Tobacco Documents Online

Volume 64 Numhcr 1 1986 THEMILBANK ~UARTERLY rnrc (:usts ol thc tlcalth E/lccts ttl Smoking, I')}{ri 481) P. IZI(-ti, THOMAS A. I'IOIX,tit)N, hPTIiR SINtiNHIMLR, iROWNI:R, and ANI)RISA N. (t01'tif1i1N iiMO Care: Issues and Options in Setting Capitation 548 '. ANI)I:RtiUN, ISARI. P. STI:INBIiR(:, JAMIiti IIOI.LOWAY, C. CANTOR tinl; for Liiascd S<It:ction in Ilr:altlt losurante 566 ;. LUIT yuity: Swedish Iicalth Policy and the Private Seuur 592 J M. {KOS(iNT11At. t ul Malingrring: Why Individuals Wincdraw linln Work csr:ncc of Illness 622 Yt:1.1N '.ty in Unsalr Jobs: Thcury, IiviJcnce, and 1'uli(y uns (i5v }.~ . K()IiINti()N ( AMRRIIH,1 UNIVI Rtilll' I'RI'+1 i! l-asr 5,01 5tnm, New Yntk, NY 11102., U S A Thr Prtt BuilJinK, TrumpmKtun Street, (:atnhrydge (.132 2RU, F.nKland Itl Surnlurd RuaJ, Uakletgh, Mellwurnc flG(i, Ausrraha Prtnted m the (Inned Stnes of Amenta I

Beyond Equity: Swedish Health Policy and the Private Sector 592 MARILYNN M. ROSENTHAL The rapid development of private medicine in Sweden poses a chal- lenge to that nation's traditional commitment to equity and a pub- lic system of health care. Economic constraints in public-sector spending, physician intent in private practice, popular interest in individual choice, and criticism of the public system have been the stimulus. The dominant Social Democrats will likely accept those elements of the private sector that are ideologically compatible, but resist others. The Myth of Malingering: Why Individuals Withdraw from Work in the Presence of Illness EDWARD YELIN Federal policy to provide income to workers disabled by chronic illness has been under attack. Data from the Social Security Ad- ministration Survey of Disability and Work reveal that anticipation of higher levels of disability income does not correlate with cessa- tion of work; the nature of illness and the structure of work do. Policies predicated on the myth of malingering will hurt millions of the disabled chronically ill. 622 Hazard Pay in Unsafe Jobs: Theory, Evi- dence, and Policy Implications 65o JAMES C. ROBINSON Critics of OSHA argue that an unregulated labor market gives firms incentives to improve working conditions. Analysis of the relation between wages and hazardous working conditions confirms that workers in hazardous jobs are paid marginally more than com- parable workers in safe jobs. But hazardous occupations are concen- trated in low-skill and low-pay strata. The empirical findings have important implications for "right-to-know" and related occupa- tional health strategies. The Economic Costs of the Health Effects of Smoking, 1984 DOROTHY P. RICE,' THOMAS A. HODGSON,2 PETER SINSHEIMER,3 WARREN EROWNER,' and ANDREA N. KOPSTEINZ ' Univerrity of Catifornia, San Francisco; Z Nationa! Center for Health Stati.rtjcs; 3 San Diego State University, San Diego C IGARETTE SMOKING IS A MAJOR CAUSE OF MORBIDITY and mortality in the United States today. It has been linked to a variety of illnesses, including heart disease, cancer, and respiratory disease. Increasing public awareness of the health risks associated with smoking has led to a decline in the proportion of adults who smoke. Yet, as of 1985, 33 percent of men and 28 percent of women smoked. Although there has been a decline in smoking in recent years, the proportion of adult male smokers who smoke 25 cigarettes or more a day has increased from 24 percent in 1965 to 30 percent in 1985; for women, the proportion increased from 13 percent to 21 percent (National Center for Health Statistics 1985, 73; 1986). The health hazards of cigarette smoking have been well documented. More than twenty years ago, the Report of the Advisory Committee to the Surgeon General of the Public Health Service was published (U.S. Public Health Service 1964). That report and a series of subsequent reports The Milbank Quarterly, Vol. 64, No. 4, 1986 ® 1986 Milbank Memorial Fund 489 I

490 1). P. Kiie et al, The Etononrii (:ortr of rhe Health i ffedt of Smokin,q, 1984 491 of the surgeon general reviewed the major prospective epidemiologic studies in the United States and abroad that established the relation between smoking and various illnesses. Recently, annual authoritative reports have been released by the surgeon general on The Health ConJeqaencu of Smoking in which cigarette smoking and its relation to cancer (1982), cardiovascular diseases (1983), and chronic obstructive lung disease (1984) were extensively reviewed. The 1985 report presented a comprehensive review of the relation between cigarette smoking and cancer and chronic lung disease in the work place (U.S. Public Health Service, 1982, 1983, 1984, 1985). In addition to the health risks of smoking, there are important economic consequences. A complete assessment of the economics of smoking requires evaluation of various health, economic, and intangible parameters, including benefits as well as costs of both the production and consumption of tobacco. In many respects the purchase and con- sumption of tobacco is similar to most other commodities and services purchased in the market place. Expenditures for purchasing tobacco cover the cost of resources used in the production process, profit, and taxes. In return, smokers obtain a certain amount of enjoyment. Thus, to a certain extent, smokers get their money's worth and the cost of resources going into the production of tobacco is offset by the benefits of tobacco consumption to smokers. On the other hand, smokers may not have complete knowledge of the harmful health effects of smoking (Warner 1985); although they know smoking is hazardous they are addicted and unable to quit, and may not consider external effects such as annoyance to nonsmokers or the cost of medical care paid `py others. In this situation, costs of smoking other than the purchase price are not fully reflected in the decision process, and benefits to smokers may be less than the combined costs to smokers and nonsmokers. In this article we focus on costs resulting from the health effects of smoking: expenditures for medical care and the value of productive output lost to morbidity, disability, and premature mortality among smokers. These are important components of an analysis of the economics of smoking. Among smokers who know smoking is hazardous to health, the prospect of quitting may be painful, and continued smoking may have become a means of avoiding the physical and psychological discomforts of withdrawal. The costs of purchasing this tobacco is not offset by the benefits of enjoyment from smoking; these costs can be considered in addition to the health effects of smoking, but they art nor quantified in this article. This article reviews alternative perspectives and studies of the economic costs of the health effects of smoking, quantifies the magnitude of the costs to the economy by employing a prevalence-based analysis, and compares our findings with those of other researchers. The appendix describes the methodology of attributable risk used in the cost estimates. Alternative Perspectives of Costs of Smoking Two distinct methodologies exist for evaluating illness and disease in economic terms, the human capital and willingness-to-pay approaches. The former method, used in this study, is called the human capital approach because an employed person is seen as producing a stream of output over the years that is valued at the individual's earnings (Rice, Hodgson, and Kopstein 19R5). The willingness-to-pay method values human life according to the amount people are willing to spend to obtain reductions in the probability of death (Schelling 1968; Acton 1975). The relative merits of these two models is a subject of continuing debate (Robinson 1986). Even within the context of human capital methods employed in this article, there are alternative perspectives from which to view the costs of smoking. Two essential characteristics that distinguish perspectives involve different views on the answer to the question "costs to whom?" and the temporal relation between smoking and costs. Costs to Whom? A prevailing view is that the costs of illness to all of society, smokers as well as nonsmokers, and the indirect costs of morbidity and mortality are the value of an individual's total output, without deductions for consumption (Mishan 1971). In accord with this view, cost-of-illness estimates measure the value of resources used (direct costs) and lost (indirect costs) and the total output lost as a result of illness or death is the value forgone. T.C. Schelling (personal communication, February 15, 1984) suggests, on the other hand, that "costs to whom?" requires two different answers: (1) costs to those afflicted by illness and disease, and (2) costs to everyone else, with the monetary impact on others being important because of the welfare gains and losses they entail for the parties to the transactions. Studies may differ in their analyses I

492 D.P. Rice et al. The Economic Co.rtl of the Health Gffects of Smoking, 1984 4493 of who gains and loses and the amount of benefits and costs to various parties. • The essential distinction between these two views is that the former counts only the value of resources used resulting in forgone alternatives, and resources lost in terms of unemployed labor, while the latter also investigates transfers of resources from one segment of society to another. We are concerned in this article with certain economic costs of the health effects of smoking, including the value of resources used to provide medical c-are and the value of labor forgone due to morbidity, disability, and premature mortality. The costs estimated are in accord with the first of the two perspectives outlined above. The distribution of a given level of output between consumption and savings and the amount of reallocation of one's output to other members of society is a function of social welfare, fiscal and monetary policy, and other means available to policy makers. The relative shares going to the ill or deceased individuals versus the rest of society are determined by the current economic policies and incentives and are a separate issue. We should keep in mind, however, that tobacco consumption and accompanying health effects, in concert with the institutional framework of the society, confer monetary benefits on one group through the imposition of monetary costs on another. On average, current and former smokers use more medical care, experience more work-loss days, and have higher mortality rates than persons who have never smoked. Although a smoker may suffer from smoking-induced illness and require medical care, the cost of the treatment may be borne, at least in part, by others. This occurs, for example, when medical care for smoking-related diseases is paid by health insurance funded by premiums collected from both other smokers and nonsmokers, or by public expenditures such as Medicare and Medicaid. Similar considerations apply to indirect costs. If a smoker loses time from work due to sickness, the real cost is the value of labor not productively employed. The monetary cost of the day lost from work may be borne in whole or in part by the sick worker and dependents, other employees, the employer, or the rest of society. The worker and dependents bear the cost of absences not covered by paid sick leave, other employees may incur costs in the form of lower wages in order to fund sick leave benefits, employers face higher costs for sick leave and additional labor costs or reduced output, and the society as a whole may have to pay higher prices to cover higher costs E of production and lose tax revenues on income lost by the sick worker. Premature mortality presents a similar situation, although the time horizon is years instead of days. Output lost is a real economic cost. There are also pecuniary transfers, including taxes forgone on income lost by the deceased, Social Security and pension benefits paid to survivors, and Social Security and pension payments forgone by the deceased to the benefit of surviving smokers and nonsmokers. The deleterious health effects of smoking generate a variety of financial flows in addition to economic costs. These financial flows have distributional effects, transferring control over the use of resources from one group to another, affecting behavior, and changing the relative well-being of individuals. Although outside the scope.of this article, which is confined to estimates of resource costs and losses, transfers such as health insurance premiums and payments, Social Security, pension, si4kness payments and benefits are important economic values in the social decision-making process. Knowledge of who benefits and who pays and the magnitudes of benefits and costs to various parties can assist in determining the societal response to smoking activities. It might be important to know, for example, the impact of smoking on Scx ial Security, Medicare payments, etc. (Office of Technology Assessment 1985). Some additional aspects of transfer payments are discussed in the section on types of costs. Finally, it is important not to view the issues in terms of smokers versus nonsmokers. Most deleterious health effects of smoking are self-inflicted on smokers by their consumption of tobacco, although there are possible health effects of passive smoking. On the other hand, economic costs and transfer payments occasioned by smoking- induced disease are shared in varying amounts by ill smokers, nonill smokers, and nonsmokers. For example, although the ill smoker receives medical care covered by health insurance, it is financed by premiums paid by ill smokers, nonill smokers, and nonsmokers. Temporal Relation between Smoking and Costs A second essential characteristic that distinguishes perspectives is the temporal relation between smoking and measured costs. Smoking presents a dynamic, time-dependent phenomenon. Some costs of smoking, such as the annoyance caused nonsmokers and property damage from smoking-related fires, are coincident in time with the I

494 D.I'. Rue el al. purchase and consumption of tobacco. The most important costs of smoking in terms of magnitude of their impact are smoking-related diseases and the attendant morbidity, mortality, medical care costs, indirect losses, and intangible losses from pain, suffering, and other quality-of-life changes. These effects result from cumulative exposure over many years and are far removed and distant in time from the tobacco use that helps cause them. In this article we present an example of a prevalence-based cost- of-illness analysis in which the current toll of direct and indirect economic costs resulting from prior smoking is estimated. That is, the health care expenditures incurred and value of economic output lost in 1980 as a result of past smoking over many years are calculated. Prevalence-based cost-of smoking estimates measure the amounts spent during a year and the value of lost economic output for deleterious health effects manifest during the year, but caused by exposure to tobacco over many previous years. In addition to knowing the current annual burden of past smoking (prevalence costs), it is important to know the future costs likely to result from current levels of smoking (incidence costs) and the reductions in costs to be expected from reductions in smoking. Prevalence costs indicate the maximum annual value of resources that could be gained for other uses as levels of smoking decrease. Even with complete and immediate cessation of all smoking, it would be a number of years before morbidity and mortality rates of former smokers returned to levels comparable to those of persons who never smoked. The total amount saved would be the sum of a series of annual reductions which rise over time to a maximum level. Examples of prevalence- and incidence-based studies of the health effects of smoking are described in a later section. . A related issue is the possible tradeoff between higher than average annual medical care use by and expenditures for smokers and longer life expectancy and additional years of medical care for nonsmokers (Leu and Schaub 1983). To the extent that smokers die prematurely, higher medical care expenditures for smoking-induced disease during the smoker's lifetime are offset to a certain degree by expenditures that would be incurred in future years if the smoker did not smoke and enjoyed longer life (Institute of Medicine 1981). The quantitative nature of total versus net direct costs of smoking, however, remains to be rigorously analyzed, and the conceptual validity of net direct The Eronomic Corr.r o j the Healeh Effecl.r of Smoking, 1984 495 costs in certain applications has been questioned (Russell 1986; Warner and Luce 1982). Types of Cost The different types of smoking costs are briefly described below. Direct Cart.r Direct costs of medical care (hospital and nursing home care, services of health practitioners, drugs, etc.) to treat diseases related to smoking result largely from illness self-inflicted on smokers by their consumption of tobacco. The costs of care of nonsmokers exposed to and ill from tobacco smoke are also included. Other direct costs of smoking include costs of cleaning clothes and air of smoke, repairing and replacing articles damaged by cigarette burns, attempts to quit smoking, fires caused by smoking, activities related to smoking and health by private and government groups, and costs to business to hire and train re- placements for ill smokers. Additional direct costs of disease borne by patients and other in- dividuals include costs of transportation to health providers, certain household expenditures, and costs of relocating (such as moving expenses). Transportation costs could be incurred not only for local transportation to hospitals, clinics, physicians, etc., but also for transportation out of state, and out-of-area living costs. lllness can force a family to incur expenses in caring and providing for the sick member of the family. Thece include extra expenditures for household help for cleaning, laundering, cooking, and babysitting; special diets; special clothing; items for rehabilitation and comfort such as exercycles, vaporizers, humidifiers, and dehumidifiers; alterations of property, such as elevators for invalids and other special housing facilities; and vocational, social, and family counseling services. Other costs originating in disease or illness are expenditures for retraining or reeducation, and care provided by family and friends. Limitations of data have hindered development of estimates of direct costs other than health expenditures, with existing information being mostly anecdotal. Luce and Schweitzer (1978) included the health care and property costs of fires caused by smoking, but these amounted

496 D.P. Rice el a!. to less than 3 percent of the total direct costs. Nonhealth direct costs have been estimated infrequently, usually for a specific disease (cancer, for example), and for relatively small samples. Although not concerned with health effects of smoking, several studies indicate the potential importance of nonhealth direct costs. Lansky et al. (1979) found mean weekly expenditures for 70 families of pediatric cancer patients totaled $56 for transportation, food, clothing, family care, and lodging. Patients receiving out-patient chemotherapy reported similar nonmedical expenses resulting from their diseases of $37 during treatment weeks and $17 during riontreatment weeks (Houts et al. 1984). Although these expenditures seem relatively high, neither study indicates how many weeks they were incurred, their relation to medical care ex- penditures, or the year of data collection. In their study of costs of caring for children with cancer, Bloom, Knorr, and Evans (1985) found nonmedical direct expenses for a six-month period in 1981 for 569 children with cancer at the Children's Hospital of Philadelphia averaged about ;4,000 annually and were almost 20 percent of the medical expenditures incurred during this same period and 15 percent of gross annual family income. One of the few studies, if not the only one, to attempt to estimate nonhealth-sector costs for the nation for all medical conditions was by Mushkin and Landefeld (1978). They estimated nonhealth direct costs between $23 billion (low estimate) and $29 billion (high estimate) in 1975, adding 19 to 23 percent to direct health care expenditures. These additional expenditures were incurred by consumers for trans- portation to providers, property losses to fire, and automobile accidents; by government for special education, vocational rehabilitation, coun- seling, added fire protection, and extra costs to the criminal justice system; and by industry for environmental and safety investments. These were the only nonhealth direct expenditures Mushkin and Iandefeld were able to estimate with existing data, and represent only a fraction of the potentially measurable costs. Although these nonmeasured costs are potentially large, their relative importance compared to health care expenditures is'uncertain. I ndirect Cott.r Indirect costs of smoking are the value of lost productivity, output, or forgone manpower resources when persons lose time from work The Economic Co.rr.r nf the Health Effect.r of Smoking, 1984 497 and other productive activities due to morbidity, disability, or premature mortality caused by smoking-induced illnesses. In this article, we estimate these indirect costs of smoking. Illness may also adversely affect productivity in addition to causing time lost from work by lessening the productivity of persons while on the job. Absenteeism also may increase costs of production with the end result that the value of output per unit of input declines. Additional indirect costs include the time a patient and/or family members spend visiting physicians, other health professionals, and hospitalized persons, and time lost from work by family members when someone in the family is ill. As for nonhealth direct costs, data for estimating indirect costs associated with lessened on-the-job productivity and time lost to various persons besides the patient are sparse. Mushkin and Landefeld (1978) estimated the cost of time spent visiting physicians, dentists, and hospitals, and days lost from work due to another person's illness at about $4 to $6 billion in 1975. This adds 5 percent to the commonly estimated indirect costs resulting from the patient's morbidity and premature mortality. The study by Lansky et al. (1979) of families of pediatric cancer patients found an average loss of pay from accom- panying the child to the hospital equal to 14 percent of family income. In the study by Bloom, Knorr, and Evans (1985), families of children with cancer lost wages amounting to 18 percent of family income. Indirect costs such as these, which have not usually been measured in cost-of-illness studies due to lack of data, very likely vary by disease and certain other parameters. These few studies indicate, however, that as for nonhealth direct costs, nonmeasured indirect costs may be a substantial portion of the economic burden of illness. Intangible Costs Direct and indirect costs are losses because they represent reduced consumption possihilities; costs result from the consumption of resources that are thus forgone to other uses. In addition to economic, that is, monetary, costs, smoking causes intangible costs. These include in- tangible costs inflicted on others in the vicinity such as the irritating effects of smoke on the visual and olfactory senses and the respiratory system, and the annoyance these cause, and also noneconomic effects of illness and disease suffered by smokers and their families, friends, coworkers, and care-givers.

498 D.P. Rice et al. f The Economic Cn.rt.r of the Health Effect.r of Smoking, 1984 499 Illness and disease are responsible for a wide variety of deteriorations in the quality of life and personal catastrophes that are not reflected in direct and indirect economic costs. Victims may suffer loss of a body part or speech, disfigurement, disability, the pain and grief of impending death. They, and those around them, may be forced into economic dependence and social isolation, unwanted job changes, discrimination in obtaining employment and health and life insurance, loss of opportunities for promotion and education, relocation of living quarters, and other undesired changes in life plans. The environment created by illness'oken induces anxiety, reduced self-esteem and feeling of well-being, resentment, and emotional problems that often require psychotherapy. Problems of living may develop, leading to family conflict, antisocial behavior, and suicide. The victims and others may experience marked personality changes and reduced sexual function. Premature mortality has direct consequences for the family, affecting, for example, duration of marriage and age at widowhood. Disrupted , development and delinquency may occur among children. The quality of life may be reduced beyond the restorative capability of current rehabilitation efforts. The combination of financial strain and psychosocial problems can be especially devastating. Psychosocial problems have been documented in numerous studies and appear to be widespread. A few examples are studies by Blanchard, Blanchard, and Becker (1976) (depression among widows), Campbell and Campbell (1978) (invasion of privacy, high insurance and interest rates, termination of employment), Cassileth et al. (1984) (mental health status), Derogatis et al. (1983) (psychiatric disorders), Goldberg (1981) (depression), Marinelli and Dell Orto (1977) (self-esteem, sexuality and sexual dysfunction). Intangibles are not easily quantified, and not easily accounted for explicitly in economic models. Consequently, it is not possible to compare the relative importance of economic and intangible costs in a common unit of measurement such as money. It is conceivable, however, that intangible costs are at least commensurate with, and may well exceed, economic costs in terms of their impact on both individual and societal welfare. Transfer Payments Smoking generates federal, state and local income and excise taxes (Harvard University lnstitute for the Study of Smoking Behavior and Policy 1985; Warner 1986). Taxes are neither benefits nor costs to the society as a whole. Rather, taxes are a form of transfer payment or reallocation of income from one segment of society to another. Although taxes are a cost to the payer and a benefit to the ultimate payee, the monetary value of the gains and losses offset each other (except for the costs which may be incurred in pperating the system for collection and disbursement). Taxes, however, undoubtedly have an impact on the welfare of payers who lose and payees who gain. Health insurance premiums paid by nonsmokers to cover the cost of medical care for smoking-related diseases incurred by smokers are transfer benefits to smokers, which are offset in monetary value by the transfer costs to nonsmokers. They occur when health insurance premiums do not reflect differential risks of disease to smokers and nonsmokers. The cost of smoking-induced disease is the value of resources devoted to medical care, whether or not paid entirely by smokers who become ill, or subsidized in part or in whole by nonsmokers. These are already counted among costs in terms of medical care expenditures. Health care premiums and out-of-pocket costs for treatment can be summed to obtain (approximately) the value of medical care resources devoted to treating smoking-caused disease, but it is important to avoid double counting. Nevertheless, it may be important to society to know the amount of subsidies involved and the extent to which nonsmokers subsidize medical care of smokers in order to decide consciously whether the society wants these to take place. Real, but as yet unexplored, benefits and costs of these subsidies are the welfare gains to smokers and the welfare losses to nonsmokers. Social Security, pension, and disability and sickness payments to ill smokers subsidized by nonsmokers (and smokers who do not suffer ill health effects), and payments forgone to smokers who die prematurely to the benefit of nonsmokers are also payments which transfer control over the use of resources from one segment of society to another. They do not represent the monetary value of resource losses caused by smoking and are not benefits or costs to society as a whole. Social Security and disability payments do result in a redistribution of income and welfare gains and losses and are important economic values. These transfer payments can be important economic values in the social decision-making process and assist in determining the societal response to smoking activities.

500 D. P. Rice ct a!. Studies of Economic Costs of the Health Effects of Smoking There are a number of studies of the costs of smoking, but no one study has addressed all aspects (Shultz 1985). Alternative perspectives of the costs of smoking differ in the focus of their concern, including: (1) aggregate costs (e.g., medical care expenditures) due to past smoking, i.e., prevalence-based costs; (2) lifetime medical care expenditures of smokers versus nonsmokers for all conditions and for specific conditions, including lung cancer, coronary heart disease, chronic obstructive pulmonary disease, i.e., incidence-based costs; (3) tradeoffs between higher than average annual medical care use and expenditures of smokers and longer life expectancy and additional years of medical care for nonsmokers; and (4) long-run reductions in smoking and its effect upon the economy, including the future impact of changes in smoking patterns on certain government receipts and expenditures, government deficit or surplus, and employment. Prevalence-based Social Costs The majority of cost-of-smoking studies have been prevalence-based analyses of social costs. They have examined costs to the society rather than private costs (accruing to participants in market transactions, such as smokers, for example) or external costs (falling on others such as nonsmokers and business and government organizations). And they have been concerned with the economic costs incurred in a period of time (most often a year) as a result of the prevalence of smoking- induced disease during this same period. Prevalence-based costs measure the value of resources used (direct costs) or lost (indirect costs) during a specified period of time (the base period), regardless of the time of disease onset. The costs of the base-year manifestations or sequelae of smoking-related disease, which may have had its onset in the base year or any time prior to the base year, are included. Prevalence-based costs assess the current costs of smoking. Current morbidity, mortality, and economic costs result from many past years of tobacco consumption, and current consumption will affect the future The Economic Cacts of the llealth lif%ctt of Smoking, 1984 5501 health of smokers. Therefore, prevalence-based, or current, costs of smoking represent the maximum annual value of resources that could gradually be shifted out of care of smoking-induced illness and into other social priorities if levels of smoking were to decrease. The impact of changes in smoking patterns would take place over a period of years, and the total amount saved would be the sum of a series of annual reductions. Alternatively, if cessation of smoking produced a larger, older population, the health care costs of smoking are resources that could in whole or in part, depending on population dynamics, provide care to an older population with longer lifetimes and lower average annual per capita health care costs. Simon (1968), Hedrich (1971), Williams and Justus (1974), Freeman et al. (1976), Kristein (1977), Luce and Schweitzer (1978), Forbes and Thompson (1983), Office of Technology Assessment (1985), and Vogt and Schweitzer (1985) have all evaluated social costs of smoking. The results of these studies cannot be compared, however, since the types of costs, diseases, and categories of smokers included, and the methodology employed vary among the studies. The study reported in this article is prevalence-based and the results will be compared with the studies by Luce and Schweitzer and the Office of Technology Assessment after the presentation of our findings. 1 ncidence-ba.red Costs In contrast to prevalence-based costs, which are the costs manifested during a period of time, usually over a year, as a result of smoking- induced disease, incidence-based costs are the lifetime costs expected to occur in a group of smokers as a result of smoking-related disease. An incidence-hased study by Oster, Colditz, and Kelly (1984a, 1984b) estimates the direct (medical care expenditures) and indirect (lost wages, salaries, and housekeeping services) economic costs of smoking and benefits of quitting among persons who smoked in 1980 for three smoking-related diseases: lung cancer, coronary heart disease, and emphysema. The economic costs of smoking are the average additional costs per smoker that will be incurred over the smoker's lifetime due to these diseases if he/she continues to smoke throughout life at the same level. Most of the total cost results from indirect losses rather than medical care at younger ages, but direct costs increase dramatically relative to indirect costs at older ages. This general pattern holds true

502 D.P. Rice et al. I The Economic Cau.r of the Health 1•ffectr of Smokink, 1984 503 for women as well as men and for each of the three smoking-related diseases. Oster, Colditz, and Kelly conclude that a smoker, over his or her lifetime, will require higher medical care expenditures for the three smoking-related diseases than will nonsmokers. Costs increase with the amount smoked, and are higher for men than women due to the higher risks of disease experienced by men (except for chronic obstructive pulmonary disease among heavy smokers 50 years of age and over). Combining Oster, Colditz, and Kelly's projections of cost per smoker and the prevalence of smoking, we estimate $500 billion as the present value of lifetime costs of smoking by current smokers in 1980 for the three diseases. The benefits of quitting are equal to the expected costs of smoking-related diseases, adjusted to take into account that ex-smokers' risks of disease slowly decline over a number of years compared with the risks faced by nonsmokers. Using a somewhat different model, Lewit (1983) analyzed the re- duction in health care costs and savings in indirect costs that would result from a gradual reduction in smoking-related disease in the United States beginning in 1980. During the first 25 years, the sum of health care costs saved was projected to be about $200 billion and the gains in indirect costs were equally substantial although realized more gradually. Lifetime Medical Care Expenditures of Smokers versus Nonsmokers 1.e4 and Schaub (1983) examine the impact of smoking on lifetime medical care expenditures of Swiss males. They estimate that although smokers have higher than average annual expenditures for medical care, the longer expected lifetime of nonsmokers means that expected lifetime medical care expenditures for males at age 35 who do not smoke will be 7 percent higher than expenditures for 35-year-old male smokers. - Comparing the methods of Leu and Schaub and Oster, Colditz, and Kelly to estimate lifetime medical care expenditures for smokers, the former includes all medical conditions, while the latter considers only the three smoking-related diseases. Leu and Schaub find lifetime medical care expenditures of 35-year-old male Swiss smokers less than expenditures for nonsmokers. Oster, Colditz, and Kelly report average lifetime costs among smokers who quit are substantially reduced. Although it appears that these two studies offer contradictory results, closer examination of the assumptions, data, and methods indicate that they may be logically consistent. On the one hand, nonsmokers, because of their longer lifetimes, might have somewhat higher or negligibly different lifetime health care expenditures over all diseases than smokers who have higher annual per capita expenditures while alive, but die earlier (Leu and Schaub 1983). On the other hand, smokers have higher expected lifetime expenditures for lung cancer, coronary heart disease, and emphysema because they are at higher risk of developing those diseases than nonsmokers (Oster, Colditz, and Kelly 1984a, 1984b). The conclusion reached by Leu and Schaub, however, that lifetime medical care expenditures of smokers are not higher, and possibly are even lower than those of nonsmokers, may be premature for at least two reasons. First, expenditures are not discounted. The effect of not discounting is to overstate expenditures of nonsmokers since a con- siderable portion of a nonsmoker's lifetime expenditures are incurred in those extra years of life granted the nonsmoker, after the age at which the smoker would die and cease to incur expenditures. This can be a considerable period of time. For example, in 1977 a male in the United States who died of cancer of the trachea, bronchus, or lung, which is typically related to smoking, on the average died at an age with an expected remaining lifetime of 14 years (Rice and Hodgson 1981). Furthermore, these expenditures will be highly concentrated in the more distant years just before the time of death. Lubitz and Prihoda (1984) have shown that, in 1978, Medicare decedents 67 years of age or older represented only 6 percent of beneficiaries but received 27 percent of reimbursements for medical care. These reimbursements were highly concentrated just before death, with one-fourth of reim- bursements in the two years preceding death for care received in the last month of life. Assuming a relatively modest discount rate of 3 percent, a dollar of medical care expenditures incurred by a nonsmoker 14 years in the future has a discounted value compared to a dollar of expenditure during the last year of the smoker of only $.66. In other words, a nonsmoker would have to incur $1.52 (52 percent more) in medical care expenditures fourteen years after the expected age of death of the smoker to offset a dollar of medical care in the I

504 D.P. Rice et al. i The Economic Coru of the Health EJfectt of Smoking, 1984 505 smoker's last year of life. Medical care expenditures of nonsmokers are deferred to the future and the appropriate comparison is between the present discounted values of the respective streams of expected annual medical care expenditures. Second, Leu and Schaub assume relatively low rates of excess medical care use and average annual medical care expenditures for smokers versus nonsmokers. We found actual excess utilization of physicians' services by smokers compared to nonsmokers 2.6 times that calculated by Leu and Schaub and excess use of hospital care 7.7 times higher. Lack of discounting and the possible underestimation of the amount by which average annual use of medical care by smokers exceeds use by nonsmokers means that l.eu and Schaub may have underestimated lifetime medical expenditures of smokers relative to nonsmokers. The amount of understatement is uncertain without further analysis, but could be substantial. Nevertheless, the concept of a tradeoff between higher than average annual medical care use and expenditures of smokers and longer life expectancy and additional years of medical care for nonsmokers is valid and an important aspect of analysis of costs of smoking. The analysis begun by Leu and Schaub should be continued in order to ascertain this relationship with greater certainty. Impact of Long-run Reductions in Smoking on the Economy Studies have examined the future impact of changes tn smoking patterns on certain economic variables, including government receipts and expenditures, government deficit or surplus, and employment. Atkinson and Townsend (1977) examined the long-run impact in Great Britain of an increase in cigarette taxes and a reduction in smoking on government tax receipts and certain transfer payments and revenues. They found that a 40 percent reduction in the number of cigarettes smoked, achieved by phasing in from 1977 to 1980 an increase in the cigarette tax, restrictions on advertising, gift coupons and sport sponsorship, and a health education program, would mean a net increase in population of 250,000 persons in 1998, with marginal change in National Health Service usage. By the year 2000, they project a substantial increase in annual tax revenues and a small net annual reduction in government spending, with savings in sickness benefits and widows' pensions more than offsetting extra costs of retirement programs and health education efforts. Gori and Richter (1978) use the Wharton long-cerm econonietric model to forecast certain economic effects of elimination of the minimum preventable portion of major causes of death, starting in 1975. Population changes resulting from reductions in mortality are introduced into the Wharton long-term model, and their effects on various economic indicators are forecast every five years from 19$Q to 2000, as mortality from preventable diseases is gradually eliminated between 1975 and 2000. Gori and Richter estimate reductions in mortality races resulting from a policy of disease prevention based on the difference between United States rates and the next-to-the-lowest rates observed in in- dustrialized countries for five major causes of death, including car- diovascular renal diseases, cancer, accidents, diseases of the respiratory system, and diabetes. The next-to-the-lowest rates were used in order to give conservative estimates. Smoking is only one of the factors responsible for observed differences in mortality; others include diet, alcohol and drug abuse, occupational hazards, air and water pollution. The relevant aspect of this analysis for our purposes is the modeling employed, which could be applied to estimate effects of reductions in mortality from smoking. A key assumption which greatly affects the projections is their restriction of the labor force to persons 16 to 65 years of age. The impact of this by the year 2000 is to increase government transfer payments by about 9 percent over what the Wharton model forecasts in the absence of disease prevention. Furthermore, under this scenario the federal deficit is more than 50 percent larger, and there are relatively minor increases in the gross national product (GNP), civilian labor force, and unemployment. Although the proportion of elderly in the labor force has been gradually declining in recent years, this assumption of no labor-force participation by persons over 65 years of age can be questioned since 23 percent of this age group had income from earnings in 1980 (Upp 1983). Current thinking leans toward raising the retirement age; starting in 2000 the age at which full Social Security retirement benefits are payable will gradually rise until it reaches 67. Reduced benefits will still be payable at age 62, but the reduction will be larger than it is now. If private pension systems follow the lead of Social Security and raise the age of eligibility for full benefits, incentives will be in place for more workers to work past age 65. The net effect on age at retirement of higher ages for full benefits and the desire of large numbers of workers to retire early is uncertain. I

`I'he Economic C.'ort.r of the Health tiffectt of SnrokinR, 1984 507 So6 D. P. Rice et al. Removing this constraint, Gori and Richter find quite different results for some variables in 2000. With disease prevention, the Wharton model predicts a federal surplus compared to a projected deficit in the absence of disease prevention, with a 65 percent difference in the two estimates. More modest increases in the GNP and government receipts and a much smaller increase in transfer payments are predicted. But with this latter assumption about labor-force participation, much larger increases in unemployment and unemployment benefits result. Gori and Richter rightly caution that the trends shown and not the numbers are important. In any case, for the purpose of our concern with longer-run reductions in smoking and their impact upon the economy, it is important to note that the direction of change in important economic variables is uncertain. The various models can be quite sensitive to assumptions about key parameters, and a good deal more analysis is required before we can be confident about long- run effects of changes in smoking patterns. Estimated Economic Costs of the Health Effects of Smoking Previous studies of the economic costs of smoking, employing the prevalence-based approach, applied global proportions attributable to smoking to illness costs (Hedrick 1971; Luce and Schweitzer 1978). For example, Luce and Schweitzer applied the following smoking percentages to updated cost-of-illness estimates originally published by Cooper and Rice in 1976: neoplasms-20 percent, circulatory system-25 percent, and respiratory system-40 percent. For this article, we have refined the estimates by using the epidemiologic methodology of "attributable risk" to calculate the direct (personal health care expenditures) and indirect (morbidity and mortality) costs associated with cigarette smoking. Attributable risk is "the maximum proportion of a disease that can be attributed to a characteristic or etiologic factor" (Lilienfeld and Lilienfeld 1980) and assumes that other factors influencing the occurrence of smoking-related diseases are equally distributed among smokers and nonsmokers. But smokers differ from nonsmokers in certain genetic, social, and economic characteristics which may contribute to disease. The prevalence of smoking 'varies by race (more blacks smoke than whites), education (fewer college graduates smoke than persons with only some high school), income (males with lower income smoke more, while the opposite holds for women), and occupation (blue- collar workers smoke more than professional or technical workers) (Vogt 1983; Warner 1983). If factors known to be related to health status and smoking habits are not controlled, the impact of smoking on health and the costs of smoking may be overstated. An interesting attempt to overcome this problem by Leu and Schaub (1983) analyzed smoking and medical care expenditures using three types of persons: smokers, nonsmokers, and nonsmoking smokers. The latter is a statistical construction having the smoking habits of a nonsmoker but like a smoker in other respects. Leu and Schaub assumed that 65 percent of smokers' excess mortality was due to smoking and 35 percent to other characteristics of smokers. Although it would be important to account for differences in mortality and morbidity between smokers and nonsmokers not due to smoking, the empirical basis for doing so is not readily apparent, and the Leu and Schaub assumption is arbitrary. The detailed methodology and sources of data for estimating the attributable risks for medical care utilization, morbidity, and mortality and their application to the direct and indirect costs of illness are detailed in the methodology appendix at the end of this article. Summary results are presented below. Disability and Medical Care Utilization Differentials Smokers are sicker and require more medical care than those who do not smoke. Table I records a comparison of the disability and medical care utilization rates for persons 17 years and over who ever smoked (current and former smokers) and those who never smoked, by age and sex; the data are from the Smoking Supplement of the 1979 National Health Interview Survey (NHIS). Higher rates in all the measures are reported for smokers compared with nonsmokers, ranging from 6 percent for physician visits to 72 percent for persons unable to work or keep house. The differentials between male smokers and nonsmokers are especially high. For example, the number of men reporting that they are unable to work is 88 percent higher for smokers compared with nonsmokers. For male smokers, hospital days are 63 percent higher, restricted-activity days are 55 percent higher, and

TABLE 1 Disability and Medical Care Utilization by Cigarette Smoking Status, Sex and Age: United States, 1979 Both sexes Mala Females A ed Smoking 17 ~ean status and over 17-44 years 45-64 years Aged 65 years 17 years and over and over 17-44 years 45-64 years Aged 65 years 17 years and over and over 17-44 years 45-64 years 65 years and over RESTRICTED-ACTIVITY DAYS PER PERSON PER YEAR All persons' 22.3 Ever smoked2 24.2 Never smoked 20.3 15.1 17.9 12.1 26.3 28.8 22.6 42.8 41.4 43.9 20.0 22.9 14.8 13.7 15.8 10.7 24.4 27.5 14.3 39.4 39.8 39.5 24.4 25.8 23.4 16.5 20.4 13.1 28.2 30.7 25.9 45.2 44.2 45.3 BED-DISABILITY DAYS PER PERSON PER YEAR All pe:sons' 7.5 Ever smoked2 7.8 Never smoked 7.2 5.7 6.5 4.7 8.1 8.8 7.1 13.8 11.7 15.4 6.1 6.7 4.8 4.3 4.8 3.5 7.1 8.1 3.6 12.1 11.1 14.5 8.9 9.3 8.6 6.9 8.4 5.7 9.1 9.8 8.5 15.0 12.7 15.7 WORK-LOSS DAYS PER CURRENTLY EMPLOYED PERSON PER YEAR All persons' 4.9 Ever smoked2 5.4 Never smoked 4.3 5.0 5.8 4.0 4.7 4.6 5.1 - - - 4.5 5.0 3.6 4.5 5.2 3.4 4.3 4.6 3.3 - - - 5.5 6.2 5.0 5.6 6.9 4.5 5.4 4.6 6.3 NUMBER OF PERSONS UNABLE TO WORK OR KEEP HOUSE3 PER 100 PERSONS4 All persons' 5.0 1.1 Ever smoked2 6.2 1.3 Never smoked 3.6 0.8 6.8 9.0 3.2 17.2 22.1 13.3 7.9 9.4 5.0 1.7 2.0 1.2 11.4 12.7 6.8 28.6 30.6 24.7 2.5 2.0 2.9 0.5 0.5 0.5 2.6 3.5 1.8 9.1 7.2 9.8 HOSPITAL DAYS PER PERSON PER YEAR All persons' 1.2 0.8 1.4 2.7 1.1 0.6 1.5 2.5 1.4 0.9 1.3 2.9 Ever smoked2 1.4 0.9 1.6 2.9 1.3 0.8 1.7 2.7 1.4 1.1 1.5 3.2 Never smoked 1.1 0.6 1.1 2.6 0.8 0.5 1.0 2.2 1.3 0:8 1.2 2.7 PHYSICIAN VISITS PER PERSON PER YEAR All persons' 5.0 4.5 5.2 6.8 4.2 3.4 4.7 6.5 5.8 5.6 5.5 6.9 Ever smoked2 5.2 4.7 5.3 7.0 4.4 3.6 4.9 6.5 6.2 6.0 5.9 7.9 Never smoked 4.9 4.4 4.9 6.6 3.7 3.1 4.1 6.6 5.6 5.3 5.3 6.6 Soxrre: Smoking Supplement of the 1979 National Health Interview Survey. Natc: These estimates will be slightly different than other published National Health Interview Survey estimates because these are computed from the one-third sample of persons who were given the smoking supplement. In addition, the variables "unable to work/keep house" and "hospital days" reflect slight definicional modifications from other published estimates. ' Excludes persons of unknown smoking status. = Includes current and former smokers. 3 Includes only females keeping house. 4 Number of persons unable to work or keep house is nor adjusted by labor-force participation, employment, or housekeeping rates. TIMN 305890

510 D. P. Rice et al. bed-disability and work-loss days are about 40 percent higher. The differentials in these disability and medical care utilization measures for female smokers compared with nonsmoking women are lower, ranging from 8 to 24 percent. For women reporting that they are unable to work or keep house, the rates are higher for the nonsmokers, except for those 45 to 64 years of age. Similar patterns are seen by age. The disability and medical care utilization rates for smokers under age 65, especially males, are sig- nificantly higher than for nonsmokers. For those aged 65 and over, the differentials are not as large and for several measures (bed-disability days for men and women, restricted-activity days for women and women unable to keep house), the rates are slightly higher for non- smokers. It is possible that some older persons suffer from a variety of chronic illnesses regardless of their smoking history, resulting in slightly higher disability rates. Also, these rates increase with age for smokers, and nonsmokers aged 65 and over tend to be older than smokers in the same age group. Morbidity and Medical Care Attributable Risks The availability of morbidity and medical care utilization rates by types of condition and smoking status enabled us to estimate for the first time the proportion of the illness measure or type of medical care used that can be attributed to smoking. We focused on the three major diagnostic categories most clearly associated with smoking- neoplasms, diseases of the circulatory system, and diseases of the respiratory system. Thus, 30 percent of the men and 17 percent of the women .17 years of age and over who suffer from these three major conditions and who report they are unable to work or keep house may be attributed to smoking (appendix table 1). Almost 3 out of 10 hospital days of care for them are estimated to be associated with smoking and the proportion is higher for men and for those under age 65. Almost I out of 5 visits by men and 1 out of 15 visits by women to physicians outside of hospitals may be attributed to smoking, while I out of 7 days lost from work is associated with smoking. Direct Costs Direct costs of smoking are the amounts spent for hospital care, physician and other professional services, drugs, and nursing home Tht Economic Goit.c of the Health Lffect.r of Smoking, 1984 51 1 TARI.E 2 Direct Costs: Total Personal Health Care Expenditures for Neoplasms and Diseases of the Circulatory and Respiratory Systems, and Amount Attributed to Smoking by Sex and Age: United States, 1980 Age Both sexes Males Females TOTAL EXPENUITURESI (millions) All ages $62,198 $27,675 $34,523 Under 65 years 32,631 15,830 16,801 65 years and over 29,568 11,845 17,722 AMOUNT A't'TR18lrr'F:1) TO SMOKING (millions) All ages ;14, 384 $8,220 $6,164 Under 65 years 8,734 5,366 3,368 65 years and over 5,650 2,854 2,796 PERCENTAGF, OF TOrAL A7TRIE- UTEC) TO SMOKING All ages 23.1% 29.7% 17.9% Under 65 years 26.8 33.9 20.0 65 years and over 19.1 24.1 15.8 Notr. Numbers and percentages may not add to totals due to rounding. 1 From HodKson and Kopstein (1984). care in behalf of current and former smokers. The data on attributable risks for medical care services enabled us to estimate the direct costs of smoking much more accurately than previous cost estimates. These factors were applied to personal health care expenditures for these three major diseases. Direct costs of smoking total $14.4 billion in 1980, accounting for almost one-fourth of the total expenditures for personal health care for neoplasms and diseases of the circulatory and respiratory systems (table 2). About $8.2 billion, or' 57 percent, are the costs of smoking for men; $8.7 billion, or 61 percent, are for persons under age 65 (figure 1). Table 3 records the direct costs of smoking by type of care. Hospital care accounts for the largest share--69 percent of the total. Professional services and nursing home care each account for 13 percent of the total, and 5 percent are for drugs.

SIZ Neoptesms. dis.ases of the circulatory end r.spbefory systems. $5.4 Under 66 66 years years and over Malet 1).1'. Rice el ul. 82.9 Under 65 66 years years and over Females FIG. 1. Direct costs of smoking by age and sex, 1980 (in billions of dollars). Morbidity Coru Morbidity costs are the value of losses in output for people who are ill and disabled and unable to work. We use average earnings by age and sex and impute a value for housekeeping services for women who are unable to keep house because of illness and disability. The attributable risks shown in appendix table I were applied to person-years lost and' to total morbidity costs for the three major diseases as described in the methodology appendix. A total of 528,000 person-years are estimated to be lost to productivity by current and former smokers, at a total cost of $7.4 billion (table 4 and figure 2). Fifty-six percent of the person-years lost and 72 percent of the morbidity smoking costs are attributed to men. The distribution by age shows that 85 percent of the person-years lost and 96 percent of the morbidity costs of smoking are for persons under age 65, reflecting the higher attributable risks for those under age 65, and their higher earnings. Mortality Co.rt.r As indicated earlier, previous studies of the economic costs of smoking applied global proportions attributed to smoking to illness costs. For '1'he liamorntr (Y the Ilealth 1 ffI7.r of Sntnkinl;, 1<)1{4 ~ N 4. o T O 'D - W ~ Y }I a ~ ~ > D v ~ ~ T ~..1 ` ~ ~o v... U q ~ YT ~ r L\ v1 ~~~r oo ~ r~ M N p N .. N ^ « N ~ N v. p~p N DD M t er N N ^ M~'r O co N N o0 ~ p N O 00 ~D r'- O ~,n0 N © N ., ^ & O O, 00 O, M 8 Q 00 ~6 N ~ O v v O O O. M v g 00 ~ O Ni o0 $ ^o o: ~ O ~D O -O 00 1_ p pp 87 N O N ~ v p O O 00 O ^ v S^~ ^ p p 00 v po N O 0~ v ^ ~' ^ N Op N M N o V 00 M { O N p M ~ O 00 «^ ~~~^'O '0 O 7 O 00 ^ 8$1; 72 z 0 z 513 I

514 D. P. Rice et al. A ~ 0 E w ~.}N N ~O O \0 r~ ~ W~~ O O N N NO~O N N~ N 79 M ~ ~ CU 0 00 g~ U -•\0 rn C~ Ia' op \oroo U J. (T ~ y~ N00.`O'- OMpN N N - u '2 E ~ O, 00 0 N Q T X S M ~ O\0 ~ OoOO~ N N ^' '~p C ~ p M ~O 1~ M ^' N N U W] N N M O, C d ~ 00 \G \O 00 N ~ z ~ E W \O + M N N w i, O N ~ .- O M N U:ji ~ O Mt~ M O M V~ ~D ^^ ~ ~ E ~. 1~ ~ N Q, M\L1 N N N N N V') 0 o a CW c . " < 0~c v c x s N N M N'G' 00 ~N N~ Q .o 10 0 u N r' 7 4 0 b N M y F+ u y N 0 0 0 y 0 00 rn The Economic Co.uu of the Health Ejfect.r of Smoking, 1984 Sex P.r.on- Morbldlty y..r.lo.t co.tU Penon- y..rs toot Age Morbidity cott. FIG. 2. Morbidity costs of smoking for neoplasms and diseases circulatory and respiratory systems, by sex and age, 1980. 515 of the this article we refined the mortality costs by estimating the attributable risks for .19 specific causes of death for males and females based on weighted mortality ratios from 4 prospective studies on smoking as described in the appendix. The attributable risks of cancer mortality from smoking among men ranges from 81 percent for cancer of the trachea, bronchus, and lung to 18 percent for stomach cancer; for women the range is from 56 percent for cancer of the esophagus to 13 percent for kidney cancer (appendix table 4). Not surprisingly, the attributable risks for emphysema and chronic bronchitis are very high-$7 percent for males and 72 percent for females. The attributable risks for aortic aneurysm are also high--66 percent for males and 49 percent for females. As indicated earlier, for mortality the cost or value to society of all deaths attributed to smoking is the product of the number of deaths attributed to smoking and the expected values of an individual's future earnings, with sex and age taken into account. This method of derivation takes into consideration life expectancy for different age and sex groups, changing patterns of earnings at successive ages, varying labor-force participation rates, imputed values of housekeeping

D.l'. 12ice el ul. '!'br liranontir uf Ihe lleul/h I fferl., r f,t'nnoking, 1984 517 services, and the appropriate discount rate to convert a stream of costs or benefits into its present worth (Rice, Hodgson, and Kopstein 1985). We used two discount rates: 4 and 6 percent. We also estimated the person-years lost, based on the number of years remaining at the time of death, from the 1980 life tables published by the National Center for Health Statistics (1984). Table 5 records the number of deaths and person-years lost to productivity for all causes of death attributed to smoking by age and cause of death., Mortality costs at 4 and 6 percent by age, and cause of death are shown in table 6. (Similar data by sex are available from the authors.) The following are highlights of our findings: • A total of 270,269 deaths in 1980 were due to smoking, resulting in 3.9 million person-years lost; • Premature deaths from smoking cost the nation $16.8 billion in 1980; • About 69 percent of the premature deaths and person-years lost arc attributable to smoking among men. Men account for 80 percent of the costs, reflecting the higher risks for men and their higher earnings compared with women; • About 31 percent of the deaths attributed to smoking occur for those aged 45 to 64; this age group accounted for almost half of the person-years lost and 70 percent of the mortality costs (figure 3); • More than half the premature deaths from smoking are caused by diseases of the circulatory system. Almost two-fifths are deaths due to malignant neoplasms. Smoking-related neoplasms, however, represent a higher proportion of person-years lost (42 percent) and of costs (46 percent), because smokers who die from cancer are usually in the younger age groups; • Of the 1.5 million deaths for persons 20 years and over in 1980 due to neoplasms and diseases of the circulatory and respiratory systems combined, 17 percent are attributed to smoking; 19 percent of the 20.9 million person-years lost and 22 percent of total mortality costs are attributed to smoking (table 7). dcher estimates of the number of deaths attributed to smoking are higlici tliari otits (ltavenholt 1985). Our estimates are conservative ® 20,2, years ® I6-61 years ~ 66.74 yars M 76 years or over 3% 7% 18% 31% 48% 70% 39i FIG. 3. Mortality losses attributed to smoking: Distribution of deaths, person-years lost, and costs, by age, 1980, for several reasons: We have not taken into account the adverse effects of passive smoking, risks of abortions, stillbirths and neonatal deaths, or deaths under age 20 that might be associated with smoking. There is a growing body of literature that has concluded that involuntary exposure to tobacco smoke represents a significant public health problem resulting in premarure deaths (Repace and Lowrey 1985; Garland et al. 1985; , National Research Council 1986). The prospective studies upon which the attributable risks were estimated were performed several years ago and did not attempt to measure the adverse effects of smoking on these additional health problems, or certain current occupational and environmental hazards that greatly increase the risk of death for smokers. The studies were based on old smoking habits. For women, whose smoking habits have approached those of men only in the past decades, the earlier epidemiologic data may well be outdated. Women currently suffering from lung cancer, whose smoking histories date back two or three decades, may have smoked more intensively than women who were in the earlier prospective studies upon whom the attributable risk estirriates were based. I

TABLE 5 Mortality Losses: Deaths and Person-years Lost to Productivity for All Causes of Death Attributed to Smoking by Age and Cause of Death: United Stares, 1980 Number of deaths Person-yesrs lost (in thousands) 75 Aged Aged years 20 years 20-44 45-64 65-74 . 75 years 20 years 20-44 45-64 65-74 and Cause of death and over years years years and over and over years years years over To'r,u. 270,269 7,130 84,700 73,426 105,013 3.940 276 t,872 998 795 MALIGNANT NEOPLASMS 103,170 2,804 40,295 35,236 24,835 1,674 108 894 475 196 Trachea, bronchus, lung 74.705 1,787 30,195 26,462 16,261 1,215 66 665 354 129 Larynx 2,603 46 1,115 903 539 42 2 24 12 4 Lip, oral caviry, pharyn: 5,382 236 2,391 1,572 1,183 93 9 54 22 9 Esophagus 4,837 110 2,079 1.510 1,138 80 4 46 21 9 Bladder 3,612 23 613 1,146 1,830 43 1 13 15 14 Kidney 1,731 69 655 551 456 28 3 15 7 4 Pancreas 5,228 105 1,652 1,753 1,718 78 4 37 24 14 Stomach 3,142 98 805 921 1,318 47 4 19 13 11 Cervix 1.930 330 790 418 392 47 15 22 7 4 DISEASES OF THE CIRCULATORY ' SYSTEM 141.546 3,796 39,718 30,687 67,345 1,948 146 875 420 506 Ischemic heart disease 86,036 2,585 31,684 18.324 33,443 1,283 96 692 246 248 Cerebrovucular disease 22,637 444 2,551 4,411 15,231 262 19 61 64 118 Hypertension 5,425 151 1,119 1,326 2,829 73 6 26 19 22 Aortic aneurysm 8,612 1,140 1,522 2,957 3,993 107 6 32 39 31 Atherosclerosis 8,993 16 376 1,077 7,524 78 1 8 15 54 Cardiac arrest 9,843 460 2,466 2,592 4,325 144 19 56 36 33 DISEASES OF THE RESPIRATORY SYSTEM 22.917 412 4,063 6,799 11,643 282 17 88 91 85 Emphysema, chronic bronchitis 14.098 112 2,989 5,230 5,767 184 4 64 71 44 InAuenu, pneumonia 8.819 300 1,071 1,569 5.876 98 13 24 21 40 OTHER CAUSES OF DEATH 2,636 118 624 704 1,190 37 5 14 10 9 Respiratory tuberculosis 536 35 182 150 169 8 1 4 2 1 Ulcer 2,100 83 442 554 1,021 29 3 10 8 8 tiate: Numbers may nor idd to rorals due to rounding. I& TINLN 305895

TABLE 6 Mortality Costs: Discounted Productivity Losses for All Causes of Death Attributed to Smoking by Discount Rate, Age and Cause of Death: United States, 1980 (in millions) Discounted at 4 percent Discounted at 6 percent Aged ears 20 20-44 65-74 75 years Aged 20 years 20-44 65-74 75 years Cause of death y and over yeus 45 -64 years years and over and over years 45 -6d years years and over mrAr.. $16,814 $3,017 $11,811 $1,565 $420 $14,836 $2,439 $10,565 $1,435 $398 MALIGNANT :VEOPLASMS 7,687 1,131 5,684 753 118 6,803 919 5,082 691 110 Trachea, bronchus, and lung 5,631 733 4,261 559 78 4,999 599 3,813 513 73 Larynx 202 20 161 19 2 180 16 144 18 2 Lip, oral cavity, pharynx 502 103 359 35 5 441 83 321 32 5 Esophagus 382 47 297 33 5 339 38 265 31 5 Bladder 113 9 74 23 7 102 7 67 21 7 Kidney 140 30 96 12 2 123 24 85 11 2 Pancreas 316 44 226 38 8 281 35 202 35 8 Stomach 172 39 106 21 6 150 31 94 20 6 Cervix 227 108 105 12 3 189 85 90 11 2 DISEASES OF THE CIRCULATORY SYSTEM 8,086 1,648 5,527 656 255 7,118 1,330 4,946 601 241 Ischernic heart disease 6,117 1,133 4,475 382 127 5,400 922 4,007 351 120 Cerebrovascular disease 667 177 328 102 60 580 140 290 93 56 Hypertension 258 63 153 31 12 225 50 136 28 li Aortic aneurysm 325 65 183 60 17 288 52 165 55 16 Atherosclerosis 96 6. 44 23 23 87 5 40 21 21 Cardiac arrest 623 203 345 58 17 536 160 308 53 16 DISEASES OF THE RESPIRATORY SYSTEM 878 185 508 142 43 775 149 455 130 41 Emphysema, chronic bronchitis 534 47 352 110 26 479 38 316 101 24 Influenza, pneumonia 344 138 156 32 18 296 111 139 29 17 OTFIER CAUSES OF DEATH 163 52 91 15 5 141 42 81 14 4 Respiratory tuberculosis 50 16 30 3 1 44 13 27 3 1 Ulcer 113 36 61 12 4 98 29 54 11 4 vote: Numbers may not add to totals 3ue to rounding TIMN 305896

TABLE 7 Mortality Losses: Deaths, Person-years Lost to Productivity, and Mortality Costs for Neoplasms" and Diseases of the Circulatory and Respiratory Systems, and Amount Attributed to Smoking by Sex and Age: United States, 1980 Mortality costs (in millions) Deaths' Person-years (in thousands) Discounted at 4 percent Discounted at 6 percent Age Both sexes Males Females Both srxes Males Females Both sexes Males Females Both sexes Males Females TOTAL Aged 20 years and over 1,535,184 806,485 728,699 20,918 10,700 10,218 $75,069 $48,738 $26,331 $65,470 $42,724 $22,746 20-64 years 376,464 235,062 141,402 9,449 5,387 4,062 63,017 43,064 19,953 54,367 37,466 16,901 65 years and over 1,158,720 571,423 587,297 11,469 5,313 6,156 12,052 5,674 6,378 11,103 5,258 5,845 AMOUNT ATTRIBUTED TO SMOKING Aged 20 years and over 267,633 185,832 81,801 3,904 2,697 1,206 16,651 13,369 3,282 14,696 11,844 2,851 20-64 years 91,088 71,364 19,724 2,130 1,581 550 14,683 12,097 2,587 12,881 10,666 2,215 65 years and over 176,545 114,468 62,077 1,773 1,117 656 1,967 1,272 694 1,814 1,179 635 PERCENTAGE OF TOTAL ATTRIBUTED TO SMOKING Aged 20 years and over 17.4% 23.0% 11.2% 18.7% 25.2% 11.8% 22.2% 27.4% 12.5% 22.4% 27.7% 12.5% 20-64 years 24.2 30.4 13.9 22.5 29.3 13.5 23.3 28.1 13.0 23.7 28.5 13.1 65 years - and over 15.2 •20.0 10.6 15.5 21.0 10.7 16.3 22.4 10.9 - 16.3 22.4 10.9 Sose: Numbers and percentages may not add to totals due co rounding. • Excludes deaths for which age is not available. TIMN 305897

524 1).P. Knr er a!. Total Economic Costs of the FI ealth Effects of Smoking The total economic costs of smoking amount to $38.6 billion in 1980. Direct costs account for 37 percent, morbidity costs for 19 percent and mortality costs 44 percent (table 8). Not surprisingly, the economic costs of smoking for men are considerably higher than for women-$27 billion and $11.6 billion, respectively. For men mortality costs are highest-50 percent of the total; for women, direct costs are highest---53 percent of the total economic costs. Smoking clearly has severe consequences for the nation, amounting to 8.5 percent of the total economic costs of all illnesses in 1980. Direct costs of smoking account for 6.8 percent of the total direct costs, and indirect costs represent almost 10 percent of the total indirect costs for all illnesses. It is evident that people who smoke die earlier, and their productivity losses are very high. We updated our figures to 1984 and the costs are even more staggering-S53-7 billion in 1984 (figure 4). To obtain 1984 values, direct costs were adjusted by the percentage change in total personal health care expenditures as reported by the Health Care Financing Administration. Indirect costs were adjusted by the percentage change in average weekly earnings as reported by the Bureau of Labor Statistics. Direct costs represent a larger share of the total-43 percent compared with 37 percent in 1980 because medical care costs have been rising faster than earnings that are the basis for estimating indirect costs. Again, mortality costs are relatively higher for males and direct costs are highest for females. Comparison with Other Cost-of-smoking Studies The studies by Luce and Schweitzer (1978) and the Office of Technology Assessment (1985).(OTA) also estimate medical care expenditures and the value of lost productivity from morbidity and premature mortality from smoking-induced disease. Their methodology is similar to that of the study reported in this article. Each of the three studies estimated costs of smoking by applying attributable risks to direct and indirect costs of neoplasms and circulatory and respiratory diseases. The costs of neoplasms and circulatory and respiratory diseases from which the costs of smoking are derived are consistent. Luce and Schweitzer inflated Cooper and Rice's (1976) estimates of costs in 1972 to 1976; '/'he licononiic Coiti o/ the Flea/th F,ffects of ,Smoking, 1984 525 0 Direct costs $53.7 Billion All persons Q Morbidity costs ® Mortality costs $36.5 Billion Males $17.2 Billion Fetnales FIG. 4. Economic costs of smoking, by type of cost and sex, 1984. OTA inflated HodKson and Kopstein (1984) and Rice, Hodgson, and Kopstein's (1985) estimates of 1980 costs to 1985; and this article utilized Hodgson and Kopstein (1984) and Rice, Hodgson, and Kopstein (1985) cost estimates. The principal sources of variation among the studies are the estimates of attributable risks. Luce and Schweitzer used attributable risks for three major diagnostic groups of diceases---•neoplasms and circulatory and respiratory diseastsr-- from Boden's (1976) study of the economic impact of environmental disease on health care delivery. For each major diagnostic group, the attributable risk was applied to total direct and indirect costs for that disease to estimate the amount due to smoking. Boden does not indicate how these attributable risks were derived. OTA improved upon this method by attributing costs of smoking according to the estimated proportion of smoking-related deaths for each disease by age and sex. By this method OTA accounted for the influence of declining attributable risks, declining per capita indirect costs, and increasing per capita health expenditures with increasing age. For our estimates, we introduce an additional refinement by estimating health care costs and indirect morbidity losses related to smoking from differences in medical care use and time lost from productive activity between smokers and nonsmokers observed in the National Health Interview Survey, rather than by differences in mortality which is characteristic of earlier studies. I

TABLE 8 Economic Costs of Smoking for All Diseases Attributed to Smoking by Type of Cost, Age, and Sex: United States, 1980 and 1984 1980 1984 Indirect costs Indirect costs ;ge and sex Total Direct costs Morbidity Mortaliry• Total Direct costs Morbidity Mortality °.MOt;NT (in millions) ~TH SEXES $38,579 $14,384 $7,381 $16,814 $53,711 $23,338 $9,286 $21,087 Under 65 years 30,678 8,734 7,116 14,828 40,241 12,872 8,935 18,434 65 years and over 7,899 5,650 264 1,985 13,471 10,466 351 2,654 MALES TOTAL 27,022 8,220 5,301 13,501 36,494 13,376 6,501 16,617 Under 65 years 65 years and over 22,669 4 353 5,366 854 2 5,086 215 12,217 1,284 29,060 7,434 7,899 5,477 6,220 281 14,941 1,676 G FEMALES TOTAL , 11 557 , 164 6 2,080 3,313 17,217 9,962 2,785 4,470 A ^ Under 65 years , 8,009 , 3,368 2,030 2,611 11,180 4,973 2,715 3,492 ~ e ~ 65 years and over 3,547 2,796 50 701 6,037 4,989 70 978 a PERCENTAGE DISTRIBUTION BY TYPE OF COST BOTH SEXES 100.0% 37.3% 19. 1% 43.6% 100.09~C 43.5% 17.3% 39.3% ~ Under 65 years 100.0 28.5 23.2 48.3 100.0 32.0 22.2 45 8 rn 65 years and . 0 a over 100.0 71.5 3.3 25.1 100.0 77.7 2.6 19.7 .1. 3 M.+LSS TOTAL 100.0 30.4 19.6 50.0 100.0 36.7 17.8 45.5 Under 65 years 100.0 23.7 22.4 53.9 100.0 27.2 21.4 51.4 65 years and over 100.0 65.6 4.9 29.5 100.0 73.7 3.8 22.5 FEMALES TOTAL 100.0 53.3 18.0 28.7 100.0 57.9 16.2 26.0 ~ Under 65 years 100.0 42.1 25.3 32.6 100.0 44.5 24.3 31.2 :z 65 years and z ~ over 100.0 78.8 1.4 19.8 100.0 82.6 1.2 16.2 a ` N ote. Numbers and percentages might not add to totals due to rounding. .. ~ • Discounted at 4 percent. ~ c.; ~ ~ A- ~ ~ ~ x TIMN 305899 N ~

528 D.P. Rice el ul. ~ The lirorroruir l;o.cli of Ihe !!eu!!h lJfer1.r of Snrokini;, /c)N4 52c Converting the costs estimated by these three studies to 1984 values (table 9) facilitates comparison of the different results. Our results are quite close to those of Luce and Schweitzer. The principal differences are with the OTA results. OTA estimates indirect costs to be 40 percent higher because of a higher estimate of mortality and a younger age distribution of the deaths due to smoking. In the OTA study 314,000 deaths were attributed to smoking versus 270,000 in our study. Deaths under 65 years of age, when indirect costs of mortality are higher, represented 41 percent of deaths from smoking in the OTA study compared with 34 percent in our study. The apparent similarities of these estimates mask substantial differences in estimated costs of the three component diseases-neoplasms and circulatory and respiratory diseases. There is, in general, a lack of consistency among the studies in terms of the magnitudes of the estimated proportions for a given medical condition. Although OTA and we calculate similar proportions for circulatory diseases (13 and 14 percent) and Luce and Schweitzer and OTA are very close on respiratory disease (40 and 41 percent), for the most part results of the three studies are quite different. The three studies differ in their application of the attributable risks to the direct and indirect costs of neoplasms and circulatory and respiratory diseases, which increases the differences in results for specific diseases. Although total costs in the three studies are fairly close, with the low estimate by Luce and Schweitzer being 85 percent of the high estimate by OTA, this is achieved through rather wide disparities in estimated costs of the various medical conditions which partly cancel out in the aggregate. The substantial amount by which OTA's estimated cost of neoplasms exceeds Luce and Schweitzer's estimate, coupled with larger costs of circulatory and respiratory diseases estimated by Luce and Schweitzer, results in nearly equal estimates of total costs. We prefer our estimates of health care expenditures and morbidity costs associated with smoking because they are based upon observed differences between smokers and nonsmokers in health care utilization and disability, including, for example, work-loss days and persons unable to work. Unfortunately, sample sizes prohibit estimation of costs by medical condition; it may be possible to overcome this by combining several years of data from the National Health Interview Survey. With respect to mortality costs, we estimated separate at- tributable risks for males and females for each specific cause of death TABLE 9 Economic Costs of and Percentage of Deaths from Neoplasms and Diseases of the Circulatory and Respiratory Systems Attributed to Smoking; Three Studies Luce and Schweitzer (1978) Rice et al. 1986 (current study) OTA (1985) AMOUNT KITRtHUTED TO SMOKING (in billions of 1984 dollars) TOTAI. $52.8 $53.7 $62.2 rype of CoJI Direct costs 21.0 23.3 19 8 Indirect costs 31.8 30.4 . 42 4 Diuate Neoplasms 8.8 N.A.* . 24.7 Circulatory diseases 26.1 N.A.• 24 3 Respiratory diseases 17.9 N.A.• . 13 1 PF.RCENTAUR ATTRInUTEt) TO SMOKING . TOTAL 22.6% 23.0% 26.6% Type nf cau Direct costs 21.2 23.5 19.7 Indirect costs 23.9 22.8 8 31 DiteaJe Neoplasms 13.1 N.A.' . 36.7 Circulatory diseases 22.3 N.A.• 20.8 Respiratory diseases 37.4 N.A.• 27.2 PERCENTAGE OF UIiATHS TOTAL 25 17 21 Diuase Neoplasms 20 25 32 Circulatory diseases 25 14 13 Respiratory disea.ces 40 18 41•• Nale: To obtain 1984 values, direct costs estimated by each study are adjusted by thca He plrhercCentaxe chanRe in total personal health care expenditures reponed by the are FictancinR Administration. Indirect costs are adjusted by the percentage change in average weekly earnings reported by the Bureau of Labor Statistics. * Data from rhe Smoking Supplement of the 1979 National Ilealth Interview Survey were used to estimare attributable risks for medical care and morbidity losses due to smoking. Irmitarinns in sample size prohibit disaggregation of health care expenditures and morbidir c~stc by medical condition. " OT y A relxuts 48 percent of deaths from respiratory diseases were due to smoking, but their taral for the denominator of the proportion excludes deaths from acute respiratory infections besides acute bronchitis and bronchiolitis, other diseases of t'Plxr respiratory tract, pneumonconioses, and other lung diseases due to external agents, and certain other diseases of the respiratory system. I

530 D. N. Rice et al. IE 7'ht l:conomic Cn.it.r of the Health Fffects of Smoking, 1984 S j-t associated with smoking rather than using one factor for each major diagnostic group of diseases. Thus, the total number of deaths attributed to smoking in our study is lower than the estimates of other researchers. Conclusions According to our estimates, smoking has severe economic consequences for the nation, amounting to a staggering $53.7 billion in 1984. We believe that our contribution to the literature on the economic costs of the health effects of smoking using the prevalence-based approach is four-fold: (1) The many conceptual issues involved in estimating the health effects of smoking were discussed and we categorized the al- ternative perspectives and methods of estimation, and compared the different cost estimates; (2) For the first time, attributable risks based on health status and medical care differentials by age, sex, and diagnosis observed in the NHIS were developed to estimate the direct and morbidity costs; (3) Attributable risks were developed by specific cause of death and by sex based on weighted mortality ratios for current and former smokers and nonsmokers from four major prospective studies; (4) Mortality costs were based on current (1980) lifetime earnings values applied to deaths by age and sex. Projections of future costs of smoking assume maintenance of current smoking behavior, including the prevalence and incidence of smoking, the amount smoked, and the type of cigarettes. Sensitivity analyses on these parameters could indicate potential changes in costs that might occur with changes in smoking habits. Beyond smoking itself, there are a number of factors influencing the health effects of smoking and attendant economic costs which are very difficult to predict and which have the potential to either increase or decrease costs. Progress in eliminating competing disease and increasing life expectancy would increase the relative risk of smoking-related morbidity and mortality. Changes in personal health practices, such as diet and exercise and exposure to chemicals in air, water, and food, may alter risks associated with smoking tcr the extent that there are synergistic relations among risks for diseases such as cancer, coronary heart disease, and pulmonary disease. Medical treatment has changed significantly over the years and changes will continue into the future. The cost of treating an illness may increase or decrease as the method of treatment changes. In a series of studies, Anne Scitovsky (Scitovsky 1968; Scitovsky and McCall 1977; Scitovsky 1985) examined changes in the costs of treatment of selected illnesses. In the years between 1951 and 1964 reduction in average length of stay was the main cost-saving change observed. Savings due to shorter lengths of stay were outweighed, however, by increases in the number of diagnostic tests and therapeutic procedures per case, greater use of specialists, and substitution of in-patient for out-patient care. During the period 1964 to 1971, average length of stay continued to decline and the number of diagnostic tests increased, raising the costs for some conditions and lowering costs of others. Between 1971 and 1981, the rate of increase in the use of ancillary services seems to have slowed, but costs were raised substantially by the introduction of several expensive "big-ticket" technologies. Although methods of treatment are certain to change, how these changes, coupled with changes in financing mechanisms that also affect medical care utilization and costs, will affect expenditures is uncertain. Advances in medical therapy may improve survival rates or lessen the severity of the condition and affect medical care expenditures and indirect costs. The future economic costs of the health effects of smoking depend on many diverse factors including smoking behavior, the incidence or prevalence of smoking-induced diseases, methods and costs of treat- ment, and valuation of losses in productivity. Nevertheless, the mag- nitude of the current costs of the health effects of smoking to the economy justifies concern over the mi.callockion of the nation's resources to harmful uses. Methodology Appendix Direct Costs Direct costs, or personal health care expenditures, in the United States in 1980 have been estimated for major diseases, including neoplasms

532 U. P. Rice et a!. and diseases of the circulatory and respiratory systems, for males and females and the two age groups under 65 years of age and 65 years of age and over (Hodgson and Kopstein 1984; Rice, Hodgson, and Kopstein 1985). Applying attributable risks as explained below to personal health care expenditures gives the estimated direct costs of neoplasms and circulatory and respiratory diseases due to cigarette smoking. The Smoking Supplement to the 1979 National Health Interview Survey (NHIS) provides estimates of utilization of hospital care and physician services for smokers (current and former) and persons who never smoked, by sex, age, and medical condition. Owing to limitations in sample size, our analysis combines the three major diagnostic categories most clearly associated with smoking-neoplasms, diseases of the circulatory system, and diseases of the respiratory system. Five steps were required to estimate direct costs of smoking: (1) Per capita rates of utilization of days of hospital care and physician visits for neoplasms and circulatory and respiratory diseases combined were calculated for smokers and nonsmokers by sex and age (appendix table 1). Differences in these rates between smokers and nonsmokers were tested for statistical significance by the standard normal test. (2) The attributable risk associated with smoking, that is, the maximum proportion of hospital days and physician visits that could be attributed to smoking, were calculated for males and females 17 to 44 years of age, 45 to 64 years, and 65 years and over. The basic formula used is that in Lilienfeld and Lilienfeld (1980): attributable risk = p (r - t) p (r - 1) + 1 rate in exposed r = relative risk = rate in nonexposed p= proportion of the population that ever smoked (current plus former smokers) (3) The number of hospital days and physician visits for neoplasms and circulatory and respiratory diseases combined that were due to smoking were calculated for males and females in the three age groups by applying the attributable risks in (2) to the total t~ 0 A ~ 4 V 0 ~ A ~ ~ > >` o 4" u T v "o ~ ~ ~ I > M M V~ M N 00 r\ y N M \.nv,W oornW rnooaN M aJ xr M op r- pN \p r, ~ONM 0~00~ O `1W M fV (V 'IT N O'v1 W N o O o"D w v~ M W 1~ 0~ 1~ ~O N O 00 O\O N \O C' Q~O f~ V" M no`~ o~ o~b~~ w <rr- N ' V \ O N N ~ N ~ ~,^~~ oo~rb< v ~rnwo rnN'a ~„ . . 0~ M 00 u. No6 O rn r, - E a 4 ~ U ~000FE ~ Q "T n hrnrnrnao~ ~ °;~ v rn rn rn a ^r0\ wrV 00 aoO~t-~ -~Mr. C w $ u `° wo o Y h v p ~ o .. fV y~ u o 0 ~~ O~oM a ~ i N Q~C~T vr~ E a o vM. ~ V oM. ~ .N-. ftlpl~~~,GGjl!!• ~ w~ M N C G ~ O ~ w`V'~O ~ 0 ~ .n^ xr ~O ~0~0 ~ N~DO \6 q6 .-~L. \Q 1~ a 1.. ~ 000 N M 00 o ~ 4. p N N 0 C u 0. a N ~ w o. a' b~ 0 oo ~o $E p oo ?~ O~ ~. 'UJ 1>1 O'V' ~D k 1~ 00 ~D 1~ M ~D NV" ~ 4 C o00 -:~ , NL 0r Mr- N Y~ M M $ z~ ~... ~~ ~ h o a a ~ z a ~ ~'wo V ~ ~~ A~ ~ p Qq N N ~ aC i ] a$ u G n > a $Z Pia !i0. ~ ~~~o q C~ w u ~, o_0 p 0-0 A p~~' ~ u ov a ~ E ~ ~ E~ ~ : oue p y;: W-K ~*Q a-X u'9 ;V ~e a- i cn hp ~ a4 o i: c1. BvBiZC zv~i2~ xv~iZ< M vBiZC . : ~ 533

534 I).l'. Kite et a!. ~ The lsconorrur Caui of the llevllh lifJrru if Smokiag, 1984 535 days and visits (smokers plus nonsmokers) in each of the sex and age groups. (4) The attributable risks in appendix table 1 for those 17 to 64 years of age were derived separately for each sex by adding together the number of hospital days or physician visits due to smoking at ages 17 to 44 years and 45 to 64 years from (3) and dividing by the total days or visits for neoplasms and respiratory and circulatory disease among smokers and nonsmokers 17 to 64 years of age. The following weighted average illustrates the method employed: AR _ ARi N; + ARI N~ '+' N; + NI i= 17 to 44 years of age j= 45 to 64 years of age AR = attributable risk (calculated in step (2) for i and j) N = number of hospital days or physician visits incurred by smokers and nonsmokers for neoplasms and circulatory and respiratory diseases. Attributable risks for all ages 17 years and over were calculated in a similar manner. Attributable risks for both sexes at a given age were obtained by summing the number of days or visits attributed to smoking for each sex and dividing by the total. (5) The direct costs attributed to smoking shown in tables 2 and 3 were derived by applying the attributed risks in appendix table I to total personal health care expenditures by type of care, age, and sex for neoplasms and circulatory and respiratory diseases. Costs of hospital care, nursing home care, and professional services (not including those of physicians) were calculated from the attributable risks for hospital care. The rationale for applying risk of hospital care to other professional services is that these services consist of home health services and private duty nursing care. Forty percent of home health services in 1980 was paid for by the hospital insurance component of Medicare and was for further treatment of a condition treated in a hospital or skilled nursing facility just prior to receiving home health services. Most private-duty nursing services were provided in the hospital. Costs of physician visits and drugs were calculated from the attributable risks for physician visits. Morbidity Costs Morbidity costs consist of the productivity losses to society, as measured by wages, salaries, and supplements, resulting from days lost from work among the currently employed, persons unable to work because of illness and disability, persons institutionalized for health reasons, and the imputed value of housekeeping services of women who are unable to keep house because of illness and disability. Using a methodology parallel to that for direct costs, attributable risks for indirect morbidity losses due to smoking were derived from the NHIS for work-loss days among currently employed persons, bed- disability days among females whose usual activity is keeping house, and persons unable to work or keep house (appendix table 1). These attributable risks were applied to the components of total morbidity losses for neoplasms and circulatory and respiratory diseases estimated by Rice, Hodgson, and Kopstein (1985) to obtain person-years lost and morbidity costs of smoking. Mortality Costs Mortality costs are the present discounted values of wages, salaries and supplements, and the imputed values of housekeeping services lost following the premature death of persons who would otherwise be productively employed or keeping house. As for direct costs and morbidity costs, the methodology consists of estimating attributable risks for deaths due to smoking and applying these to total mortality costs of specific diseases estimated by Rice, Hodgson, and Kopstein (1985). Mortality costs, however, are estimated for 19 specific causes of death attributed to smoking and are not limited to the sum of the losses from neoplasms and respiratory and circulatory diseases. Deaths in 1980 from these causes by sex and age were provided by the 1)ivision of Vital Statistics of the National Center for Health Statistics fi-om unpublished tabulations. Total person-years lost were estimated by applying the remaining years of life at each 5-year age group from the 1980 life tables (National Center.for Health Statistics 1984) to the number of deaths. Total mortality costs for the 19 causes of death discounted at 4 and 6 percent were calculated by multiplying the present value of future earnings lost from Rice, Hodgson, and Kopstein (1985) by the total number of deaths by age and sex. Attrib- utable risks of smoking taking into account two levels of exposure-

APPENDIX TABLE 2 Mortality Ratios from Prospective Studies Used to Calculate Attributable Risk of Smoking, by Sex, Smoking Status and Cause of Death ACS--25 state U.S. veterans British physiaans Swedish Maks Females ?+6iles Males Females Males Females Canse of Current death smokers Former smokers Current smokers Former smokers Current smokers Former smoken Curren smoker t Former s smokers Current smokers Former smokers Current Former Current Former smokers smokers smokers smoken 1[ALGNANT NEOPLASMS Trachl.7, bron- chus, lung 8.53 4.35 3.58 1.23 12.14 5.00 14.00 4.30 6.71 3.29 7.0 4.5 6.1 1.5 Lrytu 6.52 8.41 3.25 1.74 9.95 10.53 13.00 4.00 2.09 1.50 Lip, oral cavity, phuynx 6.52 2.25 3.25 1.74 8.25 2.05 13.00 4.00 2.09 1.50 - - - Bsophagus 3.96 1.66 4.89 1.87 6.17 1.57 4.67 1.67 - Bladder 2.55 1.59 2.00 1.94 2.15 1.55 2.11 1.22 - 1.8 1.6 2.3 Kidney 1.57 1.55 - 1.02 1.45 1.74 2.67 3.00 - Pancreas 2.14 1.37 1.42 1.15 1.84 1.17 1.57 - 1.30 1.22 3.1 2.5 4.8 5.5 Scomach - 1.26 - - 1.60 - - - - - - 2.3 - Cer.icnl - - - - - - 3.0 1.4 DISEASES OF THE CARDFOVASCUIAR SYSTEM bchemic heart disease Under 65 Mrs 2.03 1.39 1.77 1.16 1.76 1.29 1.7 1.3 1.5 1.5 65 years and over 1.36 1.17 1.28 1.27 1.61 1.30 Crrebrovascular disease 1.32 - 1.65 1.28 1.52 1.15 - - - - 1.0 1.1 - - 1.6 Hypertension 1.41 - - - 1.41 1.44 1.35 1.11 1.62 - 1.3 1.4 1.1 1.4 Aortic aneurysm 3.08 - 3.77 - 5.24 3.04 6.60 3.20 1.11 3.00 1.6 - 1.8 - Atheeosclerosis - - - - 1.86 1.15 1.38 - 1.44 1.29 2.0 2.0 1.0 2.8 DISEASES OF THE RESPIRATORY SYSTEM ..+..3^•ySeml. LhrOnlc }xoochitis 7.52 - .40 - 0.08 0.23 4.67 . 4.67 8.87 .00 Inflveary, pcxumonia 1.83 - 1.28 - 1.87 OTHER CAUSES OF DEATH Ae.spiruory tuberculosis - - - 2.12 1.26 5.00 3.67 - - - - - Ulcer 2.50 - - 3.50 2.50 2.50 1.50 - - 3.8 3.3 2.7 Sore+rer: American Cancer Society and U.S. veterans data from Hammond (1966); British physiciaris data from Doll and Peto (1976) and Doll et al. (1980); Swedish data from Cederlof, Friberg, and Lundman (1977). - Data noc available TIMN 305904

538 D.P. Rice et ul. ~rOOC~r~ M 00 N I~ V1 M O h • 3 ~ SrnC, 1--1 . ~ N N M r~. a ~ T N V~ 1~ N ••+ N O w ~ w ~ Q ~ S w > w ~ .~ ~ ~ ~ ~ . ~ a t3 ~ > O 00 V, '-O ^' N rn^o o°r`'~ N c ~ ~ ~ ~~00 00 4 ~ < c xa Q >' w ~ a o < `.~. a ~ V u ~ ~9 >v 00 C ~ E ~ s-a4 v~zzF~'~~ ~T v E .w W u C 79 4 W ~ 50 00 ~ 0 o 13 ~ .~ ~ I x [ ~ ~ qC QC Cq iq S4 S~ u u u o Zia ~ 1'!x Erononu Coit.i oJ the 1leulth 1:/Jecu of SncokinX, 1984 539 current and former smokers-were applied to the total costs to obtain mortality costs of smoking. The attributable risk formula for two levels of exposure can be expressed as (Walter 1976): Po = p, _ MR, _ AR=po+P, (MRi)+Pz(MRZ)- 1 po + p, (MR,) + P2 (MR2) proportion of those who never smoked in the population proportion of current smokers in the population mortality ratio for current smokers compared with those who never smoked proportion of former smokers in the population mortality ratio for former smokers compared with those who never smoked For males and females, overall mortality ratios for both current and former smokers were determined •for 19 diseases considered to be causally related to cigarette smoking by combining age-adjusted mortality ratios from four prospective studies on smoking: American Cancer Society 25-State Study, United States Veterans Study, British Physicians Study, and the Swedish Study (appendix table 2). Although there are other major prospective studies on smoking, these four are the only studies for which person-years of exposure data are available to estimate age-adjusted mortality ratios. Person-years are the number of persons in the study group multiplied by the number of years each person was followed until death or the study ended. Overall mortality ratios for each disease were obtained by treating each study as a separate stratum and weighting the mortality ratios by the number of deaths from each cause and the person-years exposed and unexposed in the four studies (Rothman and Boice 1979): L, ai_N../Ti Mortality Ratio = ' q' ~ b, N,,/T; a; = age-adjusted number of deaths in the exposed group in srudy i NQ; = person-years in the unexposed group in study i b, = age-adjusted number of deaths in the unexposed group in study i

~.n APPENDIX TABLE 4 -41 Weighted Average Mortality Ratios and Attributable Risk of Smoking, by Sex, Smoking Status, and Cause of Death 0 Mortality ratio Attributable risk (percenrage) Males Females Males Females C rmer F ICD Number Cause of death Current smokers Former smokers urrent smokers o smokers All smokers All smokers MALIGNA,"iT NEOPLASMS 81 8% 7% 45 162 bronchus, lung Trachea 10.02 4.47 3.67 1.29 . . 161 , Larynx 7.33 8.84 3.25 1.74 82.7 44.0 140-149 Lip, oral cavity, pharynx 6.62 2.28 3.25 1.74 71.6 44.0 150 Esophagus 4.80 1.65 4.90 1.87 62.1 56.4 188 Bladder 2.30 1.60 1.89 1.94 40.3 29.2 189 Kidney 1.47 1.63 1.50 1.02 27.0 13.2 157 Pancreas 2.00 1.37 1.48 1.26 33.0 15.5 151 Stomach 1.49 1.17 2.30 - 17.7 27.9 180 Cervix NA NA 3.00 1.40 NA 39.6 DISEASES OF THE CARDIOVASCULAR SYSTEM 410-414 Ischemic heart disease Under 65 years 65 years and over 1.88 1.49 1.38 1.20 1.67 1.28 1.17 1.27 31.7 15.4 19.3 O 7.9 ~ 4 4 lar disease sc b C 1.32 1.00 1.45 1.28 10.8 15.1 38 30- 401-404 u rova ere Hypertension 1.39 1.21 1.43 1.40 17.5 16.0 ;Q 441 Aortic aneurysm 4.46 2.95 3.19 3.01 65.6 49.2 2L 440 Atherosclerosis 1.83 1.14 1.94 2.40 26.3 33.3 a 427.5 Cardiac arrest• 3.00 - 3.00 - 43.1 37.3 ~ DISEASES OF THE RESPIRATORY SYSTEM 491-492 Emphysema, chronic bronchitis 10.13 10.97 7.40 4.89 86.7 71 6 ~ 480-487 Influenza, pneumonia 1.79 1 29 1 17 23 0 . 1 ~ OTHER CIUSES OF DEATH . . . 0.2 0 010-112 Respirato tubercul i ~ n ry os s 2.56 1.95 46 8 ; 531-534 Ulcer 2.88 2.12 3.21 2.45 . 51.3 47 0 . c'. • M li ~ orta ry ratios from U.S. P•abiic Health Service i983, 104; data for all other causes of death are calcslated irom Table 7. - Data not ;vailable ~ NA Nor applicable a TIMN 305906

542 D. P. Kia et al. - Nj; = person-years in the exposed group in study i T; = total person-years in study i(T; = No; + N,;) To illustrate this methodology, the overall mortality ratio of 10.02 (MR) for lung cancer for male current smokers was computed by applying the above formula to the data shown in appendix table 3. A similar procedure was followed to estimate the weighted mortality ratio of 4.47 (MR2) in male former smokers. Combining these overall mortality ratios for current and former smokers with the proportions of male smokers in: 1980 results in an attributable risk due to smoking of 81.8 percent for lung cancer mortality in men (appendix table 4). Ischemic heart disease (IHD) is the only disease for which age- specific mortality ratios were used to calculate separate attributable risks for smoking under age 65 and 65 years and over. Because mortality ratios for smoking for IHD decline with age (U.S. Public Health Service 1983) and deaths due to IHD rise rapidly with age, applying an overall attributable risk to all IHD deaths could have substantially overestimated the mortality losses for the elderly while underestimating the losses for those under age 65. To estimate mortality costs due to cigarette smoking by cause of death, the attributable risks in appendix table 4 were applied to the total number of deaths from the 19 specific causes of death attributable to smoking, to person-years lost, and to mortality costs discounted at 4 percent and 6 percent for males and females aged 20 and over in 5-year age groups. References Acton, J.D. 1975. Measuring the Social Impact of Health and Circulatory Disease Programa: Prelimntary Framework and listimater. Rand Report R-1967. Santa Monica: Rand Corporation. Atkinson, A.B., and J.L. Townsend. 1977. Economic Aspects of Reduced Smoking. Lancet 8036:492-95. Blanchard, C.G., E.B. Blanchard, and J.V. Becker, 1976. The Young Widow: Depressed Symptomatology throughout the Grief Process. Psychiatry 39:394-99. Bloom, B.S., R.S. Knorr, and A.E. Evans. 1985. The Epidemiology of Disease Expenses: The Costs of Caring for Children With Cancer. Journal of the American Medical Association 253:2393-97. Boden, L.1. 1976. The Economic Impact of Environmental Disease '1 he l.uonomic (;arti o% the Health 1 ffeeYS of Smoking, 1984 543 on Health Care Delivery. Journal of Occupational Medicine 18:467- 72. Campbell, J.D., and A.R. Campbell. 1978. The Social and Economic Costs of fnd-staKe Renal Disease. New F.ngland fournal of Medicrne 299:386-92. Cassileth, B.R., E.J. Lusk, T.B. Straus, D.S. Miller, L.L. Brown, T.A. Cross, and A.N. Tenaglia. 1984. Psychosocial Status in Chronic Illness. New England Journal of Medicine 3 l 1:506-11. Cederlof, R., L. Friberg, and T. Lundman. 1977. The Interactions of Smoking, I;nvironment, and Heredity and Their Implications for Disease Etiology: A Report of Epidemiological Studies on the Swedish Twin Registries. Acta Medica Scandinavica 612 (suppl.):7- 128. Cooper, B.S., and D.P. Rice. 1976. The Economic Cost of Illness Revisited. Social Security Bulletin 39:21-36. Derogaris, L.R., G.R. Morrow, J. Fetting, D. Penman, S. Piasetsky, A.M. Schrnale, M. Henricho, and C.L.M. Carnicke. 1983. The Prevalence of Psychiatric Disorders among Cancer Patients. Journal of the American Medical Association 249:751-57. Doll, R., R. Gray, B. Hafner, and R. Peto. 1980. Mortality in Relation to Smoking: 22 Years' Observations on Female British Doctors. British Medical Journal 280:967-71. Doll, R., and R. Peto. 1976. Mortality in Relation to Smoking: 20 Years' Observations on Male British Doctors. British Medical Journal 2:1525-36. Forbes, S.F., and M.E. Thompson, 1983. Estimating the Health Care Costs of Smokers. Canadian Journal of Pub/ic Health 74:183-90. Freeman, R.A., C.R. Rowland, M.C. Smith, S. Cabell Shull, and D.D. Garner. 1976. Economic Cost of Pulmonary Emphysema: Implications for Policy on Smoking and Health. Inquiry 13:15- 22. Garland, C., E. Barrett-Conner, L. Suarez, M.H. Criqui, and D.L. Wingard. 1985. Effects of Passive Smoking on Ischemic Heart Disease Mortality of Nonsmokers. AmericanJournal of Epidemiology 1121:645-50. Goldberg, R.J. 199 1. Management of Depression in the Patient with Advanced Cancer. Journal of the American Medical AJsacciation 246:373- 76. Gori, G.B., and B.J. Richter. 1978. Macroeconomics of Disease Prevention:n the United States. Science 200:1124-30. Hammond, E.C. 1966. Smoking in Relation to the Death Rates of One Million Men and Women. In Epidemiological Approachu to the Study of Cancer and Other Chronic Disraser, ed. William Haenszel, 127-204. National Cancer Institute Monograph no. 19. Wash- I

544 D. P. Rice el al. ington: U.S. Department of Health, Education, and Welfare, Public Health Service. Harvard University Institute for the Study of Smoking Behavior and Policy. 1985. The Cigarette Excise Tax. Smoking Behavior and Policy Conference Series. Cambridge. Hedrick, J.L. 197 l. The Economic Costs of Cigarette Smoking. HSMHA Health Reports 86:179-82. Hodgson, T.A., and A.N. Kopstein. 1984. Health Care Expenditures for Major Diseases in 1980. Health Care Financing Review 5:1- 12. Houts, P.S., A. Lipton, H.A. Harvey, B. Martin, M.A. Simmonds, R.H. Dixon, S. Longo, T. Andrews, R.A. Gordon, J. Meloy, arid S.L. Hoffman. 1984. Nonmedical Costs to Patients and Their Families Associated with Outpatient Chemotherapy. Cancer 53:2388-92. Institute of Medicine. 1981. Costs of Environment-related Heallh Effects. Washington: National Academy Press. Kristein, M. M. 1977. Economic Issues in Prevention. Prraentive Medicine 6:252-64. Lansky, S.B., N.U. Cairns, J.M. Clark, J. Lowman, L. Miller, and R. Trueworthy. 1979. Childhood Cancer Nonmedical Costs of the Illness. Cancer 43:403-8. Leu, R.E., and T. Schaub. 1983. Does Smoking Increase Medical Care Expenditure? Social Science Medicine 17:1907-14. Lewit, E.M. 1983. Some Economic Issues Raised by Reduced Smoking. Paper presented at the Annual Meeting of the Allied Social Sciences, San Francisco, December 28-30. Lilienfeld, A.M., and D.E. Lilienfeld. 1980. Foundations of Epidemiology. New York: Oxford University Press. Lubitz, J., and R. Prihoda. 1984. The Use and Costs of Medicare Services in the Last Two Years of Life. Health Care Financing Review 5:117-31. Luce, B.R., and S.O. Schweitzer. 1978. Smoking and Alcohol Abuse: A Comparison of Their Economic Consequences. New England Journal of Medicine 298:569-7 1. Marinelli, R.P., and A.E. Dell Orto. 1977. Tbe Psychological and Social Impact of Physical Disability. New York: Springer. Mishan, E.J. 1971. Evaluation of Life and Limb. Journal of Political Economy 79:687-705. Mushkin, S.J., and J.S. Landefeld. 1978. Nonhealth Sector Costs of Illness. Report A7. Washington: Public Services Laboratory, Georgetown University. National Center for Health Statistics. 1984. Vital Statistics of the United T he I:conomic Co.rtr of the l iealth 1 J%cts of Smoking, 1984 545 State.r, 1980, vol. 2, sec. 6, life tables. DHHS pub, no. (PHS) 84-11O4. Washington. . 1985. Health, United States, 1985. DHHS pub. no. (PHS) 86-1232. 1986. Trends in Smoking, Alcohol Consumption, and Other Health Practices among U.S. Adults, 1977 and 1983. Advance Data from Vital and Health Statistics, no. 118, June 30. DHHS pub. no. (P11S) 86-1250. Hyattsville, Md.: U.S. Public Health Service. National Reseaah Council. 1986. !s'nvironmental Tobacco Smoke: MeasurrnR lixpruurr and Aisessing Health Effects. Washington: National Academy Press. Office of Technology Assessment. 1985. Smoking-related Deaths and Financial Costs. OTA Staff Memorandum. Health Program, U.S. Congress. Oster, G., G.A. Colditz, and N.L. Kelly. 1984a. The Economic Costs of Smoking and Benefits of Quitting for Individual Smokers. Preventive Medicine 13:377-89. . 1984b. The Economic Costs of Smoking and Benefits of Quitting. Lexington, Mass.: Lexington Books. Ravenholt, R.T. 1985. Tobacco's Impact on Twentieth-century U.S. Mortality Patterns. American fournal of Preventive Medicine 1:4-17. Repace, J.L., and A.H. Lowrey. 1985. A Quantitative Estimate of Non.smokers' Lung Cancer Risk from Passive Smoking. Envimnmcnt 1 nternat ional 1 1: 3-2 2. Rice, D. P. and T. A. Hodgson. 1981. Social and Economic Implications of Cancer in the United States. National Center for Health Statistics, Vital and Health Statistics, series 3, no. 20. DHHS pub. no. (PHS) 81-1404. Waxhington. Rice, D. P. , T. A. L-Iodgson, and A. N. Kopstein. 1985. The Economic Costs of Illness: A Replication and Update. Health Care Financing Review 7:61-80. Robinson, J.C. 1986. Philosophical Origins of the Economic Valuation of Life. Milhank Quarterly 64(1):133-55. Rothman, K.J., and J.D. Boice. 1979. Epidemiologic Analysis with a Programmable (;alculator. NIH pub. no. 79-1649. Washington. Russell, L. B. 1986. I.r Prevention Better Than Cure? Washington: Brookings Institution. Schelling, T.C: February 15, 1984. Personal Communication. Schelling, T.C. 1968. The Life You Save May Be Your Own. In Problems in Public Expenditure Analysis, ed. S.B. Chase, 127-76. Washington: Brookings Institution. Scitovsky, A.A. 1968. Changes in the Costs of Treatment of Selected Illnesses, 195 1-65.'American Economic Review 57:1182-95. I

546 D. P. Rice et al. . 1985. Changes in the Costs of Treatment of Selected Illnesses, 1971-1981. Medical Care 23:1345-57. Scitovsky, A. A. , and N. McCall. 1977. Changer in the Cotts of Treatment of Selected Illnesses 1951-1964-1971. DHEW pub. no. (HRA) 77-3161. Washington. Shultz, J. M. 1985. Perspectives on the Economic Magnitude of Cigarette Smoking. New York State Journal of Medicine 85:302-6. Simon, J. 1968. The Health Economics of Cigarette Consumption. Journal of Human Resources 3:111-17. U.S. Public Health Service. 1964. Smoking and Health: Report of the Advisory Committee to the Surgeon General of the Public Health Service. PHS pub. no. 1103. Washington: U.S. Department of Health, Education, and Welfare, Public Health Service, Center for Disease Control. . 1982, The Health Consequences of Smoking: Cancer: A Report of the Surgeon General, U.S. Department of Health and Human Services, Office of the Assistant Secretary for Health, Office on Smoking and Health. DHHS pub. no. (PHS) 82-50179. Washington. . 1983. The Health Conrequencer of Smoking: Cardiovascular Disease. A Report of the Surgeon General, Office on Smoking and Health, U.S. Department of Health and Human Servicu. DHHS pub. no. (PHS) 84-50204. Washington. . 1984. The Health Consequences of Smoking: Chronic Obstructive Lung Disease. A Report of the Surgeon General, Office on Smoking and Health, U. S. Department of Health and Human Services. DHHS pub. no. (PHS) 84-50205. Washington. . 1985. The Health Conrequencu of Smoking: Cancer and Chronic Lung Di.rease in the Workplace. A Report of the Surgeon General, Office on Smoking and Health, U.S. Department of Health and Human Services. DHHS pub. no. 85-50207. Washington. Upp, M. 1983. Relative Importance of Various Income Sources of the Aged, 1980. Social Security Bulletin 46:3-10. Vogt, T.M. 1983. Medical Care and the Costs of Smoking. Public Health Reviews l 1:121-33• Vogt, T.M., and S.O. Schweitzer. 1985. Medical Costs of Cigarette Smoking in a Health Maintenance Organization. American Journal of Epidemiology 122:1060-66. Walter, S.D. 1976. The Estimation and Interpretation of Attributable Risk in Health Research. Biometrics 32:829-49. Warner, K. E. 1983. The Benefit.r and Coltr of Anti.rmoking Policie3: Final Report. Grant no. HS03634. Washington: National Center for Health Services Research. . 1985. Cigarette Advertising and Media Coverage of Smoking and Health. Neu, England Journal of Medicine 312:384-88. The Economic Co.itt of tfie Health I ffect.c of Smoking, 1984 547 . 1986. Smoking and Health lmplications of a Change in the Federal Cigarette Excise Tax. Journal of the American Medical As- sociation 255:1028-32. Warner, K. E. , and B. R. Luce. 1982. Cost-benefit and Cost-effective Analysis in Health Care. Ann Arbor: Health Administration Press. Williams, J.R., and C.G. Justus. 1974. Evaluation of Nationwide Health Costs of Air Pollution and Cigarette' Smoking. Journal of the Air Pollution Control Association 24:1063-66. Acknowledgments: The research on which this paper is based was supported in part by the Commonwealth Fund (grant no. 6516). The views expressed in this paper are those of the authors and no official endorsement by the Commonwealth Fund, the Universiry of California, the National Center for Health Statistics, or San Diego State University is intended or should be inferred. The authorc appreciate the helpful comments from two anonymous reviewers of an earlier version of the paper. Addrut correspondence to: Dorothy P. Rice, B.A., Sc.D., Department of Social and Behavioral Sciences, School of Nursing, N631Y, University of California, San Francisco, San Francisco, CA 94143.