Tobacco Documents Online

015q04 Cigarette Smoking and Lifetime Medical Expenditures THOMAS A. HODGSON National Center for Health Statisti¢~ CIGARETTE SMOKING IS KNOWN TO CAUSE LUNG cancer, coronary heart disease, cerebrovascular disease, chronic bronchitis, emphysema, and contributes to morbidity and mor- tality of these and cer~aln other diseases (U.S. Department of Health and Human Services 1989). Smoke~s at each age require more medical care than persons who have never smoked (Rice et al; 1986) and experi- ence reduced life expectancy (U.S. Department of Health and Human Services 1989). But are lifetime medical ease expenditures higher for smokers or neversmokers? Medical care use by the elderly is high and neversmokers, who live longer, might have higher lifetime medical penditures. The relationship of smoking to lifetime medical expenditures is an importam issue in terms of society's use of scarce resources, the impact on public and private health insurance programs, and which members of society bear the burden of financing medical care. Never smoking might be a cost-effective way to promote health, well-being, and a longer life even if neversmoker~ incur higher medical expenditures. However, never smoking has the greater benefit of being cost saving if expected lifetime medical expenditures are less for neversmokem I will examine this imue by estimating and comparing lifetime medical expenditures of smoket~ and neversmokem. T131060726

Studies of Medical Costs Previou~ Studies In this section I will briefly review previous studies that estimated life. time medical costs of smoking. Leu and Schaub (1983, t985) estimated that total expected lifetime medical c~re expen.ditures beginning,at age 35 for Swiss males who do not smoke will be higher than for ~mokers. Among Swiss males, the contribution of longer life expectancy to medi- cal care expenditures for neversmokers outweighed the higher average annual expenditures for smokers. In the first version of their study (1983), Leu and Schaub assumed that medical care utilization is related to smoking in the same way that motxality is related to smoking. Thus, it was estimated that the average male smoker has 8 percent more physician visits and 10 percent more hospital days per year than the neversmoker. In a revised version (1985), I.eu and Schaub analyzed the demand for medical care in Switzerland using an econometric model and concluded that smokers have somewhat fewer physician visits and slighdy more hospital days than neversmokets. The conclusion reached by Leu and Schaub in their first article, that smoking does not increase lifetime medical expenditures, was reaffirmed. In the United States. excess medical care utilization by smokers is much higher than that reported by Leu and Schaub. In the National Health Interview Survey (NHIS) Rice et al. (1986) found that the average male smoker (17 years of ag~ and over) had 19 percent more physician visits and 63 percent more hospital days per year than neversmokets. This is 2.4 times the excess physician visits and 6,3 times the excess hos- pital days reported by l.eu and Schaub for Swiss males. The higher an- nual excess medical care of smokers revealed in the U.S. data cumulated over the years a smoker is alive might mote than offset the impact of a longer life span on medical care use of aevemmokers. ldppiatt (1990) also reported that smoking lowers lifetime mediral costs. This was derived by deducting from expected lifetime medical penditures required to treat a. smoker for certain smoking-related dis- emes the additional medigal costs incurred during the longer life of a nonsmoker. Although methodologically sound, lifetime medical costs of smoking were underestimated because the data employed both underes- timate expenditures for the smoker's smoking-related diseases and over- estimate medical costs durhag the longer life of the nonsmoker. i TI31060727

[:or lifedmt costs of smoking-related dis~-s~s Lippia~r u~ed the figures of Osier er ~l. (19a4~.b) ~or ehe expected lffedme cos~ of lun~ cancer. coron2~ he~ d~sc~e, ~ad emphysema. Ahhough ~hese rh~cc condi- tions ~re impor~an~ smok~ng-rela~ed d~se~cs, m ~dd]don, smoking is a major agent tbr chronic b~onchic~, ccreb~v~ular dis~e. periphe~ atte~ ~clmivc dheme, md c~ce~ of the oral cavity, la~, and esopha~s ~ well ~ causing bladder cancer. Smoking also incre~cs the risk of pneumonia and influenza, abdominal aortic aneu~sm, and g~cdc and duodenal ulcers; it is a contributing factor in caace~ oE pancre~ and kidney: and it is ~s~iated with cance~ of the stomach and ute'ine cc~ix (U.S. Department of Health and Human So,ices 1989. 1990). Lung cancer, cocona~ he~ d~e~c, and emphysema ac- count ~ot le~ ~ half of ~e cotat short-term hospic~ days required for all d~es linked to cigarette smo~ng (Gr~ 1988). Just ~he ~ddi- dond d~e~es for which smoking ~ a maior cause require hospital days . equal to 60 percen~ of ~c total Eot the three conditions studied by Os. ccr. By limidng ~he calculation to cos~ o[ lung cancer, coronz~ dise~e, and emphysema, Lippiatc omitted substand~ morbidity, mor- tals, md health c~e utilization ~d sevetely underestimated lifetime medkd cos~ of smoking. In o~&~ co take into ~ccouat the bngct life and medic~ c~e expen- ditur~ o~ nommoke~ during these extra yea~s, I.ippiau ad)usted the timaces by Oster ec d. o~ lifetime cos~ o~ smoking.rdated dise~es. w~ done by subtracting estimated average annual per capita medic~ ex- penditures of no~moke~s over age 65 for each year of difference in life expectan~ be~een smokers and nonsmokers. Average expenditures for nonsmokers were derived from pe~ capit~ expeadku~es for ~e total pop- ulation ove~ age 65 (tmokets ~d no~mokets), the proportion o~ smok- e~ md no~moke~ in ~ population, md ~e difference in average annual medical expenditures be~een smoke~ and nonsmokers reported by Leu md Schaub (1983). Because Lcu and S~aub severely under~ti- maced ~ d~erence in annual medic~ care me md ~xpendimres be- ~een smokers md nonsmokers in ~e United States, Lippiatt's estimate of a ao~mokefs annual medical ~eadku~ k ove~caced. By excluding expenditures for dishes ~own to he ca~ed by smoking and underrating medic~ c~e utilization md cxpendi~e differences be~een smoke~ and no~mokers, ~ppiatt underestimated lifetime medi~ cos~ of smo~ng. 8e~e ~ppiact finds ~e uadeoff be~n me~c~ expenditures and life expectm~ to be only $280 per yc~ of ex- TI3"1060728

tra life for nonsmokers (in t986 dollars), we expect mote -~ccutatc esti. maces of the ~osts of smoking-related diseases and the annuzl medical expenditures of nonsmokers would produce lifetime medical costs highes for smokers than nonsmokers. Manning ecal. (1989) examined lifetime medical czre costs of smok- ing from a somewhat different perspective, but found a posit~v¢'rel:z- tionship. Their best estimate is that medical ca, e costs of smoking were $.26 per pack of cigarettes smoked in 1986 dollars discounted ~c 5 l~ettent. In our anaJysis we me a life-cycle model to veri~ the findings of ning et ~. (1989) that in ~e ~ted States smokers have higher lifetime medk~ expenditures. We ~so e~d upon ~eir ~sis to ex~ine ~e ~ing of ~eMimtcs over the life cycle, populatioa ~ well ~ iadi- vidu~ ¢~endi~, ~d sourc~ of payment for mcdic~ c~e. Data em- ployed ~e for ~¢ U.S. popularioa and include medic~ care use and mon~i~ for dl dhgnos~ ~d comes of d~. ~us overcoming the l~tadons in ~¢ Leu ~d S~aub ~d ~ppia~ studies. ~e~e medkd c~e expeMi~t~ are estimated for m~tes ~nd fc- m~ in ~6 United States whb never smoked and for moderate md h~ smoke~, including bo~ ~ent md foyer smoke~. ~foderate smoke~ ~epo~ed smo~g f~er ~an 2~ ci~rettes a day and heav smoke~ smoked Z5 or more per day. Analyzing eversmoke~ (that is, ~rreac md foyer smoke~, he~c~er called smokers) c&es i~co account • e a~bet of yem of smoking md patter~ of quitting and recidivism e~dng ~ ~e population zc ~e ~e of data collection. ~us, esti- mated e~ccted lffe~e medical ~endim~ of a smoker reflect average e~efienc¢ ~ ~e ~pularion of perso~ who take up smoking md hdud¢ ~e ~p~ oa e~ea~mr~ of dcc~ioas co quit smoking. • h ~dg. comp~on of l~edme expenditures of smokets md never- smoke~ ~ows m to ~ ~e ~pacc of beco~ng a smoke~ versus not b~oming a smoker, but do~ not addtca ~e ~pact of quitting smok- hg on me~ ~e ~endimr~. Subsequent r~e~ is plmn~d to dyz~ quic~g md lffe~e medi~ F~om ~e m~aces of ~e~e medical expenditures we determine: I. the amount of excess medical expenditures requited by smokers 2. the relationship of medical expenditures to amount smoked T131060729

4. the timing of medical expenditures during the life cycle and the phases during which expenditures of smokers exceed those of nev- crsmokcrs and vice vccsa 5- the monetary burden smoking imposc~ on private sources of fund- ;ng (for example, individuals and employers) and public sources (for example, the federal government's Medicare program) 6. the distribution of current medical care expenditures among the population of smokers and persons who never smoked 7. the ongoing bill for excess medical care required by the popula- tion of smokers 8. the aggregate furore excess expenditures of the current population of smokers Ot~er Eaonomic Costs There me other economic costs associated with smoking in addition to medical ca~e expenditures. These include expcndituses and payments lated to sick leave, disability, group life insurance, pensions, and retire- meat benefits (Manning et al. 1989: Warner 1987). The impact of smoking on Social Security benefits is among the most knporsant of these and is substantial. Shoveno Sundberg. and Bunker (1987) estimate that because o~: shorter llfe expecran~ single male smokers earning the media-q wage receive almost $18",000 less in benefits than they contrib- ute, whereas nonsmokers receive almost $3°400 mote than they pay in (in 1985 dollars). For single women with median earnings the difference between smokers and nousmokets is smaller, but still considerable. Smoking thus results in a net tranffe~ of Social Security benefits from smokers to nonsmokers. Further consideration of economic implications of smoking other than medical ca~e expenditures lies b~/ond the scope The Model The model estimating lifetime medical caxe expenditures is briefly de- scribed here, with additional derails p~ovided in the appendix. Medical care use, costs, and mor~ali~ experience of cross-sections of th~ popula- tion during each age interval ase used to generate longitudinal ptof~lm TI3q OfiOTE

of costs from age 17 to death. Expected. in the sense of average, lifetime expenditures rather than actual lifetime expenditures of any one indi- vidual a~e estimated. The principal data sources are the National Health Interview Survey for use of hospital and physician services; the National Nursing Home Survey and the National Health and Nutrition Exarfilna- r.ion Survey Epidemiologic Followup Study for nursing-home carL: the American Cancer Society's Cancer Prevention Study Ii for mortality; and the National Medical Care Utilization and Expenditure Survey and Medicare data files for charges for medical care. In general, an individual's expected expenditures during age interval t are given by: where ~, = expenditures during age interval t if the individual survives through t El, = expenditures during age interval t if the individual dies in t P~ = probability of surviving through age interval t P,, = probability of dying during age interval t" " • It is necessary to distinguish whether the individual survives or dies be- cause much higher expenditures are incurred by decedents than sur- vivors. Lifetime expenditures from age 17 are given by the sum of expected expenditures, Et, during each of the age intervals: t = ages 17-34, 35~!4, 45-54, 55-64, 65-74, 75=84, 85 and over Expenditures are discounted to obtain the present value of the stream of dollars occuuing over time. It is assumed that all persons surviving to age 85 enjoy the average remaining lifetime calculated by the National Cente~ for Health Statistics, or approximately five years for males and six years for females (National Center for Health Statistics 1990). This simplification is acquired by lack of data on life expectancy at age 85 for smoket~ and nevem'aoke~s resulting most likely in overestimates of ex- penditures for smokers and underestimates of expenditures for never- smokers at ages 85 and over. The impact on lifetime expenditures is T131060731

negligible, however, because expected expenditures at age 85 and over are a small proportion of the total, especially among smokers. Lifetime expenditures are estimated for males and females, and ac- cording to amount smoked (never smoked, moderate, or heavy smoker). Age- and sex-specific rates of medical care use and mortality according to amount smoked are employed. Medical case expenditures included ate for short-term inpatient hospital care, physicians' services (to hospi- tal inpatien~s and ambulatory patients in doctors' office~, hospital clinics and emergency morns, patients" homes, and by telephone), and nursing- home care. These medical services account for about three-fourths of to- tal personal health care expenditures (Wald~ et al. 1989), The principal services omitted from the analysis because of lack of data on how utiliza- tion relates to cigarette smoking are drags and dental services. Medical care utilization and expenditures are not evenly distributed throughout the life cycle. Variation of medical case use with age is easily accounted for by employing age-spedfic data. Equally important, dece- dents requite much more medical care and incur far greater expendi. tures than survivors among both elderly and nonetderly populations. Decedents have higher expenditures relative to survivors, not only in the year of death, but also for several years prior to death. The disparity in. expenditures of decedents versus survivors increases as the time of death approaches and may be more than six times greater in the year death curs {Lubitz and Prihoda 1984; Riley and Lubitz 1986; Roos, Montgom- ery. and Roos 1987). This phenomenon is an important aspect of lifetime medical expenditures and is included in the model. 1990 Dollars Expenditures in this analysis are in estimated 1990 dollars, with dollar magnitudes adjusted to 1990 according to increases in the medical care component of the consumer price index (Social Secutit~ Bulletin 1991). Discounting Medical care use and expenditures are highly concentrated in the later years of life, esp~dally in the several years before death. Because never- smokers live longer, their medical care expenditures are deferred to the future compared with those of smokers. The very long time horizons in this analysis (65 yearn or more in some instances), and the different dis- TI31060732

triburions of expenditures over rime for smokers and neversmokers, re- quire ~at lifetime expenditures be discounted in order not to overstate the amount for neversmokers compared with smoken. This analysis em- ploys a relatively low, but reasonable, discount rate of 3 percent. Kc*y re- suits are also presented for a 5 percent discount rare to show the irhpart of discounting. Total expected lifetime expenditures discounted at 3 percent are about one-third of nondiscounted expenditures. Causality or Association? How much of the difference in medical care use ahd expenditures is due to smoking and how much to other factors that are not equally distrib- uted among smokers and neversmokers? Smokers differ f~om never- smokers in certain genetic, social, behavioral, and economic characteristics that may contribute to use of medical care. Positive correlations have been reported between smoking and drinking alcohol. The Behavioral Risk Factor Surveys. conducted from 1981 to 1983, found that more heavy smokers (more than one pack a day) had two or more drinks a day than nevemmoke~s (Bradsrock er al. 1985). In a study at the Group HeAth Cooperative of Puget Sound, current smokers were more likely to be problem drinkem (Pearson et al. 1987). Data from the National Health Interview Survey (NHIS) show that in 1985 smokers were more likely than neversmokers to drink heavily, not exercise actively, sleep six hours or less, and skip brealffast (Schoenbom and Ben.son 1988). How- ever, smokers, especially those who smoked fewer than 25 cigarettes daily, were less likely to be overweight and to snack daily. . If factors related to heAth stares and smoking habits are ~ot con- trolled, the impact of smoking on health and medical expenditures may. be overstated. There is evidence from several sources, however, that most of the observed difference between smokers and neversmokers in mortality, medical care use, and expenditures is the result of smoking and is not just comclatcd with it. Nevcrsmokcrs, especially males, have higher income and mozc education than smokers, but the difference in medical ca~e use cannot be attributed to health habits, practices, or li.fe- styles related to incom,e and education because smokers use mote medi- eA care at all levels of income and education according to data from NH/$ as computed by the Of/ice of Analysis and Epidemiology. bhtt- son, Pollack, and Cullen (1987) estimated death rates for males in the United States in 1982 for smoking-related dLseascs and for all causes of Tl31060733

death according to age and smoking stares. Applying these estimates to ch~ number of males in the civilian noninstia~tionalized popula~on in 1985 by smoking status, it can be calcu)ared that 74 percent of excess deaths among male smokers aged 35 to 84 was due to smoking-related diseases. An interesting s~aristicaJ construct, the nonsmoklng ~moker-rype, has been employed to assess medical care expenditures due to smoking tarher than just associated with smoking (Leu and Schaub I983; Man- ning et al. 198~). The nonsmok~ng smoker-~'pe does not smoke but is like a smoker in o~her respects that distinguish smokers from never- smokers and contribute zo mothidity, mortality, and medical care use. These include education, fu.mily income, race, health insurance cover- age, and lifestyle attributes inch as drinking habits, exercise, and seat belt use. Thus, the nonsmoking smoker-type experiences medical care use and mortality that lie between those of the smoker and never- smoker. Higher medical care use and higher morzality have opposke impacts on lifetime expenditures. Thus, the higher medical care use of the non- smoking smoker-type ~ill increase lifetime expenditures relative to nev- ersmokets and decrease the excess lifetime expenditures msociated with smoking. This will be pa.~tially offer, howeve¢, by the impact of the higher morr~li~ rates of the nonsmoking smoke~-type, which reduce life expectancy and thus lifetime expenditures telarive to nevetsmokets and increase excess-lifetime expenditures of smoking. Controlling for other differences bet'~een smokers and nevetsmokets besides smoking that af- fect medical costa has,a zather small impact on excess lifetime medical cxpendi~res according to research reported by Manning ctal. (1989). Manning and his colleagues estimated lifetime medical costs per pack of cigarettes znd found that excess lifedme costs of smoke~s compared with nonsmoking smoker-types were 87 pc~cen, of excess lifedme costs of smokers comp~cd with neversmokcm. Although the 10~efcuecl comparison for asccrmlning medlcaJ ca~¢ pcndimrcs duc to smoking is between the smokce and nonsmoking smoker-type, we a~e only able to compare cversmokers and nevetsmok- e~s in our study. Nevertheless, became Manning et al. also used data from the NHIS, ir is ~asonable to conclude from thei~ rcsalts that the findings we repor~ would be only slighdy different q~andta~ivcly and no different qualitatively if formulated in terms of ~mokcm remus non- smoking smokee-tTl~s. ~ 4 TI310~073

Lifetime Expenditures Mortality Smokers have higher death rates than neversmokers at all ages over 3~ years (figure 1). The analysis begins at age 17 because data on medical care use and expenditures by smoking status are available beginning at this age. However, we lack data on mortality by smoking stems for per- sons aged 17 to 34 and it is assumed that no deaths occur until age 35. This assumption should have a negligible impact on our results because less than 4 percent of persons die before age 35 (National Center for Health Statistics 1990) and smoking should not be a major determinant of mortality between the ages of 17 and 35. Excluding. deaths prior to age 35 has a slight impact on lifetime expenditures of both smokers and neversmokers and even less of an impact on the difference in their ex- penditures. Death tares ~e steadily with age, are higher for males than females, and higher for smokers than neversmoke~s in each sex. Probabilities of survival are derived from the death rates. Table 1 shows the probability of an individual 17 yea~s old surviving to the age Dealh rate 0.7 0.6 ""~(:-- Male ell emokere • ~]'" Female ~leveremoke~'e 17-34 35~4 4~ 66-64 e6-74 75-84 TI31060735