

nflt io I>r

- 398 - 

IElrtli o' Ilv.-4 r1h r ,y . A m. miagphl ;t imke the inal;ngible henefits 

:h: h1 -:llIgc! I IIe. enotg!; tI i.tln. eigh Ihc 

¡'e' a.llgilile .s" ¡l tilie diaLg'- o Dealae asld 

¡ 1cne -ehi,¡ ¡stksl lig.si¡ius'u tv{tjiii e,- preferahl 

tu h',a :sase %sier. %t ll sesta lsu a' ¡he 

g.uif st-,1o ual¡liag.ale .ag.allast iei.t uiaiei',lm a 

791

In#unrvlVolutn XV. September 1978 

hysierectomy for olher than medical indications. 

In addition lo thes financial considerations. 

mortaliiy calculations-performed by ihe auihors-were 

found lo be i, agreement with 

those of Cole and Berlin. Our analysis showed 

thal 30-year-old women undergoing elcclive 

hysterectomy could expect an average increase 

in life expectancy of 0.32 years. assuming 

a 0.06"i operative monality. However. 

half of those addilional years of life would be 

lived after the age of 75. and 7?8f would be 

lived after the age of 65. Thus. the piocedure 

would largely benefit elderly women al the expense 

of the young uaman who would no¡ survive 

the surgery. 

Whether a given woman should have a hys- 

Iereciomy is an individual decision. Whether 

eleclive hyslereclomies should he financed by 

third panies is a policy decision which is more 

likely io be at leas¡ partially influenced by direct 

cosi considerations. If such a decision resulted 

in an increase in the nel cost of care. 

such costs would he borne by ihe public 

Iwefm;c aid N«oM 

- 399 - 

through higher insurance premiums. If the decision 

resulled in a decrease in the nel costs 

of medical care. the public might be expected 

to benefit through Jecreased premiums. The 

findings of ihis study. and the companion analyses. 

suggest that a policy of permilling eleclive 

hyslerectomies probably would not be 

cosi-beneficial when direct costs are considcred. 

However. financial crileria are necessary. 

though nol sufficienl. for making policy 

decisions on a societal level. 

Wilh the increasing cosls of medical services. 

and the increasing role of government 

and ihird parlies. limits probably will be sel 

on ¡he amount of money the country will 

spend on heallh services. and on ihe type, of 

services ihat can be reimbursed. Cost-benefil 

;tinalyses such as this can provide information 

,ich is useful in making such decisions; even 

so. many olher factors will need lo be considcred. 

regardJless of whether the decision is 

bcing made by an individual. an underwriter. 

or the governmentl. 

Mead. . S ríti.al OeWulrgnJ ém .O, .tSlu. I#ipiuh. sun J.I,,rnul e/ . t ,trJ .41t(' ¿97>t0 I)HFW. Public 

Ilc4lth Serustc. S H li .R.sllc. Md 1975 Vol I. 

S.e I ablc I-'5 pp 1-71 

1',l M hR ksJ¢II -Inflation .nJ Rcl Intcrcei 

'..,,,, ,/ 1 ,,. -.. , 7 1 -. 2i '. Junc 19h 

' J,, urnil .Ju 

) 

2i9! 

-- i ? '

- 400 - 

TALLER SOBRE ENSEÑANZA DE LA INVESTIGACION OPERATIVA 

Y EI, ANALISIS DE SISTEMAS EN PROGRAMAS DE 

ADMINISTRACION DE SALUD 

El Taller se celebró en Caracas, Venezuela, del 8 al 12 de marzo de 1982, con 

los participantes que se mencionan en el Apéndice. Fue una actividad del Programa 

de la OPS/Fundación W. K. Kellogg de Educación en Administración de 

Salud y tuvo por objeto tratar sobre la incorporación de las técnicas de la investigación 

operativa y el análisis de sistemas; en los programas regulares de capacitación 

y en los de educación continua de administradores de sistemas de salud. 

Los objetivos específicos del Taller fueron: 

* Intercambiar experiencias entre los participantes en relación con la enseñanza de la 

invcstigación operativa y el an'álisis de sistelnas aplicados a los probclemas de atención elc 

salud. 

* Desarrollar un plan básico para la enseñanza en esos campos. 

* Seleccionar una bibliografia básica para ser usada en la actualización de educadores y 

administradores de los servicios de salud así como en el proceso educacional. 

* I)isci'ar un prog'ranla básico tic dlucación, continma en investigación operativ:a y aUl',lisis 

de sistemas para la actualización de profesores, investigadores y administradores en el 

sector salud. 

En 1970 no se conocía en América Latina la aplicación de la investigación 

operativa a los servicios de salud. En 1971 la OPS organizó un simposio sobre el 

análisis de sistemas en el que se recomendó el empleo prioritario de la investigación 

operativa a nivel hospitalario. 

Ese mismo año se inició un programa de verano en el Instituto Tecnológico de 

Estudios Superiores de Monterrey, México, en el cual estudiantes de ingeniería 

industrial y profesores participaron junto con administradores de salud en la 

definición y solución de problemas hospitalarios. Desde entonces, diferentes 

modalidades de este programa se han desarrollado en otras universidades de Colombia, 

Costa Rica, Chile, Perú, México y Brasil. 

En la actualidad se tiene conocimiento de unos 11 programas de ingeniería industrial 

y de sistemas en los cuales estudiantes de dicha especialidad han participado 

en investigacion.es de servicios de salud, y 54 programas que ofrecen 

capacitación en administración de salud. 

1 EI análisis de sistenlas y la investigación operativa se deben concebir de modo amplio para que 

comuprendan las técnicas cuantitativas de la ciencia de la administración, cuya aplicaició6n a la solución 

de problemas reales puede ser de gran utilidad. Dichas técnicas pueden ser de diferente grado de 

complejidad, desde la ingeniería industrial clásica hasta la programación matemática. 

4

Problemas identificados 

- 401 - 

Durante el T.,ller los participantes examinaron los diversos problemas 2 que se 

observanm en la Arierica Latina en la capacitación de administradores de salud en 

técnicas científicas de gestión. Dichos problemas se resumen seguidamente: 

Problemas en el sector salud 

* Los médicos, que son los responsables de la toma de decisiones en el sector, por lo 

general no conocen las técnicas del análisis de sistemas y la investigación operativa. Por 

consiguiente, las decisiones se toman con base en criterios subjetivos y políticos, sin el 

beneficio de metodologías cuantitativas y de información pertinente. 

* No se cuenta con información adecuada a nivel nacional, regional o local para tomar 

las decisiones. La solución de problemas en los sistemas de salud exige el desarrollo y la 

utilización de sistemas de información. 

* Los administradores de nivel intermedio no tienen los conocimientos adecuados de 

estadística descriptiva, análisis de sistemas y técnicas de evaluación requeridos en un enfoque 

racional del proceso decisorio. 

* Algunos administradores desconfían de los profesionales capacitados en técnicas científicas 

de gestión, lo que dificulta las relaciones interinstitucionales y perjudica el proceso 

educativo. 

* En el proceso de capacitación de administradores en métodos cuantitativos (especialmente 

de médicos) se corre el riesgo de que estos, con el poco conocimiento adquirido, se 

consideren "expertos" en la materia. 

* En las instituciones de salud se han creado pocas plazas para personas capaces de 

aplicar las técnicas científicas de gestión de problemas, por lo que en el sector salud las 

oportunidades de trabajo para profesionales en investigación operativa son limitadas. 

* Los especialistas en investigación operativa muestran una tendencia a concentrarse 

en los problemas pequeños, ignorando otros más importantes que tienen importancia nacional. 

Problemas el secaor educativo 

* La mayoría de los profesores universitarios no posee la experiencia adecuada en la 

aplicación del análisis de sistemas y la investigación operativa en el sector salud para 

motivar y enseñar eficazmente a estudiantes de administración de salud. 

* La enseñanza no debe consistir simplemente en disertaciones académicas. Para que 

sea eficaz, debe mostrar a los administradores de salud cómo aplicar las metodologías y 

técnicas a la solución de los problemas. Para ello se requiere de una adecuada preparación 

por parte de los docentes y de ayudas didácticas tales como casos de estudio y de un laboratorio 

sobre toma de decisiones. 

* En el diseño de programas de administración de salud es deseable la integración de las 

escuelas de ingeniería industrial y de sistemas, administración y medicina. Un problema 

fundamental de esta integración consiste en la orientación de las escuelas de administración 

e ingeniería hacia el sector privado o industrial; en general se tiene poca experiencia o 

poco interés en el sector público. 

* La mayoría de los profesores de salud pública en América Latina son médicos y, por 

lo general, muestran resistencia hacia la participación de otras disciplinas en el proceso de 

capacitación de personal de salud. 

2 Para un ejemplo específico véase: Pérez, C. E. Improving the Managerial Capabiliy of the Colombian 

Heaklh Systm, trabajo presentado en la reunión conjunta de ORSA/TIMS, Toronto, mayo de 1981. 

5

- 402 - 

* Como las universidades latinoamericanas por lo general no poseen los recursos financieros 

para mantener personal docente de calidad, existe un continuo éxodo de profesores 

hacia el sector privado y el extran. ero. Quizás cl problema más serio con que se enfrentan 

las instituciones educativas en América Latina sea precisamentc cl de contratar y mantener 

recursos humanos calificados. 

* Existen obstáculos institucionales y políticos que dificultan el establecimiento de relaciones 

formales entre las escuelas de salud pública, ingeniería y administración. Sin embargo, 

sin estas relaciones resulta difícil desarrollar programas educativos integrados. 

Estos obstáculos solo podrán superarse si las autoridades responsables de la administración 

de las escuelas y de las universidades reconocen este problema y la necesidad de encontrarle 

solución. 

Recomendaciones 

En e'l Taller se formularon una serie de recomendaciones orientadas al desarrollo 

de programas de educaión, tanto continua como regular, para administradores 

Programas educativos 3 

- 403 - 

1. Se recomienda desarrollar tres tipos de programas para enseñar investigación 

operativa y análisis de sistemas a estudiantes de administración de salud y 

analistas de sistemas en América Latina (OPS, IE, IO/AS): 

* Cursos cortos, principalmente para administradores y otros profesionales del sector 

salud. El curso típico durará de 16 a 40 horas y tendrá como objetivo enseñar al estudiante 

el empleo de metodologías y técnicas particulares en la solución de problemas reales de importancia. 

Se recomienda desarrollar una serie de estos cursos ya que es ilusorio pensar 

que un solo curso es suficiente. No se exigirán conocimientos previos de matemática. El 

curso permitirá al estudiante entender la perspectiva del enfoque de sistemas, apreciar la 

necesidad de utilizar información en el proceso de toma de decisiones y reconocer que cxisten 

individuos capacitados que pueden aplicar las técnicas de la administración científica 

en la solución de problemas del sector salud. Deberán planificarse cursos cortos parecidos 

para investigadores operacionales y analistas de sistemas a fin de lograr que su función en 

el sector salud sea más eficaz. 

* Programa de certificacidn, con una duración aproximada de 240 horas, para administradores 

de salud en ejercicio. El programa ofrecerá al administrador un conocimiento más 

profundo de las técnicas científicas de gestión. Debe ofrecerse, de preferencia, en la noche 

y fines de semana, para permitir al administrador/estudiante continuar con su trabajo 

regular. Como requisito en el campo de las matemáticas solo se deberá exigir conocimientos 

básicos de algebra. 

* Programa de maestria, de dos años, para administradores que deseen adquirir un conocimiento 

más profundo sobre investigación operativa y análisis de sistemas. Como requisito 

de admisión se deberá exigir conocimientos de cálculo. Se deberá ofrecer un segundo programa 

a nivel de maestría para preparar individuos con orientación técnica (ingenieros industriales 

y de sistemas, matlemáticos, fisicos) en la solución de problenmas del sector salud. 

Con base en la experiencia de los Estados Unidos de América, estos individuos podrán 

asumir en períodos relativamente cortos puestos administrativos en el sector salud y se 

convertirán en una fuente secundaria de administradores con formación cuantitativa. 

2. Los estudios de investigación operativa y análisis de sistemas dcpcndcn dce 

datos. La toma de decisiones óptimas se dificulta si no se posee información 

válida y confiable. Sin embargo, en la mayoría de los programas de administración 

de salud se presta muy poca importancia a este aspecto. Se considera que la 

OPS deberla llevar a cabo un seminario regional sobre "datos e informática en 

sistemas de salud", que podría incluir temas tales como: fuente y acopio de 

datos, diseño y desarrollo de sistemas de información, procedimientos de entrada, 

procesamiento, control de calidad, clasificación y análisis de datos, intercambio 

de información y otros temas afines (OPS). 

3. En cada país se deberá realizar una serie de talleres que reúnan a las autoridades 

del sector salud y técnicos competentes en el área de sistemas e investigación 

operativa. Dichos talleres deben estar orientados a la solución de problemas 

concretos de la prestación de servicios de salud (OPS, Organismos, IE). 

4. Se deberá organizar un taller ulterior para analizar las dificultades que se 

encuentren en la ejecución y actualización de las recomendaciones aquí for- 

3 La descripción de los programas educativos recomendados aparece en la pág. 12. 

7

- 401 - 

muladas. En este taller se discutirían los problemas relacionados con la ejecución 

de programas y proyectos, prestando especial atención a los obstáculos políticos, 

culturales y sociales. Deberían participar los asistentes en el presente Taller y 

otros individuos involucrados en proyectos de investigación operativa y análisis 

de sistemas a nivel nacional o regional en los países latinoamericanos (OPS). 

Rlccur.s ,y ilalerial didlcbti o auxiliar 

Antes del Taller, se realizaron dos estudios bibliográficos (español y portugués, 

e inglés) que sirvieron de referencia para el mismo y que también podrán 

utilizarse en los cursos que resulten de sus recomendaciones. Los estudios se concentraron, 

en primer lugar, en artículos y documentos que pudieran servir de 

base para la forn. 'ación de casos de estudios, útiles en la enseñanza de análisis 

de sistemas e investigación operativa para administradores de salud, y en segundo, 

en artículos que ilustraran la problemática de la prestación de servicios de 

salud. 

Se identificaron 60 artículos en español y portugués y se presentó un resumen 

de 43 de ellos. 

El cxanmen de las publicaciones en inglés (adelantado con la ayuda de la computadora) 

reveló miles de artículos relacionados con el análisis de sistemas y la investigación 

operativa en el sector salud. Se seleccionaron 116 artículos y 24 libros 

divididos en ocho subcategorías: evaluación de proyectos, planificación y programnación 

de recursos hum,aanos, distribución de recursos, estimación de dencanda, 

análisis de costo-beneficio, control de inventarios, evaluación de tecnologías y 

análisis de áreas de capacitación. En la selección se tuvo presente el nivel 

matemático de los artículos, teniendo en cuenta que se utilizarán para el adiestramiento 

de administradores, y se eliminaron aquellos artículos puramente teóricos 

y los que exigían conocimientos de matemática superior. Aunque se insistió principalmente 

en publicaciones recientes, se incluyeron varios artículos "clásicos" y 

otros que tratan temas de importancia a nivel de los países. 

5. Después de examinar los 116 artículos en inglés, los participantes seleccionaron 

38 para traducción al español. Estos artículos servirían como fuente y 

como base para el desarrollo de casos de estudio a ser usados en los tres tipos de 

cursos recomendados. 4 

En la serie final de publicaciones que se recomiendan para los programas de 

administración en salud también se deberán incluir 12 de los 43 artículos en 

español y 10 documentos distribuidos previamente por la OPS, así como la lista 

de los 24 libros en inglés ,, la bibliografía complementaria (artículos no seleccionados 

para traducción); de esta forma, los instructores contarán con una serie 

importante de recursos bibliográficos. Si bien los libros y los artículos complementarios 

están escritos en inglés, resultarán provechosos en los programas de 

4 En la pág. 14 se incluye la lista de los artículos. 

8

- 405 - 

maestrfa, donde se supone que los alumnos poseen conocimientos básicos de ese 

idioma (OPS, IE). 

6. A fin de mejorar la calidad de la enseñanza y la pertinencia dcl contenido 

de los cursos, se debe facilitar cl desarrollo de casos de estudio y de programas 

"paquetes" de computadoras como material didáctico auxiliar (OPS). 

* Se deberán desarrollar varios casos de estudio, que varíen en complejidad desde simples 

versiones de material disponible actualmente en publicaciones, hasta casos más complejos 

que pueden requerir 8-16 horas de clase y laboratorio para resolverlos. Los problemas 

seleccionados deben ser representativos de los que enfrentan los administradores en 

salud latinoamericanos. Los organismos nacionales o los analistas que hayan participado 

en la solución de problemas en el sector deben proveer el material básico para los casos. 

Otros casos de estudio pueden obtenerse de la expansión de algunos artículos incluidos en 

los estudios bibliográficos. 

* Para mejorar las clases (y algunos casos de estudio) se deben preparar programas de 

computadoras para sesiones de laboratorio. También deben obtenerse series de datos 

reales que permitan a los alumnos el uso de los programas a fin de lograr una mejor conaprensión 

de las técnicas particulares y su empleo en la solución de problemas reales. 

Como parte de este esfuerzo, la OPS debe fomentar el intercambio de este tipo 

de información entre universidades de América Latina y el Caribc. 

D)cbcn circularsc resúmenes de proyectos y tesis sobre problemas de los sistemas 

de salud. Estos proyectos pueden servir como casos de estudio y material 

para los "laboratorios" (OPS, IO/AS, IE, Organismos). 

7. Durante el Taller se trató de identificar lagunas en la literatura. Sc consideró 

quc existen varias áreas deficientes, tanto en inglés como en español (cuadro 

1). Estas deficiencias deberán identificarse con más detalle por representantes de 

los países latinoamericanos. Posteriormente, la OPS deberá comisionar publicaciones 

o iniciar estudios que puedan reducir estas deficiencias (OPS, IO/AS, 

Organismos). 

Cuadro 1. Areas en que existei lagunas ena a libralura. 

Tipo de curso 

Ar:a Cura corio PrograRla de cenrificación Programa de nuesirla 

Español Inglés Espaiol Inglés Espaiol Inglés 

Análisis de sistemas D D S D D D 

Informática D D D D D I) 

Análisis de decisiones D E D E O E 

Investigación operativa D E D E D E 

Conceptos generales D D D D D D 

D: Deficiente. 

S: Satislacorio. 

E: Excelente. 

9

- 406 - 

8. Los propios participantes y las instituciones docentes deben considerar la 

posibilidad de colaborar en la elaboración de programas "paquetes" para microcomputadoras, 

que se puedan usar para fines didácticos y para la solución de problemas 

reales. Lo anterior implicaría que los participantes de las diferentes instituciones 

utilicen máquinas similares y un lenguaje común. Los últimos desarrollos 

tecnológicos en el campo de las microprocesadoras pone a la disposición del 

usuario computadoras capaces de resolver problemas de gran complejidad a un 

costo relativamente bajo. En general, el costo de desarrollar los programas será 

mayor que el costo de la máquina; de ahí la gran importancia de la colaboración 

(IE, IO/AS). 

9. Teniendo en cuenta la considerable superposición de los temas considerados 

en los talleres sobre administración de salud organizados por la OPS, esta 

deberá comunicar a futuros talleres los resultados de los ya realizados, en especial 

cuando el tema lo justifique (OPS). 

10. La OPS debe explorar la posibilidad de obtener fondos para la publicación 

de un libro en español sobre investigación operativa aplicada a los problemas 

de los servicios de salud y dirigido a profesionales latinoamericanos; se deberá 

prestar especial atención a problemas regionales y nacionales. El libro podría iiicluir 

algunos casos de estudio (OP/; IO/AS). 

11. La OPS debe facilitar la preparación de un artículo que describa las labores 

del presente Taller y sus resultados. El artículo se publicará en una revista 

apropiada y serviría para fomentar más actividades de este género y mejorar la 

comunicación entre especialistas de los diversos países (OPS). 

12. La OPS debe facilitar la distribución de documentos de trabajo y tesis 

preparados en los Estados Unidos de América a educadores latinoamericanos. En 

especial se recomienda circular periódicamente el Catalog of Hospital Management 

Engineering Technical Papers publicado por el centro de distribución de ingeniería 

de gestión hospitalaria de la Asociación Americana de Hospitales; citas selectas 

de Hospital Management Abstracis, y resúmenes de tesis de maestría sobre problemas 

de los servicios de salud (que podrían obtenerse por medio de la 

Asociación de Programas Universitarios de Administración de Salud (OPS, IE). 

Programas de intercambio de información 

13. La OPS debe organizar la revisión de manera regular de las publicaciones 

pertinentes sobre servicios de salud (en español e inglés) y mantenerse enterada 

de las principales reuniones nacionales e internacionales de las asociaciones de investigación 

operativa y de salud pública. La información resultante debe enviarse 

sistemáticamente a los programas de administración de salud en América Latina. 

También se deberá incluir una lista de cursos y reuniones pertinentes a nivel profesional 

en los Estados Unidos de América y en países latinoamericanos (OPS, 

IO/AS). 

14. El intercambio de información en los talleres puede mejorarse mediante la 

10 

al

- 407 - 

presentación formal de informes sobre proyectos o actividades de investigación 

por parte de algunos de los participantes. Estas presentaciones podrían programarse 

antes de la inauguración del taller (OPS). 

Rmomrndaciones generae 

15. Para lograr una presentación más eficaz de las metodologías de investigación 

operativa y análisis de sistemas a los administradores de salud, formando 

siniultáineanente un grupo de analistas competentes, se debe dar prioridad al 

desarrollo de los cursos cortos y al programa de maestría de dos años; aunque es 

importante, el programa de certificación tiene menos prioridad. Según se indicó, 

se deberán desarrollar varios cursos cortos usando varios formatos y contenidos 

para satisfacer a grupos específicos (IE, IO/AS). 

16. Todos los participantes deben promover el desarrollo de programas de investigación 

operativa y análisis de sistemas con énfasis en el sector salud en los 

departamentos y escuelas de ingeniería industrial, medicina, administración y 

administración de salud. Debe insistirse en la creación de programas interdisciplinarios, 

en particular cuando solo una universidad está involucrada (IE, 

IO/AS). 

17. Un componente importante de los programas de certificación y de maestria 

deberá ser (por lo menos) un curso de investigación operativa en salud, conjuntamente 

entre las escuelas de ingeniería, salud pública y medicina. En ese curso 

debe incluirse, como mínimo, un proyecto en el que los alumnos trabajen 

como equipo interdisciplinario (IE). 

18. Hasta que las universidades latinoamericanas no se conviertan en los 

principales centros de formación de profesionales a nivel de maestría en estos 

campos, las universidades en los Estados Unidos de América y Canadá con[inuarán 

siendo los centros de formación a ese nivel en la Región. En consecuencia, es 

conveniente que las universidades latinoamericanas establezcan acuerdos institucionales 

con programas y escuelas de esos países. Las universidades latinoamericanas 

deben procurar que los programas de Norteamérica sean relevantes y 

tengan en cuenta las necesidades educativas de los estudiantes latinoamericanos. 

Si es posible, las tesis de grado deben tratar sobre problemas que sean también 

importantes en América Latina (IE). 

19. A medida que los programas de administración en salud se desarrollen en 

América Latina, es conveniente que se establezcan vínculos firmes y formales con 

los programas de ingeniería industrial, investigación operativa y de administración. 

Sin estos vínculos, las técnicas científicas de gestión no se integrarán eficazmente 

en los planes de estudio de administración en salud (IE). 

20. La OPS puede facilitar aún más la realización de proyectos de investigación 

y programas de cooperación entre profesionales de los Estados Unidos de 

América y América Latina si se informa a los participantes y a otros individuos 

interesados sobre problemas prioritarios y posibles fuentes de financiamiento 

(OPS, IE, Organismos). 

11

- 4C -o 

21. Es preciso reiterar que la investigación operativa y el análisis de sistemas 

exigen grupos interdisciplinarios para la solución satisfactoria de los problemas 

de prestación de servicios de salud. Esta realidad debc reflejarse cuidadosamente 

en la elaboración y ejecución de todos los programas y planes de estudio (OPS, 

IE, Organismos, IO/AS). 

22. Para que los estudios y proyectos sobre investigación operativa se traduzcan 

en acciones exitosas, es preciso quc los administradores de los sistemas involucrados 

participen en todos los niveles y puedan reconocer el valor de emplear 

técnicas científicas de gestión (OPS, IE, Organismos, IO/AS). 

Descripción de los programas educativos recomendados por el Taller 

Como ya se señaló, se recomendaron tres tipos de cursos: cursos cortos, programas 

de certificación y programas de maestría. Se convino en que el énfasis 

debe estar primero en los cursos cortos, después en los programas de maestría y 

finalmente en los de certificación. 

Cursos cortos. Los cursos (de 16 a 40 horas) se ofrecerán intensivamente durante 

un periodio de dos ai -inf:l( dlis o ceslpaci;ido< s durii';eiic v.i'iais seCIililnas o iii 

SCIInCStIC. Guino Cin un curso solo sc pucdc presentar Inaterial introductorio, sc 

deben programar una serie de cursos afines que se complementen mutuamente y 

que ofrezcan a los participantes una experiencia educativa más completa. Los 

cursos deberán incluir problemas a ser resuehlos por los aluminíos, individualiniciltc 

y cii grupo. lstas sesiones de laboratorio podrían usarse para familiarizar a los 

alumnos con la computadora, ya sea por medio de un juego, por ejemplo, la simulación 

de un sistema regional de salud o de un hospital, o mediante un modelo 

sencillo de asignación de recursos o de pronósticos. 

A continuación se presentan ejemplos de cursos cortos: 

* Curso para administradores de hospitales y sistemas de salud. No se requieren conocimientos 

de matemáticas, aunque una preparación básica en esa materia es aconsejable. En el curso 

se presenta una introducción a las técnicas de la investigación operativa y del análisis de 

sistemas, insistiendo en la comprensión de las posibilidades y limitaciones de los mismos. 

Los alumnos adquieren conocimientos sobre el enfoque de sistemas, análisis de costobeneficio, 

sistemas de información, fuentes.de datos y sobre trabajo en grupos interdisciplinarios 

5 . Asimismo, aprenden a reconocer los problemas que pueden solucionarse con 

estas metodologías, a evaluar el esfuerzo y tiempo aproximados que se requieren para la 

solución y a reconocer posibles barreras políticas y posibles fuentes de ayuda para la 

realización de estos estudios. L:s temas específicos a incluirse en este curso pueden 

referirse a problemas simplificados de decisión, problemas de asignación de recursos, problemas 

de planificación (CPM/PERT). El curso debe incluir un proyecto de grupo en el 

que se formule un problema no estructurado. 

5 

site sería un curso ideal, pero se reconoce que en un solo curso no se podrán presentar 

adecuadamernte todos esos temas. 

12

- 409 - 

* Curso para ingenieros indusriale, investigadores operaionales y personal de otras diJcipliwas ¿1cnucas 

intcresados en participar en la solución de problemas del sector salud. Los alumnos se 

familiarizan con los sistemas de salud y con los tipos de problemas comunes en el sector. Se 

presentan y discuten técnicas particulares para solucionar algunos de los problemas. 

Programa de crtificación (por lo general, de 240 horas). El programa estaría diseñado 

para el administrador de salud en ejercicio que desea un conocimiento más profundo 

que el ofrecido a través de una serie de cursos cortos, y también para el 

profesional que aspira a un cambio de profesión. La participación de los administradores 

de nivel intermedio en el programa puede significar para estos un ascenso 

de categorfa. Se requiere preparación matemática en materia de álgebra. 

El egresado del programa estará en condiciones de plantear y solucionar problemas 

cuantitativos sencillos (programación lineal, teoría de colas), realizar 

análisis básicos de costo-beneficio; recopilar y reducir datos (estadística descriptiva); 

usar programas "paquetes" de computadora para análisis; formular (pero 

no solucionar) problemas más complejos. Asimismo, tendrá un conocimiento 

básico'de los costos de la prestación de servicios de salud. Dcbc cnfiatizarsc cl concepto 

de equipo interdisciplinario. El curso debe incluir un proyecto de grupo de 

2-3 meses de duración. 

Los cursos del programa deben incluir una introducción a métodos cuantitalivos; 

un curso miaás riguroso en la aplicación de métodos cuantitativos; cursos 

introductorios en economía de la salud, bioestadística, y un curso en sistemas de 

información. Pueden considerarse cursos opcionales sobre teoría de planificación 

y evaluación en salud; métodos cuantitativos de planificación y evaluación, y 

comportaicenlo de las organizaciones en el sector. 

Programas d maestría (dos años). Se recomendaron dos tipos de programa de 

maestría. El primero estaría dirigido a administradores de hospitales y sistemas 

de salud que necesitan un adiestramiento más intenso que el ofrecido en el prograUma 

de certificación. Aunque los alumnos conserven sus trabajos regulares, se 

espera que residan en una institución académica por un período adecuado. Se 

deben exigir conocimientos de matemática a nivel universitario. El programa 

debe incluir todos los cursos del programa de certificación, un curso avanzado de 

estadística, un segundo curso de investigación operativa (modelos probabilfsticos) 

y un curso introductorio en econometría. Los alumnos deben participar en 

dos o tres proyectos de grupo durante el programa; estos proyectos deben realizarse 

como parte de un curso interdisciplinario. 

El segundo programa de maestría estaría diseñado para profesionales de ingeniería 

que desean solucionar problemas de los sistemas de salud. El programa 

tiene una orientación muy fuerte hacia la metodología y técnicas de la investigación 

operativa. Además, los alumnos deben familiarizarse con los conceptos de 

contabilidad de costos, principios de economía y sistemas de información. Se 

espera que los egresados del programa puedan desempeñar varias de las siguientes 

funciones: 

13

- 4i0 - 

a) Solucionar problemas de programación lineal y de inventario. 

b) Formular y analizar modelos estadísticos. 

c) Realizar análisis estadísticos avanzados (ANOVA y regresión múltiple). 

d) Diseñar estudios para recopilación de datos. 

e) Programar en un lenguaje de alto nivel (FORTRAN, PASCAL o PL-I). 

f) Diseñar la evaluación de programas. 

g) Realizar simulaciones mediante el uso de la computadora. 

h) Presentar informes técnicos. 

i) Interactuar con las autoridades del sector y participar en equipos interdisciplinarios. 

j) Tomar un problema no estructurado, resumirlo, formular un modelo matemático, 

identificar soluciones e implantar la mejor y más aceptable de ellas. 

k) Realizar estudios de planificación a largo plazo. 

Arlículos seleccionados para traducir al español 

Los artículos que los participantes recomendaron se tradujeran al español son 

los siguientes: 

1. Berkson, D., 1. Whipple y cols. Evaluation of an automated blood pressure measuring 

device intended for general public use. AmJ Public Health 69(5), May, 1979. 

2. Escudero, J. C.: On lies and health statistics: Some Latin American examples. In.j] 

Healtit Serv 10(3):421-434, 1980. 

3. Fetter, R. B., Y. Shin y cols. Case mix definition by diagnosis-related groups. Med 

Care 18(2) Supplement:1-52, 1980. 

4.Frerichs, R. y J. Prawda. A computer simulation model for the control of rabies in an 

urban area of Colombia. Managemenl Science 22(4):411-421, 1975. 

5. Greenland, S., E. Watson y R. Neutra. The case-control method in medical care 

evaluation. Med Care 19(8), August, 1981. 

6. Hartunian, N., Ch. Smart y M. Thompson. The incidence and economic costs of 

cancer, motor vehicle injuries, coronary heart disease, and stroke: A comparative analysis. 

AmJ Public Health 70(12), December, 1980. 

7. Lcv, B., G. Revesz y cols. Patient flow analysis and the delivery of radiology service. 

Socio-Econ Plan Sci 10:159-166, 1976. 

J 8. Meredith,J. Program evaluation techniques in the health services. AmJ Public Health 

66(11):1069-1073, 1976. 

j 9. Nutting, P., G. Shorr y B. Burkhalter. Assessing the performance of medical care 

systems: A method and its application. Mcd Care 19(3), March, 1981. 

10. O'Connor, R. W. y G. L. Urban. Using a model as a practical management tool 

for family planning programs. AmJ Public Health 62:1493-1500, 1972. 

11. Reisman, A., J. Mello da Silva yJ. B. Mantell. Systems and procedurcs of patient 

and information flow. Hosp Hcalth Serv Admin Winter: 42-71, 1978. 

12. Schoenbaum, S. C., B. J. McNeil yJ. Kavet. The swine-influenza decision. NEngl 

J Med 295(14):759-765, 1976. 

13. Shuman, L. J., H. Wolfe y R. Dixon Speas, Jr. The role of operations research in 

regional health planning. Operations Rescarch 22:234-248, 1974. 

i 14. Vraciu, R. Programming, budgeting, and control in health care organizations: the 

state of the art. Health Serv Res 14(2), Summer, 1979. 

./15. Duran, L. y A. Reisman. Design of alternative provider team configurations: experience 

in both developed and developing countries. Technical memo 947, Department 

of Operations Research, Case Western Reserve University, Cleveland, Ohio, 1980. 

14 

a

- 411 - 

¢ 16. Hancock, W., D. Magerlein y cols. Parameters affecting hospital occupancy and 

implications for facility sizing. HatiM Sarv Res 13(3), Fall, 1978. 

17. Hindl, A., N. Dierckman y cols. Estimating the need for additional primary care 

physicians. Health Serv Res 13(3), Fall, 1978. 

18. Reisman, A., B. V. Dean y cols. Physician supply and surgical demand 

forecasting: a regional manpower study. Managemen Science 19(12):1345-1354, 1973. 

19. Abernathy, W. J. y J. C. Hershey. A spatial allocation model for regional healthservices 

planning. Operations Rescarch/ 20(3):629-642, 1972. 

20. Goldman, J. y H. A. Knappcnberger. How to determine the optiiltum nuniber of 

operating rooms. Modan Hospital 111:114-116, 1968. 

21. Revelle, C., D. Bigman y cols. Facility location: A review of contex-free and EMS 

models. Heltli Sav Res 12(22):129-146, 1977. 

22. Brodheim, E. y G. P. Prastacos. The Long Island blood distribution systein as a 

prototype for regional blood management. Interfacs 9(5):3-20, 1979. 

23. Duraiswamy, N., R. Welton y A. Reisman. Using computed simulation to predict 

ICU staffing needs. J Nur Admin, February, 1981, pp. 39-44. 

24. Harrington, M. B. Forecasting area-wide demand for health care services: A 

critical review of major techniques and their application. Inquiry 14:254-268, 1977. 

25. Centerwall, B. S. y M. H. Criqui. Prevention of the Wernicke-Korsakoff syndrome. 

N EnglJ Mcd 299(6):285-289, 1978. 

26. Couch, N. P., N. L. Tilney y cols. The high cost of low-frequency events: The 

anatomy and economics of surgicil mnishaps. N EnglJ Mcd 304(11):634-637, 1981. 

27. Eisenberg, J. M. y A. J. Rosofi: Physician responsibility for the cost of unnecessary 

medical services. N EnglJ Mcd 299(2):76-80, 1978. 

28. Henry, J. B. y R. L. Roenfeldt. Cost analysis of leasing hospital equipment. Inquiry 

15(1):33-37, 1978. 

29. Klarman, H. E. Application of cost-benefit ainalysis to Ithc heilth services ai1il i le 

special case of tcchnologic innovation. IntJ Healh Setv 4(2):325-352, 1974. 

30. McGregor, M. y G. Pelletier. Planning of specialized health facilities: size vs. cost 

and effectiveness in health surgery. N EnglJ Mcd 299(4):179-181, 1978. 

31. Schwartz, W. B. Decision analysis: A look at the chief complaints. N EngilJ Med 

300(10):556-559, 1979. 

32. Willems, J. S., C. R. Sanders y cols. Cost effectiveness of vaccination against 

pneumococcal pneumonia. N EnglJ Mcd 303(10):553-559, 1980. 

33. Warner, K. E. y R. C. Hutton. Cost-benefit and cost effectiveness analysis in 

health care. Mcd Care 18(11):1069-1084, 1980. 

34. Kendall, K. y S. Lee. Formulating blood rotation policies with multiple objectives. 

Ma nmlt Sccet 26(11), November, 1980. 

35. Jackson, M. N., J. P. LoGerfo y cols. Elective hysterectomy: a cost benefit analyais. 

Inquiy 15(3):275-280, 1978. 

36. Evans, J. R., K. Lashman Hall yJ. Warford. Shattuck Lecture-Healih care in 

the developing world: Problems of scarcity and choice. N EnglJ Med 305(19):1117-1127, 

1981. 

37. McNeil, B. J. y S. J. Adelstein. Measures of clinical efficacy: The value of case 

finding in hypertensive renovascular disease. N EnglJ Mcd 293(5):221-226, 1975. 

38. McNeil, B. J., E. Keller y S. J. Adelstein. Primer on certain elements of medical 

decision making. N EnglJ Mcd 293(5):211-215, 1975. 

En el cuadro 2 se indica para qué tipo de programa es aplicable cada uno de los 

38 artículos seleccionados para traducir al español. 

15

- 412 

Cuadro 2. Tipos de programa en los que resultarán más provechosos 

los artículos seleccionados para traducir al español. 

Aniculo No. Prograrma de Programa de Curso corto 

maestria certificación 

1. x X 

2. X X X 

3. X 

4. X 

5. X Xa 

6. X X X 

7. X 

8. X X X 

9. X X X 

10. X X 

11. X X 

12. X X X 

13. X X X 

14. X X 

15. X X X 

16. X X X 

17. X X X 

18I. X 

19. X 

20. X 

21. X X 

22. X X Xa 

23. X X X 

24. X X X a 

25. X X X 

26. X X X 

27. X X X 

28. X X X 

29. X X X a 

30. X X X 

31. X X Xa 

32. X X X 

33. X X 

34. X X 

35. X X X 

36. X X X 

3 7 y 38 X X X 

aOpcional. 

Artículos selcccionados en español 

Se seleccionaron los siguientes artículos, que se recomendó que la OPS distribuyera, 

al igual que sus propias publicaciones pertinentes a este campo. 

1. Ackoff, R. L. Posibilidades actuales de la investigación operativa. Administración de 

empresas 5(50):125-135, mayo, 1974. 

16 

U 

U

- 413 - 

2. Barrenechea, J. J. La selección de prioridades como integrante del proceso de decisión. 

Medicina sanitariay administración de salud (Tomo II, Parte 3: Atención de la salud), 

págs. 206-214. 

3. Dunia, W. A. de, C. Carmendia y cols. Servicio de emergencia. Consideraciones conceptuales 

sobre sufuncionamiento y organiracidn especial (Centros ambulatorios de salud y hospital general de 

200 camas. Nota técnica 76-NTe-13). Ministerio de Obras Públicas de Venezuela, Direcci6n 

General de Desarrollo Urbanístico, Secretaria Técnica, Unidad de Investigación, 

1976. 

4. Facultad de Ingeniería, Universidad de Costa Rica, y OPS, estudio de ingeniería industrial 

en el subsistema "Procesar y distribuir alimentos" del Hospital Nacional de 

niños. 1974. 

5. Grundy, F. y W. A. Reinke. Investigaciones de práctica sanitaria y métodos mnatemáicos de 

gestión: Capitulos 1, II, IV y VIII: OMS, Cuadernos de Salud Pública 51, 1974. 

6. Novaro, S. Asignación de camas de servicios de hospitalización: una técnica posible. 

Atmción médica 2(1-2), 1973. 

7. Rodríguez, R. J., L. C. Arcón y L. A. Almeida Pimentel. Estado actual del sistema 

de control de pacientes del hospital de clínicas. Bol Of Sanit Panam 84(6):493-504, 1978. 

8. Schmidt, L. Consecuencias tcnicas de investigación del sistema de información para la salud, 

Veneu/ela (SIS-V/80). Ministerio de Sanidad y Asistencia Social, Dirección dc Planilicaci6n, 

Presupuesto e Informática, Comité de Informática, 1980. 

9. Schmidt, L., E. Zorilla y M. Quintero. Sistema automático de monitoreo de señales 

EKG para una unidad de cuidados coronarios. Trabajo presentado en el Congreso Internaciuoid 

dc Sistenuls, Venezuela, julio dc 19111. 

10. OPS y Departamento de Ingeniería Industrial, Instituto Tecnológico y de Estudios 

Superiores de Monterrey, Nuevo León, México. Análisis de sistemas en lavandería de un 

hospital general. Washington, D.C., Documento HRR/13/2-B, 1976. 

En el cuadro 3 sc indican los cursos para los que son aplicables los artículos scleccionados. 

Cuadro 3. Tipos de programa en los que resultardn mnás 

provechosos los artículos en español. 

Arliculo No. ,lura,,ira ,acsrtía de cerfilicación rliglaa lc Curso colou 

1. X X X 

2. X X 

3. X X 

4. X X 

5. X X X 

6. X X 

7. X X 

8. X X 

9. X 

10. X X 

A continuación se incluyen como ejemplo los contenidos de varios cursos, que 

podrán adaptarse según las necesidades. 

17

- 4.!4 - 

EJEMPLOS DE CONTENIDO I)E LOS CURSOS CORTOS 

Introducción a la investigación operativay al análisis de sismas para profesionales de salud (40 horas) 

1. Introducción a los conceptos de análisis de sistemas, planificación, evaluación y toma 

de decisiones. Se debe presentar inicialmente una visión panorámica y un modelo conceptual, 

del análisis de sistemas aplicado a los problemas de prestación de servicios de salud. 

Este modelo debe servir de referencia a medida que se presenten los temas con mayor profundidad. 

2. Definición de un sistema de decisión, tipos de decisiones, tipos de modelos. Discusión 

sobre los niveles de toma de decisión. 

3. Introducción a la formulación de modelos; el arte de mnodelaje, modelos típicos de 

problemas importantes en los sistemas de salud. 

4. Presentación de un caso o de un microproyecto que los alumnos deben desarrollar 

durante el curso. El énfasis debe estar en la formulación del problema. La solución de una 

parte del problema puede lograrse usando programas "paquetes" para computadoras. 

5. Introducción a métodos de planificación y evaluación, incluyendo la preparación de 

diagramas, CPM (método de la ruta crítica), PERT (técnica para la evaluación y control 

de programas), gráficas de Gantt. Se debe incluir la discusión de métodos para la evaluación 

de procesos y de resultados utilizando ejemplos apropiados. 

6. Introducción a modelos de asignación de recursos. Presentación dc tan caso sencillo 

de programación lineal. Demostración de la técnica gráfica de solución. Estudio de 

supuestos en la formulación del problema. (Es deseable el uso de la computadora.) 

7. Introducción a fuentes y al acopio de datos. Deben estudiarse los problemas de la administración 

y procesamiento de datos, la confiabilidad de los datos, sistemas de información 

y posibles áreas problema. 

8. Introducción al análisis de costo-beneficio y a presupuestos. Deben examinarse el 

valor del dinero en el tiempo, cálculo de costo de proyectos, ejemplos sencillos de análisis 

de costo-beneficio, definición y cálculo de la tasa interna de retorno y relación costobeneficio. 

9. Análisis de costo-beneficio. Definición de beneficios y costos marginales, matriz de 

impacto-incidencia, ejemplo completo de costo-beneficio mediaite cálculos Imanlllu;cs. (.Es 

deseable un ejemplo utilizando la computadora.) 

10. Mercadeo de los servicios de salud. Definiciones básicas, diferencia entre venta y 

mercadeo, planificación estratégica, aplicaciones específicas al sector. 

11. Introducción al concepto de incertidumbre. Conceptos básicos de probabilidad, 

eventos, estimación y cálculo de probabilidades, el teorema de Bayes, distribuciones y 

valores esperados. 

12. Análisis de decisiones, incluyendo árboles de decisión, alternativas, resultados, experimentos, 

valores y su obtención. Análisis e interpretaciones del problema de decisión, 

estudios de cosas sencillas. Las sesiones de laboratorio deben incluir un caso completo 

sobre análisis de decisión. 

13. Técnicas de pronóstico. Concepto' de promedio móvil, suavización exponencial, 

cuadrados mínimos. Interpretación de regresión simple. Estudio de caso; es deseable el 

uso de programa dc computadora. 

14. Administración de inventarios y materiales. Costos importantes en el sistema de inventarios. 

Modelos básicos de inventario, tamaño de lote y nivel de orden. 

15. Ejecución de estudios de investigación operativa y análisis de sistemas. Grupos interdisciplinarios, 

trabajo con autoridades, validación de los resultados de los estudios, 

estudio de obstáculos en la implantación de los resultados. 

18 

dl

- 415 - 

Curwso (8 a 16 horas) para profesionales de saltud 6 

Tema Curso de Curso de Curso de dos días 

un día dos días (con matemática) 

Introducción al análisis de sistemas I hr. I hr. I hr. 

Concepto de sistemas 1 1 I 

Técnicas gráficas 2 5 5 

Técnicas matemáticas . - 8 

Acopio de

- 416 - 

9. Otras técnicas de acopio de datos: estudio de tiempos; análisis de asociación, cuestionarios. 

10. Sesión de trabajo sobre medición. 

11. Análisis de costos: medidas de costos; cálculo de datos; criterios no monetarios; 

comparación de métodos. 

12. Estadística elemental para análisis dc diferencias; prueba de hipótesis; interpretación 

de datos. 

13. Sesión de trabajo sobre evaluación de resultados. 

14. Importancia de los resultados de los estudios sobre mejoramiento de métodos. 

15. Sesión de trabajo: informes y críticas de proyectos. 

Curso/Seninario sobre invesligación de sistemas de salud (45 horas) 

Este curso está programado para estudiantes de administración de salud y de ingeniería 

industrial/investigación operativa que desean profundizar sus conocimientos en la aplicación 

de métodos cuantitativos a los iproblemas de los sistemas de salud. Además dc las discusiones 

académicas, los participantes formarán pequeños grupos interdisciplinarios para 

resolver un problema real de un hospital o del sistema de salud. Los temas a tratar en el 

seminario son: 

1. Introducción a la ingeniería industrial/investigación operativa aplicadas al sector 

salud. 

2. Evaluación del sistema de distribución de dosis unitarias de medicamentos. 

3. Modelos de costos hospitalarios: distribución escalonada de costos; microcosteo. 

4. Reembolso, modelo de incentivo: enfoque de ingeniería industrial. 

5. Modelos de evaluación de utilización de servicios. 

6. Modelos para la dotación de personal de enfermería. 

7. Modelos de reembolsos predictivos. 

8. Agrupación de hospitales para control de reembolsos y costos. 

9. Sistemas de admisión y programación de cirugías. 

10. Aplicaciones de la programación lineal: planificación de menús en hospitales, 

radiología terapéutica. 

11. Planificación regional: localización de una red de centros de salud. 

12. Servicios médicos de urgencia: sistema de información; simulación de sistemas de 

urgencia; evaluación del personal paramédico. 

13. Análisis y modelos de economía de salud. 

14. Evaluación de la eficacia de la atención médica. 

isuemas de itformación y evaluación de la calidad de la atención médica (30 horas) 

Este curso está programado para planificadores y administradores de los servicios de 

salud. Presenta una visión panorámica del uso de la información y de la tecnología del procesamiento 

de datos para la planificación, administración y control de los sistemas de 

salud. 

1. Origen, naturaleza y uso de la información en los sistemas de salud; documentación 

médica; registros administrativos. 

2. Inforinación para los procesos de torra de decisiones; importancia y objetivos de esa 

enseñanza de informática al personal de salud. 

3. Objetivos de los usuarios de la información; niveles funcionales; núcleo de centralización 

o de distribución de los sistemas de información; impacto de la información en las 

organizaciones. 

20

- 417 - 

4. Necesidades de los administradores; información para administración y planificaci6n. 

Qr 5. Análisis de sistemas; proyectos sobre análisis de sistemas; planificación y control de 

los proyectos; acopio y distribución de datos; proceso para establecer sistemas de información. 

6. Computadoras y automatización: conceptos básicos, procesamiento electrónico de 

datos; archivos y estructuras lógicas; bancos de datos; computadoras y sus componentes, 

centros de procesamiento de datos. 

7. Opciones para el procesamiento de datos: sistemas centralizados, descentralizados y 

distribuidos. 

8. Aplicaciones del procesamiento de datos en los sistemas de salud; ventajas e inconvenientes 

del uso de computadoras. 

9. Sistemas de información para la evaluación y el control; métodos de evaluación del 

uso y de la calidad de los servicios, auditoría médica. 

Mitodos cuantihtais y analílicos en administración de salud (40 horas) 

Este curso 8 está diseñado para: 1) desarrollar en el participante la apreciación y comprensi6n 

de los procesos básicos del análisis de sistemas, toma de decisiones y control; 

2) identificar la influencia de estos procesos en las actividades y responsabilidades del administrador 

de salud. Se discuten algunos modelos básicos de decisión y control así como 

la metodología del análisis de sistemas, ciencia de gestión y economía. Se recalcan las ventajas 

y limitaciones de estos modelos para la toma más eficaz de decisiones en el sector. Entre 

otros, se requieren conocimientos de álgebra a nivel universitario, álgebra lineal básica, 

probabilidad y estadística básicas. 

1. Revisión general de la literatura relacionada con el análisis de sistemas/investigación 

operativa y ciencia de gestión y el sector de la salud. 

2. Concepto y análisis de sistemas: un prólogo para la toma de decisiones en el sector. 

3. Modelos cuantitativos y sus funciones en la toma de decisiones y control en el sector 

salud. 

4. Modelos elementales para el análisis de decisiones determinísticas: inventarios y 

administración de proyectos. 

5. Modelos complejos de decisiones determinísticas: programación matemática. 

6. Solución de problemas de programación lineal por computadoras; análisis de sensibilidad. 

7. Acopio de datos apropiados para la toma de decisiones en el sector salud: técnicas de 

medidas. 

8. Aplicaciones de la programación entera y programación con metas en el sector de la 

salud. 

9. Análisis de costo-beneficio y costo-eficacia: un modo fundamental de pensar en la 

toma de decisiones en el sector de la salud. 

10. Realimentación, control y evaluación de programas para una administración más 

eficaz de la atención en salud. 

11. Temas especiales tratados por conferencistas invitados. 

8 Curso desarrollado por Barnett R. Parker, Ph.D., Escuela de Salud Pública, Universidad de 

Carolina del Norte, Chapel Hill. 

21

- 418 - 

Apéndice 

Participantes en cl Taller 

Ing. Moisés Arteaga Cabrera, Ministerio de Planificación, La Paz, Bolivia 

Carmela Calapinto de Rosales, Ministerio de Sanidad y Asistencia Social, Caracas, Venezuela 

Ing. Antonio dal Fabbro, Escuela de Salud Pública, Sao Paulo, Brasil 

Dr. Mariano Fernández, Ministerio de Sanidad y Asistencia Social, Caracas, Venezuela 

Dr. David R. Gómez Cova, Escuela de Salud Pública, Universidad Central de Venezuela, 


Dr. Pedro Guédez Lima, Dirección de Salud Pública, Ministerio de Sanidad y Asistencia 

Social, Caracas, Venezuela 

George Kastner, Instituto de Estudios Superiores en Administración, Caracas, Venezuela 

Ing. Carlos Enrique Pérez (Relator), Universidad de Pittsburgh, Pittsburgh, Pensilvania 

Ing. Victor L. Pérez, Universidad de Chile, Santiago, Chile 

Ing. Diego Prieto, Universidad de Los Andes, Bogotá, Colombia 

Ing. Carlos M. Quesada Solano, Escuela de Ingeniería Industrial, San José, Costa Rica 

Arnold Reisman, Universidad Case Western Reserve, Cleveland, Ohio 

Dr. RobertoJaime Rodríguez, PROAHSA, Hospital das Clinicas, Facultad de Medicina, 

Universidad de Sao Paulo, Sao Paulo, Brasil 

Dr. Richard H. Shachtman, Escuela de Salud Pública, Universidad de Carolina del 

Norte, Chapel Hill, Carolina del Norte 

Dr. Larry J. Shuman (Relator), Universidad de Pittsburgh, Pittsburgh, Pensilvania 

Organización Panamericana de la Salud 

Dr. Humberto Moraes Novaes, Asesor Regional y Coordinador del Programa OPS/ 

Kellogg de Educación en Administración de Salud, División de Recursos Humanos e 

Investigación, Washington, D.C. 

Ing. Jorge Ortíz, Asesor Regional en Investigación de Servicios de Salud, División de 

Recursos Humanos e Investigación, Washington, D.C. 

Ing. Jorge Peña Mohr, Asesor Regional en Administración Institucional, División de Servicios 

Integrados de Salud, Washington, D.C. 

22

MOISES ARTEAGA 

DR. DAVID GOMEZ COVA 

DR. ANTONIO DAL FABRO 

DR. CESAR HERMIDA 

DR. GEORGE KASTNER 

DR. CARLOS PEREZ 

DR. VICTOR PEREZ 

- 419 - 

LISTA DE PARTICIPANTES 

Subsecretario de Planificacion 

La Paz, Bolivia 

Escuela de Salud Publica 


PROAHSA - Hospital Das CUinicas 

Ave. 9 de Julh 2029 

01313 Sao Paulo, Brasil 

Curso de Postgrado en 

Administracion de Salud 

Facultad de Medicina 

Calle Iquique s/n 

Quito, Ecuador 

Instituto de Estudios Superiores 

Final Calle Occidente 

San Bernandino 


Inadustrial Engineering Dept. 

University of Pittsburgh 

U.S.A. 

Departamento de Industrias 

Universidad de Chile 

Santiago, Chile 

DR. DIEGO PRIETO Escuela de Ingenierja Industrial 

Universidad de los Andes 

Bogota, Colombia 

DR. CARLOS QUESADA Escuela de Ingenieria 

Industrial 

Universidad de Costa Rica 

San Jose, Costa Rica

DR. ARNOLD REISMAN 

DR. ROBERTO J. RODRIGUEZ 

DR. LARRY SHUMAN 

- 420 - 

Case Western Reserve University 

Cleveland, Ohio 

PROAHSA 

Ave. 9 de julho 2029 

01313 Sao Paulo, Brasil 

Industrial Engineering Department 

University of Pitt8burgh 

U.S .A. 

PROFESSOR RICHAR SHACHTMAN University of North Carolina 

Rosenan Hall 201H 

Chapel Hill, N.C. 27514 

DIVISION DE RECURSOS HUMANOS E INVESTIGACION 

DR. HUMBERTO M. NOVAES Coordinador 

ING. JORGE ORTIZ Investigador Operacional 

DIVISION DE SERVICIOS INTEGRADOS DE SALUD 

SR. JORGE PENA MOHR Asesor Regional en Adinnistracion 

de Salud 

OFICINA SANITARIA PANAMERICANA- CARACAS 

DR. BARRY W. WHALLEY RA-II Venezuela 

4 

1 

la 

4

BOOKS OF RtfiDIfGS - PAHO/WHO

Create successful ePaper yourself

Delete template?

Save as template?