UBC Faculty Research and Publications

Lifestyles, linkage and liabilities : theory and evidence justifying economic incentives for healthier… Barer, Morris Lionel, 1951-; Evans, Robert G., 1942-; Stoddart, Gregory Lloyd, 1948-; Labelle, Roberta J; Fulton, Jane Nov 30, 1984

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

Item Metadata


52383-Barer_M_et_al_Lifestyles_linkages_and_liabilities.pdf [ 101.99MB ]
JSON: 52383-1.0075952.json
JSON-LD: 52383-1.0075952-ld.json
RDF/XML (Pretty): 52383-1.0075952-rdf.xml
RDF/JSON: 52383-1.0075952-rdf.json
Turtle: 52383-1.0075952-turtle.txt
N-Triples: 52383-1.0075952-rdf-ntriples.txt
Original Record: 52383-1.0075952-source.json
Full Text

Full Text

Report S:l7 LIFESTYLES, LINKAGES AND LIABILITIES: Theory and Evidence Justifying Economic Incentives for Healthier Lifestyles Morris L. Barer, Robert G. Evans, Greg L. Stoddart, Roberta Labelle and Jane Fulton Division of Health Services Research and Development Office of the Co-ordinator of Health Sciences The J.F. McCreary Health Sciences Centre The University of British Columbia Vancouver, B.C., Canada V6T 1Z6 November 1984 This report has also been released as Research Report #112, Program for Quantitative Studies in Economics and Population, Faculty of Social Sciences, McMaster University. LIFESTYLES, LINKAGES AND LIABILITIES: Theory and Evidence Justifying Economic Inc-entives for Healthier Lifestyles Morris L. Barer,Ph.D. Associate Director Division of Health Services Research & Development Office of the Co-ordinator, Health Sciences & Assistant Professor Department of Health Care and Epidemiology University of British Columbia Robert G. Evans, Ph.D. Professor Department of Economics University of British Columbia Greg L. Stoddart, Ph.D. Associate Professor Department of Clinical Epidemiology & Biostatistics & Associate Member Department of Economics McMaster University Roberta Labelle, M.A. Research Associate Department of Clinical Epidemiology and Biostatistics McMaster University Jane Fulton, M.Sc. Doctoral Candidate University of British Columbia November, 1984 Acknowledgements Any project of this magnitude inevitably requires the contribution and support of individuals and organizations who just as inevitably do not get identified on the front page. We owe a particular debt to Dr. Fred Bass, for revealing to us the significance of COLD. Also offering assistance of a variety of types and at various times, were Gordon Wong, Brian McCashin, Cheryl Jackson, Susan Moloney, Mary Brunold, Annette Stark and Kent Brothers. We are grateful for the financial support provided by the Health Promotion Branch of the Ontario Ministry of Health. The research was also supported in part by National Health Research Scholar Award 6610-1231-48. TABLE OF CONTENTS Chapter I INTRODUCTION 1 .1 Background - Lifestyle Choice as a Target of Public Policy . 1 .2 Outline of the Report ...... . 4 11 II LIFESTYLES, RATIONALES FOR PUBLIC POLICY, AND ECONOMIC INCENTIVES . . . . . . . . . . . . . 14 2.1 Lifestyle Characteristics - A Working Taxonomy 14 2.2 Lifestyles and Public Intervention 18 2.3 The Rationale for Economic Incentives .21 2.4 Economic Pol icy Options . . . . . . . 25 Ill JUSTIFICATION AND CHOICE - A DECISION-THEORETIC FRAMEWORK FOR IV ECONOMIC INCENTIVES . . . . . . . . . . . . . . 30 3.1 Assumptions Underlying Efficiency Goal 3.2 Assumptions Underlying Equity Goal 3.3 A Generalized Linkage Framework ... . 3.4 Economic Incentives Revisited .... . 3.5 Some Final Thoughts on Information Requirements THE EPIDEMIOLOGIC CONNECTION 4.1 The Search for Causality - Methodologic Issues 4.2 The Epidemiologic Evidence on Smoking and Alcohol 4.2.l Smoking as Causal Agent .. . 4:2.la Lung Cancer ..... . 4.2.lb Coronary Heart Disease .... . 4.2.lc Chronic Obstructive Lung Disease . 4.2.ld Other "Causal" Links .. 4.2.2 Alcohol Consumption as Causal Agent 4.2.2a Cirrhosis of the Liver .. 4.2.2b Other "Causal" Links ... 4.2.3 Thoughts on the Epidemiologic Link V REASSESSING THE ROLE OF ECONOMIC INCENTIVES - THE ECONOMIC . . . 31 35 36 52 55 58 58 75 75 75 85 104 114 118 118 130 136 EVIDENCE . . . . . . . . . . . . . . . . . . . . . . 143 9.1 Summary and Role of Epidemiologic Evidence 143 5.2 Evidence from Upstream - Price Elasticities of Demand for Cigarettes ... 5.2.l 5.2.2 5.2.3 5.2.4 Introduction . . . . . . . . . . . General Methodology ..... . Evidence on Demand Elasticities ... . Discussion .............. . 5.3 Evidence on the Equity Objective - Do Smokers Pay 145 145 146 148 152 Their Way? . . . . . . . . . . . . . . . . . 153 5.3.l Introduction . . . . . . . . . . . . . . . . . 153 5.3.2 Methodology Employed in the Estimation of Smoking-related Health Care Costs and Tobacco Tax Revenues . . . 156 5.3.3 Sensitivity Analysis . . . 174 5.3.4 Discussion of Results 184 5.4 Economic Incentives Reconsidered 186 VI THE NOT-SO-SOVEREIGN CONSUMER - ALTERNATIVE PERSPECTIVES ON "VOLUNTARY CHOICE" . . . . . . . . . . . . . . . . . 194 VII 6.1 Alternative Perspectives to Economic Choice -A Simple Framework ............ . 6.2 Implications of Alternative Views of Lifestyle Determination . 6.3 Summary ............ . SUMMARY, 7. l 7.2 7.3 7.4 7.5 7.6 POLICY IMPLICATIONS AND ALTERNATIVE DIRECTIONS Recalling the Arguments for Economic Incentives From Theory to Reality - Assumptions as Evidential Requirements ........... . Testing the Assumptions - A Brief Recap of the Empirical Evidence ....... . The Carrots and Sticks Revisited Alternative Directions for Policy In Sum REFERENCES 195 199 214 217 217 219 223 232 234 242 247 CHAPTER 1: INTRODUCTION "Even for those conditions that undeniably have an causal component, the luxury of human innocence is a plausible defense against human accountability. we learn about disease and health and their causal the more we have the possibility of intervening to those chains of causation." organic no longer The more chains, change Veatch (1980, 51) The call to arms over individual lifestyles is reaching epic (or epidemic) proportions in this country. The proliferation of no-smoking areas, and the federally-sponsored Participaction movement are perhaps the most visible signs of a Canadian groundswell away from the "germ theory" of illness. 1 But other equally representative though less evident signs of this shift in focus emerge in public and private debate over the interpersonal consequences of unhealthy lifestyles. Canada's tax and welfare structures are a dynamic blend of altruistic and paternalistic mores. Thus, the indigent are not taxed, their health insurance premiums are subsidized, and charities and volunteer organizations flourish (and financial contributions to the former are generally tax deductible). The "health care as a right" movement was one of the pressures leading to our universal health insurance system which, although currently having to withstand a fair bit of internal buffeting from health care providers, remains a model admired outside and strongly supported within this country. Even at the level of the individual, there are few who would admit that they - 2 -are happier when any given acquaintance, friend or family member is ill than when that person is well. This individual sympathy may dissipate rapidly, however, as one moves beyond the small circle encompassing family and friends and as one focuses on illness states perceived to result from personal voluntary consumption or, more broadly, behavioural or lifestyle decisions. The drinking driver who sends himself to hospital but eliminates a carload of quite sober citizens on the way, is bound to evoke strong feelings; but they are unlikely to be feelings of sympathy. In a similar vein, an accident victim who neglected to fasten a seatbelt and a "two pack a day" smoker who contracts lung cancer may have more than incapacity in common. They, too, may have difficulty raising a quorum of sympathetic ears, and individual taxpayers may question the use of public funds to care for what are perceived as avoidable cost burdens. In short, "some people are beginning to ask why they should have to pay higher taxes or health insurance premiums to care for the victims •••• " of individual voluntary choice (Somers, 1980). This question becomes embodied in policy proposals targeting the "culprits" or their "vices". This concern (which we will argue below has actually two strands --over the absolute level of resources devoted to health care rather than other goods and services, and over the equitable distribution of those resources) and the suggestions arising out of it, manifest themselves at a number of different levels of debate. At the popular level, for example, we find elected representatives such as David Russell (Minister of Health of Alberta) dragging the "guy who's drinking too much and smoking too much" (Edmonton Journal, June 27, 1979) into the debate over direct charges or user fees. The issue receives much more extensive attention at the public ., - 3 -policy research level. As the social "price" of individual choices increases (because, for example, health care shifts to sole source (public) funding), or awareness of a price which was "always" there becomes more acute with increased information, those ohoioes may become increasingly labelled (implicitly) as "deviant" behaviour (Slaby and Tancredi, 1975). This results in pressure to adjust the individual "prices" faced by those engaging in the targeted lifestyle characteristics (Zook et al., 1981; Wilkinson et al., 1978; Heins, 1978). Policy prescriptions or suggestions have taken the form of direct charges (e.g., coinsurance or deductibles) on health care costs incurred as a result of voluntary individual choices (Fraser, 1981); structuring health insurance premiums to take account of lifestyle risk factors (Zook et al., 1981; Brailey, 1980; Warner, 1979); or taxing the offending commodities or activities (Wilkinson et al., 1978; Heins, 1978; Kristein and Grove, 1978). Of course no debate is one-sided, and many dissenting views or cautions are offered. These generally seem to relate to the ethical issues of government involvement in private decision-making (Wikler, 1978; Veatch, 1980), the invalidity or questionability of the voluntary individual choice model (Veatch, 1980; Ryan, 1971; Journal of Public Health Policy, 1980; Holtzman, 1979; Wikler, 1978), or the lack of the requisite information necessary to justify such economic interventions (Mettlin, 1979; McKeown, 1976; Evans, 1978). There are also those who seem confused as to the type of information one might require (Zook et al., 1981), or as to the current state of that informational base (Harnes, 1978; Veatch, 1980). There seems, however, to be little integration of these seemingly - 4 -diverse. themes, and no evidence of a convergence of opinion. While issues of ethics and moral fortitude are beyond the narrower scope of this report, we attempt to integrate some of these issues through a conceptual and empirical focus on the information and evidence necessary to justify economic incentives directed at efficiency and equity objectives of public policy. It turns out that such an approach provides a sufficient framework within which to integrate behavioural issues related to "lifestyle modelling". Furthermore, often ethical issues turn out to be integrally linked with issues of evidence and therefore may or may not become policy issues depending on a given state of information. In the remaining sections of this chapter we offer first some rationales for the emergence of individual lifestyles as an issue for public policy, and then our objectives and coverage for the present research. 1.1 Background Lifestyle Choice as a Target of Public Policy In all industrialized countries, the level and rate of growth of health care costs have become matters of continuing, and often urgent, public concern. The last three decades have seen dramatic increases in the share of total resources which these societies devote to health care. In the private sectors of these economies, comparable growth in sales2 would be cause for hand-shaking and poorly concealed smiles as one looked forward to the next annual shareholders' meeting. But it is now generally agreed among students of health care delivery that "more" does not necessarily imply better for this sector of the economy. - 5 -Furthermore, over the same three decades, the public sector in most countries has taken on increasing and now major responsibilities for raising the resources necessary to support those sales, and for distributing tnem throughout the health care system. Th~s, health care has emerged as a major public policy issue both because of its sheer scale expenditures in 1977 varying across industrialized countries from just over 5% of GNP in Great Britain to almost 10% in Sweden (Maxwell, 1981)--and powerful internal pressures for expansion, and because its claims on resources are now an open part of the political process, exercised in competition with all the other social priorities which place claims on the same limited pool. Canada's response to the health care financing problem has been remarkably successful, when compared with experiences elsewhere. In 1960, when the provinces were still in the process of being drawn in to the national hospital insurance program (HIDSA), Canada's health care share of GNP exceeded that of Australia, France, West Germany, the Netherlands, Sweden, Great Britain, the United States and Finland (Simanis and Coleman, 1980; Maxwell, 1981). By 1965, the United States had either caught up (Maxwell, 1981), or was nipping at our heels (Simanis and Coleman, 1980). Even in 1970 we stood third among these countries (behind Sweden and the United States) by this proxy for share of national resources devoted to the delivery of health care. But the following year saw a temporary peak (at 7.4% of GNP) in this ratio in Canada, and by 1975 only Great Britain, Australia and Finland were spending a smaller share on health care. Australia and Finland both passed Canada the following year (Simanis and Coleman, 1980). Of course countries for which early data are not easily obtained, such as Italy - 6 -and Switzerland, still ranked below Canada as recently as 1977 (Maxwell, 1981), but the message is clear. While elsewhere the problem is often conceived of as "cost-capping" pure and simple, containing by whatever means a runaway health industry whose "spiralling costs" eat up an ever larger share of society's resources, in Canada the share of the health care industry in GNP stabilized in the 7.0-7.5% range throughout the 1970s. Only very recently has it moved up over 8% (at least partially because the health care sector was relatively well shielded from the general recession of 1982-83). There are even claims by industry spokesmen that the Canadian system as a whole is underfunded, although these have not won any general acceptance and continue to appear alongside counterclaims that costs are still 'spiralling'. We have some difficulty reconciling these two positions and, although we find the evidence of runaway costs somewhat lean, it certainly does not follow that the system is underfunded. In fact, this relative stability has been achieved in the context of a delivery system which has enormous public support. In contrast to rather mixed public attitudes toward health care delivery/financing systems in some other countries, in Canada we seem to find not merely acceptance but positive pride. Both the (much less expensive) National Health Service in the U.K. and the (much more expensive) American health care industry appear to meet more severe criticism on a number of grounds. But any casual perusal of urban daily newspapers or the Medical Post should provide a ready dose of cold water for those feeling carried away by our good fortune or the success of our astute health care planning. Problems, even apparent crises, in health care finance and - 7 -delivery are reported and editorialized with numbing frequency. They remain topics of urgent public concern for several reasons. First, cost control is not an end in itself. The more fundamental questions have to do with the effectiveness with which resources are used within the health care system, and their value there rather than in the alternative uses to which they could be put. Canadian governments, provincial governments in particular, share with health care suppliers the responsibility for ensuring that the very large share of public resources devoted to health care is used effectively to promote health. As yet Canada's achievements in this area, while not negligible, are much less striking than its success in overall cost control. Second, the control process is difficult and politically bruising. Provincial governments are under continuous pressure from all sorts of interest groups to open up the flow of resources into health care. The objectives of such groups may be selfish or by their own lights public spirited, but they all come down to "more". And given the politically sensitive nature of health care, and its association with personal integrity (or existence) and basic human values, conflicts in this area receive extensive public attention and are politically very dangerous. Finally, there is no guarantee that the past decade's success in cost control can be repeated. Indeed, evidence from the early 1980s suggests the contrary. The rapid diffusion of medical technology, the aging of the population, and the long-term consequences of rather short-sighted but difficult to reverse manpower policies in the late 60 1s and early 70's are all creating powerful expansionary forces which may be difficult to contain by simply sitting on the financial lid. These growing pressures are most apparent in the increasing political conflict between provincial governments and physicians over fee schedules, but - 8 -the problems run deeper. It is in the search for responses to these current or seemingly imminent problems that lifestyles make their way into the public policy forum. Responses and suggestions for "reform" in general are numerous and varied, and most need not occupy our detailed attention here. Suffice to note that most policy initiatives may be categorized as focusing on the "supply side" or the "demand side" of the health care utilization process. The separation between the two cannot be made clearly, as it is generally recognized that "demanders" of health care, patients, are heavily influenced in their utilization choices by providers--physician advice, hospital policy and regulations. For this reason, most study of the health care system, at least in Canada, has focused on the relationship between provider incentives and behaviour and public objectives. There remains a continuing undercurrent of interest, however, in the "demand side", the factors which lead individuals to seek care and/or to accept it when recommended. In its crude form, study of the "demand side" has focused on the response of patients to the presence or absence of point-of-service charges in deciding whether or not to use health care. Some analysts, borrowing standard economic models off-the-shelf without considering the special economic and non-economic characteristics of health care, have simply pointed to public or private insurance as the causal factor driving health care cost increases (Feldstein, 1981; Pauly, 1969; etc.). Reintroduction of direct charges to patients in some form or other is then advocated as a cost-control measure (Fraser, 1981; Detwiller, 1979). This argument has a number of variants, but none appear to bear up under detailed scrutiny of logical consistency or empirical support - 9 -(Barer, Evans and Stoddart, 1979). Thus this line of policy has found little support outside the U.S. where, of course, an extensive system of direct patient charges coexists with very high and uncontrolled costs. In Canada, such policies have received unequivocal rejections in the Hall review of medicare (Hall, 1980), in the Task Force report, Fiscal Federalism in Canada (Canada, 1981b), and most recently in the new Canada Health Act. A rather different line of "demand-side" analysis, however, seems rooted (implicitly if not explicitly) in a "voluntary model" of health status determination (Veatch, 1980) which views the individual's health status over time as being largely under personal control. The health status of individuals is at least correlated with and even appears to depend to a significant extent on how they live (McKeown, 1976; Fraser, 1981; Fuchs, 1974; Pomerleau et al., 1975; are but a few of literally hundreds of references providing or reporting correlational evidence). This was certainly one of the major messages for Canadians in the federal White Paper, ! New Perspective ~ the Health of Canadians (Lalonde, 1974). Now it is obviously silly to regard the victim of a heart attack as someone who has suddenly acquired a ravenous and urgent "taste" for medical services, and who should therefore be expected to purchase these in a private market as if s/he had a sudden craving for a new car or a chocolate bar. But in a longer time perspective, one might view that individual as "choosing" a lifestyle -- more or less exercise, level of dietary fat intake, smoking or non-smoking, stress level of occupation -- which significantly influenced his/her risk of heart attack. If indeed people do to a large extent bring their illnesses and ailments on themselves by the way they choose to live then, so the argument runs, - 10 -they ought individually to bear the economic consequences of those decisions. One way of bearing such consequences would be through some degree of direct patient payment for health care services. This recruitment of the federal White Paper to the support of reintroduction of direct charges is a rather interesting logical twist. The main thrust of that paper was to argue that present information on patterns of morbidity and mortality in Canada indicated a limited payoff to further investments of resources in the health care system as such. It thus provided the intellectual basis for cost-containment. But it also indicated that there were clearly improvements in health status still to be achieved; the point was that these would require programs to improve lifestyles and environments rather than more hospitals, doctors, or medical technology. It is doubtful if the re-introduction of charges to patients was contemplated as one such program! Nevertheless, the attention drawn to individual lifestyles by the Lalonde report and other literature reporting correlations of lifestyle choice with disease/accident incidence, seems to have served the (unwitting) purpose of providing an excuse for the re-emergence of demand side cost control policies. In short, pressures on health care costs and a somewhat loosely structured but "appealing" link between a significant portion of those costs and individual decisions, seem to have dragged lifestyles to the stage, front and centre. In the following Chapter we will examine in more detail the apparent rationales for demand side incentives directed at "self-inflicted" illnesses or accidents. - 11 -1.2 Outline of the Report The major intent of this report is to identify, and then to illustrate using a number of well known examples, the information which must be available and the conditions which must be satisfied in order for a policy of economic intervention in individual lifestyles to be successful in achieving specific public policy objectives. We begin in the following Chapter by developing a working definition of lifestyle choices, and by proposing two rationales for public economic intervention in those choices: the efficient use of scarce resources, and the equitable distribution of that share of those resources devoted to health care delivery. In Chapter 2 we consider also the more general case for government involvement (of an economic or other nature), and discuss some of the possible forms economic interventions might take. In Chapter 3 we begin the task of carefully delineating the general informational requirements which seem to us necessary underpinnings to justify any economic intervention. This involves articulating the causal chain running from lifestyle-related decisions through physiological states to health care use and costs. The links in that chain are then related to specific examples of economic interventions, to demonstrate that the justifications for different interventions require different types of information. In Chapter 4, we narrow the focus to a number of lifestyle/disease linkages which are widely regarded as currently coming closest to satisfying the informational requirements identified in Chapter 3. Specifically, we review the epidemiologic literature on diseases related to smoking, with an emphasis on links to lung cancer, coronary heart disease, and chronic obstructive lung disease; and diseases - 12 -related to alcohol (cirrhosis of the liver). If current evidence on these widely studied linkages is insufficient or equivocal, then the case for economic interventions directed at other less thoroughly investigated lifestyle traits is considerably weakened. Another piece of the causal chain identified in Chapter 3 as being necessary if economic incentives are to impact on the efficiency objective, is the behavioural response to the incentives by those making the lifestyle decisions. In Chapter 5 we consider the current evidence on that issue. In turn, the equity argument for economic incentives requires that those involved in the targeted lifestyles are not already "paying the piper". Therefore, Chapter 5 also examines as a case study, net fiscal incidence relating to one of the two behaviours reviewed in Chapter 4 (smoking), and attempts to determine net gainers and losers from the current mix of taxes, subsidies and health care costs. The analysis of Chapters 3 through 5 is predicated on the "voluntary choice" view of lifestyles. In Chapter 6 we consider other models of "individual" behaviour, and argue that evidence on rational and voluntary behaviour forms yet another piece of the informational chain necessary to justify individual economic interventions. The information base necessary for public policy employing economic incentives turns out to be extensive and complex. Furthermore, this base is shown to be incomplete even for the strongest current epidemiological links. Accordingly, in Chapter 7 we discuss briefly some alternative directions for public policy, and offer some tentative conclusions regarding the role of economic incentives in lifestyles modification. - 13 -FOOTNOTES TO CHAPTER .! 1 The "germ theory" (Tesh, 1981) attributes many (most) illnesses to microorganisms, thereby implying that isolation of the organisms will lead to the development of preventive (vaccine or drug) therapies. 2 Health care costs to the public or individuals are, of course, "sales" to the industry from which come the income streams to the owners of the resources involved in the delivery of health care services. - 14 -CHAPTER ~: LIFESTYLES, RATIONALES FOR PUBLIC POLICY, AND ECONOMIC INTERVENTIONS "Our habits commonly begin as pleasures of' whioh we have no need and end as necessities in which we have no pleasure. Nevertheless we tend to resent the suggestion that anyone should try to change them, even on the disarming grounds that they do so for our own good." (McKeown, 1976, 108). 2.1 Lifestyle Characteristics - ! Working Taxonomy The term "lifestyle" undoubtedly conjures up a mix of dreamy and seamy situational visions -- cruising the Greek Isles on an 80 foot yacht, or waking to another morning of rain on Toronto's skid row. While such imagery vividly portrays different lifestyles, it is somewhat lacking in analytic content. Yet definitional boundaries of some sort are necessary to the development of a meaningful framework within which to view lifestyle-related public policy. The literature seems a mixed blessing here. A common approach is to set out a short roster of examples of what is meant by the author's use of the term "lifestyles", and to leave the reader to make the necessary extrapolation (Haggerty, 1977; Scheffler and Paringer, 1980; Pomerleau et al., 1975). At a less ad hoc level, we have suggestions that lifestyles are synonymous with cultural habits (Higginson, 1976), are "characteristic clusters of habit, customs and behaviour" (Mettlin, 1979, 559), are "related to the way people live in the social and historical context that is peculiar to them ••• [and to their] philosophy of life" (Johnson, 1976, 19), or refer "to those individual and/or societal behaviour patterns that are at least partly under individual - 15 -control, and that demonstrably influence personal health" (Somers, 1980, 1047). We have considerable trouble with the latter, as it appears to take as given the epidemiological evidence which we feel has seldom heretofore been formally delineated. "Demonstrable influence" requires evidence of causality and, as we will argue later, evidence of this sort is anything but abundant or easy to establish. The "demonstrated" lifestyle characteristics roster could be amusingly short, even though its "suspected" counterpart might remain lengthy. Here the discipline of neoclassical economics also offers no help. The main actor in that framework is the sovereign fully informed consumer making consumption decisions in the face of perfect information and a set of relative prices, and subject to a budget constraint; ergo, everything is lifestyle. We turn for guidance, then, to three oft-referenced works related to lifestyles (Fuchs, 1974; McKeown, 1976; and Lalonde, 1974). Fuchs uses the roster approach (tobacco, alcohol, stability of residence, p.53; diet, smoking, stability of family life, p .144), or the more vague "what we do and don't do for and to ourselves" (p. 151). McKeown comes closest to offering a formal lifestyle definition in his discussion of "behavioural influences" on health, but also essentially employs a roster ("smoking, exercise, diet, etc. 11 , p. 98). He does distinguish between these "micro" influences and more "macro" environmental influences such as collective shortage of food and environmental 1 pollutants. Similarly, Lalonde (1974) develops the health field concept comprising human biology, environment, lifestyle, and health care organization. More recently, Tesh (1981) identifies three generic disease causality models -- germ theory, lifestyle theory and environmental - 16 -theory. For our purposes, it is the environment and lifestyle components which are of interest. Lalonde defines "environment" to include "all those matters related to health which are external to the human body and over which the individual has little or no control" (p.32), and "lifestyle" as "the aggregation of decisions by individuals which affect their health and over which they more or less have control" (op.cit.). Personal control or voluntary choice, then, seem to be key ingredients of the Lalonde distinction between environment and lifestyle. 2 But even this does not seem to provide a watertight division, for decisions such as the choice of occupation and residence may relate to lifestyle, but at the same time embody some mix of environmental factors such as quality of water supply, and noise and air pollution. If one makes a conscious decision to work and live in Los Angeles, does the environment there become part of one's lifestyle to the extent that one exercises rational control in making the decision to move? The temporal frame may be reversed by considering the ex post building of a nuclear power plant on the periphery of a town. In this situation, the environmental risks of that plant may be argued to be part of the residents' lifestyles because (or if) they can choose to leave. The separators amidst these shades of grey, and the factors implicit in most roster separations of environmental and lifestyle factors, seem to be degree of control and frequency with which individual control must be (or is) exercised. A career decision locking in both residential and occupational environments may be a decision controlled largely or entirely by an individual, but such a decision is , seldom made. Once made, it may be relatively difficult (and/or costly) to reduce job stress or shift jobs or residence. In contrast, decisions - 17 -regarding cigarette smoking, the wearing of a seat belt, whether to go skiing or watch the football game on Sunday, generally appear to be both frequent and under the control of the individual. For our purposes in the next three Chapters, the following definition is adopted: an individual's lifestyle is represented by that aggregation of decisions over which s/he exercises significant control and which control is exercised relatively frequently. While terms such as "significant" and "relatively frequently" are conveniently lacking in clarity and precision, this provides an operational base sufficient to allow us to proceed with the analysis at hand, and seems consistent, for example, with Mettlin's (1979) "characteristic clusters of habit, customs and behaviour" approach, or Tesh's (1981) "personal behaviour factors". By this definition, then, the myriad routine decisions constituting living on a daily basis form the "lifestyle" composite. Those characteristics commonly cited in reference to their health status links (e.g. smoking, drinking, wearing seat belts, choice of diet, exercise and recreational routine) fit smoothly into this framework, while environmental pollutants are just as smoothly left outside, and factors such as job stress may be in or out depending on the degrees of freedom any given individual has in an employment setting. Before leaving this discussion, it is worth emphasizing that for the purposes of the analysis in this and the following three Chapters, we are assuming that lifestyle characteristics are synonymous with voluntary choices made by sovereign consumers. In Chapter 6 we step back from this voluntary choice framework to consider the implications of other views of lifestyle formation. If voluntary choice is not the sole or dominant factor in that process, the relative effectiveness of - 18 -economic "sticks" vis a vis other policies requires re-examination. 2.2 Lifestyles and Public Intervention Various levels of government are involved in almost all facets of individual lives in Canada today. Whether it be compulsory Canada Pension Plan contributions, Participaction, national health insurance, welfare payments, regulation of foreign investment, anti-combines investigations, or taxes on gasoline, food, cigarettes or wine, the omnipresent hand is taking, or giving, or regulating, or attempting in other ways to influence personal behaviour. Public intervention in individual lifestyles (be it through legislation, education, taxes, or moral suasion) is presumably motivated by a (rarely formal) process of comparing the aggregate social costs and benefits of a particular lifestyle characteristic. Of course one could argue rather more cynically that certain interventions (taxes on targeted "detrimental" substances) are not motivated by the results of a full-blown cost-benefit analysis, but rather by a more narrow analysis focusing on the benefits and costs to government alone. Or one could argue that it is impossible to determine the motivations of government at all (Wikler, 1978). What does seem necessary, at least intuitively, for public lifestyle interventions is a divergence between public and private evaluations of the costs and benefits of a product, service or activity. It is well known that such a divergence emerges in cases of "public consumption goods" (e.g., national defense). But it may also emerge in valuations of lifestyle choices for a number of reasons. First, private valuations may be undertaken with less knowledge about lifestyle - 19 -impacts and related costs than that held by the policy-maker. Either information that exists may not be widely disseminated, or private information filtering processes may be such that the dissemination processes in place are unsuccessful. 3 Second, external effects (benefits or, in this case, costs imposed on others by the lifestyle decision of the individual) may not be built into a private valuation, but are more likely to be incorporated at the social level. Third, if public goals do not fairly represent the collective private will, as in a situation where a narrow governmental rather than broader social perspective is taken in the "public" valuation of an intervention, a divergence is again likely to emerge. This last basis for differences is likely, however, to be the most transitory, as the collective will of the voting public may reflect and correct the inappropriateness of the narrow view at the next available voting opportunity. Thus, differences in social and private assessments of a lifestyle behaviour may emerge, and may lead to public intervention in some form. Any motivation to intervene presumably has underlying it some (more or less) well-defined goals of intervention. Following Wikler (1978) we suggest that there may be any combination of three such objectives: improvement of health, efficiency and equity. We deal briefly with all three here, and in more detail with the last two in the following section. The goal of health improvement is based on the paternalistic premise that the health of others is important to each of us, or at least to the collection of us represented by the makers of policy. If a divergence of information exists between the social and private perspectives on the health impact of a particular lifestyle, or if individuals choose to ignore the information, the paternalist argument - 20 -would be that government should intervene for the sake of the health of the individual (Weale, 1983). Both Weale (1983) and Wikler (1978) deal admirably with some of the ethical issues surrounding paternalist-moti vated intervention, and the latter appears to conclude (as an expression of his own social values) that such a motivation may be justifiable only in situations where a true information differential does exist (in which case the appropriate intervention would be simply to remove that differential!), or where the voluntary choice view of lifestyle behaviour is cast in doubt. Wikler also takes pains to point out that most arguments for paternally-motivated interventions are not supported by sufficient information on lifestyle/health status links, or on the nature of lifestyle formation. An efficiency objective for lifestyle-related interventions would be justified if too many real resources are being allocated to health care (and therefore not to other valued economic activities) because of individual lifestyle choices. In other words, if individuals bore privately the marginal lifestyle-related cost of their health care, they would live more healthy lives and consume less health care. 4 But this type of justification, too, requires information and evidence. In the following section, and then more comprehensively in Chapter 3, we lay out explicitly the types of information required. A third plausible justification revolves around the issue of equitable distribution of burdens -- "some people's distaste for having to accept both high costs and someone else's bad habits. In the view of these persons, those who indulge in self-destructive practices and present their medical bills to the public are free riders ••••• " (Wikler, 1978, 318). Again the strength of this justification rests, in any - 21 -particular situation, on the state of evidence -- in particular "that those who take chances with their health do place a significant financial burden on society" and that "the person who engages in an unhealthy life-style is responsible (our emphasis) for the costs of caring for the illness that it produces" (ibid., 320). Of course, given the necessary informational base, and presuming the information is such as to support public intervention, there are any number of ways one might actually intervene. They may be classified for convenience into economic approaches, legal approaches, and advertising or educational approaches. In the following section and for most of the remainder of this report we focus exclusively on economic incentives and disincentives, reconsidering the other possible approaches in the final Chapter. 2.3 The Rationale for Economic Incentives The arguments underlying the application of economic incentives to targeted lifestyles or their impacts, derive basically from efficiency or equity objectives. While either paternally- or efficiency-motivated goals of intervention may be pursued with a number of non-economic strategies (and, in fact, economic incentives are not particularly well suited for addressing a paternal motivation), equity objectives as outlined in the previous section can be achieved only through economic means. We suggest, then, that the case for charging patients for the care of illnesses which they have brought on themselves through voluntary lifestyle choices is, in fact, two -- an equity argument and an efficiency argument. The major distinction between the two is that an efficiency-motivated intervention rests on the assumption that - 22 -lifestyles will in fact respond to some form of economic carrot (or stick), while the equity justification does not. To see the distinction, imagine two different societies, A and B, which are identical exaept that in A, everyone engages in some health-threatening praatice, say smoking, to exactly the same extent. In B, only half the population smokes. In both societies, smoking increases the risk of various illnesses, with attendant health care costs, which are paid for from public revenue. In both societies there is a potential problem of efficiency of resource allocation. In society A, people collectively pay in taxes the full cost of their health care. But they do not, at the time of purchasing cigarettes, pay a price which reflects this health care cost. Even if an individual knows that, by smoking, s/he is adding to overall health costs, s/he cannot lower his/her taxes by quitting or cutting down. Thus cigarettes are underpriced because their prices do not incorporate the marginal impact of the product on health care costs. If cigarette use is price-sensitive, so that people in this society would smoke less if they were paying the true cost (including subsequent health care) at time of purchase, then A is using its resources inefficiently. Given the choice, A-aitizens would spend less on cigarettes and health care, and more on other things, and would feel better off. But no one of them individually can make himself/herself better off financially by cutting down on smoking, because the benefits of lowered health care use/costs would almost all accrue to other people. In effect, everyone but the individual reducing his/her smoking would be a "free-rider". Thus, A-citizens might rationally vote for some economic disincentive to smoking, even though individually they would not choose to cut down in - 23 -the absence of the incentive. In society B, however, there is also an equity problem. The non-smokers bear part of the smokers' health costs. Thus, one might argue for economic disincentives for smoking, even if they do not change smoking behaviour, because any revenue earned can be used to compensate non-smokers for the costs they bear in providing extra health care for smokers. There may also be an efficiency problem in B, in that smokers would choose to smoke less if they faced the full cost of their behaviour. But the equity case stands even if smokers' behaviour is insensitive to these measures. In A, on the other hand, since everyone smokes, there is no point in introducing economic measures if smoking behaviour is unresponsive to them -- they would have neither efficiency nor equity benefits. The interest in economic incentives directed at lifestyles as a way of moderating the growth of health care costs is, then, essentially an efficiency argument, although it also has a paternalistic version. The efficiency form focuses on "over-consumption" of unhealthy lifestyles and resulting health care, by consumers who would choose healthier lifestyles and less care if they were exposed to the full cost of those lifestyles. Once they are exposed to that full cost, however, whatever lifestyle choice they make is "right" and so by extension is the associated health status/health care cost. This "consumer sovereignty" view contrasts with the paternalistic one described in the previous section which would take improved health status and/or lowered costs as objectives in themselves. In this latter view, economic incentives should be merely part of a more general strategy to reduce illness and/or health costs as much as possible, not to find the balance which consumers appear to prefer between lifestyle and illness, and "other - 24 -things". Whether motivated by efficiency or paternal interests, however, incentives directed at moderating health care costs rest critically on an assumption that lifestyles respond to economic incentives. If they do not, then the equity argument may justify economic "incentives" on grounds of fairness between choosers of healthy and of unhealthy lifestyles, but no reductions in health costs will follow. Insofar as revenues raised from penalties on unhealthy lifestyles may be used to finance part of public sector health costs, there may be a reduction in the share borne by other taxes. 5 But this represents redistribution of health costs, not reduction. In the previous chapter we noted the rather perverse co-opting of the Lalonde (1974) report to the cause of direct consumer charges for lifestyle-related conditions or choices. This appears to be part of a broader advocacy of economic incentives motivated by lifestyle issues which is superficially justified on equity or efficiency grounds. On closer examination, this thrust appears to be an attempt to re-open or expand private funding sources for health care. Using the currently in vogue lifestyle-related social conscience in this way is viewed, probably quite correctly, as a means of relaxing the constraints on overall health care spending which arise from sole source public funding. This facilitates, of course, an expansion of the total costs of health care, to the benefit primarily of its suppliers. For example, the argument for a general reintroduction of direct charges to patients, at point of service, to be collected and retained by the providing physician or hospital, on the grounds that patients are "responsible" for their own conditions, falls into this category. The evidence seems now quite - 25 -conclusive that such direct charges at provider discretion raise the overall costs of health care (Barer, Evans and Stoddart, 1979); certainly their provider advocates seem to believe they do (Canadian Medical Association, 1981; Rigby, 1982). The advocacy of economic incentives for healthy lifestyles thus appears to further a number of different interests, depending on the form of incentives chosen and the assumptions made about individual and health care system responses. In focusing on the equity and efficiency arguments, we are assuming that expansion of health care costs, ceteris paribus, is not a general public objective, though it clearly is for some private groups. We are also assuming that control or reduction of such costs, while frequently treated as if it were an objective in itself, in fact reflects an underlying objective of more effective use of resources within health care, and a better balance of resources between health care and other things. After all, if cost minimization were the objective, one could simply close down the health care sector. Whether the effective use of resources is to be evaluated relative to some consumer sovereignty standard, or to a more paternalistic balance of health status against other objectives, is a question that does not necessarily have be to resolved, since tradeoffs must be made in either case. The institutional structure of health care in every society, however, appears to reflect political choices in favour of a more paternalistic stance and away from consumer sovereignty. 2.4 Economic Policy Options The discussion to this point has at various junctures made vague reference to premiums, direct charges and taxes. Before proceeding in - 26 -the following Chapter to formalize the nature of the information/evidence require~ to justify each type of economic (dis-) incentive, we set out in this section a taxonomy of possible interventions, and their expeoted targets (i.e., efficiency or equity). For convenience, and because it will serve to clarify the discussion, we propose separating the possible interventions into two generic classes -- (i) insurance interventions; (ii) direct interventions. A schedule of risk-rated premiums would be an example of the former; taxes on cigarettes, employer bonuses for smoking cessation, and direct charges for health care, examples of the latter. The generic distinction between the two types is that the first is not directly linked either to outcomes or behaviours, while the second consists of price or income adjustments linked directly to specific consumption or behaviour patterns. In the latter, an action must take place for the intervention to be triggered. In the former, the intervention is based on expected behaviours and outcomes. Where the incidence of a condition is uncertain at ·the individual level, individual premiums based on collective risk may be determined because the incidence is not uncertain at the collective level. The interventions may be viewed also in terms of their potential as efficiency instruments, as equity instruments, or as both. Their success in achieving either or both goals is ultimately an empirical question requiring evidence. Thus they are reconsidered in Chapter 3 after we set out the general framework for identifying informational requirements, and in Chapter 5 after we have reviewed the empirical evidence. - 27 -Risk-rated Premiums The idea of risk-rated premiums as a lifestyle policy instrument is baaed on the presumption that individuals can be allocated to relatively homogeneous groups according to their risk of incurring lifeatyle-related health care coats. Thus, if epidemiological information suggests a high relative risk of lung cancer and other diseases for smokers, one might design a health insurance premium structure which differentiates risk classes according to smoking behaviour. The number of different classes would clearly be dependent on the state of information regarding collective risk relationships and on the ability of the insuring agency to collect valid behavioural data on individuals. Taxes ~ Targeted Products £!: Activities Moat of these are as straightforward (in theory) as they sound, and could include cigarette and alcohol taxes (as examples of taxes on products), or break-a-leg taxes applied to lift-tickets at downhill ski resorts. In terms of a causal chain running from lifestyle behaviour to health care coats (developed in Chapter 3), these examples would be classified as "upstream" (dia-)incentivea. User Charges for Health Care Services These are "downstream" incentives, applied at the stage where actual health care coats are incurred as a result of prior lifestyle decisions. They would involve making the individual liable for some share (leas than 100%) of those health care coats. The actual forms they might take include those normally associated with direct charges uniform or variable co-insurance, deductibles, or combinations of the 6 two. - 28 -De-insurance This is also a downstream "disincentive" which amounts to making the individual liable for 100% of the health care costs incurred as a result of a specific lifestyle characteristic (or characteristics). It amounts to saying, if you engage in practice X, and you contract condition Y, your health insurance does not extend to treatment of that condition. Subsidies 2!!. Targeted Products 2!: Activities These upstream incentives might include government subsidies of memberships in health or fitness clubs, or of specific products which (assuming information availability) were linked to lower health care costs (e.g., bicycles, but maybe not jogging shoes). With these incentives and the objectives of applying them in mind, we turn now to a consideration of the assumptions underlying equity and efficiency arguments for the use of such instruments. This will allow us to identify the general informational requirements of economic incentives for healthier lifestyles. - 29 -FOOTNOTES TO CHAPTER ~ 1This "micro" versus "macro" terminology is employed by Higginson ( 1976). 2This dichotomy is also found, for example, in Boden's (1976) discussion of the extent of "environmental disease". 3After all, one of the major and ongoing private sector marketing dilemmas is determining strategies for getting messages past the screening blocks that determine which messages are noticed, of those noticed which are remembered, and of those remembered which induce an active response. 4Note that if in the presence of this individual risk-bearing, individuals did not change their lifestyle patterns, there would be no allocative inefficiency, or at least none attributable to lifestyles. 5we find particularly misleading the argument offered by Fraser (1981) that one needs to know the incidence of those reduced "other taxes" in order to determine the equity effect of economic incentives such as those directed at smokers in this example. It may be true that revenues raised from smokers may allow a reduction in other taxes borne largely by smokers (e.g., say the revenue allows a reduction in property taxes and in society Ball property is owned by smokers). But an equity-motivated policy decision of directing an economic "stick" at smokers so as to compensate non-smokers for higher publicly funded health care costs, is a decision independent of any subsequent or concurrent tax reduction decision. It should be viewed as such because a policy decision which happens to give back the extra revenue to smokers may be motivated by quite different public objectives entirely unrelated to the smoking or broader lifestyles issue. To tie the determination of effect of an equity-motivated lifestyle incentive to other completely unrelated policy initiatives, seems to us to muddy the waters unnecessarily. 6More detail on the variations associated with direct charges may be found in Barer, Evans and Stoddart (1979). - 30 -CHAPTER 1: JUSTIFICATION AND CHOICE - A DECISION-~~~~~~ ~- -THEORETIC FRAMEWORK !QR ECONOMIC INCENTIVES "Somebody once went to a good deal of trouble to find out if cigarette smokers make lower college grades than non-smokers. It turned out that they did. This pleased a good many people and they have been making much of it ever since. The road to good grades, it would appear, lies in giving up smoking; and, to carry the conclusion one reasonable step further, smoking makes dull minds." (Huff, 1974, 85). The discussion in Chapter 2 established two apparent rationales for government economic interventions directed at individual lifestyles. Both equity and efficiency arguments are frequently entwined in the views expressed by advocates of economic incentives and, indeed, both may apply in any given situation. But as we emphasized in that discussion, it is important to keep the two arguments separated conceptually -- the conditions for their validity are different. In particular, success with an equity objective depends not at all on whether people actually respond to an incentive; for the efficiency argument a behavioural response is critical. In both cases, however, attaining the objective requires several steps which are often telescoped in general discussion. It is important to spell out these steps, as their empirical validity is not, in fact, self-evident. Yet the arguments for economic incentives on either ground rely on the empirical validity of each step. - 31 -3.1 Assumptions Underlying Efficiency Goal The efficiency argument assumes first that the economic incentives proposed will in fact lead to a change in lifestyle. Later we shall discuss what kinds of economic incentives might be more or less likely to induce such changes. Here we note only that the choice of instrument, or form of incentive, does matter, and the generalizability of the economic incentive approach will depend on one's ability to design appropriate and effective incentives. But the assumption that any economic incentive will be effective, either absolutely or relative to alternative measures (e.g., restricting or forbidding the behaviour in question) depends on a very specific formulation of the utilization process. The implicit model of utilization is drawn from consumer demand theory, where individuals with well-defined tastes choose levels of consumption of goods and services in response to relative prices. Such consumption is isolated from any social context, and tastes are generated entirely within the consumer, independently of external forces. When the relative price of a commodi~y or activity is increased, by whatever means, the consumer reduces his/her consumption of it. As noted earlier, we will be working within the framework of the isolated consumer as treated in economic theory throughout most of this report, because that is the basis for the efficiency arguments for economic incentives. But in Chapter 6 we outline several competing, more complex and perhaps more compelling views of the lifestyle determination process which lead to different views of the efficiency question. For the moment, we merely emphasize that the efficiency argument rests on an assumption that the individual consciously and rationally "chooses" lifestyles by adding up the costs - 32 -and benefits of each. Thus by manipulating these costs and benefits, one can change the lifestyle choices made. Assuming that lifestyle choices do in fact respond to economic incentives (within a politically feasible range), the second step in the efficiency argument requires that health status respond to lifestyle change. There is extensive documentation of a correlation between various lifestyle characteristics and patterns of mortality and morbidity, but the demonstration of a causal relation is much more difficult. Certain causal relations seem more or less clearcut --smoking is generally accepted as causally linked to several illnesses, use of seat belts reduces the damage done by road accidents, too much alcohol causes cirrhosis of the liver. Even for some of these linkages, as we will illustrate in the following chapter, current evidence is such as to preclude certain economic incentives. Beyond these few links where causality has been demonstrated (more or less), the causal evidence thins out rather rapidly. More general relations between, say, diet, exercise, and heart disease, of which much was made in the federal White Paper, are still under intensive debate as to their causal significance. Certain lifestyle characteristics are risk factors for heart disease, just as a high reading on a thermometer usually corresponds to a hot room. But it does not follow ~ priori that putting an ice cube on the thermometer will cool the room. Thus the assumption that, if people respond to incentives (economic or otherwise) to improve lifestyles, they will in fact be healthier, is a substantial one which, outside a fairly circumscribed set of lifestyle characteristics, is not well grounded in empirical evidence. Of course behaviour responds to perceptions of evidence, not evidence itself, and one may argue that if perceptions can be altered, - 33 -lifestyles can be changed. This is presumably what preventive education and advertising are all about. But it is difficult (not impossible) to argue that utilization and costs of health care will also respond to perceptions. If people come to believe that jogging, or low-fat diets, reduce heart disease, then their behaviour may change. Moreover their health may improve, if we define health in terms of body weight or ability to run long distances. But if we have no firm causal evidence for a link to heart disease, we would be ill-advised to forecast a drop in the heart attack rate and to close down CCU beds. 1 To date, little such firm evidence exists. But suppose the link is there. Yet a third link in the chain of argument is the assumption that healthy people will use less health care. But this too does not follow ~priori. Quite apart from the growth in sports medicine, lifestyle counselling, and various forms of care for the "worried well" which is stimulated by increased health consciousness, there is the problem of competing risks. The Iron Law of Epidemiology is, one out of one dies; and a reduction in fatal heart attacks may keep people alive to die later of more expensive illnesses. This is not to be construed as an argument for euthanasia on cost-benefit grounds; as far as we know living is preferable to dying, and dying late to dying early. But we cannot assume that a healthier population which lives longer will necessarily need or use less health care per capita than one which kills itself off in cars or bars. It depends on the lifetime illness and associated care profile of each person. Particularly as we observe the rapid growth of the "old-old", 85+ portion of the population, with very high per capita needs for care, we might pause before assuming that healthier lifestyles - 34 -will always lead to lower health care use. Finally, even if it turns out that healthier lifestyles do in fact lead to lowered per capita need for health services, it does not follow that overall health costs will fall. If the health care service system were similar to an unregulated private market, with fully informed consumers buying well-defined commodities at competitive prices, then we ·Could be entitled to expect that a drop in "demand" would lead to lower output and probably lower (at least no higher) prices. Total sales (costs) would fall. But it is well-known that health care services depart from this model in a number of crucial respects. Most students, practitioners, and administrators of health care tend to view total costs as dependent primarily on supply-side forces, the numbers and income aspirations of people working in health care. When seat-belt legislation lowers the need (if it does) for orthopaedic care, orthopods do not cut their fees to generate new business, or leave the industry to work as taxi-drivers. They find new needs to meet. When falling birth rates lower demand for obstetricians' services, they put pressure (upward) on fees to recoup their lost income, and increase the intensity of servicing for each delivery. In this environment, a reduction in external "demand" for health care can only translate into lower expenditures if the incomes of medical care suppliers fall or if simultaneous reductions are made in hospital or medical care capacity. And this is by no means easy to do. On the other hand, efforts to reduce "needs" for health care by improving population health status may form a useful adjunct to supply-side policies of reduced capacity which have already been adopted on other grounds. But advocacy of economic incentives for healthier lifestyles, purportedly on efficiency grounds and taking the form of various types - 35 -of health care service charges, comes often from those providers and provider groups whose incomes (in the absence of capacity contraction) must fall, by simple arithmetic, if the incentives are effective in reducing health care expenditures. This should make one suspicious of too simple causal assumptions. The chain of logic from economic incentives to patients to reduced health care costs is thus long, and suspect at several points. 3.2 Assumptions Underlying Equity Goal The impact of economic incentives on an equity objective is somewhat more secure, since the behavioural and causal chain has fewer links. In particular, behaviour need not actually respond to economic measures. All that is required is that revenue from such measures be used to compensate people with more healthy lifestyles for the costs imposed on them by those with less. But it still requires that unhealthy lifestyles indeed "cause" health care cost differentials. Perhaps instead such lifestyles "risk a person's health, but risk it in such a way that he would die rapidly and cheaply at about retirement age" (Veatch, 1980, 55); or perhaps the population with "unhealthy" lifestyles is not in fact sicker because of those lifestyles. If the population with supposedly unhealthy lifestyles is not in fact sicker than the rest of the population, or if its illnesses do not in fact lead (over a lifetime) to more health care use, then the equity argument collapses. The neighbour who smokes, drinks, gets no exercise, and drops dead on the street at 62 is hardly a burden to anyone, whereas we ourselves expect through clean living to die at 90 after (probably) vast expense to the health care system. There is no obvious reason why that - 36 -2 neighbour should compensate us -- perhaps the reverse. While this discussion has focused exclusively on "unhealthy" lifestyles, "healthy" lifestyles (if in fact we accept that there are such) raise equally perplexing issues. Suppose, for example, it is true that those who partake in activities which improve cardiovascular fitness -- jogging, racket sports, hiking, cycling do indeed have lower health care costs. Is there an equity argument for, e.g., making contributions to fitness clubs tax deductible or subsidizing running shoes? An efficiency argument might hold, on the (rather dubious but testable) assumption that such subsidies would increase both participation and fitness. But is it in fact equitable to reward certain people for doing things that they like to do, independent of any assumed incentive effect? And if it is not, then the argument for lifestyle subsidies does depend critically on a detectable response of fitness to such incentives. Finally the linkage between utilization and costs in the equity argument is a bit fuzzy. Suppose it is true that people with unhealthy lifestyles use, ~ ~ result (causation, not just correlation) more lifetime health care per capita. If aggregate utilization of health care is in fact determined (at least to a considerable extent) by the structure of the supply side, as there is some reason to believe, then is there still an equity argument for requiring the choosers of unhealthy lifestyles to compensate the clean livers? The issue seems at least debatable. 3.3 A Generalized Linkage Framework The previous two sections were taken up with a general discussion - 37 -of a rather extensive set of behavioural and empirical linkages underlying the arguments for economic incentives. We go on now to formalize that discussion by considering various types of economic incentives within a decision theoretic representation of the logical and causal links between incentives and their purported objectives. The framework is built up by looking at increasingly complex causal configurations, and considering the potential for economic incentives at each level of complexity. The simplest case is that of a well-defined behaviour or activity A which always gives rise to a physiological state s, which in turn leads sooner or later to a morbidity state M and associated costs of care c, where each step in the process is necessary and sufficient for the one following it; A always leads to s, and S does not occur without A, and so on. This case is illustrated in Figure 1. The decision A leads to physiological state S with a probability of 1.00, and therefore not ending up in state S (S) has an associated probability of zero. The "necessary" condition is shown in the lower branch of the decision tree by the fact that a decision not to engage in the activity implies that the individual cannot end up with physiological condition s. Now assuming that consumer/patients are fully informed and rational, it does not matter where or how we introduce economic incentives. We could tax the activity A (upstream), or the states Sor M (mid-stream), or charge for the care C (downstream). So long as the various taxes or charges were set to yield the same revenue (suitably discounted), presumably an amount equal to c, their impact on the consumer will be the same. From the consumer's point of view, in deciding whether and how much - 38 -M s M s M s - 39 -to engage in A (how many cigarettes to smoke, kilometres to jog, grams of fat to eat) the present value of the stream of taxes or charges will be the same, and will amount to an increase in the price of A corresponding to health care cost c. We may tax cigarettes, or monitor and tax lung capacity, or charge for the health care services associated with lung cancer -- in the world of full information and certainty the effect is the same. Of course unless we are willing to extend the perfect information assumption to the policy-maker, operational complexities would affect the choice of incentive even if Figure 1 were a true representation of the role of lifestyle in the etiology of M. For example, monitoring costs (the costs of information to the policy maker) vary -- it may be easier to tax cigarettes or charge for services than to identify the intervening steps. But so long as we know that these steps follow each other with certainty, and do not follow from anything else, then we can impose the economic incentive to reduce (or increase) activity A at whatever point in the chain the costs of administration are lowest. Enforcement may constitute another operational problem which encourages upstream incentives. If C is relatively large but results from a long prior period of activity A, then the consumer/patient may not allocate, each time s/he engages in A, a sum of money sufficient to cumulate in due course to c. We as a society may then be confronted with numbers of people in state M who have not the resources for c. In a static world, where all live forever, C can always be recouped later. In reality, we may be confronted with either the ethically unacceptable denial of care, or the relief of the indigent, which of course undercuts the whole purpose of the economic incentive. One would only modify - 40 -one's level of A if one expected to have enough resources to pay for C -- i.e., to have something to lose. In this case it would clearly be more effective to tax A or the intermediate state S (assuming no monitoring problems at either stage), rather than charge C for the necessary health care. But of course this ideal picture is unrealistic. If we relax the assumptions of full information and perfect rationality for consumers, then it may well matter even in this very simplified case where we put the incentives. In general, individuals have imperfect foresight. If they overestimate the size of C as an outcome of A, charging C may have a greater impact. If they underestimate, as seems in practice more likely, then charging at A -- taxing the cigarettes rather than charging for the thoracic surgery, is more likely to be effective in making the desired equity adjustments or efficiency improvements. But this discussion is rather forced in a world of perfect certainty. Let us first retain the sufficiency side of the causal chain, but relax the necessary condition (Figure 2). This alters the lower half of the decision tree in that now even if one does not engage in activity A, one may still end up in state S and ultimately incur costs c. The upper half of the tree remains identical to that in Figure 1 -- A always leads to s. Now the expected costs attributable to A are still c, but the expected costs arising from treatment of M among those not engaging in A is not zero. It is x2c, where x2 is the probability of ending up in state S even if one does not engage in A. Say, for example, that the incidence of lung cancer among non-smokers is 1:10,000. Then the lung cancer treatment costs a non-smoker may expect to incur will be (0.0001)C, where C is assumed to be the average cost of treatment. - 41 -Figure 2 s s s M M M M c c EXPECTED COST c - 42 -In this situation the case for downstream incentives (charging for care) is dependent on our ~ post ability to monitor lifestyle behaviour. It may be relatively straightforward to pick out the smokers from among lung cancer victims, in which case (subject to the operational qualifications noted above) we will still be indifferent as to where to apply the economic incentives. We need only add the stipulation to a downstream incentive, for example, that non-smokers who incur costs C for treatment of lung cancer will not be privately liable for those costs. But assume hypothetically that inadequate fibre content in a diet has been causally linked to colon cancer (morbidity condition M) in that all who include no fibre in their diets end up with the cancer. Assume, further, that some among the "fibre-eaters" will also contract colon cancer, but that the only way of establishing the nature of an individual's diet at the time the disease is discovered is to ask that individual. The incentives for false reporting are obvious. Accordingly "downstream" incentives - charges for care for colon cancer -would have to be applied to all with the disease, despite the obvious unfairness to those "fibre-eaters" unfortunate enough to contract the illness. The larger x2, the more severe this unfairness; but even very small x2 make self-(mis)reporting possible and lead to penalizing the "innocent". In Figure 3 we continue to assume perfect certainty running from S to M and then to c, but now relax both the necessary and sufficient conditions linking A to s. Whereas in Figure 2 A always led to s, but S could occur without an individual necessarily having first engaged in A, in Figure 3 A leads to Sonly 100x1 per cent of the time (O~x 1~1.00). Of course Figure 3 is simply the more general representation, with Figure 2 setting x1:1.oo, and Figure 1 setting x1:1.00 and x2:1.00. - 43 -Figure 3 s s s M M M M M M c c EXPECTED COST - 44 -Now some who engage in A will get M and others will not, and some who stay away from A will still get M. If at stage M one can ~ post identify the A's, downstream incentives directed at A-lifestyle individuals are still appropriate. The difference now is that those engaging in A have some chance (1-x1) of avoiding incurring costs c. If such individuals can judge x1 correctly, and if there are no transactions costs and financial markets are perfect, again it will not matter where in the sequence we impose the incentives. Even if C is very large and quite improbable, if we charge C to the individual for the treatment of M, the rational consumer can buy a private insurance policy which will cover C, at a premium Cx 1c +carrying charges). Those who do not engage in A do not buy this insurance because even if they contract disease M, there is no lifestyle justification for charging them. But this depends on the possibility of monitoring behaviour A after the fact. If we cannot distinguish between A-induced M-states and other causes of M, the information base supporting downstream incentives is not there. There is a further difficulty. We have assumed to this point that all who engage in A do so to an equivalent extent. This is, of course, an unrealistic simplification. If, instead, x1 represents the average p(S/A), but there is an intensity or dose effect, private insurers may want to offer a risk-rated premium structure to the A-group. But they will often have considerable difficulty in monitoring A. If premiums do not adequately reflect A levels, any incentive effect is weakened. Historically, health insurance premiums (public or private) have not been sensitive to lifestyle choices, except insofar as group experience rating might pick these up collectively. Thus imperfect information of - 45 -insurers leads in the uncertain case to premium structures which dilute or completely negate the incentive effects of charging for c. The risk of opportunistic behaviour is likewise increased by large but uncertain C -- the rational consumer may well maintain levels of A in response to charges for C on the assumption that M will probably not occur, and if it does, s/he will receive care anyway whether or not s/he can afford it. Again, the lower one's wealth level, the less one will respond to the incentive. Finally, we need to consider briefly the relationship between evidence or information and upstream incentives. The information of probability x1 is irrelevant to the consideration of downstream incentives. What is important is the feasibility of.!!!. post monitoring. On the other hand, x1 is extremely important to upstream incentives. Is the availability of empirical evidence on the magnitude of x1 sufficient justification for upstream incentives? So long as the average value of C is known, each person engaging in A (intensity/dose response issues aside) will be expected to incur costs x1c, and a tax in that amount is defensible on both equity and efficiency grounds. Figure 4 generalizes this situation one further step toward the more realistic predisposing role of lifestyle characteristics in particular illness conditions. Now there are odds (y 1) associated with morbidity M, given s, and lifestyles (A) shift the odds of reaching S and thus of contracting M. The probability of a non-A (A) individual getting illness Mis x2y1 + (1-x2)y2• Engaging in activity A shifts those odds to x1y1 + (1-x1)y2 where, if A is an etiological factor, x1>x2• Then if y1>y2 , engaging in A increases the odds of M relative to A. As with the previous figure, activity A is neither necessary nor sufficient for the emergence of the morbid event and its subsequent costs. But - 46 -Figure 4 M s M M M s M M s . M c c c c EXPECTED COST x 2 y1C (1-x 2 )y2 C - 47 -Figure 4 allows, for example, the possibility that a smoker will contract lung cancer for reasons quite unrelated to the smoking. This might be portrayed as a move along the path ASMC, a path precluded by Figure 3. Now a policy of downstream charging for care costs C will penalize not only those in the A group who have the misfortune of contracting M, but also those in the A group who contract M for reasons quite unrelated to A. This has little to recommend it in the absence of extremely strong causal evidence running from A to c. It further requires either no links from A to C, or the ability to differentiate A from A individuals at stage M. As for upstream incentives, the expected cost of treatment for illness M among A-group individuals is [x1y1+(1-x1)y2]c, but the expected cost attributable to activity A is x1y1c. In the unlikely event that (1-x1)y2 is large (or at least not very small) relative to x1y1, the informational basis for justifying even upstream incentives becomes rather weak. In short, a policy of taxing A will impose costs on many who will not in fact develop M. It may also impose costs on some who engage in A, develop M, but take a route to M independent of A. Whether such taxing is inequitable or inefficient depends again on the state of available information. Clearly we will not have ~ ante information classifying A-group individuals with illness M into causal sub-groups. But assume y2:o.oo. Then if there were any way of knowing which people engaging in A would develop M and which would not, a general tax on A becomes less defensible. If all in the A-group are (as far as we know) equally at risk of M then each one can be thought of as generating a social cost of x1y1c as a result of lifestyle choice, and a general tax on A is appropriate. If different A-group individuals have different probabilities of developing M and we are able to identify - 48 -those probabilities, then a risk-adjusted tax schedule becomes more attractive than a general tax on both efficiency and equity grounds. The dose-response evidence on smoking (see Chapter 4) comes to mind here. A flat rate tax on cigarettes serves the purpose of dose-related risk-adjustment, at least for equity purposes, if the probability of state M is a linear function of cigarettes consumed. If it is not, the flat rate may over-tax any consumption, but progressively under-tax increasing consumption. Throughout the four Figures we have maintained an implicit p(C/M) of 1.00. In fact, for any illness state M, there will be a frequency distribution of costs (Figure 5) which could be tacked on to the right hand sides of each of the four previous Figures. Thus, C in Figures 1 through 4 represents the expected treatment cost of condition M, and we may apply the less-cluttered frameworks without significant loss of generality. However, even a concatenation of Figures 4 and 5 may be lacking an important piece of policy information. The -->M--->C node may be different for different complete branches of the trees. Smokers and non-smokers may get lung-cancer, but the severity (and thus costs) may be greater, on average, for the smokers. Using the A-group, rather than overall, distribution of C could then result in a different magnitude of front-end incentives, or affect the size of the private premiums in the case of an individual insuring in the face of a downstream dis-incentive. We resist the temptation of presenting the full-blown decision tree, as the marginal information generated at this stage would in all likelihood be outweighed by the psychic cost to the reader. To summarize, we have attempted to illustrate that the appropriate point (if any) of imposition of economic incentives depends on the - 49 -,.. N () () ,.. I c () c () - 50 -relative strengths of the necessity and sufficiency conditions. If a chosen lifestyle activity A always leads to M, but many other uncontrollable (under present knowledge) events do as well, then we have a good case for taxing A (if we can identify it) but a very bad case for charging for c. Conversely, if M always results from activity A, even though many of those who engage in A do not develop M, we have a good case for charging for C while a much weaker case for taxing A. The time sequence and the possibility of opportunistic behaviour, however, modify this conclusion. As noted, we may tax A anyway if we think we will not be able to collect for C -- then we are imposing the purchase of a sort of compulsory insurance policy along with activity A. Of course, the more complex linkages of Figures 3 and 4 create problems of knowledge and perception as well. If the causal relationship between A and C is as simple as that portrayed in Figure 1, it is very likely that research will be able to establish the existence and strength of that relationship with some precision, and that choosers of lifestyles will become aware of it. But a relationship of Figure 3 or 4 form is much more difficult to identify, and hence the difficulty in achieving consensus among epidemiologists about lifestyle-health linkages. Under such circumstances, it is even more difficult for consumers to make any realistic assessment of the conditional probability of C given A. Moreover, the nature of the statistical problems in measuring the conditional probability P(M/A), combined with a certain natural and quite healthy optimism in the average consumer, is likely to lead to underestimates such that the impact of any economic incentive at the C or M end of the process will be greatly attenuated. If "it can't happen to me", then why should I worry about costs? Again, the case is - 51 -strengthened for placing economic incentives at the "upstream" or front end of the process. If consumers are imperfectly informed about the dependence of M on A, upstream incentives work better; if they are fully informed, then it does not matter. But to the extent that A is neither necessary nor sufficient for M, then the case for any form of economic incentive disappears. In particular, there is no warrant whatever for an argument that since X% of health care costs are generated by illnesses which are to some degree correlated with lifestyles, therefore X% of costs should be imposed on all patients as direct charges. When we refine the argument down to focus only on those illnesses which do in fact seem to be causally related to lifestyle choices, the strength of the causation and the extent to which other factors intervene still remain important qualifying considerations. Few indeed are the illnesses which are both necessarily and sufficiently caused by particular activities or lifestyles. The efficiency and equity, as well as the feasibility, of policies to link lifestyle choices with health care costs do not seem to be issues which can be analysed independently of the characteristics of specific disease processes. For that reason, we have chosen in this study to selectively examine two "lifestyle" activities which have been identified as having a clearly established (or at least generally accepted) causal relationship with illness status and related costs. Before turning to the epidemiological literature in the following Chapter, however, we return briefly to some of the specific economic incentives described in Chapter 2, to re-examine them in light of the discussion in this Chapter. - 52 -3.4 Economic Incentives Revisited In the previous Chapter we argued that economic incentives could be classified as either insurance-based or direct. Direct incentives were further classified in this Chapter according to their staging -- the place in the causal chain at which they are brought to bear. Thus, risk-rated premiums are insurance-based, taxes on activities are upstream direct (dis-)incentives, physiological state taxes or illness penalties would be mid-stream interventions, and direct patient charges (including total de-insurance) are downstream incentives. As a result of the extensive development in this Chapter of the general informational requirements for generic classes of incentives, we are now in a better position to consider the potential impact of the various types of incentives on equity and efficiency objectives. Risk-rated Premiums Modifying uniform premiums to reflect risk classes is clearly an initiative designed to achieve an equity (wealth transfer) objective. Any potential effect of such a policy on aggregate health care costs (the efficiency objective) is dependent on at least three phenomena or types of information. First, the A----> C linkages developed in the Figures earlier in this Chapter must be "strong". That· is, the targeted lifestyle must in fact affect health status and health care costs. Second, the lifestyle behaviour in question must be monitorable by the insurer so that risk classes may be established. Third, lifestyles must respond to the resulting risk-adjusted premium structure. Even if A is a necessary and sufficient condition for M and (by extension) c, and easily monitorable, the high-risk A-group individuals may derive utility - 53 -from A in excess of their marginal premium cost. Because they would then be unlikely to alter their behaviour, no efficiency gains would be obtained. Achieving equity depends, as well, on the first two conditions --epidemiological links and monitorability -- but not on the third. Given sufficiently strong A----> C causal links and an A behaviour that is monitorable and a basis for risk class identification, the premium option appears to serve an equity objective rather well. While there will be intra-class wealth transfers, ~post, from those in each class who do not incur costs C to those who do, that does not strike us as a wealth transfer about which public policy should be particularly concerned. Consistency with general insurance principles requires only that the ~ ante intra-risk-class expected wealth gain or loss is zero. After all, the same types of ~ post of transfers occur continuously under our present universal insurance plan and are, in fact, one of the reasons for the existence of the plan. Downstream Incentives Policies involving some degree of individual financial risk-bearing, whether they be co-payment, deductibles or complete de-insurance of C where C is a result of A, raise efficiency effect problems noted earlier. Because of individuals' likely tendency to underestimate x1y1c (where x1y1 is their own probability of incurring costs Casa result of action A) through underestimating x1, y1, C or some combination, any incentive effect may be attenuated. While we might guess that the efficiency incentive effect of such a policy would be somewhat stronger than that entailed in risk-rated premiums, such a hypothesis is clearly testable and in need of empirical evidence. In - 54 -the presence of private insurance against a public policy of de-insurance, for example, the efficiency impact would be identical to that for risk-rated premiums. But there are even more serious potential information-related problems associated with attempting to reduce health care costs with downstream incentives. As pointed out repeatedly in the previous section of this chapter, unless the epidemiological links are extremely strong in the sense that A is a necessary condition for M and c, the case for downstream incentives depends on the separability at stage M of A-induced and other-induced M-states. As for the equity objective, similar conditions hold. The effect of these policies is to transfer wealth from those who do incur treatment costs for a given condition, either to taxpayers as a whole or to providers of care. If the group from whom wealth is being transferred includes any with condition M for reasons unrelated to A, this becomes a less-justifiable wealth transfer policy. There are further informational requirements associated with this set of incentives. One's success in achieving the desired wealth transfers, or in reducing overall health care costs will clearly depend on the size of the incentives. The behaviour-modifying impact of a policy of de-insurance, for example, may be strengthened if information on the frequency distribution of C is made available. Similarly, a wealth transfer objective requires information on the marginal costs attributable to action A. Upstream Incentives This group of incentives involves taxing or rewarding specified lifestyle choices. Such policies' impact on efficiency will clearly be - 55 -dependent on the price elasticity of the lifestyle choice, and the strength of the lifestyle -----> cost linkages. The latter topic is taken up in Chapter 4 while we deal with some evidence relating to the former in Chapter 5. The equity effect of upstream incentives is a transfer of wealth from those engaged in A to those not so engaged. Again, however, we have an additional set of informational requirements associated with this set of policies. How, for example, does one determine the appropriate tax rate? Information must be available on the costs incurred in treating disease state M, the share of those costs attributable to A, and the "units" of A "consumed" by the population. Then the tax rate would be equal to [(proportion of C attributable to A) x C/units of A "consumed"], where C is the total cost of treatment for M. The denominator will clearly be a function of the size of the tax. # This suggests that the "optimal" tax rate would be determined through solving a system of simultaneous equations (an issue that is, thankfully, beyond the scope of this paper). 3.5 Some Final Thoughts 2!!. Information Requirements While the information and evidence necessary to justify a range of different economic policies is beginning to seem almost forbidding in most cases, the prerequisites of a useful policy may not be satisfied even if all informational components are, miraculously, available. As Mettlin (1979) and others point out, if one moves any distance away from a voluntary choice theory of lifestyles, a successful intervention on efficiency grounds would need to affect not only the individual but also any other factors implicated in the "lifestyle" decisions. This - 56 -point is raised only as a caution here, as the topic receives detailed attention in Chapter 6. This Chapter has made apparent, among other things, the role of epidemiological links in the underpinnings of any economic lifestyle interventions. Accordingly, in the next chapter we turn to a consideration of the process of identifying epidemiologic causality. This is followed by selective examination of the two "lifestyle" activities -- smoking and alcohol consumption -- commonly acknowledged to provide some of the strongest currently available causal evidence of this nature. These activities by no means exhaust the range of possible lifestyles in which one might be interested, but we intend the analysis to be illustrative of the process one would have to go through in determining whether a particular lifestyle-illness link was either secure enough (Chapter 4), or significant enough (Chapter 5), to justify any sort of economic incentive policy. The information requirements, relative to our present knowledge, turn out to be substantial even in these areas where our information is greatest -- the implications for extension to the much less well understood areas of diet, stress, exercise, or general social relations are not encouraging. - 57 -FOOTNOTES !Q. CHAPTER ]. 1rn early 1984 results of a randomized controlled trial have finally shown a causal link between cholesterol intake and mortality from heart disease. For over twenty years previously, medical recommendations and popular wisdom were based on, at best, suggestive correlations. 2 A study done by the Department of Health .and Social Security (Great Britain, n.d.) estimated that a sustained 40% reduction in smoking, starting in 1971, would by the year 2001 raise projected health care costs by£5 million, due to increases in the elderly population. Social Security payments were also estimated to rise by i24 million. - 58 -CHAPTER 4: THE EPIDEMIOLOGIC CONNECTION " ••• the only indisputable way to settle this matter [of causality] would be to put the effect of smoking to experimental test in human beings. Since this is impossible we cannot disprove [the alternative hypothesis that] constitutional differences between smokers and non-smokers ••••• explain [the] apparent effects of cigarette smoking". (Hausner & Bahn, 1974, 107-8). The analysis of the previous Chapter allowed us to establish the general extent of the epidemiologic evidence necessary to justify economic interventions. For example, downstream (dis-)incentives such as direct charges or de-insurance require satisfaction of the necessity condition (Mis always a result of A), or the ex post information to differentiate A-induced from other-induced M-states. Upstream (dis-) incentives are justifiable given "sufficiently strong" epidemiological causality. In other words, if there is strong evidence of a differential probability of M for A group as opposed to A group individuals (even if the actual risk is small) a case can be made for upstream incentives (of some magnitude). We have seen, then, that the state of information regarding the etiological role of A in causing M should be a key determinant in any policy decisions. We turn in the following sections to interpreting and reviewing some of that epidemiologic evidence. 4.1 The Search for Causality - Methodologic Issues While Figure 1 of the previous Chapter was both conceptually straightforward and easy to analyze in terms of economic incentives, it was also clearly an oversimplified representation of any lifestyle-- 59 -illness linkage. It portrayed 'absolute• causality -- A was a necessary and sufficient condition for M. It is safe, we think, to suggest that such relationships would make themselves known without the benefit of natural experiments or other sophisticated •staged' research designs. Equally, if M were an unpleasant enough prospect, A would disappear shortly after the population became aware of the absolute nature of the causality. In fact, eradication or marked reduction of A would likely result even from 'weaker' evidence. If, for example, smoking were a sufficient (but not necessary) condition for lung cancer - all smokers get it - and given that lung cancer is both painful and fatal, we suspect there would be a far smaller number of smokers in the world. Of course individuals would still be faced with a temporal tradeoff between the eventual lung cancer and the satisfaction generated by the smoking for some unknown period of time -- they would know with certainty that lung cancer was in their futures, but they would still be faced with having to decide how large a discount factor to apply to the disutility of that future state. Given that lifestyle choices in general are probabilistic risk decisions, how does one go about determining whether in fact some causal links are sufficiently strong to satisfy the epidemiological information requirements for specific economic interventions? Since the information is not waiting for the taking, epidemiological methods have been developed for identifying causality, and causal criteria have been established with which to evaluate the results of applying those methods. There are, then, two dimensions associated with causal determination -- method of investigation, and criteria for causal strength. Within each, there are more and less important elements in - 60 -determining the etiologic role of any factor A in any morbid state M. These two dimensions are portrayed in Figure 6. Each row of the matrix denotes an alternative method of investigation, and the rows are sorted (from top to bottom) by commonly agreed relative strength of the method for identifying causal relationships (see, e.g., McMaster University, 1981). Similarly, the columns contain alternative criteria for establishing causal strength. In this case there is somewhat less 1 unanimity about the relative importance of the different criteria. We have arbitrarily chosen the ranking from the recent Canadian Medical Association Journal's Clinical Epidemiology Rounds (ibid.) for the purpose of discussion in this section. We discuss in succession, then, data requirements, research designs, and causal criteria, and conclude with some general thoughts on using this framework to evaluate currently available epidemiological evidence. Data Requirements We are interested in the strength and nature of causality running from a lifestyle decision A (the independent variable), to a morbidity state M (the dependent variable), and ultimately to the health care costs, if any, resulting from treatment of M. our basic unit of observation is the individual. For each individual the data elements are: whether or not s/he is an A-group or A-group individual; if an A-group, intensity or dose information; whether or not the individual ends up in state M, and when in relation to activity A. Of course causality cannot be determined by observing any individual. Thus we require these data for groups of individuals. Data Figure 6 Method Evidence from true Strength experiments in humans Randomized Contro 11 ed Trial Cohort Study Case-Control Study Descriptive Study Internal Temporality Dose-Consistency Response Biological Specificity Plausib11 ity External Consistency C\ ~ - 62 -can be collected in two ways, distinguished by the degree of linkage between the data generation and the research question. "In the first way, (each) person under study was originally observed by people who ~not performing!. specific investigation (our emphasis), and who reported the observations in routine records. Afterward, to get the research data, the investigator extracts the information available in those records. In the second way, the investigator makes special plans beforehand for the techniques with which each person is to be examined and the data recorded. The first way of collecting research data can be called retrolective, and the second way, prolective" (Feinstein, 1973). 2 Thus, data recorded "with investigative intent" may be distinguished from those which happen to be available and to (at least partially) satisfy investigative data requirements. As we will see shortly, data collection for experimental research designs is definitionally prolective; for cohort and case control studies it may be prolective or retrolective. Descriptive or "case" studies could theoretically use prolective, but almost always appeal to retrolective data compilation. One other point bears mention. Epidemiological research often reports mortality data. Indeed, the causal links between lifestyles and mortality resulting from specified conditions are undoubtedly highly interesting. The perspective of this particular research, however, being the scope for economic incentives, dictates an interest in morbidity states and the costs incurred in treating them. If everyone over the age of 60 who shovels snow is found to die suddenly of heart attacks, a public intervention may be warranted on altruistic or paternalistic grounds. But a death penalty, or a tax on snow shovels ,• - 63 -would do little to further equity or efficiency goals (unless funeral parlours are run by the State). In our discussion below of the "strength" criterion, we will be describing comparative measures such as relative and attributable risk. The relative mortality risk will not, except by coincidence, be identical to the relative morbidity risk. Because mortality is often easier to monitor than morbidity, it is generally reported more frequently. While we may use mortality data to proxy morbidity information, the distinction noted above is critical to the discussion of economic incentives. Research Designs and Methodologic Considerations The search for causality is a quest for evidence in support of "asymmetrical statements that have direction" (Susser, 1973, 65). Thus, we are not interested in relationships of the A = M (symmetric) variety, but rather of the A --> M configuration. By A --> M we mean more specifically that a change in lifestyle leads to (or is followed by) a change in health status. Furthermore, we are not particularly interested in statistical associations between A and M, except to the extent that they may be used as points of initiation for causal investigations. Causality is generally reflected in correlation, but correlation does not imply causality. Two general reasons for associations not being causal in nature may be labelled "spurious correlations" and "indirect associations" (Hausner and Bahn, 1974). The former may arise by chance (more likely with small than with large samples), or because of any number of biases which may creep into a research design or analysis (Sackett, 1979). By contrast, a lifestyle - 64 -characteristic would be "indirectly associated" with a disease state if both the characteristic and the disease state were statistically correlated with a common third, or set of other, variables or factors. Thus evidence of an apparently causal relationship may in fact reflect a failure to control or standardize in the research design and/or analysis for the effects of confounding variables. The strongest of the research methods for determining etiology or causation is the randomized control trial. While not without its problems (in that the potential of introducing bias exists if extreme care is not taken in design), it allows the investigator to control the independent variable of interest and standardizes for all other potentially confounding factors. In effect, "random allocation converts confounding variables into uncontrolled variables" which are "equally distributed between groups" (Susser, 1973, 96). Of course randomized trials may be conducted on human or non-human subjects, the former trials providing the stronger human-related evidence. A randomized trial in the lifestyles context would involve the random allocation of persons to the A-group and the A-group. Then those in the A-group would be required to engage in activity A, and the researcher would follow the two groups and monitor the incidence of M. If the activity has an intensity dimension (e.g., number of cigarettes smoked per day), a spectrum of A-groups differentiated by A-intensity might be required. Not surprisingly, the "lifestyle"-related literature is not overrun with results from randomized trials. The ethical problems associated with imposing lifestyles (or parts thereof) on individuals are hardly likely to disappear in the foreseeable future, particularly when the reason for conducting such trials would presumably be a suspicion of - 65 -detrimental effect. The likelihood of designing such trials will undoubtedly be inversely proportional to the strength of currently available causal evidence on lifestyle-related disease etiology. Thus, we suspect that a longitudinal trial investigating the aggregate disease effects of jogging would pose more design than ethical problems. On the other hand, the ethical would far outweigh the design problems in a proposed trial to investigate the smoking ---> lung cancer link. It is worth noting as well, that although the randomized trial is commonly regarded as the definitive research design, the policy maker may encounter problems of generalizability. A clinical procedure or intervention may be established as efficacious in an experimental or controlled setting, yet be ineffective in a natural setting (for reasons of inadequate patient compliance, inadequate practitioner skills, or other factors in the therapeutic setting not accounted for in the design). Of course the potential policy inapplicability of human trial results is not the reason we discard them here, for the same may be true of cohort or case-control studies. Rather, it is the general prospects for coercing individuals to adopt specific lifestyle traits for the cause of etiological inquiry that dictate that the method will not concern us further here. Next on the list of research designs using human subjects is the "cohort" study. It shares "direction" with the randomized trial in that individuals are followed prospectively from A (or A) to M or M. The data as noted earlier, may be collected prospectively or retrospectively. Thus, an example of a "prolective" cohort study would - 66 -be the purposive selection of two groups of individuals, appropriately matched in as many dimensions as possible, with the exception of activity A. 3 One group would contain individuals already involved in A (note the distinction from the experimental design), the other would be the I "control". Then the groups would be monitored for evidence of M. In a "retrolective" cohort study, the direction of population monitoring would be the same, but the researcher would resort to an already existing data base for determination of the A-group and A-group individuals. In the context of lifestyles, "retrolective" data collection for cohort studies is likely to be the exception rather than the rule. A sufficient retrolective data base would require not only the A vs. I separator information, but also information for matching of the populations. It will be the rare occasion when data collected for purposes unrelated to the research at hand provide, ex post, the necessary information base and are accessible (i.e., not confidential). The potential sources of bias which make the cohort study a weaker design then the randomized trial are well-documented in any epidemiology reference. The general problem is, of course, that once one moves away from randomization, it is much more difficult to control for all conceivable confounding factors. In addition, the temporal nature of the research will mean that individuals in each group may change their lifestyle status over time, necessitating their removal (if monitorable) from the research study. The time requirements of such studies are, in general, problematic. The third type of research design is most commonly labelled "case-control 11. 4 By this method, Mand R would first be identified, and then - 67 -the researcher would follow each group back in time to compare the incidence of A-type lifestyle behaviour within each group. As with cohort studies, data collection may be "retrolective" or 11prolective 11 • Results from case-control studies are of much less potential etiological value than those from cohort studies, because of the overwhelming possibilities for the introduction of bias (Feinstein, . 1973; Sackett, 1979), which may be cumulatively severe enough to completely negate the results. The confounding variable problem is, of course, again pervasive. A major problem not experienced with cohort analyses is that of chronology bias (Feinstein, 1973, 297) -- the M and R groups will of necessity be comprised only of the survivors of the original underlying cohorts. "For this and other similar reasons, the results of case-control studies are tenuous at best in sorting out the etiology and causation of human illness" (McMaster University, 1981, 987). 5 Finally, we mention briefly the generic observational study --alternatively labelled as descriptive studies, case studies, or cross-sectional studies. This type of study is confined in its etiological contribution to identifying statistical associations. Thus a study might report for example, that among 10,000 lung cancer patients in Country X at time t, 8,ooo or 80% indicated that they had smoked at least half a pack of cigarettes a day for consecutive periods in excess of five years. Unfortunately such studies tend often to be sensationalistic and are used to infer causality where no such inference is possible from the data viewed in isolation. For the purposes of establishing causal lifestyle relationships, - 68 -then, we will prefer results from cohort studies (in the absence of randomized control trials). Failing those, we may look to case-control design, and from descriptive studies we will be looking primarily for consistency with results based on other research methods. Causal Criteria Following the CMAJ Clinical Epidemiology Rounds (McMaster University, 1981), we will discuss briefly the seven causal criteria included in Figure 6. The most important criterion will be the strength of the association. An association will be strong if, for example, the ratio of the probabilities x1/x2 in Figure 3 of the previous chapter is large. This has, in fact, two components -- an epidemiological component and a policy relevance component. While the former is of primary interest in this Chapter, from the perspective of rationales for public policy interventions the latter may be equally important. To see the difference, consider two hypothetical lifestyle characteristics, A and B, and the two corresponding disease states M and N. Say as a result of repeated cohort studies it has been established that individuals engaging in activity A are five hundred times as likely to end up in state M as those in the A group. Suppose, in contrast, that the incidence of N is three times as high among B as among B individuals. Finally, suppose that within the population as a whole, the incidence of Mis 1:1,000,000, while that of N is 1:5,000. There is little doubt about the relative epidemiologic importance of A and B. But the policy importance of M (depending, of course, on the cost of treating M may be trivial. In contrast, the incidence of N may be sufficient to draw - 69 -policy attention to B despite the much lower relative risk. In short, incidence can be more important than the epidemiologic strength of the 6 targeted etiology. The two most commonly applied indicators of causal strength are the relative risk ratio and the concept of attributable risk. The relative risk ratio is as it sounds, the relative probability of ending up at an "effect", with and without the purported causal agent. Attributable risk is that share of the total "effect" incidence among the A-group, for example, that is attributable to the A-behaviour. Cohort studies provide a direct method of calculating relative risks, as shown in the following lifestyles context: M A Say that two cohorts are drawn up, one containing persons who engage in lifestyle activity A, the other containing "abstainers". The total number of persons in the A group is (m1 + m1), of whom m1 end up in morbidity state M, and m1 end up "not-morbid". Similarly, m2 of a total of Cm2 + m2) from the A group become afflicted with condition M. Then the risk of contracting M for an A-group individual relative to that for an A-group member is given by: [m,/Cm, + m,)J/[m2/(m2 + m2)]. Note that in Figure 3 of the previous chapter, the relative risk is x1/x2• - 70 -Squeezing excess risk information from case control studies is much more difficult. Note that now, rather than starting with Cm1 + m1) persons with A-group behaviour and (m2 + m2) persons with A-group behaviour, the starting point is Cm1 + m2) persons with condition M, and Cm1 + m2> persons with condition M. The computation of relative risk requires the denominators (m1 + m1) and (m2 + m2). But in a case-control study, these sums have no meaning. They "are wholly contrived values formed by the cases included in the arbitrary selection of diseased and non-diseased people that form the trohoc" (Feinstein, 1973, 301), because "the incidence of cases can be ascertained neither among those exposed nor among those not exposed" (Susser, 1973, 88). One can salvage some causal strength information from case control studies under certain conditions. Recall that the relative risk in a cohort study is given by [m1/(m1 + m1)]/[m2/(m2 + m2)]. If m1 is small relative to m1, and m2 is small relative to m2, then this ratio may be approximated by [m1/m1J/[m2/m2J. This is of more than passing algebraic interest, since the absolute incidence of many diseases in which we might be interested may be rather small, and may still be small (although perhaps not as small) for an "exposed" A-group. Thus, the odds ratio or cross-product ratio provides a relative risk proxy in case-control studies which satisfies certain conditions. Not only must the incidence of the condition M in the general population be small, but the observations making up the cases and controls must be representative of the general population. - 71 -In general, relative risk information is more informative than attributable risk information, in assessing the strength of a putative causal relationship, whereas attributable risk information will be more informative from the policy impact perspective. While we have gone on at some length about the strength criterion, this is not inappropriate given its rank among the causal criteria available in the absence of randomized experiments. It is also important to realize why relative risk information from cohort studies constitutes stronger evidence than similar information from case-control studies, as suggested by the placement of the two cells in our Figure 6. We move on, then, to the internal consistency criterion. Here we will be looking for repeated reporting of similar results from different settings and using different methods. There is always a chance, of course, that a particular bias may be common to all designs suggesting a particular linkage, i.e., spurious consistency. With increased replication of a given result, the chance of such an occurrence decreases. We distinguish between "internal consistency" which refers to repetitive evidence addressed to the same etiological phenomenon, and external consistency. The latter, a far less important criterion, would require consistency of the linkage under investigation, and other peripherally related linkages. For example, "this yardstick would link the scrotal cancer of chimney sweeps in a former era with the more recent appearance of lung cancer among persons who inhale, rather than wear, the products of combustion" (MQMaster University, 1981, 989). A third criterion to be used in assessing cohort and case-control evidence is temporality. The postulated causal agent must precede temporally the occurrence of the "effect", or A must not only precede M, but the intervening period must be long enough to allow "for any necessary - 72 -period of induction and latency" (Hausner and Bahn, 1974, 1010). Temporality will be more difficult to establish for conditions with long latency periods. The dose-response test of causality amounts to checking that increased intensity of activity A is associated with increased (or decreased) incidence of state M, or increased (decreased) costs of treatment for state M (if, for example, the A-intensity is reflected in M-severity). Evidence of a reverse dose-response would also tend to support a causal relationship. If A-group individuals who reform and switch over .to I-group membership show an M-state incidence inversely proportional to the length of time since the switch, a reverse gradient would be said to exist. In short, causality is supported if a change in A is reflected in a change in M. The dose response criterion goes beyond A vs. A, to consider proportional or reverse A gradients or intensities. What is of interest is that these changes in lifestyle get reflected in a consistent form in the M stage of the chain. If the results of an epidemiological investigation are consistent with what is deemed plausible physiologically or biologically, we are likely to give them greater weight in our causality assessment. This criterion is not of great importance for two reasons. First, it is dependent on the state of another body of knowledge at any given point in time; and second, biological plausibility (like economic theory) may be as much a function of imagination as of hard fact. Finally, specificity refers to a one-to-one relationship between A and M. Thus, if A were both necessary and sufficient for M, and A were not implicated in any other M-states, a one-to-one relationship would be said to exist. This criterion is of limited use because A as a - 73 -potential causal agent in a number of different M-states need not necessarily refute its causal role in any particular M-state. Lack of specificity will not refute a causal hypothesis. In conclusion, the evidence which we will examine for causal linkages will be comprised largely of cohort and case control studies. The major criteria of interest will be strength of the association (relative and attributable risks), internal consistency, temporality and dose-response. Of course even after taking account of all these criteria, "the causal significance of an association is [still] a matter of judgement which goes beyond any statement of statistical probability." (United States, 1982, 20). Indeed, the epidemiologic evidence must be but one input (necessary but not sufficient) into the process of designing appropriate public policies directed at lifestyle-related issues. This discussion has allowed a relatively brief overview of the nature of the epidemiologic link in the lifestyles --> health care costs chain. In particular, we have attempted to provide a concise epidemiological roadmap of methods and causal criteria. As a result of this discussion, we take a slightly constricted Figure 6 into the following sections as a framework. Figure 7 reflects our elimination of the human randomized trial as a potential methodology in lifestyle-related etiological investigations. In the following sections we review the state of causal evidence for a number of diseases linked to smoking and alcohol consumption. For each prospective causal link, we will be looking for concentrations of causal "confirmation" toward the upper and left side of the unshaded portion of Figure 7. Figure 7 Method Randomized Cantrall ed Trial Cohort Study Case-Control Study Descriptive Study Evidence from true experiments in humans Strength Internal Consistency Ternporality Dose-Res pons e Biological Plausibility Specificity External Consistency -...J ~ - 75 -4.2 The Epidemiologic Evidence We attempt in the following sections to synopsize key epidemiologic studies relating to four putative lifestyle links: smoking------> lung cancer; smoking-----> coronary heart disease; smoking chronic obstructive lung disease; and alcohol----> cirrhosis of the liver. The reference list at the conclusion of the report provides the reader with an indication of the literature examined. Of the numerous studies reviewed, only a select few for each causal investigation explicitly receive attention in this chapter. This approach was adopted for two reasons -- the objective of this chapter was not a comprehensive epidemiologic review; and we have restricted our close attention to those studies providing the 'best' evidence based on the criteria described in the previous section. 4.2.1 Smoking as Causal Agent 2.1a Lung Cancer The most comprehensive source of information on the nature of the linkage between smoking and morbidity and mortality is undoubtedly the U.S. Public Health Service's series of reports of the Surgeon General. A recent report in that series, The Health Consequences of Smoking=. Cancer (United States, 1982), covers lung cancer in great detail. The Report adheres closely to the causal strength framework adopted in the previous section of this chapter7, and provides a comprehensive review of cohort and case control studies examining relative and - 76 -attributable risks as well as dose-response patterns. The following discussion is a synopsis of key information from that source. "Lung cancer, first correlated with smoking over 50 years ago, is the single largest contributor to the total cancer death rate. [It] ••• accounts for fully 25 percent of all cancer deaths in [the United States]; it is estimated that 85 percent of lung cancer cases are due to cigarette smoking. Overall, smokers are 10 times more likely to die from lung cancer than are nonsmokers. Heavy smokers are 15 to 25 times more at risk than nonsmokers. The lung cancer death rate for women will soon surpass that of breast cancer •••• , currently the leading cause of cancer mortality in women •••• The 5-year survival rate for cancer of the lung is less than 10 percent. This rate has not changed in 20 years. Early diagnosis and treatment do not appreciably alter this dismal survival rate." (ibid, vi) The strongest non-experimental (or sub-experimental) design for epidemiological research is the cohort study. While we noted in the previous section that both cohort and case-control studies could be prospective or retrospective in temporal direction, the Report (ibid.) implies that the cohort studies reviewed were all prospective, and the case control studies undertaken retrospectively. Most of the prospective studies reviewed reported relative mortality statistics. But both our equity and efficiency objectives require morbidity evidence - efficiency being directed at changes in patterns of utilization, and equity - 77 -at changes in the financing of those patterns. In this situation our only recourse was to use data on five year survival rates to "justify" our use of mortality data. The extremely low five year survival rate for lung cancer, and the fact that mortality is most often preceded by some morbidity pattern (i.e. sudden death is not the rule) suggest that mortality-based epidemiological evidence suitably proxies relative risk of morbidity for this disease. The Report (ibid.) identifies eight cohort studies providing relative smoker/non-smoker mortality rate information. Tables 1 and 2 (taken from the Report) provide synoptic descriptions and relative risk summaries respectively. Further details may be found in the Report and in the respective references (provided therein). While there is considerable variation in the mortality ratios in Table 2, even the low ends of the male (3.76) and female (2.03) ratios suggest a significant relative risk of mortality. At the high end (14.2 for male Canadian veterans; 5.0 for British female physicians) there seems little doubt about the pervasiveness of the effect of smoking on lung cancer mortality rates. Retrospective investigations of this relationship (using various methodologies) not surprisingly have been more numerous. The Report tabulates 35 such pre 1971 studies (Table 3) and notes a continuation of the pattern of results for post-1971 findings. This genre of studies provides an even wider Table 1: OUTLINE OF EIGHT MAJOR PROSPECTIVE STUDIES Dall Dora &e.l Allthon Hill Hamll!Dlld Kahn Hira,._ JOiie Humnand Pelo Ham Pille ~ Walker Mala and Total population llrltilh remala U.S. or Call9dian White --S.bjedl dodan in ftl.enns 29 t-IUI in 25 diltricili in penaiaMn lline Stala Stale ,,.,.. l'llpulatiuft .. 40.000 J,000,000 290,000 21i6,000 9'l,l100 187,000 Feinale l,000 562,671 <l'I. 142.1151 14,000 •n• a.as+ Z-84 ..... 40 .,._90 50--48 and ap Y-ol 1!151 19SJ 1954 1• 191111 1!162 """llllllll 1951 y..,."' l>-22 followap J2,.,.. 1619'9 JS ynn .,_. 4 ,.an ,.,.. ,_. NH•'-., u.1• 150,000 ICrJ,,llOO 39,ICIO 11.000 12,000 .. u. re,_,.,.. or IOD,llOO 8,000,000 uoo.ooo 3,000,000 500,000 1'111,000 ·~ Source: U.S. (1982), p.33. Weir Dunn Linden Bralow c.Ji(amia a.in in ftrioul omapaliolll 18,DOO IMC 1954 1-1 ,_,. 4,'lllO 480,000 ~ Fnliers Rn bee Lllrich ~1 lmllpleol the hedilll ......... J5,00ll n,'lllO a.a .,. .,... 4,500 11511,000 ..... CX> - 79 -Table 2: LUNG CANCER MORTALITY RATIOS - PROSPECTIVE STUDIES Population Siu Number Nonsmokers Cigarette of deaths smokers British 34,000 males 441 1.00 14.0 Physicians 6,194 females 27 1.00 5.0 Swedish 27 ,000 males 55 1.00 7.0 Study 28,000 females 8 1.00 4.5 Japanese 122,000 males 940 1.00 3.76 Study 143,000 females 304 1.00 2.03 ACS 25-State 358,000 males 2018 1.00 8.53 Study 483,000 females 439 1.00 3.58 U.S. Veterans 290,000 males 3126 1.00 11.28 Canadian . Veterans 78,000 males 331 1.00 14.2 ACS 9-State Study 188,000 males 448 .1.00 10.73 California males in 9 occupations 68,000 males 368 1.00 7.61 Source: U.S. (1982), p.36. - 80 -Table 3: RELATIVE RISK RATIOS FOR LUNG CANCER MORTALITY, RETROSPECTIVE STUDIES, 1939•1970 Year/Author Mate·· Female'" 1939 Millier (/9/J 5.4+ 1943 Schairer and Schoniger (237) 5.7• 1945 Potter and Tully (2/3) 4.1 • 1948 Wassink (288) 4.7 1950 Schrek et al. 124'11 1.8 1950 Mills and Port.er 11811 5.7 1950 Levin el al. 1155al 1.5 1950 Wynder and Graham 1315) 13.0 2.9 1952 McConnell et al. 11781 1.2 2.8 1952 Doll and Hill (6/l 9.4 2.1 1953 Sadowsky et al. (230) 3.9 1953 Wynder and Cornfield (311) 6.1+ 1953 Koulumies (/47) 36.0 1953 Lickint (/561 10.4• 5.3 1954 Breslow et al. (34) 3.2 1954 Watson and Conte (289) 5.6• 3.3 1954 Gsell (90) 26.8+ 1954 Randig <215) 5.1 + 2.2 1956 Wy:ider et ai. (308) 1.4 1957 Segi et al. (2481 1957 Mills and Port.er (/82) 4.2 0.6 1957 Stocks 1259) 4.9 1.6 1957 Schwartz and Denoix 1245) 10.4 1958 Haenszel et al. (94) 2.5 1959 Lombard and Snegirerr (/6/l 7.9 1960 Pernu 12091 8.4 1.9 1962 Haenszel et al. (931 5.2 1962 Lancaster <152) 9.8 1964 Haenszel and Taeuber (95) 1.3 1966 Wicken (295) 3.9 1968 Gelfand et al. (87) 25.3+ 2.9 1968 Hitosugi (/2/l 2.6 2.3 1969 Bradshaw and Sch"nland (33) 1969 Ormos et al. (205) 9.3 0.2 1970 Wynder et al. (319) 20.8+ 6.78 • Compui..d according lo method or Cornfield <491. • • Ratio or smoker to noru;moker. + Baaed upon fewer than 5 cue nonamohrs. Source: U.S. ( 1982), p. 3 5. - 81 -range of relative risks for males (from as low as 1.2 to upwards of 30), and a range for females quite comparable to that from the prospective studies (most results falling in the 1.5 to 7 range). What are we able to conclude about causality from these results? The most important causal criterion was strength of the relationship or association. While there are no 'gold standard' relative risks which permit unequivocal conclusions (causality after all can never be proved, only refuted), it should be clear to all but the most dyed-in-the-wool smoker that male smokers, in particular, are five to ten times as likely to become lung cancer mortality statistics as their non-smoking counterparts. While the comparable figures for female smokers are somewhat less likely to instill panic in prospective smokers, relative risks of 2 to 5 are still difficult to ignore. It is the consistency of the evidence which is perhaps most striking about this association. No ratios for males of 1.0 or less than 1.0 were reported in any of the studies synopsized in the Report, and very few (and then only retrospective) studies estimated male relative risks between 1.0 and 2.0. Among studies of the association in females, two of those cited reported relative risks less than l.O, but the vast majority (including all the prospective studies) reported ratios larger than 2.0. Relative risk of lung cancer mortality has, then, consistently been found to be greater for smokers than for non-smokers. - 82 -Temporality is more difficult to assess definitively from retrospective than from prospective data since it is not always possible to establish that a physiological state S (see p~evious chapter) did not exist at the time of onset of the behaviour of interest. However the prospective studies of this association "in which an enormous number of [presumably] initially disease-free subjects were followed over varying time intervals" (ibid, p.39) do provide temporal support. Further support comes from investigations of histologic changes in the bronchial epithelium of cigarette-smokers (ibid, pp. 55-9), which have suggested that such changes "progress toward cancer in continuing smokers but reverse in ex-smokers" (ibid, p. 42). In a similar vein is Doll and Peto•s (1976) finding of a reverse relationship between lung cancer mortality and years of smoking cessation among former smokers. A causal relationship would be expected to exhibit dose response characteristics; that is, causality could be cast in doubt if relative mortality or morbidity rates were dose-invariant. 'Dosage' has, of course, both intensity and temporality dimensions. At any particular point in time, the number of type-standardized (e.g. for filter differences) cigarettes smoked per day and the smoker's inhalation behaviour would represent smoking intensity. But smokers' patterns may change over time and, at the very least, there was a time in every smoker's life when (s)he was not a smoker. Thus, information on duration of smoking may be of equal - 83 -importance. Table 4 provides information from the eight prospective studies on the intensity aspect of 'dosage'. A clear, dramatic, consistent and direct relationship may be observed between number of cigarettes smoked per day and mortality ratios. The 'age at start' evidence provides further corroborative weight, as three of the pros.pective studies which report this information show an inverse relationship between age at start and mortality ratios. In general, then, the dose response evidence supports evidence on the strength of the association. While . this discussion has covered the four most important causal criteria, it is worth noting as well that the dose-response evidence, consistent evidence that cessation is associated with reduced mortality, and the earlier-noted evidence on histologic changes in bronchial epithelia related to ~ntensity of smoking, all support the biological plausibility of the association. In summary, the strength, consistency, temporality and biological plausibility of the association between cigarette smoking and lung cancer are such as to provide strong evidence of a causal link from this lifestyle behaviour to the subsequent morbid state of lung cancer. The relationship is not "specific" -- that is, even if smoking were both a necessary and sufficient condition for the inducement of lung cancer, it is also associated with other disease states. Furthermore, the evidence indicates clearly that the joint necessity/sufficiency conditions do not hold -- there are smokers who die of other causes,and there are - 84 -Table 4: LUNG CANCER MORTALITY RATIOS FOR MEN AND WOMEN, BY NUMBER OF CIGARETTES SMOKED PER DAY - PROSPECTIVE STUDIES Men Women Cigarettes Mortality Cigarettes Mortality Population smoked per day ratios 1moked per day ratios ACS 25-State Nonsmoker 1.00 Nonsmoker 1.00 Study 1-9 4.62 1-9 1.30 10-19 8.62 10-19 2.40 20-39 14.69 20-39 4.90 40+ 18.71 40+ 7.50 British Nonsmoker 1.00 Nonsmoker 1.00 Physicians 1-14 7.80 1-14 1.28 Study 1~24 12.70 1~24 6.41 25+ 25.10 25+ 29.71 Swedish Study Nonsmoker 1.00 Nonsmoker 1.00 1-7 2.30 1-7 1.80 8-15 8.80 8-15 11.30 16+ 13.70 16+ Japanese Study Nonsmoker 1.00 Nonsmoker 1.0 All ages 1-19 3.49 < 20 1.90 20-39 5.69 20.29 4.20 40+ 6.45 U.S. Veterans Nonsmoker 1.00 Study 1-9 3.89 10-20 9.63 21-39 16.70 ;'2:40 23.70 ACS 9-State Nonsmoker 1.00 Study 1-9 8.00 10-20 10.50 20+ 23.40 Canadian Nonsmoker 1.00 Veterans 1-9 9.50 10-20 15.80 20+ 17.30 California males Nonsmoker 1.00 in nine about 1/ 1 pk 3.72 occupations about 1 pk 9.05 about 1 1/ 1 pk 9.56 Source: U.S. (1982), p.38. - 85 -(admittedly few) oases of lung oanoer unrelated to smoking. None of this is particularly surprising and in faot it does not detract from the powerful body of evidence pointing to a causal relationship. These factors do, however, take on some importance at the policy stage. We return to issues of policy after examining the remaining putative causal relationships. 2.1b Coronary Heart Disease (CHD) Unlike lung oanoer, coronary heart disease is not represented by a single manifestation. Atherosclerosis, myocardial infarction and angina peotoris are all regarded as indicative of existence of CHD broadly defined, although risk factors may not bear equally on eaoh manifestation. Often, though by no means always, these three conditions follow the temporal relationship, atherosolerosis-->isohemia-->angina peotoris-->myooardial infarction (MI) although the absence of angina but fatal MI, or the presence of angina but death from causes other than MI are not uncommon. In the following discussion we focus primarily on myocardial infarction (fatal or non-fatal). This treatment is governed partly by relative inoidenoe of presenting conditions, but more fundamentally by relative availability of epidemiological evidence bearing on the issue of causality. While coronary atherosclerosis is not an uncommon diagnosis, at 8 least for acute hospitalizations, there seems a rela'tive dearth of epidemiological evidence linking this condition to various risk factors. This may be because without pathologic - 86 -examination, atherosclerosis is less precise a labelling than myocardial infarction. Similarly, definitive diagnosis for angina is more difficult than for MI. In any case, the little evidence on atherosclerosis which does exist is drawn largely from autopsy studies (U.S., 1979). But only rarely has smoking behaviour been available for cohorts whose deceased members were examined for pathology. The HEW 1971 edition of~ Health Consequences of Smoking (U.S., 1971), brought together data from 6 major autopsy studies (Table 19; repeated as Table 1, Ch. 4 in U.S., 1979). Four of the six suggest a dose-related positive relationship between degree of atherosclerosis and smoking history, although the data precluded multivariate analysis. The 1979 report reviews the results of a number of more recent studies, including one prospective study. These additional data tend to reinforce the earlier findings. The epidemiologic evidence bearing on smoking as a causal agent in the incidence of myocardial infarction is far more extensive. However, the studies most often cited in this regard have been around for some time. In fact, tables synopsizing the major prospective studies of CHD mortality and morbidity appeared first in the 1971 edition of the HEW series (U.S., 1971), were repeated in the 1976 edition (U.S., 1976) (as part of the repetition of the entire 1971 CHD chapter), and again in the 1979 edition. Recognizing a good thing when we see one, we have replicated the mortality evidence as our Table 5, and the relative risk of morbidity table as our Table 6. Discussion of these tables may be found in U.S. (1971, PP• A•IW, .,.,, -"' l'•ffl'ftl .... ~"'al. l!lill. ru Taylar ~I al.. 1970. ru ..... o.-. 11111, U.s.A. .... "9jld. A..-hrt "--. 1"111, U.1.A. Table 5: Na•borud lJPf" ........ 3.llfi.1 .... .....,...,._ nwn 1' 61 ~~ .... ..r ..... u;1.,.... , ......... ......... 'Y_ 40 59 y..an nf ..... , •111~ .. 151 Cllil-faniamlt ....... ..... ,... ., .... •llJ. 1,ffrwllilit .... ... ,.,. ., ... •llJ. CORONARY HEART DISEASE MORTALITY RATIOS RELATED TO SMOKING - PROSPECTIVE STUDIES. (Actual number of deaths shown in parentheses)l 0.la mllectioa la11..i Blulti-1o1wir ..-m'llin,r """, .. ~ ... "''"' ..... th .. .. ,1r ....... lnl.t'f\·W-1 ...... ftJUlar t••lua·up ........ i-. q..c;.. Mire ... ra11ow..., "dMdi mtifal& lleclicll ... ......... ... f ...... FallD•· Nu1a ..... .. ol U-01 ..... " ~· ~ ~ M l,Tll .. Zll CipftllaldaJ SS anrl • :!II 1m11:1i1 !\:II ·:!II :!W-1 1!'41 ., •• ·111111 SS ······ 1111 UI '31 .. ··-·· l.!li t:!lll ·:!II . ... :Ifill 1:!!1 NS •• •••• •• UID All..._ I.Ill ~IO ••••• •• 1.39 ~· ....... l.&7 ;>311 ••••••• 1.74 NS •• ••• ••• 1.111(27) <10 ... . ... U6 Ptl 211 . . . • • . • • • 1.70 11161 >211 • • . . . . • uo llill Cipn, pipes l.Clll (%1) 1.31 fZt) [SM=Smokers NS=Nonsmokers] ,.. •lrialioa r.-44 ~ lif>-6' .... NS .•• •••••. I.Ill I.Ill I.Ill I.Ill ~•o ........ U2 2.05 UI 1.11 ~· ........ &.It 1.17 LM I.JI ~30 . ••• ••• . &.Sl UI ... UI >• ........ 7.S'I us uz uz AR ••• •• •••• l1A 115 UI I.JI (con't) r-.-&1 llala ai1•y ... ~ ... th.or '"" "111D ml t1llt)" NS ..... ..... ... cisan-Sii illduda a...,.__ •Uai.. .u.rw-. epecit" ... m.p.riU. bet- the '°'81 au111ber ol deaU. and the au111 of the individual iunokinc calqarim are due lo the esduaiDn of either-maal, .m-IW-. miaed, or ea-........_ Pp" n1- epecit"_. oalJ far dime proriM 111 aulhan. Source: U.S. (1979), pp.4-22 to 4-26. I 00 -..J Aulhnr, ,nr. cwnl~ llaha ,,. l:.S.A Hn~-.ma. l!l&i, Ja .. n i.an ... 1 •.• al .• l!llill, l'SA Hamn111111f allll (iarfi""rf. 19&!1, l'llA. ... rr .... 1 ... ....... 1 ..... 1!111!1 l':\A Source: Table 5: CORONARY HEART DISEASE MORTALITY RATIOS RELATED TO SMOKING - PROSPECTIVE STUDIES. (con't) (Actual number of deaths shown in parentheses)l Sulltlorr a..t IH., .. r ,,. .. ....... ,_ "~ftlinll rs . .w. ........... ... - ......... Z.-.6'14 r.n...r-ap ......... oldrath .... an. c•·rtmrai.. :!ID.II~ Trainrd Ill-Jai- l.l'F\irtltn adult.""' and,.,, ..... Ill" 40. "JI ..r rlnth rmifnk. 5.12i..U.. :11...tftl H• ..... , ......... ami11a1inll "II\' 30 59 .... ,.~ .......... ~ill,,!1.'14 Qlk'!ltii• ........ nail\' ... 1 U.\,lliS , ............ , ....... "dn1h .,....,19 rmirira1r. •• ftlll')·. 'ill.1111 mat. 11-~llW ''""".,. 1ntt1'\· .. • •luok·nr.. .,.1 ~\am-1naliolft and fnl~••·u1• lo~ ok·•lh n'f'l1fir1tr. U.S. (1979) I pp.4-22 t'o~•••. Su1111 • .,. a11 111' I 'ipn ti•'" rla~ 1~1-ar..I 1k111h. 1112 10.11!111 Sll ........ 11111~1 AH .-on 1.14 141~1 I 9 ..... . 1.39 1'1!11 10 211 .....• 1.111 1210'll 2139 .. . .. . 1.114 112!1'll ">39 ... .. .. 2.0ll 121&&1 I 91 Sll .. . .. . . . llJl llil I~ .... 1131~1 .~ ... .. 111111:11 I:! !i:! Sll . . .. . . 1011 l:!il lp0 , llll'il llM ,>:91 ... :!311~1 6 l~.1'1!1 ,. ........... ~ .... SS ....... 1111 1.1•1 I 9 ... .. l.Zi (I.~ Ill 19 .. . . . l.$1 l.t! :lll :ti - Ii:! I :,i ~I • .. I.ii lllil 1; 51 I.I- SS 1111 m;a11·ht1I ll'.11 "~' 1:11<."ol 11 .. n 1111 ••lh :!.:....,..! rontrof" to 4-26. ,~tp'JI. 1• ..... r,..n NS .. t.m SM • 1.04 l&Z11 ~ NS .. I.Ill SM • I.Ill 131151 [SM=Smokers NS=Nonsmokers] Art" •anal•111 r .......... r. Pft'limie-Ill')' """"' '"'"'" 1a....1 ... -111-1!1 50 !i9 11111!1 'lll i'J 5!1rhtho Sll .. . ... . 1111 11111 1m tm I 9 .. . . . . . trill l.!19 ... u~ 10 19 .. . :!.59 2.13 lie:! UI :9131 . . . . !!iii Ul 1.91 l8 .... .. .... 5.51 :!.i9 I ':9 Ui F.malt-s Sll .... I.Ill 1111 tm I.I• I 9 ·- - · Ill 1.15 1111 1.-;& 111 19 •. . . • . !I .. :!.li li9 0911 lll :111 . . •. . :I.Ii:! 2.&I :!.111 I.ti .-111 .. . ... 1331 3.1:1 t!rrl :II~ ~ .. 54 :n.&9 Sll ... I Ill Im I ••I •1·· 111111 • ll!lt 1111111011 I 81 111'11 131 IL'\41 continued ••• I 00 00 I Table 5: CORONARY HEART DISEASE MORTALITY RATIOS RELATED TO SMOKING - PROSPECTIVE STUDIES. (Actual number of deaths shown in parentheses)l [SM=Smokers NS=Nonsmokers] A4111o ... Su•i...,..,.1 ......... s ......... )nr, 1)1"' or llala "'' or C'...,..tLrs da) ('"ipn_ ....... . ......... l'Gnln111> ftlUllll) ....... llull ...... '<"lion l)'Mlnl .... 1i.. ......... 1117.ilCI Qiatinn- 3 12 !>2!11 NS ....•••• 1.00 (11191 c..,,. 50M M4!1 81 && a• and •-hi~ulal .. .., .. All SDllcn 1.70 IDill 'll•<D.001) NS •. l.llO h"S .... .. ... I.Ill 19111 l.llO 1142! l.llO 13111 Ull lti'31 "- ia ' .... "" , ............ <.ID .. .. .. . 1.29 119'ZI SM • UJI (Giii All """*n l.!1311&il 1.115 19621 1• C9:lll IAI Cll!ll 19511, 511&1~ "'d.alll 10 31 .. .... 1.11!1 (11611 l'ipts <10 ........ l.311 (a'il l.3111511! 1.n1491 1.%7 l!'illl l:.S.A. "'•· Mlirira"'- 3140 ...... %.al lliMI NS •• l.llO ID l1I ....... 2.111 r..it:11 2.DI 1!5111 u11m1 Uill 115111 >40 .... .. . 2.41 (llRI Siii • I.ID (312) >211 ........ UI 13111 U711991 19211211 L!i 1131 °"Jlr ZJllZ..i... O..lailrd 10 t:I NS ........ Ull(311 n.ta "11">" tt al .• ,,.... ..... AD-.krn 2.411 ff.II .., la..., 1-.. ....... rumi .. <31 ....... 2.Cll (171 ....... .,, l:.S.A. 31162~ tm.ud 211 ......... 1.711 f2111 _.,_ er.,. fallow-up. · 11 >211 . .... .. l.50 (35) .r CHO .a 1.11:1..i... mtry. SS AU.111. illrludr ... 39:1.\,_. riprand CX> .... d.....__ \D Dal ud Appn11i- Qustioit- 10 1,316 NS ........ l.llO 3.\-'t 6" a11& Hill, -1e11 aaift ud All ..... .n l.3$ NS ....... .. l.llO' 1.1111 I.ID .... Cl.IOI , ....... ..., I It ....... 1.29 I 14 ........ 3.13 1.40 I.Tl r.-t .. Bri!M aldmll IUA ...... 1.%1 1ne ....... Ua 1.13 l.!r am- ..,..... mtifale. >Z't ....... l.G >Z .. .. .... 1.3& LIZ UI :ltrul...t 3.7•.U. Qu<'llio• ' lfiZ SS • .. . 1.00 NS .. .. . .. . 1.00 a111l litt-ll i; ........ ,. ...... ..c S9' ....... I.~ '""~ ........ r••lu.--up I 31 ... . . .. UR li•llJo.'f'· ol dralh ... :ai ...••. 1.76 ...... n:rt1ra1r . llt"SI. A1'1'"'•i· Qunt- 6 ;?.Int SS . . ... . 1no r;pn :IH9 50-e 70 ..ta .. r 1966 maLrl)' ... .,.. .... All ....... ,.,,. 1.611 .. ~ • SS .. 1.00 SS .. ... .. .. llJI I.Oii I.Oii ......... 111.WO , .......... , • 10 ..... . I» 1:tr.1 SY . 0.911 1161 <10 ...... 0.9i 11"1 1.56 IZlll 1.11 1991 .... irr .... of ilcalh IU 31 ... • IAA li&lil l'i .... ID 2D .. .. . Ua llL~I 1.6715571 l.l!I 19'1 ..... ftrlifirat.e. ,31 . .• • I 1M 1tl11 NS I Ill >lll .. .. .... I.II.\ 1&~1 1.16 CINI l.T.112111 \t-\d'lllll s... . 0.96 1951 Source: U.S. (1979)' pp.4-22 to 4-26. cont1nued .•• Table 6: CORONARY HEART DISEASE MORBIDITY AS RELATED TO SMOKING. (R;sk ratios-actual number of CHO manifestatfons shown ;n parentheses) 1 [SM=Smokers NS=Nonsmokers EX=Ex-smolkers] rRosrECTtVE S1'UDIF.S Authar, N11111brr and Data Follow. Nu111htt of Jftl', t~of callmian up inrioknls Cipnotlcs:da7 ~in. riian Ace tfrialioa fCllllllltllts nun try papulatioll ~ ~ 2JJIZ males Drtailtd 10 KJ 111)'1>- NS ..•••.•..•... ... .. ·~SZ) Data illrludr ft al~ Fra111iapa111, lll!llital rardial All •lllllliftS .. . .. ... 2.3(,(1911 CHO dalhl, 1964, 3M2~ e111111ina- inf•~- <31 ........•... . . ... 1.9111441 an!)- on malts t:.S.A. of •. lion and tmand 31 .. ...... •.... .. . ... 111516'1 •• ,_s"' l,913 main ro11ow-up. rno >31 ................. 3.11411131 • and rree ol Albany, ikaths. rHO on enlrJ. ~Jiran HS illdudrs of•· pipes. a.an. andn~ I \0 Stamllr 1.329 CHO- 1111.erYiew • t6 CHO NS ... .. .•• . .•.. ..... I.mm NS illduclm 0 tt al., ,_mate. and eulftitt. < 10 riprdlcs .•..... 1~61 e1-smalien. ••• en1ploJftl or ation wilh <hips ··········· lnrludes all t:.5.A. """""Ga dinic <5 pipes •....••..••. rHO. f.ompuy ro11ow-. 10 19 ric-lta .... 3.61(8) t0--51 ye.n >31 riprdln . .. ... 3.R3(29) °'•· >5 riian ·········· >5 piJe •. .......... Jrnla111•. 3,1112 main Initial t l ·Z UM lllJI>" HS ... . ... ..•. ...... l.~211 ttnr1.i..-rt al, 39 51 ye.n -lical rardial El ... ........ .... .. Ultllll Cp<0.11111 •• »• llllllkenand 19&11, of at" at e111111ifta. infarrl._ furftllt .. ..... .... . . 2.'IM(&ll) HS ••••••••••• 1.1111•1 11111111 n-~ tSA. tllllJ. lion and D tll ·tlay .... . . .. ... . tl.391.SI lr<0.11111 Cumnt •.•..•. U3(3S) 12&C331 NS inrluds ro11ow..ap >16 ················· 3.061591 (nimpari"I( farnwr pipe .., ... , t-15 and I&+ I ...... e1a111ina- ..... en. tionl. Source: U.S. (1979)' pp.4·-27 to 4-33 cont; nued ••. A•lllar. ,.,, -·llJ 1 ..... ll ... 1919, l!.S.A. Epotaia, 1967, l!.S.A Shapi"' et al., 198!1, l'.S.A Source: Table 6: CORONARY HEART DISEASE MORBIDITY AS RELATED TO SMOKING. (can't) N•mbrr and lJPI' of populat.. 5,12'1 ...... .... ,_.... 301111 )Un "'• 1,5&1 male ud rema~ Nlidtall .r-Teeu..ti. MidL 110,000 male and rema1e dnlllfts or Health lnsura-Plan or Greak!r New Yon IHIPI 3S 6t ~an or. (Risk ratios-actual number of CHO manifestations shown in parentheses)l [SM=Smokers NS=Nonsmokers EX=Ex-smokers] l'IWSPtrl'IVE h"Tl!IJlt::S Daia Fullcnr· ~:umbrr or Nllldioft up iacidtall fiiaret1&s.da7 Pipts, ciian ~ ........ Jan Medical 12 zal myo- llyacardial lnramion 1111111iaalion canlial llales Femalea and ro11ow. in rare- NS ....... . .....••.. 1.11111211 1.11111311 up. lions. All Sii ............. Ul(ls.11 1.71(231 3M1 rHD. HeafJ SM .......... l.115C:l91 Rask or CHO (O\walll .... Female! NS ....••••..•..•••. • 1.1111161) 1.11111891 I 10 ........ .. . ..••. 1.34(~1 0.8&(181 11-211 ... ....••.•.. 1.1111901 1.29(181 >211 ... .............. 2.41(7&) 11.93(3) Initial • 96 male, llalei llales llltdical 92r~ ..., Ill al and- 40.1111 esam1aa- CHDin- NS .••..........••••• 1.1111111 1.111(71 SM .....•..•.•. l.Nll2) lion and duding F.X . •. ..••..•..•.•••. 6.Sll 101 1.27(111 81 and.,..,,.. ~at .... lhs, C°lpfttla ........... S.311361 l.9fiC231 SM ..... . .. .. .. 11.116(6) ro11o.--up U«ina. and Female! esamiaa- myocanlial NS .. ..... .. ..... 1.11111211 1.11111471 ..__ inramions. EX •.•... .. .. . .... 0.119131 l.311SI Ciprettei .... .. . . ... 1.Ql41 O.Q2) llurliae nwd- 3 T"tal ....... Females lllaleonl7 .... al inltt· umpfti- SS ..... .. .. ....•.. l.Clll 1.1111 NS ......... l.Clll 3.\-44 654 S.Ut 3.\-44 wiew and ried. AD cumal. .. .. ...... Z.14 2.Clll SM ......... l.&! 1.1111 l.Clll l.Clll l.Clll examination ripnolta .......... (p<0.01) h•>0.011 (p<0.011 Z.47 3.111 ... ~ and ~lar <31 ... .. . . . ....... . . I.SO 0.52 2.IS 1.32 rolk1w-up. >31 ......... .. ... .. Z.3.1 1.71 3.IN 3.29 1.81 115 ~ ... ..... .. ... ... 6.36 10.CB 79 5.30 21115 S.92 U.S. (1979). pp.4-27 to 4-33 eo.uiwa1a Rersaanmliuii or .. llenll -.-i ID ower I I Z~Jar ..... period. but dala arc~ par1md in "'-"' ~)- iari-....., rala Attal number al CHD illO-cletill d.rifed r ..... dala• ---ul l.IUI ...... ..,r1aa Fnmta Tolal •Y. 654 ~ -.lial ia-l.Clll 1.1111 ramian • U1 I.Ill indud... thca 2.31 Ui5 dad witkui 411 huun 11.'19 4.07 NS iadude H-llllllbn. continued ••• A•tliar, ,..., -""' ,., .. 1970 ,.,... ...... F1•nd llal7 Nether-landa Gmm TaJlar, d al. It'll U.S.A. Daytoll " al~ lt'lll, U.S.A. Table 6: CORONARY HEART DISEASE MORBIDITY AS RELATED TO SMOKING. (con't) N_..,.ud tnr or pap11latiali t,tll5111Mt ill Sew• lries eM ,an or • at ntrJ. 1.571 ... nitra.I _...,_ .. Jan of •Ill "'"'· 4ZZ male U.S. Ydera•pu-tiripmlillf .. lllllltrob in a dinical trial ar a did hip ia •-lu-rated rat (Risk ratios-actual number of CHO manifestations shown in parentheses)l [SM•Smokers NSaNonsmokers EX=Ex-smokers] l'ROOPfl:'l'JVE srum•~ !>ala Fallow- N11..ti.r Dr aillediam .., illcidrnt. C'"apmlalda7 Pi ..... cican qe varialDI Jan 111\tnirwa 5 &5 ileach:t. t.'S, EX ....... NI myorar- CSM <311 --······ '-~3111) .., ra11ow. dial in- All rurrt'nt llflHAlniU- ramians. (>311 ············• 1.31(1001 .... ., 1211 ancina latal 1mrl.aris. pllyiiciara. IS.\ Dlhrr tOl l.olal. ,....,,.,... 5 ... u.. NS and El .......•• l.~IZ) .... ... 33 lllJDCU'• AU cum:at . ...•.•.• l.'J'.11150) ..,. ro11ow- lfial.f• 11f1 eumilla- rardionl. tiDft_ 11 allliM pKloril. 55 other CHD. Zl2 Iota!. lal«Wiewl up lo II 11 IUddrn <10 ......... ..•.. •.. l.CQZI) and nmtine desO.. 10 31 ······•••······ l.04(22) rona......, '4 defmile >211 ·---············· t.11(131 , .... ma. ~ .. tionL infudions. Source: U.S. (1979), pp.4-27 to 4-33 r..-tlt. lllrlucla .n CHD1~ inrlurli111 EKG cmc-. C.0.dS all -ntriis;. -·....r el~ l:SA. trur-bet-- tolal CHD ud lhe -or_.iar ,_.. .. 1odifremn •r.-. pianl<!d., •than. -All CHD i9dudi11t EKG cliai--No data 1111 NS a a t!paral.I! K""'P· continued ••• \0 N Aulllar, ,.,, -·llJ Duaa It al., '"" l!.S.A. Paol1111 Ptojld, Alllff'ican Heart AmmatiDll '""· U.5.A. hul cl ... 11111, l'.S.A. Table 6: CORONARY HEART DISEASE MORBIDITY AS RELATED TO SMOKING. (con't) Numb.rud lJPR of papulatillll 13,148malc .. bell .. ill penadir i.e.Jlll cn•••boa cliaica. T.4Z'1 •llM inala».sl Jan., .. al HllJ. l,911t Wateni tledilr r.o. male woncn panlri .. li. .. .....,.. am 1111111 far 4 I•% JCBl'L (Risk ratios-actual number of CHO manifestations shown in parenth-eses)l [SM=Smokers NS=Nonsmokers EX=Ex-smokers] PROSPn;TIVE Sl'UDIES Data Follow· Nalllb.r of eallectiaD ... iacidmla CilUftlalclaJ Pipes. a.us ~ warialiDa C--la Jan Dala aa17 up lo 14 Tolal •• ••• ~mat. illiridrnla calnded f..-clillic ..... .... ID i.'111 cnmina&ian lad...S.. Ncwcf smoUd ... .. . ud follow· fatal and <10 •• •... . ... . . . .. •• • -fatal 31 ················· myacudial >31 ....... ... .. ..... infamion ud ...WCa -~ Smnill( ca..U...tioll ud llill«J. NS ...... ....... ..•.. 1·7 . .•• . .•.. •.. •••• • g 12 . . .•. . . ..• . .•. . 13-17 ... ... . . ... . . .. 111-%2 . •. ••.•.• •. •••• Zl-21 ... .. ..... ... .. >28 . ... . . .. .. . . .. .. l.OO(s:I) ··~ 2.0M(3ll) Utl(15t) N~ c-r, l'Olllrals -1111 (1,1116) Zl 33 % 1 ' II ' 12 41 30 3 2 9 ' lp<0.006) 1.111(5.1) l.%1(5t) .... ti.-Sii UIG(25) tffisll SM U1UDI ...... tlad...S.. NS. El. .... tclll(tai 1.111(15'1) <1111 ~.,. ~II UIC»I t >211 ap. m1a1da7 . ladlllla.U CHO but _..... -~ No clala .... ... Clllllp8rilw _._ .... --... Mdmluped din al _, --47 ...,.; ... porlanl. 28 111yucardial warm.. 13 dratlll CHD. •Unlw olhenrile mpecil"iecl, di.,ariliel between the lotal number of manifalaliona and the .um of the individual amokintr i:alqioriea are due lo lhe uduaion of eilher-ioaal, am..etia-_ miaed, or H4111CJken. Source: U.S. (1979), pp.4-27 to 4-33 '° w - 94 -38-40). The major prospective studies of CHD mortality show relative risks for all cigarette smokers falling in the range 1.35 to 2.4, with considerable variation by age. !n partiouiar, the elderly were found in some studies to have lower relative risks (given smoking intensity) than their younger counterparts. There are clear dose-response patterns evident in the large majority of these studies, and these patterns are reinforced by selective data indicating an association of CHD mortality with degree of inhalation and initial smoking age. Of more relevance to the focus of this project is the epidemiologio evidence linking smoking to MI-related morbidity. Table 6 summarizes the major prospective studies which had been reported prior to 1971. CHD morbidity includes myocardial infarction and angina peotoris in this Table. The relative risk of morbidity displays a dose-response relationship similar to that in the majority of the mortality rate studies. The 1971 Surgeon General's Report (U.S. 1971, Table A6) also summarizes information from a number of case control studies which uniformly indicate a higher proportion of smokers among CHD oases than among controls. In the previous section on lung cancer, we noted the relevance of smoking cessation information to the temporal dimension of causality. Considerable evidence and debate surrounds the issue of the effect of smoking cessation on CHD morbidity. The evidence as assembled in the 1971 Report (U.S., 1971, Table 15; repeated as Table 4, Ch. 4 in U.S., 1979) is reproduced here as - 95 -Table 7, and we consider the debate in our discussion below. The four sets of results in Table 7 clearly illustrate the reduced risk both of CHD and death from CHD among ex-smokers relative to continuing cigarette smokers. To complete the synopsis of epidemiologic information provided in the 1971 report, we consider angina pectoris as an isolated manifestation. Table 8 contains morbidity data from four prospective studies (U.S., 1971, Table 5; repeated as Table 6, Chapter 4, in U.S. 1979). The evidence is at best equivocal. Only one of the studies was able to establish statistically significant differences between smokers and non-smokers in relative risk of CHO. While this evidence on the relative risk of CHD morbidity for smokers may appear somewhat dated, evidence emerging since then has not appreciably altered whatever conclusions one could draw in 1971. Much of the research effort has attempted to adjust for confounding variables and synergistic effects, and much of the debate has focused on the role of the ex-smokers evidence in the pursuit of causality. The major post-1971 and pre-1979 evidence is reviewed in U.S. (1979). The post-1971 evidence relating to angina (ibid. 4-46 to 4-48) continues to be indicative of a positive association between smoking and angina, although strongly significant differences between smokers and non-smokers continue to be the exception rather than the rule. Selected literature appearing since 1979 is included in our bibliography. - 96 -Table 7: THE EFFECT OF THE CESSATION OF CIGARETTE SMOKING ON THE INCIDENCE OF CHO. (Incidence ratios-actual number of cases or events are shown in parentheses) Author, year, country Jenkins, et al., 1968 U.S.A. Hammond and Garfinkel, 1969, U.S.A. Shapiro, et al., 1969, U.S.A. Pooling Project, American Heart .Association 1970, U.S:A. Results All CHD events Never smoked .. . . .. .. .. .. • .. .. . . . .. 1.00(30) Current cigti.rette smokers ................ 2.36(84) Former cigarette smokers ................ 2.15(19) Death from CHD Smoked 1-19 cigarettes/day Never All myocardial infarction 1.00(21) 2. 78(68) 2.47(15) Smoked >20 cigarettes/day Comments smoked regularly ............... 1.00(1,841) 1.00(1,841) 2.55(2,822) 1.61(62) 1.51(154) 1.16(135) 1.25(133) 1.05(80) 1.2.8(564) Male data only Current cigarette smokers ............... 1.90(1,063) Stopped <1 year .................. 1.62(29) 1-4 ............................... 1.22(57) 6-9 ................................ 1.26(55) l(}-19 .............................. 0.96(52) >20 ............................... 1.08(70) All ex-cigarette smokers ......... 1.16(253) Total definite myocardial infarction Never smoked , ............. , ....... , ..... , . , ..................... 1.00 Current cigarette smokers ...................................... 1.87 Stopped 55 years ............................................... 0.76 All CHD deaths Never smoked ..................... . >112 pack/day ..................... . 1 pack/day ...................... .. >l pack/day ....................... . Ex-smokers ..........•.•.•.........•. First major coronary event 1.00(27) 1.65(34) 1.70(86) 3.00(68) 0.80(19) 1.00(53) 1.65(72) 2.08(205) 3.28(154) 1.25(51) Source: U.S. (1979), p.4-34. Autlicw, ,..,, -·"7 .,.,.., ft .. ~ llM, U.S.A. .loalli-. el .. ~ l ... U.1.A. a.a.I, elal.. U.S.A. ....... el .. i;: 1•. U.S.A. Table 8: CORONARY HEART DISEASE MORBIDITY AS RELATED TO SMOKING-ANGINA PECTORIS-PROSPECTIVE STUDIES. {Risk ratios-actual number of CHO manifestations shown in parentheses)l [SM=Smokers NS=Nonsmokers] Number ud tn- of papulatiaa UllZ ........ r ... ......, 80-G,..,. of mp. l,lll ....... Alt.a1, .... ,..,. ofap. 8,182 ....... ........ at •"7· l,l2T ..... ...if-a. ,..,. "' .. •• 110,DOD male Dela mlla-tion IJ.!ailed ...tieal Haminatioll and follow-up .Initial medieal HuNMtioll ...... follow-•p .,, . repoat Ham Ina-tao.. lilodiral Haminatioll and fnllow-up. Bueline Number ro11aw-up of CipfttlmldaJ ,..... inriMnl.I JD II 4''1 12 Ill NS 1.00(301 All . . ....... . • l .D!IC511 <20 . ..•......•.........••..••••... 1.17(15) 20 ... •.•.•...•.. .. ..•. . . .••....•. D.lllCIBI >20 . . ... •... . . . . . . ..•.. ···-······. l.15(181 29 NS 1.00(91 All tunTnl dtrvetla .. ............ . ....... l .44(111 >16 .. ........ •.. ... •. .•.•. •.• •• ••. l.1.'1(141 11'7 Malm NS .... . ..•. •••. •... •. •. . . ••.. .. . 1.00(161 Hl!&YJ SM, >211 dll"ft\119 ........•. . ..... . .. .. . 2.IM( 17) NS r-1.00(581 Cipfttte Sii •.•. .. . . .. . .. . •.. . .. 0.1111( ll) Cipn and pipm I Total ...... remai. 11a1m ,. ...n.tiDB ..... eon-.1.1 NS nodllde H· _._..and ........ <ipr _._.. NS iadude ,_pipe udcicw ....... t(p<DDl) Mllf-- ...sical u,.._.. NS . . . . I.DD . I.DD NS .• . •.... I.DD iir..4' I.DD 66' ~ llp<DCl&I enrallom of i11tttwie• irlod Cumtnl ,..., Yorti and na-i.. tUI CilJ e•&ft'lina- <40. ..•.•. l.51 HIPa.el tioll and >40 ..•.••.•..•....•••.•• 4.11& ,....., ,.ular ... follow-up. uo l.20 Sii . •. •... . tl.11 NS •. .. I.ID I.DD Curftnl eipftll.le... . .. . .. uo U1 Z.DI <40 . . .•• . . .•• .• ... .. •..•... Z.Sli 1.40 1.54 >40 ... ............. .. ... . 1015 1.511 1.15 .. ._._ NS . .. ••.. .... ••. ...... .... •. I DD I.DD I.DD <Aaftel dpfttl.le . ........ I.Ml 1.11 D.t'J < 40 . . ... •...•••. ... ... .•. . . l .fT l.51 I.CM >40 uz NB iadllde .,....,,..__ •V..- olllerwill 11•irled, dilparitim bel-n the total numi.r of 11W1uf•lation1 Mii the aum of the indiridual amoliiq .. ......,._Aft due to the esd...1..n of eithcroacMional, mileolia-, miaod,"' o ....... -. Source: U.S. {1979), p.4-47. '° ..... - 98 -In a controversial recent article, Seltzer (1980) raised some serious questions about the validity of much of the epidemiologic evidence bearing on the smoking-----> CHD causal link. He focused on the ex-smoker data reported by Doll & Hill (1964), Hammond (1966), Kahn (1966) and Gordon et al. (the Framingham Study, 1974). The thrust of his argument was that ex-smokers are not representative of smokers (i.e. the same as smokers in all possible confounding respects except that they quit smoking). This argument was backed up by data generated by the author and others (Friedman et al., 1979) suggesting that differences did indeed exist "in a number of cardiovascular symptoms, socio-personal characteristics •••• " etc., and that the nature of those differences was such as to indicate lower CHD risk among smokers destined to quit than among continuing smokers. A randomized controlled trial of smoking cessation (Rose and Hamilton, 1978) in which no significant differences were found in mortality rates, apparently supported Seltzer's contention. In addition, Seltzer pointed out a number of inconsistencies and equivocations in the 1979 Surgeon General's Report; most notable among those appeared to be the lack of a convincing biological rationale for a causal relationship between smoking and CHD. He further noted the lack of causal support implied by the evidence on angina pectoris, the lack of evidence on the role of smoking duration, and the lack of support for a causal relationship emerging from studies of twins. While not coming directly out in support of the alternative constitutional hypothesis, 9 Seltzer makes clear his feeling that the epidemiological evidence is insufficient to bridge the gap from statistical correlation to causality. - 99 -Since then, further evidence has been brought to bear on the issue of smoking cessation. Kannel (1981) reported on an investigation within the Framingham Study, which "failed to reveal any important differences which could account for the prompt reduction in risk observed in those who gave up the habit" (p. 325). In addition, he commented that "the [baseline] differences noted [by Seltzer (1980)] have only a speculative rather than a demonstrated relationship to CHD incidence" (ibid). Finally, he pointed out that the Rose and Hamilton (1978) trial was fraught with group switching, smoking habit switching, and lack of an objective, verifiable method of classifying quitters. Of even more harm to the Seltzer contention, however, are more recent results from the empirical work in which Seltzer himself was originally involved (Friedman et al., 1979). This work had established differences in baseline characteristics of ex-smokers and continuing smokers. In a follow-up article (Friedman et al., 1981), not including Seltzer as an author, the authors report on an analysis of relative mortality rates of the same study population after adjusting for baseline differences. Their finding that after "adjustment for major base-line differences in risk characteristics, the relative risk of dying from coronary heart disease among persistent smokers was 2.2 times that among quitters (p = 0.004 ••• ) 11 (p. 1407) led them to conclude that smoking cessation "appears to result in a substantial reduction in coronary and total mortality that cannot be explained by the characteristics of quitters before they quit" (p. 1410). More - 100 -recently, this debate has been put in perspective by Tweed (1982). While apparently overstating Seltzer's support for the constitutional hypothesis (at least as implied in his 1980 paper), Tweed neatly summarizes the objections to the causal theory, notes that some of those objections appear to have been undermined by more recent evidence (such as that cited above), and finally concludes that "the inconsistencies in the evidence ••• cast doubt on a simple causal relation between cigarette smoking and coronary heart disease". More likely, he suggests "that cigarette smoking enhances the risk of coronary disease through a complex interaction of toxic effects and constitutional susceptibility. In addition, psychosocial factors may influence both the risk of coronary heart disease and the susceptibility to smoking" (pp. 104-5). We return, then, to our causal criteria and attempt to assess the strength of the epidemiologic link in our information chain for CHD. What seems clear is that the strength of the relationship as implied by relative risks of morbidity and mortality, is considerably less than was evident for the smoking---- lung cancer relationship. Here we found most of the aggregated relative risks of MI in the range of 1.5 to 3. The relative risks of morbidity tended to be slightly lower than those of mortality in the studies which reported both. Relative risks appear to be inversely related to age, although the relationship was far from linear in most studies, and not even monotonic in some. - 101 -The evidence on CHO is also not all as consistent as that for lung cancer. While the consistency of the relative risks of morbidity and mortality from MI is really beyond question all studies in the synopsis show overall relative risks greater than 1.0, and most are greater than 1.5 -- the consistency of the evidence does not carry over to other manifestations (particularly angina), and much of the evidence based on ex-smokers and twins is still subject to some methodological question. The prospective study evidence on temporality is once again not as strong as that from lung cancer. The strength of the cohort type of analysis is presumably its ability to take initially disease-free groups and track them over time. But the earliest manifestation of CHO - atherosclerosis - is, of course, impossible to monitor accurately in large groups at the beginning of a study period. In other words, one cannot be sure that those who continued to smoke during a cohort study were as free of atherosclerosis at the beginning of the study as those who subsequently ceased smoking. This is nothing more, of course, than the constitutional hypothesis. One has to conclude, however, that the repeated findings of relative risks from prospective studies in the 1.5 - 3.0 range is at least strongly consistent with the required temporal direction. Furthermore, despite the debate, evidence from the smoking cessation literature (if it continues to hold up to methodological scrutiny) is strongly indicative of a temporal relationship from smoking to CHO. - 102 -We noted in the previous seotion that dosage oan be thought of both in terms of intensity and temporality. In one of these dimensions - intensity - the evidence seems relatively olear, at least for aoute MI. The other - temporality - provides no real support for a dose-response effect. As Seltzer (1980) notes, no temporal dose-related association is evident in four of the major prospective studies. In contrast, most of the studies cited in this section provide strong evidence of an intensity effect (inhaling and type of cigarette problems aside). In fact, among the large CHD population studies, only Hirayama (1967) fails to produce an overall positive relationship between number of cigarettes smoked per day and relative risk of morbidity or mortality. For angina alone, two of the four studies in Table 8 suggest no intensity-related dose-response. The biological plausibility criterion is, as inferred above, still out to jury. In addition, its strength appears to vary with CHD manifestation. With respect to atherosclerosis, the process of atherogenesis itself is not clearly understood, and the possible role of the various ingredients constituting cigarettes on that process is even less clear (U.S., 1979, pp. 4-8 to 4-10). As for the biological role of smoking in incidence of the more narrowly-defined CHD, attention has been focused primarily on carbon monoxide and nicotine. Some research (e.g. Aronow, 1976a, 1976b) suggests strongly that carbon monoxide and nicotine each play a role in promoting the development of CHD. Nicotine is purported to cause an increase - 103 -in myocardial oxygen demand which cannot be met in instances of impaired pulmonary function or myocardial ischemia, as well as to increase the adhesiveness of blood platelets leading to the promotion of build-up in atherosclerotic lesions in the coronary blood vessels. Furthermore, nicotine may promote irregular heart beat and increased beat rate. Carbon monoxide is postulated to work in quite a different manner, through its role in the formation of carboxyhemoglobin. Carboxyhemoglobin interferes with the delivery of oxygen to the myocardium and reduces the threshold of the heart to ventricular fibrillation in the presence of myocardial ischemia. (More detail may be found in U.S. (1971, 56-66). But there is not unanimous agreement among clinical researchers on these promoting roles. Seltzer (1980), for example, suggests that "facile postulations have been the order of the day, with nicotine and carbon monoxide featuring among the guilty culprits" (p. 277). Since the clinical detail is far beyond the scope or interests of this project, we simply conclude by adducing the weakness of this criterion from the existence of the continuing debates over the aetiology of atherosclerosis, the role of cigarettes in that aetiology, and the role of cigarette ingredients on the incidence of the more acute forms of CHD. The final criterion we consider, and then only briefly, is specificity. Hypertension, physical activity, and obesity are but a few of the other major factors which have been linked ·to CHD. While the strength of causal evidence on each of these is at least as equivocal as that for smoking, the point is that - 104 -we are very likely not dealing here with a specific one-to-one relationship. We know that many CHD cases cannot be linked back to smoking behaviour, and we also observe many smokers who live to ripe old ages and may or may not die of CHD . In summary, one is led to the inescapable conclusion that support for the smoking causal link hypothesis is far weaker for the CHD-complex than for lung cancer. However, the potential policy importance of CHD vastly exceeds that of lung cancer due to the sheer differences in incidence rates. For example, in 1980 in British Columbia, there were over 6,ooo deaths due to diseases of the heart, and about 1,000 from cancer of the lung, bronchus and trachea (B.c., 1983). On the morbidity side, the earlier-noted 12,500 acute hospital separations and 126,000 day for angina, atherosclerosis and acute myocardial infarction in 1981-82 may be contrasted with the 3210 separations and 50364 days attributable by primary diagnosis to malignant neoplasms of the bronchus and lung. In short, weaker causal evidence may be counter-balanced somewhat by the importanc~ of this disease complex in health care utilization and costs. But again, we leave the policy discussion to a later section. 2.lc Chronic Obstructive Lung Disease (COLD) The two disease states comprising the generic category of chronic obstructive lung disease - chronic bronchitis and emphysema -- 105 -comprise the majority of illnesses classified more broadly as "chronic non-neoplastic bronchopulmonary diseases" (U.S., 1979). Both refer to situations "characterized by chronic obstruction to airflow within the lungs" (U.S., 1971), and are often indistinguishable in patients with both. Dyspnea on exertion is a common symptom of both, with accompanying cough in bronchitis but not normally in emphysema. Chronic bronchitis "refers to a condition associated with the prolonged exposure to nonspecific bronchial irritants and accompanied by mucous hypersecretion and certain structural alterations in the bronchi.... In epidemiological studies, the presence of cough or sputum production on most days for at last three months of the year has sometimes been accepted as a criterion for the diagnosis" (U.S., 1979, 6-9). "Pulmonary emphysema is that anatomically defined condition of the lung characterized by an abnormal, permanent increase in size of the distal air spaces (beyond the terminal bronchiole) accompanied by destructive changes" (U.S., 1971, 139). Unlike lung cancer and particularly coronary heart disease, COLD is not a major cause of mortality. 10 It is, however, widely held to be the major cause of lost productive time at work, and a major contributor to aggregate ambulatory and 11 hospital services utilization (see, for example, Bass (1973)). Our major focus here will, therefore, be on the evidence linking smoking to COLD morbidity. We look first, however, at the epidemiological evidence on relative risks of mortality. - 106 -Table 9 is taken from U.S. (1971) and reports mortality ratios from the major prospective and retrospective studies. Table 10 provides updated data from U.S. (1979) on six of those studies. The relative (cigarette smoker/non-smoker) mortality ratios for COLD range from about 2.3 to 10.1. But more of the data refer specifically either to bronchitis or emphysema. For the former, the range is 4.5 to 11.6, and for the latter, 5.9 to 14.2. These figures are much closer to the orders of magnitude found for lung cancer than for coronary heart disease, and appear to provide dramatic evidence of increased risk of mortality from COLD among current smokers of cigarettes. Also very evident in Table 9 is the uniformity of evidence suggesting a strong positive dose-response relationship for both bronchitis and emphysema. The effect of smoking cessation on mortality does not appear to be a direct and •spontaneous' reduction in relative risk. This may be because of bias in the epidemiological data, or because much of the damage underlying COLD symptoms is irreversible or slow to repair. The potential bias is similar to that described in the context of CHD -- baseline differences in continuing smokers and smokers who quit. In particular, among those in the ex-smokers' groups may be a considerable number who quit because of ill health. Thus, Doll and Peto (1976), and Dirksen et.al. (1974) found COLD related mortality ratios for ex-smokers in excess of those for continuing smokers. Fletcher and Horn (1970) looked at "doctors age 35-64 in England and Wales, many of whom (had) stopped smoking cigarettes •••• to protect their health Author, year, country, reference Hammond and Hom, 1968, U.S.A. (JOS). Doll and Hill 1964 Great Britain (70). Best, 1966, Canada (10). Table 9: CHRONIC OBSTRUCTIVE BRONCHOPULMONARY DISEASE MORTALITY RATIOS (Actual number of deaths shown in parentheses)1 [SM=Smokers NS=Nonsmokers] Number and Data Follow-up type of collection 7e&rl population 187,783 white Questionnaire 3¥.r males in 9 and follow-up states 60-4i9 of death 7ears of age. certificate. Approximately Questionnaire 10 41,000 male and follow-up British of death ph11ician1. certificate. Approximately Questionnaire 6 '18,000 male and follow-up Canadian of death PROSPECTIVE STUDIES Number of deaths 338 SM ...••• 308 NS •..... 30 292 Chronic bronchitia 111 Other 181 124 Cigarettes/ day pipes, cigara Ciga.rettea NS ..••.•• 1.00 (80) <10 •••••• 1.67 (10) 10-20 ••••• 3.00 (67) >20 •..•• . 3.6' (40) All •...••. 2.85 (231) Pipu NS ...... 1.00 (80) SM •••••• 1.77 (23) Ciga.re NS ••..•• 1.00 (30) SM •••••• 1.29 (18) NS 1-H Chronic bronchitis Ciaarettee ••...• 1.00 ... • . 6.80 15-24 • •. 12.80 >25 . . •. . 21.20 All . . . . . 11.60 Pi11,.. and Ci11nr• ~M . . .. . :1 .no Cigarettee NS ••.•.. 1.00 <10 ••••• 7.02(17) NS <10 Emphysema Cigarette. ••••••• 1.00 •••••• 4.81 (9) veterans. certificate. 10-20 . •. 13.65(49) 10-20 .• • •• 6.1!2(21) >20 • ••• 14.63(12) >20 •••••• 6.93 (7) All • ..•. 11.42(78) All ••••.••.. 6Jl5(37) Pipea · Pi pee SM • ••••. 2.11 (6) SM •.••••• 0.'15 (2) Cigar• Cigar• SM •.•••• 3.57 (1) SM • ~ •..••• 3.SS (1) --Hammond, 440,658 male11 Interviews by ' 389 Al alee 1966, 562,671 ACSvolun- SM •. ...• 369 NS ••.•.•. 1.00 (20) U .S.A. females teers. NS • . . . • • 20 SM (atre (101). 35-8' years .(5-64) • . 6.65(194) of a.ire in SM (age 25 states. 65-79) .11.(1(175) Source: U.S. (1971), pp.142-144. Other Cigarettes NS ••••• 1.00 ...... 1-1' • . •• 0.66 0 ...... 15-24 •• 1.08 >26 •... 0.63 AU . .... 0.81 Pi1••·• atul Cigar• .SM . .. .. 0.711 continued ... Author, year, country, reference Table 9: CHRONIC OBSTRUCTIVE BRONCHOPULMONARY DISEASE MORTALITY RATIOS (con't) (Actual number of deaths shown in parentheses)l [SM=Smokers NS=Nonsmokers] Number and type of population Data collection Follow-up yea.rs Number of deaths Cigarettes/ dal· pipes, cigars PROSPECTIVE STUDY Chronic bronchi ti:< Emphysema Other ~~~~--~~~--~~~--~~~~~~~~~~~~~--~~~~~~~~~~~~~~~~~~~~~~~~ Kahn, U.S. male 1966, veteran• U.S.A. 2,265,67, ( Jll). person yean. Weir and 6ll,153 males Dunn, in ,·arious 1970, occupations U.S.A. in California. (.l?i?S). Wicken, 1,189 males. 1966, North-ern Ireland (127). Questionnaire and follow-up of death certificate. Questionnaire and follow-up of death certificate. Personal inter-view with relatives of Individuals listed on death register. 8% 5-8 Bronchitia SM ••••.•• 64 NS ••••••• 13 Emphr1•cma SM •••••• 284 NS ••••••• 18 58 NS •••••• 1.00 (31) All SM .•• 6.49 ( 348) Current ciga-rettes .10.08(229) Pipe a SM ••••.• 2.36 (9) Ciga.ra SM •••••• 0.79 (5) HETROSl'ECTlVE STLlIW 1,188 obtained retrospec-tively. SM •••. 1,064 NS •..•. 124 CurrC11t dg11-rettca onlr1 NS •.•••• l.00(13) 1-9 •••.•. 3.63 (5) 10-20 •••• 4.51 (22) 21-39 •••. 4.57 (12) >SS .•••. 8.31 (4) All ..•..• 4.49 (43) Cigarette a 011111 NS ...... 1.00 (124) 1-10 •.... 2.95(245) 11-2:? .•.. 3.43(300) >23 ..... 4 ·" (168) Mi:.rcd SM •••••. 1.55 (62) Pipc11 or cigar11 SM •••••• 1.84(289) Cxrrnat do11-rdte11 Otllv NS .•••••• 1.00 (18) 1-9 •.••••• 5.33 (10) 10-20 •••• 14.04 (93) 21-39 •••• 17.04 (62) >39 ••.•• 25.34 (17) All .••••• 14.17 (186) Ciga.rcttea NS . . .... :1.00 :::10 .••.. -.&.18 ::!:20 ...•. 11.80 1>30 ..... 20.86 All ..... :12.33 1 Unless otherwise specified, disparities between the total number of deaths and the sum of the individual smoking categories are due to the exclusion of either occasional, mlsceJlaneous, mixed, or e.'C•smokers. :: NS includes pipe and cigar smoke1·s; SM includes ex-smokers. Source: U.S. (1971), pp.142-144. ....... 0 00 Table 10: COLD MORTALITY RATIOS IN SIX PROSPECTIVE STUDIES British Men in 2.5 States U.S. Canadian Men in California Doctors 45-64 6&-79 Veterans Veterans 9 States Occupations Emphysema and/or bronchitis 24.7 - - 10.08 - 2.30 4.3 ....... 0 l.O Emphysema with-out bronchitis - 6.55 11.41 14.17 7.7 Bronchitis - - - 4.49 11.3 Source: U.S. (1979), p.6-10. - 110 -rather than as a response to ill health •••• (and found) a 24 percent reduction in bronchitis mortality between 1953-57 and 1961-65, as compared with a reduction of only 4 percent in all men of the same age in England and Wales, among whom there was no reduction of cigarette smoking" (U.S. 1971, 145). Thus, ex-smokers may in fact have lower relative risks than those who continue to smoke, but the reductions may be limited by irreparable cellular damage, or may be obfuscated by baseline differences in epidemiologic analyses. As we noted earlier, COLD becomes important from a policy perspective not so much because of high relative risks of mortality, but because of the pervasiveness of the COLD symptom complex as a cause of health services utilization and lost productive time. Morbidity related to COLD is commonly identified through use of a symptom complex including the symptoms noted earlier by which chronic bronchitis and emphysema are identified in many epidemiological studies. The 1971 Surgeon General's Report (U.S. 1971, Table A2, pp.195-205) provides a synopsis of evidence on relative symptom prevalence. The evidence provided there is again consistent and dramatic symptom prevalence rates among smokers tend to be at least twice those of non-smokers, frequently three to seven times as high, and have been found to be 25 times as high. The degree of 'control' exercised in these studies is likely highly variable, but it is difficult to argue with the consistency of the data even if they do not represent 'pure' relative risks. Where dosage information is reported, a clear dose-response - 111 -relationship is again evident. Symptom prevalence has also been reported to diminish in ex-smokers (U.S., 1971, 146). The other generic approach to establishing a relationship between smoking and COLD has involved investigations of physiological changes and functional capacity. These studies have attempted to identify changes which would show up as functional disabilities or relative functional impairment, and have investigated the potential of the functional indicators as predictors of subsequent COLD symptom complex incidence. It is this latter step which makes those investigations of interest in our examination of relative risks of morbidity. The literature is reviewed in U.S. (1979, 6-12 to 6-22) and suggests that one cannot in general extrapolate from early small airway functional changes to predict reliably subsequent COLD symptoms, and that the symptoms themselves are not always indicative of small airway function abnormalities. On the other hand, there is a significant body of literature suggesting a higher prevalence of lung function abnormalities in smokers than in non-smokers. Despite the imprecision in identifying COLD morbidity, the evidence is such as to be strongly indicative of a causal relationship. Most dramatic is evidence on the most important criterion -- strength. Relative risks or proxies have tended to be of orders of magnitude similar to those for lung cancer. Again as with lung cancer the consistency of the evidence is beyond question. A large number of studies using varying methodologies and subjects from varying age groups and - 112 -locations has produced time and again evidence suggesting relative risks of morbidity and mortality in excess of 2.0 and more likely approaching at least twice that level. Evidence on temporality most often comes from investigations of smoking cessation or investigations of pathology and physiological changes at various pre-morbid states. But smoking cessation will form a basis for temporality judgements only if smoking does not cause irreversible damage. As we noted above, the evidence from ex-smokers on COLD morbidity and mortality seems to us some way from being sorted out But U.S. (1979, 6-22) seems rather unequivocal in stating that "smoking cessation results in a reduced prevalence of symptoms in all age groups and in reduced mortality rates". That statement appears to be based solely on the evidence relating to functional capacity rather than symptom prevalence. As for pathologic studies, Auerbach et al. (1972) found a positive relationship between smoking and emphysema in lung sections, and subsequently (Auerbach et al. 1974) between smoking and physiological changes in a histologic tissue study. In addition, there was evidence of less physiological damage among ex-smokers than continuing smokers, but also of some permanent pathological . change among the ex-smokers, particularly the ex-heavy-smokers (U.S., 1979, 6-24). The dose-response evidence has already been alluded to. Both the mortality and morbidity ratios showed strong dose-related effects (at least as proxied by cigarettes per day). In - 113 -addition, Auerbach et al. (1972) found a dose-related deterioration in physiological condition from autopsied lung sections and Mitchell et al. (1976) have produced similar evidence relating to functional capacity. Other studies are reviewed in U.S. (1979). Much of the COLD-related research has focused on what we would consider to be investigations of the biological plausibility of a smoking---> COLD causal relationship. In fact, the functional, physiological and pathologic investigations noted in various contexts above, all bear indirectly on this criterion. But in addition, there has been increasing attention lately to "the mechanisms by which smoking might induce lung damage". We refer the interested reader to U.S. (1979, 6-25 to 6-32) for a review of this literature. The mechanisms ("altering protease-antiprotease bal~nce in the lungs, compromising immune mechanisms, and interfering with pulmonary clearance mechanisms "(p. 6-26) appear as yet not to be definitively implicated, although evidence pointing to the first of these appears strongest and most consistent. There is also a considerable amount of literature dealing with the specificity criterion. It may be divided into clinical and social science research. The former body of research has focused its attention primarily on genetic and occupational and environmental exposure factors, whereas the latter has attempted to adjust morbidity rates for the potential confounding effects of socioeconomic variables. The evidence suggests that there are some genetic susceptibilities to COLD - 114 -(ibid, pp. 6-33 to 6-36), that there is a strong possibility of synergies with certain occupational pollutants and general air pollution (although cigarette smoking remains the dominant independent factor) (ibid, pp. 6 -36 to 6-38, and Chapter 7), and that socioeconomic variables play a minor independent role, if any, in the development of COLD (ibid, pp. 6-38). 12 In summary, the smoking----> COLD relationship, even if causal, is clearly not specific. Smoking has been implicated in many disease states other than COLD, and COLD symptoms do appear (although not often in relative terms) among non-smokers. 2.ld Other "Causal" Links Cigarette smoking has been at least tentatively causally implicated in a wide range of other illnesses. The three dealt with in detail above were chosen because of the strength of the evidence suggesting a causal relationship and the potential policy significance deriving from attributable risks. Links to other illnesses are reviewed in two recent Reports of the Surgeon General, (U.S., 1979, 1982) and therefore receive only brief note here. While lung cancer has produced the strongest epidemiological evidence of a causal link to cigarette smoking, cancers of a number of other sites have also been linked to this lifestyle characteristic. The 1982 Report, The Health Consequences of Smoking=.::. Cancer (U.S., 1982) lists the larynx, oral cavity, esophagus, bladder, kidney, pancreas and possibly the stomach - 115 -as sites with increasing susceptibility to malignancies in smokers. The relative risks vary considerably. For laryngeal oanoer relative risks ranging ~rom 2 to 4o and a strong dose-response relationship have been reported. Those for cancers of the oral cavity have not been quite as dramatic (1.3 to 13), but again dose-response evidence has supported these relative risks in suggesting a causal link. Both prospective and retrospective studies show heightened risk of esophageal cancer among smokers relative to non-smokers. The major prospective studies have reported mortality ratios in the 1.8 to 6.5 range, and since this disease is virtually always quick and fatal, these may proxy relative morbidity ratios. Again a dose-response relationship has been found, and "histological data [suggest] that smoking antedates premalignant and malignant transformation of esophageal epithelium" (U.S., 1982, 96 - 7). What is of particular interest with this site is the frequency of studies suggesting a tobacco/alcohol synergy in esophageal carcinoma. Relative risks for cancer of the bladder and kidney are closer in general magnitude to those reviewed earlier for CHD, than to those for lung cancer, COLD, or cancer of the larynx. The 1982 Report (U.S., 1982) classifies cigarette smoking as a "contributory factor" in both. Similar conclusions were reached with regard to the pancreas. Only a weak association has been found between smoking and stomach cancer. - 116 -The prospective studies which have examined mortality from cancer of any site suggest a mortality ratio of about 2.0 for male smokers, and somewhat lower (about 1.3) for females. But as we have found already, the effects of cigarette smoking are far more pervasive than would be implied simply by examining relative risks of cancers. In addition to COLD and CHD, retrospective studies have suggested elevated relative risks of peptic ulcer disease, and there is a growing body of evidence suggesting that those who do not smoke may not escape heightened smoking-related risk levels. In particular, smoking mothers are now thought to be placing themselves and their unborn infants at increased risk (U.S., 1979, Ch. 8) of lower birth weights, spontaneous abortion, peri- and neonatal mortality, complications of pregnancy, etc.; and non-smokers may be put at elevated risk by the smoke of others in the non-smoker's proximate environment, making such individuals "passive smokers" (U.S., 1979, Ch. 11; Shephard, 1982). Finally, there is some reason to believe that smoking behaviour may increase the difficulty or cost of treatment for morbid conditions otherwise unrelated to smoking. Smokers may take longer to recover from anaesthesia, for example, or have more difficulty with post-surgical pulmonary function. It would be interesting to compare incidence of complications or lengths of stay in hospital, by smokers and non-smokers for particular conditions or procedures, adjusted for age and sex, but this has not to our knowledge been done. - 117 -2.le Summary This section of the report has attempted to review cursorily the nature and strength of the epidemiological information linking cigarette smoking to a number of morbid states. Our attention focused on three illnesses which we felt, a priori, were of the greatest policy significance or for which the epidemiological evidence suggesting a causal relationship was strongest. In addition, we looked briefly at other diseases in the previous sub-section. This review suggests a number of things. First, in no instances did we find smoking to satisfy both the necessity and sufficiency conditions. Even where relative risks were the highest -- cancers of the lung, esophagus, larynx, and COLD -- cigarette smoking was not a necessary condition •. That is, these illnesses are found, admittedly not often, among non-smokers. It was also not a sufficient condition -- almost everyone can relate tales of an aged relative who smoked "like a chimney" and died of unknown causes at 103 (give or take 15 years). Second, the condition or condition-complex of greatest policy significance among those reviewed, is surely coronary heart disease. Yet the epidemiological evidence on relative risks is considerably weaker (although not insignificant) than for, say, lung cancer or COLD. Third, the condition providing the greatest justification for public intervention may be COLD -- not only do smokers show symptom complex prevalence ratios on the order of magnitude of lung, oral cavity and laryngeal cancer relative risks, but the - 118 -prevalence of the complex in conditions presenting for medical care is much higher than for any of the cancers (although lower than for CHD). The evidence here suggests, then, that the epidemiological information is insufficient to justify the application of direct economic incentives at point and time of care, but may be sufficient to support economic incentives 'levied' at other points in the putative causal chain. In the following section we take a similar look at the nature of the alcohol consumption--- cirrhosis of the liver association. 4.2.2 ALCOHOL CONSUMPTION AS CAUSAL AGENT 2.2a Cirrhosis of the Liver ~ ~-Here again we have utilized the method of "secondary review" (i.e. using existing reviews of the literature). We draw extensively on Ashley and Corey (1980) who undertook an exhaustive review of the epidemiologic literature in order to develop risk factors for cirrhosis of the liver, for use in the Health Hazard Appraisal instrument. Other more limited or dated reviews include Lelbach (1976), Spritz (1979), and United States (1978). Alcohol has been widely found to be associated with at least three distinct liver abnormalities -- fatty liver, alcoholic hepatitis, and alcoholic cirrhosis. Many researchers hold that the degenerative progression to cirrhosis almost always - 119 -passes through the fatty liver and hepatitis phases first, although cirrhosis is rarely preceded by acute alcoholic hepatitis (Lieber, 1978). Both fatty liver and hepatitis appear to be (at least partially) reversible upon cessation of alcohol ingestion, while cirrhosis presents as "irreversible destruction of the liver parenchyma and fibrosis" (Ashley and Corey, 1980, 108). Not all alcoholic cirrhosis is fatal, of course, and drinking cessation has been found to increase survival prognosis in mild to moderate cases. We noticed a dramatic difference between the epidemiologic literature addressing the alcohol-----> cirrhosis linkage, and the smoking literature. In particular, in the case of alcohol much of the evidence is in the form of general population correlations between the incidence of death due to alcoholic cirrhosis, and per capita consumption of alcohol. Well-designed smaller sample prospective cohort studies seem non-existent, and even case control studies are not abundant. This may be due partially to measurement problems specifically related to alcohol intake and effects (Turner, et al, 1977). Spritz (1979) notes that the evidence suggesting a causal link is of three generic types -- epidemiologic, animal studies, and human experimentation, or what we would consider to be human population epidemiology, animal epidemiology, and human clinical epidemiology. As noted above, the epidemiologic research consists largely of population based associative evidence. Part of that evidence derives from such 'natural' - 120 -experiments as prohibitions and alcohol control programs which induced significant changes in population consumption patterns. Additionally, Ashley and Corey (1980) review in detail a number of case-control studies. We return to the human epidemiologic evidence momentarily. The strongest evidence arising from animal studies appears to be recent work with baboons (Lieber and De Carli, 1974) in which cirrhosis was produced by administering soi of calories as alcohol, despite maintenance of" the nutritional adequacy of the alcohol-containing diet. Fatty liver, alcohol hepatitis, and cirrhosis were all produced in the animals receiving alcohol ••• the intake levels were ••• comparable to those seen in severe human alcoholics" (Spritz, 1979, 2735). The population-based epidemiologic research has examined the relationship between cirrhosis mortality rates and geographic or intertemporal variation in consumption. Figure 8, taken from Schmidt (1977, 13) is one of the more dramatic portrayals of the geographic alcohol <---> cirrhosis association, given other evidence that "the general level of alcohol consumption in a population is closely related to the prevalence of heavy use" (ibid, p. 6). Figure 9, a Canadian equivalent, again illustrates a generally positive relationship between cirrhosis mortality and per capita consumption. Figure 10 provides one intertemporal example, for Ontario over the period 1928-72. Of course none of this does much to suggest anything other than consistent evidence of a strong positive association. - 121 -Figure 8: CIRRHOSIS MORTALITY PER 100,000 POPULATION 25 YEARS OF AGE AND OLDER AND ALCOHOL CONSUMPTION PER CAPITA. DEATH RATES ARE FOR 1972 EXCEPT THE U.S.A. AND BELGIUM (1971); CONSUMPTION FIGURES ARE THE 1968-1970 AVERAGE CONSUMPTION. FRAHtE 57.2 16.4 PORTUGAL 55.1 14.1 ITALY 52.1 14.0 AUSTRIA 49.1 11.4 WEST GERMANY 39.6 11.3 SPAIN 38.8 11.7 U.S.A. l!l.a 2ll.6 tZEtHOSLOVAKIA ------·--· 28.1 ....__ _______ 1 a.o SWITZERLAND ·-------1,24.6 ...__ __________ 10.0 HUNGARY ·------20.7 .__ _____ ___.... 8. 5 BELGIUM ·-----11, 20.5 L.....- ----~8.3tANADA ·----·· 19.6 .....__ ______ , 6.5 POLAND ·---··17.2 ____ _.15.5 DrnHARK ·----lJ16.2 _____ _,6.7 SWEDEtl 15.6 - L fver t f rrhos is 5.7 D tonsumptfon NORWAY 7.6 I 3.4 "" •• 94 FINLAND 7.S --, 4. 1 NETHERLANDS r 7.4 14.8 SOUTH IRELAND , 7.0 14.S UNITED KINGDOM 5.7 I 6.2 Source: Schmidt {1977), p.13. - 122 -Figure 9: CIRRHOSIS MORTALITY PER 100,000 POPULATION 25 YEARS OF AGE AND OLDER AND ALCOHOL CONSUMPTION PER CAPITA 15 YEARS OF AGE AND OLDER, 1971 FOR THE PROVINCES OF CANADA. - Liver Cirrhosis 0 Consiinption A. • • 82 BRITISH COLUMBIA •••••••••••••••• Z2.7 ___________ __,! 2.39 ONTARIO QUEBEC MANITOBA NOVA SCOTIA ALBERTA PRINCE EDWARD ISLAND l~EW BRUtrSWICK SASKATCHEWAN NEWFOUNDLAND ••••••••••••••zi.5 ....__ _______ ___, 2.15 ••••••••••••• 19.9 .._ _______ ___,! 1.88 ••••••••••• 17.7 ------------JI 2.06 15.5 .__ ______ ._ii 1. 71 ••••••••••• 15.2 '------------.J 2. 12 •••••••••• 15.2 .__ ______ __.11.64 .___ ____ ._ii 1.45 13.8 ••••••••12.2 ..._ _______ ___. 1. 76 •••••L9.2 .___ ____ ....... 1.38 Source: Schmidt (1977), p.14. - 123 -Figure 10: CIRRHOSIS MORTALITY PER 100,000 POPULATION 20 YEARS OF AGE AND OLDER AND ALCOHOL CONSUMPTION PER CAPITA POPULATION 15 YEARS OF AGE AND OLDER, ONTARIO 1928-1972. DEATH RATES WERE CORRECTED TO ALLOW FOR THE EFFECTS OF THE 6TH REVISION OF THE I.L.D.C.D. £ l 10..-~~~~~~~~~~~~~~~~~~~~~--21 • 11 • 17 7 lS ' u II I 7 ~L1ver tirrhoi1s De1th Rite (L) 2 s A• .12 ........................... ---------------.-..--_..., ......... .._... __________ .,....... ............ J 1928 1933 J9J8 I "'I llSJ IHI lHJ 1961 1972 Source: Schmidt (1977),. p.16. - 124 -Spurious correlations appear to be more convincingly ruled out by the associative evidence which has been provided by a number of 'natural experiments' involving sudden and dramatic shifts in consumption patterns, ceteris (assumed) paribus (Schmidt, 1975). Both world wars provided such experimental settings because of abrupt, severe diminution in alcohol supply. Schmidt (1977, 11) notes, in his review of epidemiological analyses of these phenomena, that "these reductions were always accompanied by a remarkably rapid and very substantial drop in the cirrhosis death rates", and that "sharp changes in cirrhosis mortality are not inconsistent with the clinical course of the disease. Often the cirrhotic process can be halted by abstinence and conversely a previously established liver pathology can be reactivated in a short time when drinking is resumed". Spritz (1979) notes additional evidence from prohibition in the U.S. and alcohol control programs in Scandinavia, and Ashley and Corey (1980) review other studies which have examined consumption/mortality lag times. The literature based on smaller samples is a source of dose-related information bearing on the etiologic linkage. Dose-response has, as always, both intensity and temporality dimensions. Ashley and Corey (1980) review evidence based on a number of dose-related measures -- daily consumption, duration of drinking, and the combinative "lifetime consumption" (ibid, p. 45). While Schmidt (1975, 19) notes that "the level of consumption [usually] increases with the - 125 -number of years of alcohol abuse", so that "it is usually not possible to determine the effect of quantity independent of duration of drinking", at least two studies standardized for duration in examining the effect of consumption, and vice-versa. Using a non-abuse/abuse critical point of 160g/day intake, Lelbach (1975) found highly significant differences in cirrhosis incidence. Eghoje and Juhl (1973) were reported by Ashley and Corey (1980) to have generated comparable evidence. A number of case-control studies which have not attempted any temporality standardization have, nevertheless, also produced corroborative dose-related evidence. Ashley and Corey (1980, 63-72) review in detail three such studies from France (Pequignot and Tuyns, 1975, 1978; Pequignot et al., 1978; Durbec et al., 1979), all of which strongly suggest a positive dose-response relationship, and which even suggest heightened risk at low (<40g/day) consumption levels. Tables 11 and 12 are taken from Ashley and Corey (1980, pp. 64-68), and illustrate a quadratic or exponential relationship between intake and relative risk. Ashley and Corey (ibid) also computed relative risks from the Durbec et al. (1979) data, and these are reported in Table 13. - 126 -Table 11: Evidence on Relationship Between Alcohol Intake and Relative - Risk of Asai tic Cirrhosis (I) -Alcohol Intake (Daily Average) Males --Females 0 - 20 g. 21 - 40 g. 41 - 60 g 61 - BO g. 81 -100 g. 101 -120 g. 121 -140 g. 141 -180 g. 181 -240 g. 241 -560 g. 0 - 20 g. 21 - 40 g. 41 - 60 g. 61 -220 g. Relative Risk 0.7 o.B 1. 7 3.2 11.5 14.1 26.7 40.3 115.5 1.0 3.0 12.0 204.0 Source: Adapted from Ashley and Corey (1980, pp.64-65), based on Pequignot and Tuyns (1975). - 127 -Table 12: Evidence on Relationship Between Alcohol Intake and Relative Risk of Ascitic Cirrhosis (II) Alcohol Intake (Daily consumption in grams) Males 0 - 20 21 - 40 41 - 60 61 - 80 81 -100 101 -120 121 -140 141 + Relative Risk 1. 0 3. 1 6.2 13.8 29.6 41.0 124.3 659.3 Source: Ashley and Corey (1980, p.67), based on Pequignot et al. ( 1978). Table 13: Evidence £!!. Relationship Between Alcohol Intake and Relative Risk of Cirrhosis (III) Average Daily Consumption i&2. 0 - 20 21 - 40 41 - 60 61 - 80 81 -100 101 -120 121 -140 141 -160 161 -180 181 -200 201 -220 221 -240 241 -260 261 -280 281 -300 300 + Relative Risk 1. 0 3.5 5.6 4.5 4.3 15.6 24.0 56.0 36.0 61.8 51.6 108.2 52.4 19.7 137.7 49.2 Source: Ashley and Corey (1980, p.171), based on Durbec et al. (1979). - 128 -Schmidt and de Lindt (1969) studied about 5,400 males and 1,100 females and established cirrhosis mortality ratios of 11.5 for men and 25 for women. There exists, then, a consistent body of evidence indicating not only a strong association, but a significant relationship between daily intake and cirrhosis incidence or mortality. While the nature of consumption information derived from surveys may be such as to bias the relationships toward upward concavity, the suggestion of a positive monotonic relationship at least over a significant portion of the observed range (0-160g/day) is difficult to dispute. The temporality dimension of 'dose-response' appears to have been investigated independently even less often. Ashley and Corey (1980) cite four studies which have addressed the issue, although apparently mis-reporting the results from Eghoje and Juhl (1973). The evidence, such as it is, in all cases suggested a positive relationship between duration of heavy ingestion, and likelihood of cirrhosis. Finally, a number of studies by Lelbach which consider the "joint" dose variable, "lifetime consumption", are reviewed in Ashley and Corey (1980). "Very stong correlations were found between "lifetime consumption 11 •••• and both severe liver damage and cirrhosis". (ibid, p. 45). In summary, the strength and dose-response criteria both point strongly to a causative link from alcohol abuse to cirrhosis. Furthermore, evidence from international and intertemporal - 129 -population studies on the one hand, and from sample-based case control studies on the other, is highly consistent, in direction even if not precisely in functional form, in terms of dose relationship. Evidence bearing on the temporality criterion comes from the animal experimentation noted earlier, and from the "cessation" literature. The natural experiments described above during which consumption dropped dramatically and was followed by precipitous drops in cirrhosis mortality rates certainly suggest a temporal direction running form consumption to disease state. Furthermore, pathogenic research has apparently established the staging of cirrhosis development, through fatty liver and often alcoholic hepatitis. This relationship, in fact, appears to have the strongest and best understood temporality and biological plausibility evidence from among the relationships we have examined in this chapter. Finally, the alcohol consumption-----> cirrhosis relationship is not specific for a number of reasons. First and most obvious is the fact that not all cirrhotic liver disease is alcohol-related. The proportion of cirrhosis attributable to alcohol is apparently in the 60% - 95% range (Ashley and Corey, 1980; Saunders et al., 1981). Second, alcohol has been at least anecdotally (and often more rigorously) implicated in disease states other than cirrhosis. We have mentioned two here (fatty liver and alcoholic hepatitis), and now turn very briefly to non-liver-related conditions. - 130 -2.2b Other "Causal" Links While a detailed epidemiological examination of associations between alcohol and other illnesses is clearly and deliberately beyond the scope of the present project, we note briefly in this section some of those putative causal and not-so-causal links. Turner et al. (1977a, 1977b) and U.S. (1978) have reviewed some of the more common alleged effects of excessive alcohol intake on various body organs and systems other than the liver. Within the gastrointestinal tract, alcohol has been implicated in esophageal damage (varicose veins, ulceration, chemical irritation, cancer), stomach lesions (erosive gastritis, ulcerations, other inflammations), small intestinal abnormalities (malabsorption, motility changes, cell damage, enzyme disruption), and pancreatic disease (chronic pancreatitis, although the independent role of alcohol as opposed to diet is still a .subject of some controversy (Turner et al., 1977a). While U.S. (1978) approaches the effects of alcohol on the nervous system from a biochemical perspective, Turner et al. (1977b) devote most of a paper appraising the epidemiological evidence on disease states -- alcohol withdrawal syndrome, psychosis and hallucinations, brain damage, and memory loss and blackout. They note that "many of the most serious neurological and mental disorders commonly associated with alcohol use arise as a consequence of prolonged excessive - 131 -intake but not necessarily from its direct effect" (p.274). Included in this category are the withdrawal syndrome, secondary nutritional disorders which can, however, arise in oases not involving alcohol use, and psychosis. More directly related to alcohol abuse appear to be hallucinations and brain damage, although good quantitative evidence on the relationship between intake and effect is scarce. Similarly, memory loss and blackout have been associated with acute episodes of intoxication, or chronic heavy intake, but here there appears to be somewhat more confirmatory evid~nce implicating rapid ingestion of upwards of 150g, or chronic ingestion of 300 or more grams. Alcohol withdrawal has been associated in experimental settings with sleep disturbance. In general, the evidence on direct effects is not all that strong and is often confounded by unknown interactive nutritional effects. Cardiovascular disease is considered by U.S. (1978), Turner et al. (1977a) and Spritz (1979) (atherosclerosis only). "An unresolved paradox exists concerning the relation of heavy alcohol consumption to myocardial disease. On the one hand, numerous references in the literature associate the two; on the other, overt cardiomyopathy (damage to the heart muscle) appears uncommon even in longstanding alcoholism" (Turner et al. 1977a, 241). Any relationship between alcohol intake and atherosclerosis has, if anything, suggested a protective rather than degenerative effect from moderate consumption, but difficulties associated with separating the effects of smoking - 132 -and alcohol consumption have confounded most of the research of this particular relationship (Spritz, 1979). In general, the evidence appears to suggest no particular role for alcohol consumption as a risk factor for atherosclerosis. Similar conclusions are suggested by the literature on acute myocardial infarction. The evidence suggests either no effect or a protective effect (Turner et al., 1977a, 241-2). The relationship between alcohol consumption and hypertension has been variously found to be "positive" or "no effect". In short, epidemiological analyses of coronary heart disease provide an insufficient basis from which to implicate alcohol consumption in a causative manner. Although limited research has addressed the role of alcohol in endocrine system disruption, U.S. (1978) is able to conclude at the end of its review that alcoholism causes reduced levels of testosterone in males, and that alcohol inhibits oxytoxin production and increases pancreatic insulin secretion. No epidemiological evidence linking alcohol abuse to specific disease states is provided, with the possible exception of hypoglycemia. Alcohol appears to be implicated in cancers of a number of sites. We already noted earlier the putative link to esophageal cancer. Also mentioned in U.S. (1978) are cancer of the larynx, pharynx, mouth, lung, stomach, colon, rectum, liver and pancreas. Table 14, taken from that report, synopsizes the evidence derived from prospective studies. The Table 14: RESULTS OF SELECTED PROSPECTIVE STUDIES ON THE RELATIONSHIP BETWEEN ALCOHOL CONSUMPTION AND CANCER lnve~tigator Population Size Deaths Number of Deaths from Specif ic Cancers and and of Total Deaths from Mouth Pharynx Larynx Lung Esopha- Stom- Colon Rectum Date of Years of Study All 1451 149' 161' 1621 KUS ach 1531 1541 Publication Observation Group Cancers 150' 1511 Sundby, 1967 (IOU Alcoholtcs treated on 1.722 Observed 1.061 204 4' 9' 5' 19' 40' 45 9 12 Oslo, 1925·62 Eapected 496.9 96.6 0.5 07 0.6 54 34 36.8 62 42 Schmidt and De l int. Alcoholtcs treated in 6,478 Observed 738 87' 6' 7' 30' 7' 7 8 1972 (92) Toronto. 1951-63 Eapected 346.23 68.11 - 175 078 1421 1.35 832 665 Nicholls et •I .• Abnormal drinkers on 4 mental 935 Observed 3091 50' - - - - 20' 5 4 3 1974(80) hospitals in London, 1953-67 Eapected 112 7 29.30 - - - - 1143 Hakulten et al., Alcohol mosusers on Chronic Observed - - - - 200" 101• - 82 -1974(40) f inland (males), Alcoholtc - - - 992 597 - 86.6 -1965·68 205,000 Eapected Alcoholic males 30+ 1n 4.370 Observed - Bl' - 3• 3 33' 4' 6 3 -Hels1nk1. 1967-70 Eapected - 633 - 0.53 216 2017 098 799 163 -Hirayama. 197 4 ' (481 Dally users of alcohol on 1265.1181 - 121.1671 15.5601 - - - - - - - -sample of male populatlllft Standard in Japan, 1966· 73 Mortality Rate 1.071 353' 151 140 1.28' 182' 097 122 082 Monson and Lyon, Alcoholics 1n Massachusetts 1.382 Observed 894 105 - 13 6 19 5 15 7 4 19751481 mental hospitals, 1930-71 Eapected - 137 3 - 3.9 16 141 2.6 146 112 57 lemon et al. Seventh·Day Adventosts 1n 47,866 Observed 3.456 608 - 3 - 39 19 47 - 97 1964 (621 California. 1955-59 Eapected 4.919 800 - 16 - 11 50 66 - 121 Enstrom. 1975 1271 Mormons 1n Caltlornoa - Observed 4.036 816 - 22 - 145 12 40 79 27 135+) 197072 Eapected 5.4394 1.1368 - 30.7 - 2405 211 528 1092 34 Lyon et at. 1976 1661 Mormon Non Mormon ratio - - 09 - - 042 051 042 076 086 -lmalesl. 1966-70 Hole Numbefs in pi&fenlhe.ses. •efer to reference list at ttw end ct ,,,,, chaplet 1 Eacns1ve alcohol cancer rNted mortality ills.a found 1n mfl11ast1num •nd thrro.d. cancer and leukemia we-re ea<. eSS1\lt' in V1rt 11~k t 'f' dmit..ers 'S11n.t1ei1nt at the 005 level. • S1an.ftant •t the 0.01 level • S11!"fteant i1t the 0001 level 'Three d111t uta1ortH 1n the E16hfh R~t"~'°"• lnl~rrytl(JIYI Clan1f1c.JIKH'I of D1sHses. Adapl!d 1965 ltCOA) Source: U.S. (1978). p. 138. IJver Pan-1551 creas 1561 6 5 31 31 2 66' 443 2 4 077 221 I-' w - - w 122 106 4 3 42 51 - 64 - 63 16 54 199 629 - 099 - 134 -report also reviews the retrospective studies. These generally provide corrobation for the evidence based on the prospective designs. The link appears to be strongest for esophageal cancer, followed closely by cancer of the pharynx and larynx. Alcohol and smoking are thought to work synergistically in these sites, but the evidence is far from unequivocal. At the very least, the relationships are non-specific -- "in Iran, an area with an unusually high incidence of esophageal cancer, Kmet and Mahboubi (1971) found alcohol consumption very rare" (U.S. 1978, 141). While we could go on at some length to look at the role of alcohol in traffic and other accidents, and the social effects of alcohol (crime, violence, family disruption, etc.), the interested reader may consult U.S. (1978) for a cursory review and extensive references on most of these areas. Instead, we conclude this section by discussing briefly the evidence bearing on fetal alcohol syndrome. Ashley (1981, 141) notes that the 1970's have witnessed the "rediscovery of the adverse effects of heavy maternal drinking on the fetus", the relationship having been noticed as far back as the writings of Aristotle. Since 1973 the syndrome, characterized by a number of physical and behavioural features (U.S., 1978, 172; Little and Streissguth, 1981), has been the subject of extensive study. Because its rediscovery was retrospective, the bulk of the research has been of the retrospective case review variety. More recently, a number of large prospective studies are beginning to generate preliminary results. The ten years - 135 -since the 'rediscovery' have been filled with population and clinical epidemiology, and studies in animals. "The accumulated data ••• indicate that the fetal alcohol syndrome is at the far end of the spectrum of possible effects of alcohol on the fetus. It has been found only in the offspring of mothers with chronic alcoholism ••• Among the offspring of women with severe chronic alcoholism, possibly a third or more will have sufficient stigmata to warrant the diagnosis of fetal alcohol syndrome. In contrast, among the offspring of those who abuse alcohol less severely, the risk appears to be much lower" (Ashley, 1981 , 141-2) • It is also worth noting that from a public policy perspective, this syndrome has to date remained almost a non-issue. As of 1980, less than 500 cases had been documented in the literature. Incidence rates appear to be from 1 in 300 to 1 in 2,000 infants (Sokol, 1981). Nevertheless, the literature is beginning to generate sufficient strength and consistency, to suggest a causal link from heavy chronic alcohol consumption by pregnant women. A limited number of the actual epidemiological studies are reviewed in Sokol (1981), and more extensive reviews may be found in U.S. (1978), and Turner et al. ( 1977a). Attempts to 'implicate' alcohol abuse in fetal impairment are hampered not only by the small numbers of infants presenting at birth with the syndrome, but also by potential maternal confounding variables -- poor nutrition, heavy smoking, drug use, and age (Turner et al., ibid). Evidence "from animal - 136 -studies is quite compelling and clearly suggests a risk for human infants when daily alcohol consumption is 3 ounces (six drinks) or more ••• 11 (U.S., 1978, 189). In summary, diseases of the liver have the strongest epidemiological links back to alcohol abuse. Risks of fatty liver, hepatitis, cirrhosis and cancer are apparently enhanced by exposure. Additionally, cancers of the esophagus, pharynx and larynx have generated epidemiological evidence suggestive of a causative relationship, and links of a causal nature between very heavy exposure and damage to the unborn fetus seem destined to be confirmed by the continued development of cohort-based evidence over the coming decade. 4.2.3 Thoughts On The Epidemiologic Link We began this chapter by reviewing the necessary role of strong and specific epidemiological evidence in the development of the informational justification for lifestyle-related economic policy initiatives. In particular, we noted that economic incentives such as user fees directed at the end stages of the putative causal chain (presentation of the disease states to the health care delivery system) would require that disease state M was always attributable to or identifiable with lifestyle choice A. The justificatory requirements for upstream incentives such as taxes (levied at the time of A) were somewhat less restrictive, requiring only significantly elevated risks of M given A. Our examination of the literature in this chapter was based on a - 137 -strategy of assessing the strongest existing epidemiologic evidence related to voluntary exposure. That strategy led us to examine putative causal relationships between cigarette smoking and lung cancer, coronary heart disease and chronic obstructive lung disease, and between alcohol abuse and cirrhosis of the liver. A number of key points coming out of this examination bear emphasis in this concluding section. First, it is not possible to separate clearly "those who do from those who do not" engage in A. That is, an individual who religiously smokes one cigarette per day is not necessarily best thought of as belonging to the 'smokers' in a smokers vs non-smokers analysis. All the epidemiological evidence reviewed above suggests that the issue of "does vs doesn't" is less important than attempting to determine exposure intensities at which risks appear to change. This search for threshold effects is illustrated neatly in Figure 11, adapted from Turner et al. (1977a). We have added to that figure, a quadrant D in which we attempt to portray the hypothetical "does vs does not" scenario. In that quadrant, of course, there is an immediate effect on health from engaging in the implicated lifestyle activity to any extent. The curve may take on a number of shapes thereafter. More realistic, apparently, are the possibilities portrayed in quadrants A and B and, for alcohol consumption, c. Clearly, the actual shape will vary with the type of activity, and the shapes will be determinable from repeated epidemiologic research. The key point from all this is that if quadrant D is not a realistic schema, then downstream (stage M) incentives would require evidence that indicated all incidents of state M were attributable to some threshold level of activity A. - 138 -Figure 11: HYPOTHETICAL RELATIONSHIPS BETWEEN LIFESTYLE ACTIVITY AND HEALTH STATUS :c t-...J < LL.I ::c z: 0 t- Neg u LL.I LL. LL. LL.I LL. 0 LL.I 0 LL.I c::: c.o L&J Pos c ::c t-...J < L&J ::c z: 0 Neg t-u L&J LL. LL. L&J LL. A c B Neg 0 Pos LEVEL OF ACTIVITY , Neg ~ D I ~ / ~, ,/ ~ / __ _, ------ -- -0 0 r----~-------- or----------------------L&J L&J c::: ~ Pos c Pos LEVEL OF ACTIVITY Source: Adapted from Turner et al. (l 977a), p.237. - 139 -But investigations of this nature are made difficult by the fact that two phenomena may be at work, and may not be separable in empirical research. The first and most obvious is that even if a threshold exists below which no one's health status is impaired due to A, it may not be that all oases of M derive from A-exposure above the threshold. In addition, however, assume all M cases do result from activity A at or above the threshold level. The second problem is then one of self-reporting bias. Stage M (downstream) incentives would require not only that we know what the threshold level is, but that we are able retrospectively to monitor individuals' exposure history so as to assign him/her to the above- or below-threshold group. The fact that thresholds may be complex functions of more than one variable (e.g. daily intake and duration of activity) simply makes even more remote the likelihood that we will ever see situations in which downstream economic incentives could be justified. Certainly the evidence reviewed in this chapter, from some of the strongest putative causative associations, is not sufficiently strong or of the nature necessary to constitute that justification. Heavy alcohol use is implicated in up to 90% of alcoholic cirrhosis cases, but problems of reporting alcoholic vs other cirrhosis, deciding what is 'heavy', and obtaining reliable drinking behaviour information all serve to confound any policy initiative. Similarly, it may turn out that severe alcohol abuse always leads to fetal alcohol syndrome. But again, where is the 'severity' threshold, and how does one go about monitoring the drinking behaviour? Even where we find the strongest smoking-related epidemiological evidence, for COLD and lung cancer, there seem to be sufficient cases not attributable to smoking as to eliminate downstream - 140 -incentives. On the other hand, some of the evidence is sufficiently suggestive of oausal relationships as to warrant our continued examination or upstream incentives. Here we might include lung cancer, COLD, and laryngeal cancer with smoking, and cirrhosis with alcohol abuse. But the shape of the curve in Figure 10 becomes particularly important in the design of an appropriate economic incentive, for achieving either an efficiency or equity objective. Assume, for example, that a smoking---> lung cancer relationship looks like quadrant B. Here the true cost of smoking behaviour depends on the extent of exposure, 13 so that achievement of either objective would require taxes on cigarettes individually tailored to each person's exposure. The mind boggles. In conclusion, then, the evidence reviewed in this chapter seems sufficient to allow us to focus our attention in the following chapter exclusively on upstream (stage A) economic incentives. Not only are we some distance from having epidemiologic evidence of a type necessary to support downstream charges or penalties, but problems of monitoring and opportunistic behaviour discussed earlier, not to mention ethics (Kramer, 1979), may in any event be severe enough to undermine any such policies even with the epidemiologic evidence in hand. - 141 -FOOTNOTES TO CHAPTER 4 - -1The absolute importance of any one criterion will, of course be reduced as one moves down its column. 2 Alternatively, one can think of a research investigation based on retrolective data collection as a "secondary analysis" (Susser, 1973). 3cohort studies are described either as "observational" methodologies (e.g., Susser, 1973), or as "sub-experimental" designs (e.g., McMaster University, 1981). 4As Feinstein (1973) notes, however, "case-control" does a poor job of conveying the notion of backward observation (from effect back to cause). For lack of a suitable label, he adopts the term "trohoc". 5Much of the empirical work of economists seems to fit rather loosely into the generic case-control method. In general, such estimation techniques as logit analysis may be likened to investigating retrospectively the causality behind an M vs. A variable, while regression analyses using continuous dependent variables are analogous to relating a continuum of M-state intensities to one or more putative causal agents. These statistical techniques are commonly applied to historical data, usually collected "retrolectively", and attempt to move in a backward direction, from effect to cause. Much effort and energy is expended on attempting to account for confounding factors, through statistical standardization. Thus, while no overt prior matching of cases and controls takes place, the economist ends up investigating the causal interaction of factors which contribute to cases being cases, and controls being controls. For the "biological consistency" criterion, read "theoretical consistency" -- an economist's results will only be as reliable as the underlying causal theory is plausible. 6 It is this combination of incidence and absolute risk in the relative risk ratio, for example, that induces Feinstein (1973) to label it an "ablative risk ratio". We are inclined to be somewhat less critical in the current lifestyles context. While it is true that relative risk ratios hide individual risks, relative risks are rarely all that are reported. In other words, one can go back and retrieve information on the magnitude of a given problem without undue strain. We feel that it is, in fact, conceptually useful in this context to consider separately the epidemiologic information requirement, and then the policy significance of the epidemiologic evidence. For the former purpose, the relative risk ratio provides a useful indication of etiological importance. - 142 -7In fact, the Report uses "coherence" where we have referred to "biological plausibility" and embodies dose-response within the strength criterion. Other criteria are directly comparable to those adopted in this Report. 8 In 1981~82 in British Columbia, acute care hospital separations with a primary diagnosis of coronary atherosclerosis were 4,436 representing about 46,000 days stay. In contrast, angina pectoris accounted for 2,377 separations and 14,200 days, while acute MI was recorded as primary diagnosis in 5,613 separations entailing 66,000 days of acute hospitalization. 9The "constitutional hypothesis" maintains that the relationship between cigarette smoking and CHD is 11 ••• largely fortuitous and that the significant relationships are between the genetic make-up of the individual and CHD and between the genetic make-up of the individual and his becoming a cigarette smoker". Thus, "people with certain constitutional make-ups are likely to develop CHD, and the same people are more likely to smoke cigarettes". (U.S., 1971, 48) 10 In B.C. in 1979, chronic bronchitis and emphysema were the listed causes of death in 452 cases, as contrasted with malignant neoplasm of the bronchus and lungs (1018), and acute MI (32199) and coronary atherosclerosis (1968), (B.C., 1982). Of course COLD may be a contributing factor in many more deaths than those in which it is identified as the principal cause. 11 rn B.C. in 1981-82, 1535 acute and extended care separations showed COLD (chronic bronchitis and emphysema) as a primary diagnosis. These separations accounted for 18,122 days of care. 12Table A6, p.216 in U.S. (1971) summarizes a number of the early socioeconomic studies. 13we are assuming, of course, that health status deterioration is directly proportional to health care costs. - 143 -CHAPTER ..2,: REASSESSING THE !!Qb! OF ECONOMIC INCENTIVES -,!!!! ECONOMIC EVIDENCE ON SMOKING 5.1 Summary and Role of Epidemiological Evidence We noted at the conclusion of the previous chapter that the epidemiologic information relating to lifestyle choices was such as to warrant the narrowing of our focus on economic incentives to those incentives which may be brought to bear at the time individuals are "exposed" to or engage in activity A. In Chapter 3 we described the assumptions which must be satisfied if efficiency or equity objectives are to be achieved. The efficiency case required first a change in lifestyle behavior as a result of the incentive. The equity case required that the incentives serve a redistributive function - to compensate those not engaging in A for the marginal collective costs attributable to behavior A. Accordingly, we devote this chapter to two tasks. The first section addresses the efficiency goal by selectively reviewing the economic literature which has attempted to establish demand elasticities - that is, the lifestyle responses of individuals to upstream disincentives or taxes. The second section is an analytical attempt to assess the current evidence relating to the equity objective. More specifically, we examine the current relationship between smoking-related insured health care costs in Ontario and smoking-related tax revenues. In both of the following sections, as the title of this chapter - 144 -suggests, we consider only smoking. This further narrowing of focus was motivated by a number of considerations related to the equity objective. Since we looked in detail at only one putative alcohol-related link, that with cirrhosis, much additional work would have been required to assemble evidence on cost attribution for the many other illnesses and conditions to which alcohol has been linked in at least an associative manner. · But even more important was our strong feeling that any attempt to analyze the economic costs of alcohol abuse would take us far beyond the major intent of this project. Since it is often argued that alcohol-related social costs are at least as significant as direct health ~costs, a major and complex separate costing project would have been required. In contrast, we felt that Chapter 4 had covered in sufficient detail the major health care problems related to smoking, and that the social costs attributable to smoking were not as significant/substantial as those attributable to alcohol. In short, we are sufficiently confident in our smoking-related information necessary to estimate the cost side of the equity issue, and sufficiently uneasy about our alcohol-related information, that we chose in this Chapter to restrict attention to smoking. As in previous chapters, our emphasis is on the development of methodological approaches and frameworks rather than exhaustive content coverage. While we do not claim that this chapter contains the final word either on the efficiency or equity-related evidence, we feel that the results are sufficiently robust to form the basis for some rather thought-provoking perspectives on the whole lifestyle issue. In addition, we believe the framework developed here is suitable for similar analyses of other linkages. The results in - 145 -section 5.3 surprised even us. 5.2 Evidence from Upstream = Elasticities of Demand for Cigarettes 5.2.1 Introduction If economic incentives are to satisfy an efficiency objective, we noted in Chapter 3 that the necessary (although as should be clear by now, not sufficient) first impact must be an appropriate lifestyle behaviour adjustment. If tripling the price of a package of cigarettes (ceteris paribus) leaves cigarette consumption unchanged, the relationship between health status and per capita health care costs (and other downstream links in the causal chain) become considerably less interesting, at least from the perspective taken here. Thus, the first assumption underlying the efficiency case for point-of-'consumption' (upstream) economic incentives is that consumers do, indeed, respond to such incentives. In this section we review briefly the evidence on elasticities of demand for cigarettes, in an attempt to assess the sensitivity of this lifestyle activity to relative prices. Unlike section 5.3 of this chapter, which develops new evidence relating to the equity objective, this section restricts itself to drawing together primary and secondary literature sources which have reported or reviewed research addressing on elasticities for cigarettes. While we felt that there was a significant gap in the available information necessary to address the question of redistribution and - 146 -equity, and that we could (within reasonable time and space) generate some such information for the purposes of this study, neither was the case for demand elasticities. There is already a considerable body of literature from which one may adduce information sufficient for our limited purpose here; and generating new evidence of this kind would have significantly lengthened (both in time and paper) an already long report. 5.2.2 General Methodology Attempts to estimate elasticities of demand for cigarettes generally follow one of two approaches - multivariate analyses or quasi-experimental studies. A multivariate analysis is a retrospective examination of the impacts of putative explanatory factors in generating inter-observational variation in demand. Observations may be countries, regions, cities, families, etc. in a cross-sectional analysis, or different points in time for a single country/region/group of families, etc. in a time-series analysis. Among the explanatory factors may be age and sex mix, socioeconomic status, ethnic mix, geographic variables, other exogenous factors such as public education campaigns, and of course, price. Price variables should represent 'real' (purchasing power) prices since, for example, a price of 2p in region X relative to region Y (price p) would not be expected to have any consumption impact if all other prices, and incomes, in region X are also twice those in region Y. Analyses of intertemporal variation will generally involve some sort of price deflation, and cross-section analyses may adjust for inter-regional - 147 -variation in the relative price of cigarettes by, for example, including real income variables (or a disposable income index) to reflect relative economic status. In fact, one might expect price elasticity to be a function of income levels - an individual's or region's demand response may well depend on income in a manner distinct from any pure income effect. That is, an income variable would represent the independent impact of income level on demand, but would not take account of the fact that price effects may themselves be functions of income. Single price and income variables impose identical price elasticities irrespective of income level, plus an independent income effect. Interdependencies may be incorporated using interactive variables which are some combination (e.g. multiplicative) of the price and income variables. Similar interdependencies are likely at the individual level, between price and absolute consumption levels (Harris, 1980). Price elasticities may be derived, directly or indirectly, from the estimated coefficient(s) of the price variables(s). The second approach, quasi-experimental, is much less often the approach of economic analysts, primarily (we presume) because of lack of familiarity with the techniques, but also because of the difficulties in building sufficient standardization into a prospective but non-randomized pseudo-experiment. The general approach is to assess prospectively the impact of price increases on community consumption patterns. Time-series analyses are plagued by measurement and statistical problems not found to the- same degree in cross-sectional studies. In particular, intertemporal product heterogeneity may undermine the most - 148 -straightforward methods of measuring oonsumption (demand). It is oommonly acknowledged that the weight of tobaoco per oigarette has declined over time (Ontario Council of Health (O.C.H.), 1982), imparting a predictable bias to use of the "number of oigar~ttes per capita or per day" as dependent variables. Filter tips, tobacco content of butts, tobacco consumption in hand-rolled cigarettes and other more or less obscure factors all tend to militate against accurate measurement of price elasticities. Furthermore, independent variable values in time-series tend often to display multicollinearity. Since there has been a relatively comprehensive recent review of the demand elasticities literature in an Ontario context (Ontario Council of Health (O.C.H.) 1982), the remainder of section 5.2 draws extensively and relatively briefly from it, supplementing equally briefly with a few more recent studies. Evidence on Demand Elasticities In addition to providing a oogent review of the factors whioh may undermine relatively straightforward attempts to estimate elasticities using time-series analysis (cigarette content and construction, hand-rolled cigarette share of market), the O.C.H. (1982} report provides a brief sUimDary of some of those past attempts. Table 15 is taken from that report, and summarizes the estimates of price elasticity reported in twelve souroes over the period 1933-80. As might be expected from economic analyses of a particular phenomenon, the twelve studies have produced almost as many distinct estimates of elasticity. From estimates of no price impact (elasticity - 149 -Table 15: VARIOUS ESTIMATES OF ELASTICITY OF DEMAND FOR CIGARETTES Year Elasticity Estimates Author Published Count!Z, Price !!!££!!!!. Method Schoenfeld 1933 U.S.A. -0.68 Time-series Tennant 1950 U.S.A. 0.00 +0.46 Time-series Lyon and Simon 1968 U.S.A. -0.51 Quasi-experimental Laughhunn and Lyon 1971 U.S.A. -0.81 +0.42 Cross-sectional Russell 1973 U.K. -0.59 +0.53 Time-series Atkinson and Skegg 1974 U.K. 0.00 Time-series Peto 1974 U.K. -o.501 Time-series Miller 1974 U.S.A. -0.40 Time-series Thompson & Macleod 1976 Canada (- +0.991 Cross-sectional (-0.75' +0.47 Time-series Sehm 1977 Finland -0.30 Time-series Leu 1979 Switzerland -0.802 +0.802 Time-series Harris 1980 U.S.A. -0.44 Cross-sectional I Various values were given in these articles. Given here are the medians of five or six estimates of elasticity. Unlike the other estimates in this Table, Leu's estimates do not appear to have been based on deflated (i.e., real) prices and incomes. The author notes that the elasticity based on real dgarette prices "never proved to be significant". Source: U.S. (1982), p.66. - 150 -of zero), the figures range as high as -0.8, implying an 8 per cent decline in demand in response to a 10 per cent price increase. The differences may be attributed to methodology (different extents of standardization for contemporaneous effects; time-series vs cross-section; different variable constructions, etc.), data (different sampling frames, geographic locations, accuracy of data, etc.), and other more or less random factors. The Council report (O.C.H. 1982) describes briefly one methodological exchange (Russell; Atkinson & Skegg; Peto), but makes no further attempt to reconcile the figures. Nor is such an attempt made here. Suffice it to note that price elasticities are likely to vary with income and cigarette consumption levels, but that these interdependencies notwithstanding, elasticities seem to range from about -0.3 to -0.8. Other more recent or unreviewed evidence provides further corroborative results. Lewit and Coate (1982) undertook a cross-sectional analysis of adult smoking habits based on data from the 1976 U.S. Health Interview Survey, and report a price elasticity of -0.42. They note that the price response manifests itself more strongly in the decision to smoke at all than in the quantity-once-smoking. The authors also provide a brief critique and review of past U.S. studies, noting in particular the upward bias to cross-sectional studies resulting from the use of sales-based dependent variables aggregated to the state level which have not taken account of inter-state differences in excise taxes. (The resulting inter-state bootlegging makes sales based on taxes paid unreliable estimates of true consumption). Those cross-sectional estimates are reported to have ranged from -0.4 to -1.3. Their analysis, using the individual as the dependent variable - 151 -observation, is intended to circumvent this type of bias, and not surprisingly finds an elasticity at the lower end of the range. The interdependencies of publicity campaigns and taxation increases are explored in more recent work by Leu (1982), in an examination of experience with concurrent policies in Switzerland. Despite the fact that one would expect demand to respond to real rather than nominal price changes, Leu found a nominal price response, and argues that it resulted from concurrent nominal price increases and a strong publicity campaign. Thus, the nominal price result is interpreted as an indirect publicity effect superimposed on the real price effect. Real price elasticities of -0.5 to -0.6 were found in Leu's time-series analyses. Perhaps the most interesting of Leu's findings is the apparent synergistic effects of price and publicity, suggesting that an unheralded tax increase plus a separate publicity campaign unaccompanied by price change, are jointly likely to produce less consumption response than concurrent equivalent price and publicity policies. Finally, Warner (1977) estimated a price elasticity of about -0.5 from a time-series analysis of U.S. cigarette consumption. While by no means either a critical or exhaustive review, this discussion suggests that (methodological and data differences aside) cigarette consumption does tend to respond to changes in real price. Considerable work appears to remain in refining elasticity estimates, differentiating between decision to smoke and quantity-once-smoking elasticities, and examining age/sex, income and quantity interdependencies. Furthermore, the temporal erosion of price effects warrants further study. But for our purposes in this report, the knowledge of a range of aggregate real price elasticities of about -0.3 - 152 -to -0.8, with expected values more likely in the -0.4 to -0.5 range, seems sufficient. We conclude section 5.2 with an examination of the implications of this evidence. 5.2.4 Discussion Almost all available evidence suggests that consumers do, indeed, respond to real increases in the price of cigarettes. While elasticities around -0.5 cannot be termed "dramatic", equally we have no basis for presuming such responses to be insignificant. The significance of any particular elasticity is ultimately beyond the purview of the economist-cum-researcher, and rests with the policy maker. A one per cent real price increase appears to have generated reductions in consumption averaging about 1/2 per cent. Responses to relatively larger price increases could significantly exceed this. For example, Lewit and Coate (1982) found a price elasticity of -0.42 based on a price variable with a range of approximately 35 cents to 60 cents. It is conceivable that larger increases might induce larger consumption effects. Or they may not. But even an elasticity in the neighbourhood of -0.4 does suggest that the first assumption underlying an efficiency case is sufficiently grounded empirically as to justify more detailed examinations of the other links in the causal chain. The epidemiological evidence reviewed earlier eliminated many putative causal relationships from consideration. Furthermore we eliminated downstream (point of service) incentives from the feasible policy set even for those conditions for which causal evidence did exist, because of the nature of that evidence. - 153 -This review of the impact of upstream incentives brings us to the point, then, of being able to argue that there are a limited set of disease conditions which appear to be causally linked to lifestyle choices, and that, among those choices, smoking at least appears to be sensitive to upstream economic incentives. We turn attention in section 5.3 to evidence related to the arguments underlying the equity justification for upstream economic incentives. 5.3 Evidence ~ the Equity Objective - Do Smokers Pay Their Way? 5.3.1 Introduction This section is an analytic attempt to assess the current evidence on the equity argument as it pertains to smoking. The analysis is undertaken by comparing smoking-related insured health care costs to the revenue derived from tobacco taxes. Before commencing the analysis, however, a brief review of the assumptions underlying the equity goal is warranted. As noted in Chapter 3, the role of economic incentives in the context of the equity objective is not to reduce health care costs, but rather to redistribute the burden of costs in accordance with the burden of illness responsible for generating those costs. Thus the first assumption which must be fulfilled in order to justify levying economic penalties is that unhealthy lifestyles indeed contribute to higher 1 health care costs • (It is not sufficient that unhealthy lifestyles - 154 -merely contribute to higher rates of utilization, as demonstrated in Chapter 3). For smoking, there is extensive evidence indicating that the assumption is valid; however, the degree to which smoking is responsible for various illnesses remains an issue for debate (as evidenced by the variety of relative risk factors cited in Chapter 4). The second assumption underlying the equity objective is that revenues from economic measures are sufficient (and are in fact transferred) to compensate people with more healthy lifestyles for the costs imposed on them by those with less healthy lifestyles. The implications of this for smokers in a system of publicly financed health care are fairly obvious. Smokers alone must provide the funds required to finance the portion of health care costs attributable to smoking; in other words, non-smokers must somehow be compensated for higher public health care costs resulting from the aforementioned illnesses. But the equity argument, strictly speaking, would require that the compensating payments are exactly sufficient to offset all smoking-related health care costs. Any payments above or beyond these costs would technically violate the equity argument as smokers would then become the disadvantaged group. That is, they would bear not only the costs of smoking-related illnesses, but also the costs of illnesses or other public services not causally linked to the behavior for which they are taxed. The purpose of section 5.3.2 is to determine the extent to which the equity objective is realized, using smoking as a case study. We assessed the evidence by examining the direct costs of insured medical and hospital services attributable to smoking compared with the excise tax revenue collected from tobacco sales in Ontario in 1978. In - 155 -principle we would like to examine all publicly financed health service expenditures attributable to smoking; however this analysis deals with only hospital and medical expenditures. This is because while data are available on smoking induced morbidity and mortality rates, it is much more difficult to quantify, say, the extent to which publicly financed expenditure on pharmaceuticals is attributable to smoking. Moreover, hospital and medical expenditures constituted over 68% of all public sector health expenditures in 1978 _(Canada, 1982a), hence the majority of public spending on health care is captured by these two items alone. While we do not feel that the individual items in the remainder are quantitatively significant, the omission will nonetheless be addressed in a sensitivity analysis later in this section. We also chose to limit the analysis to direct health care costs only. The rationale for this follows from the nature of the issue which we seek to examine; the equity argument is concerned with assessing health costs incurred from smoking, relative to the revenue which it generates. Consequently, items such as the loss of property resulting from fires caused by smoking, or decreased productivity due to smoking-related morbidity, do not rightly belong in the analysis because they have no impact (at least no direct impact) on publicly financed health care expenditures. It should be emphasized at this point that the present analysis is not an attempt to estimate the social costs of smoking or to evaluate the costs and benefits of the activity; in fact a variety of studies have already been undertaken with that goal in mind. (See, for example, Shillington (1977), Thompson and Forbes (1982), or Collishaw and Myers (1983). Rather, ours is essentially a financial analysis - 156 -of the impact of smoking on health services expenditures by government. The Canadian system of public health insurance was designed to protect the individual from the consequences of uncertain activities. Given the putative oausal link between smoking and several illnesses, one surely cannot regard the consequences of the activity as wholly uncertain. As such, health care precipitated by smoking should not qualify as a fully insured benefit - hence the need to address the question 'do smokers pay their way'? Methodology Employed in the Estimation of Smoking-Related Health Care Costs and Tobacco Tax Revenue -- -- -- -- ---The calculation of direct health care costs attributable to smoking and the revenue derived from tobacco taxes would appear to be a straightforward exercise. Unfortunately the inherent limitations of the currently available data restrict the methodologic options available for use. Ideally, publicly financed health care costs for each of the smoking-related diseases would be multiplied by the proportion of the disease attributable to smoking to derive a direct cost figure. However, neither health care costs by disease, nor attribution ratios, are directly obtainable from existing data sources. Taxation data for both provincial revenues and the Ontario contribution to federal revenues is also required; only the former, however, is directly available from existing data sources. A series of calculations and estimations are thus required to transform the available data into the desired information. The following analysis proceeds in five stages: first, the proportions of each of the smoking-related diseases directly - 157 -attributable to smoking are calculated using both morbidity and mortality data. Second, health care utilization attributable to each of the disease categories is combined with data from the first stage to determine the amount of utilization attributable to smoking. Next, the associated health care costs (both hospital and medical) directly 2 resulting from smoking are computed. The fourth stage then calculates the total contribution of Ontario smokers to both federal and provincial tobacco tax revenues. Finally, the estimated tax revenue is compared to the aforementioned health care costs. Statistics Canada provincial data for the year 1978 were used whenever possible. i) Calculation of Proportions Attributable The methodology employed in the first three stages of the analysis closely follows that used by Shillington (1977) in his national study on the economic consequences of smoking. Several modifications were made, however, in order to incorporate certain additional data relevant to the present study. Wherever these modifications appear, they are flagged as such. Thus, unless otherwise indicated, the reader is referred to Shillington for a detailed exposition of the data sources and methods employed in stages i - iii. Following Shillington, four smoking-related diseases were selected for analysis (based on the strength of their putative causal link and/or on the policy significance of the disease) and defined in terms of the eighth revision ICDA codes3 as follows: Disease Category ICDA-8 Code Lung Cancer 162, 163.0 - 158 -Coronary Heart Disease Bronchitis Emphysema 410, 411, 412, 414 490, 491 492 Estimates of the attributability of smoking to each of these disease categories were calculated using both morbidity and mortality ratios since it is not always obvious, as we noted in Chapter 4, which of the two alternative measures provides a more accurate indicator of health care utilization. Shillington suggests that costs due to physician services should be apportioned according to morbidity ratios while costs resulting from hospital care should be calculated using mortality rates, since the latter takes into account the greater severity of illness among smokers. He concedes however, that the issue remains sufficiently nebulous to warrant the inclusion of both measures in the analysis. The morbidity and mortality ratios by disease are shown in Tables 16 and 17 respectively. The morbidity rates for lung cancer were taken from Kohn and White (1966), for coronary heart disease from the U.S. Department of Health, Education and Welfare Framingham Study (1973), and for bronchitis and emphysema from the U.S. Health Interview Study (1965). Similar ratios for mortality were extracted from Kohn's analysis of the Dorn study (1966). (All data are summarized in Shillington). A standard formula for calculating proportions attributable4 was then applied to the morbidity and mortality ratios to derive the percentages in Tables 16 and 17. While we prefer not to rely on any one source for the above data, there are, as Chapter 4 indicates, almost as many estimates of relative - 159 -Table 16: MORBIDITY RATIOS AND PERCENTAGE ATTRIBUTABLE TO SMOKING Lung Coronary Heart Bronchitis and Cancer Disease Emphysema Morb. % 11orb. % Morb. % Ratio Attrib. Ratio Attrib. Ratio Attrib. Male --15-24 -.,': -..': -.': '" 2.2 34.7 25-34 -;': ..,·: -.': )~ 2.2 39,5 35-44 3,9 61.2 )~ 'I': 2.2 39,5 45-54 3.9 59.3 1. 7 26.0 2.2 37,6 55-64 3.9 59,3 1. 3 13.2 2.2 37.6 65 + 3,9 47.2 1. 2 5. 8 2.2 27.0 Female 15-24 -;': ;': '" o.o 2.5 33.4 25-34 -;': ;': -;': 0.0 2.5 3 7. 1 35-44 2.9 42.8 ,., 0.0 2.5 37. 1 45-54 2.9 37.2 . 92 o.o 2.5 31. 9 55-64 2.9 37.2 .Bo o.o 2.5 31. 9 65 + 2.9 16.6 1. 10 o.o 2.5 13.6 *No data available. Source: Shillington (1977) - 160 -Table 17: MORTALITY RATIOS AND PERCENTAGE ATTRIBUTABLE TO SMOKING Lung Coronary Cancer Heart Disease Bronchitis Emphysema Mort. % Mort. % Mort. % Mort. % Ratio Attrib. Ratio Attrib. Ratio Attrib. Ratio Attrib. Ma 1 e --< 35 - o.o - o.o - 0.0 - 0.0 35-39 ;I: Bo.a 1. 7 27.6 5.0 68.5 7,9 79.0 40-44 -.': 80.0 6. 1 73.5 5.0 68.5 7.9 79.0 45-49 4.3 62.4 3.0 50. 1 5.0 66.8 7,9 77.6 50-54 5.6 69.8 2.7 46. 1 5.0 66.8 7,9 77.6 55-59 13.4 86.2 1.8 28.7 5.0 66.8 7.9 77.6 60-64 14.6 87.2 1 • 5 20. 1 5.0 66.8 9.0 80. 1 65 + 8.5 69.8 1. 3 8.5 5.3 57.0 1 1. 3 76.0 Female Mort. % Ratio Attrib. 15-34 - 0.0 - o.o - o.o 35-39 2.8 41.5 - o.o - 0.0 40-44 2.8 41. 5 - o.o - 0.0 45-49 2.8 36.0 2.0 23.8 4.9 54.9 50-54 2.8 36.0 2.0 23.8 4.9 54.9 55-59 1. 9 21.9 1. 7 17.9 4.9 54.9 60-64 1. 9 21. 9 1. 7 17.9 4.9 54.9 65 + 1. 9 8.6 1.4 4.o 7.5 40.6 Source: Shillington (1977) - 161 -risk and percents attributable as there are studies to determine them. Thus we have decided to use the numbers from Shillington and to employ sensitivity analysis to assess the impact on the calculations of 5 alternative estimates. ii) Calculation of Hospital Utilization Attributable to Smoking Although we must obtain utilization rates for both hospital and ambulatory care attributable to smoking, only the former can be derived directly from Ontario data. In Ontario, visits to a physician are not disaggregated along a disease-specific, or even episodic, basis. Consequently, a different methodology was employed to calculate the smoking-related costs of physician services, the discussion of which is deferred to the next section. In order to estimate the amount of hospital utilization attributable to smoking, we first identified the number of 6 patient days associated with each of the smoking-related diseases. To be used with the data on proportions attributable this information must also be differentiated on an age-sex basis. However, the age categories used in reporting the proportions attributable differ for both the morbidity and mortality estimates. Thus two separate age-sex breakdowns of patient days are required. The 1978 data for Ontario are presented in Table 18. The number of patient days in each age-sex category were then multiplied by the associated percentage attributable for that category to determine the number of patient days attributable to smoking. This information is shown in Table 19. - 162 -Table 18: PATIENT DAYS BY AGE AND SEX, ONTARIO, 1978 (Age-Sex Categories Required for Use with Morbidity Data) Coronary Hurt Bronchitis and Lung Cancer Disease Emphysema ~•e 15-2'4 189 230 '405 25-3'4 361 3,641 87'4 35-4'4 1 ,401 20,403 943 '45-5'4 18,6'40 85, 173 7,633 55-6'4 18,6'40 85,173 7,633 65+ '47,'480 160, 1'48 32,359 !!!!!!!.! 15-2'4 0 185 823 25-3'4 211 4'46 1,058 35-'4'4 I, 177 '4,'499 1,560 '45-5'4 6,912 31,851 '4,663 55-6'4 6,912 31,851 '4,663 65+ 12,679 179,911 16,252 Age-Sex Cateqorles Required for Use with Mortality Data Coronary Heart Bronchitis and Lung Cancer Disease Emphysema Male Bronchitis Emehysema <35 581 3,938 79'4 '485 35-39 701 10,202 36'4 109 '40-'4'4 701 10,202 36'4 109 '45-'49 9,320 '42,586 2,0'43 I, 77'4 50-5'4 ,,320 '42,586 2,0'43 1,77'4 55-59 9,320 '42,586 2,0'43 1, 77'4 60-6'4 9,320 '42,586 2,0'43 1,77'4 65+ '47,'480 160, 1'48 15,227 17, 132 Female 15-3'4 211 631 I ,881 35-39 589 2,251 781 '40-'4'4 SB~ 2,251 781 '45-'49 3,'456 15,927 2,331 S0-5'4 3,'456 15,927 2,331 55-59 3,'456 15,927 2,331 60-6'4 3,'4!i6 15,927 2;331 65+ 12,679 179,911 16,252 ~: Statistics Canada, unpublished data. Note: Statistics Canada does not provide as detailed an age breakdown as Is required for some of the age categories Identified In the morbidity and mortality percentages attributable. Therefore when necessary, the ni.nber of patient days In the broader Statistics Canada age categories were divided equally among the age groups required according to the morbidity and mortality data. - 163 -Table 19: PATIENT DAYS ATTRIBUTABLE TO SMOKING, BY AGE AND SEX, ONTARIO, 1978 Source: Tables 16, 17 and 18. Calculated Using Morbidity Data: Coronary Heart Lung Cancer Disease Ha le IS-24 - -25-34 - -35-44 857 -45-54 11,054 22,14S 55-64 1 I ,OS4 11,243 6S+ 22,411 9,289 TOTAL: 4S,376 42,677 !.!.'!!ili IS-24 - -2S-34 - -3S-44 soi. -45-54 2,571 -SS-64 2,S71 -65+ 2, IDS ---TOTAL: 7.7SI 0 Calculated Using Mortality Data: Coronary .Heart Lung Cancer Disease Hale --<3S - -3S-39 561 2,816 4o-44 S61 7,498 45-49 5,81~ 21,336 so-sit 6,505 19,632 55-S9 8,034 12,222 60-64 8, 127 8,560 6S+ 33,141 13,613 TOTAL: 62,745 8S,677 ~ lS-34 - -3S-39 244 -40-44 244 -45-49 1,244 3, 791 so-sit 1,244 3,791 55-59 757 2,851 60-64 757 2,851 65+ 1,090 7, 196 TOTAL: S,S80 20,480 Bronch It Is and Emphysema 14 I 345 372 2,870 2,870 8,737 IS,33S 275 393 579 1,487 1,487 2,210 6,431 Bronchi t Is and Emphysema Bronc:hltls Emphysema - -249 86 249 86 1,36S 1,377 1,365 1,377 1,365 1,377 1,36S 1,421 8,679 13,020 14,637 18,744 ---1,280 1,280 1,280 1,280 6,598 11,118 - 164 -iii) Calculation of Smoking-Related Health Care Costs Publicly financed direct health care costs attributable to smoking were calculated aa the sum of the ooat of' physician services and hospital care attributable to smoking7. As a different methodology was employed to compute each of the above costs, a separate presentation is warranted for each. Smoking-Related Hospital Costs The cost of hospital care was calculated as the product of the number of patient days attributable to smoking in each disease category and the per diem hospital rate for that disease, summed over the four disease categories. The number of smoking-related patient days was computed in the previous section; but the derivation of disease-specific per diem rates required some additional calculations. Hospital per-diem costs are not available on a disease-specific basis; rather, an aggregate number representing the gross operating cost per day for all diseases is provided by the Ministry of Health. That figure for Ontario in 1978/79 was $165.06 (Ontario 1979). It is possible, however, that smoking-related diseases may be, in general, more or less costly to treat than other inpatient diseases. We therefore used disease-specific marginal inpatient per diems from Barer (1982) to determine the extent to which the costs of smoking-related diseases deviate from the average. Since the data in Barer all pertain to marginal and not average costs, we were unable to use the actual figures. Instead, we used the marginal costs to derive a measure of the cost of smoking-related diseases relative to the average cost for all - 165 -diseases. The following results were obtained: DISEASE CATEGORY Lung Cancer Coronary Heart Disease Bronchitis Emphysema DISEASE-SPECIFIC MARGINAL COST PER DAY AS A PROPORTION OF NON-DISEASE SPECIFIC MARGINAL COST PER DAY 1.14 .89 .68 .Bo The interpretation of these results is that the per-diem treatment costs for lung cancer run an estimated 14% higher than the mean per diem treatment cost for all diseases. 8 The results with respect to the other diseases can be similarly interpreted. 9 The per diem rate of $165.06 was then adjusted using the relative costs of the diseases as shown above. The resulting disease-specific per diems are as follows: Lung Cancer Coronary Heart Disease Bronchitis Emphysema $188.17 146.90 112.24 132.05 The adjusted per diem rates are likely overestimates of the true costs of treating the smoking-related diseases, primarily because they 10 include approximately 15% of costs not attributable to inpatient care - 166 -Thus by using the per diem measure in our calculations we are likely upward-biasing the inpatient cost of care for these diseases. But that potential upward bias will materialize only if the disease-specific portions of non-inpatient activity are different from the inpatient days mix. In other words, we are implicitly assuming that the disease-related activity proportions are the same for non-inpatient as for inpatient activities. Reassuringly, we have no evidence which leads us to believe otherwise. 11 Having calculated the disease-specific per diem rates , we then multiplied these rates by the corresponding number of patient days to derive the total hospital costs attributable to smoking. The results are presented in Table 20. Smoking-Related Physician Costs Due to the limited amount of data available on ambulatory care, a different methodology was required to calculate the cost of physician services. Despite their limitations, data gathered by the Saskatchewan Medical Care Commission provide the cost per patient, by ICDA-8, as well as the cost of services per 1,000 population. Since the data pertain exclusively to costs and utilization in Saskatchewan, several adjustments were required to take account of differentials in fee schedules and utilization patterns between the two provinces. The calculations are provided in Table 21. First, the total cost per 1,000 population for Saskatchewan was adjusted to account for the fee schedule differential between Saskatchewan and Ontario, using an adjustment factor of 13.71% - the difference in fees between the two provinces in 1971 as calculated by Table 20: COST OF HOSPITAL CARE ATTRIBUTABLE TO SMOKING CALCULATED USING MORBIDITY AND MORTALITY DATA 1978, ONTARIO Patient Days Hospital Attributable Attr i outab 1 e At tr i bu tab 1 e to Per Diem Cost ($'000) Cost ($ '000) Smoking ($) Using Using Morbidity Est. Mortality Est. Morbidity Data Mortality Data Lung M 45,376 62,745 188. 17 8,538 11 '807 Cancer F 7,751 5.,580 188.17 1,459 1,050 Coronary M 42,677 85,677 146.90 6,269 12,586 Heart Disease F 0 20,480 146.90 0 3,009 Bronchitis H 15,335 33,381 122.15 1 ,873 4,077 & Emphysema F 6,431 11,718 122.15 786 1 ,431 TOTAL: 18,925 33,960 Source: Table 19 and text. I ...... CJ' --.J Total Cost, Saskatchewan, 1971 1 ($/000 population) Coronary Lung Heart Cancer Disease Bronchitis Emphysema Hale 15-2'4 .oo 2.6'4 8.59 7 .2} 25-3'4 .00 212.72 32.17 9.68 35-ltlt '48.75 212.72 32.17 9.613 '45-5'4 11'4. 78 2102.70 215.76 173.69 55-61t 1t62.19 2102.70 2i5.76 173.69 65+ 642.05 1t829.55 596.21 648 . 81t ~ 15-21t .oo 6.61 7.20 J.60 25-31t .oo 61t.61t 29.52 9.06 35-ltlt 19. llt 61t.61t 29.52 9.06 lt5-5lt lt8.75 687.27 61t.03 31t.61t 55-6'4 51. 51 687.27 61t.03 31t.61t 65+ I04.lt9 3897.52 115.10 88.90 lsource: Sh1111ngton (1977) Table 21 COST or PHYSICIAN SERVICES, ONTARIO 1978 Estimated Total Cost, Ontario, i n I 971 do I I a rs ( S ' 000) Coronary Lunq Heart r.ancer DisE>ase Bronchitis F.mphysema 0 2 9 8 0 179 27 9 JO 131 20 7 66 1216 125 . 101 203 923 94 76 262 1968 2~3 264 0 7 8 3 0 55 25 8 12 39 18 5 28 397 37 20 25 328 30 16 59 2205 65 so Estimated Total Cost, Ontario In 1978 dollars ($ 'OOO) Coronary Lunq Heart Cancer Cllsease Bronchitis 0 2. 7 12.3 0 244.2 36.8 40.9 178.7 27.3 90.0 1658.6 170.5 276.9 1259.0 128.2 357.4 2684.4 331.5 0 9.5 10.9 0 75.0 34. l 16.4 53.2 24.6 38.2 541.5 50.5 34. l 447.4 40.9 80.5 3007 .6 88.7 Emphysema 10.9 12.3 9.5 137 .8 103.7 360. l 4.1 10.9 6:8 27.3 21.8 68.2 ...... 0\ 00 - 169 -Shillington. Next, the costs were further inflated to reflect the inter-provincial difference in real per capita utilization. This difference was estimated by comparing the 1971 per capita public sector health expenditures for both provinces. Saskatohewan spent $208.79 per person while Ontario spent $255.42, or 22.3% more (Canada, 1979). As mentioned above, 13.71% of this difference can be attributed to higher fee schedules in Ontario; the remaining 8.6% is an estimate of the difference in real per capita utilization. Thus the 1971 Saskatchewan physician costs per thousand population were augmented by a total of 22.3% to yield the estimated cost per thousand population for Ontario in 1971. These estimates were in turn multiplied by the total Ontario population in the corresponding age/sex groups to derive the estimated total cost of physician services for Ontario in 1971 dollars. Next, the costs in 1971 dollars were adjusted using the Ontario physician fee index from Wolfson, Evans, and Lomas (1980) (1971 = 100, 1978 = 136.4) to yield the estimated total cost in 1978 dollars. Finally, total costs from Table 21 were multiplied by the appropriate percentage attributable in each age/sex/diagnostic category to produce the estimated costs of physician services attributable to smoking as presented in Table 22. Hospital care costs from Table 20 and physician services costs from Table 22 were then summed to determine the total (hospital and physician) direct health care cost attributable to smoking. This information is summarized in Table 23. - 170 -Table 22 COST Or PHYSICIAN SERVICES ATTRIBUTABLE TO SMOKING, ONTARIO 1978 DOLLARS ($'000) Attributable Cost Using Morbidity Data: Coronary Heart Bronchitis and Lung Cancer Disease Emphysema Ka le 15-24 - - 8.0 25-34 - - 19.4 35-44 25. l - 14.5 45-54 53.4 431 .2 115. 9 55-64 164.2 166.2 87.2 65+ 168. 6 155. 7 186.8 -- --- --TOTAL: 411 . 3 753 . l 431 .8 ~ 15-24 - - 5.0 25-34 - - 16. 7 35-44 7.0 - 11. 7 45-54 14.2 - 24.8 55-64 l 2. 6 - 20.0 65+ 13. 3 - 21.4 -- --- --TOTAL : 47 .1 0 99.6 Attributable Cost Using Mortality Data: Coronary Heart Bronchitis and Lung Cancer Disease Emphysema Bronchitis Emphysema Hale 15-24 - - - -25-34 - - - -35-44 32.7 90.4 18.6 7.5 45-54 59.5 797 .8 113. 9 l 06.9 55-64 240. l 307.2 85.6 81.8 65+ 249.5 228.2 188.9 273.7 -- -- -- --TOTAL: 581 .8 1423.6 407 .0 469. 9 remale ---15-24 - - -25-34 - - -35-44 6.8 - -45-54 13.7 128.9 15.0 55-64 7.4 80. l 12 .o +65 7.0 120.3 27.7 -- -- --TOTAL: 34.9 329.3 54.7 Source : Tables 16, 17 and 21 - 171 -Table 23 TOTAL DIRECT COSTS (HOSPITAL CARE AND PHYSICIAN SERVICES) ATTRIBUTABLE TO SMOKING, ONTARIO 1978 (S'OOO) Total Cost Estimated from Morbidity Data: Hospital Physician Total Costs Servlce ·Costs Costs Lung Cancer M 8,538 411 8,949 F 1 'i,59 47 1,506 Coronary M 6,269 753 7,022 Heart Disease F 0 0 0 Bronchitis M 1'873 432 2,305 & Emphysema F 786 100 886 TOTAL: 18,925 1'743 20,668 Total Cost Estimated from Mortality Data: Hospi ta1 Physician Total Costs Service Costs Costs M 11'807 582 12,389 Lung Cancer F 1, 050 35 1,085 Coronary M 12,586 1,424 14 ,01 o Heart Disease F 3,009 329 3,338 Bronchitis M 4,077 8.77 4,954 & Emphysema F 1'431 55 1,486 TOTAL: 33960 3,302 37,262 Source: Tables 20 and 22 - 172 -iv) Calculation of Revenue Derived from Tobacco Taxes Tobacco excise taxes are levied at both the federal and provincial levels. Although the present analysis is concerned with only those data pertaining to Ontario, it is important to recognize that a significant portion of federal tobacco tax revenue is generated in Ontario; to omit the contribution of Ontario residents to federal tobacco excise taxes would therefore constitute a gross underestimate of the true value of tobacco tax revenue. Data on the provincial tax revenue collected is easily obtainable; 12 for the fiscal year ending March 31, 1979 the Ontario figure was $258,569,000 (Canada, 1982a). The Ontario contribution to federal revenues is more difficult to secure, however. The federal government collected $710,579,000 nationally from tobacco taxes in 1978/79 (Canada, 1981a). Unfortunately there is no simple way of determining the portion of these revenues originating in Ontario; estimation of the Ontario contribution .was therefore necessary. Two types of specific taxes were levied on tobacco products by the federal government in 1978/7913 : an excise tax of $.03 per 5 cigarettes and excise duty of $5.00 per thousand cigarettes (Canadian Tax Foundation, 1979), resulting in a total tax of $11.00 per thousand cigarettes. Although it might be feasible to calculate the portion of the total federal tax collected in each province, this figure would be of little use since the taxes are collected at the manufacturer or distributor level, not at the point of purchase. Consequently the estimates of provincial contributions would be biased in favour of - 173 -Ontario and Quebec where the majority of manufacturers are located. We therefore had to base our estimates on data regarding average per smoker consumption and the number of smokers in the population. Two alternative approaches to the estimation were employed. Method A: This method uses the number of regular cigarette smokers in Ontario relative to that for Canada as a whole to approximate the Ontario contribution to federal revenues. In 1981, Ontario had 2,000,916 regular cigarette smokers or 33.2i of the total national number of 6,023,140 (Millar, 1983). 14 If it is assumed that the per capita consumption of cigarettes in Ontario is no different from the national average 15 , then the ratio of smokers in Ontario to Canada as a whole may be used to approximate the relative contribution of Ontario residents to federal tobacco tax revenue. Accordingly, it is estimated that $235,912,000 or 33.2i of the total federal revenues of $710,579,000 were generated by Ontario residents. Method B: This second method uses the estimated daily per smoker consumption of cigarettes to calculate federal tobacco excise taxes collected. In 1979, the average smoker in Canada consumed 28.2 cigarettes daily (ibid., Table 48), or 10,293 cigarettes each year. In 16 Ontario alone there were in excess of 2,000,000 regular smokers , contributing to a total of 20,586,ooo,ooo cigarettes consumed annually. The federal government collected, in excise taxes and duties, $11.00 per thousand cigarettes manufactured. This results in an estimated contribution of $226,446,000 from Ontario residents to federal tobacco tax revenues. The estimates derived using the two alternative methods of - 174 -calculation are fairly similar in magnitude. It should be remembered, though, that these estimates are conservative because they exclude the 12~ general manufacturers' sales tax which is levied on all tobacco products (see footnote 13). v) Comparison of Smoking Related Health Care Costs to Revenue Derived from Tobacco Taxes The results derived in the previous sections are summarized in Table 24. Smoking-related health care costs computed using both morbidity and mortality data are presented along with provincial and federal tobacco tax revenue. It quickly becomes evident that the health care costs attributable to smokers are more than offset by the tax revenue they generate. Regardless of the method of estimation, smoking-related costs never exceed ten percent of total tobacco tax receipts. Even if these costs are compared to provincial revenue alone, the resulting ratio remains below fifteen percent. The sensitivity of these estimates to variations in the underlying assumptions used to derive them is considered in the next section. Sensitivity Analysis Many of the estimates employed in the above analysis hinge on specific sets of underlying assumptions. While we believe those assumptions to be generally valid, we nonetheless tested the stability of the results to variations in the data and the assumptions. Sensitivity analysis is a necessary component of any research involving uncertainty or imprecision in the data, or in which there - 175 -Table 24 COMPARISON OF ESTIMATED SMOKING RELATED HEALTH CARE COSTS TO REVENUE DERIVED FROM TOBACCO TAXES, ONTARIO , 1978 Smoking related health care costs: morbidity estimate mortality estimate Provincial revenue from tobacco taxes Estimated contribution of Ontario residents to federal revenue from tobacco taxes TOTAL revenue from tobacco taxes METHOD A 235,912,000 494,481,000 20,668,000 37,262,000 258,569,000 METHOD B 226,446,000 485,015,000 Smoking related health care costs as a percentage of TOTAL tobacco tax revenue: Morbidity est., Method A 4.2% Morbidity est., Method B 4.3% Mortality est., Method A 7.5% Morta 1 i ty est., Method B 7.7% Smoking related health care costs as a percentage of PROVINCIAL tobacco tax revenue: Morbidity est., 8.0% Morta 1 i ty est., 14. '4% - 176 -is controversy over the methodologic approaches employed. Using the above criteria, we chose three areas for testing: the percentages attributable, the range of smoking-related diseases and the range of publicly financed health care costs. Percentage Attributable As mentioned earlier, we were uneasy about some of the percentages attributable employed by Shillington; specifically, we were concerned by the number of age/sex cells for which data were not available and with the zero entries for the percentage attributable (using morbidity data) for the female coronary heart disease category. Accordingly, for hospital costs we used the maximum percentage attributable in each disease/sex category to re-estimate the number of patient days attributable to smoking for each of the age cells in that category. As well, the percentage attributable cited for male coronary heart disease was substituted for the zero entries in the female category. Thus the following figures were used to derive alternative high scenario estimates of the health care costs attributable to smoking: Percentage of Care Costs Attributable to Smoking - Selected Illnesses Calculated Using Morbidity Ratios: Lung Cancer Coronary Heart Bronchitis Emphysema Disease Male 61.2% 26.0 39.5 39.5 Female 42.8 26.0 37.1 37.1 - 177 -Calculated Using Mortality Ratios: Lung Cancer Coronary Heart Bronchitis Emphysema Disease Male 87.2% 73.5 68.5 80.1 Female 41.5 23.8 54.9 54.9 These percentages are unequivocally overestimates of the true figures and generally lie outside the range implied in Chapter 4. Notwithstanding the upward bias in subsequent calculations imparted by these extreme estimates, the health care costs attributable to smoking were recalculated using the new percentage estimates. The calculations are presented in Tables 25 to 29 which correspond in format to Tables 19, 20, 22, 23 and 24. The results in Table 29 indicate that even though smoking-related health care costs as a percentage of total tobacco tax revenue doubles using the alternative estimates of proportions attributable, the costs nonetheless remain below 16% of the revenues. If the contribution of Ontario residents to federal taxes are excluded, the costs still remain below 30% of the provincial revenues. The original results are thus not dependent upon the estimates of percentages attributable employed in the calculations. Expansion of the Range of Smoking-Related Diseases The smoking-related diseases identified in the analysis were selected based on the strength of their putative causal link. We recognize, however, that at least part of the burden of some additional diseases can be attributed to smoking17 • Unfortunately most of the epidemiologic evidence supporting this assertion does not meet the methodologic standards which are required to establish a conclusive - 178 -Table 25: PATIENT DAYS ATTRIBUTABLE TO SMOKING, BY AGE AND SEX, ONTARIO, 1978 (Calculated Using Upper Estimates of Percentage Attributable) Using Morbidity Date: Coronary Heart Bronchitis and Lung Cancer DI se11e Emphysema ~ 15-24 116 60 160 25-34 221 947 345 35-44 857 5,305 372 45-54 11,408 22, 145 3,015 55-64 11,408 22,145 3,015 65+ 29,058 41,638 12,782 TOTAL: 53,068 92,240 19,689 Female 15-24 - 48 305 25-34 90 116 393 35-44 504 I, 170 579 45-54 2,958 8,281 1,730 55-64 2,958 8,281 1,730 65+ 5,427 46,777 6,029 TOTAL: 11,937 64,673 10,766 Using Mortality Data: Coronary Heart Bronchitis and Lung Cancer DI sease Emphysema Bronchitis Emehysema Male <35 507 2,894 544 388 35-39 611 7,498 249 87 4o-4lt 611 7,498 249 87 45-49 8, 127 31,301 1,399 1,421 so-sit 8,127 31,301 1,399 1,421 SS-59 8, 127 31,301 1,399 1,421 60-64 8, 127 31,301 1,399 1,421 65+ 41,403 117' 709 I0,430 13,723 TOTAL: 75,640 260,803 17,068 19,969 Female 15-34 88 lSO 1,033 35-39 244 S36 429 40-44 244 S36 429 45-49 1,434 3,791 1,280 so-sit 1,434 3,791 1,280 55-59 1,434 3,791 1,280 60-64 1,434 3.791 1,280 65+ 5,262 42,819 8,922 TOTAL: 11,574 59,205 lS,!33 Source: Tables 16, 17 and 18. Table 26: COST OF HOSPITAL CARE ATTRIBUTABLE TO SMOKING CALCULATED USING UPPER ESTIMATES FROM MORBIDITY AND MORTALITY DATA 1978, ONTARIO Patient Days Hospital Attributable Attributable to Per Diem Cost ($ 1 000) Smoking ($) Using Morbidity Est. Morta 1 i ty Est. Mor bi di ty Data Lung M 53,068 75,640 188. 17 9,986 Cancer F 11,937 11 '574 188.17 2,246 Coronary M 92,240 260,803 146.90 13,551 Heart Disease F 64,673 59,205 ll16. 90 9,500 Bronchi t Is M 19,689 37,037 122.15 2,405 & Emphysema F 10,766 15,933 122. 15 1 '315 TOTAL: 39,003 Source: Table 25 and text. A tt r i bu tab 1 e Cost ($'000) Using Morta 1 i ty Data 14,233 2' 178 38,312 8,697 4,524 1 ,946 69,890 ....... -...J \0 - 180 -Table 27 COST nF 'HYSICIAN SERVICES ATlRIBUTABLE TD S"OKING, ONTARIO 1978 DOLLARS (S'ODO) Attrlbut1ble Cost U'lng Upper £1tl1111te1 of Morbidity D1t1: Coronuy Heart lronchltls ind Lung C1ncer Dlsuu Emphyse1111 ~ 15·2lt - 1.0 9.1 25·3'4 - 63.5 19.4 35·4'4 25 .1 46.4 14.5 '45·5'4 55.1 431 .2 121.8 55·6'4 169.5 327.3 91.6 65+ 218.7 698.0 273.2 - - --TOTAL: 468.4 1566.4 529.6 ~ I 5-2'4 - 2.4 5.6 25·34 - 19. 5 16. 7 35·'4'4 7.0 13.8 11. 7 '45-54 16.4 140.8 28.8 55·64 14.6 116.3 23.3 65+ 34.4 781.9 58.2 -- --TOTAL: 72.4 '1074.7 144.3 Attributable Cost Using Upper Estimates of Mortality Data: Coronary Heart 8ronchiti1 ind Lung Cancer Disease Emphysema Bronchitis Emphi::seMa Male 15-24 - 2 .1 8.4 8.7 25-34 - 179.5 25.3 9.8 35-44 35.7 131 .4 18 .6 7.6 45-54 78.5 1219. 1 116.8 110.3 55-64 241.4 925.3 87.8 83.0 65 + 311.6 1973. 0 227.0 288.5 -- -- --TOTAL: 667.2 4430.4 483.9 507 .9 Fema 1 e 15-24 - 2.3 8.2 25-34 - 17 .8 24.7 35-44 6.8 12.6 17.2 45-54 15. 9 128.9 42.6 55-64 14 .1 106.5 34.5 65 + 33.4 715.8 86.1 -- -- --TOTAL: 70 .2 983.9 213.3 Source: Table 21 and text - 181 -Table 28 TOTAL DIRECT COST (HOSPITAL CARE AND PHYSICIAN SERVICES) ATTRIBUTABLE TO SMOKING, ONTARIO 1978 (S'OOO) Total Cost Estimated Using Upper Values from Morbidity Data: Hospital Physician Total Costs Service Costs Costs Lung Cancer M 9,986 468 10,454 F 2,246 72 2,318 Coronary M 13,551 l '566 15,117 Heart Disease F 9,500 l ,075 l 0, 57 5 Bronchitis M 2,405 530 2,935 & Emphysema F 1, 315 144 l ,459 TOTAL: 39,003 3,855 f 42,858] Total Cost Estimated Using Upper Values from Mortality Data: Hosp i ta 1 Physician Total Costs Service Costs Costs Lung Cancer M 14,233 667 14,900 F 2, 178 70 2,248 Coronary M 38,312 4,430 42,742 Heart Disease F 8,697 984 9,681 Bronchitis M 4,524 992 5, 516 & Emphysema F 1,946 213 2' 159 TOTAL: 69,890 7,356 f 77 ,246 I Source: Tables 26 and 27 - 182 -Table 29 COMPARISON OF UPPER ESTIMATES OF SMOKING RELATED HEALTH CARE COSTS TO REVENUE DERIVED FROM TOBACCO TAXES, ONTARIO, 1978 Smoking related health care costs: morbidity estimate mortality estimate Provincial revenue from tobacco taxes Estimated contribution of Ontario residents to federal revenue from tobacco taxes TOTAL revenue from tobacco taxes METHOD A 235,912,000 494,481,000 42,858,000 77' 246, 000 258,569,000 METHOD B 226,446,000 485,015,000 Smoking related health care costs as a percentage of TOTAL tobacco tax revenue: Morbidity est., Mo r b i d i t y es t. , Mo rt a 1 i t y es t. , Mortality est., Method A Method B Method A Method B 8. 7% 8.8% 15 .6 % 15 .9 % Smoking related health care costs as a percentage of PROVINCIAL tobacco tax revenue: Mor bi d i t y est. , Mor ta 1 i t y e s t. , 16.6% 29.9% - 183 -causal link. As well, research is lacking on a variety of morbid conditions which may be peripherally linked to smoking. Therefore rather than attempting to identify the range of diseases for which smoking is potentially a contributing factor, we instead chose to determine how much more expenditure on smoking-induced health care would be required to support an equity case for further taxation. Our most liberal estimate of smoking-related health care costs is $37,330,000, or 7.5J of total tobacco tax revenue. Costs would therefore have to be increased thirteen-fold before they would exceed revenues. It is difficult to envision a range of diseases which could reasonably be included in the analysis so as to augment the costs by thirteen times the original estimates. Even if costs are compared to provincial revenue alone, the required increase would be in excess of six times the original amount. It seems highly unlikely that the diseases not identified in the present analysis could be responsible for such large increases. Expansion of the Range of Health Care Costs The health care costs considered in the analysis are limited to insured hospital and medical care. In principle at least, smoking-related diseases may be responsible for public expenditures on other types of health services (e.g. drugs, appliances, public health, health research, etc.). Since there is no method of determining with even reasonable accuracy the proportion of such expenditures attributable to smoking, we chose to perform an extreme sensitivity analysis by attributing all remaining public expenditures on health care to smoking-related diseases. This is obviously a gross overestimate of non-- 184 -hospital and medical costs attributable to smoking. Specifically, in 1978, Canadian public sector expenditures on general/allied special hospitals and physicians constituted 68% of all public sector health care expenditures (Canada, 1982b). The cost estimate in Table 24 (mortality data basis, since these are highest) was therefore increased to reflect the assumption that, for Ontario, all of the remaining 32% of expenditures on other services were made for smoking-related diseases. The resulting cost was $54,875,000 or 11% of total revenue from tobacco taxes (21% of the corresponding provincial revenue). Thus even with an overexpansion of the range of health care costs included in the analysis, the costs attributable to smoking-related diseases remain well below the revenue generated from tobacco taxes. The sensitivity analysis could be extended (almost indefinitely) to include variations in other data used for the estimates. But there should be a valid reason for subjecting variables to sensitivity testing; items such as the per diem calculations or the number of disease-specific patient days may be questioned, the former because of well-known problems with (even diagnosis-specific) per diems as estimators of hospital resource costs, the latter because of inaccuracies in diagnostic coding, multiple simultaneous diagnoses, etc. We chose not to undertake sensitivity testing of these variables because we felt they were sufficiently accurate, and the changes required to alter our basic results so extensive, as not to warrant the effort involved. Discussion of Results The sensitivity analysis serves to verify our original findings: - 185 -on equity grounds, smokers already pay their way. Although this exercise was approached as a positive, rather than normative, financial analysis of the impact of smoking on health service costs (relative to tobacco excise tax revenue generated by the activity) we attempted, wherever possible, to avoid imparting any bias by ensuring that we overestimated costs and underestimated revenue. Nonetheless, costs never exceeded fifteen percent of revenues. The policy significance of these results is twofold. First, if government targets smoking as a lifestyle which warrants modification, they should not appeal to the equity objective to justify their actions. A strong case may be made either for protecting non-smokers from the harmful and unpleasant consequences of smoking (external effects), or for protecting smokers against the consequences of their own acts (paternalism). But in the narrow financial setting, smokers are 'pulling their weight'. Second, governments (both federal and provincial) should be aware that if they are successful in reducing smoking, the impact on the health care system is not as predictable as it may appear to be. Given the paramount importance of the supply side of the health care delivery system in determining utilization, and the inevitable "Law of Competing Risks", that people who do not die of smoking will die of, and be cared for, something else, there is no reason to assume that a decline in the smoking-related diseases will necessarily be translated into a corresponding decline in overall utilization. 18 In conclusion, it is important to emphasize that we are in no way encouraging smoking. There are many excellent reasons for public policy - 186 -to discourage smoking, but financial reasons are not among them. our results demonstrate that responses to contentious issues, which· may appear intuitively obvious to some, are often not empirically supportable. 5.4 Economic Incentives Reconsidered The case for economic incentives targeted at lifestyle choices has been argued to be two cases - an efficiency case, and an equity case. The former argues that current patterns of health care consumption exceed what one would find in a world where rational, fully informed individuals personally bore lifestyle-related marginal health care costs. Furthermore, that argument is critically dependent not only on the lifestyle -----> illness linkage, but on behavioural responses to those economic incentives. That is, if there are disease states linked to lifestyle choices, but consumers do not respond to the proposed economic incentives, the efficiency case for those incentives disintegrates. After all, if prices are adjusted to reflect marginal lifestyle-related costs, and consumers do not correspondingly adjust their behaviour, then in fact the putative over-allocation of resources to health care, attributed to lifestyles, is non-existent. Having individuals bear the marginal health care costs of their choices would not reduce health care costs. The fact that some segment of health care costs may be attributed to lifestyle choices would remain. But if private risk-bearing reveals no allocative inefficiency, the case for economic incentives based on efficiency arguments evaporates. That does not imply, of course, that the equity justification disappears with it. - 187 -The equity case does not require a consumer response to the incentives. Rather, it is an argument that smokers should compensate non-smokers for the marginal costs which fall (through taxation-supported health care) to the non-smokers, irrespective of whether the vehicle for compensation causes a behavioural response among smokers. The revenue raised through the incentives may be used to compensate the · non-smokers. But both cases do depend rather critically on evidence of causal relationships between lifestyle behaviours and on the allocation of resources by society to the health care sector. The selective epidemiological evidence reviewed in Chapter 4 revealed a limited set of behaviour -----> disease l~nkages of a causal nature. Even for this limited set, the epidemiological evidence was never sufficiently strong to justify downstream incentives such as de-insurance or user fees. It was, however, sufficiently. strong to support the equity case, given appropriately constructed incentives, and the efficiency case pending evidence on behavioural responses to those incentives. With that as background, this chapter set about two tasks. The first was to assess the behavioural response evidence pertaining to the efficiency case for smoking. As we summarized in section 5.2.4, the evidence on price elasticities suggests that consumers do, indeed, respond to price changes by reducing cigarette consumption. Unfortunately as we pointed out in Chapter 3, the road from there to reduced health care costs is rather rocky. Nevertheless, both the economic and epidemiological evidence support the efficiency case for upstream incentives directed at cigarette smoking. The precise nature - 188 -of those incentives requires more precise information on attributable costs. The second objective of this chapter was not only to estimate those costs attributable to smoking (which information could be used in structuring upstream incentives for both efficiency and equity cases), but more broadly to assess the current status of the equity case. If, as was suggested by Chapter 4, there are diseases which are causally linkable back to lifestyle decisions taken by some subset of the population, and if to those diseases may be attributed marginal health care costs, then the equity case for incentives argues that smokers should compensate non-smokers for the proportion of attributable costs borne (through public funding) by the non-smokers. The obvious question this raised in the Canadian world of publicly financed universal health insurance and privately borne tobacco taxes, was the relationship between current taxation levels and current health care costs attributable to smoking. Section 5.3 attempted to draw together data germane to both cost attribution and taxation revenue. We showed that even under a set of assumptions overtly charitable to the equity case (i.e. over-estimates of health care costs attributable and under-estimates of cigarette-related tax revenues), smokers appear already to be more than paying their way. This does not undermine the equity case in principle, but indicates that taxes currently in place are more than adequate. Therefore, the findings of the study to this point could be swmnarized as follows: (i) there appears to be a valid efficiency case for upstream dis-incentives directed at cigarette smoking; - 189 -(ii) such incentives probably serve to reduce health care costs attributable to smoking, but may or may not reduce aggregate health care costs; (iii) there is also a valid case in equity for upstream charges for smoking -- tobaooo taxes; (iv) the current Ontario and federal tax structure already imposes charges more than sufficient to meet both equity and efficiency objectives; (v) there are, however, strong non-financial arguments -- external effects and paternalistic interests -- for public programs to discourage smoking; (vi) as a component of such programs, further increases in tobacco taxes would both lower smoking levels and increase tax revenues, while successful smoking cessation programs would lower both smoking levels and tax revenues. Thus, from the perspective of pub~ic budgeting, there may be a good case for a combined strategy. Anti-smoking campaigns legitimate tax increases; tax increases protect public revenues from the consequences of successful cessation programs. To this point in the study, we have been proceeding with analyses built on a particular underlying theory of lifestyles - that choices such as smoking and drinking are just that, lifestyle choices. That assumption is, for example, essential to the equity argument for economic incentives which is based on principles of voluntary generation of financial externalities by those choosing particular lifestyle attributes. Similarly, the efficiency argument rests on assumptions of rational consumers choosing more of particular lifestyle attributes because they face less than the 'full' price consequences of those choices. We have pushed that traditional framework about as far as it can be pushed, and our results depend critically on it. It seems prudent, - 190 -then, to step back briefly to re-examine the assumptions inherent in that framework. - 191 -FOOTNOTES TO CHAPTER 2 1As noted in Chapter 3, the converse of this is that economic "rewards" should be provided to compensate people who choose relatively healthier lifestyles than average. 2In calculating the impact of smoking behaviour on health care costs, we have not calculated any cost savings due to the premature death of smokers. As noted above, insofar as smokers die earlier from lung cancer, CHD, or other illnesses, they do not generate costs of care for other forms of morbidity and mortality. By failing to account for these costs, we have overstated the total impact of smoking on health costs generated by smokers. On the other hand, we have also neglected the indirect health costs generated by "passive smoking" of non-smokers. 3The ninth revision of the ICDA Codes was not used because it was introduced on April 1, 1979; our data all pertain to 1978, however, and thus the eighth revision is more appropriate for the present purposes. 4 See Shillington, op.cit., Appendix c, for the formula used to calculate attributable risk and the assumptions underlying it. 5we were especially concerned with the number of cells for which data are unavailable and with the zero entries for the morbidity ratios associated with coronary heart disease in females. We have addressed these deficiencies in two ways: first, employing mortality ratios in lieu of morbidity ratios provides additional information, as well as non-zero entries for the above category. Second, the sensitivity analysis uses alternative (non-zero) estimates of the percents attributable when testing the robustness of the results. 6The number of patient days is a commonly accepted measure of utilization and is appximately equal to the number of admissions multiplied by the average length of stay. 7we recognize that there might be some non-medical, non-hospital publicly financed health care costs that get excluded from our current measure of "health care costs". In our sensitivity analysis we have therefore expanded costs to include more than just insured hospital and medical services. The reader is referred to section 5.3.3 for the results. 8This is assuming, of course, that the differences between disease-specific and non-disease specific marginal costs closely parallel the associated differences in average costs. - 192 -9The reader may find the value of 0.89 for coronary heart disease --11% less costly than the "average" day of stay -- disturbingly inconsistent with images of high technology coronary care units. These values were calculated from 1969-74 Ontario data, and it may be that the spread of CCU facilities in the last decade has raised the CHD value. But as the reader can check, a doubling of this value (to 1.78) would still leave the conclusions secure. 10see Barer, op.cit. 11 For the purposes of subsequent calculations, the separate per diem rates for bronchitis and emphysema were aggregated for an average rate of $122.15. 12 Ideally, we would want tax revenue calculated on a calender year basis, since that is the accounting period used to calculate health care costs attributable to smoking. Tax revenues, however, are reported on a fiscal year basis ending March 31. We chose to use data from the fiscal year 1978/79 because we felt they best approximated revenues collected in the calender year 1978. 13The federal government also levies an "ad valorem" tax on the sale of tobacco products. The general manufacturers sales tax on tobacco was 12% of the selling price in 1978/79. However since revenues collected from general sales taxes are not disaggregated by commodity, it is impossible to determine what proportion of total sales tax receipts were attributable solely to sales tax on tobacco products. 14 This figure, broken down on a provincial basis, is available for 1981 only. Although we would prefer 1978 data, we do not feel that utilizing data from 1981 biases the estimates since it is unlikely that the geographic distribution of smokers changed significantly between 1978 and 1981. 15 Although five provinces have a higher percentage of regular cigarette smokers in the population than Ontario (Millar, op.cit.), we can find no evidence to indicate any inter-provincial differences in the per smoker consumption of cigarettes. 16In 1981 there were 2,000,916 regular smokers in Ontario, or 33.1% of the male population and 27.2% of the female population. In 1979, 34.4% of the male population and 27.9% of the female population were regular smokers. (Millar, op.cit., Tables 3 and 4). Thus there were more (in percentage terms) smokers in the population in 1979 than in 1981. Given that the population has increased over the period 1979-81 and the number of smokers has decreased, the 1981 figure of 2,000,916 smokers is probably a good (albeit rough) approximation of the 1979 figure. - 193 -17 It is important to emphasize that although we have identified the smoking-related diseases as being lung cancer, coronary heart disease, bronchitis and emphysema, these are very broad categories which actually encompass several diseases. For example our definition of coronary heart disease includes acute myocardial in~arotion, other acute and subacute forms of ischaemic heart disease, chronic ischaemic heart disease (including aneurysm of the heart) and asymptomatic ischaemic heart disease. Similarly, lung cancer includes malignant neoplasms of the trachea, bronchus, lung and pleura. 18rf the reduction in prevalence of smoking were achieved by increased cigarette taxes, however, the estimates of a relatively low price elasticity of demand indicate that total tax revenue could be increased (at least beyond current levels) at the same time. The burden of these taxes would, however, fall disproportionately on lower-income people. Chapter .§.: - 194 -The Not-So-Sovereign Consumer - Alternative Perspectives on "Voluntary Choice" 11If the smoker's behavior is the result of toilet training rather than rational decision-making, then to blame the smoker for the toilet training seems odd". Veatch (1980, 52) The first five chapters of this report have been rooted in the assumptions of the economic theory of the consumer. Lifestyles, or the constellations of activities, goods, and services of which they are composed, have been viewed as objects of choice, which independent, rational consumers select in a process of comparison of their respective costs and benefits. Preferences over different lifestyles and their effects which consumers bring to this choice process are assumed determined independently of the analysis, "immaculately conceived" in Boulding's term. The role of economic incentives in such a world is to shift the balance of perceived costs and benefits so as to lead the consumer to choose more healthy lifestyles, or to compensate for financial externalities imposed by the choices of one set of consumers on another. This behavioral theory, while fundamental to much of economic analysis, may however be open to question in the context of both the individual components which constitute a lifestyle, and the more general and loosely defined composite that we label "lifestyles". It seems, as Veatch (1980) argues, at least an incomplete, even if not actually incorrect, view of the processes which underly the development of individual lifestyles. Accordingly, this chapter is devoted to a - 195 -relatively brief review of several competing and more general behavioral views of lifestyle 'choice' which move away from the emphasis on conscious moment by moment choice by sovereign, rational individuals. Integrated into that review is a consideration of the implications of such alternative views for economic policy incentives directed at lifestyles. 6.1 Alternative Perspectives on Economic Choice=~ Simple Framework In a recent article on ethical issues raised by the too hasty acceptance of the •voluntary health risks' view of the world, Veatch (1980) offers five possible models of health risk determination -- the voluntary model, the medical model, the psychological model, the social structural model, and a multicausal model. While his emphasis is on health risks, and ours in this chapter is on lifestyle causality, we show in the following discussion that his uni-dimensional models or health risks serve to confuse two distinct sets of hypotheses. Whether or not these distinct sets are in fact interdependent depends on the biological and epidemiological evidence. To illustrate, consider his "voluntary model". Being a model of health risks ("the model that considers the individual as personally responsible for his health" (p.51)), it must combine a voluntary model of lifestyle choice with epidemiologic evidence linking lifestyles (however determined) with health status. In effect, this model represents a single cell in a two-dimensional framework - one dimension representing models of lifestyle determination, the other, models of health status determination. This alternative framework is illustrated in Figure 12, where Veatch's FIGURE 12: MODELS OF THE LIFESTYLES-+ HEALTH STATUS LINKAGES MODELS OF HEALTH STATUS DETERMINATION MEDICAL/GERM GENETIC LIFESTYLE ENVIRONMENTAL VOLUNTARY CHOICE VCM VCG VCL VLE MODELS OF PHYSIOLOGICAL PHM PHG PHL PHE ' LIFESTYLE PSYCHOLOGICAL PSM PSG PSL PSE DETERMINATION SOCIAL STRUCTURAL SSM SSG SSL SSE ECONOMIC/ POLITICAL EPM EPG EPL EPE MuL T 1_FACTOR1 AL MFM MFG MFL MFE MULTI CAUSAL VCMC PHMC PSMC SSMC EPMC MFMC ...... \0 (J'\ - 197 -(1980) voluntary model of health risks may be portrayed as cell VCL. When the determination of health risks is viewed in its two contributory dimensions, the intermix of lifestyle and health status models represented by Veatoh's taxonomy becomes evident. As noted above, his voluntary model is really a combination of a voluntary choice lifestyle model with a lifestyle-determined health status model. The "medical model" is actually an entire column of our framework; that is, it is uni-dimensional -- lifestyle determination is, under this model, irrelevant to health risks. Similarly, his social structural model seems to be a loose gathering of evidence relating social class to health status, and is therefore apparently intended as a model of health status determination. We return to this 'model' below. On the other hand, Veatch 1s psychological models concentrate exclusively on the other (vertical) dimension representing models of lifestyle determination. Any of the health status determination models may be superimposed on either the psychological or voluntary choice models, and the most appropriate overlay is an empirical question. When Veatch considers the possibility that "so-called voluntary health risks [might be] psychologically determined" (p. 52), for example, he is implicitly settling in the PSL cell of Figure 12. That is, he is assuming as given the lifestyle -----> health status linkage, and then concentrating on 1 psychological determinants of lifestyle • We do not mean to imply, of course, either that Veatch's framework is without value, or that Veatch was the first and last author to muddle the two dimensions underlying the relationships between lifestyles and disease or health status· causality. In an examination of the political implications of alternative theories of chronic disease causality, Tesh - 198 -(1981, 379) correctly notes that "the lifestyle hypothesis dismisses with a wave of a hand most environmental toxins and it ignores the crucial connection between individual behavior and social norms and rewards". This, too, combines theories of disease causality (lifestyle vs environmental) and lifestyle causality (individual vs social structural). But the social impact on individual choice is irrelevant if environment is the sole determinant of disease and, conversely, a focus on social determinants of lifestyle implies irrelevance of the environmental disease theories. The point of this discussion is merely to offer a framework within which to separate two types of theories. Our focus in the rest of this chapter will be on the lifestyle column of Figure 12, and this framework should clarify the manner and extent to which our focus differs from those of Tesh and Veatch. That is, we accept for the moment that sufficient evidence exists to implicate certain lifestyle characteristics in health status deterioration, and then consider the implications of alternative models of lifestyle determination. In fact, we restrict our attention even further, since the entire report to this point has assumed "voluntary choice". Thus, we now turn to cells PHL, PSL, SSL, EPL and MFL of Figure 12. These may not exhaust the possible rows in that matrix, but are more than sufficient to illustrate the implications of movement away from assuming voluntary choice. - 199 -6.2 Implications of Alternative Views of Lifestyle Determination Physiological and Psychological Models The most obvious counter-factual to voluntary choice would be provided by evidence that certain lifestyle characteristics are physiologically determined. While addiction to certain chemical substances is commonly regarded as being part physiological and part psychological (and even part social), and varies in degree with different substances (Jaffe and Kanzler, 1979) and individual metabolic tolerances, we may usefully illustrate the implications of physiological and psychological lifestyle determination using addiction. In fact, the exact categorization of addiction (as a physiological or psychological phenomenon) is relatively unimportant, as we will see as this discussion develops, because the implications of each view for economic incentives (even if not for broader social policy (Tesh, 1981)) are relatively similar, and because addiction is often offered as evidence against the voluntary choice view of the world without thought to its being but one example of a class of models. Addiction can be thought of as an extreme form of 'learning', or endogenous taste formation -- past consumption or lifestyle choices affect present preferences. But addiction has additional content. The non-addict may learn to enjoy certain activities, or certain commodities may be an "acquired taste." Having acquired such tastes the individual feels better off for having done so and would not wish to go back to his/her former self. But the addict, while also now having a greater - 200 -"taste" for the substance or activity of addiction, feels worse off and usually expresses a wish that the physiological or psychological process could be reversed. If able to exhibit voluntary choice, the addict wouJ.d choose reversal. In a world of perfect information, and taking a long-term view, this may not matter. The cigarette smoker, addicted to nicotine, may be relatively (not completely) insensitive to economic incentives. But in this hypothetical world people choose to take up addicting activities in full knowledge of the consequences. They choose to become addicted, after balancing the costs and benefits thereof. Thus economic disincentives to unhealthy but addictive activities, like smoking, will have a long-term effect as they discourage people from becoming addicted. In the short run, the disincentives have no efficiency impact, but, over time, the addicts die off, and in the long term the economic incentives are not weakened by the phenomenon of addiction. The equity argument, however, is a bit more dubious. Those who "chose", in this analysis, to become addicted prior to the introduction of the economic incentives will now have difficulty reacting. Depending on the strength of the addiction, they may be permanently penalized. The argument that those whose earlier "voluntary" (but now difficult or impossible to reverse) lifestyle choices impose costs on others, should pay compensation, becomes difficult to defend. Those now addicted were not informed of all the costs at the time they 11chose 11 addiction. Requiring compensation changes the rules of a game from which the addict cannot easily extricate himself. But the "perfect information" world is rather dubious as well. In - 201 -the case of cigarette smoking, addiction during childhood is common and, in the case of female children, increasingly so. In these circumstances, to speak of a "rational, fully informed" choice to become addicted seems indefensible. Rather it appears that parts of lifestyle patterns are formed early in life, often without informed (or even conscious) choice, and when such patterns are addictive, either physiologically or psychologically, they may be very difficult to change later by choice. In such circumstances, economic incentives intended to stimulate "individual responsibility" and to shift conscious and deliberate lifestyle choices which are not in fact being made, may be a good deal less effective than programs to assist in breaking addictions or changing habit patterns. In addition, since the equity argument seems solidly grounded in the concept of personal responsibility for the marginal effects of one's choices on the collective other, that argument also collapses if the marginal costs are not spinoffs of personal utility maximization by fully-informed, rational consumers. It would be a peculiar social conscience, indeed, which insisted that those addicted to alcohol because of psychological problems dating back to childhood, should bear the marginal costs of treatment for the health implications of that alcoholism. There may be a case for restitution, but one must be careful that one's information is sufficient to identify unequivocally the party(ies) responsible. This brings us to the next set of models. Without having established that lifestyle characteristics are or are not determined by physiological or psychological factors, we can conclude that validation of either model would undermine such economic incentives as are still - 202 -under consideration. Social Structural Models2 A second apparent major problem with the consumer choice model of lifestyles is that it isolates such choice from its social context. This suppresses two important linkages, one influencing choice-making behavior, the other resulting from it: Social Structural 1---> I Individual I ---> I External I Factors I I 'Choice' I I Effects I The positive or negative effects of lifestyle choice may impinge on other members of society (my drunken driving risks your life and limb; my cigarette smoking pollutes your air and injures your lungs) in ways additional to the pure financial externalities which an equity objective attempts to address. These are the standard problems of "external effects" - if my behavior has a negative (positive) effect on you, and you cannot impose these costs (confer these benefits) on me, then I will over (under)-engage in such behavior relative to some more general social optimum. Economic incentives are of course one way of responding to such externalities - tax (subsidize) the behavior in question and compensate (charge) the victim (beneficiary). But such externality-pricing systems often have great difficulty in practical application, and as a result one frequently sees prohibition (mandation) of harmful (beneficial) activity. How, after all, does one compensate the drunken driver's victim? But additional to these widely studied external effects, of which, - 203 -as noted above, health care program expenditures are just one form, it is important to recognize the impact of collective behavior on individual choices. To be the only smoker in a group, or the only jogger in town, is much less attractive than to be one of many. If most employees exercise regularly, the employer may provide convenient facilities inducing even more employees to partake. In Amsterdam, where much of the population cycles, there are convenient dedicated bicycle lanes with their own traffic control. In Toronto, the cyclist takes life in hand. The set of phenomena of which the above are examples, have been labelled by Schelling (1978) as "critical mass" behaviors characterized by "the way people's behavior depends on how many are behaving a particular way, or how much they are behaving that way" (p.94). The phenomena, as expressed diagrammatically by Schelling (pp. 104-106), are adapted to our lifestyles context in Figure 13. Suppose that the willingness of individuals to engage in an activity depends (ceteris paribus) on their perceptions of the proportion of the rest of the population who engage in it. There will presumably be some rugged individualists (point A) who will each "do their own thing" even if they believe no one else ·is doing it. As the perceived percentage of participators increases, a few more people become willing to participate, so the curve rises from A to B. As the activity becomes more and more socially acceptable, people are drawn in at a more rapid rate (B - C - D). But eventually increases in participation slacken off as the more socially responsive have been drawn in, until at E we have a different group of "rugged - 204 -FIGURE 13: CRITICAL MASS BEHAVIORS AS DETERMINANTS OF LIFESTYLE* % OF POPULATION WILLING TO ENGAGE IN AN ACTIVITY % 100 80 __,;.D._ ____ .. 1 E I 0--------------------------------------i.-... % OF POPULATION PERCEIVED AS AS ENGAGING IN AN ACTIVITY 100% *Adapted from Schelling (1978, 104) - 205 -individualists" left - those who will not participate even if (they think) everyone else in society is doing so. Points B, c, and D are important, however, as equilibria - at each of them the proportion perceived to participate just equals the actual number willing to do so. Thus if perceptions are accurate, the participation rate, once steady at B, c, or D, will stay there. But B and D are stable, in that a small fluctuation in perceived participation rates away from either will generate a move back to equilibrium. Say, for example, that the population, having been at point B, for one exogenous reason or another feels that (B+£)% of the population now smokes. Then (B+E-')%, (where O<S<£) will actually end up smoking, causing a subsequent downward revision in perception of proportion smoking. This in turn will lead to a further decrement back toward B in the number engaging in the activity, and so on. In contrast, point C is an unstable equilibrium -- if perceived participation increases beyond c, then a rise in actual behavior will be triggered which, when perceived, stimulates further increases ••• until Dis reached. On the downside, a drop from C stabilizes at B. Of course such a curve need not have three equilibria. In fact, it need not be stable at all, in the sense of portraying a critical mass phenomenon -- the only "equilibrium" for a curve which lies under the diagonal in its entirety, is at 0%. Examples of this subclass of critical mass phenomena are Schelling's (1978) "dying seminars", and Akerlof's (1970) "lemons" as applied to used cars, or voluntary insurance markets (see, for e.g. Rothschild and Stiglitz (1976) and Evans (1983). In each case, the social or market interaction "collapses" to an "equilibrium" where no one participates. More generally, the number of - 206 -non-zero/non-100% equilibria will be determined by the number of intersections of the curve with the 45 degree diagonal. So long as the individualists at A and E exist, and everyone else's activity depends on perceptions, then at least one equilibrium point must exist. This view of lifestyle behavior has a considerable degree of plausibility. The idea of stable and unstable equilibria has been expressed by Bass (1978) in the anti-smoking context as "health epidemics." If most people smoke, smoking reduction is very difficult and backsliding among quitters in a smoking community is the norm. But if enough people can be convinced to quit, then quitting may take hold as a dominant behavior. Thus if one is at point D - most people smoke -cessation campaigns toward C tend to fall back to D. But if one could get the rate of smoking down past C, then group behavior would carry one all the way to B, - a "healthy epidemic." Similarly a large enough group of cyclists may create enough political pressure to get proper bike paths - at which point cycling will increase further. Schelling (1978) proposes divide and conquer strategies to move smoking behavior, for example, from D down past C. Say, for example, we were able to segregate the upper and lower halves of the population willing to smoke, by willingness to engage in the activity. The upper half then are those who would be most easily induced to quit smoking (i.e. the half in Figure 13 who require the highest % of others to be perceived to be engaged in the activity before they themselves will participate.) But inducing a health epidemic among this upper half is relatively easy, for the influence of each in their segregated half has doubled. That is, the perceptions of those in the upper half have not changed, but the curve for that half may lie entirely below a 22.5 degree "diagonal". - 207 -Onoe that half is 'conquered', similar incremental partitions may facilitate a rapid move down below c. Of course this all depends in praotioe on being able to shift the social frame of reference for sub-groups of the population. This more extended view of lifestyle choice does not rule out economic incentives ~ priori. Increasing the cost of the activity in Figure 13 could move the whole curve down, for example, by reducing the willing participators at every perceived level. A large enough shift in the curve could bring points D and C together, causing the "healthy epidemic" drop to B, and really prohibitive taxes could bring A down to O so that ABCDE lay below the 45 degree line throughout its length and only O was an equilibrium. But the emphasis on the social framework within which lifestyles emerge or are "chosen" leads one to consider other mechanisms for change as complementary to, or substituting for, economic incentives. The creation of smoke-free areas in public buildings or transportation not only responds directly to the health concerns of non-smokers, it also serves as a continuing reminder of the community concern with smoking. The unhealthy lifestyle is labelled as socially undesirable, equivalent to moving downward the ABCDE link in Figure 13, and lowering the levels of smoking behavior perceived by smokers or potential smokers in the community around them. Just as marketers of a product (including cigarettes) show it as being used by people whom the potential buyer is believed to respect, emulate, or identify with, so the restriction of smoking in public places may serve to reduce the number of others seen to smoke. - 208 -The Participaction program similarly represents an attempt to enhance the respectability of exercise and fitness, and thus modify individual behavior through revising perceptions of what "everybody else" is doing. Regardless of how one judges the success of that particular program (and one gets the sense that a few well-marketed "respected bodies" such as Jane Fonda and Victoria Principal have significantly hastened the plodding successes of Participaction), it seems fairly clear that within the last decade peoples' perceptions of the extent to which other people exercise, and public attitudes in general towards fitness, have undergone a dramatic change which strongly reinforces changes taking place in individual priorities. Economic incentives in the form of public (or private) support for various forms of exercise and game facilities may also be important, but it is unlikely that similar effects could have been achieved by subsidies for sports equipment, bonuses for fitness levels, or charges for coronary care units. The change in public attitudes toward particular lifestyles seems to be the critical change, backed up by either advertising (Participaction) or advertising and regulation (smoking). Seatbelt use has also been promoted by advertising and regulation -fines for non-use represent another form of "economic incentive" although they are not really a "price" charged for non-compliance. Experience in various jurisdictions suggests that legislation is more effective than education in promoting seat-belt use, but it may be that the "social disapproval" expressed in legislation and embodied in the police is a good deal more effective than economic deterrence alone. The possibility that failure to fasten seat belts may constitute contributory negligence in accident cases, thus impairing a victim's - 209 -ability to recover damages, could in theory represent a very powerful economic sanction favouring use, and in principle this could be extended to charging unbelted victims for part of their medical and hospital care. It is doubtful if this could be enforced after the fact, however, still less could care be denied those unable to pay. One might imagine recovery from the victim's insurer, which then would encourage differential premiums for "belters" and "non-belters", but monitoring and enforcement problems probably rule this out. On the whole, energetically enforced legislation seems a good deal more practical in this area than economic incentives, particularly at the "downstream" end. This serves to distinguish the social from the physiological and psychological models in their apparent implications for economic incentives. Whereas physiological and psychological determination of lifestyle calls into question the justification and/or targeting of such incentives, the social view is a different sort of departure from voluntary choice. The implications of that departure are not that economic incentives are difficult to sustain as a policy, but rather that there are probably other far more effective policies available! Some may be most effective in conjunction with economic dis-incentives, others not. But the social view seems a less marked departure from voluntary choice (modified voluntarism), therefore leaving some scope for independent or conjunct roles for economic incentives. Economic/Political Models The social context approach to lifestyles emphasizes the extent to - 210 -whioh one person's behavior depends on what everyone else is doing, either because of some sort of "herd instinct" or because there are scale economies in lifestyles - the cyclist effect. But there is another class of "modified voluntarism" models, which explores the origins of individuals' preferences rather than the peer influences on those preferences once developed. In other words, voluntarism may be affected in more than one way. While preference modification and peer activity perception are often mingled in advertising campaigns, they are conceptually separate. "Epidemic" effects arise from the latter, not the former, for example. This malleability of preferences gives rise to what we have labelled the economic/political models, but which may less charitably be coined the "villains" theories of lifestyle choice. People with unhealthy lifestyles are led astray by the advertising of large corporations which profit from those lifestyles. On the other side, "health education" can be thought of as merely one form of public relations campaign to encourage health behavior or discourage consumption of certain products. This viewpoint represents a step beyond the traditional epidemiological approach to illness. Health can be promoted by destroying pathogenic agents, or encouraging hosts to behave in ways which increase their resistance. But in general one does not expect "the bugs" to engage in political activity to cripple campaigns against them, or to retain public relations firms to encourage hosts to behave in ways which make them susceptible to illness. Yet in the lifestyle case, this is of course what we find. The role of the pathogen is frequently played by - 211 -the supplier of the product (tobacco, alcohol, fatty or low-nutrition foods) whose use constitutes the unhealthy lifestyle. Suppliers engage in overt choice modification through large advertising budgets, and lobby (particularly in the U.S.) to deflect any public policy measures which would undermine their collective bottom line (Pinney, 1979; Sapolsky, 1980; Fielding, 1978; Goodman, 1981). Where a product can be associated with a healthy lifestyle, of course, we find its producers acting as reverse pathogens, promoting healthy lifestyles in the interest of profit. The point is that the market mechanism as a moulder of tastes is neutral with respect to health effects. But since lifestyles correlate with product consumption patterns, any attempt to modify lifestyles is a threat to someone's sales. Thus we find the absurdity of simultaneous advertising expenditures by the private sector to sell cigarettes and by the public sector to discourage smoking - efforts which in aggregate have to imply wasted resources. A more effective public policy of legislating against cigarette advertising, for example, would not only remove a major source of choice modification, but could also free up those public resources currently devoted to stop smoking campaigns. It would not, of course, be popular with advertising agencies or the media, who are at present able to sell their services to both sides. More absurd still is the public subsidization of those private tax deductible advertising expenses. What quickly becomes clear is that the industry "villains" have a number of willing accomplices in both public and private sectors. From this viewpoint, economic incentives for healthy lifestyles, particularly at the downstream end, point up an obvious hypocrisy. If one argues that the linkage from smoking to lung cancer is sufficiently - 212 -well established that smokers should pay for their own treatment - since they brought on themselves an entirely predictable illness - then how can one justify continuing permission to tobacco companies to advertise cigarettes? If such advertising is legal, one must assume either that it is ineffective in promoting smoking (in which case why does it continue?) or that the smoking behavior it promotes is not so tightly linked, causally, to morbidity and mortality as to constitute incitement to self-mutilation or suicide. The latter is presumably the justification - but then as noted earlier in this report the case for downstream economic incentives disappears. But what of upstream incentives? One finds in this case a seeming conflict of interest. Raising taxes on an activity or product while simultaneously encouraging that activity (through tax deductibility of advertising expenses, or agricultural assistance to tobacco farmers) not only undermines the efficiency case for the economic 'incentives', but calls into serious question the targeting of the incentives suggested by the equity case. If the "accomplices" are implicated in the marginal health care costs, raising the necessary revenue from those encouraged by the "villains" can only be justified (in a peculiarly perverse manner) as an attempt to outrun the health care costs with tax revenues. Such policies have little to do with equity or efficiency. The search for "villains" is important because it re-emphasizes that lifestyle formation is a social process, and that attempts to improve health through lifestyle modification threaten powerful economic interests in a way which the eradication of smallpox, say, does not. Policies to modify lifestyles, if they are to be effective, must - 213 -recognize and deal with these interests. Economic incentives based on theories of individual responsibility do not respond to these issues. And while Departments of Health within governments tend to see lifestyle issues strictly in terms of health, Departments of Agriculture or Industry have different views. In fact a voluntary choice as opposed to "villains" view of lifestyles may be used to justify an institutional configuration which, for example, separates health from agriculture (Tesh, 1981). In some parts of Canada, and more extensively in the U.S., tobacco is not a health hazard but a cash crop. A serious public program to reduce smoking would need to recognize the interdependent interests of Health and Agriculture, thereby explicitly loosening the grip on the voluntary choice view of lifestyles. An effective policy might require a complementary program of compensation for tobacco growers. In fact a reasonable long-term approach might simply be for governments to buy up, over time, all the current and future (appropriately discounted) production quotas of the Tobacco Marketing Board at their market prices. Thus production would be terminated voluntarily. We return to this and other policies in the final chapter. And so the search for "villains" which led to the discovery of "accomplices" rapidly leads one toward the conclusion that the "accomplices" may in fact be the "villains". When we observe that unhealthy lifestyles are difficult to change because powerful groups profit from, and thus promote, such lifestyles, we are naturally led to ask, which groups, and how do they promote unhealthy lifestyles? In the case of smoking, tobacco companies and tobacco farmers make easy and natural "villains." But if we ask who actually profits most from tobacco use, the obvious answer is, as we have seen, government. The - 214 -same holds for alcohol. But of course "government" as an entity separate from society does not keep these revenues; they do not flow into the pocket of the Minister of Finance. These tax sources reduce the burden of other taxes; if they did not exist the general income or sales taxes would be higher. We have just seen that any hope of significant offsetting reductions in government expenditure, due to e.g. lowered health costs, is largely illusory. So the largest single beneficiary from unhealthy lifestyles, at least in the form of tobacco and alcohol use, is the general taxpayer. As Pogo remarked, "We have met the enemy and he is us". The implications of this observation are somewhat disturbing. Would any Canadian government, or the taxpayers behind it, be prepared to accept the consequences of a truly effective program of smoking cessation? If lifestyle improvement comes accompanied by significantly lower revenues or higher taxes, it will certainly be met with mixed enthusiasm. And the resistance will not all come from short-sighted smokers or industry "villains". Since economic/political factors play such a major role in lifestyle choice, one ought to look beyond simple economic policies. We turn our attention to some of the alternatives in the summary chapter. But first, it seems prudent to pause and digest the implications of the discussion in this chapter. 6.3 Summary Even with sufficiently strong causal epidemiological evidence to link particular lifestyle attributes to health status deterioration, - 215 -justification for economic incentives directed at those individual lifestyles requires that those attributes resulted from the exercising of consumer choice. In this chapter we have considered the implications of a number of different types or departure from a voluntary choice view of lifestyle formation. Four generic alternative perspectives were considered which generate two types of implication for economic incentives. To the extent that either of the first two theories - physiological and psychological - may be shown to be sole or participatory determinants of lifestyle, both the efficiency and equity cases for economic incentives are seriously undermined. Questionable, indeed, are penalizing policies intended to shift lifestyle choices which are not, in fact, voluntary choices, or to compensate losers by penalizing those engaging in financial-externality-generating activities which they did not choose of their own volition. With the social and economic/political theories of lifestyle determination, one ends up in quite a different spot. These do not represent such total departures from voluntary choice, but rather represent 'modification of voluntary choice' models. They do not necessarily or in all situations lead to the undermining of economic policies, but they do suggest other incentive policies applied either independently or in conjunction with individual economic incentives. Because we find a multifactorial theory of lifestyle creation far more compelling than the voluntary choice model, we turn in the final chapter not only to a summary of the ground covered by our analysis, but also to some of those other policy possibilities. - 216 -FOOTNOTES TO CHAPTER ~ 1. We do not mean to imply that health risks cannot be psychological in nature (e.g. psychosomatic illness). Rather, Veatch's psychological model is articulated in the context of health risk behaviors, and lifestyle characteristics, so is in fact a psychological model of lifestyle determination. 2. 'Social Structural' is applied here with a quite different intent than by Veatch (1980). His social structural •model' of health risks appears to be little more than an empirical observation of the correlation between socioeconomic class and disease incidence, mortality and the like. It appears, by our framework, to be offered as a model of health status rather than of lifestyle determination, whereas we are using it in the latter context. In addition, our Figure 6.1 does not include a social structural column because this portrayal of health status determination strikes us as somewhat vacuous. After all, it is surely not socioeconomic status per ~ which is a determinant of health status, but rather the attributes of that socioeconomic status, whether they be lifestyle, environment or some combination of the column headings in Figure 6.1. In contrast, social structural models of lifestyle determination do seem to constitute a distinct approach to explaining lifestyles (although there will undoubtedly be those who would argue that the social •effects' we are about to describe can only influence lifestyles through psychological channels; that they are in fact socio-psychological effects. So be it, for again the watertight-ness of the framework turns out to be relatively unimportant). - 217 -CHAPTER 7 CARROTS AND STICKS REVISITED: SUMMARY, POLICY IMPLICATIONS AND ALTERNATIVE DIRECTIONS "We have met the enemy and he is us" Pogo The justification for economic incentives directed at individuals in the pursuit of their 'own' lifestyles has, we believe, been shown to be both complex and subject to empirical question at almost every turn. Because of that complexity this concluding chapter begins with a synopsis of the original theoretical case(s) for economic incentives. We follow with a review of the evidential requirements associated with such incentives. The third section briefly summarizes our reviews of the state of that evidence. In the fourth section we bring forward the possible economic incentives which have been mentioned at various points in the report, and examine each in light of the discussion in the first three sections. The chapter concludes with a brief look at other public and private policy options which have evolved as responses to individual lifestyle choices. Focusing primarily on options related to smoking, the concluding section serves to elucidate the objectives (not always efficiency or equity) of some of those options. 7.1 Recalling the Arguments for Economic Incentives In a succinct and well-reasoned recent editorial, Kramer (1979, 138) argued that there were two justifications for making "people who avoidably contribute to their own ill health ••• pay for the drain which they thereby make on health resources ... The first, ••• is the - 218 -deterrence which this shifting of costs would effect. The second ••• is simply fairness". In this report we have attempted to develop a somewhat more detailed theoretical rationale for economic incentives directed at voluntary choices which contribute to nealth costs, but Kramer's two justifications reflect the underpinnings. The first justification, "deterrence", is equivalent to our case for intervention based on principles of efficiency. Here the argument went that if a particular voluntary behavior were both health-threatening and price-sensitive, and if the resulting deterioration in health increases health care costs, then resources are being inefficiently allocated. The "right" use of resources will follow if people voluntarily cut back on the activity (and if their improved health leads to lower health care use and lower health care costs - two separate and non-trivial assumptions) in response to a requirement that they pay personally for all its costs, direct and indirect. The second justification, "fairness" in Kramer's words, was more fully developed here as the equity argument for economic incentives. This argument is based not on allocative inefficiency, but on another of economists' favourite concepts - externalities. But this case does not represent the traditional polluting factory or noisy stereo type of external effect, but rather a financial externality - whether or not your behaviour affects me directly, it costs me money. The equity argument arises from the institutional/financing environment. Insurance mechanisms, whether public or private, fund individual care costs from collective resources. In principle, an insurance system (public .2!. private) might differentiate premiums to reflect self-induced costs; in - 219 -practice, costs of monitoring and enforcement make this discrimination trivial or impossible. While each of these theoretical cases for economic intervention is compelling, each is also dependent on long and at points fragile chains of assumed causation. Such interventions cannot be supported on theoretical grounds alone, but require evidence that the theoretical linkages on which they rest are in fact accurate representations of the world the interventions are designed to affect. Accordingly, we devoted a long chapter (the third) to articulating in detail both the theoretical links in those causal chains, and the implied evidential requirements for validating those links. 1.2 From Theory to Reality= Assumptions as Evidential Requirements The efficiency argument for public intervention begins in consumer demand theory - the first assumption is that lifestyles reflect the rational choice-making behavior of fully informed sovereign consumers. By that view, such individuals balance relative prices and marginal utilities so as to 'consume' a basket of goods and services which maximizes the utility of each consumer. Then economic incentives designed to enhance efficiency require, as a second assumption, price elasticity of the lifestyle activities or consumption patterns. If alcohol consumption, for example, is price inelastic, then confronting individuals with the true cost of their consumption choice will leave unchanged their purchase (and, by assumption, ingestion) of alcohol. They may (in fact will, ceteris paribus) purchase less of other things, but that is not exactly what the efficiency case had in mind. If alcohol consumption, and by extension health care costs, remain - 220 -' unchanged, then the allocation of resources to health care vis a vis other sectors is already the "correct" (or at least efficient) allocation. If they faced the true relative prices, consumers would choose to leave that component of their consumption pattern unchanged. There is, by implication, no allocative inefficiency and no efficiency case for economic intervention. Taking as given the first two assumptions, we then require a third assumption, that the price-induced shift in behavior results in a health status change. Continuing with the alcohol example, if doubling the price causes all who consume to halve their consumption, but lower consumption is known to leave those individuals' health status unchanged, ceteris paribus, then the efficiency case is once again in trouble. But even granting the first three assumptions is not enough. The efficiency case is based on an assumed over-allocation of resources to the health care sector. The fourth assumption, then, is that those people who are made healthier as a result of their lifestyle adjustments, will in fact need less health care. And finally, assumption five requires that individuals' lowered health care needs will be translated into lower aggregate health care costs. The observation, inter alia, that lowered health care costs implies lowered health care provider incomes suggests that this assumption as much as any of the others requires careful examination. This set of five assumptions represents the logical chain of events underlying the efficiency justification for public economic intervention. An unsupported or unrealistic assumption at any point in the chain will cause the case to collapse. By implication, we may - 221 -summarize the requirements for evidence as follows: (i) (ii) (iii) that lifestyles are the result of rational, fully-informed choices by individuals in response to relative costs in the context of independent utility maximization; that those choices are price elastic and, therefore, would shift in response to a shift in relative costs and benefits; that health status is causally linked to the lifestyle choices in question; (iv) that health care needs are a direct and monotonic function of health status; and (v) that aggregate health care costs follow directly from health care needs. The case for economic incentives based on the observation that individual behaviors may result in financial externalities, the equity case, may be pieced together in a similar manner. Like the efficiency argument, it requires as a first assumption that choices of lifestyle are in fact choices made by those same fully informed rational consumers choosing on the basis of marginal utilities and relative prices. But here the similarity ends temporarily. The equity case does not require that consumers respond to economic incentives, since it is based on a rationale of compensating non-participants for the collectively funded marginal financial consequences of the participants' behavior. If those participants' behavior does not change, the economic instrument permits that compensation; if it does change, then some combination of compensation and lowered requirements for compensation will obtain. Thus, behavior shifts are not necessary to the equity case. Instead, the second assumption in the equity argument is the third in the efficiency case. One again requires the causal link running from lifestyles to health care costs, the first component of which is - 222 -the lifestyles -------> health status link. Not surprisingly, the third assumption is then that improved health status implies lowered health care requirements. If it does not, then there cannot be marginal lifestyle-induced health care costs requiring compensating transfers. Finally, the fourth assumption is, as in the efficiency case, that per capita health care needs translate directly into aggregate health care costs. If health care needs turn out to be only one of several contributory factors in health care costs (in particular if, as much evidence suggests, aggregate costs are determined by types and numbers of manpower and facilities available), then requiring individuals with particular lifestyles to compensate those without, raises some serious ethical questions. Lifestyle choice may not affect global costs at all; only the distribution of costs across the population. The evidential requirements underlying an equity-based justification for economic interventions, then, include numbers (i), (iii), (iv) and (v) from the list developed above for the efficiency case. Even if the available evidence will in fact support the logical chain from (i) to (v), it does not follow that a public policy response is called for. It may be that, as appears to be the situation in Ontario, policies are already in place which respond (or over-respond) to a case in principle for intervention on either equity or efficiency grounds. "Something should be done" - and it is. Exploration of the current policy situation might be considered a sixth class of evidence necessary to support ~ initiatives. Thus, we have found that the practical justification for public economic intervention on efficiency grounds requires six types of evidence, and on equity grounds, five. 1 Four of the evidential - 223 -requirements ((i), (iii), (iv) and (v)) are common to both cases. The equity case does not require that individuals respond to such intervention by changing their behavior (ii). While this project's major objective was to articulate the extent and types of evidence required to support public economic incentives directed at lifestyles, a secondary objective was to assemble the key components of evidence. We turn, then, to a synopsis of the evidential ground covered in this report. 7.3 Testing the Assumptions=! Brief Review of the Empirical Evidence Of the requirements for evidence summarized in the previous section, earlier chapters were devoted to examining four: (i), (ii), (iii) and (vi). Chapter 6 dealt with (i), alternative theories of lifestyle formation. In Chapter 5 we examined (ii) and (vi), and Chapter 4 was devoted in its entirety to (iii). Our treatment of (i) was different from that of (ii), (iii) and (vi) in that we reviewed, not empirical evidence, but rather the implications of competing theoretical perspectives. The choice of both emphasis and treatment throughout this report was influenced partly by the availability of evidence and partly by our perception of the relative importance of various components of the evidence. The epidemiologic evidence was so fundamental to both justifications for economic policy and so much in question as to warrant detailed review. The oft-heard popular call to arms over the cost of peoples' lifestyles always refers to this link in the chain, while rarely (explicitly) to the others. As Chapter 4 shows, the available epidemiological evidence is extensive. In contrast, evidence on (iv) - 224 -and (v) is less available and more problematic. But lack of evidence is as important to this analysis as its presence, since the former implies that evidential requirements for intervention are not satisfied. Our approach was to take assumption (i) as given ~or the purposes of examining assumptions (ii), (iii) and (vi). Furthermore, our examination of evidence for assumption (iii) was restricted to a small set of putative causal relationships between lifestyle characteristics and health status. One lifestyle characteristic (smoking) was studied in examining assumptions (ii) and (vi). Finally, we back-tracked in Chapter 6 to consider the implications of various forms of evidence bearing on assumption (i). The synopsis here roughly follows that route. Four relationships were examined in Chapter 4: cigarette smoking ------> lung cancer, chronic obstructive lung disease, and coronary heart disease; and alcohol ------> cirrhosis of the liver. The rationale for these choices was twofold: (a) our collective prior perusal of the literature and discussions with various informed individuals suggested that these four relationships represented some of the most causally robust current evidence. It remained to establish an analytical framework within which to determine whether the causal strength was sufficient to satisfy assumption (iii), even for those four apparently outstanding sets of linkage; and (b) examining those relationships with the strongest causal links would eliminate any necessity to undertake a more wide-ranging and comprehensive review, for the strength of causation in those relationships would form an "upper bound" on the extent to which assumption (iii) is satisfied empirically. - 225 -Since summaries for each linkage appear in Chapter 4, here we bring forward only selected general and policy relevant results. our review makes clear the absence of evidence that smoking, even above some threshold intensity, is a necessary precursor to any disease state. Even the strongest causal evidence, for conditions such as cancer of the lung and larynx, indicates that these diseases will present occasionally in non-smokers. Without reference to any of the other assumptions listed above, this finding essentially eliminates justification for one class of economic interventions -- universal downstream user fees, or de-insurance of selected procedures. Kramer (1979, 140) may feel that the inhumanity of withholding "medical care from those sick persons who cannot pay" rates as the "most distressing" of her four "very serious problems" with economic interventions. In fact the more serious prior problem is that care would inevitably be withheld from, or charged to, a significant number of people who had never engaged in the targetted lifestyle. There may exist lifestyles which lead inevitably to specific morbid conditions, but to date they have not been identified. Nor has smoking been found to be a sufficient predisposing factor for any disease state. Most people who smoke heavily escape even the conditions with the highest relative risks while living long enough to eliminate the argument that 'it may have got them if they'd lived long enough•. This finding places the second and final nail into the coffin of downstream charges. To see this, suppose smoking were sufficient though not necessary for morbid condition M. Assume also that at the downstream end of the link one has sufficient information to distinguish smoking-caused condition M from other-caused condition M. Everyone who - 226 -smokes ends up with M. Information retrieval problems aside, once M presents, it is possible to establish responsibility. If the individual so affected smoked, then one could charge for care; if not, not. The absence of necessity in the relationship eliminates the justifioation for blanket downstream charges. But the absence of sufficiency eliminates that for selective downstream charges. Nevertheless, the evidence for a number of diseases (particularly lung cancer, chronic obstructive lung disease and laryngeal cancer) clearly satisfies the causal criteria in a probabilistic sense. That is, after all, the limit to the implications of relative risk evidence. The evidence from the alcohol literature yielded similar conclusions. While almost all alcoholic cirrhosis can be attributed to heavy alcohol use, there remains that small unattributable segment to preclude universal downstream charges. Selective charges are confounded by the state of threshold evidence and behaviour monitoring problems. Sufficiency is obviously violated - most drinkers do so with impunity. Again, however, this relationship and that with other diseases of the liver and fetal alcohol syndrome are probabilistically causative. The implications of our extensive but selective review of epidemiological research would appear to be that there are some disease states which are causally linked to lifestyle characteristics, but that those linkages are strong enough only to justify a particular subset of economic intervention options. We return to this theme in the next section. What needs to be taken forward is the finding that evidence on some putative linkages is currently sufficiently strong to provide qualified support for assumption (iii), but that those linkages are few - 227 -and far between. By implication, justification for economic interventions directed at many other lifestyle characteristics comes up against serious problems at assumption (iii); and if assumption (iii) is not satisfied, there is no need to continue to look f'urther for evidence on the other assumptions. For those few conditions which have been causally linked to lifestyle characteristics, an efficiency case requires that the lifestyle choices be sensitive to relative costs and benefits. Accordingly, Chapter 5 utilized smoking as an example and contained a brief review of evidence on price elasticities for cigarettes. While elasticities for different consumption activities (smoking, alcohol consumption, skiing, consuming junk foods) will undoubtedly vary, none is likely to reveal either perfect elasticity (any price increase would eliminate the activity) or inelasticity (the activity is insensitive to price). Thus the findings for smoking are likely to be sufficiently representative as to allow some generalization of the policy implications. For cigarette smoking, elasticities appear to range from -0.3 to -0.8 for the percentage price changes and price ranges reviewed in the literature. But considerable methodological problems continue to hamper accurate estimation. For our purposes here it may be safer to look at the implications of different elasticities rather than of any particular •true' elasticity. Since perfect (infinite) elasticity is a highly unlikely eventuality for any lifestyle-related choice, we might reasonably focus on the remaining range. Perfect inelasticity would, of course, - 228 -undermine assumption (ii), but perfect inelasticity even over a limited price range makes intuitive sense only in the context of circumstances not related to voluntary choice, e.g. addiction, and then only in the case of widespread (in fact universal) absence of voluntary choice. Once one backs off universality, one finds products such as alcohol and cigarettes, widely thought to have some addictive properties. For these products we find non-zero price elasticities. The question then becomes, how much price response constitutes sufficient evidence on assumption (ii). For an answer we need to look back to the development of the efficiency case. An efficiency case for economic policies requires that, if faced with different relative costs, individuals would choose different consumption patterns. By implication, any non-zero price elasticity will satisfy this condition. The larger the absolute value of the price elasticity, and the greater the 'true' versus actual price differential, the greater the extent of allocative inefficiency. A very low price elasticity may not warrant the administrative and political costs of implementation, even in the face of a(n assumed small) inefficiency. While the elasticities found in the smoking literature are not dramatic, they have been found consistently to range from -0.3 to -0.8. This suggests sufficient price sensitivity for cigarette consumption to satisfy assumption (ii). While we did not review similar literature for other lifestyle attributes, elasticities significantly different from zero are, we would contend, equally likely for other goods, services and activities. This seems, in fact, the least problematic of the assumptions. - 229 -A more problematic assumption, perhaps, is (i), which implies that lifestyles result from a set of choices made by sovereign, rational, fully-informed individuals reacting only to relative prices, budget constraints, and relative benefits as implied by the shape and form of their utility functions. Our review of the literature in Chapter 6 revealed a number of competing, and in some cases more convincing, views of the lifestyle development process. Competing view~ of lifestyle formation include physiological, psychological, social structural, and economic/political models. The physiological and psychological models represent fairly substantial departures from a voluntary choice view of the world, whereas the social and political models may be thought of more as "modified voluntarism". This distinction, we argued, takes on some importance when we come to address assumption {i). If addiction or childhood-related psychological processes determine or affect adult lifestyle formation, policies based on assumptions of voluntary fully-informed choice require careful re-examination. If, instead, those choices are 'voluntary' in the sense intended here, but are shaped or influenced by social or economic/political environments, economic policies may not be entirely undermined, but targeting may require care, or other conjunctive non-economic policies may be desirable. The evidence, such as it is, suggests that each of these competing models is likely to play a role in subsets of individuals and lifestyle attributes. Certainly there is now little question about the addictive properties of certain substances constituting parts of certain individuals' lifestyles. Talk of individual responsibility for marginal - 230 -costs, and shifting of informed consumption patterns through economic incentives rings hollow in such a context. There is also sufficient anecdotal evidence (if nothing else) to suggest that social critical mass phenomena do exist in lifestyle-related activities. But unless economic policies are better able than non-economic to induce desirable social (rather than individual) trends, there is no particular policy imperative requiring economic policies in isolation or even as parts of a package of policy initiatives. The implications of an economic/political view of the world are similar. It, too, undermines assumption (i), and again suggests some combination of collective (rather than individual) and carefully targeted economic incentives, or non-economic initiatives. Some of these options are explored again in the final section of this report. But let us assume for a moment that assumptions (i) through (v) do in fact hold. Then we must ask whether policies already in place may have met the efficiency or equity objectives. To this end, part of Chapter 5 was taken up with an analysis of the equity balance in Ontario as it relates to cigarette smoking. There we found that the current tax structure generates sufficient smoking-related revenue to compensate for the marginal health care costs. Indeed, even under a very conservative set of assumptions, gross over-compensation (based on an equity case alone) seems more the reality. This supports one of the economic/political model variants described in Chapter 6 -- that the 'collective we' are politically unwilling to part with the handy revenues generated by cigarette consumption, although those revenues appear long since to have left the attributable health care costs in the dust of the tobacco farms. Consumers of these products face prices far - 231 -in excess of the true discounted combined health oare plus product cost. This finding simultaneously undercuts any argument for increased charges on allocative efficiency grounds. The extensive set of assumptions requiring empirical support and the nature of the empirical evidence leave a remarkably weak case for new economic incentives directed at individual choices. The epidemiological evidence eliminated downstream incentives and suggested that only for a small number of products or activities is the causal evidence strong enough to justify upstream incentives. Doubts about the extent to which those few products or activities find their way into lifestyles because of voluntary choice cast doubt on the case for such incentives even directed at those attributes. And if one is willing to grant the plausibility of voluntary choice, doubts remain about the complex health status ----> health care costs linkage which has not received attention in this report. Certainly evidence of a significant role for the supply side of that 'market' does exist (see, e.g., Evans and Wolfson, 1978). And finally, we find in Ontario that policies which take assumptions {i) through (v) as valid, at least in the case of smoking, have already been put in place. Furthermore, they appear to have been overzealously applied in either efficiency or equity contexts. This seems, then, an appropriate juncture at which to review the plausibility of alternative forms of economic incentives with the individual as a target. - 232 -7.4 The Carrots~ Sticks Revisited At the conclusion of Chapter 2 we set out a tentative roster of economic incentives or disincentives which had two things in common --they were all policies which could be targeted at individuals, and they were all policies which could be designed and implemented in the lifestyles context. Now, far better equipped to assess alternatives, we re-examine each. We can eliminate certain options without further discussion. These are the downstream charges for health care services which, to be justified, would require the ~post identification of those conditions caused unequivocally by the lifestyle behaviour. Neither user fees (of less than 100 per cent of actual treatment cost) nor complete de-insurance is justifiable for any linkages that we are aware of today. Nor are partial user fees any more justifiable than complete de-insurance. If a particular condition is not always caused by behaviour A, then charging partial fees for its treatment will be both unfair and relatively ineffective. The other policy options are targeted not at the actual incidence of illness, but at relative probabilities of contracting illness. These are in general more consistent with evidence that certain apparently voluntary choices increase the probability of later incurring identifiable marginal health care costs. But their justification in any particular situation is dependent on the quality of the evidence on causality and voluntary choice. The first of these options mentioned in Chapter 2 was risk-related premiums. If individuals can, ~ante, be segregated into risk groups - 233 -based on their lifestyle behaviours and causal epidemiological evidence, then evidence linking various health states to health care costs could be used to construct differential health insurance premiums. There are a number of problems with this type of policy, the most obvious of which is the intractable challenge of collecting behavioural information. If consumers/individuals hold an information advantage over the insurers, this type of scheme would quickly unravel due to adverse selection. Individuals would assign themselves to too-low-premium classes, driving up those premiums and driving away the better risks in a chain of events which would eventually leave many individuals uninsurable and others not desiring any of the available options (Evans, 1983). Risk-related premiums could serve an equity objective rather well, but only under very stringent conditions -- sufficiently strong causal evidence on assumptions (iii), (iv) and (v), a conviction that assumption (i) is realistic, and monitorability of lifestyle behaviour. Alternatively, policy makers might tax a wider variety of products or activities negatively linked to health status, or subsidize those positively linked. Justification for such policies is dependent, of course, on the strength of the series of links forming the lifestyle behaviour ----> health care costs chain. Then their impact on efficiency, although not equity, further depends on price elasticities. While these policies are the most easily justified on the current evidence, they too are not without problems. Our analysis of cigarette taxes and smoking-related health care costs in Chapter 5 points up one -- that of determining the appropriate ad valorem tax rate -- and the others draw from the equivocal evidence bearing on each of the key - 234 -assumptions outlined earlier. But this brings us to some sort of closure on the issue of economic incentives directed at individuals' lifestyle choices. Beyond circumspect point-of-choice taxes or subsidies on a remarkably narrow set of products or activities, current evidence is insufficiently strong to support this genre of public policies. This analysis also allows us to hammer the one outstanding nail into the coffin of user fees. The evidence justifying user fees on the grounds of individual responsibility for components of publicly funded health care costs can only charitably be described as rarefied, or more realistically as ectoplasmic. Rather, user fees seem justifiable now only as a weak-kneed mechanism for taxing the ill to avoid confronting some very basic algebra. All this is not, however, meant to suggest that with the exception of selective taxes we should sit on our hands. The framework developed in the past chapters can be brought off the shelf as new evidence on the role of lifestyles in illness etiology comes to light. And the state of epidemiological evidence, while probabilistic rather than deterministic, is sufficient to justify exploration of options . other than financial (dis)incentives directed at individuals. We conclude this report with a brief examination of some of those possibilities. 7.5 Alternative Directions The economic options which have consumed our attention in this report shared two characteristics: they were directed at individual choice; and they were intended as instruments toward goals of efficiency - 235 -or equity. In this concluding section our interest at the outset continues to be the efficiency and equity issues, but we attempt (without pretending to be comprehensive) to expand the scope beyond financial carrots or sticks directed at the individual. This seemed an appropriate way to conclude a study which has found that (i) voluntary choice may not be the most compelling behavioral description of all aspects of lifestyle formation, and (ii) even granting voluntary choice, the strength of the epidemiological link is sufficiently strong only to support upstream (dis)incentives (product or activity taxation) for a small number of lifestyle attributes. Taking these in reverse order suggests a need to look not only at non-economic policy options which might be directed toward individuals, but options of all types more .cognizant of the social, economic and political influences on lifestyle formation. Any policy, economic or other, which is to achieve goals of efficiency or equity must impact on one or more points or linkages in the putative causal chain from lifestyles to costs. If one is able to isolate and eliminate the lung cancer pathogen in cigarettes, the linkage from the lifestyle activity to health status collapses (Haggerty, 1977), and with it both efficiency and equity problems. Since the true cost of smoking would fall, allocative inefficiences currently related to smoking would be eliminated and, since attributable costs would disappear, there would no longer be a case for any policy on financial grounds. The problem of the directly offensive impact of smoke on non-smokers would, of course, remain. Similarly, a miraculous palliative which largely eliminated the health care resource requirements of liver cirrhosis, for example, would - 236 -weaken the health status --> health care resources linkage necessary to both efficiency and equity problems. It would not follow that we (collectively) should ignore alcohol abuse, but the case for economic disincentives on financial grounds would no longer merit consideration. But in the absence of breakthroughs affecting most lifestyle --> health status linkages, and given evidence inconsistent with lifestyle-related policies of any form at the downstream end (health care costs), what other upstream (lifestyle formation) policy options exist? We offer one possible framework within which alternatives may be classified in Figure 14. The two dimensions are generic policy choices, and the targets of those choices, respectively. Following Quelch (1978), we suggest that 2 there are three broad policy classes -- financial, legal and message. Within each, policies may be designed with the individual, the social structure/environment, or the political/economic environment in mind. The target will presumably depend on the policy-makers' philosophical/evidential view of lifestyle formation. In addition, policies or initiatives may be public or voluntary/private. Of course the upper left hand cell in this framework should be extremely familiar, for it is what most of this report has been about. Moving down the first column takes us beyond economic policies while still targeting individual choice, while movement along the first row would be justified by an other-than-individual-choice view of lifestyles coupled with a desire to stay with financial/economic policies. Our approach here will be to begin at the bottom and work up in rows, offering as we go along some examples and discussion for each cell. Figure 14 Generic Policy - 237 -Beyond Individuals and Economic Incentives--Lifestyle Theories and the Expanded Policy Basket Lifestyle Determinants I I Individual Social I Structural I I I Taxes, User Fees, Financial I De-Insurance, etc. I I I Worker Bonuses Options Legal Smoking Prohibition in Public Places; Message Seat Belt Legislation I Health I Practitioner I Counselling; I Conventional I Media Campaigns;! Anti-smoking I publicity I I I Social Learning; Participaction Political/ Economic Cigarette Advertising Ban - 238 -In addition to moving outside purely economic policies, the following discussion makes two other marked departures. In the interest of comprehensive coverage we relax our focus on public intervention to include potential voluntary private initiatives. Of course the alert reader will immediately realize that this also necessitates a departure from our focus up to now on goals of efficiency and equity only. It would be difficult to argue, for example, that a widespread movement by private practitioners to smoking counselling would be motivated either by efficiency or equity considerations. Their objectives will be some blend of promoting patients' health and adding to their own product line and fees. Message Policies Message policies include education and advertising, both of which use the communications media. There the similarity ends. Education may be thought of as the process of reducing factual information gaps; advertising represents the set of activities which alter information perceptions, "attitudes, beliefs, [and] motivation" by providing "selected information and perspectives", (Hoffman, 1979) for the sole purpose of affecting behaviour. Thus, education may affect behaviour, but that is not its raison d'etre. Rather, education policies or initiatives have the goal of ensuring that behaviour reflects informed choice. Health education directed at lifestyle choices has an uneven track record, leaning if anything to the unsuccessful. Haggerty (1977) reviews a number of examples and suggests that "passive information transfer" is far less likely to affect behaviour than education policies which require active individual involvement and which employ, indeed - 239 -focus on, a social context. Thus, social learning (or behaviour modification) has become a popular and relatively successful form of education directed at the individual but cognizant of the social influences on lifestyle choices (Pomerleau et.al., 1975, Haggerty, 1977). Foraie (1980) reviews a number of education initiatives, and notes that the more successful among "traditional" programs combine mass media techniques with counselling or voluntary self-help groups. Bass (1978) promotes the concept of initiating "wellness epidemics" by recognizing the social context of lifestyle formation, while Warner (1977) and Leu (1982) have attempted to evaluate the impact of mass media anti-smoking campaigns. The latter found synergistic effects of such publicity with individual financial disincentives. Rimer (1979) acknowledges the social context of lifestyle determination by advocating the use of role models in anti-smoking campaigns. In addition to mass media and social learning, a third education policy solidly rooted in the individual choice view of lifestyles has received considerable attention in the literature. This is the use of respected health professionals, primarily physicians, for lifestyle counselling (Holleb, 1979; Baumgartner, 1982; Wilson et.al., 1982; Rosen & Ashley, 1978). For example, one of the key agents in promoting Bass' (1978) wellness epidemic would be the family physician. Senior (1982) points out that hospitals are far behind physicians in promoting smoking cessation, primarily because the other health professionals do not yet (and may never) see themselves in a role model context. The reference to role models in the context of counselling underlines the inevitable blend of educational and advertising - 240 -techniques. If education is the process of communicating information, then advertising may, perhaps must, be used as a complementary vehicle for increasing individuals' likelihood of absorbing that information and acting . upon it. Thus the Participaction program has both education and advertising elements, as do seat belt campaigns, drinking driver messages and the like. There is nothing quite like the image of a mangled body and a shattered windshield to propel home factual information about the impairing influence of alcohol consumption. Legal Legal initiatives are a double-edged sword. They tend to be the most successful at effecting behaviour change, so long as they are not too onerous, but at the same time they represent the most intrusive policies from the perspective of individual autonomy (Haggerty, 1977; Wikler, 1978). In the lifestyles context, perhaps the best known regulatory initiatives have been those directed at smoking in public places (Lewis, 1979). These policies not only target individual choice but, in altering perceptions of social acceptability also promote cessation through the social underpinnings of choice formation. SUch "legal" policies are not solely the domain of the public sector, however. Private restaurants• and airlines• designation of smoke-free areas may equally be thought of as pseudo-legal initiatives with the same dual targets. Falling into the same slot in our framework is legislation making seatbelt use mandatory. While primarily a legal policy rooted in individual choice, the fact of the legislation may produce social influence on behaviour, and tied to the legislation may be financial penalties of one sort or another. Cigarette advertising bans have been contemplated, even implemented, - 241 -in various jurisdictions over the years. overwhelming successes (Hamilton, 1977). They have not generally been This type of policy most comfortably fits the "legal" row of our framework, although one could also make a case for its inclusion in the "message" row. It falls, in either case, in the political economy row, as it is a policy not targeted directly at individuals or their social environments but rather at the economic influences on lifestyle formation as represented by the tobacco industry. One could dream up a variety of other legal initiatives for the political economy cell which reflect the fiscal information reported earlier. If, in fact, governments are net financial beneficiaries (and the primary beneficiaries) of a lifestyle choice such as smoking, perhaps a policy of limiting taxation of cigarettes would not be inappropriate! Financial We conclude with a look at financial policies. While the bulk of this report has examined economic (dis)incentives directed at individual choice from a public policy perspective, such initiatives need not be restricted to public policy. Shepard (1980), for example, reports on an apparently successful anti-smoking initiative undertaken by a small corporation in the United States. This particular policy, a bonus for no smoking at the workplace, was directed at individuals but was likely aided by the social environment represented by the workplace. In the previous chapter we noted that truly successful policies aimed at lifestyle modification could only be those that take into account the political and economic context of lifestyle formation. Policies to reduce smoking may also require compensation for those who - 242 -would most lose by the success of such policies. In the extreme, this could entail the "buying out" of tobacco producers, cigarette manufacturers and distributors, hops growers, distillers and brewers, etc. But the extreme may be too extreme. After all, dose was shown to be a key variable in the epidemiological evidence on alcohol consumption; negative effects may develop only above a threshold. If some levels of consumption do not generate downstream effects, where is the justification for a total buy-out? (Abstracting, of course, from black market problems and the like). 7.6 In Sum While the upper right hand cell in our policy framework, represented by financial incentives directed at the political/economic factors in lifestyle formation, may not house many viable policies, it does lead rather naturally to what seems an appropriate conclusion. There are a variety of policies, both public and private, which may be directed at efficiency, equity or even paternalistic goals deriving from lifestyle behaviours and their consequences. Private initiatives (e.g. health practitioner counselling) will in general, be paternalistically motivated and we have little trouble with those. But the paternalistic case for public intervention is more contentious (Wikler, 1978), because it rests ultimately not on analysis alone, but on some concensus or convergence of political and social values which may differ across societies. While the efficiency and equity cases for public intervention are more compelling, the problem of designing appropriate and effective public policies to address them is - 243 -anything but straightforward. To the extent that lifestyles are formed in a broader social/political context, economic incentives targeted at individuals may be directed at the wrong targets. In an environment where government at all levels does not discourage, in fact overtly encourages, particular lifestyle attributes (e.g., through tax deductions for cigarette advertising, and liquor consumption for "business purposes" (Mosher, 1983)), it makes little sense to talk of publicly supported legal, financial or media initiatives directed at those lifestyle attributes. Such policies only add to the social cost of promoting the attributes. When the very activities which are of ostensible concern .are in fact being publicly supported, one must nervously conclude that the collective concern is outweighed by the collective benefits of the activities being supported. If that is not the case, then a good place to start might be in the upper right hand cell of Figure 14, through working toward a consistent set of public policies. In the smoking context this may mean acknowledging the political/economic realities by making some necessary changes to corporate tax regulations and fine tuning cigarette taxation to ensure that governments are not net fiscal beneficiaries of cigarette consumption. Only then should we contemplate a balanced set of other policies recognizing both the social and individual roles in lifestyle formation. A more limited set of objectives, less threatening to the revenues of both government and industry, might include the extension of policies to protect non-participants, or involuntary participants, in unhealthy life-styles -- such as smoke-free areas to protect the health and comfort of non-smokers, determined efforts to keep drinking drivers off - 244 -the road, and restrictions on advertising to protect children in particular from the habit-formation or addiction process. In addition, programs to assist those who want to escape from addiction should, if demonstrably effective, be expanded. In such programs, economic incentives may or may not play a valuable role -- the issue is inherently empirical. But for governments to contemplate general economic penalties for the consequences of unhealthy lifestyles, on the argument that people should be more "responsible" for their own health, is cynical hypocrisy when these same governments not only assist the marketing of these lifestyles, but also are the principal profiters from such unhealthy lifestyles. The "responsibility" argument looks rather like yet another way of deriving revenue from unhealthy lifestyles rather than a serious interest in health. This discussion may appear to focus too much on "smoking and drinking", the traditional evils, rather than lifestyles in general. The justification is that these activities are those where the health consequences of lifestyle choices are best established -- the same reason why they were chosen as the focus for this study. In the more general areas of diet, exercise, or stress, we simply have very little secure causal information about health implications, much less a basis for estimates of quantitative effect or conditional probabilities. But it is worth noting that in the cases of seat belt or motorcycle helmet use where health implications of lifestyle choice are clearcut and governments have ~ economic stake in unhealthy lifestyles, they act not through "economic incentives" but by direct regulation. And their - 245 -actions are effective. Similarly, immunization campaigns are carried out through publicity and direct provision, not by providing bonuses to the immune or taxing the remaining susceptible. In these cases, economic incentives would probably represent an excuse for inaction rather than a policy for action. - 246 -FOOTNOTES TO CHAPTER 1 1. It is interesting to note that Kramer (1979) included two of these requirements in her very much shorter editorial, in the context of "four very serious problems" with these sorts of policies. The two she included were the epidemiological requirements (iii), and the evidence on fully informed independent choice (i). 2. Both Evans (1978) and Wikler (1978) imply the same three-pronged generic choice set, although the labels differ somewhat. - 247 -REFERENCES Akerlof, G.A. (1970), "The Market for 'Lemons•: Qualitative Uncertainty and the Market Mechanism", Quarterly Journal of Economics 84:3. American Cancer Society (1978), ! National Dilemma: Cigarette Smoking 2!'.:. the Health o~ Americans, Repo~t of the National Commission on Smoking and Public Policy to the Board of Directors, American Cancer Society, Inc., January. American Heart Association (1978), "Value and Safety of Diet Modification to Control Hyperlipidemia in Childhood and Adolescence", Ad Hoc Committee of the Steering Committee For Medical and Community Programs, American Heart Association Committee Report, August, 381A-3B5A. Aoki, K. et al. (1979), "An Epidemiologic Approach to Host Factors in the Etiology of Cancer", National cancer Institute Monograph, no. 53, 17-23. ·Aronow, W. s. ( 1976a), "Carbon Monoxide and Cardiovascular Disease", in Smoking and Health: !.:. Modifying the Risk for the Smoker, pp. 321-330, DHEW (NIH) 76-1221, Washington, D.C.: U.S. Government Printing Office. Aronow, w.s. (1976b), "Introduction to Smoking and Cardiovascular Disease", in Smoking and Health: !.:. Modifying the Risk for the Smoker, pp. 231-241, DHEW (NIH) 76-1221, Washington, D.C.: U.S. Government Printing Office. Aronow, w.s. (1976), "Effect of Cigarette Smoking and of Carbon Monoxide on Coronary Heart Disease", Chest 70:4, 514-518. Aronow, w.s. (1980), "Effect of Non-Nicotine Cigarettes and Carbon Monoxide on Angina", Circulation 61:2, 262-265. Arthes, F.G. and A.T. Masi (1976), "Myocardial Infarction in Younger Women. Associated Clinical Features and Relationship to Use of Oral Contraceptive Drugs", Chest 70:5, 574-583. Ashley, M.J. (1981), "Alcohol Use During Pregnancy: A Challenge for the 180s 11 , Canadian Medical Association Journal 125:2, 141-154. Ashley, M.J. et al. (1981), "Smoking or Health in the 80•s 11 , Canadian Medical Association Journal 125, 1077-1078. Ashley, M.J. and P.N. Corey (1980), Health Hazard Appraisal Risk Factor Update, Cirrhosis of the Liver, Department of Preventive Medicine and Biostatistics, University of Toronto, Health and Welfare Canada: Division of Health Systems. Associated Press (1981), "AMA Butts Out Cigarette Stock", The Province September 30, C14. - 248 -Associated Press ( 1981), "Cigarettes Can Be 'Good' For You, Research Shows", The Province August 26, B3. Associated Press (1982), 11 430,000 To Die From Smoking", The Province February 23, A3. Athanasou, J.A. (1981), "Sickness Absence and Smoking", The Medical Journal of Australia 1, 211-212. Atkinson, A.B. and J.L. Skegg (1974), "Control of Smoking and Price of Cigarettes - A Comment", British Journal of Preventative and Social Medicine 28, 45-48. Auerbach, o. and L. Garfinkel (1980), "Atherosclerosis and Aneurysm of Aorta in Relation to Smoking Habits and Age", Chest 78:6, 805-809. Auerbach, o. et al. ( 1979), "Changes in Bronchial Epithelium in Relation to Cigarette Smoking, 1955-1960 vs. 1970-1977", New England Journal of Medicine 300:8, 381-386. Auerbach, o. et al. (1976), "Cigarette Smoking and Coronary Artery Disease", Chest 70:6, 697-705. Auerbach, o. et al. (1974) "Relation of Smoking and Age to Findings in Lung Parenchyma: A Microscopic Study", Chest 65:1, 29-35. Auerbach, o. et al (1972), "Relation of Smoking and Age to Emphysema; Whole-Lung Section Study", New England Journal of Medicine 286:16, 853-857. Ayres, S.M. (1975), "Cigarette Smoking and Lung Diseases: An Update", American Thoracic Society 3:5, 1-6. Baghurst, K.I. (1980), "Nutritional and Health Aspects of Alcohol Consumption", Medical Journal of Australia 2:4, 177-180. Barer, M.L. (1981), Community Health Centres and Hospital Costs in Ontario, Occasional Paper 13, Ontario Economic Council. Barer, M.L. and R.G. Evans (1984), "Prices, Proxies and Productivity: An Historical Analysis of Hospital and Medical Care in Canada", in Diewert and Montmarquette, eds., Price Level Measurement: Proceedings from~ Conference Sponsored ~ Statistics canada, pp. 705-777, Ottawa: Minister of Supply and Services Canada. Barer, M.L.; R.G. Evans and G.L. Stoddart (1979), Controlling Health Care Costs ~ Direct Charges to Patients .= Snare or Delusion, Toronto: Ontario Economic Council. Bass, F. (1978), "Using Health Care Systems to Help People Quit Smoking, Facilitating a Welcome Epidemic", Paper presented at International Conference on Smoking Cessation, New York City, June 21-23. Bass, F. (1973), Medical Care Use Attributable to Cigarette Smoking thesis, Baltimore, Maryland: John Hopkins University. - 249 -Baumgartner, K. (1982), "EDITORIALS - Smoking cessation: physician --make a stand" I Canadian Medical Association Journal, 126:January, 101-102. Ballet, s. and L. Roman (1969), "The Predictability of Coronary Heart Disease", in A.N. Brest, Coronary Heart Disease, pp. 80-97, Philadelphia: ~.A. Davis Company. Berry, R.E. (1976), "Estimating the Economic Costs of Alcohol Abuse", The New England Journal of Medicine 295:11, 620-621. Berry, R.E. and J.P. 'Boland (1977), The Economic Cost of Alcohol Abuse, London: Collier Macmillan Publishers. Best, E.W.R. and C.B. Walker (1964), "A Canadian Study of Smoking and Health - Second Report", Canadian Journal of Public Health 55:1, 1-11 • Blumenthal, J.A. et al. (1978), "Type A Behavior Pattern and Coronary Atherosclerosis", Circulation 58:4, 634-639. Bock, F.G. (1968), "Dose Response: Experimental Carcinogenesis", National Cancer Institute Monograph No. 28, Toward a Less Harmful Cigarette, pp. 57-63, u.s. Dept. of Health, Education, and Welfare, Washington, D.C.: U.S. Government Printing Office. Boden, L.I. (1976), "The Economic Impact of Environmental Disease on Health Care Delivery", Journal of Occupational Health 18:7, 467-472. Borgatta, E.F. and R.R. Evans (1968), Smoking, Health, and Behavior, Chicago: Aldine Publishing Company. Boyce, B.A. (1976), Nursing Practise: Respiratory Care Terminology, New York: American Lung Association. Brailey, A.G. (1980), "The Promotion of Health Through Health Insurance", The New England Journal of Medicine 302:1, 51-52. Brand, R.J. et al. (1979), "Work Activity and Fatal Heart Attack Studied By Multiple Logistic Risk Analysis", American Journal of Epidemiology 110:1, 52-62. ~ Brenner, M.H. (1975), "Trends in Alcohol Consumption and Associated Illnesses. Some Effects of Economic Changes", American Journal of Public Health 65:12, 1279-1292. Brenner, M.H. (1979), "Mortality and the National Economy, A Review of the Experience of England and Wales, 1936 to 197611 , The Lancet Sept. 15, 568-573. Breslow, L. (1980), "Cigarette Smoking and Health", Public Health Reports 95:5, 451-455. - 250 -Brest, A.N., ed. (1969), Coronary Heart Disease, Philadelphia: F.A. Davis Company. British Columbia (1982, 1983), Vital Statistics of the Province of British Columbia, Victoria: Queen's Printer for British Columbia. British Columbia Lung Association, "C.O.P.D.: Chronic Obstructive Pulmonary Disease", !2!!!:. Health 62:1, 1-14. Burch, P.R.J. (1978), "Smoking and Lung Cancer: The Problem of Inferring Cause", Journal of the Royal Statistical Society 141, 437-477. Burch, P.R.J. (1979), "Coronary Disease: Risk Factors, Age, and Time", American Heart Journal 97:4, 415~419. Burch, P.R.J. (1980), "Smoking and Lung Cancer. Tests of a Causal Hypothesis", Journal of Chronic Diseases 33, 221-238. Burch, P.R.J. (1981), "Smoking and Mortality in England and Wales", Journal of Chronic Diseases 34, 87-103. Burns, D.M. (1975), "Consequences of Smoking- The Involuntary Smoker", in J. Steinfeld et al., eds., Smoking and Health. II. Health Consequences, Education, Cessation Activities, and Governmental Action, Proceedings of the Third World Conference on Smoking and Health, pp. 51-57, DHEW (NIH) 77-1413. Butler, N.R. (1975), "Smoking and Pregnancy", in J. Steinfeld et al., eds., Smoking and Health. II. Health Consequences, Education, Cessation Activities, and Governmental Action, Proceedings of the Third World Conference on Smoking and Health, pp. 43-49, DHEW (NIH) 77-1413. Cala, L.A. and F.L. Mastaglic (1980), "Computerized Axial Tomography in the Detection of Brain Damage", The Medical Journal of Australia 2:4, 193-198. - -Campbell, R.R. (1971), Economics of Health and Public Policy, Washington, D.C.: American Enterprise Institute. Canada (1979), National Health Expenditures in Canada 1960-1975, Ottawa: Health and Welfare Canada. Canada (1981a), Federal Government Finance, Statistics Canada Catalogue No. 68-211, June. Canada (1981b), Fiscal Federalism in Canada, Report of the Parliamentary Task Force on Federal-Provincial Fiscal Arrangements (Breau Report), Ottawa: House of Commons, August. Canada (1982a), Provincial Government Finance, Revenue and Expenditures, Statistics Canada Catalogue No; 68-207 Annual, June. Canada (1982b), National Health Expenditures in Canada 1970-1979, Ottawa: Health and Welfare Canada. - 251 -Canadian Medical Association (1981), Evidence Presented to the Special Committee on the Federal-Provincial Fiscal Arrangements, House of Commons, Canada, Minutes of Proceedings and Evidence, Issue No. 10, pp. 10-3 to 10-54, and 10A-1 to 10A-44, Tuesday, May 12, First Session, Thirty-Second Parliament, 1980-81. The Canadian Tax Foundation (1979), The National Finances, 1978-79, Toronto. Carlena, E. (1979), "Smoking and Occupation", in L.M. Ramstrom, ed., The Smoking Epidemic, ! Matter of Worldwide Concern, Proceedings of the Fourth World Conference on Smoking and Health, Stockholm. Chovil, A.C. (1979), "Occupational Lung Cancer and Smoking: A Review in the Light of Current Theories of Carcinogenesis", Canadian Medical Association Journal 121, 548-555. Christianson, R.E. (1979), "Gross Differences Observed in the Placentas of Smokers and Non-Smokers", American Journal of Epidemiology 110:2, 178-187. Clarke, J.H. et.al. (1982), "Cigarette Smoking and External Locus of Control Among Young Adolescents", Journal of Health and Social Behavior 23:Sept., 253-259. Collishaw, N.E. and G. Myers (1983), "Dollar Estimates of the Consequences of Tobacco Use in Canada, 1979", paper presented at the annual meeting of the Canadian Public Health Association, June 3. Crawford, W.A. (1980), "Photochemical Smog Standards- Practicability or Purism: An Evaluation of Health Risk in Sydney", The Medical Journal of Australia 1, 8-12. ~-Cullen, K.J. et al. (1980), "Alcohol Consumption and Cigarette Smoking in Busselton, 1966-197811 , The Medical Journal of Australia 2:2, 87-92. Dalen, J.E. and R.B. Hickler (1981), "Oral Contraceptives and Cardiovascular Disease", American Heart Journal 101:5, 626-638. Davies, J.W.; C.B. Walker and E.W.R. Best (1964), "The Magnitude of the Lung Cancer Problem in Canada", Smoking and Health, Ottawa: Health and Welfare Canada, 74833-5-1. Dayton, s. (1979), "Smoking and Coronary Heart Disease", The Lancet January 13, 111. de Lint, J.E.E. (1980), "Alcohol Control Policy as a Strategy of Prevention", Journal of Public Health Policy March, 41-49. Detwiller, L.F. (1979), "Consumer Charge: Key to Hospital Autonomy and Consumer Responsibility", Hospital and Health Services Administration Spring, 88-97. - 252 -Dirksen, H., L. Janzon and S.E. Lindell (1974), "Influence of Smoking and Cessation of Smoking. A Population Study of Closing Volume and Nitrogen Wash-Out", Scandinavian Journal of Respiratory Disease 85 (Supplement), 266-274. Doerken, H. (1968), "The Etiology of Myocardial Infarction - With Special Reference to Cigarette Smoking Among Young Coronary Patients and Those With Seoond Heart Attacks", National Cancer Institute Monograph No. 28, Toward ~ ~ Harmful Cigarette, pp. 21-27, U.S. Dept. of Health, Education, and Welfare, Washington, D.C.: u.s.· Government Printing Office. Doll, R. and A.B. Hill (1964), "Mortality in Relation to Smoking: Ten Years' Observation of British Doctors", British Medical Journal 1, 1399-1410. Doll, R. and A.B. Hill (1966), "Mortality of British Doctors In Relation to Smoking: Observations on Coronary Thrombosis", National Cancer Institute Monograph No. 19, Epidemiological Study of Cancer and Other Chronic Diseases, pp. 205-217, U.S. Dept. of Health, Education, and Welfare, Washington, D.C.: U.S. Government Printing Office. Doll, R. and R. Peto (1976), "Mortality in Relation to Smoking: 20 Years' Observation on Male British Doctors", British Medical Journal 2, 1525-1536. Doll, R. et al. (1980), "Mortality in Relation to Smoking: 22 Years' Observation on Female British Doctors", British Medical Journal April 5, 967-971. Donovan, J.W. (1979), "Review of Mortality and Morbidity Due To Smoking", in L.M. Ramstrom, ed., The Smoking Epidemic, A Matter of Worldwide Concern, Proceedings of the Fourth World Conference on Smoking and Health, Stockholm. Doyle, J.T. (1968), "Importance of Dose Response In Terms of Total Cigarette Smoke, 'Tar', and Nicotine: Cardiovascular System", National Cancer Institute Monograph No. 28, Toward a Less Harmful Cigarette, pp. 43-63, u.s. Dept. of Health, Education, and Welfare, Washington, D.c.: U.S. Government Printing Office. Doyle, J.T. et al. (1976), "Factors Related to Suddenness of Death From Coronary Disease: Combined Albany-Framington Studies", American Journal of Cardiology 37, 1073-1078. Durbec, J.P., J.M. Bidart and H. Sarles (1979), "Etude des variations du risque de cirrhose du foie en fonction de la consommation d'alcool", Gastroenterol. Clin. Biol. 3, 725-734. Dwyer, T. and B.S. Hetzel (1980), "A Comparison of Trends of Coronary Heart Disease Mortality in Australia, U.S.A. and England and Wales With Reference to Three Major Risk Factors - Hypertension, Cigarette Smoking and Diet", International Journal of Epidemiology 9: 1, 65-71. - 253 -Eckardt, H.J. et al. (1981), "Health Hazards Associated With Alcohol Consumption", Journal of the American Medical Association 246:6, 648-666. - --Eghoje, K.N. and E. Juhl (1979), "Factors Determining Liver Damage in Chronic Alcoholics", Scandinavian Journal .2f Gastroenterology 8, 505-512. Elmes, P.c. (1981), "Relative Importance of Cigarette Smoking In Occupational Lung Disease", British Journal of Industrial Medicine 38, 1-13. Elwood, J.M. (1981), "Wood Exposure and Smoking: Association With Cancer of the Nasal Cavity and Paranasal Sinuses in British Columbia", Canadian Medical Association Journal 124, 1573-1577. Enstrom, J.E. (1979), "Rising Lung Cancer Mortality Among Nonsmokers", Journal of the National Cancer Institute 62:4, 755-760. Evans, R.G. (1978), "The Cost-Effectiveness of Preventive Health Services", Paper prepared for the Alberta Public Health Association Convention, Red Deer, Alberta, April 12. Evans, R.G. (1983), "The Welfare Economics of Public Health Insurance: Theory and Canadian Practice", in L. Soderstrom, ed., Social Insurance, p.p. 71-103, Amsterdam: North Holland. Evans, R.G. and A.D. Wolfson (1978), "Moving the Target to Hit the Bullet: Generation of Utilization by Physicians in Canada", paper prepared for the NBER Conference on the Economics of Physician and Patient Behaviour, Stanford, California, January 27-28. Fagerhaugh, S.Y. (1973), "Getting Around With Emphysema", The American Journal of Nursing 73:1, 1-6. Farber, E. (1973), "Biochemical Pathology of Liver Cell Injury", in M. Eliakim et al., eds., International Symposium~ Hepatotoxicity, pp. 3-6, New York: Academic Press. Farber, E. and M.M. Fisher (1979), Toxic Injury of the Liver. Part!' New York: Marcel Dekker, Inc. Feinleib, M. and R.R. Williams (1973), "Relative Risks of Myocardial Infarction, Cardiovascular Disease and Peripheral Vascular Disease by Type of Smoking", in Smoking and Health: .!:. Modifying the ~ for the Smoker, 243-255, DHEW 76-1221. Feinleib, M. et al. (1979), "A Comparison of Blood Pressure, Total Cholesterol and Cigarette Smoking in Parents in 1950 and Their Children in 1970 11 , American Journal of Epidemiology 110:3, 291-303. Feinstein, A.R. (1973), "Clinical Biostatistics. XX. The Epidemiologic Trohoc, the Ablative Ratio, and 'Retrospective' Research", Clinical Pharmacology and Therapeutics 14:2, 291-307. - 254 -Feldstein, M. (1981), Hospital Costs and Health Insurance, Cambridge, Mass.: Harvard University Press. Ferris, B.G. Jr. (1975), "Smoking and Lung Function: Epidemiological Evidence", in J. Steinfeld et al., eds., Smoking and Health. II. Health Conseguences, Education, Cessation Activities, and Governmental Action, Proceedings of the Third World Conference on Smoking and Health, PP• 115-129, DHEW (NIH) 77-1413. Fielding, J.E. (1978), "Successes of Prevention", Milbank Memorial Fund Quarterly 56:3, 274-302. Flamant, R. et al. ( 1979) , "Lung Cancer", in L. M. Ramstrom, ed. , The Smoking Epidemic, ! Matter of Worldwide Concern, Proceedings of the Fourth World Conference on Smoking and Health, Stockholm. Fletcher, C.M. (1979), "Smoking and Chronic Lung Disease", in L.M. Ramstrom, ed., The Smoking Epidemic, ! Matter of Worldwide Concern, Proceedings of the Fourth World Conference on Smoking and Health, Stockholm. Fletcher, C.M. (1975), "The Consequences of Smoking- Pulmonary Disease", in J. Steinfeld et al., eds., Smoking and Health. II. Health Consequences, Education, Cessation Activities, and Governmental Action, Proceedings of the Third World Conference on Smoking and Health, pp. 35-41, DHEW (NIH) 77-1413. Fletcher, C.M. and D. Horn (1970), "Smoking and Health", WHO Chronicle 24, 345-370. Foote, E. (1981), "Advertising and Tobacco", Journal of the American Medical Association 245:16, 1667-1668. Foraie, D. (1980), Report to the Thrust Committee, Employee Health Programs, Policy Research and Management Services Branch, Saskatchewan Health. Foraie, D. (1980), Report to the Thrust Committee, Lifestyles and Modification of Behaviour, Policy Research and Management Services Branch, Saskatchewan Health, April. Fraser, R.D. (1981), discussion paper in R.M. Bird and R.D. Fraser, Commentaries £!! the Hall Report, p.p. 26-88, Toronto: Ontario Economic Council. Friedman, G.D. (1980), "Cigarette Smoking and Coronary Heart Disease: New Evidence and Old Reactions", American Heart Journal 99:3, 398-401. Friedman, G.D. et al. (1979), "Characteristics Predictive of Coronary Heart Disease in Ex-Smokers Before They Stopped Smoking: Comparison With Persistent Smokers and Non-Smokers", Journal of Chronic Diseases 32, 175-190. - 255 -Friedman, G.D. et al. (1979), "Mortality in Middle-Aged Smokers and Non-Smokers", New England Journal of Medicine 300:5, 213-217. Friedman, G.D. et al. (1981), "Mortality in Cigarette Smokers and Quitters", New England Journal of Medicine 304:23, 1407-1410. Friedman, K.M. (1975), "Cigarette Smoking and Public Policy", American Journal of Public Health 65:9, 979-980. Fuchs, V.R. (1974), Who Shall Live? Health Economics and Social Choice, New York: Basic Books, Inc. Galambos, J.T. (1975), "The Course of Alcoholic Hepatitis", in J.M. Khanna, Y. Israel and H. Kalant, eds., Alcoholic Liver Pathology, pp. 97-111, Toronto: Alcoholism and Drug Addiction Research Foundation of Ontario. Galambos, J.T. (1979), Cirrhosis, Toronto: W.B. Saunders Company. Gallagher, R.P. and J.M. Elwood (1979), "Cancer Mortality Among Chinese, Japanese, and Indians in British Columbia, 1964-73", National Cancer Institute Monograph, no. 53, 89-94. Garfinkel, L. (1979), "Cardiovascular Mortality and Cigarette Smoking", in L.M. Ramstrom, ed., The Smoking Epidemic, ! Matter of Worldwide Concern, Proceedings of the Fourth World Conference on Smoking and Health, Stockholm. Glueck, C.J. et al. (1978), "Diet and Coronary Heart Disease: Another View", New England Journal of Medicine 298:26, 1471-1474. Goldsmith, J.R. (1975), "Health Effects of Air Pollution", Basics of RD 4:2, 1-6. Goodman, E (1981), "A Political Smokescreen That Shrouds Health", The Vancouver Sun June 25, A6. Gordon, T. (1981), "Diet and Its Relation to Coronary Heart Disease and Death in Three Populations", Circulation 63:3, 500-515. Gordon, T. et al. (1974), "Death and Coronary Attacks in Men After Giving Up Cigarette Smoking: A Report From the Framingham Study', Lancet ii, 1345. Great Britian (n.d.), "Smoking and Health", Department of Health and Social Security. Greenwald, P. (1980), "Assessment of Risk Factors for Cancer", Preventive Medicine 9, 260-263. Grieg, M. et al. (1980), "A Prospective Study of the Development of Coronary Heart Disease in a Group of 1202 Middle-Aged Men", Journal of Epidemiology and Community Health 34, 23-30. - 256 -Hackland, s. (1976), Smoking Habits of canadians, 1965-1974, Research Bureau, Non-Medical Use of Drugs Directorate, Health Protection Branch, Health and Welfare Canada, Technical Report Series No. 1. Haggerty, R.J. (1977), "Changing Lifestyles to Improve Health", Preventive Medicine 6, 276-289. Hall, E.M. (1980), Canada's National-Provincial Health Program for the 1980s, Ottawa: Health and Welfare Canada. Hallan, J. and I. Linne (1970), "Cirrhosis of the Liver in One Community", in A. Engel and T. Larsson, eds., Alcoholic Cirrhosis and Other Toxic Hepatopathias, pp.336-352, Stockholm: Nordiska Bokhandelns Forlag. Hamby, R.I. (1981), "Hereditary Aspects of Coronary Artery Disease", American Heart Journal 101:5, 639-649. Hamilton, J.L. (1977), "The Cigarette Advertising Ban as a Public Health Policy", working paper no. 5, Wayne Economic Papers, Dept. of Economics, Wayne State University, Detroit, Michigan. Hammond, E.c. (1966), "Smoking in Relation to the Death Rates of One Million Men and Women", National Cancer Institute Monograph No. 19, Epidemiological Study of cancer and Other Chronic Diseases, William Haenszel, ed., pp.127-171, U.S. Dept. of Health, Education, and Welfare, Washington, D.C.: U.S. Government Printing Office. Hammond, E.C. (1968), "Quantitative Relationship Between Cigarette Smoking and Death Rates", National Cancer Institute Monograph No. 28, Toward ~ Less Harmful Cigarette, pp. 3-8, U.S. Dept. of Health, Education, and Welfare, Washington, D.C.: U.S. Government Printing Office. Hammond, E.C. and H. Seidman (1980), "Smoking and Cancer in the United States", Preventive Medicine 9, 169-173. Hammond, E.C. and L. Garfinkel (1980), "General Air Pollution and Cancer in the United States", Preventive Medicine 9, 206-211. Hammond, E.c.; I.J. Selikoff and H. Seidman (1979), "Asbestos Exposure, Cigarette Smoking and Death Rates", Annals New York Academy of Sciences 330, 473-491. Harnes, J.R. (1978), letter to the editor, The New England Journal of Medicine 299:11, 607. Harrington, J.T. (1980), "Alcohol, Arsenic, and (Rapidly) Old Kidneys", Archives of Internal Medicine 140:2, 167-168. Harris, J.E. (1980), "Taxing Tar and Nicotine", The American Economic Review 70:3, 301-311. -Harris, J.E. (1982), "Increasing the Federal Excise Tax on Cigarettes", Journal of Health Economics 1, 117-120. - 257 -Hartroft, w.s. (1975), "On the Etiology of Alooholio Liver Cirrhosis", in J.M. Khanna, Y. Israel and H. Kalant, eds., Alcoholic Liver Pathology, pp. 189-197, Toronto: Alcoholism and Drug Addiction Researoh Foundation of Ontario. Hasin, Y. et al. (1979), "Relationship Between Extent of Coronary Artery Disease and Correlative Risk Factors", American Heart Journal 98:5, 555-561. Haynes, S.G. and M. Feinleib (1980), "Women, Work and Coronary Heart Disease: Prospective Findings from the Framington Heart Study", American Journal of Public Health 70:2, 133-140. Health and Welfare Canada (1976), Canadian Trends in Smoking-Related Diseases: Chronic Bronchitis and Emphysema Mortality, Technical Report Series No. 4, Research Bureau, Non-Medical Use of Drugs Directorate. Health and Welfare Canada (1976), Canadian Trends in Smoking-Related Diseases: Ischemic Heart Disease Mortality, Technical Report Series No. 5, Research Bureau, Non-Medical Use of Drugs Directorate. Health and Welfare Canada (1977), Smoking and Health in Canada, Staff Papers, Long Range Health Planning and Non-Medical Use of Drugs Directorate. Health and Welfare Canada (1976), Canadian Trends in Smoking-Related Diseases: Lung Cancer Mortality, Technical Report Series No. 3, Research Bureau, Non-Medical Use of Drugs Directorate. Health and Welfare Canada (1978), Final Report: Federal-Provincial Task Force .2!!. Alcohol Statistios, Ottawa: Health and Welfare Canada. Health and Welfare Canada (1982), "Disability Attributable to Smoking -Canada, 1978-79", Chronic Diseases in Canada 3:3, 61-62. Heins, M. (1978), letter to the editor, The New England Journal of Medicine 298:22, 1263. Henry, W.T. (1979), "Unhealthy Lifestyles and Their Resulting Health Conditions - An Objective Review of the Purportive Causal Links", unpublished paper. Higgins, M.W. et al. (1977), "Smoking, Socioeconomic Status, and Chronic Respiratory-Disease", American Review of Respiratory Disease 116, 403-410. Higginson, J. (1976), "A Hazardous Society? Individual Versus Community Responsibility In Cancer Prevention", American Journal of Public Health 66:4, 359-366. Higginson, J. (1980), "Proportion of Cancers Due to Occupation", Preventive Medicine 9, 180-188. - 258 -Hirayama, T. (1967), Smoking in Relation to the Death Rates of 265,118 Men and Women in Japan, Tokyo: National Cancer Centre, Research Institute, Epidemiology Division. Hirayama, T. (1975), "Smoking and Cancer: A Prospective Study on Cancer Epidemiology Based on a Census Population in Japan", in J. Steinfeld et al., eds., Smoking and Health. II. Health Consequences, E~ueation, Cessation Activities, and Governmental Action, Proceedings of the Third World Conference on Smoking and Health, pp. 65-72, DHEW (NIH) 77-1413. Hirayama, T. (1981), "Non-Smoking Wives of Heavy Smokers Have a Higher Risk of Lung Cancer: A Study From Japan", British Medical Journal 282:17, 183-185. Hoffman, W.E. (1979), "The Impact of Tobacco Advertising and Promotion on Cigarette Consumption", in L.M. Ramstrom, ed., The Smoking Epidemic, ! Matter of Worldwide Concern, p.p. 249-254, Proceedings of the 4th World Conference on Smoking and Health, Stockholm. Holleb, A.I. (1979), "How Doctors Can Influence Public Attitudes", in L.M. Ramstrom, ed., The Smoking Epidemic, ! Matter of Worldwide Concern, p.p. 260-263, Proceedings of the 4th World Conference on Smoking and Health, Stockholm. Holmes, D.R. et al. (1981), "Coronary Artery Disease in Twins", British Heart Journal 45, 193-197. Holtzman, N.A. (1979), "Prevention: Rhetoric and Reality", International Journal of Health Services 9:1, 25-39. Hrubec, z. et al. (1976), "Background of Angina Pectoris: Social and Environmental Factors in Relation to Smoking", American Journal of Epidemiology 103:1, 16-29. Huff, D. (1974), How to Lie with Statistics, Middlesex, England: Penguin Books. Israel, L. and A.P. Chahinian, eds. (1976), ~Cancer: Natural History, Prognosis, and Therapy, New York: Academic Press. Isselbacher, K.J. and E.R. Feller (1977), "The Natural History and Management of Alcoholic Liver Disease", in M.M. Fisher and J.G. Rankin, eds., Alcohol and the Liver, pp.349-364, New York: Plenum Press. Jaffe, J.H. and M.B. Kanzler (1979), "Smoking as an Addictive Disorder", National Institute of Drug Abuse Research Monograph Series 23, 4-23, January. Jarvik, M.E. et al., eds. (1977), Research£.!! Smoking Behaviour, NIDA Research Monograph 17, Washington, D.C.: Superintendent of Documents, U.S. Government Printing Office. - 259 -Johnson, T.R. (1976), The Ethical Aspects of Government Intervention into Individual Behaviour, p.19, Ottawa: Health and Welfare Canada. Johnson, T.R. (1978), "Additional Evidence on the Effects of Alternative Taxes on Cigarette Prices", Journal of Political Economy 86:2, pt.1, 325-328. Jones, K.R. and T.R. Vischi (1979), "Impact of Alcohol, Drug Abuse and Mental Health Treatment on Medical Care Utilization", Medical Care 17:12, supplement, 3-81. Journal of Public Health Policy (1980), "The Lifestyle Approach to Prevention", editorial, Journal of Public Health Policy, March, 6-9. Kahn, H.A. (1966), "The Dorn Study of Smoking and Mortality Among U.S. Veterans: Report on Eight and One-Half Years of Observation", National Cancer Institute Monograph No. 19, Epidemiological Approaches to the Study of Cancer and Other Chronic Diseases, William Haenszel, ed., pp. 1-27, U.S. Dept. of Health, Education, and Welfare, Washington D.C.: U.S. Governmen~ Printing Office. Kannel, W.B. (1979), "Preventive Trials and Tribulations", American Journal of Epidemiology 109:4, 405-407. Kannel, W.B. (1981), "Update on the Role of Cigarette Smoking in Coronary Artery Disease", American Heart Journal March, 319-328. Kannel, W.B. and D.L. McGee (1979), "Diabetes and Cardiovascular Risk Factors: The Framingham Study", Circulation 59:1, 8-13. Kannel, W.B. et al. (1968), "Cigarette Smoking and Risk of Coronary Heart Disease. Epidemiologic Clues To Pathogenesis. The Framingham Study", National Cancer Institute Monogragh No. 28, Toward .! Less Harmful Cigarette, pp. 9-20, U.S. Dept. of Health, Education, and Welfare, Washington, D.C.: U.S. Government Printing Office. Kannel, W.B. et al. (1979), "An Investigation of Coronary Heart Disease in Families. The Framingham Offspring Study. 11 , American Journal of Epidemiology 110:3, 281-290. Kato, H. et al. (1973), "Epidemiologic Studies of Coronary Heart Disease and Stroke in Japanese Men Living in Japan, Hawaii and California", American Journal of Epidemiology 97:6, 372-385. Keller, M. et al., eds. (1974), Second Special Report to the U.S. Congress £!!. Alcohol and Health, June 1974, U.S. Dept. of Health, Education and Welfare, Washington, D.C.: U.S. Government Printing Office. Kleiman, K.; J.J. Feldman and M.A. Monk (1979), "The Effects of Changes in Smoking Habits on Coronary Heart Disease Mortality", American Journal of .Public Health 69:8, 795-802. - 260 -Klenckner, M.S. Jr. (1960), Cirrhosis of the Liver, Springfield, Illinois: Charles c. Thomas. Kmet, J. and E. Mahboubi (1971), "Oesophageal Cancer Studies in the Caspian Littoral of Iran: Initial Observations", in D.J. Jussawalle and R. Doll, editors, International Seminar ~ Epidemiology of Oesophaseal Caneer Monograph No. 1, 31-44. Kohn, R. and K.L. White (1966), Health Care, Report of the World Health Organization/International Collaborative Study of Medical Care Utilization, London. Kolbye, A.C. (1980), "A U.S. Viewpoint: Legislative and Scientific Aspects of Cancer Prevention", Preventive Medicine 9, 267-274. Korsten, M.A. and c.s. Lieber (1979), "Hepatic and Gastrointestinal Complications of Alcoholism", in J.H. Mendalson and N.K. Mello, eds., The Diagnosis and Treatment of Alcoholism, pp. 20-43, New York: McGraw-Hill. Kramer, H.J. (1979), "Self-Inflicted Disease: Who Should Pay For Care?", Journal of Health Politics, Policy and Law 4:2, 138-141. Kristein, M.M. (1977), "Economic Issues in Prevention", Preventive Medicine 6, 252-264. Kristein, M.M. and D.A. Grove (1978), letter to the editor, The New England Journal of Medicine 299:11, 606-607. Labarthe, D.R. (1979), "Methodologic Variation in Case-Control Studies of Reserpine and Breast Cancer", Journal of Chronic Disease 32, 95-104. ~ Lalonde, M. (1974), ! New Perspective~ the Health of Canadians: ! Working Document, Ottawa: Department of National Health and Welfare, Government of Canada. Lange, D.E. and B.A. Schacter (1977), "Prevalence of Alcohol-Related Admissions to General Medical Units", unpublished paper, University of Manitoba: School of Medicine. Lee, P.N. and L. Garfinkel (1981), "Mortality and Type of Cigarette Smoked", Journal of Epidemiology and Community Health 35, 16-22. Leevy, C .M. ( 1969), "Hard Liquor and Cirrhosis", in A. Engel and T. Larsson, eds., Alcholic Cirrhosis and other Toxic Hepatopathias, pp. 283-295, Stockholm: Nordiska Bokhandelns Forlag. Leevy, C.H.; H. Popper ands. Sherlock (1974), Diseases of the Liver and Biliary Tract, Proceedings of the fifth quadrennial meeting of the International Association for the Study of the Liver, Acapulco: Castle House Publications Ltd. - 261 -Leavy, C.M.; T. Chen and R. Zetterman (1975), "Alcoholic Hepatitis, Cirrhosis, and Immunologic Reactivity", New York Academy of Science 252, 106-115. Lelbach, W.K. (1975), "Cirrhosis in The Alcoholic and its Relation to the Volume of Alcohol Abuse", Annals New York Academy of Science 252, 85-105. Lelbach, W.K. (1975), "Quantitative Aspects of Drinking in Alcoholic Liver Cirrhosis", in J.M. Khanna, Y. Israel and H. Kalant, eds., Alcoholic Liver Pathology, pp. 1-18, Toronto: Alcoholism and Drug Addiction Research Foundation of Ontario. Lelbach, W.K. (1976), "Epidemiology of Alcoholic Liver Disease", in H. Popper and F. Schaffener, eds., Progress in Liver Diseases V.5, pp. 494-515, New York: Grune and Stratton, Inc. Leu, R.E. (1982), Modirying Risk-Taking Behavior Through Public Policy: The Case of Cigarette Smoking, Discussion Paper No.49 B, Institut fuer Sozialwissenschaften der Universitaet Basel, Switzerland, May. Leviton, A. (1973), "Definitions of Attributable Risk", American Journal of Epidemiology 98, 231. Lewinski, T. et al. (1975), "Smoking Habits in Lung Cancer Patients of the Institute of Oncology, Warsaw", in J. Steinfeld et al., eds., Smoking and Health. II. Health Consequences, Education, Cessation Activities, and Governmental Action, Proceedings of the Third World Conference on Smoking and Health, pp.109-113, DHEW (NIH) 77-1413. Lewis, O.D. (1979), "Establishing Smoke-Free Environments", in L.M. Ramstrom, ed., The Smoking Epidemic, ! Matter of Worldwide Concern, p.p. 255-259, Proceedings of the 4th World Conference on Smoking and Health, Stockholm. Lewitt, E.M. and D. Coate (1982), "The Potential for Using Excise Taxes To Reduce Smoking", Journal of Health Economics 1, 121-145. Lieber, c.s. (1975), "Liver Disease and Alcohol: Fatty Liver, Alcoholic Hepatitis, Cirrhosis, and Their Interrelationships", Annals New York Academy of Sciences 252, 63-83. ~ Lieber, c.s. (1976), "The Metabolism of Alcohol", Scientific American 234:3, 25-33. Lieber, c.s. (1978), "Pathogenesis and Early Diagnosis of Alcoholic Liver Injury", New England Journal of Medicine 298, 888-893. Lieber, c.s. and L.M. DeCarli (1974), "An Experimental Model of Alcohol Feeding and Liver Injury in the Baboon", !!.!.. Med Primatol 3, 154. Lindell, S.E. (1979), "Passive Smoking", in L.M. Ramstrom, ed., The Smoking Epidemic, ! Matter of Worldwide Concern, Proceedings of the Fourth World Conference on Smoking and Health, Stockholm. - 262 -Little, R.E. and A.P. Streissguth (1981), "Effects of Alcohol on the Fetus: Impact and Prevention", Canadian Medical Association Journal 125:2, 159-164. Luce, B.R. and s.o. Schweitzer (1978), "Smoking and Alcohol Abuse: A Comparison of Their Economic Consequences", New England Journal of Medicine 298:March 1 569-571. Magnus, K.; A. Matroos and J. Strackee (1979), "Walking, Cycling, or Gardening, With or Without Seasonal Interruption, in Relation to Acute Coronary Events", American Journal of Epidemiology 110:6, 724-733. Mann, G.V. (1977), "Diet-Heart: End of an Era", New England Journal of Medicine 297:12, 644-650. Marks, v. and J. Wright, eds. (1978), Clinics in Endocrinology and Metabolism: Metabolic Effects of Alcohol, London: W.B. Saunders Company Ltd. -Martini, G.A. and C.H. Bode (1970), "The Epidemiology of Cirrhosis of the Liver", in A. Engel and T. Larsson, eds. Alcoholic Cirrhosis and Other Toxic Hepatopathias, pp. 315-335, Stockholm: Nordiska Bokhandelns Forlag. Marvin, M.w. (1976), "Alcohol Problems in Canada: A Summary of Current Knowledge", paper for the Federal-Provincial Working Group on Alcohol-Related Problems, Non-Medical Use of Drugs Directorate, Health and Welfare Canada. Matroos, A.; K. Magnus and J. Stracker (1979), "Fatal and Non-Fatal Coronary Attacks in Relation to Smoking in Some Dutch Communities", American Journal of Epidemiology 109:2, 145-151. Hausner, J.S. and A.K. Bahn (1974), Epidemiology: An Introductory Text, Toronto: W.B. Saunders Company. Maxwell, R.J. (1981), Health and Wealth, Toronto: Lexington Books. McGinnis, J.M. (1982), "Targeting Progress in Health", Prevention 97:4, 295-307. McKeown, T. (1976), The Role of Medicine, London: Nuffield Provincial Hospitals Trust. McMaster University Health Sciences Centre, Department of Clinical Epidemiology and Biostatistics ( 1981), "How to Read Clinical Journals: IV. To Determine Etiology or Causation", Canadian Medical Association Journal 124:April 15, 985-990. McMichael, A.J. (1979), "Cirrhosis as an Index", Proceedings, International Medical Advisory Conference, October 22-24, Melbourne. - 263 -McMillan, G.C. (1973), "Evidence for Components Other Than Carbon Monoxide and Nicotine as Etiological Factors in Cardiovascular Disease", in Smoking and Health: .!!_Modifying the Risk for the Smoker, P•P• 363-367, DHEW 76-1221. Mendelson, J.H. and N.K. Mello (1979), The Diagnosis and Treatment of Alcoholism, Toronto: McGraw Hill. Mattlin, c. (1979), "Prerequisites of Successful Lifestyle Intervention", Social Science and Medicine 13A, 559-562. Mezey, E. et al. (1975), "Hepatic Fibrogenesis in Alcoholism", in J.M. Khanna, Y. Israel and H. Kalant, eds., Alcoholic Liver Pathology, pp. 145-156, Toronto: Alcoholism and Drug Addiction Research Foundation of Ontario. Miettinen, o.s. et al. (1976), "Cigarette-Smoking and Nonfatal Myocardial Infarction: Rate Ratio in Relation to Age, Sex and Predisposing Conditions", American Journal of Epidemiology 103:1, 30-36. Millar, W.J. (1983), Smoking Behaviour of Canadians 1981, Ministry of Supply and Social Services Canada, Statistics Canada Catalogue No. H39-66/1983E. Miller, A.B. (1977), "Recent Trends in Cancer Mortality in Canada", Canadian Medical Association Journal 116, 28-30. Miller, A.B. (1980), "Identification of Adults at High Risk of Lung Cancer", Canadian Medical Association Journal 122, 985-987. Ministry of Health (1976), Bibliography: of Selected Information~ Education and Prevention of Alcohol and Drug Misuse, and Teenage Drinking, Province of British Columbia: The Alcohol and Drug Commission Library. Mitchell, P.I. et al. (1980), "Role of Alcohol in the Aetiology of Hypertension", Medical Journal of Australia 2:4, 198-200 • . Mitchell, R.S. et al. (1976), "The Morphologic Features of the Bronchi, Bronchioles and Aveoli in Chronic Airway Obstruction: A Clinicopathologic Study", American Review of Respiratory Disease 114:1, 137-145. Moeschlin, s. and P. Righetti (1969), "Wine and Cirrhosis", in A. Engel and T. Larsson, eds., Alcoholic Cirrhosis and Other Toxic Hepatopathias, pp. 301-311, Stockholm: Nordiska Bokhandelns Forlag. Mosher, J .. F. (1983), "Tax-Deductible Alcohol: An Issue of Public Health Policy and Prevention Strategy", Journal of Health Politics, Policy and Law 7:4, 855-888. Mosteller, F. and J.W. Tukey (1977), Data Analysis and Regression: A Second Course in Statistics, Massachusetts: Addison-Wesley Publishing Company. - 264 -Mulvihill, J.J. (1976), "Host Factors in Human Lung Tumors: An Example of Ecogenetics in Oncology", National Cancer Institute Monograph No. 52, 115-121. Neubauer, D and R.Pratt (1981), "The Second Public Health Revolution: A Critical Appraisal", Journal of Health Politics, Policy and Law 6:2, 205-228. Oakes, T.W.; G.D. Friedman and c.c. Seltzer (1973), "Mail Survey Response by Health Status of Smokers, Nonsmokers, and Ex-Smokers", American Journal of Epidemiology 98:1, 50-55. Ontario (1978), Hospital Statistics 1978-1979, Toronto: Ministry of Health. Ontario Council of Health (1982), Smoking and Health in Ontario: ! Need for Balance, Report of the Task Force on Smoking, Toronto: Ontario Council of Health. Ouellet, B.L.; J.M. Remeder and J.M. Lance (1979), "Premature Mortality Attributable to Smoking and Hazardous Drinking in Canada", American Journal of Epidemiology 109:4, 451-463. Pauly, M.V. (1969), "A Measure of the Welfare Cost of Health Insurance", Health Services Research, Winter. Peguignot, G. and A. Tuyns (1975), "Rations d 1Alcool Consommees 'Declarees• et Risques Pathologiques", Anglo-French Symposium on Alcoholism 1-15, Paris, Inserm. Peguignot, G., A. Tuyns and J.L. Berta (1978), "Ascitic Cirrhosis in Relation to Alcohol Consumption", International Journal of Epidemiology 7, 113-120. Peterson, D.R. (1981), "The Sudden Infant Death Syndrome - Reassessment of Growth Retardation in Relation to Maternal Smoking and the Hypoxia Hypothesis", American Journal of Epidemiology 113:5, 583-589. Peto, J. (1979), "The Health Effects of Smoking in Women", in L.M. Ramstrom, ed., The Smoking Epidemic, ! Matter of Worldwide Concern, Proceedings of the Fourth World Conference on Smoking and Health, Stockholm. Pinney, J.M. (1979), "Governmental Policies on Smoking and Health', in L.M. Ramstrom, ed., The Smoking Epidemic, ! Matter of Worldwide Concern, Proceedings of the 4th World Conference on Smoking and Health, Stockholm. Piper, G. W. ( 1981), "Health Risks to Women Who Smoke", Canadian Medical Association Journal 124, 1250-1251. Policy Research and Management Services Branch, Saskatchewan Health (1981), Opportunities for Prevention and Promotion in the 1980•s, Report of the Task Force, April. - 265 -Pomerleau, o. et al. (1975), "Role of Behavior Modification in Preventive Medicine", New England Journal of Medicine 292:24, 1277-1282. Popham, R.E. (1956), "The Jellinek Alcoholism Estimation Formula and its Application to Canadian Data", Quarterly Journal of Studies of Alcohol 17, 559-593· Popper, H. (1976), "The Pathogenesis of Alcoholic Cirrhosis", in M.M. Fisher and J.G. Rankin, eds., Alcohol and the Liver, pp. 289-307, New York: Plenum Press. Popper, H. and F. Schaffner (1970), "Structural Studies in Alcohol and Drug Induced Liver Injury", in A. Engel and T. Larsson, eds., Alcoholic Cirrhosis and Other Toxic Hepatopathias, pp. 15-37, Stockholm: Nordiska Bokhandelns Forlag. Potter, B.J. et al. (1980), "Experimental Production of Growth Retardation in the Sheep Fetus After Exposure to Alcohol", The Medical Journal of Australia 2, 191-193. Quelch, J.A. (1978), Marketing and Preventive Health Care: Towards a Broader Perspective, Working Paper Series No.200Rs, School of Business Administration, Univ. of Western Ontario. Ramstrom, L.M., ed. (1979), The Smoking Epidemic: ! Matter of Worldwide Concern, The Fourth World Conference on Smoking and Health, Stockholm: Almqvist & Wiksell International. Rankin, J.G. et al. (1975), "Epidemiology of Alcoholic Liver Pathology", in J.M. Khanna, Y. Israel and H. Kalant, eds., Alcoholic Liver Pathology, Toronto: Alcoholism and Drug Addiction Research Foundation of Ontario. Reese, L. and P. Uksik (1981), "Radioimmunoassay of Serum Myoglobin in Screening for Acute Myocardial Infarction", Canadian Medical Association Journal 124, 1585-1588. Reid, D.D. (1973), "Epidemiological Studies in the United Kingdom on Smoking and Coronary Heart Disease", in Smoking and Health: .!!_ Modifying the Risk For the Smoker, p.p. 275-279, DHEW 76-1221. Rigby, N. (1982), letter to the editor, The Province January 8, B1. Rimer, I. (1979), "New Approaches to the Mass Media", in L.M. Ramstrom, ed., The Smoking Epidemic, ! Matter of Worldwide Concern, p.p. 267-271, Proceedings of the 4th World Conference on Smoking and Health, Stockholm. Rix, K.J.B. (1977), Alcohol and Alcoholism, Montreal: Eden Press. Robinette, C.D. et al. (1979), "Chronic Alcoholism and Subsequent Mortality in World War II Veterans", American Journal of Epidemiology 109:6, 687-700. - 266 -Rogers, I.; G. Piper and G. Kleisinger (1980), "Smoking and Health", report of the task force on smoking and health, Policy Research and Management Services Branch, Saskatchewan Health. Rogot, E. (1978), "Smoking and Life Expectancy in U.S. Veterans", American Journal of Public Health 68:10, 1023-1025. Room, R. (1977), "Measurement and Distribution of Drinking Patterns and Problems in General Populations", in Edwards et al., eds., Alcohol-Related Disabilities, Geneva: World Health Organization. Rose, G. and P.J.S. Hamilton (1978), "A Randomised Controlled Trial of the Effect on Middle-Aged Men of Advice to Stop Smoking", Journal of Epidemiology and Community Health 32, 175. Rose, G. and M.G. Marmot (1981), "Social Class and Coronary Heart Disease", British Heart Journal 45, 13-19. Rosen, c. and H.J. Ashley (1978), "Smoking and the Health Professional: Recognition and Performance of Roles", Canadian Journal of Public Health 69:5, 399-406. Rosenberg, L. et al. (1980), "Cigarette Smoking in Relation to the Risk of Myocardial Infarction in Young Women. Modifying Influence of Age and Predisposing Factors", International Journal of Epidemiology 9:1, 57-63. Rothschild, M. and J.E. Stiglitz (1976), "Equilibrium is Competitive Insurance Markets: An Essay on the Economics of Imperfect Information", Quarterly Journal of Economics 90:4. Royal College of Physicians (1977), Smoking£!'.: Health: The Third Report from the College of Physicians of London, Kent, England: Pitman Medical. Russell, M.A.H. (1979), "Smoking as a Dependence Disorder", in L.M. Ramstrom, ed., The Smoking Epidemic, ! Matter of Worldwide Concern, Proceedings of the Fourth World Conference on Smoking and Health, Stockholm. Ryan, w. (1971), Blaming the Victim, New York: Vintage Books. Sackett, D.L. (1979), "Bias in Analytic Research", Journal of Chronic Disease 32, 51-63. Sackett, D.L. et al. (1968), "Relation Between Aortic Atherosclerosis and the Use of Cigarettes and Alcohol", New England Journal of Medicine 279:26, 1413-1420. Sapolsky, H.M. (1980), "The Political Obstacles to the Control of Cigarette Smoking in the United States", Journal of Health Politics, Policy and Law 5:2, 277-290. ~ Saunders, J.B. et al. (1981), "A 20-year Prospective Study of Cirrhosis", British Medical Journal 282:Jan. 28, 263-266. - 267 -Scheffler, R.M. and L. Paringer (1980), "A Review of the Economic Evidence on Prevention", Medical care XVIII:5, 473-484. Schelling, T.C. (1978), Micromotives and Macrobehavior, New York: Norton. Schmidt, w. (1975), "Agreement, Disagreement in Experimental, Clinical and Epidemiological Evidence on the Etiology Of Alcoholic Liver Cirrhosis: A Comment", in J.M. Khanna, Y. Israel and H. Kalant, eds., Alcoholic Liver Pathology, pp. 19- 30, Ontario: Alcoholism and Drug Addiction Research Foundation of Ontario. Schmidt, w. (1976), "Some Comments on Alcohol Control Policies with Particular Reference to Canada", in Epidemiology: Public Policy and Alcohol Problems in Canada, pp. 43-57, Health and Welfare Canada. Schmidt, w. (1977), "The Epidemiology of Cirrhosis of the Liver: A Statistical Analysis of Mortality Data With Special Reference to Canada", in M.M. Fisher and J.G. Rankin, eds., Alcohol and the Liver, Proceedings of the Third International Symposium, pp. 1-30, New York: Plenum Press. Schmidt, w. (1980), "Effects of Alcohol Consumption on Health", Journal of Public Health Policy March, 25-40. Schmidt, w. and J. de Lint (1969), "Mortality Experiences of Male and Female Alcoholic Patients", Quarterly Journal of Studies ~Alcohol 30, 112-118. Schneiderman, M.A. and D.L. Levin (1975), "Epidemiology of Cancer and Tobacco Use: Trends and Trend Indicators", in J. Steinfeld et al., eds., Smoking and Health. II. Health Consequences, Education, Cessation Activities, and Governmental Action, Proceedings of the Third World Conference on Smoking and Health, pp. 73-84, DHEW (NIH) 77-1413. Schneiderman, M.A.; P. Decoufle and c.c. Brown (1979), "Thresholds for Environmental Cancer: Biologic and Statistical Considerations", Annals New York Academy of Science 329, 92-130. Seltzer, c.c. (1972), "Critical Appraisal of the Royal College of Physicians' Report on Smoking and Health", The Lancet January 29, 243-248. -Seltzer, c.c. (1980), "Smoking and Coronary Heart Disease: What Are We To Believe?", American Heart Journal 100:3, 275-280. Senior, S.L. (1982), "Study of Smoking Habits in Hospital and Attitudes of Medical Staff Towards Smoking", Canadian Medical Association Journal 126:January, 131- 133. Shepard, D.S. (1980), Incentives for not Smoking: Experience at the Speedcall Corporation, ! Preliminary Report, Center for the Analysis of Health Practices, Harvard School of Public Health. - 268 -Shephard, R.J. (1982), "Some Hazards of Passive Smoking", Canadian Family Physician 28:February, 299-302. Sherman, R.M.; S. Reiff and A.B. Forsythe (1979), "Utilization of Medical Services by Alcoholics Participating in an Outpatient Treatment Program", Alcoholism: Clinical and Experimental Research 3:2, 115-120. -Shillinglaw, G. (1968), "The Concept of Attributable Cost", in D. Solomons, ed., Studies in Cost Analysis, pp. 134-147, Richard D. Irwin Inc. Shillington, E.R. (1977), Selected Economic Consequences of Cigarette Smoking, Monograph Series No. 2, Health and Welfare Canada, December. Simanis, J.G. and J.R. Coleman (1980), "Health Care Expenditures in Nine Industrialized Countries, 1960-76 11 , Social Security Bulletin 43:1, 3-8. Singer, M. (1978), "How to Reduce Risks Rationally", The Public Interest 51:Spring, 93-112. Skog, O.J. (1979), "Liver Cirrhosis Mortality as an Indicator of Prevalence of Heavy Alcohol Use: Some Methodological Problems", mimeograph no. 22, Oslo: National Institute for Alcohol Research, ISBN 82-7171-027-3. Skog, O.J. (1980), "Liver Cirrhosis Epidemiology: Some Methodological Problems", British Journal of Addiction 75, 227-243. Slaby, A. and L. Tancredi (1975), Collusion for Conformity, New York: Jason Aronson Inc. Smith, G.C. et al. (1981), "Sickness Absence, Respiratory Impairment and Smoking in Industry: An Australian Study", The Medical Journal of Australia 1, 235-237. Smith, R. (1981), "Alcohol, Women, and the Young: The Same Old Problem", British Medical Journal 283, 1170-1172. Smith, R. (1981), "The Relation Between Consumption and Damage", British Medical Journal 283, 896-898. Sokol, R.J. (1981), "Alcohol and Abnormal outcomes of Pregnancy", Canadian Medical Association Journal 125, 134-148. Somers, A.R. (1980), "Life-style and Health", Chapter 26 in J.M. Last, ed., Public Health and Preventive Medicine 11th edition, pp. 1046-1065, New York: Appleton-Century-Crofts. Spritz, N. (1979), "Appraisal of Alcohol Consumption as a Causative Factor in Liver Disease and Atherosclerosis", The American Journal of Clinical Nutrition 32, 2654-2658. - 269 -Spritz, N. (1979), "Review of the Evidence Linking Alcohol Consumption with Liver Disease and Atherosclerotic Disease", The American Journal of Clinical Nutrition 32, 2734-2738. ~ Stallones, R.A. (1980), "The Rise and Fall of Ischemic Heart Disease", Scientific American 243:5, 43-49. Stamler, J . (1978), "Lifestyles, Major Risk Factors, Proof and Public Policy", Circulation 58:1, 3-19. Stamler, J. (1979), "Research Related to Risk Factors", Circulation 60:7, 1575-1587. Sterling, T.R. (1975), "A Critical Reassessment of the Evidence Bearing On Smoking as the Cause of Lung Cancer", American Journal of Public Health 65:9, 939-953. Sterling, T.R. (1978), "Does Smoking Kill Workers or Working Kill Smokers? OR The Mutual Relationship Between Smoking, Occupation, and Respiratory Disease", International Journal of Health Services 8:3, 437-452. Susser, M. (1973), Causal Thinking in the Health Sciences: Concepts and Strategies of Epidemiology, New York: Oxford University Press. Tager, I.B. et al. (1979), "Effect of Parental Cigarette Smoking on the Pulmonary Function of Children", American Journal of Epidemiology 110:1, 15-26. Takeuchi, J. and A. Takada (1975), "Alcohol and its Hepatotoxic Effect", in J.M. Khanna, Y. Israel and H. Kalant, eds., Alcoholic Liver Pathology, pp. 199-224, Toronto: Alcoholism and Drug Addiction Research Foundation of Ontario. Terris, M. (1967), "Epidemiology of Cirrhosis of the Liver: National Mortality Data", American Journal of Public Health 57:12, 2076-2088. ~ Tesh, s. (1981), "Disease Causality and Politics", Journal of Health Politics, Policy and Law 6:3, 369-390. Thelle, D.S. and O.H. Forde (1979), "The Cardiovascular Study in Finnmark County: Coronary Risk Factors and the Occurrence of Myocardial Infarction in First Degree Relatives and in Subjects of Different Ethnic Origin", American Journal of Epidemiology 110:6, 708-715. Thompson, M.E. and W.F. Forbes (1982), "Costs and Benefits of Cigarette Smoking in Canada", Canadian Medical Association Journal 127:November 1. Thurlbeck, W.M. (1974), "Chronic Bronchitis and Emphysema- The Pathophysiology of Chronic Obstructive Lung Disease", Basics of RD, 3:1, 1-6. - 270 -Toufexis, A. (1981), "Taming the No.1 Killer", Time, June 1, 44-55. Townsend, J.L. and T.W. Meade (1979), "Ischaemic Heart Disease Mortality Risks for Smokers and Non-Smokers", Journal of Epidemiology and Community Health 33, 243-247. Turner, T.B. et al. (1977a), "Measurement of Alcohol-Related Effects in Man: Chronic Effects in Relation to Levels of Alcohol Consumption. Part A", The Johns Hopkins Medical Journal 141:5, 235-248. Turner, T.B. et al. (1977b), "Measurement of Alcohol-Related Effects in Man: Chronic Effects in Relation to Levels of Alcohol Consumption. Part B", The Johns Hopkins Medical Journal 141:6, 273-286. Turpeinen, o. (1979), "Effect of Cholesterol-Lowering Diet on Mortality from Coronary Heart Disease and Other Causes", Circulation 59:1, 1-7. Tweed, W.A. (1982), "Does Cigarette Smoking Cause Coronary Heart Disease?", Canadian Medical Association Journal 126:Jan. 15, 102-107. Tytherleigh, M. (1983), editorial, The Province, January 5. U.S. Department of Health and Human Services (1982), The Health Consequences of Smoking. Cancer., a report of the Surgeon General, Washington, D.C.: U.S. Government Printing Office. U.S. Department of Health, Education, and Welfare (1971), The Health Consequences of Smoking, a report of the Surgeon General, Washington, D.C.: U.S. Government Printing Office. U.S. Department of Health, Education and Welfare (1978), Alcohol and Health, Washington, D.C.: U.S. Government Printing Office. u.s. Department of Health, Education, and Welfare (1979), Smoking and Health, a report of the Surgeon General, Washington, D.C.: U.S. Government Printing Office. Veatch, R.M. (1980), "Voluntary Risks to Health, The Ethical Issues", Journal of the American Medical Association 243:1, 50-55. Vladeck, B.c. (1975), "Policy Alternatives for Alcohol Control", American Journal of Public Health 65:12, 1340-1342. Wald, N.J. and A. Bailey (1981), "The Importance of Tar and Nicotine in Determining Cigarette Smoking Habits", Journal of Epidemiology and Community Health 35, 23-24. Walter, S.D. (1976), "Methodological Developments in the Use of Attributable Fraction For Health Priorities and Strategies in Canada", staff papers, Long Range Health Planning, Health and Welfare Canada. _,<j - 271 -Warner, K.E. (1977), "The Effects of the Anti-Smoking Campaign on Cigarette Consumption", American Journal of Public Health 67:7, 645-50. ~ Warner, K.E. (1979), "The Economic Implications of Preventive Health Care", Social Science and Medicine 13C:4, 227-237. Watson, C.J. (1970), "Forms of Hepatic Cirrhosis: Variations in Basic and Clinical Features and in Factors Affecting Prognosis", in A. Engel and T. Larsson, eds., Alcoholic Cirrhosis and Other Toxic Hepatopathias, pp. 219-232, Stockholm: Nordiska Bokhandelns Forlag. Weale, A. (1983), "Invisible Hand or Fatherly Hand? Problems of Paternalism in the New Perspective in Health", Journal of Health Politics, Policy and Law 7:4, 784-807. Weber, A.; c. Jermini and E. Grandjean (1976), "Irritating Effects on Man of Air Pollution Due to Cigarette Smoke", American Journal of Public Health 66:7, 672-676. ~ Weiss, N.S. (1981), "Inferring Causal Relationships, Elaboration of the Criterion of 'Dose-Response'", American Journal of Epidemiology 113:5, 487-492. Weiss, W. (1978), "Lung Cancer Mortality and Urban Air Pollution", American Journal of Public Health 68:8, 773-775. White, J.R. and H.F. Froeb (1980),- "Small-Airways Dysfunction in ; Nonsmokers Chronically Exposed to Tobacco Smoke", New England Journal of Medicine 302:13, 720-24, 742-43. Wigle, D.T. (1977), "The Distribution of Lung Cancer in Two Canadian Cities", Canadian Journal of Public Health 68, 463-468. Wigle, D.T. et al. (1980), "Relative Importance of Smoking as a Risk Factor for Selected Cancers", Canadian Journal of Public Health 71, 269-275. Wikler, D.I. (1978), "Persuasion and Coercion for Health, Ethical Issues in Government Efforts to Change Life-Styles", Millbank Memorial ~ Quarterly/Health and Society 56:3, 303-338. Wilkinson, R. et al. (1978), "Health Costs of Alcohol and Tobacco: Who Pays?", New England Journal of Medicine 298:22, 1262-1263. Willett, w. et al. (1961), "Cigarette Smoking and Non-Fatal Myocardial Infarction in Women", American Journal of Epidemiology 113:5, 575-582. Wilson, D. et.al. (1982), "FAMILY PRACTICE - Randomized clinical trial supportive follow-up for cigarette smokers in a family practice", Canadian Medical Association Journal, 126:January, 126-129. 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items