Elsevier

Social Science & Medicine

Volume 48, Issue 3, February 1999, Pages 375-391
Social Science & Medicine

Smoking-attributable medical care costs in the USA

https://doi.org/10.1016/S0277-9536(98)00344-XGet rights and content

Abstract

Medical care costs attributable to cigarette smoking are estimated using an econometric model of annual individual expenditures for four types of medical services: ambulatory, hospital, prescription drug, and other (which includes home health and durable medical equipment and excludes dental and mental health). The model follows the two-part specification of Duan et al. (1983). Estimation is carried out using the 1987 National Medical Expenditure Survey. Fitted values are used to calculate smoking-attributable fractions (SAFs) of expense by type of service and by age and gender category. The overall weighted average SAF is 6.54%. SAFs are generally largest for ambulatory and smallest for hospital expenses. They are larger for males and for the older age categories. The model is analyzed for heteroscedasticity and goodness of fit. Additional analysis using the National Health Interview Survey is conducted to test for the possible effect of not being able to include alcohol consumption in the primary model. A balanced repeated replication analysis is conducted to evaluate the variance of the SAFs. Variances are found to be acceptably small. An extension of the model to support evaluation of smoking-attributable costs for special populations such as individual states, and special insurance pools such as Medicaid recipients, is described. Results for the fifty states are presented. Conclusions and subjects for further research are discussed.

Introduction

Policy initiatives intended to mitigate the public health and public finance consequences of widespread tobacco consumption require reliable, accurate assessments of the portion of aggregate medical care expenditures that is attributable to past and present smoking behavior. These initiatives include programs to reduce the incidence of smoking among teenagers and efforts to promote and assist smoking cessation, which must be motivated by a clear cost/benefit advantage. They also include legal efforts to recoup costs attributable to smoking from cigarette manufacturers, which require a fair assessment of damages.

This paper presents an assessment of smoking-attributable costs that differs from prior efforts in that it relies on econometric statistical methods applied to person-level cost data rather than on epidemiological formulas applied to aggregate data. This approach avoids three problems inherent in prior approaches. First, it permits more complete control for the effect of factors other than smoking on health care costs. Second, it avoids assumptions about the comparability of expenditures across smoking categories for condition-specific mortality or utilization. Third, it permits inclusion of costs that are due to the general increase in morbidity cited in the US Surgeon General's Report on Smoking and Health (US Department of Health and Human Services, 1989). Thus, the present approach should lead to more reliable estimates.

The general conceptual framework is as follows. There are problems with defining, and therefore quantifying, `health', the ultimate output of health care. These problems preclude estimation of standard neoclassical production and cost functions. Thus, the marginal effect of prices or quantities of inputs on `health', or its cost, cannot be evaluated in the usual way. However, if optimizing-agent behavior can be assumed, then observed annual individual health care expenditures may be regarded as minimum average costs. This allows estimation of average individual costs in terms of inputs, or attributes of individuals that affect costs, such as age, gender, cigarette consumption history, and other factors, without recourse to the amount of `health' purchased.

This average cost framework underlies all prior efforts to assess smoking-attributable costs. These efforts can be divided into two groups, the `prevalence' approaches and the `incidence' approaches. Prevalence approaches analyze current experienced costs into components, such as a smoking-attributable component, while incidence approaches attempt to estimate the expected lifetime medical care costs attributable to smoking.

Among the prevalence approaches, perhaps the first published attempt to estimate the total direct medical care costs attributable to smoking is that of Luce and Schweitzer (1978). They simply multiplied the costs of treating smoking-related diseases as detailed in Cooper and Rice (1976) by the smoking-attributable fraction of the cases of these diseases developed by Boden (1976) to estimate that, in 1976, smoking was responsible for 7.8% of total national direct health care costs. This approach was improved upon by the United States Congress, Office of Technology Assessment (OTA, 1985) and by Rice et al. (1986). Using an attributable risk formula with age/gender-specific mortality ratios by disease category, OTA estimated that smoking accounted for about 6% of national direct health care expenditures in 1985. Using medical care use data rather than mortality data, Rice and Hodgson classified individuals into several categories, including categories based on smoking behavior. Calculated attributable risk figures were multiplied by the estimated total expenditures for that subset of the population to produce smoking-attributable expenditures. These were then summed over demographic and utilization categories and divided by the corresponding totals to produce a relatively small number of `smoking-attributable fractions' of expenditure, or SAFs, that may be applied to populations or expenditure figures other than those upon which they were derived. They estimated the overall SAF for the United States to be 6.8% in 1980. This attributable risk approach has been incorporated into a software package called SAMMEC (for smoking-attributable mortality, morbidity and economic costs; see Schultz et al., 1991) that has been distributed by the Centers for Disease Control (CDC) of the United States Public Health Service.

To resolve problems with the epidemiological aggregate approaches, an econometrics-oriented effort was initiated in 1991, sponsored first by the CDC and later by the Robert Wood Johnson Foundation (RWJF). A summary interim report from this effort appears in Morbidity and Mortality Weekly Report (Bartlett et al., 1994), where the aggregate SAF was reported to be 7.1%. An initial extension of this early model, which intended to estimate state-level SAFs for Medicaid populations, appeared in Public Health Reports (Miller et al., 1998), where the aggregate SAF was reported to be 14.36%. It is not clear why these figures are so different, especially since smoking is less prevalent among Medicaid recipients than among the general population. The present work represents a refinement and improvement on these earlier reports, as well as the first complete technical documentation of the entire line of research.

The incidence approach was taken by Oster et al. (1984), Leu and Schaub (1985), Manning et al., 1989, Manning et al., 1991 and Hodgson (1992), among others. This approach can be expected to lead to lower estimates of direct medical care costs than does the attributable risk approach, since it accounts for the fact that, on average, smokers do not live as long as do nonsmokers. So, while smokers may consume more services while they are alive, they don't consume these services for as many years. The primary difficulty of the incidence approach is that appropriate data is harder to obtain. The time lag between smoking and its consequences, in terms of morbidity and medical expense, may be twenty years or more. Therefore, a proper incidence study must monitor the medical expenditures of smokers and nonsmokers over this sort of time frame. As a result, considerable disparity exists in the results of the few studies that have attempted to implement it.

The following sections describe, respectively, the model specification, the data used, the basic results, several additional analyses that were conducted to evaluate the validity of the model, an extension of the model to compute SAFs for specific populations, and conclusions and subjects for further research.

Section snippets

Specification

This section describes a `reduced form' model used to assess smoking-attributable medical care costs. The term `reduced form' is enclosed in quotes because it is not, in the strictest technical sense, the reduced form of the accompanying structural form. However, the general relationship between the two is similar to the textbook relationship: the structural form is a system of equations, each of which models a hypothesized causal relationship between values of variables, while the reduced form

Data

The 1987 National Medical Expenditure Survey (NMES) was conducted by the Agency for Health Care Policy and Research in the United States. The Household Survey Component of the NMES consisted of a 4-round series of face-to-face and telephone interviews over the course of the year involving 38,446 participants in roughly 14,000 households drawn from the civilian, noninstitutionalized population. The survey participants were asked for detailed information about their medical care expenditures in

Results

Table 1 gives descriptive statistics for the variables included in the model, first for smokers then for nonsmokers. Table 2, Table 3 give the estimates and the test statistics for each parameter in each of the eight expenditure equations. For the smoking coefficients, 39 of the 48 are positive; 19 of these are significant at the 90% level or higher. Of the nine negative coefficients, only two are significant.

Former smokers tend to have higher expenditures than do current smokers. This may be

Sample selection bias

Potential problems with sample selection bias were addressed in the specification section above. While a sample selection bias equation was not deemed necessary, a Heckman–Lee specification was nevertheless estimated as a check on the sensitivity of the results to this issue. While the inverse Mills ratio was significant in five out of eight equations, the SAFs were essentially the same as those generated by the simpler specification.

Heteroscedasticity

In the calculation of the SAFs, there is an implicit

An extension to simulate SAFs for specific populations

The NMES contains data (with appropriate population weights) on 34,459 individuals nationwide. This sample size is not large enough to allow subsetting of the data by state without compromising confidentiality. Therefore, estimation of smoking-attributable medical expenditures for a given state requires integration of the NMES with additional data sets containing information at the state level. One primary data set (Tobacco Use Supplement (TUS) of the Current Population Survey (CPS)), two

Conclusions and subjects for further research

The first conclusion of this paper is that cigarette smoking affects medical expenditures to a significant and substantial degree. Overall smoking-attributable fractions in the 4 to 7% range translate currently into over fifty billion dollars per year in medical care expenditures in the United States. Much of these expenditures are paid through the public sector, especially through Medicare and Medicaid. Thus, individuals who smoke impose a substantial externality cost on the public at large.

References (32)

  • W.E. Griffiths et al.

    A Monte Carlo evaluation of the power of some tests for heteroscedasticity

    Journal of Econometrics

    (1986)
  • R.E. Leu et al.

    More on the impact of smoking on medical care expenditures

    Social Science and Medicine

    (1985)
  • J.C. Bartlett et al.

    Medical-care expenditures attributable to cigarette smoking, United States, 1993

    Morbidity and Mortality Weekly Report

    (1994)
  • L.I. Boden

    The economic impact of environmental disease on health care delivery

    Journal of Occupational Medicine

    (1976)
  • T.S. Breusch et al.

    A simple test for heteroscedasticity and random coefficient variation

    Econometrica

    (1979)
  • Chow, G., 1983. Econometrics. McGraw Hill, New...
  • B.S. Cooper et al.

    The economic cost of illness revisited

    Social Security Bulletin

    (1976)
  • N. Duan

    Smearing estimate: a nonparametric retransformation method

    Journal of the American Statistical Association

    (1983)
  • N. Duan et al.

    A comparison of alternative models for the demand for medical care

    Journal of Business and Economic Statistics

    (1983)
  • W.H. DuMouchel et al.

    Using sample survey weights in multiple regression analyses of stratified samples

    Journal of the American Statistical Association

    (1983)
  • Efron, B., Tibshirani, R., 1982. An Introduction to the Bootstrap. Chapman Hall, New...
  • H. Glejser

    A new test for heteroscedasticity

    Journal of the American Statistical Association

    (1969)
  • S.M. Goldfield et al.

    Some tests for homoscedasticity

    Journal of the American Statistical Association

    (1965)
  • Greene, W.H., 1993. Economic Analysis, 2nd ed. MacMillan, New...
  • T.A. Hodgson

    Cigarette smoking and lifetime medical expenditures

    The Milbank Quarterly

    (1992)
  • J.G. Kovar et al.

    Bootstrap and other methods to measure errors in survey estimates

    The Canadian Journal of Statistics

    (1988)
  • Cited by (99)

    • U.S. healthcare spending attributable to cigarette smoking in 2014

      2021, Preventive Medicine
      Citation Excerpt :

      Medicaid and Medicare paid for approximately half of the total smoking-attributable healthcare spending in 2010 (Xu et al., 2015). Existing studies show that cigarette smoking-attributable healthcare spending has historically accounted for 5–14% of U.S. annual healthcare expenditure through 2010 (U.S. Department of Health and Human Services, 2014; Xu et al., 2015; Congressional Budget Office, 2012; Miller et al., 1999; Centers for Disease Control and Prevention, 2002. However, less is known about changes in such spending by payers in recent years, especially following changes to health insurance coverage since 2010 (Cohen and Martinez, 2014), and a period of Medicaid expansion beginning in January 2014 and early adoption of expanded Medicaid in a few states between 2010 and 2013 (Cohen and Martinez, 2014; Black and Cohen, 2015; Frean et al., 2017; Wherry and Miller, 2016a).

    • Cost of Tobacco-related Cancer Hospitalizations in the U.S., 2014

      2018, American Journal of Preventive Medicine
    • Tobacco dependence treatment in the emergency department: A randomized trial using the Multiphase Optimization Strategy

      2018, Contemporary Clinical Trials
      Citation Excerpt :

      For QALY loss, the most recent estimates [51–53] are better than older estimates [54,55]. For medical costs, we expect to use the latest update to CDC's SAMMEC model [1] which builds from the methods pioneered by V. Miller [56]. We will inflate all cost savings to the same year's dollars as the program costs and compute present value of savings in future years at a 3% discount rate.

    • Smoking-attributable medical expenditures by age, sex, and smoking status estimated using a relative risk approach

      2015, Preventive Medicine
      Citation Excerpt :

      Following the latest evidence on smoking-attributable fractions in the U.S. healthcare spending (Solberg et al., 2006; Xu et al., 2015), we use 8.66% of projected 2013 Personal Health Care Expenditures (PHCE) for adults 19 years of age and older less dental care in our example, which is consistent with previous cross-sectional studies that reported smoking-attributable fractions ranging from 6.5% to 14% (Miller et al., 1999; Warner et al., 1999). However, any estimated attributable fractions based on studies published within the last 15 years can be used with this approach (Centers for Disease Control and Prevention, 2002; Congressional Budget Office, 2012; Levy DE, 2011; Miller et al., 1999; Solberg et al., 2006; Warner et al., 1999). We used the 2012 National Health Interview Survey to determine the distribution of the US population by age, sex, and smoking status.

    • Annual healthcare spending attributable to cigarette smoking: An update

      2015, American Journal of Preventive Medicine
      Citation Excerpt :

      Using data from the 2006–2010 MEPS linked to the 2004–2009 NHIS, this analysis reveals that 50 years after the first Surgeon General’s report on smoking and health was released in 1964, cigarette smoking continues to be a major contributor to annual healthcare spending in the U.S. Across all payers, cigarette smoking was associated with 8.7% of annual aggregated healthcare spending. This finding is consistent with previous cross-sectional studies21,22 that reported smoking-attributable fractions ranging from 6.5% to 14%. This conclusion also remains fairly stable across the sensitivity analyses.

    View all citing articles on Scopus
    View full text