UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An episodic approach to the demand for medical care Stoddart, Gregory Lloyd 1975

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1975_A1 S83.pdf [ 7.83MB ]
Metadata
JSON: 831-1.0100074.json
JSON-LD: 831-1.0100074-ld.json
RDF/XML (Pretty): 831-1.0100074-rdf.xml
RDF/JSON: 831-1.0100074-rdf.json
Turtle: 831-1.0100074-turtle.txt
N-Triples: 831-1.0100074-rdf-ntriples.txt
Original Record: 831-1.0100074-source.json
Full Text
831-1.0100074-fulltext.txt
Citation
831-1.0100074.ris

Full Text

AN EPISODIC APPROACH TO THE DEMAND FOR MEDICAL CARE by Gregory Lloyd Stoddart B.A., University of Western Ontario, 1971 A Thesis Submitted in Partial Fulfilment of the Requirements for the Degree of Doctor of Philosophy in the Department of Economics. The University of British Columbia July, 1975 We accept this thesis as conforming to the required standard. In presenting th i s thes is in p a r t i a l fu l f i lment o f the requirements for an advanced degree at the Univers i ty of B r i t i s h Columbia, I agree that the L i b r a r y sha l l make it f ree ly ava i lab le for reference and study. I further agree that permission for extensive copying of th i s thes is for s cho lar ly purposes may be granted by the Head of my Department or by his representat ives . It is understood that copying or p u b l i c a t i o n of th i s thes is f o r f i n a n c i a l gain sha l l not be allowed without my wri t ten permiss ion. Department of 6 C O / U J m<-S, The Univers i ty of B r i t i s h Columbia 2075 Wesbrook Place Vancouver, Canada V6T 1WS Date A - v y - f t & ("i?^ i ABSTRACT The t h e s i s i s an a n a l y t i c and empirical i n v e s t i g a t i o n of factors Influencing the demand f o r medical care. I t proceeds from the premise that the u t i l i z a t i o n process i s best conceptualized as c o n s i s t i n g of two stages, a p a t i e n t - i n i t i a t e d stage and a physician-generated stage. Furthermore, due to the nature of medical care i t s e l f , the economic concept of "demand" can only meaningfully be applied to the former stage while the term " u t i l i z a t i o n " encompasses both stages. I t i s then demonstrated that previous demand models, which p o s t u l -ated autonomous consumer decision-making and were tested upon u t i l i z a t i o n data rather than i n d i c e s of demand, have been both a n a l y t i c and empirical m i s s p e c i f i c a t i o n s . Consequently a measure of p a t i e n t - i n i t i a t e d u t i l i z a t i o n ( i . e . demand), l a b e l l e d "the episode of medical s e r v i c e " , i s defined, developed and employed i n regression a n a l y s i s of demand f o r and u t i l i z a t i o n of medical care by a sample of Vancouver f a m i l i e s . The major thrusts of the empirical p o r t i o n of the t h e s i s , both of which are supported by the r e s u l t s , are f i r s t that an e p i s o d i c approach should prove more appropriate f o r explanation of demand d i f f e r e n t i a l s than u t i l i z a t i o n d i f f e r e n t i a l s , and second that the conventional micro-economic model of demand must be re- i n t e r p r e t e d with respect to a medical marketplace characterized by p u b l i c medical insurance. The main empirical f i n d i n g s are the existence of a consistent, inverse, and approximately l i n e a r income-demand r e l a t i o n s h i p , the existence of a non-linear education-demand r e l a t i o n s h i p ( p o s i t i v e at lower education l e v e l s but l e v e l l i n g off a f t e r high school), and the i n s i g n i f i c a n c e of i n d i r e c t p r i c e e f f e c t s . i i TABLE OF CONTENTS LIST OF TABLES . . iV LIST OF FIGURES ACKNOWLEDGEMENTS v i CHAPTER 1. INTRODUCTION 1 CHAPTER 2. THE UTILIZATION PROCESS . 10 The Nature of Medical Care 10 Discretionary Behaviour of Physicians 14 Existing Measures of Ut i l i z a t i o n 17 CHAPTER 3. THEORIES OF THE "DEMAND" FOR MEDICAL CARE: THE CONVENTIONAL•MICRGECONOMIC MODEL 22 Price Effects .... 24 Income Effects 29 Taste and Need Variants 32 CHAPTER 4. THE HUMAN CAPITAL MODEL 44 CHAPTER 5. THE HEALTH BELIEF MODEL 59 CHAPTER 6. MEASUREMENT OF DEMAND THROUGH EPISODES OF MEDICAL SERVICE 69 Definition of Episodes 69 Determination of Episodes 73 Applications and Further Development 77 CHAPTER 7. HYPOTHESES CONCERNING DEMAND BY INDIVIDUALS, FAMILY HEADS AND FAMILIES FOR EPISODES OF MEDICAL SERVICE 81 Price Effects 81 Income Effects 82 Taste Factors 83 Significance of the Episodic Approach 87 i i i i CHAPTER 8. EMPIRICAL RESULTS: INDIVIDUALS 90 CHAPTER 9. EMPIRICAL RESULTS: FAMILY HEADS 105 CHAPTER 10. EMPIRICAL RESULTS: FAMILIES 114 CHAPTER 11. SUMMARY 122 BIBLIOGRAPHY .125 APPENDIX A. DATA BASE AND VARIABLE MEASUREMENT 133 APPENDIX B. ALGORITHMS FOR DETERMINATION OF EPISODES OF MEDICAL SERVICE 139 APPENDIX C. EXAMPLE OF RESULTS PROVIDING BASIS FOR GROUPING AND/OR DELETION OF DUMMY VARIABLES .... 142 APPENDIX D. VARIABLE LISTS, MEANS AND STANDARD DEVIATIONS 143 LIST OF TABLES Table Page I Comparison of Predicted and Estimated Results of the Human Capital Model 52 II Demand for Episodes of Medical Service: A l l Individuals gi III U t i l i z a t i o n of Medical Services: A l l Individuals 93 IV Demand for Episodes of Medical Service: Heads 106 V Ut i l i z a t i o n of Medical Services: Heads ]_']_]_ VI Demand for Episodes of Medical Service: Families 115 VII Uti l i z a t i o n of Medical Services: Families V LIST OF FIGURES Figure Page 1. Determination of Optimal Health Stock for Any Age i 48 2. A Schematic Interpretation of the Health Belief Model 61 ACKNOWLEDGEMENTS To properly acknowledge everyone who significantly assisted me in the course of this research would probably require two pages. But why not? I wish to begin by thanking my thesis supervisor, Robert G. Evans, for generously making his time available whenever needed, during even the most hectic of weeks. Although I cannot say that he initia t e d my interest in health economics, he has certainly heightened and shaped i t s development significantly. His advice on this research and other matters over the past four years has, in retrospect, been consistently sound, though sometimes unheeded. Perhaps most important, he has at c r i t i c a l times forced me to think • and then demonstrated how to do i t better! For the past three years i t has been my good fortune to be asso-ciated with Project T.E.A.M.,'a multidisciplinary research group co-ordinated by Morton M. Warner of the University of British Columbia Department of Health Care and Epidemiology. I hope that this experience has been as productive, interesting and enlightening for my associates as i t has been for me. For their overall assistance and specific advice regarding the development of episodes, I would like to thank both Morton Warner and Dr. Michael Coles of Project T.E.A.M. Lewis James i s one member of the Project who deserves special thanks. As programmer/analyst he was responsible for the organization and any transformations of the T.E.A.M. data set, which provided the data samples for this thesis. I suspect that Lewis' consumption of aspirin could be modelled quite well as an exponential function of my v i s i t s , requests v i i and comments concerning the data set; however, his assistance has been invaluable. * I have also benefited from discussion with Terence J. Wales of the U. B. C. Economics Department and Anne Crichton and Tim Cramer of the Health Care and Epidemiology Department during the latter stages of the thesis. Financial support was provided by Health and Welfare Canada through Student Fellowship No. 610-22-17. Economists are fond of stressing the importance of behaviour at the margin. Had i t not been for the assistance of Tim Day, Vera Anne Stoddart, Kerri E l l i o t t , Estelle Murray and Tina-Lynn Chow, I fear that the thesis might have stalled at c r i t i c a l stages. Shelagh Penty, in addition to typing parts of the f i n a l and an earlier draft, has advised on and assisted in the production of various materials at a l l stages of the research. I also wish to thank Janet Whiteley, who typed a large part of the f i n a l draft, for tolerating a typing schedule which was in a state of constant revision. My warmest thanks, however, are to Vera Anne, my silent partner in this project, for her financial support, her friendship and her silence. Chapter 1 1 INTRODUCTION Over the past two decades the level of national health expenditures i n Canada has risen steadily and dramatically to a point where i t now approxi-mates an unprecedented eight percent of GNP^. This growth in expenditures, and associated problems in the organization and distribution of medical • . s . . . services, have resulted i n a plethora of commissions and proposals to change 2 the health care delivery system ; however, our information base is a l l too often inadequate for major policy decisions. In particular, questions such as "what accounts for the recent growth in health care expenditures?" and "what have we achieved i n terms of health care system output with such ex-penditures?" remain largely unanswered. 3 Research by Fuchs (1973) and Evans (1974) on the f i r s t question i n -dicates that there are no simple answers price changes, quality changes, growth in per capita income, population growth, and the introduction of hospital and medical insurance have apparently interacted to produce most of the growth in expenditures, while a residual remains unexplained. One message that clearly emerges from the work of both authors, however, i s that we must increase our understanding, at the micro level, of the decisions and factors involved in the process whereby consumers u t i l i z e medical services i f we are to be able to account for changes in the quantity of services consumed. It is to this general objective that the present thesis relates. This thesis is an analytic and empirical investigation of factors related to the demand for medical care, in particular the demand for physician services. It proceeds from the premise that the u t i l i z a t i o n process is best conceptua-lized as consisting of two stages, a patient-initiated stage and a physician-4 generated stage . The former consists of the consumer's i n i t i a l contact with the physician for a given health problem, request or situation. The latter 2 consists of subsequent contacts and services provided during the remainder of the treatment episode for the purpose of satisfying the consumer's re-quest or resolving his/her problem or situation. Further, due to the nature of the good called "medical care" and the dual role of the physician as both supplier of services and effective demander of services on behalf of the patient, the economic concept of consumer demand can only meaningfully apply to the patient-initiated stage of u t i l i z a t i o n . Even then the mean-ing of the term is not entirely clear since, as we discover below, neither price nor the product being demanded is particularly well-defined in the medical care market. Although this two-stage view of the u t i l i z a t i o n process is potential-ly useful in the formulation and testing of models of both patient and physician behaviour, existing u t i l i z a t i o n measures do not allow i t s oper-ationalization since they are not based upon episodes of illness or treat-ment and thus do not permit identification of the patient-initiated stage, i.e. the i n i t i a l contact of any episode. In this regard, the current re-search differs significantly from previous work on the demand for medical care in that the i n i t i a l objective i s the development of a medical care u t i l i z a t i o n measure permitting identification of the patient-initiated stage and thereby recognizing the distinction between demand for and u t i l i z a t i o n of medical services. While the newly developed measure has several potential applications, i t s a b i l i t y to provide an index of patient-initiated demand is of primary interest here, since the second objective of the thesis i s to empirically investigate the relevance of the conventional microeconomic model of the demand for medical care to a marketplace characterized by public medical insurance. As we shall demonstrate below, an index of patient-initiated demand"*, though lacking in previous studies, i s essential to the proper specification of the conventional micro model for empirical work. The importance of understanding the u t i l i z a t i o n process was alluded to above but warrants elaboration since u t i l i z a t i o n of medical services has both economic and c l i n i c a l significance. Under our existing medical care delivery system most personal medical care expenditures result from i n i t i a l contact with a physician. Given this situation, knowledge of factors influencing the demand for medical care i s potentially useful in controlling u t i l i z a t i o n and, under public medical and hospital insurance plans, the cost of providing care. Moreover, such knowledge is of general importance in predicting the reactions of consumers to changes i n the medical care delivery system whether the changes are primarily structural, such as the community health centre movement i n Canada, or financial, such as proposed national health insurance schemes currently inspiring substantial interest in health economics in the United States. A clear understanding of consumer demand may also assist efforts to model physician behaviour, a pivotai and controversial topic in health economics, by isolating the physician-generated stage of u t i l i z a t i o n . Finally, an understanding of demand and u t i l i z a t i o n i s essential to an assessment of the va l i d i t y of claims that shortages of medical personnel exist.^ From a c l i n i c a l viewpoint, knowledge of factors influencing demand is useful in attempts to encourage u t i l i z a t i o n of certain services by specific target groups as part of an overall objective of increasing the health status of a population. Assuming that the health benefits of u t i l -ization are positive at the margin — - an assumption which incidentally re-mains largely unconfirmed^ removal of known economic and non-economic barriers to u t i l i z a t i o n should positively influence health status-. Until recently this c l i n i c a l orientation was responsible for the greater portion of research on the u t i l i z a t i o n of medical services and i t was through the c l i n i c a l avenue that such research influenced public policy. For example, in response to evidence indicating that financial barriers prevented people from consuming supposed optimal amounts of medical care, attention was and s t i l l i s devoted to the potential role of health insurance plans in guaranteeing financial access to services. The Hospital Insurance and Diagnostic Services Act of 1957 and the Medical Care Act of 1966 are prime examples of legislation arising out of this concern in 8 Canada while the Americans have recently enacted Medicaid and Medicare plans to assist the poor and the elderly, respectively. When ignorance of basic health information was thought to be res-ponsible for gaps between potential and actual health levels, efforts were concentrated on health education programs for the public, both at the com-munity and national levels, ranging from instruction in proper methods of toothbrushing to the creation of awareness of the danger signals of cancer. Similarly, now that spatial accessibility is considered a key determinant of demand, increasing emphasis is being placed upon the location of new f a c i l i t i e s , an emphasis which has been partly responsible for the creation of neighbourhood health centres by the U.S. Office of Economic Opportunity. . Prior to describing the organization of the ensuing chapters, two qualifying remarks are in order. F i r s t , caution must be exercised in em-ploying the terms "health care" and "medical care" since the latter i s only one component of the former. Health care a c t i v i t i e s , broadly defined, i n -clude any and a l l a c t i v i t i e s influencing health, from specific medical services to pollution control programs and community physical fitness and recreation programs. Similarly, physician services are but one component of medical care, which also includes the services of other medical professionals 5 (dentists, nurses, etc.), hospital services, drugs, and medical appliances. The thesis concentrates upon physician services because, structurally, the physician i s undoubtedly the quarterback of the medical care system while, empirically, Canadian data do not yet permit the linkage of records necessary to provide a comprehensive statement of the medical care consumption of individuals or families. Second, i t should be emphasized that the term "physician-generated u t i l i z a t i o n " i s not intended to possess any normative significance, and in particular should not be interpreted as implying that such u t i l i z a t i o n i s by definition unnecessary. Rather the term simply reflects the obser-vation that, after i n i t i a l l y contacting a physician, a patient usually relinquishes effective control over the quantity and type of services he/she receives and accepts the physician's advice regarding a treatment 9 regimen . The physician thus becomes an agent who is normally assumed to be acting on behalf of the patient for the remainder of the u t i l i z a t i o n process. Models of physician behaviour and discussions of physician objectives may question whether this agency relationship is always foremost in the physician's mind; fortunately, however, conclusions regarding alter-native objective functions of physicians are not central to the demand analysis undertaken in the following chapters. Chapter 2 expands on the proposition that demand and u t i l i z a t i o n d i f f e r due to the nature of the good "medical care" and the resulting agency role of the physician. Empirical evidence is cited to support the contention that physicians possess discretionary power over the demand for their services and consequently u t i l i z a t i o n i s the product of both patient and physician preferences and decisions. The most common ut i l i z a t i o n measures are examined and found deficient for demand analysis. Finally, questions 6 of product definition, in the medical marketplace are raised, remaining un-answered u n t i l Chapter 6. If demand and u t i l i z a t i o n d i f f e r , then this difference should be re-cognized in analytic frameworks employed to test models of either; however, this i s not the case with most previous literature. Chapters 3, 4, and 5 review empirical work originating from three theories of the demand for medical care. The review indicates that consistent model misspecification has occurred, i.e. while theories of autonomous consumer demand have been tested upon u t i l i z a t i o n data and attempts have been made to explain u t i l i z a -tion rates without including supply side variables, we lack both a theory of demand tested upon an appropriate index and an appropriate general theory of patient-physician interaction capable of explaining existing u t i l i z a t i o n data. This curious, yet understandable (in ligh t of the blurred demand-ut i l i z a t i o n distinction), state of the art may in part account for the con-f l i c t i n g empirical results which dot the literature. Results of previous work on the conventional micro model of the demand for medical care are examined i n detail since they are especially relevant to hypotheses tested in subsequent chapters. Assembly of a demand model begins in Chapter 6 with the introduction of a demand and u t i l i z a t i o n measure termed "the episode of medical service". The key feature of the "episodic" framework i s that i t permits the i d e n t i f i -cation of i n i t i a l contacts.for health problems and situations while retain-ing the capacity to generate existing u t i l i z a t i o n measures. This i s s u f f i -cient to enable us to make a distinction between demand and ut i l i z a t i o n for empirical purposes. The measure i s defined and algorithms employed to operationalize the measure on a particular data base are presented. Drawing from the earlier examination of the conventional micro model, and taking into explicit account the effects of public medical insurance 7 upon the role of price and income variables, Chapter' 7 outlines hypotheses concerning the demand behaviour of individuals, family heads, and families. The model specified by Chapters 6 and 7 is then tested upon a data base described in Appendix A, and empirical results are reported in Chapters 8 through 10. Chapter 11 summarizes the thesis and i t s main conclusions. 1 8 FOOTNOTES 1 . This estimate is based on s t a t i s t i c s published by the Health Economics and Statistics Division of Health and Welfare Canada in National Health  Expenditures in Canada 1960-71 (Ottawa, 1973) p.10, Table 4. The trend is readily apparent in the main component of total health expenditures, expenditures on personal health care, which recorded average annual percentage increases of 11%, 10% and 13% for the periods 1953-59, 1959-65, and 1965-71 respectively. (Evans, 1974) 2. For example, the 1964 Report of the Royal Commission on Health Services (Hall Commission), the 1969 Task Force Reports on Costs of Health Care  in Canada, the 1970 Report of the Ontario Committee on the Healing Arts, the 1970 Castonguay-Nepveu Commission Rapport du Commission d'Enquete  sur l a Saute et le Bien-Etre Social, The Community Health Centre in  Canada (Hastings Report, 1972), and the 1973 Foulkes Report Health Security  for British Columbians. 3. Fuchs analyzed factors contributing to growth in expenditures for medical care in the United States during the period 1947-1967, for which the average annual percentage increase was approximately 8%. 4. This view of u t i l i z a t i o n was suggested almost a decade ago by Feldstein (1966), although he did not envision the stages as being analyzed sequen-t i a l l y ; however, few subsequent investigators have followed up his suggestion in their empirical demand studies, or even ex p l i c i t l y noted the distinction between demand for and u t i l i z a t i o n of medical services which the suggestion implied. (For exceptions see Berki (1972), Fuchs (1973), and Evans (1974). 5. Unless otherwise noted, the term demand hereafter refers to the patient-initiated stage of u t i l i z a t i o n , i.e. the demand for i n i t i a l contacts in treatment episodes. 6. During debates concerning the alleged shortage of personnel, especially physicians, the point is often raised that consumers cannot obtain as much care as they want or need, or that the right service is not a v a i l -able at the right time in the right place. . For present purposes i t w i l l suffice to note that i) this popular definition of a shortage does not correspond to our economic definition, and i i ) both want and need are non-economic concepts. Want may crudely be interpreted as the quantity of services desired by the patient in the absence of any price, and need refers to the quantity of services which expert medical judgement would prescribe for a patient to achieve a given health state. Interested readers can review a lengthy discussion of these concepts in the l i t e r a -ture, beginning with Boulding (1966) and including comments by Jeffers et a l . (1971) (1974), Lloyd (1971), Longhurst (1971), Fuchs (1973), and Mueller et a l . (1974). 7. A number of studies are currently planned or in progress to test the assumption. One of the largest, the Rand Health Insurance Study, de-scribed in Newhouse (1974), proposes to assess the effects of changes in u t i l i z a t i o n on the health status of individuals. 9 8. For evidence on the success of Canadian legislation in reducing i n -equality of access to medical services by income level see Beck (1973) and Enterline et a l . (1973). 9. Patients do retain some control after the i n i t i a l contact however, as evidenced by the phenomenon of patient non-compliance with regimens, broken appointments, failure to follow up a referral, etc.. 10 Chapter 2 THE UTILIZATION PROCESS The process whereby a physician prescribes and supervises a treatment regimen in response to a problem or request presented by a patient was referred to in the preceding chapter as the u t i l i z a t i o n process. Furthermore, i t was suggested that the process be separated into two stages for analytic purposes: i) a patient-initiated stage consisting of the patient's f i r s t contact with the physician in any treatment regimen, and i i ) a physician-generated stage consisting of the remainder of the contacts and services involved in the treatment regimen. This chapter presents the view that the economist's concept of consumer demand can only meaningfully relate to the patient-initiated stage of the u t i l i z a t i o n process. This is due to the nature of the good "medical care" i t s e l f , which is responsible for both problems of price and product definition and the agency role of the physician. The agency role implies that physicians possess discretionary power over the demand for their own services. If so, then the volume of medical services con-sumed by a patient during a given time period, (i.e. the patient's u t i l i -zation), is.a function of both the nature and number of patient requests and physician behaviour in response to those requests. Demand, iri the economist's sense of an autonomous consumer making quantity decisions under a budget constraint and according to his/her preferences, can there-fore refer only to i n i t i a l patient requests for treatment or service. The Nature of Medical Care One of the most popular and controversial discussions in health economics surrounds the view that medical care is a unique market good. 11 Participants in the debate appear to hold one of three positions some, like Mushkin (1962) claim medical care possesses unique characteristics, while others like Lees (1961) claim that medical care has no characteris-tics that are not found in other market goods, and a third group, i n -cluding Klarman (1965) and Wirick and Barlow (1964) argues that while any particular characteristic of medical care may be found in another market good, medical care possesses a unique combination of characteristics. Since the characteristics of medical care have been discussed at length by these and several other authors^", the present discussion is restricted to significant characteristics and, more important, to their implications for models of the demand for medical care. It i s generally agreed that, hypochondria excepted, medical care i s not desired for i t s own sake, but rather because i t contributes toward good health. An individual's need for care i s said to depend upon an unpredictable incidence of i l l n e s s , and few substitutes exist for the necessity called medical care. (This statement should be tempered by the observation that a significant portion of u t i l i z a t i o n does not occur in response to life-threatening conditions, and though few market alterna-tives exist to seeking medical care, one always has the "do-nothing" option, a significant alternative for self-limiting conditions.) External effects i n medical care consumption (e.g. immunizations), entry restrictions on suppliers of care, the existence of not-for-profit institutions (hospitals) in the medical care industry, and the mixed consumption-investment nature of medical care expenditures have also been cited as important characteristics, yet these are perhaps not the most significant ones, and with the possible exception of not-for-profit institutions these characteristics can be found in other goods and i n -12 dustries. Of much more fundamental significance are problems of medical care market equilibrating mechanisms and consumer ignorance. With regard to equilibrating mechanisms, even in the absence of medical insurance the price of medical care is not a well-defined concept. Consider a situation i n which an uninsured consumer requests treatment for a specific set of symptoms. Although the consumer is demanding treat-ment, prices refer only to specific procedures or components of the treat-ment packages and the consumer has no way of knowing ex ante which or how many procedures w i l l ultimately be required i n the course of a given treatment. Consequently, prior to contact with a physician the price of medical care (i.e. the eventual cost of treatment) i s undefined. For some common problems or requests the consumer may develop expertise at predicting cost but i n general this cannot be assumed to be the case. The patient may also exert some control over the u t i l i z a t i o n process, and thus over the cost of care, through non-compliance with the physician's advice or self-medication, though at the risk of prolonging recovery. Even in the. limiting case of f i r s t contacts i t seems incomplete to relate demand to the nominal or l i s t e d price of an office v i s i t , for example, 2 since this i s l i k e l y only the beginning of a treatment package. The conventional role of prices i s further affected by the existence of medical insurance. Though some plans may involve co-insurance or de-ductibles, most plans require the consumer to pay a fixed premium each year and i n return guarantee the consumer access to a range of specified medical services. The premium paid by any given consumer is not directly related to his/her u t i l i z a t i o n although premiums may be adjusted in response to changes in aggregate u t i l i z a t i o n . In this situation neither the l i s t e d price of a particular procedure nor the potential cost of 1 3 treatment can be considered a relevant measure of the price of medical care to the consumer since insurance insulates against both. This i n -sulation has prompted the suggestion that the introduction of medical insurance creates a "moral hazard" effect, i.e. consumers w i l l react to a significant reduction in the effective price of medical care by con-3 suming more care than they in some sense "ought" to.• Many of the d i f f i c u l t i e s in defining the price of medical care are rooted in the problem of product definition. What product is being de-manded by the consumer and supplied by the physician in the medical marketplace? Moreover, do the transactors'perceptions of the product even coincide? One might argue, as Grossman (1972) does, that the consumer's 4 demand is for "good health" ; however, that commodity clearly cannot be obtained solely via a market transaction with a physician. As casual empiricism suggests, recent studies have documented"*, and Grossman him-self implicitly acknowledges, good health i s a function of several variables including l i f e s t y l e and environment in addition to the consump-tion of medical services. Just as good health is too broad a definition of the product de-manded by consumers, so is an office v i s i t or a specific medical procedure too narrow a definition. In effect, what is demanded and can be supplied by a physician is a "treatment", which may be defined as a package of medical services which the physician feels will, resolve the patient's problem or satisfy the patient's request. D i f f i c u l t i e s in product and price definition stem from a common source, namely the consumer's ignorance of his/her own medical condition and the effectiveness and a v a i l a b i l i t y of alternative medical technologies. It i s this information differential between the patient and physician 14 that i s the fundamental distinctive characteristic o.f the medical care market. The consumer's engagement of a physician to act as his/her agent is a direct attempt to eliminate the d i f f e r e n t i a l . The fact that the physician-agent is both the seller of information and advice and the pro-ducer of medical procedures results in effective dual control by the physician of both demand for and supply of medical services after the i n i t i a l contact by the patient. Consequently, the duration, components and eventual cost of a treatment episode are to a significant extent at the discretion of the physician. Discretionary Behaviour of Physicians Before br i e f l y reviewing literature indicating that physicians can and do influence the u t i l i z a t i o n of their services through discretionary behaviour (thereby rendering the demand-utilization distinction an im-portant one), i t should be noted that the issue here is not what motivates physicians to influence u t i l i z a t i o n but rather that, given some motivation, they can influence i t . Although income-generation may be one explanation for discretionary behaviour which increases u t i l i z a t i o n , i t i s by no means the only one. Another hypothesis, forwarded by Fuchs (1973), is that physicians operate under a binding "technologic imperative" which requires them to do "everything possible" for a patient, regardless, of cost to society. Fuchs asserts that while in most markets potential product improvements are weighed against their cost before imple.mentation, in the medical care market the physician feels obliged to produce the best product that existing technology w i l l allow. In reality, both i n -come-generation considerations and the technologic imperative are pro-15 bably operative. Moreover, both imply that physicians exert discretionary power over utilization. The classic investigation on this topic was conducted by Monsma (1970) who compared utilization data from two comprehensive health insurance plans operating in New York city, the Health Insurance Plan of Greater New York (HIP) and Group Health Insurance, Inc. (GHI). Patient samples drawn from subscribers to the plans were closely matched by demographic, social and economic characteristics as well as perceived health status and at t i -tudes towards care. The plans themselves were almost identical. Under both plans the marginal cost of physician services to the patient was effectively zero; however, due to differences in payment mechanisms, GHI physician services had a positive marginal revenue while HIP physician services had a zero marginal revenue. Monsma found that overall utilization was higher in the GHI sample, and that the effect was most obvious in surgery rates. GHI enrollees "demanded" almost twice as many surgical procedures per hundred persons per year as HIP subscribers. Furthermore, the differences between the two groups were most pronounced for procedures such as tonsillectomies and appendectomies which involved the removal of non-essential tissue at l i t t l e risk to the patient. As evidenced by the references he cites, Monsma's conclusion that financial return was at least one consideration in physician advice only lent empirical support to a proposition that had been frequently forwarded by knowledgable observers of the medical care delivery system. Moreover, the Monsma results strengthened similar earlier findings by Densen et al.(1958) (1960) (1962). The Monsma results were also consis-tent with the utilization experience of members of the United Steelworkers 16 of America and the United Mine Workers under insurance plans providing for salaried vis-a-vis fee-for-service physicians.^ In the United Mine Workers study, evidence indicated that surgical u t i l i z a t i o n by patients attending salaried physicians was 37% lower than similar u t i l i z a t i o n by those attending fee-for-service physicians, with an even greater dis-crepancy in the rate of appendectomies. More recently, the significance of discretionary behaviour on the part of the physician emerges from the work of Enterline et al.(1973) in Canada and Fuchs and Kramer (1972) in the U.S.. . The Enterline study examined physician u t i l i z a t i o n rates in Montreal before and after the introduction of Medicare. Findings were consistent with the hypothesis that physicians reacted to the introduction of insurance by adjusting their service mix, thereby generating increased u t i l i z a t i o n from a given number of i n i t i a l contacts. Fuchs and Kramer undertook a comprehensive examination of the determinants of expenditures for physician services in the U.S. during the period 1948-1968. Employing cross-section, state data for 1966, they analyzed variations in quantity of services per capita, physicians per capita, quantity of services per physician, and insurance coverage-. They concluded in general that supply factors (technology and number of physicians) appeared to be of decisive importance in determining the u t i l i z a t i o n of and expenditures for physician services. In particular, they found that income, price and insurance e l a s t i c i t i e s of u t i l i z a t i o n were small compared to the direct effect of number of physicians, even after this effect was adjusted for the possible effect of number of physicians on level of fees. They further found that states with high quantity of service per capita had relatively low quantity of service per physician. One hypothesis which would explain this finding is that where physicians are relatively numerous they reduce the impact of service a v a i l a b i l i t y on volume of services provided by performing more services for each patient.^ Further evidence of discretionary behaviour on the part of Canadian physicians under a fee-for-practice, third-party payment system is, found g in a recent report on health services in Nova Scotia. The report notes that when Medicare severely restricted the conditions under which physicians could b i l l for an office v i s i t and a "pap smear" test separ-ately, the number of tests performed immediately began to decline. Although Evans (1975) mentions the poss i b i l i t y of incomplete reporting under a combined payment for the v i s i t and test, i t i s s t i l l too early to ascertain the extent of this effect. To sum up regarding discretionary behaviour, i t w i l l suffice to say that whether due to some profit or u t i l i t y motive, a technologic im-perative, or as the sincere agent of the consumer, physicians generate u t i l i z a t i o n in the process of legitimizing and satisfying patient re-quests. Furthermore, i f they are able to generate differing amounts of u t i l i z a t i o n from a given number of i n i t i a l contacts, as i t appears, then u t i l i z a t i o n data are inappropriate for the testing of models which postu-late autonomous consumer decision-making and some index of demand i s re-quired. The following note on u t i l i z a t i o n measures currently in use re-inforces this conclusion. Existing Measures of Uti l i z a t i o n Dollar expenditures for medical care and the number of physician 9 v i s i t s , contacts or procedures during a given time period are the two most common types of dependent variable employed in attempts to explain 18 u t i l i z a t i o n , although Aday and Eichhorn (1972) catalogue several variations of these measures. Dollar expenditures for care are usually defined to exclude insur-ance premiums but to include expenditures made by insurance plans on be-half of individuals or families. The most frequent criticism of the dollar expenditure measure is that i t i s subject to bias from variation i n the price of the same services received by different consumers. Systematic price variation could be due to regional price differences, price dis-crimination (including the provision of charity care by the physician), or institutional arrangements through which services are provided at zero or raduced cost to patients with special characteristics or dis-a b i l i t i e s . Price variation may also be due to the provision of similar services (e.g. pediatric advice) by both specialists and general prac-titioners, normally at a higher price by the specialist. Of course, the services of specialists may represent a different product than the ser-vices of general practitioners. In particular, a claim that i s often made but d i f f i c u l t to assess, i s that price differences reflect quality differences. If so, then dollar expenditure measures of u t i l i z a t i o n would offer a convenient method of adjusting for the quality of care. Even in the absence of systematic price variation, however, the dollar expenditure measure i s an inappropriate index of the demand for care. As discussed above, in most cases the consumer cannot be assumed to know the eventual cost of treatment resulting from an i n i t i a l contact. Further, under most public medical insurance plans there i s l i t t l e i n -centive for the consumer to care about the S i z e of the b i l l submitted by the physician to the plan. The v i s i t , contact, and procedure measures share this shortcoming 19 with the dollar expenditure measure, since i t i s not the patient who determines the eventual number of v i s i t s , contacts, or procedures. Con-sequently, these measures are not suitable for use i n models which attempt to explain variations in demand among individuals or families. In ad-dition, however, the heterogeneous nature of v i s i t s , contacts, or pro-cedures makes even u t i l i z a t i o n comparisons between patients or groups of patients misleading. The next three chapters analyze three existing models of the demand for medical care. In each case the studies under review consistently f a i l to recognize the nature of the u t i l i z a t i o n process and overlook the demand-utilization distinction, with the result that either i ) i n -teresting and important hypotheses about consumer demand are tested upon ut i l i z a t i o n measures such as those mentioned in this section, or i i ) attempts to explain u t i l i z a t i o n data proceed without regard for supply side variables. Predictably, the results are unconvincing and at times confusing. 20 FOOTNOTES 1. For example, Fuchs (1966) (1972) (1973), Arrow (1963), Lloyd (1971), Berki (1972), Culyer (1973), Ruchlin and Rogers (1973), and Migue* and Belanger (1974). 2. The relationship between demand and price may be more meaningful i f the patient's request is l i k e l y to be completely satisfied in one v i s i t or i f we consider the price effects of indirect costs to the patient of making the v i s i t e.g. earnings foregone, travel expenses, child care expenses, etc. 3. The "moral hazard" discussion is part of a more general, theoretic interchange on the welfare costs of health insurance, beginning with Arrow's classic a r t i c l e (1966), and continuing with Pauly (1968), Zeckhauser (1969), and others. 4. This doesn't necessarily c l a r i f y the product definition, since i t is extremely d i f f i c u l t to define "good health". In spite of the World Health Organization definition: Health means more than freedom from disease, freedom from pain, freedom from untimely death. It means optimum physical, mental, and social efficiency and well being. (from the Preamble to the W.H.O. Constitution) i t may be better in the present context to interpret good health in a Popperian sense as the avoidance of mortality and morbidity. 5 . Auster et a l . (1969) found U.S. state mortality rates to be more sig-nificantly related to variables such as alcohol and cigarette con-sumption per capita and education level then variables describing consumption or provision of medical care. More recently, Health and Welfare Canada has released a working document (Lalonde, 1974) which analyzes causes of mortality and morbidity in Canada, seriously questioning the impact upon health of the medical care delivery system relative to factors of human biology, environment and l i f e s t y l e . 6. See the Special Study on the Medical Care Program for Steelworkers  and their Families, United Steelworkers of America, (Pittsburgh, 1960) and Roemer and Shain (1959). 7. Whether such services are necessary, or productive, i s an entirely separate question. It may well be that in areas where physicians are relatively scarce, patients have a backlog of unmet need. Con-sequently as the number of physicians increases, the increased services per patient represent a more comprehensive treatment of patients' conditions. At the present time, and until more sophisticated measures of quality are derived, there does not appear to be a d e f i n i -tive way of testing this backlog-of-unmet-need hypothesis. 21 8. Report of the Royal Commission on Education, Public Services, and  Provincial-Municipal Relations Vol. II Chapt. 9 Queen's Printer, (Halifax, 1974). 9. A v i s i t w i l l usually involve at least one contact between patient and physician, but may involve more i f , for example, the physician performs two procedures on the patient during the same v i s i t but with a significant time interval between the two. Patients may also have other types of contacts (e.g. telephone) or may have a contact for a procedure with someone other than the physician (e.g. office nurse). Procedures are specific items such as pelvic exams, i n -jections, etc. which the physician b i l l s for. 22 Chapter 3 THEORIES OF THE "DEMAND" FOR MEDICAL CARE: THE CONVENTIONAL MICROECONOMIC MODEL . * Although health economics as a f i e l d of study remains in i t s childhood, there exists a substantial bank of literature on the demand for and u t i l i z a t i o n of medical services. In fact, even surveys of the topic are becoming numerous. In addition to Lloyd's selective review of economic literature on the demand for. medical care (1971), and a recent bibliography by Aday.and Eichhorn (1972), taxonomies of various approaches to the topic have been suggested by Anderson (1973), Greenlick et a l . (1968), Andersen (1968) , and McKinlay (1972). Unfortunately, a common and fundamental deficiency of these surveys is their failure to focus on the analytic structure underlying empirical efforts to explain the u t i l i z a t i o n process or identify the determinants of demand.^ The usual modus operandi of reviewers i s to f i r s t specify a number of categories of independent variables believed to account for variations in individual or family expenditures on medical care or volume of medical services used, and to then discuss existing empirical results relating to each category. The most popular categories of independent variables have been labelled "economic", "social-psychological", "socio-economic", "socio-cultural", and "organizational", and in this way five corresponding approaches to the problem of "explaining" the demand for medical care have been identified. Of course, by regressing some measure of demand or u t i l i z a t i o n up-on a l i s t of independent variables regardless of how lengthy or sophisticated one has not explained consumer behaviour with respect to medical care. For this, one requires theory supported by agreement between theoretic predictions and empirical observations. Furthermore, 23 failure to specify analytically distinct theories of demand not only com-plicates the interpretation of results of the various approaches, but also camouflages the question of whether or not the problem has been i n i t i a l l y conceptualized or specified accurately. Although i t i s neither obvious from nor recognized e x p l i c i t l y in the literature, three basic theories of the demand for medical care do exist. Most demand and u t i l i z a t i o n studies either implicitly rely on or represent variants upon one of the three. This chapter c r i t i c a l l y reviews the conventional microeconomic model of the demand -for medical care, noting potential improvements in the specification of both independent and dependent variables for later empirical work. Suceeding chapters examine the human capital model and the health belief model respectively. With the exception of the Grossman (1972)-human capital model of the demand for health, a l l economic literature on the demand for medical care is based either exp l i c i t l y or implicitly upon the conventional micro-economic theory of consumer behaviour. Furthermore, many attempts to demonstrate the importance of non-economic variables such as "tastes" and "need" are actually variants of the standard theory. On the assumption that medical care can be treated similarly to 2 any other market good for purposes of demand analysis , the conventional model views the demand for medical care as rational, informed choice under a budget constraint. The relevant explanatory variables are of course prices and income, and to a lesser extent tastes. Although most 3 empirical studies begin by assuming a demand function , i t should be remembered that such demand functions are rooted in a u t i l i t y maximization framework. Maximization of a u t i l i t y function U = U (x^, x^,. . . ,x^,.. ., x defined over n goods (where x is medical care) and satisfying certain 2 4 axioms, subject to a budget constraintp^.Xj=Y, results in a set of f i r s t order conditions which can be solved for x^, thereby yielding a demand function for medical care of the form x^=f(p^, p^, ...» p^, Y), where p_. i s the price of the j*"* 1 good and Y represents income. The significant predictions of the model are that ceteris paribus the quantity of medical care demanded i s inversely related to i t s own price and positively related to the consumer's income. Although empirical evidence supports both predictions, a review of studies emphasizing the role of price and/or income variables indicates that existing evidence must be interpreted cautiously. Price Effects Several studies have confirmed the negative own-price e l a s t i c i t y of physician u t i l i z a t i o n predicted by the conventional model; however, the magnitude of the estimates ranges from 0 to -1. Feldstein and Sever-son (1964) calculated the price e l a s t i c i t y of physician u t i l i z a t i o n to be -.2, while more recent estimates by Rosett and Huang (1973) for a l l ambulatory care l i e between -.5 and -1.0. Newhouse and Phelps (1974), however, found e l a s t i c i t i e s of less than (in absolute value) -.1 for physician office v i s i t s . The finding of e l a s t i c i t i e s approaching -1.0 i s somewhat surprising since the demand for physician services has traditionally been thought to be relatively inelastic due to the essential nature of such services and the a v a i l a b i l i t y of few substitutes. Klarman (1965) has emphasized that observed low price e l a s t i c i t i e s may be the result of price discrimination by physicians and their provision of free care to those who can not afford the going rate. The existence of medical insurance w i l l also tend to produce low price e l a s t i c i t i e s since insurance 25 insulates the consumer against the cost of obtaining care. There are other reasons, in addition to the existence of price 4 discrimination and medical insurance , why one should regard estimates of the magnitude of price e l a s t i c i t y with suspicion. The normal method of estimation relates u t i l i z a t i o n observations (either dollar expenditures on medical services, or number of v i s i t s , contacts or procedures) to observations on the prices of selected physician services (e.g. general examination), or some average price calculated from a selected set of fre-quently used services. In terms of the two-stage view of the u t i l i z a t i o n process already described, estimates derived i n this way are not estimates of the sensitivity of consumers to the. price of physician services. They are instead estimates of the sensitivity of u t i l i z a t i o n , jointly determined by the patient and physician, to the price of services and as such depend as much upon the behaviour of physicians and characteristics of prevailing medical practice as they do upon the behaviour of consumers. Variations i n the quality of care provided also undermine estimates of price e l a s t i c i t y . Since the development of quality measures remains an unsettled area i n the health care evaluation literature, quality i s reluctantly and by default assumed constant across samples, physicians, and procedures. Yet the heterogeneous nature of medical care makes i t easily possible for a physician to provide services differing in quality, though nominally the same, to different patients. To the extent that quality differences account for price differences and higher quality care results in lower u t i l i z a t i o n in a given period, then part of the variation in u t i l i z a t i o n attributed to price effects i s actually due to quality differences. If so, then estimates of the magnitude of price e l a s t i c i t y w i l l be biased. 26 Perhaps the most serious limitations to attempts to estimate demand functions and price e l a s t i c i t i e s are imposed by our earlier conclusion that the price of medical care i s not well-defined, particularly when medical insurance exists. Under public or private medical insurance the effective price of medical care to the consumer may be defined as the out-of-pocket cost for a given treatment package. Out-of-pocket cost con-sists of the sum of the share of premium price attributable to the package, any additional deterrent or u t i l i z a t i o n fees, the cost of any uninsured services (e.g. drugs) prescribed as part of the package, and the indirect costs involved, such as travel costs, child care expenses and foregone earnings. Ut i l i z a t i o n fees, i.e. charges levied for services i n addition to the insurance premium, are sometimes thought to be useful in controlling u t i l i z a t i o n and preventing abuse of services, particularly hospital emergency and outpatient services. Beck (1971) analyzed the impact of Saskatchewan's $2.50 charge for a physician office v i s i t , which was intro-duced in 1968 and removed in 1971, concluding that in this case an increase in the price of medical care did reduce overall physician u t i l i z a t i o n ; how-ever, the overall reduction was the result of significant decreases in u t i l i z a t i o n by lower income families while upper income families exhibited higher u t i l i z a t i o n rates than they had previous to the charge. By now there should be no doubt that insurance is positively and significantly correlated with u t i l i z a t i o n . Wirick and Barlow (1964) , Andersen and Anderson (1967), Andersen (1968), Andersen and Benham (1970), Greenhill (1971), Bice and Eichhorn (1972), Enterline et a l . (1973), and Wan and Soifer (1974) have a l l demonstrated this relationship, to mention only a few such studies. The reasons for this correlation are not so \ 27 clear. Under private insurance plans, i t is d i f f i c u l t to determine whether insurance coverage induces consumers to use a greater quantity of services than they otherwise would or whether people purchase insurance because they have a strong health orientation and recognize themselves as poten-t i a l high utilizers."* It i s also possible that both possession of insur-ance and relatively high u t i l i z a t i o n of medical services are related to some third, and fundamental factor, such as income or education. In any event, the point to be made regarding the price effects of insurance upon the demand for medical care is that in an era of comprehensive and universal public health insurance the persistence of u t i l i z a t i o n differentials, even though Canadian consumers possess standard and approximately equal coverage, indicates that we must look elsewhere for explanation of differences in the behaviour of families and individuals with respect to consumption of medical care. Insurance definitely i n -fluences the level of demand and u t i l i z a t i o n ; however, i t can no longer be held to account for variations in either. On the other hand, the indirect costs of consuming medical care appear to be exerting an increasingly important influence on demand and u t i l i z a t i o n , as evidenced by the present concern of hospital authorities over the use of convenient emergency departments by people with non-emer-gent problems. Indirect or transactions costs of consuming medical care consist of time and money costs incurred over and above the cost of pre-scribed treatment. Travel and child care expenses are two of the more significant money costs incurred by families,. while time may be valued in i t s e l f or for the foregone earnings i t represents. In this regard, the ava i l a b i l i t y of sick leaves, the possession of di s a b i l i t y insurance, and the a b i l i t y to obtain time off work, without loss of pay, to v i s i t a 28 physician are l i k e l y to be significant determinants of an individual's demand, though very l i t t l e empirical research has been completed on these variables to date. It may also be the case that intangible or psychological transactions costs of u t i l i z a t i o n exist for certain groups, such as unpleasant registration procedures in c l i n i c s etc.; however these are d i f f i c u l t concepts to operationalize, as our discussion of the health belief model below il l u s t r a t e s . Research on the indirect costs of u t i l i z a t i o n has concentrated upon the effect of distance travelled to the source of medical care rather than the indirect dollar cost of care per se. While we would expect the former to be correlated with the latter, they may differ due to differences in mode of transit, for example, so that some dollar measure may be a more accurate explanatory variable. Aday and Eichhorn (1972) summarize the work of Bashshur et a l . (1971) , Salber et a l . (1971) and others by concluding that distance exerts a more significant effect upon the choice of a medical care source than i t does upon the volume of services consumed. Working with data on the u t i l i -zation of services by insured, low-income families in two New York neigh-bourhoods, Acton (1973a) has recently suggested that travel and waiting time may have already replaced money prices as a significant determinant of u t i l i z a t i o n for certain groups. In the most rigorous investigation of this to date, he arrives at estimates of travel time price-elasticity of -.25 to -.38 for privately provided care and -.6 to -1.0 for care pro-vided in public c l i n i c s or outpatient departments. The corresponding estimates of waiting-time price e l a s t i c i t y are, as expected, smaller, and are -.05 and -.12 respectively. Later work by Acton (1973b) also demonstrates, contrary to the Aday and Eichhorn summary, that distance 29 travelled to a source of care significantly affects the volume of services u t i l i z e d . Income Effects As in the case of price e l a s t i c i t y estimates, empirical research on the income el a s t i c i t y of medical care expenditures has also tended to support the conventional micro model. The volume of physician services consumed has been found to be positively related to income, although the relationship i s weakening over time due to the spread of health insurance. Early estimates by Stigler (1946), (1952), Feldstein and Severson (1964), and Feldstein and Carr (1964) placed income el a s t i c i t y in the .6 to .8 range. Silver (1970) calculated an income el a s t i c i t y of .85 for employed persons in various U.S. region-age-sex c e l l s . Recently Grossman (1972) has calculated an income e l a s t i c i t y of .7 for total medical care expendi-tures, while Rosett and Huang (1973) present estimates ranging from .25 to .45 for incomes ranging from $4,000 to $10,000.^ Estimates of the income e l a s t i c i t y of medical care consumption are plagued by the same d i f f i c u l t i e s that beset attempts to calculate price e l a s t i c i t i e s . Quality differences and price discrimination undermine the r e l i a b i l i t y of the estimated coefficients, while use of expenditure data produces an income e l a s t i c i t y of u t i l i z a t i o n rather than demand. These and other problems are discussed by Andersen and Benham (1970) in one of the more comprehensive investigations of the relationship between income and medical care consumption. Andersen and Benham examined the effects of ^ selected independent variables and alternative measures of income and u t i l i z a t i o n upon estimated income e l a s t i c i t i e s . Working with 1964 cross-section data from a survey of 2,400 American families, they 30 assumed a demand function of the form D = D(P, Q, Y,. M, V, I) where P represents price, Q quality, Y family income, M demographic characteris-t i c s , V attitude toward preventive care, and I illness level. Two indices of family medical care consumption were employed: i)dollar expenditures on physician services, including charges paid by voluntary health insur-ance programs but excluding health insurance premiums, and i i ) quantity of services consumed, a measure derived by weighting specific services by standard prices based on the "relative values" assigned to such services in the California Relative Value Fee Scale, a scale which enables the stan-dardization of fee schedules for similar services across physicians. In effect, the quantity of service measure is a revised dollar expenditure measure designed to remove any bias present due to systematic price v a r i -ation. Price was measured by the premium cost of a l l health insurance paid directly by the family and by a series of dummy variables intended to capture the effects of different types of health insurance. Quality was proxied by dummy variables representing whether or hot a family had a regular source of care, and i f so, whether the regular source was a general practitioner, specialist, or clinic/hospital outpatient depart-ment. Family income equalled earned and other income as reported by the family; however, an alternative permanent family income series was estimated and later substituted for reported family income. Demographic variables included age, sex, education and employment status of family head, social class, family size, urban vs. rural residence, and southern U.S. vs. northemU.S. residence. The attitude toward preventive care variable was used to group families in which a l l members had had a physical exam with-in the preceding five years. Illness level was determined by the total number of symptoms (from a given checklist) experienced by the family 31 during the survey year. The Andersen and Benham results i l l u s t r a t e the complex nature of the income-medical care consumption relationship. By adjusting for P, Q, M, and V they were able to reduce the estimated income e l a s t i c i t y from .41 to .22, with the price variable (and specifically insurance coverage) having the most significant depressing effect.. The authors concluded that higher income families had higher expenditures because they had more comprehensive coverage and larger families, and made more use of specialists than lower income families. Substituting permanent for reported income, they found, surprisingly, that income el a s t i c i t y dropped from .22 to 17; however, when the estimations were adjusted for illness level, and the accompanying adverse effect of illness on earn-ings, the revised estimate (.30) was in line with expectations. Finally, since the dollar expenditure measure of consumption generally under-stated the volume of services consumed by low income families, Andersen and Benham found that estimates of income e l a s t i c i t y derived using dollar expenditures as the dependent variable consistently exceeded estimates using the quantity of services measure, sometimes by a s i g n i f i -cant amount. Although Andersen and Benham purport to account for variations in family medical care consumption, there is no direct role in their framework for the provider of services. The different organizational settings in which care may be provided enter through the quality measure, but no other supplier characteristics (e.g. place of graduation, years since graduation, prescribing habits) are considered. This conceptual discrepancy could be obviated in one of two ways, either by adding supplier characteristics to the l i s t of independent variables or by retaining the demand function as assumed 32 and developing a more accurate index of the demand for medical care than their dollar expenditure or quantity of service measure. Moreover, what is the rationale for the extended l i s t of explanatory variables i n the Andersen and Benham demand function? They themselves offer a rather weak ju s t i f i c a t i o n for the inclusion of variables in addition to price and income, and the only explanation afforded by appeal to our conventional micro model is that the additional variables somehow measure tastes. This, in fact, i s the avenue that most attempts to explain the demand for medical care have travelled. A core of price and income variables is normally augmented by a taste matrix in an attempt to control for "non-economic" influences. As we shall discover from taste and need variants of the micro model, these "non-economic" independent variables are frequently the most significant empirically. Taste and Need Variants The assumption underlying attempts to estimate demand functions for medical care which include socio-demographic and socio-psychological variables in the estimating equation is that such variables serve as proxies for consumer tastes, although the process of taste formation remains rather fuzzy. Consumer tastes are broadly interpreted to include attitudes, perceptions, values and sometimes need for care. An i l l u s t r a t i o n of the taste variant of the conventional micro model is found i n Wirick and Barlow (1964)^, where the authors begin by assuming that the demand for any market good, including medical care, depends upon the consumer's "wants" in addition to his/her income and market prices. A want i n turn requires three conditions, beginning with the existence of a real physio-logical need and followed by perception (by the consumer) of real or sup-posed physiological or psychological needs, and a willingness to meet f e l t needs by taking action, whether or not the want Is translated into 3 3 a demand depends upon the consumer's a b i l i t y to meet the need, and any additional economic or other constraints. Before proceeding with the Wirick and Barlow argument i t should be noted that they have postulated a demand function of the form x^=x^(p^, p^> p ^ p n, Y, W ) where x± is any market good, p.j^  the price of Y income and W consumer wants. Since the term "want" is not specifically defined i t is not clear whether the demand for x^is influenced.only by want for x. or by a l l wants. Moreover, the theoretic foundation of this 1 8 function i s quite fuzzy; indeed, i t is d i f f i c u l t to imagine how a function of this form could be derived from any u t i l i t y maximization framework. The core of Wirick and Barlow's paper, is their attempt to estimate a consumption function for medical services by regressing the expendi-tures of individuals for medical care on eight explanatory variables proxying the preconditions of wants and the individual's a b i l i t y to translate wants into demands. Thus age and sex were used to represent the existence of needs, while willingness to meet needs was proxied by education, region (urban-rural, north-south) where head grew up, and attitude towards early care. Family income, family size, and health 9 insurance coverage determined a b i l i t y to meet needs. Using this formu-lation of the consumption function they were able to account for approxi-mately 13% of the variation in medical expenses across individuals, with higher expenditures being positively and significantly correlated with age, possession of health insurance, and females. Both high and low income individuals had greater expenditures than middle income individuals, a result attributed to the observation that illness simultaneously reduces income and increases medical expenditures. Perhaps surprisingly, attitudes toward seeking early care were of only slight importance in explaining variations i n expenditures. The authors concluded a discussion of their results by suggesting that the consumption of medical care depended primarily upon needs, although the a b i l i t y to secure care was of some significance at the margin. Such a conclusion i s not surprising when we note that need and ability-to-pay variables dominated the set of explanatory variables and, not coincidentally, were the easiest to measure. The omission from the consumption function of any measure of the perception of needs i s particularly disconcerting when contrasted with the authors' statement that"the consumption of medical care i s based upon f e l t needs, and i f no needs are f e l t , there w i l l be no consumption".^The omission of perception variables illustrates that normal measurement pioblems are compounded when attempting to measure tastes; however, ± n defence of attempts such as this to account for the effect of consumer tastes, i t should be noted that the empirical results do identify relationships and correlations which subse-quent theories w i l l presumably be called upon to explain. Socioeconomic status (SES) or social class i s another surrogate for consumer tastes that has been employed in regression analysis of medical care expenditures, especially expenditures for preventive services. Although there are several specific techniques for constructing socio-economic status scales most rely on some weighted combination of education, income and occupation variables."^The major drawback in employing SES variables to explain the demand for medical care i s the d i f f i c u l t y in se-parating empirically the various channels through which SES affects demand. For example, we may hypothesize that SES effects are primarily income effects, in which case higher SES would be positively correlated with demand due to greater a b i l i t y to pay. Yet we might alternatively hypo-thesize that SES effects are primarily occupation or education effects. In the former case the argument may be that SES and demand for care are positively correlated due to the increased stress accompanying higher 35 status occupations, while in the latter case a positive correlation may-be predicted due to the increased awareness of symptoms and service a v a i l -a b i l i t y accompanying higher education. Wan and Soifer (1974) represent the only attempt that this writer is aware of to analyze the interrelationships of income, occupation and edu-cation in addition to the effects of each upon u t i l i z a t i o n . Their path analysis of the determinants of physician u t i l i z a t i o n in a sample of New York and Pennysylvania households indicates that income indirectly affects u t i l i z a t i o n through i t s relation to health insurance coverage, but has no significant direct effects. Education, on the other hand, appears to be the primary explanatory variable among the three, influencing both income and occupation. If the Wan and Soifer results held in general, then one might wish to replace SES by education alone as an explanatory variable; however there i s some evidence that education and income diff e r in their impact upon physician u t i l i z a t i o n by type of medical service. Coburn and Pope (1974) recently tested several hypotheses concerning the link between SES . and preventive health care u t i l i z a t i o n on a sample of 1,000 employed Canadian males. Although they were unable to provide any convincing results concerning the explanatory significance of SES they did find that income was relatively more powerful in explaining u t i l i z a t i o n variations for general physical exams while education was better able to account for variations in polio vaccinations. Occupational status was by far the least important of the three SES components. 12 Summarizing this and other literature on the relation of SES to the demand for and u t i l i z a t i o n of physician services, we find that higher SES groups use relatively more services than lower SES groups and that this is largely due to greater use of preventive services by higher SES groups. The causal relationship between SES and u t i l i z a t i o n , the relative importance 36 of SES components, and whether or not SES i s an appropriate proxy f o r consumer tastes, s t i l l remain d i f f i c u l t questions about which we know very l i t t l e . Moreover, r e s u l t s concerning the s i g n i f i c a n c e of SES i n explaining u t i l i z a t i o n d i f f e r e n t i a l s may w e l l be heavily influenced by p a t i e n t -physician i n t e r a c t i o n . Pauly (1974), for example, has suggested that physician-generated u t i l i z a t i o n i s l i k e l y to be greater the lower i s the education l e v e l of the patient and the l e s s routine the medical problem, since both f a c t o r s increase the information gap between physician and patient. A second v a r i a n t of the micro model r e s u l t s from the view that con-sumer tastes are i n part determined by an exogenous state of health depen-dent upon the incidence and s e v e r i t y of i l l n e s s . This v a r i a n t , hinted at by Wirick and Barlow (1964), emphasizes the consumer's "need" f o r medical care as the key determinant of demand and u t i l i z a t i o n . Several indices of need appear i n the l i t e r a t u r e , i n c l u d i n g measures of health status, diagnosis, perceived i l l n e s s symptoms, number of d i s a b i l i t y or r e s t r i c t e d a c t i v i t y days, days l o s t from work, and physician-rated urgency of reported medical conditions. While some of these indices involve tech-n i c a l d i f f i c u l t i e s , they convey the basic idea of the need v a r i a n t , which i s that the consumer must be motivated to consume medical care by some divergence between his/her a c t u a l and desired health l e v e l before p r i c e and income v a r i a b l e s become operative. In general, the v a r i a n t hypothesizes that demand f o r physician services i s p o s i t i v e l y c o r r e l a t e d with the f o l -lowing: subjective evaluation of poor health status by i n d i v i d u a l s , an i n d i v i d u a l ' s perception of p a r t i c u l a r symptoms as serious, and r e s t r i c t -ion of the i n d i v i d u a l ' s normal functioning. 13 In an extremely i n f l u e n t i a l book, Andersen (1968) proposed a taxonomy c l a s s i f y i n g f a c t o rs a f f e c t i n g medical care u t i l i z a t i o n i n t o three groups 37 l a b e l l e d "predisposing", "enabling", and "need". He postulated that for u t i l i z a t i o n to occur a family must be predisposed to receive care, pre-v a i l i n g conditions must enable the family to obtain care, and the family must perceive a need f o r care. Family p r e d i s p o s i t i o n was assumed to be a f f e c t e d by family composition (e.g. family s i z e , age of head), s o c i a l structure (e.g. employment of main earner, s o c i a l c l a s s of main earner, education), and health b e l i e f s (e.g. value of good health, knowledge of disease). Both family resources (e.g. income, health insurance) and com-munity resources (e.g. physician/population r a t i o , urban residence) were assumed to a f f e c t the a b i l i t y of the family to obtain care. Need was a function of the incidence of i l l n e s s (e.g. d i s a b i l i t y days, health l e v e l ) and the response of the family to i l l n e s s (e.g. l i k e l i h o o d of seeing a doctor for symptoms). Andersen separated medical care u t i l i z a t i o n i n t o physician, h o s p i t a l , and dental u t i l i z a t i o n noting that d i s c r e t i o n a r y behaviour by the family i s l i k e l y to be greatest f o r dental s e r v i c e s , smallest for h o s p i t a l services and somewhere between the two extremes for physician s e r v i c e s . He further hypothesized that need v a r i a b l e s would make t h e i r greatest c o n t r i b u t i o n i n accounting for v a r i a t i o n i n h o s p i t a l u t i l i z a t i o n , whereas predisposing and enabling v a r i a b l e s were expected to perform better on dental u t i l i z a -t i o n , where there was more scope for d i s c r e t i o n a r y behaviour on the part of the family. He suggested that a l l three categories of v a r i a b l e s would contribute to an explanation of physician u t i l i z a t i o n since physician services represented a mixture of d i s c r e t i o n a r y and non-discretionary s e r v i c e s . Physician u t i l i z a t i o n was measured by both d o l l a r expenditures and quantity of services i n a manner s i m i l a r to that of Andersen and Benham (1970) noted above. 38 A n a l y s i s o f t h e u t i l i z a t i o n e x p e r i e n c e o f a p p r o x i m a t e l y 2,400 A m e r i c a n 14 f a m i l i e s was c o m p l e t e d u s i n g an AID program. P r e d i s p o s i n g v a r i a b l e s were f o u n d t o a c c o u n t f o r r e l a t i v e l y more o f the v a r i a n c e i n p h y s i c i a n use t h a n i n e i t h e r d e n t a l use o r h o s p i t a l u s e , i n t h a t o r d e r . E n a b l i n g v a r i a b l e s were a l s o found t o a c c o u n t f o r r e l a t i v e l y more o f t h e v a r i a n c e i n p h y s i c i a n use t h a n h o s p i t a l o r d e n t a l u s e . A s i m i l a r r e s u l t h e l d f o r need v a r i a b l e s , w h i c h a c c o u n t e d f o r a h i g h e r p r o p o r t i o n o f p h y s i c i a n t h a n h o s p i t a l u s e , and a c c o u n t e d f o r none o f t h e v a r i a t i o n i n d e n t a l u s e . Thus A n d e r s e n ' s h y p o t h e s i s t h a t need v a r i a b l e s w o u l d be most i m p o r t a n t f o r n o n - d i s c r e t i o n a r y s e r v i c e s , and p r e d i s p o s i n g and e n a b l i n g v a r i a b l e s more i m p o r t a n t t h e more d i s c r e t i o n a r y t h e s e r v i c e , was o n l y w e a k l y s u p p o r t e d . However, w i t h r e g a r d t o t h e p h y s i c i a n u t i l i z a t i o n r e s u l t s o n l y , need v a r i a b l e s a p p e a r e d t o a c c o u n t f o r a p p r o x i m a t e l y h a l f t h e e x p l a i n e d v a r i a t i o n w h i l e e n a b l i n g components were l e a s t e f f e c t i v e . S u r p r i s i n g l y , s p e c i f i c v a r i a b l e s s u c h as e d u c a t i o n o f h o u s e h o l d h e a d , o c c u p a t i o n o f main e a r n e r , s o c i a l c l a s s , income and p o s s e s s i o n o f h e a l t h i n s u r a n c e were r e l a t i v e l y i n s i g n i f i c a n t when compared t o need v a r i a b l e s . The o b v i o u s q u e s t i o n a b o u t t h e s e r e s u l t s i s "why?", and t h e f i r s t p l a c e t o s t a r t l o o k i n g f o r o u r e x p l a n a t i o n i s i n t h e measurement o f f a m i l i e s ' need. The p r i m a r y measure o f need was the number o f d i s a b i l i t y days e x p e r i e n c e d by f a m i l y members, i . e . days when th e y were k e p t " i n b e d , i n d o o r s o r away f r o m u s u a l a c t i v i t i e s " . ^ A s i g n i f i c a n t p o s i t i v e r e l a t i o n -s h i p between t h i s measure and p h y s i c i a n u t i l i z a t i o n i s n o t s u r p r i s i n g ; however, t h e r e i s a c a u s a l i t y p r o b l e m . I t may w e l l be t h a t p h y s i c i a n u t i l i z a t i o n l e a d s t o d i s a b i l i t y days r a t h e r t h a n v i c e - v e r s a as t h e A n d e r s e n framework s u g g e s t s . M o r e o v e r , i f d i s a b i l i t y days a r e a r e s u l t r a t h e r t h a n a d e t e r m i n a n t o f p h y s i c i a n u t i l i z a t i o n , A n d e r s e n ' s e m p i r i c a l r e s u l t s 39 could be substantially altered by omission of the d i s a b i l i t y days variable from the explanatory variables in the AID analysis, especially since this variable was at least twice as powerful as any other of the approximately thirty explanatory variables. The Andersen results raise an important point regarding the need variant of the conventional model. If we are to e x p l i c i t l y assume that needs must be present for price and income variables to be operative, then we had best be certain to measure need prior to i n i t i a l contact with a physician. If need i s defined and labelled through contact with a physician, then ambiguous results are unavoidable i n estimated demand functions including need variables. Furthermore, the hypothesis that some families use more physician services because they are "sicker" than other families, (i.e. have a higher incidence of i l l n e s s ) , seems to beg the question of why these families are "sicker". If most u t i l i z a t i o n s t a t i s -tics can in fact be explained by the incidence of i l l n e s s , then the l o g i -cally prior question becomes that of explaining and predicting incidence. On the other hand, the actual incidence of illness may not be as important as the severity of i l l n e s s , or the family's perception of the need for care. In the former case i t may be instructive to analyze the determi-nants of u t i l i z a t i o n by. diagnosis, while the latter case i s dealt with by the health belief model below. Regardless of which revisions are made, the ambiguities associated with the need variant must be removed i f the variant i s to be seriously pursued in future studies. ; Though perhaps the most influential a r t i c l e on the role of need as a determinant of demand and u t i l i z a t i o n , Andersen's is not the only such study .^Results of a study by Monteiro also i l l u s t r a t e the ambiguous causal relation between variables measuring the need for medical care and u t i l i z a t i o n of medical services. Using data from a 1971 survey of approxi-40 mately 1,100 Rhode Island r e s i d e n t s , Monteiro found an i n v e r s e - r e l a t i o n -ship between income and the l i k e l i h o o d that an i n d i v i d u a l had contacted a physician for any reason i n the previous s i x t y days. Adjustments f o r age, w i l l i n g n e s s to see a physician, and mode of d e l i v e r y of care (public c l i n i c vs. p r i v a t e o f f i c e ) did not a f f e c t the r e s u l t ; however, the number of r e s t r i c t e d a c t i v i t y days ( d i s a b i l i t y days) experienced by respondents was found to be negatively co r r e l a t e d with income. From t h i s Monteiro concluded that, although d i f f e r e n t income groups i n her sample exhibited equal contact rates i n response to i l l n e s s , lower income groups experienced a higher incidence of i l l n e s s and therefore appeared more l i k e l y to be u t i l i z e r s . Thus the i n c o m e - u t i l i z a t i o n r e l a t i o n s h i p was dominated by the n e e d - u t i l i z a t i o n r e l a t i o n s h i p . Two observations should s u f f i c e here. F i r s t , the measurement of need by r e s t r i c t e d a c t i v i t y days again begs the question of whether need causes u t i l i z a t i o n or u t i l i z a t i o n defines need. In e i t h e r case, the r e -s t r i c t e d a c t i v i t y days v a r i a b l e i s worthless as a pr e d i c t o r of demand or u t i l i z a t i o n rates due to i t s ex post nature. Second, i n the Monteiro study i t appears that the i n t e r e s t i n g and relevant r e l a t i o n s h i p i s neither i n c o m e - u t i l i z a t i o n , nor n e e d - u t i l i z a t i o n , but rather income-need, which i s not analysed. The Wan and So i f e r study, c i t e d e a r l i e r i n t h i s s e c t i o n , also found need f o r medical care to be a more s i g n i f i c a n t determinant of physician u t i l i z a t i o n than enabling v a r i a b l e s . In t h i s case, however, need was proxied by measures of a household's health status and i t s tendency to react to i l l n e s s by seeking care. The authors assumed that enabling and predisposing v a r i a b l e s a f f e c t e d u t i l i z a t i o n through t h e i r influence on need, an hypothesis which appears to be an improvement over the s p e c i -f i c a t i o n s of Andersen and Monteiro. Consequently, Wan and S o i f e r attempted 41 t o i d e n t i f y t h e f a c t o r s i n f l u e n c i n g h o u s e h o l d h e a l t h s t a t u s and t e n d e n c y to r e s p o n d t o i l l n e s s . A l t h o u g h t h e i r r e s u l t s were i n c o n c l u s i v e , t h e y r e i t e r a t e d t h e common f i n d i n g t h a t p o o r h e a l t h s t a t u s i s p o s i t i v e l y and d i r e c t l y r e l a t e d t o age ( p a s t s i x t y - f i v e y e a r s ) and s e x ( f e m a l e ) . N e v e r -t h e l e s s , by e x a m i n i n g the i n t e r r e l a t i o n s h i p s among th e e x p l a n a t o r y v a r i -a b l e s t h e m s e l v e s , Wan and S o i f e r a p p r o a c h e d t h e need v a r i a n t i n a po-t e n t i a l l y more f r u i t f u l manner t h a n t h e i r p r e d e c e s s o r s . The e x i s t e n c e o f t a s t e and need v a r i a n t s and t h e i m p l i c i t r e l i a n c e upon t h e c o n v e n t i o n a l framework by numerous r e s e a r c h e r s a t t e s t t o t h e w i d e s p r e a d a c c e p t a n c e o f t h e m i c r o model. Y e t t h e model needs t o be r e -examined i n l i g h t o f t h e f o l l o w i n g t h r e e o b s e r v a t i o n s . F i r s t , i t i s a p p a r e n t f r o m t h e p r e c e d i n g r e v i e w t h a t t h e model has been a p p l i e d a l m o s t e x c l u s i v e l y t o u t i l i z a t i o n d a t a even though t h e c o n v e n t i o n a l t h e o r y assumes t h a t o n l y p a t i e n t s ' r e s o u r c e s and p r e f e r e n c e s a r e t o c o u n t . Second, e m p i r i c a l r e s u l t s d i s a g r e e on t h e e f f e c t s a n d / o r s i g n i f i c a n c e o f c e r t a i n v a r i a b l e s s u c h as income and e d u c a t i o n , ^ w h i l e t h e e f f e c t s o f c e r t a i n o t h e r t a s t e and need v a r i a b l e s have n o t been f u l l y s p e c i f i e d o r i n v e s t i g a t e d . F i n a l l y , t h e r e i s a p a u c i t y o f i n f o r m a t i o n r e g a r d i n g t h e p e r f o r m a n c e o f t h e c o n -v e n t i o n a l m i c r o model under c o n d i t i o n s o f p u b l i c m e d i c a l i n s u r a n c e , where one w o u l d e x p e c t to f i n d a r e d u c t i o n i n t h e s i g n i f i c a n c e o f t r a d i t i o n a l p r i c e and income e f f e c t s . The r e s p e c i f i c a t i o n and r e t e s t i n g o f t h e model s u g g e s t e d by t h e above o b s e r v a t i o n s b e g i n s i n C h a p t e r 6; however, two i n t e r v e n i n g c h a p t e r s a r e n e c e s s a r y to c o m p l e t e o u r r e v i e w o f t h e o r i e s o f t h e "demand" f o r m e d i c a l c a r e . 42 FOOTNOTES Lloyd's 1971 review represents an exception to this statement; how-ever, since his review was restricted to economic literature, and preceded Grossman's (1972) important research on human capital his comments relate to only one of the three theories discussed in the following chapters. It i s precisely this assumption which the interdependence of demand and supply i n the medical care market violates. Given this inter-dependence, the basic assumption of the conventional model requires the model to be tested upon an index of demand rather than u t i l i z a t i o n . Nordquist (1970) has proposed a model in which a consumer maximizes a u t i l i t y function defined over present and future consumption and future health states, subject to current and future budget constraints; however, as Lloyd points out (1971; p-81) the model is a complex one since both future income and future health states are random variables. In addition, i t appears that many of the model's predictions are untestable. Grossman's work i s perhaps a clearer example of a model beginning e x p l i c i t l y from a u t i l i t y maximizing framework, yet i t also has serious limitations which w i l l be discussed below. It i s unlikely that these two influences would be operating simul-taneously to depress the calculated price e l a s t i c i t y in any given sample. The practice of price discrimination i s rapidly decreasing as physicians' associations bargain for and participate in the setting of fixed-fee schedules under medical insurance plans. Although hard data bearing on the second hypothesis are not conclusive, Anderson and Feldman (1956) found that families with insurance i n -curred higher out-of-pocket expenses as well as- higher t o t a l expenses for medical care than did uninsured families at every income level. The motivations and factors influencing consumers' purchases of private health insurance i s perhaps one of the few interesting questions remaining in the health insurance literature. Several economists are currently pursuing the topic, most notably Phelps (1973) and Nordquist and Wu (1974). Under public insurance plans, such as the Canadian provincial plans, another interesting question concerns the effects of the introduction of insurance on physician behaviour in selecting components of the patient's treatment package. From an Evans-Fuchs viewpoint, one hypothesis is that the introduction of insurance a l -lowed physicians to increase the services generated from an i n i t i a l contact without fear of worsening the economic welfare of the patient. Recall that low price e l a s t i c i t i e s have been explained by the observation that medical care is a necessity. In fact, i f a l l medical conditions were life-threatening we might expect a zero price e l a s t i c i t y . And in such a case, the income e l a s t i c i t y of demand would presumably be undefined. (e.g. "I only see the doctor when i t ' s important and then money is no object.") In rea l i t y , the supportive and preventive 43 nature of much care lessens the significance of this "health f i r s t " attitude. 7. For another example see Long (1969). 8. In his review of selected economic work on the demand for medical care, Lloyd has also noted the hazy theoretic foundations of early work (1971; p. 35, 40, 47, 81). 9. Wirick and Barlow attach theoretic significance to several other variables (intensity of need, veteran status, neighbourhood income, neighbourhood education, liquid assets, e l i g i b i l i t y for free care, sick leave provisions, race and proximity to services); however, these variables do not appear in the estimated consumption functions. 10. Wirick and Barlow (1964; p.103) 11. For examples see Blishen (1958) (1967), Gordon 0-969), and Green (1970). In this thesis the word "education" refers to formal education, i.e. years of schooling, although this i s often regarded as indicative of the overall amount of general information or knowledge possessed by an individual. In particular, the term "education" does not mean health knowledge or "health education", although knowledge of health matters may be significantly correlated with a person's formal edu-cation level. 12. Notably Koos (19.54), McBroom (1970) , and Bice and Eichhorn (1972). 13. The Andersen framework has been adopted by most subsequent researchers, e.g. Wan and Soifer (1974), Bice and Eichhorn (1972), and Andersen and Newman (1973), and i s the basis for a research bibliography by Aday and Eichhorn (1972). 14. AID stands for Automatic Interaction Detector and is a s t a t i s t i c a l technique which orders independent variables according to answers to the question "What single predictor w i l l give a maximum improve-ment in a b i l i t y to predict values of the dependent variable at any stage of the analysis?" The program consecutively works through sub-sets of the independent variables created by dropping variables i d e n t i f i e d in answer to this question. 15. Andersen (1968; p.100, Table B-3) 16. In addition to the Monteiro (1973) and Wan and Soifer (1974) references discussed here, the role of need variables is also emphasized by Richard-son (1970), Bice and White (1969), Kisch and Kovner (1969) and Gaspard and Hopkins (1967). 17. Although extremely d i f f i c u l t to prove, the disagreement among empirical results may be directly related to the f i r s t observation, since the various samples may have contained differing degrees of physician influence over u t i l i z a t i o n . 44 Ghapter 4 THE HUMAN CAPITAL MODEL A second theory of the demand for medical care, based upon the human capital approach, has recently been forwarded by Michael Grossman (1972). In the Grossman model, medical care is viewed as i f i t were any other market good and is assumed to be combined with the consumer's own time to produce the commodity "good health". The demand for medical care is therefore derived from the demand for good health, which is in turn influenced by the shadow price of health. The quantity of health chosen by the consumer i s the result of a constrained u t i l i t y maximization pro-cess. The model i s referred to as a human capital model for two reasons. F i r s t , expenditures on medical care are undertaken primarily to increase one's stock of health capital. Although health does enter the consumer's u t i l i t y function directly, in the sense that sick time is a source of d i s u t i l i t y , health i s viewed primarily as an investment commodity with the return to investment i n health capital being the increased amount of healthy time available to the consumer for work or leisure a c t i v i t i e s . Second, more educated consumers are assumed to be more efficient producers of health capital in the Grossman world. Increases in a consumer's education level enable increased production of health capital from given inputs of medical care and time in a manner similar to the role of technology in increasing output from given capital and labour inputs. 45 The model may be summarized more formally as follows^: max. U = U (<5 H , .... «5 H , Z , Z ) ( 1 ) o o n n. o n s.t. H i + 1 " H i = Z i " ^ i H i ( 2 ) I. = \ (M., TH±; E.) ( 3 ) Z. = Z. (X., T.; E ±) (4) TW. + TH. + T. + TL. = C (5) 1 1 X 1 n n £ P.M. + F.X. / W.TW. + A 1 1 i i = S 1 l o 4 _ • ( 1 + r ) 1 ( 1 + r ) 1 (6) !=o x ' ' i=o ( 1 ) The consumer's intertemporal u t i l i t y function (subscripts .denote time periods) i s defined over the services yielded by the stock of health capital (H represents the stock of health capital and «5 the flow of health services per unit stock) and the consumption of a l l other commodities (Z). (2) The stock of health capital may be augmented by investment, where net investment equals gross investment in health stock (I) less depreciation. The depreciation rate (£) is assumed to be exogenous and positively correlated with age past some point in the l i f e cycle. ( 3 ) The consumer combines medical care (M) and own time (TH) to produce gross investment. Human capital (E), proxied by education, is assumed to be distinct from health capital (H). Increases in education are assumed to raise the marginal products of medical care and own time proportionally, thus the more educated are more 2 efficient producers of health , and education i s "factor neutral". The production function i s assumed to be homogeneous of degree one in medical care and own time inputs. 46 (4) As in (3) the consumer combines other goods .(X) and own time (T) to produce other commodities (Z) under similar assumptions regarding human capital and homogeneity. (5) The total amount of time in any period (C) must be exhausted by a l l uses, i.e. time spent working (TW), time spent producing health (TH), time spent producing other commodities (T) and time lost from market or non-market activities due to illness or injury (TL). TH i s defined as time spent directly producing health (e.g. v i s i t i n g the doctor) and is assumed to be distinct from TL. Also, —- \ 0 by assumption. If C equals one year then the number of healthy days in year i i s given by ~ TW^  + + TH_/ (6) The present value of outlays on goods, including medical care, must equal the present value of earnings income over the l i f e cycle plus i n i t i a l assets ( A q = discounted non-earnings income) where P i s the price of medical care, F the price of other goods, W the wage rate, and r the rate of interest. Maximization of the u t i l i t y function subject to the constraints given in equations (2) - (6) yields equilibrium quantities of RY and Z^. Since H q and S^ are assumed given, i t is apparent that optimal quantities of gross investment determine optimal health stocks, and attention i s thus shifted to the determinants of gross investment in health. As one might expect, the model suggests that the consumer invests in his/her health un t i l the present value of the marginal cost of making the investment i s equal to the present value of the marginal 47 benefits of the investment. That i s , at the margin:-- ^ - + S i = r - ^ i - l + ^ i <7> I T 1-1 marginal product of health stock marginal cost of gross investment percentage rate of change in marginal cost between period i - l and period i psychic rate of return If s^ = 0, thereby implying that health is solely an investment commodity, then condition (7) states that the monetary rate of return to gross investment in health must equal the price of making the investment, which Grossman terms the user cost of capital. Noting that r,"tC, and ^ are a l l independent of the health stock,^ we can see that the right-hand side of condition (7) defines an i n -f i n i t e l y elastic supply curve for health capital. With the assumption of diminishing marginal productivity of the health stock, the left-hand side defines a downward-sloping demand curve for health capital, since is assumed independent of H . The determination of a consumer's optimal health stock for any age 1 is shown in Figure 1. Where G = "tf -s i = Figure 1 Determination of Optimal Health Stock for Any Age i Source: Grossman (1972; p. 12) 49 In this human capital model of the demand for health, the shadow price of health depends upon the price of medical care (P) and the consumer's age ( i ) , wage rate (W), and education (E). Since the demand for medical care (M) i s derived from the demand for health (H) there are two sets of qualitative comparative static predictions indicating the effects of shifts in P, i , W and E upon both M and H. Consider f i r s t the effects of an increase in the price of medical care. Such a price increase raises the marginal cost of gross invest-ment in health and ceteris paribus adversely affects both the equilibrium health stock and, through the gross investment function, the demand for medical care. Technology permitting, the latter effect would be reinforced by a substitution of time for medical care in the gross investment production function. In order to generate unambiguous predictions with regard to shifts in age i t is necessary to assume that E and W, in addition to TV and G, are independent of age.. Under these conditions 0 since rising depreciation rates increase the marginal cost of investment. (In fact, the Grossman consumer dies when health stock f a l l s below some minimum level, presumably due to the "price" of health becoming too high as depreciation rates rise.) The effect of growing old on one's demand for medical care i s more complex. From the J T V \ 0 result, and the knowledge that optimal health stock i s determined by optimal gross investment, we would expect a consumer to demand less medical care as he/she ages; however, as he/she grows old a given amount of gross investment produces a smaller and smaller net addition to the health stock due to rising depreciation rates. Therefore the consumer has an incentive to offset part of the. reduction i n health stock, caused by rising depreciation rates, by increasing gross investment and demand for medical care. It can be seen from Figure 1 that, to the extent that the demand curve for health capital i s inelastic, the negative effect on demand for medical care caused by rising marginal cost w i l l be dominated by a positive effect as the consumer attempts to compensate for the smaller net additions to health of a given gross investment. Thus — > 0 as 0 1 D Shifts i n the wage rate affect both the cost and the benefit side of the consumer's decision to invest in health. An increase in the wage rate raises the monetary return to investment by increasing the value of healthy time, whether i t be work or leisure time, since the wage rate i s the opportunity cost of leisure. On the other hand, since time i s one of the inputs in the ;gross investment production function, an increase i n the wage rate increases the marginal cost of investing in health; however, as long as some of each input M and TH must be used in the production process, the increase in the value of healthy time w i l l exceed the increase in marginal cost and / 0, ? MvO. Technology permitting, the second result would be reinforced as medical care i s substituted for time i n the production process. With regard to education, Grossman assumes that W, G, r, and % are a l l independent of E. The assumption that the wage rate i s i n -dependent of education, lik e the assumption of independence between age and both education and the wage rate, i s rather d i f f i c u l t to accept, consequently we shall relax these assumptions momentarily. Retaining ^H v them for the moment, however, i t can be seen that rz—=rs 0 since increases • E in education (human capital) reduce the marginal cost of investment by raising the marginal products of medical care and time in the gross investment production function. This result, and the knowledge that H is determined by I, imply that increases in education w i l l have a positive effect upon the demand for medical care; however, this over-looks the fact that given amounts of medical care and time now produce greater additions to the health stock due to the higher education level of the consumer. Consequently, some of the investment called for by the ^ ^  ^ 0 result i s already supplied due to the change in production tech-nology, and the pressure to increase demand for medical care i s lessened. To the extent that the demand curve for health capital i s inelastic, this incentive for the more educated to offset part of the education-induced increase in their health by reducing their purchases of medical M services w i l l dominate. Therefore, the model predicts ~. „ Co as <1. & bi D Although by far the most elegant piece of work to date on the demand for medical care, the Grossman model warrants substantial c r i -ticism. The empirical performance of the model w i l l be examined f i r s t . This w i l l be followed by specific comments on analytics and criticism of the fundamental approach of the human capital model. Grossman tests the model on survey data obtained from U.S. families, selecting a subsample of 1800 whites i n the labour force, of whom 550 comprised the f i n a l data set by virtue of the fact that they reported sick time during the preceding year. The relevant- dependent variables for the estimations were health, as measured by restricted activity days or work-loss days, and purchases of medical services, as measured by dollar expenditures on a l l types of medical care. The relevant independent variables were age of the individual ( i ) , number of years of formal schooling (E), weekly wage rate (W) adjusted for weeks 4 worked and family income (Y) adjusted for transitory components. Price variables were omitted since there was assumed to be no significant price variation across the sample, and income was included to indicate an individual's command over resources, in the absence of wealth data. The following table provides a comparison of Grossman's predicted and estimated relationships."* Table I Comparison of Predicted and Estimated Results  of the Human Capital Model numerator ^H bM predicted estimated predicted estimated denominator b i - - (*) + as ED<1 + (*) bW + + (*) + -* E + + (*) - as €-D<l + &P - -7) Y - + (*) * significant at .05 level. The performance of the model on the medical care equation i s of particular interest. Results support the already known fact that age and medical care u t i l i z a t i o n are positively and significantly related. On the other hand, the predictions concerning education and the wage rate are not supported. Both estimated relationships are insignificant and have the wrong sign. The education result is particularly inter-esting since the results of other studies support the finding of a positive relationship between education and u t i l i z a t i o n . ^ In both cases Grossman cites measurement errors as a potential explanation for the unexpected findings. The estimates for the health stock equation f u l l y support the predictions of the model, and contain the interesting result "o Y In terms of conventional theory this indicates that health i s an inferior good; however, Grossman offers another explanation. He sug-gests that the income e l a s t i c i t y of goods such as alcohol, tobacco, etc. which affect health adversely may exceed those of medical care and other positive influences on health. If so, health w i l l be found to be inversely related to income. Furthermore, Grossman demonstrates that a strength of the model i s i t s ability, to predict, under certain conditions, that an individual w i l l demand less health but more medical care. This underlines what i s undoubtedly the main strength of the model i t s a b i l i t y to generate testable predictions. Yet an exami-nation of several of the assumptions of the model illustrates the high price, i n terms of the restricted application of the model, that i s paid for those predictions. Recall that the wage rate was assumed to be i n -dependent of the stock of human capital, and both were assumed to be independent of age. Relaxation of the f i r s t assumption to allow a positive correlation between W and E does not affect the qualitative comparative static predictions of the model regarding health since the effects of W and E are mutually reinforcing; however to the extent that the health capital demand curve is inelastic, this relaxation ^ M renders predictions for -^-^ and -^-g- ambiguous. Table I also indicates that i f education and/or the wage rate i s positively correlated with age, the predictions for 4*-y- and, in the case of education, are ox • o i weakened. 54 Two further restrictions are the assumption of constant returns to scale i n the production of health capital and the assertion that time spent producing health can be distinguished from time lost due to illness or injury. Regarding the f i r s t assumption, i t would be interesting to examine the decreasing returns case, which may be a more r e a l i s t i c assumption about the production of health capital. In the second case, i t would certainly seem that time spent v i s i t i n g the doctor for a specific condition i s both TH and TL. Further, i f time lost from work but spent convalescing i s considered to be spent producing health, then the cost of producing health i s increased and presumably the optimum health stock i s reduced. With regard to the gross investment production function, Grossman acknowledges that i f desired gross investment equals zero, the model breaks down since: W i G i <T - f f . . + i . for I. = I. = 0 Furthermore, gross investment cannot be negative since there are no capital markets for health. A different problem arises i f gross i n -vestment requires only time inputs, which is not an unrealistic s i t u -ation for many common conditions. If time and apple juice are sufficient to raise one's health stock (and expenditures on apple juice are not considered expenditures on medical care) then the model predicts £H - 2>M ~JTw = ^' "bW = ^ s i n c e the change in the value of healthy time i s completely offset by the change in the cost of producing such healthy time. Another interpretation of this situation might be that depreciation rates are negative, (i.e. by doing nothing one increases his/her health stock), yet the model does not seem to allow for negative values of ^ . Finally, i t has been suggested by Pollak and Wachter (1973) that i f household decisions regarding time allocation reflect not only produc-tion considerations but also direct household preferences as to the use of time, then the household production function approach does not provide a satisfactory account of time allocation. Consequently, to the extent that an individual has direct preferences regarding TH, TW, T, and TL the theoretic underpinning of the Grossman model is weakened. The model i s weakened further by the amount of information that the consumer must be assumed to possess. In order to make appropriate investment decisions the consumer must be aware of the range of medical services available and know their relative effectiveness, i.e. their a b i l i t y to augment his/her health stock and consequently increase healthy time available i n future periods. It has already been noted that this i s precisely the information the consumer lacks and is part of the reason a physician is retained. The human capital model also makes a subtle shift from health to healthy time as the product being demanded by the consumer. The analytic differences between the two are significant. Health, or health stock, i s presumably physically unbounded. It i s constrained only by the cost of acquiring health capital and the consumer's re-sources. Yet healthy time i s bounded in the model by 365*5 days. In the human capital model, two individuals would apparently be defined as possessing the same amount of health capital i f TL = 0 for both. A definition such as this neglects the important effect of health upon quality of time and l i f e , yet illustrates the d i f f i c u l t y associated with defining the product being demanded by the consumer, a point raised 56 in Chapter 2 and one to which we shall return in Chapter 6. The fundamental approach of the Grossman model may be c r i t i c i z e d on the grounds that the importance of health extends beyond i t s effect on one's earning a b i l i t y or amount of leisure time. Note that the model's applicability i s reduced with respect to several significant categories of both consumers and medical services. Since the wage rate i s a crucial variable i n the model, determining the monetary return on i n -vestments in health, the model w i l l experience d i f f i c u l t y accounting for the health and medical care behaviour of children, the unemployed, the retired, and housewives groups which, collectively, l i k e l y account for the bulk of medical care u t i l i z a t i o n ! Since many medical services do not, and are not intended to, directly increase the amount of healthy time available, the model presumably does not apply to purely supportive or informative medical services, nor to pain or discomfort relieving services or such services as allergy injections, treatment for minor burns or lacerations, maintenance of chronic conditions, or obesity counselling. Yet one may well argue that services of this nature, (i.e. services with l i t t l e investment potential), comprise the major part of a family physician's practice. Although i t i s recognized that Grossman posits ah "as i f " model, one suspects that the consumer is not motivated to purchase medical services by return on investment in health capital considerations, but rather by a desire to alleviate pain and discomfort or to relieve anxiety and uncertainty regarding symptoms. Even in the case of ser-vices which resolve life-threatening problems i t Is d i f f i c u l t to accept the notion that attempts to extend one's l i f e are investment-oriented. On the topic of the consumer's motivation, Berki (1972) has commented that, for the consumer faced with a medical problem, i n i t i a t i o n of a uti l i z a t i o n episode i s a move taken reluctantly, but taken because the physician i s the "only game in town". Maximization of health outcome i s not foremost i n the consumer's mind. He/she i s quite l i k e l y satisfied to minimize the loss and/or regret involved in the game. One f i n a l and important observation w i l l conclude this chapter. Like i t s predecessors in the conventional micro framework, the human capital model i s a model of autonomous consumer decision-making which implicitly, prohibits demand-supply interdependence. And, also like preceding models, i t has been empirically misspecified• The demand for medical care i s said to depend, via the demand for healthy time, upon the price of care and the consumer's age, education, wage rate and income. Yet these independent variables are regressed upon ut i l i z a t i o n data, even though u t i l i z a t i o n i s determined by physician behaviour in addition to demand. 58 FOOTNOTES 1 . Since readers may not be familiar with the Grossman model, and since i t does receive substantial criticism in this chapter, i t seems only f a i r to summarize the model in some detail. Complete mathematical derivations of the results and predictions reported here can be found in Grossman (1972; pp. 13-30 and pp. 8 4 - 1 0 1 ) . Those familiar with the model may wish to skip directly to the discussion on page 4 9-2. Note that this i s only one possible explanation for the frequently observed positive correlation between health and schooling. It may also be the case that good health enables students to achieve higher educational levels or that the health-education relation-ship i s the result of the relationship of each to some third variable, e.g. income. 3. £. and r are independent of H. by assumption; "if i s independent of H due to constant returns to scale in production and given input prices. 4. A detailed explanation of the empirical formulation of the model is offered by Grossman i n Chpt. IV pp. 39-54. 5. For detailed results, co-efficients, t-ratios, etc. see Grossman pp. 55-57. 6. For a review of studies finding a positive relationship between education and u t i l i z a t i o n of physician services see Aday and Eichhorn (1972; pp. 1 9 - 2 0 ) ; however, they also note that average length of hospital stay is shorter for the more educated. Since the Grossman results are based on expenditures for a l l types of medical care the estimated positive relationship between demand and education could be attributed to hospital expenditures dom-inating total medical care expenditures. The result could also be explained by a relatively elastic demand curve for health capital, but Grossman finds evidence that the curve i s relatively inelastic. 59 Chapter 5 THE HEALTH BELIEF MODEL Whereas the preceding model can be attributed to a single author, the health belief model of the demand for medical care, as described by Rosenstock (1966), i s a loose synthesis of the efforts of many social-psychologists to understand and explain the process of seeking care. Although the variables involved in the model are for the most part non-economic, the model deserves attention since i t has influenced the taste variant of the conventional micro model^ and examined the patient-initiated stage of the u t i l i z a t i o n process; i n some detail, viewing this stage as a series of sequential decisions on the part of the patient and his/her family. Unfortunately, compared to the preceding models, the health belief model i s by far the least rigorously formulated, which makes review of i t s analytic structure and empirical performance more d i f f i c u l t . Consequently this chapter begins with a general statement of the model, and follows this with a schematic explanation of i t s components, an examination of selected empirical results relevant to the components, and a brief critique. The health belief model was originally formulated to predict families' preventive health behaviour, such as obtaining vaccinations and annual check-ups; however, two of i t s foremost advocates, Becker (1972) (1974) and Rosenstock, imply that i t can also be applied, with some 2 adjustment, to illness and sick-role behaviour . It concentrates upon how people come to realize they are i n need of medical services, and how they decide to seek care. The model specifies that for demand to occur an individual or family must 1) recognize and accept that they have a 60 health deficiency, 2) perceive that the benefits of seeking care i n the formal medical care delivery system outweigh: a l l economic, social and psychological costs, and 3) be prodded or triggered into action by some specific stimulus, called a "cue". The likelihood of these conditions being satisfied i s increased to the extent that the individual or family harbours "positive" health beliefs. For example, the l i k e -lihood of an individual demanding medical services i s increased i f he/ she believes any or a l l of the following: i ) that medical science can cure most problems that people experience; i i ) that one should see a doctor for a l l symptoms; and i i i ) that a person should have regular phys-i c a l examinations, even i f he/she feels healthy at the time. Demand i s also hypothesized to be positively related to the range and accuracy of the consumer's knowledge of medical symptoms and service a v a i l a b i l i t y . While i t may be tempting for an economist to dismiss a model of this nature by saying that i t i s only an explanation of taste formation, a social-psychologist, on the other hand, may claim that economic variables are but a minor and relatively unimportant part of the entire model. Considering the loose formulation of the model, there is no way to assess these competing claims. Since the model does not easily lend i t s e l f to algebraic formulation, a crude schematic interpretation of the major components is presented in Figure 2. Perhaps the best description of the core of the model i s that the prospective patient i s assumed to perform an a l l - i n c l u s i v e , medical cost-benefit analysis before seeking care. The benefits of u t i -l ization are the economic and non-economic costs of the condition which 3 are thus avoided, plus any other benefits of becoming a patient , while the perceived barriers to u t i l i z a t i o n may be psychological, financial or Figure 2 A Schematic Interpretation of the Health Belief Model Perceived benefits of u t i l i z a t i o n less perceived barriers to u t i l i z a t i o n Perceived susceptibility to particular medical condition times perceived severity of that condition Modifying Factors Demographic Social-psychological Health Beliefs and Knowledge age social class prior experience with sex peer group pressure condition race personality knowledge of symptoms ethnicity etc. belief in effectiveness of etc. treatment attitude towards physician etc. Cues to Action mass media campaigns illness of relative or friend reminder c a l l from physician etc. , Likelihood of Demand-ed 62 institutional. The negative consequences of not seeking care are best thought of as the product of the individual's perceived probability that he/she actually has a particular condition and his/her perception of the severity of the condition i n question. Perceptions of benefits and costs are assumed to be modified by an array of demographic, social-psychological, and health belief variables. On the assumption that the outcome of the modification process leaves the individual psychologically ready to act (i.e. benefit-cost ratio^>1) a l l that i s required to cause demand i s some stimulus or trigger. Two aspects of the model i n particular should be noted. F i r s t , the model almost exclusively emphasizes the subjective states of the prospect-ive patient. What is important i s the individual's perceptions of susceptibility, severity, benefits and barriers rather than the actual measurements. Second, the model specifies entry into the medical care system as the result of a sequential decision-making process on the part of the individual. Although the relevant variables would have to be red-uced to manageable proportions, and significant causal relationships specified in much more detail, the dynamic tone of the health belief model may be what i s ultimately required for a general theory of the u t i l i z a -tion process. Though pure speculation at this stage, i t might prove f r u i t f u l to model u t i l i z a t i o n as a series of decisions regarding entry, continuation of treatment, and termination of treatment. It may be possible to erect a general framework around two sets of interacting variables, one set under the control of the physician, one under the control of the patient. Decisions regarding entry, continuation and termination might then be modelled separately, perhaps employing different combinations of variables 63 from the two original sets for each type of decision. From a soc i a l -psychological viewpoint, Suchman (1965) has already broken the u t i l i z a t i o n process into five stages, analyzing social and medical factors which influence the behaviour of the patient at various stages. It may ultimately be useful to attempt a redefinition of the stages of the process in economic terms and to then examine the factors affecting the behaviour of both the physician and the patient at each stage. Although the health belief model remains i n i t s formative stages, several weaknesses are apparent. Put simply, the model i s at once both too vague and too grand. It contains such a large number of variables that almost any demand or u t i l i z a t i o n s t a t i s t i c s can be rationalized by the model. Yet the relationships among the components of the model and the specific roles of explanatory variables are too ill-defined to allow the model to generate testable predictions. For example, the model pre-dicts that people who have positive health beliefs are l i k e l y to demand more medical care than those who do not. Yet i f one could design an experiment to test the proposition, and found no supporting evidence, the finding could be explained by the fact that the individuals in the experi-ment were not subject to "cues to action" of sufficient strength. Given the range of possible cue mechanisms, one problem in ever designing a crucial experiment becomes obvious. And i t i s therefore d i f f i c u l t to know where to grasp the model in order to shake out testable predictions. Under these circumstances, i t i s not surprising that researchers have attempted to isolate small pieces of the grand synthesis for further 4 study of the relationships between specific variables. Of particular importance are studies of the manner and direction in which "modifying factors" affect perceptions of susceptibility, severity, benefits and 64 barriers, since the model contains few behavioural assumptions and yields no testable predictions concerning this component. In a classic investigation, Koos (1954) found clear differences by social class in the sensitivity of individuals to symptoms of conditions requiring medical care. Business or professional persons were much more lik e l y to define a given condition as requiring care than were ski l l e d or semi-skilled workers, and similarly for skilled or semi-skilled workers vis-a-vis unskilled labourers. Freidson (1961) has demonstrated that members of different social groups obtain both lay and professional diag-noses and professional care through different . networks of relatives, friends and other information channels. Culture and ethnicity, i n addit-ion to social class, have been shown to be significant modifying variables. Studies by Zborowski (1952) and Zola (1966) have convincingly demonstrated that personal recognition of and response to symptoms may be culturally conditioned. Although these variables are not in the economist's normal sphere of operations, i t i s important that economists recognize the potential of such non-economic variables to dominate empirical results. Evidence presented by Berkanovic and Reeder (1974) reinforces this caveat. While presenting their reservations concerning models which rely upon either a b i l i t y to pay or perception of symptoms to explain u t i l i z a t i o n , the authors cite a sample in which 30% of the respondents reported not having gone to a doctor recently although they f e l t they should have. Only 10% of the reasons offered for failure to see a doctor involved economic considerations. Most reasons given related to negative a t t i t -udes toward physicians and medical care f a c i l i t i e s , or other negative health beliefs. 65 Although the studies^ mentioned above, and other studies of specific relationships within the synthesis, point to the importance of positive health beliefs, conflicting evidence has been presented by Wirick and Barlow (1964), Andersen (1968), and Aday and Eichhorn (1972). In their attempt to allow for the effects of consumer tastes as well as price and income variables, Wirick. and Barlow specified variables measuring an individual's willingness to seek care. They found these variables to be relatively insignificant in explaining variation in expenditures on medical care. Andersen went even further, specifying variables designed to measure family knowledge of disease and the value a family placed upon health services, physicians, good health and health insurance. These variables, singularly and collectively, were also insignificant in explaining expenditures on medical care. Finally, in a review of approx-imately twenty studies on the relationship of health beliefs to utilization, Aday and Eichhorn similarly concluded that health belief variables were of l i t t l e significance in explaining utilization differen-tials . In defence of the health belief model, i t should be noted that the model purports to predict entry into the medical care delivery system, not expenditures; however, in the absence of any demand-utilization distinction or suitable measures of demand i t is not surprising to find the model applied to utilization. Conflicting results on the importance of health beliefs are under-standable given the imprecise nature of the model and measurement difficulties associated with several of the key variables. As Rosenstock (1966) himself concedes, experimental manipulation of health beliefs and perceptions of costs and benefits of utilization has not been undertaken to any great extent, empirical work on the nature and role of cues is rare, 6 6 and many studies supporting the importance of health beliefs have been based upon small and possibly non-representative samples. Moreover, measurement of health beliefs has often been conducted retrospectively, (i.e. individuals were surveyed at the end of a u t i l i z a t i o n period about their beliefs), thereby admitting possible bias due to beliefs changing through contact with the medical care system during the u t i l i z a t i o n period. The inescapable subjectivity of survey responses regarding health beliefs further renders comparison of empirical results between samples less meaningful; however, this bias could be reduced by standardizing questions concerning beliefs across studies. Although the health belief model forwards several specific hypotheses concerning factors influencing the demand for medical care, at the present time the overall model does not appear to be a testable construct. Several revisions and detailed specifications would be required to make i t testable. F i r s t , specific elements comprising the range of benefits and costs would have to be enumerated and a method devised to quantify each. Next a set of behavioural assumptions would have to be formulated to indicate the manner in which modifying variables act. At this stage i t would be helpful i f the number of relevant modifying factors could be reduced.. .. It would also be necessary to define the set of acceptable cues, i.e. cues assumed to be sufficiently strong to trigger, entry, given that the person was psychologically ready to i n i t i a t e care. Assuming that some unambiguous predctions could be generated from such a tightened model, what information would be necessary to test these predictions? Ideally, one would begin with baseline information on the socio-demographic and economic characteristics, health status, and specific health beliefs of both a sample and a control group. Then and here i s the catch for a given set of medical services one would 67 l i k e to be able to alter the beliefs of the sample group, confirm the alteration, and observe resulting differences i n demand between the sample and control groups when both were subjected to identical, controlled cues. A crucial experiment would also insure that the groups received no other demand stimuli. The health belief model completes our review of theories of the "demand" for care. In summary, the health belief model, and economic models represent the two basic approaches to explanation and prediction of the demand for medical care. To date, each has played a somewhat limited role i n the other. Partially because price and income variables are more convenient policy levers than perceptions and health beliefs, greater attention has been devoted to the economic than the social-psycho-logical approach. This thesis certainly continues the trend, but recognizes the potential relevance of alternative, non-economic approaches. 68 FOOTNOTES 1. For example, Andersen (1968), and Wirick and Barlow (1964). 2. Kasl and Cobb (1966) define, health behaviour as action taken by a healthy person to remain healthy, whereas illness behaviour refers to action taken to identify one's health state and seek treatment i f necessary. Sick-role behaviour refers to action taken toward recovery by persons who consider themselves i l l . 3. These "other benefits" could be economic (e.g. tax deductions), but are usually thought of as non-economic in nature. For example, Shuval (1970) concluded that a significant portion of u t i l i z a t i o n in Israel could be explained through non-health related behaviour, i.e. people went to physicians because having oneself labelled as " i l l " in some way legitimized failure, or v i s i t i n g the doctor gave one ascribed status in the community. Similar themes are touched on by Suchman (1965), Cole and Lejeune (1972), and Antonovsky (1972). 4. McKinlay (1972) provides an excellent review of the social-psycho-logical literature on this topic. Since the health belief model does not e x p l i c i t l y enter the thesis research, only a few art i c l e s are mentioned here in order to i l l u s t r a t e the general nature of existing empirical work on the model. 6 9 Chapter 6 MEASUREMENT OF DEMAND  THROUGH EPISODES OF MEDICAL SERVICE The f i r s t half of this thesis has demonstrated that the construct-ion and testing of any model of consumer demand for medical care must begin with an accurate view of both the u t i l i z a t i o n process and the product demanded by the patient and supplied by the physician. The second half proceeds to estimate demand functions for medical care, based upon the conventional microeconomic model as i t might be inter-preted under prevailing conditions of universal and comprehensive public medical insurance in Canada. Since the dependent variable for these estimations must be a measure of demand rather than u t i l i z a t i o n , this chapter develops both a definition and a measurement technique designed to produce an index of demand referred to as the "episode of medical service". Definition of Episodes An episode of medical service i s defined as a set of medical services received continuously by a patient in response to a particular request. Three aspects of this definition require elaboration and j u s t i f i c a t i o n . F i r s t , an episode begins with a patient request. It may be a request for information, care of some acute condition or preventive care. The consumer may be quite aware of the problem or may only suspect that some-thing i s "wrong". The situation presented by the consumer for the physician's attention may vary in degree of complexity and/or severity. But there i s presumably some reason for each i n i t i a l contact which manifests i t s e l f in a consumer request. In other words, the consumer's demand i s for 70 what we have, up to now, referred to simply as "treatment" A, i.e. the resolution of a specific problem or the satisfaction of a specific request. The consumer does not normally demand a specific medical procedure nor $X of medical care, as i f medical care was oranges. In fact, there i s every reason to believe that'in the absence of knowledge regarding the effectiveness and relative discomfort of alternative procedures, the consumer Is either indifferent to the combination of services used to produce the treatment or, for any given request, prefers a treatment involving less procedures to one involving more procedures. Furthermore, assuming the existence of transactions costs of v i s i t i n g the physician, unless medical care i s defined as treatment, or satisfaction of patient requests, normal economic agents would be irrational i n preferring more to less medical care, ceteris paribus. Even i f the consumer demands something more than an episode of medical service, say good health for example, the most a physician can provide during any given time period i s a package of preventive or diagnostic services, or a package of services designed to relieve symptoms or provide information, and possibly reassurance, i n the absence of symptoms. Acknowledging the uncertain effectiveness of many medical procedures and the role of factors such as patient l i f e s t y l e s , willingness to. follow instructions, and biologic characteristics, i t therefore seems reasonable to suggest that, in addition to being a measure of demand, the episode of medical service i s an acceptable definition of the product supplied by the physician^. Note also that i t is the concept of the set of services, rather than the total number of services comprising the set, that i s Important. Since patient-physician interactions vary in duration and occur periodically as 71 determined by patient requests, the product demanded and supplied i s termed an episode of medical service. Since a request i s presented i n  an i n i t i a l contact, and since the number of episodes by definition bears  a one-to-one correspondence to the number of requests, the episode of  service i s a measure of demand consistent with a two-stage view of the ut i l i z a t i o n process. Specifically, the number of episodes of medical  service experienced by a family or individual i s an index of that consum- ing unit's demand for medical care. The provision of services i n a relatively continuous flow i s the third important aspect of the definition of an episode of medical service, for this continuity distinguishes an episode of service from an episode of il l n e s s . While the development of the episode of service measure in relation to models of demand and u t i l i z a t i o n represents an analytic and empirical innovation on the economic side, the notion that illness (and consequently care of illness) i s episodic has been previously noted on • 3 the c l i n i c a l side . Indeed, one of the strongest endorsements for the use of episodes of medical service as a measure of demand is that they correspond to the natural history of medical care. It has long been recognized that il l n e s s i s episodic i n nature, and, as Richardson (1970) noted, the use of episodes of illness as a health measure allows meaning-f u l comparison of instances of "getting sick". In his study of the urban poor's use of physician services, Richardson employed episodes of i l l n e s s , which he retrospectively defined from survey data on the basis of at least two consecutive days withdrawal from usual activity by respondents. If our focus i s upon the economic significance of demand and u t i l i z a t i o n rather than the contribution of services to health status, then service episodes are more useful than illness episodes; however, i t should be 72 noted i n any event that service episodes d i f f e r from i l l n e s s episodes in two significant respects. First,the beginning and end of illness episodes are more d i f f i c u l t to identify, and do not necessarily coincide with, the i n i t i a t i o n and termination of service episodes. For a given episode, the onset of illness presumably precedes service, while i l l n e s s may terminate either before or after service. An example of service extending past the illness episode would be a patient returning for a call-back v i s i t after completing a course of antibiotics. Had the same patient been told to take the antibiotics and return only i f they did not clear up the problem, failure on the part of the patient to return would i l l u s t r a t e a case i n which the illness episode was resolved after the service episode had nominally terminated. Second, service episodes may occur i n the absence of illness episodes and vice-versa. As previously noted, many medical services are informative or supportive rather than curative. On the other hand, consumers may cope satisfactorily with i l l n e s s episodes without contacting a physician.or otherwise entering the medical care system. It i s worth noting that the episode of medical service remedies the deficiencies of previously mentioned u t i l i z a t i o n measures while retaining the f l e x i b i l i t y to generate them should they be relevant to a particular question. Measurement of u t i l i z a t i o n and demand by episodes of service is not (affected by price variation, as are dollar expenditure measures, and is more homogeneous than measurement by simple aggregation of a l l services used by a patient during a year, since the episodic approach relates services consumed to the continuous treatment of a specific problem or request. Although the homogeneity of episodes of medical ser-vice could be further improved by adjusting for both quality variation 7 3 and severity of the patient's problem, neither of these i s easily accomplished. Therefore the homogeneity of episodes remains dependent upon the homogeneity of systems and codes employed to classify the patient's problem and/or the physician's diagnosis. The most significant feature of the episode of medical service, however, is i t s a b i l i t y to provide data on just the patient-initiated stage of the u t i l i z a t i o n process, thereby removing the fundamental d i f f i c u l t y with u t i l i z a t i o n measures, and enabling appropriate testing of demand models. Whereas dollar expenditures on care and total v i s i t s , contacts, or procedures during a given period are functions of both patient and physician behaviour, the number of original medical requests made by the patient i n a given period i s a function of- patient behaviour alone, and thus the number of episodes of medical service i s a preferable indicator of the demand for medical care. Determination of Episodes In order to operationalize the concept of an episode of medical service, several pieces of information are required. The continuous u t i -l ization records of patients must be available and include information on a l l physician v i s i t s and contacts, and any procedures performed during those contacts. It i s also necessary to know the reason for each v i s i t or contact. If the patient-has originated the contact then the nature of his/her request or problem i s relevant. If the physician has generated the contact then i t is important to know in relation to what patient request or problem. Finally, i t i s essential that there be some method of identifying the beginning and end of episodes of service in order to transform a patient's aggregate u t i l i z a t i o n data into a series of specific problems or requests. 74 For any contact, the nature of a patient's request can be identified in three ways by presenting problem, by diagnosis, or by procedure performed. Presenting problem is a c l i n i c a l term indicating, usually on the basis of the patient's own words, the reason the patient has presented himself/herself for care. In most situations the physician makes and records a diagnosis; however, for procedures such as allergy injections or obesity counselling the nature of the request i s obvious. Services provided during any given contact, and over a series of contacts, can thus be linked to specific requests or problems. For a given patient, identification of the beginning of a new episode often amounts to confirmation that the previous episode has terminated. The i n i t i a t i o n of episodes can be identified by the presence of new presenting problems or diagnoses (Which normally represent the recog-nition or interpretation of new symptoms) or, i n the absence of pathology, by new medical care objectives as expressed by the patient. Episodes of service may be terminated by either the physician or the patient. The physician may indicate, at the completion of a contact, that ^he patient requires no further treatment, while patients often terminate episodes by f a i l i n g to return for appointments. Often, however, the termination of the service episode i s not clearly defined. Patients are commonly instructed to return " i f necessary", an option which necessitates further guidelines for the determination of episodes. Consider the p o s s i b i l i t i e s for a case in which a patient i s treated for a throat infection and told to return " i f necessary". If the patient f a i l s to return, then in retrospect the episode terminated with the f i r s t and only contact. If the :patient returns i n two days s t i l l com-plaining of a throat infection, then the episode of service continues 75 unt i l the problem i s resolved. If the patient returns i n three days with a broken arm, but no throat infection, then presumably the episode of medical service for the throat infection has terminated and a new service episode for the broken arm has commenced. If the patient returns in a month, again with a throat infection, then presumably the problem has recurred and therefore the f i r s t episode of service i s considered to be terminated, although a second episode of medical service for the same presenting problem has begun. These examples i l l u s t r a t e the basic c r i t e r i a for determination of episodes. In general, whenever the patient presents a new problem or request, a new episode commences. Conversely, as long as the patient presents the same problem or request during contacts with the physician, the original service episode continues; however, the last example i l l u s -trates that elapsed time between contacts i s a key factor i n determining episodes of service for recurring presenting problems. Although these examples i l l u s t r a t e general guidelines, certain types of problems require special attention. For diagnoses of chronic conditions, i t is unrealistic to expect the problem to ever be resolved. At best, medicine can minimize the extent to which the condition affects the patient's functioning, i.e. restrict the condition to a "normal" level. In this situation, the c r i t e r i a for determining episodes must be adjusted accordingly. Services involved i n monitoring and/or maintenance of a chronic condition at a given level of d i s a b i l i t y constitute a single episode of medical service, whereas i f a chronic condition should suddenly worsen, then the set of services required to bring the condition back to i t s iHormal" level constitutes an episode in i t s e l f . Similarly, for procedures such as desensitizations and immunization series, a l l 76 repetitions of the procedure on an uninterrupted basis constitute a single episode, in keeping with the principle of matching the empirical construct of an episode to the product the patient demands. Of course, these guidelines do not cover a l l possible types of service episodes, Considerations for further development and extension of the episode of medical service measure are discussed i n the next section; nevertheless, with three further extensions the guidelines do cover most situations in which ambulatory care i s provided by general or family practitioners. To return to our throat infection example, i f the patient had returned in three days with a broken arm, and s t i l l complained of a throat infection, then two distinct episodes of service would be taking place concurrently, one for each problem. These service episodes could quite easily develop into overlapping episodes, i f one began and terminated prior to the other, or into interleaved episodes, i f the patient returned for treatment of the problems alternately. Multi-problem contacts and related presenting problems also complicate the determination of service episodes. If a patient presents a number of problems in a given contact, then i t i s often necessary to examine the specific procedures performed in order to determine which problems resulted i n the provision of episodes of service. If two different problems are treated j o i n t l y i n one contact through one procedure, i t is often d i f f i -cult to determine that two episodes were provided. Consequently, the number of episodes may be underestimated. Moreover, returning to our patient treated for a throat infection and told to return " i f necessary", i f the patient returns in three days with no throat infection but an ear infection, does a new service episode commence? The correct algorithm i s 77 that i f the new problem i s obviously closely "related" to the old problem, then the original service episode continues. But i f the new problem i s unrelated, then the original episode has terminated and a new episode commences. Unfortunately, there can be substantial disagreement among any randomly selected group of physicians over which medical problems are like l y to be related. Fortunately, however, many conditions have a number of symptoms, consequently the physician's diagnosis may remain constant even though the patient's presenting problem changes. Thus diagnostic information provides a secondary avenue for the determination of episodes. Prior to outlining the intended application of the episodic approach, a qualification i s i n order. To the extent that physician u t i l i z a t i o n involves a learning process on the patient's part, episodes may not bear a perfect correspondence to demand; however, i t i s extremely d i f f i c u l t to distinguish the effects of such a learning process from ordinary changes in consumer tastes. Furthermore, the effect of the learning experience on demand cannot be unambiguously predicted. Applications and Further Development The stimulus behind development of the episodic approach was the desire to generate demand data allowing a proper empirical specification of the conventional micro model for investigation of hypotheses concerning factors influencing the demand for medical care. Consequently, a specific set of algorithms for the determination of episodes of medical service was developed and applied to a data set containing the u t i l i z a t i o n records of approximately 1,300 Vancouver families for a one year period. Via these algorithms, a l l v i s i t s , contacts and procedures for each individual were transformed into a series of service episodes provided by the staff of the family practice unit from which the individuals received their 78 ambulatory, primary medical care. In this way, measures of individual and family demand for medical care were obtained for use as dependent variables. With regard to independent variable data, a combination of family practice unit records and a mailed household survey provided information on socio-demographic and economic characteristics of patients and their families. A f u l l description of the origin, content and construction of the FPU data set appears as Appendix A. The algorithms for determination of episodes are lis t e d in Appendix B. Hypotheses to be investigated using the episodic approach and the conventional micro model are outlined i n the following chapter. Although the episodic approach is employed here to refine demand analysis, i t opens several other research avenues as well. Three such avenues are b r i e f l y mentioned to i l l u s t r a t e the v e r s a t i l i t y of the approach. The episode of service is perhaps the most appropriate basis for cost-effectiveness analysis of alternative treatment patterns. The extent to which an episode of service satisfies a patient's request, removes symptoms, or restores a normal level of functioning can be viewed as a measure of effectiveness. Further, for a given diagnosis, i f the costs of achieving a specified level of effectiveness via alternative episodes of service can be calculated, then well-defined episodes of ser-vice with documented effectiveness at acceptable cost might be viewed as desirable treatment technologies. Moreover, data collected on an episode of service basis would allow much more detailed analysis of the characteristics of medical care produc-tion processes than present data. It was suggested above that the episode of medical service Is a useful definition of physician output. For 79 example, had episodic data been available, i t may have been possible for Evans (1974) to determine what proportion of the increase i n u t i l i z a t i o n i n the post-Medicare period was due to changes i n patient demand and what proportion was due to changes i n the mix of services employed by physicians to satisfy demand. In general, episodic datarwould be useful for analysis of changes i n technology or service mix over time, type of provider, and organizational setting. Finally, on a related topic, the cost of a standardized episode of service for some common medical problem (or some combination of weighted costs of standard episodes for several common problems), would provide an alternate, and perhaps improved, index 4 of the price of medical care. Service episodes are multi-dimensional and potentially rich units of analysis. (Beginning with a particular presenting problem or diagnosis, an episode may consist of one or more v i s i t s , with each v i s i t consisting of one or more contacts, and each contact consisting of one or more procedures.) Yet i n order to f u l l y develop and u t i l i z e the episodic approach, algorithms for determination of episodes should be generalized beyond their specification in this thesis. A comprehensive^operational-ization of the episode concept would require, i n addition to ambulatory medical services, the inclusion of a l l procedures performed i n hospital, drugs (both prescribed and over-the-counter), telephone, contacts, extended care services, etc. Although d i f f i c u l t i e s might eventually arise in separating medical from health services, the potential analytic contri-bution of episodes more than warrants the attempt. In addition, medical practitioners should be encouraged to consider the possi b i l i t y of stand-ardizing a classification for episodes similar to the classifications which exist for presenting problems and diagnoses. 80 FOOTNOTES 1. Again,(refer to Ch. 3, f.4,), people such as Feldstein (1966) have recognized this, but have stopped short of suggesting ways to define and operationalize the concept of treatment. Klarman (1965) noted that the consumer may simply be assumed to be demanding to be seen by a doctor, believing that the doctor w i l l then handle the entire situation. This would appear to be a crude version of the episode of service approach. In this regard, i t can be noted that the episode of medical service, as a measure of demand, i s consistent with the Ch. 2 observation that specific medical services are not desired for their own sake, but rather for their collective a b i l i t y to satisfy the the consumer's problem. 2. Although our concern i n this thesis i s restricted to the demand side of the medical care market, the potential use of the episode of medical service as a measure of output should not be overlooked. Cooper and Culyer (1973), for example, have stated that the measure-ment of medical care output i s the most fundamental problem i n health economics. 3. See especially Solon (1960) (1969), and also Richardson (1970) and Fink (1973). Solon (1969) used a sample of nursing students to i l l u s t r a t e the c l i n i c a l significance of viewing u t i l i z a t i o n i n terms of episodes. Certain of the algorithms mentioned in the next section rely on considerations f i r s t suggested, though not operation-alized, by Solon. 4. See Scitovsky (1964)(1967) for discussion of problems in the constru-ction of medical care price indices. 5. With respect to the generalization of empirical results obtained using the current algorithms and data.sample, i t should be noted that although a more sophisticated set of algorithms and a random sample of patients drawn from the records of several general practitioners may be preferable, neither was available. Moreover, demand results presented in the following chapters are of substantial interest i n themselves, since i t i s l i k e l y that an increasing number of Canadians w i l l obtain their primary health care i n a health centre setting such as the one employed for this research. For discussion of how the FPU patient population may d i f f e r from that of a more general setting see Appendix A. 81 Chapter 7 HYPOTHESES CONCERNING DEMAND BY INDIVIDUALS, FAMILY HEADS AND FAMILIES FOR EPISODES OF MEDICAL SERVICE By defining and operationalizing the episode of medical service, the previous chapter met one of our objectives. This chapter outlines hypotheses and expectations regarding factors influencing the demand for episodes. The conventional micro framework of price, income and taste variables is retained*, and several hypotheses are drawn from our earlier discussion of the conventional model; however, indirect price variables replace the nominal price of medical care and alternative hypotheses regarding the role of income are discussed. Beth adjustments stem from the observation that public medical insurance insulates consumers against financial barriers to i n i t i a t i o n of the u t i l i z a t i o n process. Further, explicit "need" variables are dropped from the model specification due to the causality problems associated with them, while special attention i s paid to the range of possible education effects. Wherever appropriate, hypotheses are outlined for three consuming units - individuals, family heads, and families. . Price Effects 1 Since a l l but an insignificant number of families i n the sample were covered by the provincial medical insurance plan for a l l care included i n episodes of medical service (as defined in this thesis), the l i s t e d fees for procedures - even for f i r s t office v i s i t s - were of course meaningless. Furthermore, since the family practice unit charged no deterrent fee, and since data do not exist on the share of premium price attributable to an 8 2 episode, (because families normally receive varying premium contributions from employers and perhaps government), the definition of the price of medical care on page 26 reduces to the indirect or transactions costs of consuming care. Of particular interest are variables affecting i n i t i a l contacts for problems or requests, since these determine the number of episodes, and therefore demand. Three variables were employed to investi-gate these indirect price effects distance from family residence to source of care, corresponding travel time, and a b i l i t y of head to obtain 2 paid time off work to seek medical care for himself/herself. Distance (combining both the effect of distance per se and i t s effect on dollar cost of care) was expected to be inversely related to demand in the case of a l l consuming units, although a rather weak effect was expected for family heads since the family residence may not be the rele-vant place of origin. Similar expectations were held with respect to the alternate variable "travel time from family residence to source of care", while a positive relationship was predicted between demand by heads and the a b i l i t y of employed family heads to get paid time off to seek medical care. Income Effects On the basis of i t s conventional enabling effect, income would normally be expected to be positively related to demand; however, since insurance premiums for families in the sample were independent of the quantity of services consumed, this influence was postulated to be insignificant i n 3 a l l cases. Unfortunately, alternative hypotheses regarding income effects yield ambiguous predictions. If we could make the assumption that a l l income was wage income, earned by the family head, then a Grossman view might lead to the prediction of an inverse relationship between demand by 8 3 heads and income, due to the opportunity cost of time. Yet, Grossman acknowledged that, i f income acts as a proxy for l i f e s t y l e , i t i s not inconsistent to find consumers demanding more medical care though less health. In the FPU sample, income was measured by gross family income as reported by family head. While information on disposable, permanent and earnings versus non-earnings income might have been preferable in theory, i t was unavailable; however three further variables were employed to investigate income effects. An attempt was made to allow for transitory influences on income by specifying variables for the presence of a household head who was either a full-time student or unemployed. In both cases, lower opportunity costs of time would suggest a positive relation-ship with demand, although this effect was expected to be stronger for unemployed heads than for student heads, whose demand behaviour was expected to be influenced more by education. A positive relationship was also expected between demand and the e l i g i b i l i t y of a family head for dis a b i l i t y benefits, the hypothesis being that heads who must worry about income loss due to time off work are less l i k e l y to present their problems and symptoms to physicians. Taste Factors Whereas hypotheses concerning price and income effects have some definite origin in conventional theory, - hypotheses concerning taste factors do not. They are perhaps best viewed as attempts to adjust for socio-demographic factors which would otherwise be confounded with economic effects. Nevertheless, results concerning taste factors, when compared to the results of economic influences on demand, do convey important 84 information to policy makers. The relative significance of socio-demographic versus economic variables i s of particular interest since socio-demographic variables provide less policy leverage than economic variables, which can be relatively easily manipulated to change incentives, at least on the demand side. In the FPU sample, socio-demographic information allowed investigation of the effects of age, sex, family size, l i v i n g situation, education and status on the demand for episodes of medical service. Age and sex were to be expected to be important influences on demand, for obvious reasons. Specifically, the very young, the elderly and females (and especially those i n the childbearing years) were expected to 4 exhibit greater demand than other individuals. Family size was expected to be positively related to family demand, but i n addition was expected to be inversely related to individual demand. The hypothesis underlying this inverse relationship was that individuals i n larger families would delay seeking care since they had other family members to care for symptoms, at least in the early stages for minor problems which might be self-resolving. Since family size revealed l i t t l e about the composition of families, an alternate variable labelled " l i v i n g situation" was specified. This variable indicated whether the family unit consisted of only one individual, an individual plus child(ren), two individuals, two individuals plus child(ren), a group of individuals or a group of individuals plus child(ren). Hypotheses regarding l i v i n g situat-ion were similar to those for size effects. It was anticipated that individuals l i v i n g alone or with one other person would demand more episodes than individuals i n a nuclear family, while individuals l i v i n g i n groups would demand less. 85 The effect of education on demand and utilization has been the source of much conflict in previous literature. Both hypotheses and results have been in conflict, sometimes within the same study. Consequently, there remains a strong incentive to continue the investig-ation of the role of education. The significance of the investigation is further heightened by the claims of the human capital model in Chapter 4 and the results of an earlier study by Auster et. a l . (1969). In the Auster study, the authors analyzed correlates of U.S. state mortality rates, finding that education levels exhibited a much more significant negative correlation to mortality rates than did expenditures for medical care! This, and other similar evidence, led them to suggest that the future health status of the population might be enhanced by spending less on the provision of medical care and more on education; however, this suggestion ignores two distinct obstacles to spending less on medical care. First , and of particular significance under public medi-cal insurance, medical care spending is jointly determined by patients and practitioners, as we have continually emphasized. However, a logically prior question is whether the two directives spend less on medical care and more on education are compatible. For the sake of Auster et. a l . , education had best be negatively related to demand, yet this has not be established. On the other hand, neither has a positive relation-ship. Although Feldstein and Severson (1964), Wirick and Barlow (1964) and Long (1969) found some evidence of a negative relationship between education and utilization, opposite findings have been forwarded by Andersen (1968) and Grossman (1972), and summarized by Aday and Eichhorn (1972). Moreover, none of the results have been conclusive. In the Grossman work, \ • 86 for example, the results contradicted the predicted negative relation-ship between demand and education. These results mirror the conflicting hypotheses regarding education effects. It may be argued that the more educated demand fewer episodes because they are also l i k e l y to possess more information about health matters that allows them to self-medicate; however, the health knowledge hypothesis can be bent both ways. It i s at least equally l i k e l y that the more educated have a greater awareness of which symptoms require attention in addition to their l i k e l y greater knowledge of the av a i l a b i l i t y of medical services. A longitudinal study might allow the testing of one reconciliation of these positions, namely that more educated people demand less care over their lifetime because they make greater use of preventive services; however, such data are not presently available. A negative relationship between demand and education i s also consist-ent with the hypothesis that the more educated have a higher opportunity cost of time since education and income are positively correlated; however, i f income effects are allowed to enter, then d i f f i c u l t y arises in obtaining unambiguous predictions due to the competing " l i f e s t y l e " and "opportunity cost of time" hypotheses mentioned i n the previous section. Furthermore, within a family unit, whose education i s to count? While recognizing that parents exert a significant influence over demand by children, i t i s not clear a p r i o r i whether the wife's or the husband's education i s the relevant variable. Both were investigated in the FPU sample. The investigation of socioeconomic status, the f i n a l taste factor in the FPU sample, posed problems similar to those of education. The basic question i s whether status introduces another dimension into the analysis 8 7 or whether the apparent effects of status can be explained by a combi-nation of income and education effects. The question i s complicated by certain shortcomings of the scale employed to measure status, as described in Appendix A. Assuming that status does introduce a new dimension, and since the scale employed in the FPU sample i s based on occupation, we can restric t our consideration of hypothesis to occupa-tion effects. We might then expect to find a greater demand for episodes on the part of individuals engaged in occupations involving direct physical hazards or high levels of stress, although the episodes would presumably be triggered by quite different types of problems. Significance of the Episodic Approach If the episodic approach represents an improved approximation to reality in comparison to previous "demand" models, (i.e. i f the premise underlying this thesis i s correct and the demand-utilization distinction is important), then we would expect a given array of independent variables relating exclusively to patient characteristics to explain a relatively greater proportion of the variance in number of -episodes than in either number of contacts or dollar expenditures for care, for any consuming unit. Since patients in the FPU sample were covered by insurance, i t was necessary to construct an index of dollar expenditures based on the procedures performed and the list e d fees for.them.^ While i t i s recognized that this i s not the exact equivalent of dollar expenditure measures employed in previous literature, i t i s the best approximation possible. 88 FOOTNOTES 1. Although we do not attempt to test the Grossman model • and cannot, since the FPU data base lacks wage rate information i t w i l l nevertheless be interesting to compare the results for income and education variables to the predictions of the human capital model. 2. The household survey asked two further questions which might have generated useful information on the role of indirect prices. One question asked, "How much does i t cost, on average, to make a v i s i t to the Family Practice Unit? Please include a l l expenses incurred for one normal v i s i t , such as transportation expenses, child care expenses, loss of earnings, etc." Although the question was aimed at the head of the household, who was supposed to be completing the questionnaire, i t is suspected that the ambiguous wording of the question led many respondents to indicate the cost of v i s i t s by family members other than themselves. This would explain the distribution of responses, in which over 70% of heads indicated that a v i s i t cost less than $2.00 (the lowest possible answer on the response scale). These results are examined again, on the assumption that they are valid, in Chapter 9. The other question asked i f heads were able to obtain paid time off work to take other members of their family to the family prac-tice unit. The high number of missing responses to this question (some because respondents didn't know or the question wasn't applicable) l e f t the remaining cells too small to proceed with meaningful analysis. Information which wasn't obtained in this survey, but which is relevant to hypotheses on indirect price effects i s length of wait-ing time at source of care, total time required for a v i s i t , and perhaps the number of allowable sick days per year. 3. S t r i c t l y speaking, enabling effects might s t i l l be operative under insurance to the extent that consumption of episodes of medical service involves necessary purchases of complementary and uninsured services. 4. Age and sex may be considered to be implicit measures of biologic need. In this sense the model retains need variables; however, no attempt is made to include explicit need variables such as number of restricted-activity or di s a b i l i t y days. 5. For any patient, "dollar expenditures on care" was calculated as the sum, across procedures, of procedure performed times l i s t e d fee. Physician procedures and fees appropriate to the FPU can be 89 found in the Schedule of Minimum Fees (1973) of the British Columbia Medical Association. No fee was included for nursing or lab procedures which were recorded separately from an office v i s i t , for FPU purposes, but which would normally be included in a v i s i t . For the small number of procedures performed by other FPU health professionals, fees were constructed on the basis of prevailing rates in private practice or a consideration of the cost of time required for the procedure, with an appropriate allowance for overhead. 90 Chapter 8 EMPIRICAL RESULTS: INDIVIDUALS This chapter presents the results of demand equations estimated for a l l individuals in the FPU sample, where " a l l " refers to those i n -dividuals who had at least one episode of medical service during the study year. Missing data, and patient migration into and out of the FPU population, necessitated further reductions in sample size.* Since sample sizes are yet smaller for family heads and families, and since results for these two consuming units are similar to those for individ-uals, the present chapter i s the focus of empirical work in this thesis; however the results for heads and families are summarized in the follow-ing two chapters, with discussion of particularly interesting findings on variables unique to either consuming unit. In a l l three chapters, demand results are contrasted to u t i l i z a t i o n results obtained by re-taining the independent variable specification and replacing episodes by contacts and/or expenditures as the dependent variable. An ordinary 2 least squares multiple regression program was employed for a l l estimations. Although abbreviated variable names are explained as encountered below, variable l i s t s (with accompanying summary statistics) appear in Appendix D. The f i r s t equation in Table II reports the basic results of the d^emand equation for individuals, while the remaining equations are variations of the f i r s t employed for purposes outlined below. Our i n i t i a l obser-vation, with regard to equation one, i s that price variables did not significantly contribute to the explanation of variation in demand by individuals for episodes of medical service. As equation two illustrates., Table II Demand for Episodes of Medical Service: A l l Individuals (Dependent Var. = IES; N = 418) eq, . 1 eq. 2 eq. 3 variable coeff. t - value coeff. t - value coeff. t - value DIST -0.007 (0.32) INCA 0.719 (1.83) 0.709 (1.79) INCB 0.986 (2.62) 0.974 (2.57) INCC -0.457 (1.70) -0.465 (1.72) FY2 -0.003 (2.56) STUHEAD -0.746 (1.54) -0.748 (1.54) -0.650 (1.34) UNEMPLOY 0.565 (1.74) 0.549 (1.67) 0.773 (2.68) SEX 0.545 (2.05) 0.546 (2.06) 0.559 (2.09) SAG 0.988 (3.39) 0.989 (3.39) 1.005 (3.42) AG1 2.389 (2.35) 2.435 (2.37) 2.343 (2.29) AG6 LSB EDSA -0.526 (1.84) -0.519 (1.82) -0.468 (1.65) EDSB -0.985 (3.11) -0.975 (3.06) -1.000 (3.14) EDS 2 EDS 3 EDS4 EDS 5 EDS6 EDUCATS BL1 0.666 (2.14) 0.664 (2.13) 0.645 (2.05) BL2 0.534 (1.98) 0.583 (1.98) 0.540 (1.81). CONSTANT 2.478 2.524 2.979 I 2 .133 6.35' .130 5.86 .118 6.64 VO Table II con't. eq. 4 eq. 5 eq. 6 eq. 7 variable coeff. t - value coeff. t - value coeff. t - value coeff. t - vali DIST INCA 0.737 (1.87) 0.826 (2.06) 0.725 (1.84) 0.664 (1.68) INCB 1.033 (2.71) 0.981 (2.61) 1.025 (2.72) 1.032 (2.72) INCC -0.477 (1.76) -0.374 (1.36) -0.435 (1.61) -0.439 (1.61) FY 2 STUHEAD -0.775 (1.59) -0.829 (1.70) -0.586 (1.19) -0.681 (1.39) UNEMPLOY 0.638 (1.89) 0.505 (1.55) 0.559 (1.72) 0.499 (1.53) SEX 0.576 (2.15) 0.507 (2.15) 0.573 (2.15) 0.480 (1.79) SAG 0.928 (3.07) 0.930 (3.17) 0.987 (3.36) 1.027 (3.49) AG1 2.354 (2.31) 2.375 (2.34) 2.390 (2.35) 2.240 (2.17) AG6 -0.338 (0.78) LSB 0.432 (1.46) EDS A -0.528 (1.85) -0.480 (1.68) EDSB -0.986 (3.11) -0.954 (3.00) EDS2 -1.284 (1.57) EDS3 -0.632 (1.84) EDS4 -1.154 (3.13) EDS 5 -0.539 (1.57) EDS 6 0.044 (0.12) EDUCATS 0.043 (0.95) BL1 0.671 (2.16) 0.648 (2.08) 0.702 (2.22) 0.587 (1.87) BL2 0.599 (2.02) 0.571 (1.94) 0.551 (1.80) 0.618 (2.05) CONSTANT 2.485 2.365 2.608 1.694 .132 5.90 .135 6.04 .135 5.38 .114 5.91 93 when distance from residence to source of care (DIST) was entered, i t did have the expected sign but was extremely weak. A further attempt, not shown, to enter the effect of travel time from residence to source of care (TRAVTIM) produced similar results (coeff. -0.006 and t-value 0.84). Although such results are i n i t i a l l y surprising, reflection upon them suggests a potentially significant qualification to the hypothesized role of indirect prices. If we assume that time and distance have no effects per se, then the results of equation two imply that either ex-penses such as travel costs are insignificant, or individuals do not place as much value on their time as economists do, or both. It could be the case that the results vare dominated by those individuals who, not being employed nor recognizing the opportunity cost of their time, do not value their time as hypothesized. If this i s the case, and indirect price effects are operative only for individuals with a wage rate, then we might expect to find DIST and/or TRAVTIM entering more significantly in the equations for family heads reported in the next chapter. Contrary to time and distance prices, income does appear to s i g -nificantly affect demand. Equation one indicates that individuals in the $0-4,999 (INCA) and $5,000-8,999 (INCB) family income brackets had more episodes than individuals in the $9,000-19,999 range, while i n d i -3 viduals in higher family income brackets (INCC) had less episodes. As equation three demonstrates, by replacing the income dummies by FY2 (family income in hundreds of dollars), there is a monotonically de-creasing, approximately linear relationship between family income and 4 demand. Although this negative relationship i s (considering price re-sults) paradoxically consistent with an opportunity cost of time hypothesis, i t i s most definitely inconsistent with the hypothesis that income has an enabling effect, thus confirming our expectations regarding the role of income under public medical insurance. The relationship also f a i l s to support the l i f e s t y l e hypothesis stemming from the human capital model. In addition, equation one indicates that individuals in families with a student head had fewer episodes, and individuals in families with an unemployed head had more episodes, than other individuals."' The latter result i s again consistent with an opportunity cost of time hypothesis; however, given that hypothesis, we would expect to find a positive re-lationship between unemployment and demand with respect to heads behaviour more so than i n the case of a l l individuals. As expected, females exhibit higher demand than males, especially females between the ages of 18 and 40, as the interaction term SAG i n equation one i l l u s t r a t e s . Age variables provide some interesting re-sults, however, in that the expected positive relationship between demand and age past some point i n the l i f e cycle i s absent from equation one. In fact, as equation four shows, the dummy for individuals over 60 yrs. (AG6) enters with a negative sign, although very weakly. This result might be due to a number of factors. F i r s t , since i t was not possible to standardize demand results for diagnosis or presenting problem, i t could be that the elderly have less episodes, but more contacts per episode, than other individuals since episodes for the elderly are more l i k e l y to involve chronic conditions. This would explain both the present result and the findings of previous investigators. If this i s the case, then we would expect to see AG6 enter positively i n an equation in which episodes were replaced by contacts as the dependent variable. Such a regression was run, however, and AG6 continued to be weak and negative. This negative relation of old age to both episodes and contacts might be the result of higher hospitalization rates for 95 the elderly. This factor is particularly relevant i f i t causes the FPU sample to be biased toward the relatively more healthy elderly, by virtue of the fact that they are not already in hospitals or, more important, extended care f a c i l i t i e s . In contrast to the AG6 results, AG1 (under 1 yr.) entered the equation positively, as expected, although there were less i n -fants in the sample than expected. Neither family size nor l i v i n g situation variables appear in equation one. The reason for this i s evident from equation five. Although a number of family size and l i v i n g situation groupings were specified, only in the case of two adults l i v i n g together (LSB) was there any effect;^however, to the extent that LSB picked up two individuals of the same sex as opposed to childless couples, i t may be duplicating the hypothesized positive effect on demand of the l i v i n g alone situation. With regard to socioeconomic status, the relatively strong showing of dummies BL1 (unskilled and blue collar) and BL2 (professional) in equation one indicates that status does introduce another dimension into the analysis. The positive signs on the dummies are consistent with the hypothesis that occupation affects demand in a non-linear manner. In-dividuals at both ends of the Blishen status scale exhibit higher demand than individuals in the middle status range.^ Non-linearities are also important with respect to the role of education. As equation one indicates, compared to individuals who have completed high school, both individuals with less education (EDSA) and individuals with more education (EDSB: technical-vocational-nursing school or community college) demand fewer episodes. The second result requires some explanation, however. The dummy variable series EDSA-G, 8 arid i t s continuous counterpart EDUCATS , related to the education of 96 individuals 18 yrs. or older. For younger individuals, these variables took on the value of the education level of the mother, i n an attempt to both adjust for education in progress and investigate the influence of mother's education. Upon investigation of the EDSB variable by occupation, i t was found that this educational level was dominated by individuals who either were nurses themselves or were in families in which the mother was a nurse. Consequently, this knowledge and av a i l a b i l i t y of care may account for the effect of the EDSB variable in equation one. The above explanation cannot account for a l l evidence of non-linear-it y in education effects, as equations six and seven demonstrate. A l -though, on the basis of EDUCATS (number of years of formal schooling) we might conclude that education and demand are positively, even i f weakly, related a conclusion which has been increasingly found i n the literature, yet one which contradicts the human capital model prediction 9 i t i s apparent from consideration of the EDS2-6 series that this positive result i s the product of a positive and approximately linear relation between demand and education up to completion of high school (EDS2 and EDS3) and a rather irregular relation thereafter. Even allow-ing for the bias in EDS4, the results do not support the hypothesis of a positive linear relation throughout. This finding i s of particular importance, for i t suggests that conflicting results reported in previous literature may be extremely sensitive to i) the definition of education categories employed, i i ) other characteristics of patients, such as occupation, which are not adjusted for in regressions of demand on education and i i i ) restrictions that the estimation technique and speci-fication place on functional forms. For a l l equations reported in Table II, the F sta t i s t i c s are 2 significant at the .01 level, while the corrected R st a t i s t i c s i n -dicate that the equations explain a normal, or slightly larger than normal, amount of variance, given the nature of the data base and the area of investigation. For example, Wirick and Barlow (1964) explained less than 10% of the variation in medical expenditures of Michigan i n -dividuals, even after partitioning the sample on the basis of sex. 2 Similarly, Acton (1973a) reported uncorrected R st a t i s t i c s of between .06 and .14 for his equations with physician v i s i t s as dependent variables. Grossman (1972), i n his estimations of the demand for medical care by 2 whites with positive sick time, reported corrected R stati s t i c s of .05 - .07. In each of these studies, the dependent variable was a u t i l i z a t i o n measure rather than a demand measure. Equations three and four of Table III i l l u s t r a t e that a similar reduction in explained variance occurs in the FPU sample when number of physician*"^contacts or expendi-tures are substituted for number of episodes as the dependent variable In equation one. This result i s exactly what we would expect i f u t i l i z a -tion i s supplier-influenced, since equation one contains no supply side variables. Furthermore, the simple correlation coefficients of episodes and v i s i t s (.70), contacts (.67), procedures (.67) and expenditures (.34) also support the hypothesis that while patients effectively determine the number of episodes (by the number of problems they present), once an episode begins they relinquish much of their control over u t i l i z a t i o n to the physician. Moreover, within any particular v i s i t , patients exert even less control over the number of contacts or procedures. Finally, the weak correlation between episodes and expenditures i s not surprising, Table III Uti l i z a t i o n of Medical Services: A l l Individuals eq. 1 IES (episodes) eq. 2 ATTEND (attendance) eq. 3 ICQJTCS (contacts) eq. 4 IEXPINS (expenditures) variable coeff. t - value coeff. t - value coeff. t -. value coeff. t - val DIST -0.001 (0.03) INCA 0.719 (1.83) 0.115 (1.66) 0.781 (0.60) -8.204 (0.28) INCB 0.986 (2.62) 0.027 (0.39) 1.293 (1.04) -9.322 (0.34) INCC -0.457 (1.70) 0.085 (1.80) -0.742 (0.84) -24.401 (1.23) STUHEAD -0.746 (1.54) -0.015 (0.19) -1.603 . (1.00) -51.247 (1.44) UNEMPLOY 0.565 (1.74) -0.073 (1.32) 1.345 (1.35) 20.329 (0.85) SEX 0.545 (2.05) 0.079 (1.75) 1.776 (2.03) 24.282 (1.25) SAG 0.988 (3.39) 0.037 (0.69) 2.317 (2.41) 77.495 (3.62) AG1 2.389 (2.35) 0.314 (1.55) -0.679 (0.20) 2.362 (0.03) AG6 -0.107 (1.40) LSA 0.224 (2.86) LSB 0.083 (1.51) LSC 0.052 (0.85) EDS A -0.526 (1.84) 0.071 (1.25) -2.009 (2.14) -27.647 (1.32) EDSB -0.985 (3.11) 0.043 •(0.71) -2.651 (2.54) -12.192 (0.52) EDSC 0.081 (1.62) BL1 0.666 (2.14) -0.063 (1.22) 4.064 (3.96) 61.539 (2.69) BL2 0.584 (1.98) -0.113 (2.28) 1.196 (1.23) 16.028 (0.74) CONSTANT 2.478 0.588 4.802 37.948 I* 418 .133 6.35 580 .034 2.16 418 .075 3.84 418 .057 3.10 VO 00 99 considering the range of procedures per episode''" and the absence of i n -come constraints under public medical insurance. With a few interesting exceptions, the results of the u t i l i z a t i o n equations echo the demand results in equation one of Table III, though the general reduction of significance in the coefficients renders them much less interesting. The number and complexity of procedures involved in prenatal care and deliveries is reflected in the SAG (females 18-40) coefficient in equation four, while the UNEMPLOY result in the same equation provides further evidence, i f such is s t i l l required, of the insignificance of income effects under public medical insurance. A comparison of AG1 coefficients from equations one, three and four i s consistent with the frequently forwarded hypothesis that episodes for infant care are primarily reassurance episodes for mothers. The reduction of significance in the EDSB (tech.-voc.-nursing school or comm. college) between equations one and three is also of interest.. In further u t i l i z a t i o n regressions adjusted for l i v i n g situation, the significance of this educational level f e l l much further, suggesting two possible effects. The f i r s t i s that children whose mothers are nurses have less episodes and contacts than other children; however the effect of their mother's education and occupation is weakened with respect to uti l i z a t i o n . The second is that nurses l i v i n g alone or with one other person also have less episodes and contacts than other individuals; how-ever, their education and occupation again affects the former more than the latter. It is particularly interesting, as these results indicate, that supplier-influence on u t i l i z a t i o n surfaces even in the case of individuals with extensive health training and knowledge. The f i n a l result of interest regarding u t i l i z a t i o n i s actually a 100 non-result. At f i r s t glance, i t appears that low status individuals XBL1) have significantly more contacts per episode than high status (BL2) individuals, but this is not necessarily the case. Further inves-tigation revealed that the BL1 result was heavily influenced by two outlying observations with atypically high numbers of contacts. The remaining equation in Table III is reported for general i n -terest only. The dependent variable ATTEND (defined as 1 i f individual had at least one episode - zero otherwise) was specified in an attempt to examine the likelihood of an individual being a u t i l i z e r , given that he/she was in the FPU patient population. Several independent variables of potential interest were added to the basic demand specification; however, the results do not warrant detailed attention since, as pre-viously mentioned, the number of non-utilizers may be overestimated. To pursue this line of analysis properly, Tobit estimation procedures should, probably be adopted. Summarizing the results of this chapter, i t appears that demand by individuals for episodes of medical service i s inversely related to i r i -12 come and not significantly influenced by time or distance price variables, although more precise measurement and matching of indirect price variables to the hypothesized transactor is definitely required. With regard to taste factors, age and sex are found to be important, though this is neither a new nor surprising result. What Is^ interesting about taste factors, however, is that one must be prepared to dig further into the results of regression analysis in order to avoid being mislead by inter-action effects or functional form restrictions. Education effects are a prime example of this caveat. In addition, a comparison of demand and 101 ut i l i z a t i o n results supports the claim that the episodic approach is both a r e a l i s t i c and f r u i t f u l framework for analysis of the medical marketplace. 102 FOOTNOTES 1. Missing data problems arose in the case of families who did not return the household survey. In order to avoid serious under-estimation of demand, and consequently misleading results, i t was necessary to further r e s t r i c t the sample to only those individuals who were FPU patients for the entire study period. This excluded those who registered at or l e f t FPU during the study period and those who died during the study period. Although i t is never possible to know exactly which or how many families are obtaining primary medical care from other sources than the FPU (especially for those people apparently having no episodes during the year!) a very good approxi-mation was obtained through examination of patient charts and follow-up of surveys that were returned undelivered. In addition, practi-tioners often noted a patient's intention to leave the FPU on record-ing forms for contact information. 2. The St a t i s t i c a l Package for the Social Sciences program REGRESSION • was employed. Multiple mode, and sta t i s t i c s 1 and 2 were specified. No options were specified. (Among other things, this defaulted to . listwise deletion of missing data.) Note that the episodic approach, by eliminating supply-demand interdependence in the analytic speci-fication, removes the need for more sophisticated estimating tech-niques such as those employed by Acton (1973a) and Newhouse and Phelps (1974). 3. INCA, INCB and INCC were defined on the basis of previous regression runs with dummies which represented smaller income brackets. Similar operations were performed on other dummy variables discussed in this chapter. For an example of results upon which regrouping and deletion (i.e. combination with the base) were carried out, see Appendix C. Specific regroupings and deletions are apparent from the variables l i s t e d in Appendix D. 4. Similar results were obtained with a specification in which the FY2 equation was compared to an equation employing the INC1-6 series instead of the INCA-C series of dummy variables. 5. Although the percentage of individuals in families with unemployed heads (22%) may seem high, i t probably dsn't, since 5% of individuals were in families with a student head, 9% of individuals were over 60 years old, and the unemployment rate for the province was above 8% at the time of the survey. Simple correlations involving these variables, that may be of interest are given in the following table: INCA INCB INCC FY2 STUHEAD .12 .04 -.16 -.18 UNEMPLOY .54 .09 -.16 -.35 103 6. FAMSIZE, in particular, added l i t t l e to equation one. (coeff. .056; t-value 0.69) A dummy variable specification failed to improve this result. 7. As noted in Appendix A, the Blishen scale assigns status on the basis of occupation. In this respect, the dummies have been labelled rather loosely "unskilled and blue collar" and "professional" in contrast to their base, which was labelled "white collar". Labels are arbitrary. The point is that unskilled and blue collar (25.25 to 33.76), white collar (33.80 to 67.78) and professional (68.32 to 76.44) indicate progressively higher status levels. 8. The transformation of education levels to years of formal schooling was as follows: none (0), elementary (6), some high school (9), high school (12), technical-vocational-nursing school or community college (14), some university (14), university degree(s) (16). 9. There were no individuals in the no education CEDS1) category. Remaining categories are elementary school (EDS2), some high school (EDS3), technical-vocational-nursing school or community college (EDS4), some university (EDS5) and university degree(s) (£DS6). The base for the series was "completed high school". A regrouping of these dummies yielded EDSA (less than high school), EDSB (tech-nical-vocational-nursing school or community college) and EDSC (at least some university) for the same base. 10. As mentioned previously, these measures include some contacts with other FPU health professionals; however, non-physician u t i l i z a t i o n i s insignificant for present purposes. 11. The following table i s from preliminary work on a detailed analysis of characteristics of episodes. It indicates the range in the total number of procedures for individuals with 1 - 15 episodes during the study period. For example, the f i r s t column shows that there were 113 individuals, (27% of the sample), who had only one episode, yet some individuals experienced a single procedure episode while at least one of the 113 individuals had 21 procedures performed for him/her during the episode. 104 number of episodes absolute frequency relative frequency (%) range in number of procedures 1 2 3 4 5 6 "7 8 9 10 11 12 13 14 15 113 87 73 50 38 14 16 13 2 4 3 3 1 0 1 27.0 20.8 17.5 12.0 9.1 3.3 3.8 3.1 0.5 1.0 0.7 0.7 0.3 0 0.2 1 2 3 4 5 9 10 9 26 16 18 20 24 29 21 40 25 41 42 54 37 31 31 25 28 74 Totals 418 100.0 This table illustrates the extra dimension that the episodic approach provides for analysis of u t i l i z a t i o n . It also suggests that the FPU data base can be mined more vigorously than i t has been in this thesis; however, in order to r e s t r i c t the thesis to a reasonable length, only results central to the objective of the thesis demand analysis are discussed. Certain other results of general or tangential interest, such as the ATTEND results later in this chapter, are reported without discussion. 12. It is possible, although i t is not discussed in the text, that the i n -verse relation between demand and income is attributable to the fact that both are related to a third factor, incidence of i l l n e s s , though in opposite directions. If so, then causality might run from illness to low income, yet we would s t i l l observe the inverse income-demand relation. The household survey data, however, do not appear to support this interpretation. Responding to a question asking i f family income had been any higher or lower than normal in the past year due to i l l -ness, unemployment or other unexpected factors, 14% of family heads indicated lower than normal family income. Considering that i) only 24% indicated they were not eligible for d i s a b i l i t y benefits i f unable to work due to illness or injury (and 14% didn't know), and i i ) many of those heads reporting negative transitory effects could be in the 5 - 10% of non-student, less than 60 yr. old heads that were unemployed, i t appears that illness may not be the important income-demand link. Unless, of course, most unemployment was also due to i l l n e s s . Chapter 9 1 0 5 EMPIRICAL RESULTS: FAMILY HEADS Methodologically and s t r u c t u r a l l y , both t h i s chapter and the following one on f a m i l i e s p a r a l l e l the preceding chapter on i n d i v i d u a l s . Both samples were adjusted for migration and missing data, the same estimation techniques were employed, relevant v a r i a b l e l i s t s and summary s t a t i s t i c s again appear i n Appendix D and the r e d e f i n i t i o n and regrouping of dummy v a r i a b l e s proceeded i n a s i m i l a r fashion. Basic demand equations are again reported f i r s t , followed by discussion of a l t e r n a t e s p e c i f i c a t i o n s designed to i n v e s t i g a t e s p e c i f i c p r i c e , income or taste e f f e c t s . P r i o r to concluding each chapter, u t i l i z a t i o n r e s u l t s are reported and compared to demand r e s u l t Equation one of Table IV conveys the b a s i c f i n d i n g concerning the factors i n f l u e n c i n g family heads' demand f o r episodes of medical care. Only age, sex and income e f f e c t s appear to a f f e c t the demand behaviour of heads to any s i g n i f i c a n t degree. Notably absent from the basic equation for heads, compared to the i n d i v i d u a l s equation, are education and socioeconomic status v a r i a b l e s * , while time and distance p r i c e v a r i a b l e s continue to perform poorly. The heads sample provides the most appropriate place to t e s t f o r time and distance p r i c e e f f e c t s , since time involved i n consuming medical care presumably has some e x p l i c i t d o l l a r value f o r employed . heads, as opposed to i t s i m p l i c i t opportunity cost i n the case of house wives, for example. Unfortunately, though, attempts to explore the r o l e of such i n d i r e c t p r i c e s were again undercut by measurement d i f f i c u l t i e s . As equation two i l l u s t r a t e s , the expected sign on DIST Table IV Demand for Episodes of Medical Service: Heads (Dependent Var . = IES; N = 202) e eq . 2 eq . 3 variable coeff. t - value coeff. t - value coeff. t - value DIST -0.029 (0.78) PTOSA 0.148 (0.36) PTOSB 0.499 (1.06) INCD 1.226 (3.21) 1.055 (2.55) 1.248 (3.18) INCE -0.520 (1.21) -0.605 (1.38) -0.533 (1.22) STUHEAD -0.816 (1.26) -0.714 (1.08) -0.829 (1.27) DISBENL 0.086 (0.26) AG4 -0.875 (1.21) -0.926 (1.28) -0.886 (1.22) SEX 0.685 (2.02) 0.625 (1.82) 0.681 (2.00) BL1 BL2 EDA EDB EDC ED 2 ED 3 ED 4 ED5 ED 6 CONSTANT 2.503 2.531 2.463 R2 .118 .114 .114 F 6.38 4.24 5.31 o Table IV con't. eg- * eg. 5 variable coeff. t - value coeff. t - value DIST PTOSA PTOSB INCD 1.337 (3.37) 1.319 (3.34) INCE -0.694 (1.55) -0.623 (1.41) STUHEAD -1.004 (1.51) -1.031 (1.52) DISBENL AG4 -0.849 (1.16) -0.933 (1.27) SEX 0.802 (2.21) 0.755 (2.16) BL1 0.071 (0.16) BL2 0.479 (0.87) EDA -0.346 (0.69) EDB -0.085 (0.15) EDC 0.292 (0.69) ED 2 -1.173 (1.29) ED 3 „ -0.114 .(0.22) ED 4 -0.054 (0.10) ED 5 0.439 (0.90) ED 6 0.381 (0.81) CONSTANT 2.344 2.371 .111 3.52 .113 3.58 108 i s of l i t t l e Interest considering the t-value. A similar specification J for TRAVTIM produced equally uninspiring and ambiguous results. The results are ambiguous because i t is s t i l l not clear that the independent variables have been measured correctly. Since DIST and TRAVTIM represent distance and travel time from place of residence to the FPU they may not be relevant to the behaviour of family heads who travel to the FPU from their place of employment. On the other hand, r e c a l l from footnote 2 of Chapter 7 that over 70% of heads indicated that an average v i s i t to the FPU cost less than $2.00, including foregone earn-ings, child care expenses, and transportation costs. If this distribut-ion of responses is valid, and time and distance per se do not have any deterrent effects, then i t i s not at a l l surprising to find that DIST and TRAVTIM failed to contribute significantly to explanation of the variation i n episodes. Furthermore, i f we assume i ) that heads are able to obtain paid time off work to obtain medical care and i i ) that the only other expenses are transportation costs, then in order to exceed $2.00 for a round trip (assuming travel by car at a cost of 20C a mile — including depreciation!) a person would have to travel more than five miles. In the FPU sample the mean distance from residence was slightly less than this. A l l this suggests that indirect prices may only have an effect i f they are binding. In other words,paid time off work may have an insulating effect upon indirect prices similar to that of public medical insurance on nominal or direct prices. Consequently, the variables PTOSA (head able to get paid time off to obtain medical care) and PTOSB (head does not know i f able to get paid time off, or no response) were also included in equation two. Although PTOSA i s positively related to 109 demand as expected, i t s effect i s both weak and small. Ironically, the effect of the second variable i s both larger and stronger; however, the strong positive correlation (.89) between PTOSB and UNEMPLOY implies that PTOSB is serving as a proxy for unemployed heads. This interpretation is reinforced by the observation that the coefficient for low income heads (INCD: $0 - 8999) decreases i n both magnitude and significance when PTOSB enters. Although not presented here, the same effect upon INCD was apparent when UNEMPLOY was entered into equation one. Conversely, i n the absence of INCD, UNEMPLOY became significant. The coeffecients for INCD and INCE ($20,000 or more) indicate that the inverse relation between income and demand found in the individuals sample i s present in the heads sample as well, again confirming our expectations of income effects under public medical insurance. Equat-ion one also reports that student heads have less episodes than other heads, yet this may be due to the " l i v i n g alone" situation of many students or the fact that they are enjoying some of their healthiest years of the l i f e cycle. E l i g i b i l i t y for d i s a b i l i t y benefits (DISBENL) was also postulated to be a positive influence on demand, since i t allegedly removed fears of income loss due to work limitations imposed by a physician i n response to reported symptoms. Equation three indicates that this influence was not of any significance i n the FPU sample, though the sample was almost equally s p l i t between those eli g i b l e and in e l i g i b l e 2 for benefits. Other than age (AG4: heads 18-22) and SEX (female), which had their expected negative and positive effects on demand respectively, taste factors played no role i n the basic demand equation for heads. FAMSIZE and a l l liv i n g situation effects were almost i n v i s i b l e , though the estimations of the latter were hampered by small c e l l sizes. As equation four displays, status effects were of approximately the same size and sign as i n the individuals equations, though much less significant, which raises an interesting question. Occupation of head defines status. In the individuals equations status variables were significant, apparently due to the nature of occupations involving stress or physical hazards. Yet i n equation four of Table IV status does not show significance, and no explanation immediately comes to mind. Equation four also indicates that education effects for heads are similar to those for Individuals, though again much less significant. Moreover, i t i s apparent from equation five that the weak positive relation that does exist between demand and education levels off after completion of high school. As i s the case with the rest of the Table IV equations, equation five i s s t a t i s t i c a l l y significant at the .01 level, and explains a "normal" amount of variance. With regard to u t i l i z a t i o n , Table V demonstrates, as expected, that replacement of episodes by either contacts or expenditures reduces -2 both the R and F st a t i s t i c s substantially. In addition, i t further weakens the significance of a l l but two of the coefficients, AG4 for contacts and SEX for expenditures. The simple correlation of number of episodes with number of v i s i t s (.78), contacts (.77), procedures (.72) and expenditures (.60) i s yet another instant replay of the results for individuals which supports the hypothesis that patients exert progress-ively less control over v i s i t s , contacts, procedures and expenditures compared :to episodes. Table V Ut i l i z a t i o n of Medical Services: Heads 202) variable eq. 1 IES (episodes) coeff. t - value eq. 2 ICONTCS (contacts) coeff. t - value eq. 3 IEXPINS (expenditures) coeff. t - value INCD INCE STUHEAD AG4 SEX CONSTANT 1.226 -0.520 -0.816 -0.875 0.685 2.503 (3.21) (1.21) (1.26) (1.21) (2.02) 1.204 -0.978 -1.715 -2.970 1.688 5.358 (1.28) (0.b2) (1.07) (1.66) (2.01) 3.553 -4.407 -18.812 -21.905 25.425 42.679 (0.34) (0.37) (1.06) (1.12) (2.76) R 2 .118 6.38 .046 2.93 .037 2.53 112 An attempt was made to further explore influences on family heads' ut i l i z a t i o n by standardizing for number of episodes. Equations two and three were rerun with IES added to the l i s t of independent variables in each case. The only particularly interesting result of this exer-cise was that i n both cases the sign of INCD ($0 - 8999) changed, becoming -1.12 (t-value 1.78) i n the contacts equation and -16.03 (t-value 1.86) i n the expenditures equation. This suggests that low income heads have more episodes but less contacts and expenditures per episode than high income heads. This finding that higher income heads have fewer, more contact and procedure-intensive episodes could be due to their higher opportunity cost of time, (assuming that a l l or most income i s wage-income); however, conclusions are premature in the absence of standardization for presenting problem or diagnosis. Overall, the results of this chapter suggest that economic influences are more important to the demand behaviour of heads than other individuals. Furthermore, these economic influences, primarily income effects, are relatively more important to the demand behaviour of heads than are taste factors. Finally, the use of an episodic approach for modelling demand is again supported. '' 1 1 3 FOOTNOTES 1 . Also notable i s the fact that not one of the simple correlation coefficients among the group INCD-E, EDA-C and BL1-2 exceeds .4. 2. DISBENL showed rather low correlations with INCD (-.25) and INCE ( . 1 9 ) . 114 Chapter 10 EMPIRICAL RESULTS: FAMILIES Although many variables in the FPU sample are individual-specific, several apply to the demand behaviour of entire family units. Furthermore, some individual-specific variables (e.g. mother's education) have been hypothesized to significantly affect the behaviour of remaining individuals i n a given family unit. Consequently, estimation of family demand and u t i l i z a t i o n equations, i n addition to those for individuals and family heads, is worth attempting wherever data permit. In the FPU sample, however, r e c a l l that families are defined i n a manner that allows for several other l i v i n g situations i n addition to the nuclear family. This definition should be kept in mind for interpretation of the results that follow. Partly for this reason and to the extent that previous trends tend to be repeated, we restri c t our analysis in this chapter to results of particular interest. In general, the trends established i n the preceding two chapters do continue. The basic family demand equation, Table VI equation one, suggests that, in addition to the obvious effect of family size, low family income (INCD) and young family heads (HAGl) were significantly related to demand, though in opposite directions. Beginning our analysis again with price effects, we find none entering the basic family equation. As equation two of the same table illustrates,when distance (DIST) was entered i t was very insignificant. The same was true of travel time. Since these variables would seem, a p r i o r i , to be quite relevant to the demand of remaining family members as opposed to employed heads, the results suggest that both pure distance and time effects and their indirect dollar costs are of minimal 1 influence on family demand. Table VI Demand for Episodes of Medical Service: Families (Dependent Var = FES; N = 236) eg. 1 eg. 2 eg. 3 coeff. t - value variable coeff. t - value coeff. t - val DIST -0.013 (0.19) STUHEAD -1.303 (1.12) • -1.307 (1.12) UNEMPLOY 0.873 (1.14) 0.853 (1.09) INCD 2.211 (2.87) 2.206 (2.92) INCE -0.308 (0.38) -0.323 (0.39) HAG1 -2.267 (1.46) -2.261 (1.45) HAG 2 1.360 (2.05) 1.364 (2.05) HAG 3 -0.843 (0.81) -0.846 (0.81) FAMSIZE 1.987 (8.94) 1.995 (8.79) EDS A -1.227 (1.45) -1.207 (1.41) EDSB -1.979 (2.29) -1.957 (2.24) EDSC 0.200 (0.27) 0.204 (0.08) BL1 BL2 CONSTANT 0.432 0.474 -1.371 (1.18) 0.198 (0.27) 1.825 (2.45) -0.635 (0.78) 2.030 (9.28) -0.911 (1.07) -1.595 (1.83) 0.508 (0.68) 0.924 R2 .305 .302 .279 F 10.39 9.49 12.44 Table VI con't eg. 4 eq. 5 eq. 6 variable coeff. t - value coeff. t - value coeff. t - vali STUHEAD -1.771 (1.49) -1.505 (1.26) -1.345 (1.16) UNEMPLOY -0.064 (0.09) -0.067 (0.09) 0.266 (0.36) INCD 1.986 (2.57) 1.998 (2.52) .1.840 (2.46) INCE -0.029 (0.03) -0.584 (0.71) -0.604 (0.56) FAMSIZE 2.292 (6.73) 1.998 (8.97) EDSA -0.598 (0.69) -0.758 (0.86) -1.001 (1.17) EDSB -1.481 (1.67) -1.480 (1.66) -1.630 (1.86) EDSC 0.545 (0.72) 0.571 (0.76) 0.452 (0.59) BL1 0.668 (0.89) BL2 0.458 (0.52) FS1 -6.026 (8.01) FS2 -3.146 (4.14) FS3 4.501 (2.60) -LSI 0.966 (0.79) LS2 0.535 (0.41) LS3 1.479 (1.47) LS4 1.701 (0.97) LS5 1.251 (0.65) LS6 -0.249 (0.13) CONSTANT 8.525 -0.478 0.827 R 2 .258 .272 .276 F 9.23 7.27 9.99 117 With regard to income effects, and again recognizing that a causality problem may exist, i t i s interesting to note that while demand is again inversely related to income, this effect i s not as strong at the higher income levels as i t was in the case of individuals and heads. Demand is also inversely related to STUHEAD and positively related to UNEMPLOY, again familiar results. Of much more interest is the dummy variable series HAG1-3, repres-enting the age of the head of the family (18-22, 23-40 and over 60 years respectively), of which the f i r s t two have relatively large (in absolute value) coefficients. While age of head i s sometimes considered to proxy the stage of a family in the li f e - c y c l e , and thus the family's "need", i t might also be suggested that age of head i s really picking up the influence of education, income, or family size effects. Equation three of Table VI, in which the HAG1-3 series i s deleted, does not support this suggestion, however, since INCD-E, FAMSIZE and EDSA-C 2 coefficients do not change appreciably. (We would not expect much change i n EDSA-C however, since i t emphasizes the education of mothers). With regard to family size effects, equation four presents an alternate, dummy variable specification while the more interesting equat-ion five attempts to examine the effect of li v i n g situation per se by adjusting for family size. The LS3 (two adults) and LSI (living alone) results weakly support the hypothesis that individuals l i v i n g alone (or as in the case of LS3, with another person of the same sex) present more problems for treatment because care via the nuclear family unit i s unavail-able. Further, with regard to family size and li v i n g situation effects i t should be noted that a specification which replaced these effects by number of children performed relatively poorly, which was not unexpected 118 considering the low proportion of nuclear family situations in the sample and the AG2 (1-5 yrs.) and AG3 (6-15 yrs.) results from the individuals demand equations in Chapter 8. We earlier hypothesized that mother's education would be an important influence on family demand. Equation one of Table VI tends to support this hypothesis, with demand being especially reduced by EDSB, which i s the education level into which diploma school nurses were classified. Of further interest i s the observation that this result holds up i n the u t i l i z a t i o n equation for contacts (equation two, Table VII), although not i n i t s expenditures counterpart (equation three). Apparently, mothers with medical backgrounds reduce not only.family 3 demand but also family u t i l i z a t i o n . With regard to the u t i l i z a t i o n equations, note the general reduct-ion i n significance which again accompanies substitution of contacts or expenditures for episodes as the dependent variable. Whereas the F st a t i s t i c s in Table VI a l l easily exceeded their .01 level c r i t i c a l values (though this i s somewhat misleading due to the role of FAMSIZE), the F for equation three of Table VII barely accomplishes this even though the equation-also contains FAMSIZE. Furthermore, the corrected 2 R of equation three a l l but disappears. The trend for episodes to be more strongly correlated with (progressively) expenditures (.51), procedures (.83), contacts (.84) and v i s i t s (.86) also continues. Other notable results of equations two and three, to the extent that the, equations are believable, are that income loses significance in the expenditures equation as predicted and age of head 23-40 yrs. gains significance, suggesting that i t has been picking up pregnancy episodes a l l along. When both equations were standardized for the Table VII U t i l i z a t i o n of Medical Services: Families (N = 236) eq. 1 eq. 2 eq. 3 FES FCONTCS FEXPINS (episodes) (contacts) (expenditures) variable coeff. t - value coeff. t - value coeff. t - val' STUHEAD -1.303 (1.12) -2.628 (0.83) -70.579 (1.12) UNEMPLOY. 0.873 (1.14) 1.885 (0.89) 35.246 (0.84) INCD 2.211 (2.87) 4.006 (1.95) 15.064 (0.37) INCE . -0.308 (0.38) -0.274 (0.13) -20.888 (0.47) HAG1 -2.267 (1.46) -6.084 (1.44) -23.923 (0.28) HAG2 1.360 (2.05) 3.311 (1.83) 84.325 (2.34) HAG 3 -0.843 (0.81) -1.498 (0.53) -13.133 (0.23) FAMSIZE 1.987 (8.94) 4.329 (7.13) 39.845 (3.29) EDSA -1.227 (1.45) -2.177 (0.95) -36.865 (0.80) EDSB -1.979 (2.29) -5.231 (2.22) -43.466 (0.92) EDSC 0.200 (0.27) 0.816 (0.41) -10.297 (0.26) CONSTANT 0.432 0.322 7.742 R2 .305 .223 .045 F 10.39 7.13 2.03 120 number of family episodes, the only variable even approaching s i g n i f i -cance was HAG2 i n the expenditures equation, lending further support to the pregnancy conclusion above. The results of this chapter and previous ones generally support the main hypotheses of this thesis that i ) the conventional micro model has been misspecified through the use of u t i l i z a t i o n rather than demand measures, and i i ) even when properly specified on episodes of medical service, the conventional micro model needs a conceptual overhaul or i t is inappropriate for investigation of demand under conditions of public medical insurance. Hypotheses concerning the effects of some specific factors such as education, age, sex and l i v i n g situation have received varying degrees of support, while those concerning other variables such as time and distance prices and socioeconomic status have received less support, although conclusions are hampered in the latter two cases by suspected measurement errors. The following, and f i n a l , chapter summarizes the overall investigation. 121 FOOTNOTES 1 . Unless, of course, the family results are dominated by single-individual family units, in which case DIST and TRAVTIM may not be the proper measures of distance and time effects, as already noted i n Chapter 9. 2. It might be worth noting that the removal of the series HAG1-3 increases the linearity ( i f "linear" can be said to have degrees) of the income-demand relationship. 3. A further reason for caution in interpreting the family u t i l i z a t i o n results i s provided by the observation that there are end of range problems in these equations. Although this was also the case with Table III (individuals' u t i l i z a t i o n ) , these appear more serious. 122 Chapter 11 SUMMARY This thesis has attempted to ex p l i c i t l y demonstrate that the distinction between demand for and u t i l i z a t i o n of medical services necessitates a more r e a l i s t i c view of the medical marketplace than is to be found in previous literature. Toward this end, i t has for-warded a critique of existing theories of the "demand" for medical care, (after d i s t i l l i n g numerous models into their three theoretic origins), which illustrates that most existing attempts to model demand are misspecified since they test models of autonomous patient decision-making upon u t i l i z a t i o n data, which is significantly supplier-influenced. In order to remedy this deficiency, a new measure of demand, labelled the "episode of medical service", has been developed and operationalized upon the u t i l i z a t i o n records of families attending a Vancouver family practice unit. Demand and u t i l i z a t i o n equations have then been estimated for individuals, family heads and families. A consistent, inverse, and approximately linear relation was observed between demand and income, which contradicts the enabling effect postulated by the conventional micro model but supports the hypothesis that the opportunity cost of time affects demand behaviour of higher income individuals; however, this was predicted to be the case under conditions of public medical insurance. The existence of insurance was also hypothesized to render nominal or lis t e d price of services meaningless, and to increase 1 2 3 the significance of time and distance prices, both .in their pure form and their dollar translations. Hypotheses concerning these indirect price variables, however, were only weakly supported. This could be due to the peculiar characteristics of the sample employed, but is more li k e l y accounted for by measurement error induced by the definitions of the variables. In particular, the suspicion is that i f we can match time and price variables more accurately to the transactor involved, their significance w i l l increase. With regard to taste factors, a variety of results of differing magnitude and significance were found. Education was perhaps the most interesting single factor, the finding being that demand was positively related to education up to completion of high school, but levelled off thereafter. This result implies that previous specifications not allowing for non-linearities in taste effects may be responsible for conflicting results in the literature concern-ing the role of education. Finally, equations specified solely on the basis of a patient decision-making model consistently explained a significantly larger proportion of the variance in demand than u t i l i z a t i o n . This suggests that supply side variables are indeed necessary to explain variations in u t i l i z a t i o n , and confirms the importance of the demand-utilization distinction mentioned above. The main contributions of the thesis are then.seen to be as follows: i) demonstration that the u t i l i z a t i o n process should be viewed analytically as a two-stage process in which patients present problems or requests that are subsequently resolved through treatment 1 2 4 regimens prescribed by practitioners; i i ) the f i r s t comprehensive review of the demand for medical care literature since the introduction of a human capital model of the demand for care; and i i i ) the develop-ment and operationalization of the episodic approach for economic analysis of the market for medical care. Although i t i s recognized that much re-mains to be explored through an episodic approach, even within the data sample already analyzed, and substantial refinement of che definition algorithms i s necessary, i t i s f e l t that the methodology and results of this thesis represent an important building block in the construction of a r e a l i s t i c framework for analysis of the medical marketplace. 125 BIBLIOGRAPHY A. Books, Articles and Papers Acton, J. P., Demand for Health Care Among the Urban Poor, with Special Emphasis on the Role of Time. RAND Corporation, R-1151 OEO/NYC, April 1973a Acton, J. P., Demand for Health Care When Time Prices Vary More than  Money Prices. RAND Corporation, R-1189- OEO/NYC, May 1973b. Aday, L. A. and Eichhorn, R., The U t i l i z a t i o n of Health Services: Indices and Correlates. Dept. of Health, Education and Welfare Publication No. (HSM) 73-3003, December 1972, Rockville, Maryland. Andersen, R., A Behavioral Model of Families' Use of Health Services. Chicago, Center for Health Administration Studies, Research Series Number 25, 1968. Andersen, R. and Anderson, 0, W., A Decade of Health Services: Social Survey Trends in Use and Expenditure. University of Chicago Press, Chicago, 1967. Andersen, R. and Benham, L., "Factors Affecting the Relationship Between Family Income and Medical Care Consumption", in H. E. Klarman, ed., Empirical Studies in Health Economics. Proceedings of the Second Conference on the Economics of Health, The Johns Hopkins Press, Baltimore, 1970. Andersen, R. and Newman, J.F., "Societal and Individual Determinants of Medical Care Uti l i z a t i o n in the United States". Milbank  Memorial Fund Quarterly, Vol. 51, Winter 1973, pp. 95-124. Anderson, J. G., "Health Services Utilization: Framework and Review", Health Services Research. Vol. 8, F a l l 1973, pp. 184-199. Anderson, 0. W. and Feldman, J. J., Family Medical Care Costs and Voluntary Health Insurance: A Nationwide Survey. McGraw H i l l , New York, 1956. Antonovsky, A., "A Model to Explain Visits to the Doctor", Journal of  Health and Social Behaviour. Vol. 13, No. 4, December 1972, pp. 446-454. Arrow, K. J., "Uncertainty and the Welfare Economics of Medical Care", American Economic Review. Vol. 53, December 1963, pp. 941-973. Auster, R. et a l . , "The Production of Health, An Exploratory Study", Journal of Human Resources. Vol. 4, No. 4, pp. 411-436, 1969. 126 Bashshur, R. L. et a l . , "Some Ecological Differentials in the Use of Medical Services", Health Services Research. Vol. 6, Spring 1971, pp. 61-75. Beck, R. G., The Demand for Physicians' Services i n Saskatchewan, . Ph. D. Thesis, University of Alberta, 1971. Beck, R. G., "Economic Class and Access to Physician Services Under Public Medical Insurance", International Journal of Health  Services. Vol. 3, No. 3, 1973, pp. 341-355. Becker, M. H. et a l . , "Motivations as Predictors of Health Behaviour", Health Services Reports. Vol. 87, November 1972, pp. 852-862. Becker, M. H. et a l . , "A New Approach to Explaining Sick-Role Behaviour in Low-Income Populations", American Journal of Public Health, Vol. 64, March 1?74, pp.205-216. Berkanovic, E. and Reeder, L. G., "Can Money Buy the Appropriate Use of Services? Some Notes on the Meaning of U t i l i z a t i o n Data", Journal of Health and Social Behaviour. Vol. 25, June 1974, pp. 93-99. Berki, S. E., Hospital Economics. D. C. Heath and Co., Lexington, Mass., 1972. Bice, T. W. and Eichhorn, R. L., "Socioeconomic Status and Use of Physician Services: A Reconsideration", Medical Care. Vol. 12, May-June 1972, pp. 261-271. Bice, T. W. and White, K. L., "Factors Related to the Use of Health Services: An International Comparative Study", Medical Care. Vol. 7, March-April 1969, pp. 124-133. Blishen, B. R., "The Construction and Use of an Occupational Class Scale", Canadian Journal of Economics and P o l i t i c a l Science. Vol. 24, November 1958, pp. 519-531. Blishen, B. R., "A Socio-Economic Index for Occupations in Canada", Canadian Review of Sociology and Anthropology. Vol. 4, February 1967, pp. 41-53. Boulding, K. E., "The Concept of Need for Health Services", Milbank  Memorial Fund Quarterly. Vol. 64, October 1966, pp. 171-202. Coburn, D. and Pope, C. R., "Socioeconomic Status and Preventive Health Behaviour", Journal of Health and Social Behaviour. Vol. 15, June 1974, pp. 67-78. Cole, S. and Lejeune, R., "Illness and the Legitimation of Failure", American Sociological Review, Vol. 37, June 1972, pp. 347-356. Cooper, M. H. and Culyer, A. J. , Health Economics. Penguin, Baltimore, 1973. 1 2 7 Culyer, A. J., "Is Medical Care Different?", in M. H. Cooper and A. J. Culyer, ed., Health Economics. Penguin, Baltimore, 1973. Densen, P. et a l . , Prepaid Medical Care and Hospital Ut i l i z a t i o n . Hospital Monograph Series No. 3, American Hospital Association, Chicago, 1958. Densen, P. et a l . , "Prepaid Medical Care and Hospital Utilization in a Dual Choice Situation", American Journal of Public Health. Vol. 50, November 1960, pp. 1710-1726. Densen, P. et a l . , "Comparison of a Group Practice and a Self-Insurance Situation", Hospitals. Vol. 36, November 1962, pp. 63-68, 138. Enterline, P. E. et a l . , "The Distribution of Medical Services before and after 'Free' Medical Care - The Quebec Experience", New  England Journal of Medicine. Vol. 289, No. 22, November 29, 1973. Evans, R. G., "Beyond the Medical Marketplace: Expenditure, U t i l i z a t i o n and Pricing of Insured Health Care in Canada", paper presented at the Universities - N.B.E.R. Conference on the Role of Health Insurance in the Health Services. Sector, Rochester, N. Y., May-June 1974. Evans, R. G., "Health Services in Nova Scotia", Canadian Public Policy, Vol. 2, No. 3, Summer 1975, pp. 355-366. Feldstein, P. J., "Research on the Demand for Health Services", Milbank Memorial Fund Quarterly. Vol. 64, July 1966, pp. 128-162. Feldstein, P. J. and Carr, W. J., "The Effect of Income on Medical Care Spending", Proceedings of the Social Statistics Section  of the American S t a t i s t i c a l Association. Vol. 7, 1964. Feldstein, P. J. and Severson, R. M., "The Demand for Medical Care", in Report of the Commission on the Cost of Medical Care. American Medical Association, Chicago, 1964. Fink, R., "Analysis of Util i z a t i o n Data", Medical Care. Vol l l , No. 2, Supplement, March-April 1973, pp. 109-116. Freidson, E., Patients' Views of Medical Practice. Russell Sage Foundation, New York, 1961. Fuchs, V. R., "The Contribution of Health Services to the American Economy", Milbank Memorial Fund Quarterly. Vol. 44, October 1966, pp. 64-102. Fuchs. V. R., "Health Care and the United States Economic System - An Essay in Abnormal Physiology", Milbank Memorial Fund Quarterly. Vol. 50, April 1972. 128 Fuchs, V. R., "The Growing Demand for Medical Care",-in V. R. Fuchs, ed., Essays in the Economics of Health and Medical Care. National Bureau of Economic Research, New York, 1973. Fuchs, V. R. and Kramer, M. J., Determinants of Expenditures for Physicians' Services in the United States 1948-68. National Bureau of Economic Research, Occasional Paper 117, New York, December 1972. Gaspard, N.- J. and Hopkins, E. E., "Determinants of use of Ambulatory Medical Services by an Aged Population", Inquiry. Vol. 4, . March 1967, pp. 28-36. Gordon, J. B., "Socioeconomic Status: A Re-Examination of i t s Dimensions", Journal of Human Resources. Vol. 4, No. 3, 1969, pp. 343-359. Green, L. W., Status Identity and Preventive Health Behaviour. University of California School of Public Health, Berkeley, 1970. Greenhill, S., "Alberta Health Care Utilization Study", Canadian  Journal of Public Health. Vol. 62, January-February 1971, pp. 17-22. , Greenlick, M. R. et a l . , "Determinants of Medical Care Uti l i z a t i o n " , Health Services Research. Vol, 3, Winter 1968, pp. 296-315. Grossman, M., The Demand for Health: A Theoretical and Empirical Investigation. National Bureau of Economic Research, New York, 1972. Jeffers, J. R. et a l . , "On the Demand vs. Need for Medical Services and the Concept of Shortage", American Journal of Public Health. Vol. 61, January 1971, pp. 46-63. Jeffers, J. R. et a l . , "Rejoinder", American Journal of Public Health. Vol. 64, January 1974, pp. 58-60. Johnston, J.,' Econometric Methods (2nd ed.) McGraw-Hill, New York, 1972. Kasl, S. V. and Cobb, S., "Health Behavior, Illness Behavior and Sick Role Behavior", Archives Environmental Health. Vol. 12, February 1966, pp. 246-267. Kisch, A. I. and Kovner, J. W., "The Relationship Between Health Status and Utilization of Outpatient Health Services", Archives Environmental Health. Vol. 18, May 1969, pp. 820-833. Klarman, H. E., The Economics of Health. Columbia University Press, New York, 1965. 129 Klein, L. R., An Introduction to Econometrics. Prentice-Hall Inc., Englewood C l i f f s , N. J., 1962. Koos, E. L., The Health of Regionville: What the People Thought and  Did About It. Columbia University Press, New York, 1954. Lalonde, M., A New Perspective on the Health of Canadians. Queen's Printer, Ottawa, 1974. Lees, D. S., Health Through Choice. Institute of Economic Affairs, London, 1961. Lloyd, C , "The Demand for Medical Care: A Selective Review of the Literature", Working Paper No. 71-79, Bureau of Business and Economic Research, College of Business Administration, University of Iowa, Apr i l 1971. Long, S. G., Demand for Medical Care. Ph.D. Thesis, University of Iowa, 1969. Longhurst, P. Jr., "Demand Versus Need for Medical Services", American Journal of Public Health. Vol. 61, June 1971, pp. 1075-1076 McBroom, W. H., "Illness, Illness Behavior and Socio-economic Status", Journal of Health and Social Behavior. Vol. 11, December 1970, pp. 319-326. McKinlay, J. B., "Some Approaches and Problems in the Study of the Use of Services: An Overview", Journal of Health and Social  Behavior, Vol. 13, 1972, pp. 115-152. Migue', J. L. and Belanger, G., The Price of Health. Macmillan, Toronto, 1974. Monsma, G. N. Jr., "Marginal Revenue and the Demand for Physicians' Services", in H. E. Klarman, ed., Empirical Studies in Health Economics, Proceedings of the Second Conference on the Economics of Health, The Johns Hopkins Press, Baltimore, 1970. Monteiro, L., "Expense is No Object Income and Physician Vi s i t s Reconsidered", Journal of Health and Social Behaviour. Vol. 14, June 1973, pp. 99-115. Mueller, B. S. et a l , , "Medical Services - Demand Versus Need: A Comment", American Journal of Public Health,Vol. 64, January 1974, pp. 54-58 Mushkin, S. J., "Health as an Investment", Journal of P o l i t i c a l  Economy. Vol. 70, October 1962, pp. 129-157. Newhouse, J. P., "A Design for a Health Insurance Experiment", Inquiry. Vol. 11, March 1974, pp. 5-27. . 130 Newhouse, J. P. and Phelps, C. E., "New Estimates of Price and Income El a s t i c i t i e s of Medical Care Services", paper presented at the Universities - N.B.E.R. Conference on the Role of Health Insurance in the Health Services Sector, Rochester, N.Y., May-June 1974. Nordquist, G. L., "A Model of the Demand for Preventive Medicine Under Uncertainty", Working Paper No. 70-6, College of Business Administration, University of Iowa, 1970. Nordquist, G. L. and Wu, S., "The Joint Demand for Health Insurance and Preventive Medicine", paper presented at the Universities -N.B.E.R. Conference on the Role of Health Insurance i n the Health Services Sector, Rochester, N. Y., May-June 1974. Pauly, M. V., "The Economics of Moral Hazard", American Economic  Review. Vol. 58, June 1968, pp. 531-537. Pauly, M. V., "Information and the Demand for Medical Care", Discussion Paper, Bureau of Health Manpower Education, National Institute of Health, August 1974. Phelps, C. E., Demand for Health Insurance: A Theoretical and Empirical Investigation. RAND Corporation R-1054 - 0E0, 1973. Pollak, R. A. and Wachter, M. L., "The Relevance of the Household Production Function and i t s Implications for the Allocation of Time", Discussion Paper No. 262, Department of Economics, University of Pennsylvania, September 1973. Richardson, W. C , "Measuring the Urban Poor's Use of Physicians' Services in Response to Illness Episodes", Medical Care. Vol. 8, March-April 1970, pp. 132-142. Roemer, M. I. and Shain, N., Hospital Utilization Under Insurance. American Hospital Association, Chicago, 1959. Rosenstock, I. M., "Why People Use Health Services", Milbank Memorial  Fund Quarterly. Vol. 44, July 1966, pp. 94-124. Rosett, R. N. and Huang, L., "The Effect of Health Insurance on the Demand for Medical Care", Journal of P o l i t i c a l Economy. Vol. 81, March-April 1973, pp. .281-305. Ruchlin, H. S. and Rogers, D. C , Economics and Health Care. Charles C. Thomas, Springfield, 111., 1973. Salber, E. J. et a l . , "Utilization of Services at a Neighborhood Health Center", Pediatrics. Vol. 47, February 1971, pp. 415-423. Scitovsky, A. A., "An Index of the Cost of Medical Care - A Proposed New Approach", In S. J. Axelrod, ed., Proceedings of the  Conference on the Economics of Health and Medical Care. University of Michigan, Ann Arbor, Michigan, 1964. 131 Scitovsky, A. A., "Changes in the Costs of Treatment of Selected Illnesses, 1951-65", American Economic Review. Vol. 57, December 1967, pp. 1182-1195. Shuval, J. T., Social Functions of Medical Practice, Jossey-Bass, San Francisco, 1970. Silver, M. , "An Economic Analysis of Variations in Medical Expenses and Work-Loss Rates", in H. E. Klarman, ed., Empirical  Studies in Health Economics. Proceedings of the Second Conference on the Economics of Health, the Johns Hopkins Press, Baltimore, 1970. Solon, J. A. et a l . , "Delineating Patterns of Medical Care", American  Journal of Public Health. Vol. 50, August 1960, pp. 1105-1113. Solon, J. A. et a l . , "Episodes of Medical Care: Nursing Students' Use of Medical Services", American Journal of Public Health. Vol. 59, June 1969, pp. 936-946. Stigler, G. J., The Theory of Price. (1st ed.), Macmillan, New York, 1946. Stigler, G. J., The Theory of Price. (2nd ed.), Macmillan, New York, 1952. Suchman, E. A., "Stages of Illness and Medical Care", Journal of  Health and Human Behavior. Vol. 6, F a l l 1965, pp. 114-128. Wan, T. T. H. and Soifer, S. J., "Determinants of Physician Uti l i z a t i o n : A Causal Analysis", Journal of Health and Social  Behaviour, Vol. 15, June 1974, pp. 100-108. Wirick, G. and Barlow, R., "The Economic and Social Determinants of the Demand for Health Services", in S. J. Axelrod, ed., Proceedings of Conference on the Economics of Health and  Medical Care. University of Michigan, Ann Arbor, 1964. Zborowski, M., "Cultural Components in Responses to Pain", Journal  of Social Issues. Vol. 8, F a l l 1952, pp. 16-30. Zeckhauser, R., "Medical Insurance: A Case Study of the Tradeoff between Risk Spreading and Appropriate Incentives", Journal of  Economic Theory, March 1969. Zola, I. K. , "Culture and Symptoms: An Analysis of Patients' Presenting Complaints", American Sociological Review. Vol. 31, October 1966, pp. 615-630. B . Reports and Government Documents The Community Health Centre i n Canada Vol. I-III. Queen's Printer, Ottawa, 1972 Health Security for British Columbians. Victoria, December 1973. National Health Expenditures in Canada 1960-71. Health and Welfare Canada, Ottawa, 1973. Rapport du Commission d'Enquete sur, l a Sante^et le Bien-Etre Social. Gouvernement de Quebec, Quebec, 1970. Report of the Ontario Committee on the Healing Arts; Queen's Printer, Toronto, 1970. Report of the Royal Commission on Health Services. Queen's Printer, Ottawa, 1964. Report of the Royal Commission on Education, Public Services and Provincial-Municipal Relations. Queen's Printer, Halifax, 1974. Special Study on the. Medical Care Program for Steelworkers and their  Families. United Steelworkers of America, Pittsburgh, 1960. Task Force Reports on Costs of Health Care in Canada. Queen's Printer, Ottawa, 1969. 133 Appendix A DATA BASE AND VARIABLE MEASUREMENT The data base was obtained from Project T.E.A.M. (The Effectiveness of a Multidisciplinary Health Care Team in Meeting Patient Needs), a federally-financed and multidisciplinary research project based i n the University of British Columbia Department of Health Care and Epidemiology. Project T.E.A.M was designed to evaluate the overall effectiveness of the Fairmont Family Practice Unit (FPU) a centrally located family practice unit operated by the Department. The FPU is a clinic-type setting, staffed by physicians,residents > in family medicine and nurses, which provides primary health care to approximately 1,300 Vancouver families. The FPU also has social work, physiotherapy, and nutrition services available in order to meet the needs of individuals and families in their entirety; however, over 85% of a l l contacts involve physicians and nurses. Further, episodes which consisted exclusively of ancillary services were excluded from the samples employed for demand estimates. It should be noted that FPU personnel are salaried, which restricts the comparison of u t i l i z a t i o n results to those of other studies; how-ever, since demand analysis is the raison d'etre of this thesis, and the episodic approach is developed specifically to remove supplier influence on demand, potential bias from the method of financing and organization of the FPU is minimized, i f not eliminated. (Morever, i f the objective functions of physicians are dominated by a technologic imperative, then the demand - u t i l i z a t i o n distinction w i l l be impor-tant even in the case of salaried practitioners.) 134 Since the composition of the patient population of the FPU differs slightly from that of a typical general practitioner, the evolution of the FPU population i s worth noting. The FPU opened i t s doors i n 1969 and, since the unit was sponsored by the university and practitioners do not i n general advertize, the i n i t i a l patient population was rather select, consisting mainly of persons who had learned of the existence of the FPU through their university contacts. This situation changed i n the next few years, however, as the FPU became known throughout the city and as i t accepted referrals from community agencies, general practitioners and nearby Vancouver General Hospital. Consequently, with the exceptions of a high proportion of single-person "families" and higher education and income levels than one might expect i n cross-section data of this type (see Appendix D ), the present FPU patient population i s similar to that of most Vancouver general practitioners with respect to socio-economic and socio-demographic characteristics. . Some bias may remain i n the FPU sample with regard to health beliefs and attitudes, however, since the FPU does have teaching a c t i v i t i e s occurring simultaneously with patient care. Patients who object to the presence of health science students presumably w i l l change their source of care. Such potential attitudinal bias i s extremely d i f f i c u l t to measure. Furthermore, i n order to make conclusions on i t s significance i n this sample, a complete survey of a l l health beliefs and attitudes would be required. The d i f f i c u l t i e s associated with such a survey, and the testing of the overall health belief model, have been discussed i n Chapter 5. In the process of evaluating the effectiveness of the FPU, Project T.E.A.M. has assembled a substantial bank of data concerning the u t i l i z a t i o n experiences of patients, their socio-demographic and economic characteristics, and their needs for medical care. Data pertain to the year July 1, 1973 - June 30, 1974 and were collected par t i a l l y through direct contact between patients and FPU staff members and partially via a mailed household survey. A l l information was f i l e d by both family and individual identification numbers, thus permitting individuals, family heads or families to be employed as the unit of analysis; however, in the FPU data a "family" is defined to be a l l persons having the same family identification number. Since the family identification number i s the f i r s t four digits of the individual iden-t i f i c a t i o n number, i t i s possible that a family may consist of one person. Consequently, the term "family" does not necessarily indicate the presence of more than one person, or children. The following sections summarize the FPU data relevant to the empirical results reported i n Chapters 8, 9 and 10. Utilization Data During the study period for Project T.E.A.M., special recording procedures were instituted i n the FPU to gather extensive information on each contact (excluding telephone contacts) between staff members and patients. It was in part this recording structure that allowed the determination of episodes of medical service. In contrast, epi-sodes cannot be delineated i n the major source of Canadian u t i l i z a t i o n data, medical insurance plan records, because the procedures performed cannot be accurately linked to specific requests by, or problems of, patients. 136 For each v i s i t of every patient during the study period, the date and the number of contacts were recorded. For each contact, the following pieces of information were noted: i ) patient and FPU staff member identification numbers. i i ) presenting problem(s) - categorized on the basis of the Symptom Classification System for Health Services Research developed by the Kaiser Foundation. i i i ) diagnosis or diagnoses - categorized on the basis of the Canuck Classification System and revised and up-dated through the International Classification for Health Problems in Primary Care. iv) procedures performed - physician procedures coded as per the B. C. physician fee schedule. Internal c l a s s i -fication systems were devised for procedures performed by other FPU professionals and are available from the author upon request. v) disposition of patient - a coding which indicated at the conclusion of the contact, whether the practitioner f e l t that treatment for the patient's problem(s) termi-nated with that contact or whether the patient had been instructed to return. vi) de facto supervision period - a number indicating the number of days within which the patient was l i k e l y to return i f necessary. If no further contacts ensued for this problem, then the treatment was considered to be finished at the end of the supervision period. Data on Patient and Family Characteristics Information on the socio-demographic characteristics of patients and their families was obtained by the FPU receptionist at the time of a family's f i r s t contact and registration. Information on economic characteristics was obtained from responses to a survey mailed to a l l FPU families in May, 1 9 7 4 . Registration information included the name, age, sex and educational attainment of each family member, the occupation of the 137 household head, and the family's "situation" (see below). Survey data provided further information on family income, and distance and travel time from family residence to the FPU. Measurement of these variables is indicated by the scales below. In addition, the survey yielded "yes-no" information on whether: a) the household head was a student; b) the head was unemployed at the time of the survey; c) i f employed, the head's employer allowed the head paid time off work to seek medical care for himself/herself; and d) the head was eli g i b l e for d i s a b i l i t y benefits in the event of i l l n e s s . Variable Scale i ) educational attainment i i ) family situation (no formal schooling/elementary/ some high school/ completed high school/completed tech. or voc. school ( i n c l . nursing school and community college)/ some university/university degree ('living alone/alone with child or children/ two adults l i v i n g together/two adults with child or"children/group of adults/ group of adults with child or children). i i i ) head's occupation 2 iv) family income v) distance to FPU (coded by Blishen (1967) scale) (less than $999/1,000 - 2,999/ 3,000 - 4,999/ 5,000 - 6,999/ 7,000 - 8,999/ 9,000 - 10,999/ 11,000 - 13,999/14,000 - 16,999/ 17,000 - 19,999/ 20,000 - 22,999/ 23,000 - 25,999/ $26,000 or more) (in miles - less than 1/1 - 2.9/ 3 - 4.9/5 - 6.9/ 7 - 8.9/ 9 - 10.9/ 11 - 13.9/ 14 - 16.9/ 17 - 19.9/ 20 - 22.9/ 23 - 25.9) vi) travel time to FPU (in minutes - less than 5/ 5 10 - 19/ 20 - 29/ 30 - 59/ 60 or more). 9/ 138 Notes 1. The Blishen scale assigns a number between 25.25 and 76.69 to occupations categorized on the basis of 1960 census information. Although i t is the most commonly used scale, i t does have serious limitations. It is a cumbersome scale to apply, and is rapidly becoming outdated some inconsistencies i n the rankings of occupation by status are presently obvious. Most important, however, i s the fact that the numbers themselves are, other than ordinally, meaningless. Consequently, the status score must be converted to dummy variables for use in regression analysis. Decisions on the number of such dummies are somewhat arbitrary; however, in this case three status groups were identified after examining patterns in the frequency distribution of the raw Blishen score. 2. This scale did not extend to a high enough income level, with the result that the t a i l of the frequency distribution rose sharply. Although this does not impair the use of the income variable i n dummy form, results employing the continuous form of the variable may be biased depending upon the value chosen for the f i n a l c e l l . Consequently, for the few estimations that i t affected, the value selected for the f i n a l c e l l was arrived at by estimating a Pareto distribution as per Klein (1962; pp. 151-154). 139 Appendix B ALGORITHMS FOR DETERMINATION OF EPISODES OF MEDICAL SERVICE Since the following algorithms operate on data described i n the preceding appendix, readers are advised to begin by reviewing Appendix A. For each patient i n the Fairmont Family Practice Unit Data Base the following algorithms were applied to his/her record of v i s i t s , contacts, and procedures: 1. Occurrence of a new presenting problem ini t i a t e s a new episode, unless the new presenting problem i s closely related to a problem or problems for which episodes either are currently i n progress or have recently (see alg.2.) been completed. If d i f f i c u l t y i s encountered in deciding whether the new problem i s a related one, consult the diagnosis accompanying the new problem. If the diagnosis has not changed from the previous prob-lem(s) then the original episode continues. If the diag-nosis has changed, then i n i t i a t e a new episode. 2. Repetition of the same or related presenting problem(s) calls for the continuation of an episode, unless the contacts repeating the problem are separated by more than 15 days. Exceptions to this 15-day rule are per-missible over Christmas and other extended holiday periods. Exceptions are also permissible in the case of i ) con-tacts separated by more than 15 days, but s t i l l within the de facto supervision period, or i i ) contacts for conditions for which extended periods are required to assess treatment effectiveness (e.g. fractures, birth control advice). 3. A l l contacts, on a regular basis, for desensitization constitute a single episode. Similarly for obesity counselling. If the regularity of such contacts ceases, and resumes at a later date, then resumption in i t i a t e s a new episode. 4. A l l contacts for monitoring and/or maintenance of chronic conditions (e.g. diabetes, hypertension, a r t h r i t i s ) , on a regular basis, constitute a single episode. If " f l a r e -ups" of chronic conditions occur, then contacts for such "flare-ups" are subject to the algorithms l i s t e d above. 140 C o n t a c t s f o r d e p r e s s i o n , p a r a n o i a , and o t h e r e m o t i o n a l p r o b l e m s a r e h a n d l e d s i m i l a r l y t o c h r o n i c c o n d i t i o n s . I n p a r t i c u l a r , t h e y a r e n o t s u b j e c t t o t h e 15-day a l g o r i t h m . 5. A l l c o n t a c t s f o r r e l a t e d p r o b l e m s l e a d i n g t o t h e d i a g -n o s i s o f pregnancy c o n s t i t u t e a s i n g l e e p i s o d e . From t h e d i a g n o s i s o f p r e g n a n c y t h r o u g h n o r m a l d e l i v e r y t o immediate p o s t n a t a l c a r e c o n s t i t u t e s a n o t h e r e p i s o d e . C o m p l i c a t i o n s o f t h e p r e g n a n c y a t any p o i n t a r e h a n d l e d s i m i l a r l y t o " f l a r e - u p s " . D u r i n g p r e g n a n c y , c o n t a c t s f o r u n r e l a t e d c o n d i t i o n s and pr o b l e m s a r e h a n d l e d t h r o u g h t h e b a s i c a l g o r i t h m s l i s t e d above. These a l g o r i t h m s were s u f f i c i e n t t o d e t e r m i n e e p i s o d e s i n t h e m a j o r i t y o f c a s e s , w h i l e s t i l l a l l o w i n g some l a t i t u d e f o r i n t e r p r e t a t i o n and r e c o n s i d e r a t i o n o f e x t r e m e l y a t y p i c a l c a s e s . I n c e r t a i n i n s t a n c e s t h e a d v i c e o f members o f P r o j e c t T. E. A. M. and/or t h e F a i r m o n t F a m i l y P r a c t i c e U n i t was so u g h t and a p p r e c i a t e d . A l t h o u g h t h i s a p p e n d i x does n o t employ t h e p r e c i s e names o r number o f c o n d i t i o n s as l i s t e d i n o f f i c i a l p r e s e n t i n g p r o b l e m and d i a g n o s i s c l a s s i f i c a t i o n s , t h i s i n f o r m a -t i o n c a n be o b t a i n e d f r o m t h e a u t h o r on r e q u e s t . Mock examples o f e p i -sode d e l i n e a t i o n , and f u r t h e r i n f o r m a t i o n on i n t e r p r e t a t i o n o f r e l a t e d p r o b l e m s , de f a c t o s u p e r v i s i o n p e r i o d s , and m u l t i p l e p r o b l e m c o n t a c t s may a l s o be o b t a i n e d on r e q u e s t , b u t w o u l d occupy an amount o f s p a c e h e r e q u i t e o u t o f p r o p o r t i o n t o t h e i r c o n t r i b u t i o n t o t h e t h e s i s . N o te t h a t , s i n c e e a c h c o n t a c t has t h e p o t e n t i a l t o i n i t i a t e a new e p i s o d e , and e p i s o d e s may t e r m i n a t e i n d e p e n d e n t l y o f one a n o t h e r i n a c c o r d a n c e w i t h t h e a l g o r i t h m s , c o n c u r r e n t and i n t e r l e a v e d e p i s o d e s do n o t n e c e s s i t a t e any f u r t h e r , o r s p e c i a l , a l g o r i t h m s . They a r e f u l l y d e t e r m i n e d i n t h e c o u r s e o f a p p l y i n g a l g o r i t h m s 1 - 5 above. The r o l e o f d i s p o s i t i o n i n f o r m a t i o n , however, r e q u i r e s some e x p l a n a t i o n . O r i g i n a l l y i t was t h o u g h t t h a t s u c h i n f o r m a t i o n w o u l d be c e n t r a l t o a t t e m p t s t o d e l i n e a t e e p i s o d e s ; however, t h e e v e n t u a l s p e c i f i c a t i o n o f t h e a l g o r i t h m s s u b s t a n t i a l l y r e d u c e d t h e i m p o r t a n c e o f t h e d i s p o s i t i o n c o d i n g . I n a d d i t i o n t o f a c i l i t a t i n g d e c i s i o n s i n t h e few c a s e s where a l g o r i t h m i c r e s u l t s were ambiguous, and i n some complex c a s e s o f r e l a t e d p r e s e n t i n g p r o b l e m s , d i s p o s i t i o n i n f o r m a t i o n a s s i s t e d i n m o n i t o r i n g t h e p e r f o r m a n c e o f t h e a l g o r i t h m s . I t i s r e c o g n i z e d t h a t t h e d e t e r m i n a t i o n o f e p i s o d e s o f m e d i c a l s e r v i c e as o u t l i n e d I n t h i s s e c t i o n i s s t i l l s u b j e c t t o s e v e r a l i m p r o v e -ments c a s e i n p o i n t , t h e number o f days s p e c i f i e d i n a l g . 2 s h o u l d be more s e n s i t i v e t o the p r e c i s e p r o b l e m however, i f t h e above p r e s e n -t a t i o n , and C h a p t e r 6 , have a t l e a s t e s t a b l i s h e d t h e e p i s o d i c a p p r o a c h and i l l u s t r a t e d i t s m e r i t s , t h e n o u r p r e s e n t o b j e c t i v e s have been a c h i e v e d . 142 Appendix C EXAMPLE OF RESULTS PROVIDING BASIS FOR GROUPING  AND/OR DELETION OF DUMMY VARIABLES Dependent Variable: IES: A l l Individuals Variable Coefficient t - value INC1 0.710 (0.932) INC2 ,0.797 (1.734) INC3 1.191 (2.752) INC4 0.067 (0.190) INC5 -0.607 (1.467) INC6 -0.250 (0.620) STUHEAD -0.600 (1.155) UNEMPLOY 0.608 (1.729) SEX 0.529 (1.801) SAG 1.186 (2.912) AG1 2.063 (1.984) AG2 0.100 (0.214) AG3 -0.370 (0.879) AG4 -0.413 (0.883) AG5 -0.522 (1.308) AG6 -0.788 (1.543) LSI 0.122 (0.276) LS2 -0.821 (1.637) LS3 0.455 (1.296) LS4 0.502 (0.639) LS5 0.614 (0.903) LS6 -1.506 (1.791) BL1 0.758 (2.180) BL2 0.518 (1.574) BL3 0.199 (0.556) EDS 2 -1.168 (1.368) EDS 3 -0.513 (1.432) EDS4 -1.018 (2.690) EDS 5 -0.401 (1.150) EDS 6 0.172 (0.467) CONSTANT 2.693 N = 418 R, = .20 R = .135 F = 3.180 143 Appendix D VARIABLE LISTS, MEANS, AND STANDARD DEVIATIONS I. EQUATIONS FOR ALL INDIVIDUALS (N=418) Standard Name Variable Mean Deviation IES number of episodes of medical service 3.23 2.39 ICONTCS number of contacts 6.79 7.61 IEXPINS dollar expenditures 71.09 67.86 DIST dist. from family residence to FPU (in miles) 5.64 4.91 INCA family income $0 - 4,999/base 9,000 - 19,999 .16 .36 INCB family income $5,000 - 8,999 • .11 .32 INCC family income $20,000 or more .29 .45 INC1 family income $0 - 999/base 9,000 - 13,999 .02 .17 INC2 family income $1,000 - 4,999 .13 .34 INC 3 family income $5,000 - 8,999 .11 .32 INC4 family income $14,000 - 19,999 .20 .40 INC5 family income $20,000 - 25,999 .14 .34 INC6 family income $26,000 or more .16 .37 FY2 family income (hundreds of dollars) 160.23 108.65 STUHEAD household head a student .06 .23 UNEMPLOY household head unemployed .23 .42 SEX females .58 .49 SAG females 18 - 40 yrs. .28 .45 AG1 under 1 yr. .02 .11 AG2 1 - 5 yrs. .09 .29 AG3 6 - 1 5 yrs. .12 .33 AG4 16 - 22 yrs. .12 .32 AG5 23 - 40 yrs. .38 .49 AG6 over 60 yrs. .09 .29 BL1 unskilled & blue collar/base white collar .17 .37 BL2 professional .20 .40 BL3 missing .19 .39 EDSA3 less than high school/base high school .23 .42 EDSB tech.-voc. school, comm. college, nursing sch. .15 .36 EDSC at least some university .41 .49 EDS1 no education/base high school 0.0 0.0 EDS2 elementary school .02 .15 EDS3 some high school .21 .40 EDS4 tech.-voc. school, comm. college, nursing sch. .15 .36 EDS 5 some university .21 .40 EDS 6 university degree(s) .20 .40 EDUCATS years of formal schooling 12.77 2.59 LSA l i v i n g alone/base two adults plus child. .12 .32 LSB two adults .19 .39 LSC other fam. s i t . except two adits. plus child. .14 .35 LSI l i v i n g alone/base two adults plus child. .12" .32 144 Name Variable Mean Standard Deviation LS2 l i v i n g alone with child(ren) .06 .25 LS3 two adults .19 .39 LS4 group of adults .02 .14 LS5 group of adults plus child(ren) .03 .18 LS6 missing .02 .14 FAMSIZE family size 3.18 1.61 ATTEND* individuals with at least one episode 1.0 0.0 * N = 580 for equations employing ATTEND a For individuals under 18 yrs., a l l EDS variables take on the value of the mother's education. If mother's education i s missing then the head's value i s assumed. Name II. ADDITIONAL OR REPLACEMENT VARIABLES FOR FAMILY HEADS EQUATIONS (N=202) Variable Mean Standard Deviation IES * IC0NTCSA IEXPJNS DIST PTOSA PTOSB INCD INCE ^ STUHEAD i UNEMPLOY DISBENL A G V S E X * B L 1 * BL2 EDA EDB EDC EDI ED2 ED3 ED4 ED5 ED6 able to get paid time off to obtain med. care answer to PTOSA is don't know or no response family income $0 - 8,999/base 9,000 - 19,999 family income $20,000 or more eligi b l e for d i s a b i l i t y benefits 18 - 22 yrs. tech.-voc.-nursing school or comm. college at least some university no education elementary school/base high school some high school tech.-voc.-nursing school or comm. college some university university degree(s) 3.07 2.35 6.10 5.57 52.00 50.95 4.82 4.20 .49 .50 .29 .46 .38 .48 .18 .39 .06 .25 .28 .45 .43 .50 .05 .22 .44 .49 .17 .37 .11 .32 1 .20 .41 .12 .33 .41 .49 0.0 0.0 .03 .18 .17 .38 .12 .33 .18 .39 .22 .42 * Variable identical to l i s t I above. 145 III, ADDITIONAL OR REPLACEMENT VARIABLES  FOR FAMILY EQUATIONS* (N=236) Standard Name Variable Mean Deviation FES no. of episodes of medical service for a family 6.07 5.07 FCONTCS number of contacts for a family 12.66 12.09 FEXPLNS dollar expenditures for a family 131.31 130.14 DIST A 5.21 4.52 STUHEAD A .07 .25 UNEMP.LOY .26 .44 I N C D * .34 .48 INCE .19 .40 HAG1 age of head, 18-22 yrs./base 41-60 yrs. .04 .20 HAG 2 age of head, 23-40 yrs. .50 .50 HAG3 # age of head, over 60 yrs. .14 .35 FAMSIZE 2.45 1.48 FS1 family size of 1/base 3-5 .35 .48 FS2 family size of 2 .25 .43 FS3 A family size of 6-8 .03 .17 L S 1 * .21 .41 L S 2 * .07 .25 L S 3 * .25 .43 L S 4 * .04 .19 LS5* .02 .16 LS6 .03 .17 EDSAa less than high school/base high school .22 .41 EDSB tech.-voc.-nursing school or comm. college .17 .38 EDSC. at least some university .35 .47 BL1 A .18 .38 BL2 .15 .36 * Variable identical to l i s t I and /or II a For a l l EDS variables, value for family determined by mother. If no mother, value determined by head. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0100074/manifest

Comment

Related Items