Tobacco Documents Online

I SAMMIEC II Smoking-Attiibutable Mortality, Morbidity,-and Econom.i: Costs Computer Software _ and - Documzntation Project Staff 40 James M, -Shultz, M.S., Ph.D. - Department of Epidemiology and Public Health University of Msami School of Medicine 1029 N.W. 15th Street (R~69) Miami, Florida 33136 (3J5) 347-6972 - - Thomas E. Novotny, M.D. ChieL Program- Services Activity _ Office on Smoking and Health Center nter for Chronic hronic Disease Prevention and Health Promotion - Centers for Disease Control Public Health Service U.S. Department of Health and Human Services Rochville; Maryland 201i57_ (301) 443-1575 Dorothy P. Rice, M.S., Sc.D. (Hon.) -- Department of-Social and Behaviorai Sciences School of Nursing - University of California San-Francisao, California 94143 The authors wish to aclmowledge- the editorial and production assistance provided by HCR and- iu staH, including Donna Grande, Project Director, and William Thomas, Jr., Information Support Coordinator.

A ., , SAMMEC II Smoking-Attributable Mortality, Morbidlty,_ s€nd Ec°onomic Costs Computer Software and Documentation Module 2: t1'fethodology -and Conceptual Issues James M. ShtlYta, M.S., Ph.D. Thomas E Novotny, M.D. Dorothy P. Rice, M:S.,- Sc.D. (Hon.) Prepared for. __ - - Otfiee on Smoking and Health Center for Chronic Disease Prevention and IIleeltb Promotion Centers for Disease Control Public Health Service - US. Departenent of Health and Human Services October O40

SAAMIEC II 1Vf odule 2: Iviethodology-ard Conceptual Issues Table of Contents Page Chapter 1: Ile Disease Impact of Cigarette Smoking _ 1 -Chapter s:- Smoking-Attributrble Disease Impact Estimation Rationale - - - 3 Data for the Group Under Study - 3 Epidemiologic Measures 4 Economic Cost-of-Illness lvieasures 4 The Role of Computer Software 5 Chapt.er 3: Smokirtg-Attribntable Desease Impact Estimation: Conceptual Issues - 7 ibe Scope of the Calculations 7 Interpretation and Use of the Estimates - 9 Disease Impact Estimation as a Health Policy Intervention 10 Conceptual Issues for Cost-0f-Ill.•g= Studies -_ - - - 10 Estimation ls.sues for Cost-of-Illness Studies - - 12 Chapter 4: Sretoking-Attributable Cost-of-fllness Studies: Review of the :aterature 17 Cost-of-fllness Studies - 17 Cost-0s i lness Guidelines - - 17 National Calculations of Economic Gosts-Attributable-to Smcldn; United States 18 Natioaal Calculations of Economic Costs Attributable to Smoking: Other Gouutriea - 20 State Calculations of Econosnic Costs Attrshutable to Cigarette Smoking 20 Economic Costs to the Individual Annbutarbie to Cigarette Smoking 21_ Cost Offsets: Extended Iafe Expectancy for Nonsmokers- and Former Smokers 22 Summary - - - 22 L J./ CG't QD b.I Xh

0 j Chapter S: I'be Disease Impact Measures_ - 23 Definitions of Smoking-Related Disease Impact Measures 23 Chapter 6: Smoking-Related Diagnoses - Chapter- 7: Smohng-Attributable Fractions 27 - (,hapter 8: Calculation of Smoking-Attributable Mortality 31 - Smoking-Attributable Mottality. An Over-view= 31 - Estimation of Smo1~`ng-Attributable Mortality - 31 Comments on Smoking-Attributable Mortality-Related Measurm 31 Smoldng :4?trtbutable Mortality Rates 32 Chapter 9: Calculation of Smoking-Attributable Years of PotentW bife Lost 33 Smoking :Attributable Years of Poter~tial -Life Lost:- An Overview 33 Fstimation of Smc~g-Attr,~uta.bl euta.ble Years of Potential Life Lost 33 Smoldng-Attnbuta-ble Years of Potential Life Lost_ Rates - 33 Chapter 10: Calculation of Smolcing-At-t:ibutable Indirect Mortality Costs 37 Smoldng-Attn'but~kle Indirect Mortality Costs: An Overview 37 Estimation oi Smoking-Attributable Indirect Mortality Costs -- Smoking-Attributable Indirect Mortality Cost Rates - 37 38 lK.N Chapter 11: Calctraat:an of Srnoking-Attributable Diaect Health-Care Costs _ 41 - Smolr:og-Atta-ibutable Direct Coats: An Overview ' 41 Estimation of Smoking-Attributable pirw Health-Care Costs - 41 Chapter 12: - Calculation of Stnoking-Attributable Indirect Morbidity Costs 43 = Costs: Ap Overview ble indirect Morbidit but Smola Att 45 y ng- ri a Estimation of Smoltir.g-Attnbutable Indirect Morbidity Codts_ -- - -- 45 Iteferences - - 49

Chapter _ I The Di-sease_ Iznpact of C:gaa ett=e Smoking Cigarette smoking is the chie& single preventable cause of premature mortality in the Unit~ States (U.S. Deparu~ent of Health and ]`Ier~man Services, 1986). A series of reports by t.he Surgeon General on tb E- health consequences of smoking has documented the contribution of _ cigarette smoking to deaths e.aused by cancers (USDHHS, 19182), cardiovascular diseases - (USDHHS, _ 1983), and chronic obstructive pu!ffionarq diseasm (USDFi..~i,S, 1984). From these three reports, national smoking=attnbutab1= mortality was estimated at 350,000 deaths in 1980 (equivalent to 17% of total national mortality). Other estimates of mortality attnbutable to cigarette smoking have-been reported, including-27Q,C00 deaths for 1980 (l~ce et_aL, 1986), 314,000 deaths- for 1982 (Office of Technology Assessment, 1985), 320,000 deaths for 1984 (Centers for Disease Control, 1987) and 390,000 deaths for 1985 (L3SDH~iS, 1989). On average, each smoker_ who dies ftom-a smoking-related disease forfeits 15 years- of life compared with his or her- nonsmaking counterparts (.Varner,-1987). -Morbidity rates are higher for-cigarette smo-kers- throughout their life, particularly from respiratory diseases,- than for persons who have never smoked. - Moreover, both ac the workrite and at home, passivcsmoking (the inhalation of sidestream emissions from cigarettes smoked by co-workers or family members) -increases the risk for lung cancer, coronary l±eart- diseau, and respiratory disease am ong nonsmokers= (LJSDHHS, 1986). In the United States,- direct medical costs #or the detection, treatment, and rehabilitation of persons with smoking=attribu3ablc clinical diseases have- be-en estimated to exceed S23 billion _ annually (1984 dollars, Rice et al., 1986). Indirecc morbidity costs, defined as the costs for excessive 3ick leave days and disability days for smoldng-linlced illnesses, are regarded as a negative productiviry-facsor estimated to -total $9 billion-annually. Indirect mortaiiry costs, defined as the economic value of forfeited futurc :.arnmgs for persons who die prematurely from smoldng-relatecl causes,- are-valued -at -$2 1 btllion annually: Tae- total of these three cost measures (S53 billion) is -equivalznt to 7% of total national direct costs of illness plus 1~0 of illness-associated indirect costs. Not included in this estimate are the psychosocial oosts associated with smoking-related disease& These data provide compellang reasons for the active promotion of nonsmoDdr_g. Such activities include health education among youth for primary prevention of smoking°, smoking cessation programs for persons addicted-to zigarettes; economic disincentives to smoking; and-regulatory controls on the production, marketing, and sale of-tobacco products. "I-be development of smoking-control policies and educational programs relies heavily on State, local, and corporate decision-makers. For each jurisdiction, the following questions can be -- asked: 'How many people die from smoking-caused illness?' 'How much- is smoking costing tu?' 'What would be our return on- investment for a nonsmoldng initiative?' For each question, the concept of cost includes both tangible -and intangile costs. Answers to some of I

these questions are .~s~le frotF.~ existing data sou~. •3uribg the past decadc, incrcasingly sophisticated methods have been applied to estimate the eeonomic costs of smoking (Luce and Schweitzer, 1978; Forbes and 'fbompson, 1983; Rice and 1-Iodgson, 198?; Oster et aL, 1984; O6ce of iechnology Assessment, 1985; Rice et a3., 1986). - - 'Disease impact estimation' is used here as-an expression for the process of quantifying a health problem from several vantage- points, including -morbidity, mortality, and economic costs. The combination of epidemiologic estimates of the human costs of smola.ng-(mQrbidity and mortality) with indicators -of sn;oking-attrib_ utable econorn:c-costr provides a comprehensive-pictung of d.isaase impact. Although= disease impact estimation methods have been used with national data, we of these methods -vvith local data ha-s been technologicall~y d.ifficult. _ As a=r~ul~ thc-h~nesota - Department of He~~h (IyIDPl~ developed mici•ocomputer software for estimating amol4ng- attributable disease impact (Sbultz, 1986a, 1988). Ile- software evolved during development of a statewide smoleng-;ontrol plan _(INIDK-1984; Dean et aV, 1985, 1985), -the passing of-oon= smoldng legislation (Shultz et a! , 1986), and the implementation of the program atic phase of the plan (the kii.an_esotd -Nonsm~okdbg Initiative). 'Ile- software-was named SAMMEC _ (Smoking-Attnbutablc Mortality, Morbidity, and Economic Costs). - SAMMEC was used_ to produce a national estimate of sraoking-attri'cutable mortaiity and yeztrs of potential li:e iost-for 1984 (CDC, 1987). In additiQn, -State-speciFic calculations -for -198r were performed-by State-based_CLC-personnel and State employees in all 50 States, Puerto Rico, and Washington, D.C. (C7 C, 1988). These data are pubiished in=the 1990 bicnnial status report to- Congress pi=oduccd by the O~'i~ on Smoking and Health ,USDF€HS, 1~). - SAMMEC was also -Used-to develop disease impact estimates for other nations, including Austraua, France, the People's FcepubhC of China, and-selected provinces in Canada (Health Department of Western Australia, 1937;-Hi_11 and Giron,-1982+World Banlt,_ in press; Collishaw and Myera, 1.984). ~ Tbe current and -expected disease irnpact of Eigarette_ smoking for less developed countries (LDCs) has- not been-adequately quantified (Ltwit, 1988). Smoking prevalence rates, on which predictions of smolcng-attn-butable disease depend, are increasing in many LJ>CL- 4lthough - smo;<ing-attnbutable d:sea_se in;pact cstunation is important for LDCs, calculation procedures L the lack of mortality, and software tools su ch as SAIvsNEC _P may be inappropr'-ate because of prevalence, and -econdmic datd in these countries. -- - SAMMEC II is the second phase of SA,'+INffiC applicatiori,s. - SnMRMC II is a new software product-that improves on the 5nethodology of its predecessor and sccornmodates additional types of health data. OD _ VI AD -- - ~ ~ !-a 2 7~ Q #N1

- - Chapter Smoking-Attributable Disease Impact Estimation ~~. the tobacco industry (Schelling, 1987; Warner, 1 RadonskA formidable cass against tobacco use has -been made by estimating the disease burden of cigarette smolang (IJ.S. -Department-of Health, Education, and Welfare, 1979). .is burden has- tieen quant~ed in te~s -of smo~ng-~t~'btstable mortality, years of pot.ential ;ife iost (YPLL), excess medical care, =and e;cess disability (Rice et aL Economic measures belp define-the impact of cigarette-smolcing on health-care syitems-snd on the productivity of the population (Shulta, 1985a; Office of 'f'echnology Assessment, 1995; Rice et aL, 1986). Estimates of smoking-attributable econotn.ic costs are useful for developing health policy and for planning smoking-control initiatives. - In additiona cost ertimates may help -- polieyaFakers make decision-, about- tobacco-control activities (Shulu and Moen, -19$6; St,ultz et al.,-1986,_Smitl:i et aL, 1990). - Previous -smolting-related disease impact studies do not descn'be al! dimensions of smoking and disease. For ez:ampie, the pain and- sufi'enng, decreased quality of life, and related- Psychosocial - aspects o, physical illness are not- measured- (Abt, 1975; HHodgson and- Meiners, 1979). Prevalence-based =t-o; -t7iness calculatioas do not account for economic factors such as Social Secutitv disbursements, pension claims, changes in the demand for bealtb- specialties related to the treatment of smoh~g-associated illness,- and the,"empl-oy3nent by or monetary dividends from Data for the Group Under Study Five -sets_ of data are necessary for computing disease impact measures: mortality data,_smokiDg prevalence rates, b+eaith-care cost data, earnings data, and populatioa data.- These data are available -for all States and for some large municipalities. Outside the United States, the quality and availability of these data vary considerably. State-specific smoldng-prevalence rate data are available for States p$rticipating in_tbe telephone-based Behavioral Risk Factor Surveillance System (Bft.) conducted by the Centers for Disease Control (CDC) (Remington et al., 1985; CDC, 1989a j. The BR~'SS provides comparable data for all participating States. Tle Office on Smoking and Health used smoking prevalence data-from the Current Population Survey (CPS) in its State-specife calculations for 19&5 (Marcus et al., 1989). State-speciflc CPS data for 1989 are now avai7able. Estimates of direct bealtb-care costs attributed to smoking require an estimate of total personal health-care expenditures for the prevention, diagnosis, treatment, and rehabilitation of all i7lnes3es for the group under study (Hodgson, 1983; FRice et aL, 1986). In the United States, health-care cost estimates used in SAJvIIv€EC II may be available fiom State health agencies, the 'A

Health Care Financing Admiaistration (Levit, 1985; Lazenby, 19W, or university-based health services researchesa. For comparabdity among States,-co3ts can be prorated using national personal health-care expezsditures compiled by the Health Care-Financing f5dministrationfor five cost centera: hospita izz-tion, physicians' fee's,-nursing home fees, medications, and other professional servicm Epidemialogic Measures The pivotal calculation in di;easeim impact estimation isthe att-rsnutable-risk, or'smoking= anttbutable frac€ion' ~el~, 1987). This measure, -de~ne + as the maximal proportion of casc~ of a disease causally=liaed to cigarette smoking, is a function of two otber-- measures: current and former smoking prevalence rates and relative risks (Lilienfeld and Menfeld, 1M,-Walter,- 19i6). Diagnosis-specific rela - tive risks for smolcing-related diseases, -defined as the ratia of mortality- -among current or former smokers to that of never -smokers, have been developed -from several large prospective studies-of-smoking and mortality (Hammond, 1966; Kahn, 1966;-Doll and - Peto, 1976; Cederlof ec aL, 1977; Doll et aL, 1980; Stellrnan and-Garfinkel; 1985; U.S. Department of-Healtn and =Human Ser•~is.r.s, 1989). Relative risk estimates for smoking diffzr by sex and study population (USDHEW, 1979; ??SDHHS, 1989): Relative rWa are typically lower for former smokers than for current smokers; for former smokers, relative risks decrease as the number ofyears after cessation increases. However, age-specific relative r-isk estimates are not avai7abl-e for most smoldng-related diagnose& S?tlvD:ffi-C H software benefits hrom-the inclusion of updated _relative risk estimates derived from the most recent American Cancer Society data, the Cancer Prevention-Study lY (CPS-I~) (USDHI:S, 1989). These.-relativerislt- estimates are based on a fou3-year follo°•a-up study (1982 -to 1986)of 1.2 million entrants of the CPS-IL For smoking-related-diagnoses in the CFS-H, the relative risks-yvere calculated by - comparing the age-adjusted mortalit•j rates for current -and former smcrkers- with those of never smokers. For a specific population, the smoking-attributable fraction may be an overestimation or an underestimation- if the physiological ,and behavioral sharacteristics of the populat;otk are not comparable to those of the- study populations -from which the relative risk measures were - derived. Use of smo--lring prevalence rates for the current year also may result in an under- estr,mate of smoking-attn'butable diseaseand-death. Overall smoking prevalence rates are lower in recent years_ than in the- previous 30 years (fJSDHHS, 1989), and the burden of most chronic diseases linked to- smoking, such as cancer and obstructive -pulmonary-disease, reflects previous decades of higher smoking prevalence. Economic Cost-of-Rlness Measures- Health economics measures calculated by SA1viN.EC II also use a prevalence-based method- ology (Rice et al, 1986) that estimates the current annual costs of the lifetime smoking = behavior of persons who receive medical treatment, are absent-fromwork,-os succumb to smoking-induced Wness during the year under study (Scheiling, 19S7). Grossly aggregated

) econom,ic data on health-care costs do not permit diagnosis=;pecific estirtmatrs.. Personal health - expenditure data are typically sum- mariaed-by type of cost (e.g., bospital costs, m_edi6ation costs) and only occasionally by disease catepa:y- (H9dgson and Kopsseirc, 19£~4',.' Data are incomplete for patterns- of morbidity and patterns of inedieal use by disease status. The National Center - for Health Statistics has documented excess rates of hospitalization and use of physician services for smokers compared with never smokers (Rice et aL, 1986). However, a study of medical use, by smoking status, for participants in the Kaiser-Peririane8te health plan indicated that rates of outpatient service -utilization did not differ between current smokers and persons who had never smoked; smokers were less likely to seek preventive medical services (Vogt and Schweitzer, i98"Z). Availability s~d ac~i.*aey of estimates of total person$1 health-care crpenditures also vary by State. Ile entire array of health econoaics data, as well as its completeness and accuracy, may -v°ary considerably in other nations. Tae Role of Computer Softva:+g Spreadsheet soft-waPZ=incorporates both epidemiologic and ealth_eco:iorni~ data (Shultz, 1985b, 1986a, 1988; Sbuliz~ Rice, and Ir:odesoz:, 1986). - Calculati~ i~nr can be reproduced rapidly and accurately. Standardized software can generate reports spccific to populations- under study and provide comparable estimates for similar populations. Ho•vever, unercical use of disease irnpact data produced by SAN'wfEC II may lead to- misinterpretation or overinterpreta:ion_ of findings. 'rhese Iimitations_ ~re discussed later-in-this docun:ent. - 39 - -- - CJY CX' ~ ~ N ~ 5 O

r Chapter 3 S-rnoking-Attributable Disease Impact Estimation:_ Conceptua=l Iss-ues The Scope of tbe- Calculations Tbe disease impact calculations dcscribed here include direct health-care costs, indirect mortality costs, and indirect morbidity costs. These thre_e types of costs are routinely u3ed by the National Center for-Health Statistics-to defirle the costs of illness (Hodgson, 1983; Hodgson and Meiners, 1979, 1982). '1'ney do not _represent all economic interactions. Sever~l important questions about compensating economic effects-are presented below. - ."_. _"y (1j -Ane smoi'dng-atzributable dir_ect medical costs a net liabi;liry to society? Warner (1987) notes that calculations-by Rice and colleagues (19S6) focus on the readily measurable, tangible zosts-of smoking. ':bese calculations regard smoking-attributable direct medical costs as a liability. Warner notes-that, from a purely economic viewpoint, health-care expenditures attnbutable to cigarette-smoking also represent tobacco-related jobs-created and incomes paid to workers. Another perspective is that the provision of medical eare services for coping witlr the consequences of cigarette smoking requires that society forego alternative goods and-services (an argumcnt used by Berry and Boland, 1977, against alcohol abuse). Rice and colleagues (1986) state: On average, current and -former smokers use more medical care ... than persons who have never smoked. Although a smoker may suffer from smoking-induced illness and require medical care, the cost of 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 from both other smokers and nonsmokers, or by public expenditures such as Medicare and Medicaid. [2] Do smokers generate larger lifetimt medical cosir than do n6nsmokes? Lewit (1983) suggests that movement toward a tobacco-free society may reduce heaJth-care costs. Conversely, a Swiss study indicated that higher annual medical costs for smokers were offset by the longer lifetime of medical utilization by nonsmokers (Lzu and Schaub, 1983, 1985): Lifetime medical costs per individual were_found to be-equal for smokers and nonsmokers. Wareer (1987) discusses the offsetting effects of longer life expectancy for nonsmokers, which leads to a delay, but not necessarily a reduction, in lifetime health-care expenditures:- ..t it is clear that, in financial terms, all smking-related health-care xats do not represent a pure social burden. At least some of these costs would be offset by the later health-care costs associated with a tobacco-free society, even xi CA OD .~ CD ipa