Correspondence
The Health Care Costs of Smoking
N Engl J Med 1998; 338:470-472February 12, 1998
- Article
To the Editor:
The conclusion by Barendregt et al. (Oct. 9 issue)1 that smoking cessation would eventually lead to increased health care costs is questionable because the authors underestimate differences in health care use between smokers and nonsmokers. They attribute to smoking the leading smoking-related diseases, including heart disease, stroke, lung and some other cancers, and chronic obstructive pulmonary disease. According to the Surgeon General's reports, smokers are also at higher risk for a variety of other diseases, including pneumonia and influenza, abdominal aortic aneurysm, gastric and duodenal ulcers, and cataracts. Adding these conditions increases by more than 30 percent the estimate of short-term hospital days for diseases linked to smoking.2
Smokers also have higher utilization rates and expenditures for coexisting diseases not related to smoking. Coexisting disorders increase the resources required to treat many smoking-related conditions; hospital stays are longer, more tests and procedures may be required, and so forth. The authors' report that per capita health care costs are up to 40 percent higher among male smokers and up to 25 percent higher among female smokers, as compared with nonsmokers. However, in the United States, peak differences in costs between smokers and nonsmokers can be more than 100 percent for men and almost 70 percent for women.3 Limiting the calculation of smoking-related costs to the diseases noted in the article omits substantial morbidity and health care utilization and severely underestimates the lifetime medical expenditures associated with smoking.
The authors emphasize the results at lower discount rates that support the conclusion that a nonsmoking population would eventually have higher health care costs. But their sensitivity analysis of the discount rate renders this conclusion ambiguous, since they calculate lower costs for nonsmoking at discount rates of 4.5 percent and higher. Barendregt et al. reject discounting, however, citing differences of opinion on discounting lifetime costs. There is controversy about discounting health effects such as years of life gained, but not about discounting monetary sums. Discounting future monetary values is universally accepted by economists. Furthermore, given the underestimates in the authors' analysis, it is likely that the costs for the nonsmoking population are also lower at the lower discount rates, including the range of 3 to 5 percent generally accepted as most likely to include the social discount rate. Three studies cited by the authors each calculated higher lifetime costs for smokers in the United States. But the authors offer no explanation for the difference between their results and these findings.
3 ReferencesThomas A. Hodgson, Ph.D.
Centers for Disease Control and Prevention, Hyattsville, MD 207821
Barendregt JJ ,Bonneux L ,van der Maas PJ . The health care costs of smoking. N Engl J Med 1997;337:1052-1057
Full Text | Web of Science | Medline2
National Center for Health Statistics, Graves EJ, Gillum BS. Detailed diagnoses and procedures, National Hospital Discharge Survey, 1994. Vital and health statistics. Series 13. No. 127. Washington, D.C.: Government Printing Office, 1997. (DHHS publication no. (PHS) 97-1788.)
3
Hodgson TA . Cigarette smoking and lifetime medical expenditures. Milbank Q 1992;70:81-125
CrossRef | Web of Science | Medline
To the Editor:
Barendregt et al. estimate that the difference in life expectancy from birth between smokers and nonsmokers is 7.3 years for men and 6.0 years for women. These estimates were apparently calculated from rate ratios for five categories of tobacco-related diseases, without consideration of competing risks and with the use of fixed rates for all ages. It is of course common for smokers to have more than one of these diseases, and rate ratios should be expected to decrease at older ages as competing risks increase, as should differences in life expectancy with advanced ages. Although sensitivity ranges are given, no analyses of the effects of assuming the lower bounds are provided.
The article cited as reporting similar mortality differences,1 and much of the other literature, compares male “continuing cigarette smokers” with “lifelong nonsmokers,” without adjustment for confounding risk factors (smokers have more stress, use more alcohol, exercise less, have higher dietary fat intake, and use seat belts less than nonsmokers). Furthermore, an intention-to-treat analysis in 1951 would have included the 50 percent of smokers who subsequently stopped smoking and who had mortality rates similar to those for lifelong nonsmokers. In addition, these differences in mortality rates should narrow at higher ages because of competing risks and in more recent cohorts because of a generally longer life expectancy from birth.
The results reported are extremely sensitive to changes in assumptions about mortality rates. Small reductions in mortality rates can make enormous differences in long-term projections, increasing and perpetuating the initial cost savings.
1 ReferencesJames F. Fries, M.D.
Stanford University School of Medicine, Palo Alto, CA 94304-18081
Doll R ,Peto R ,Wheatley K ,Gray R ,Sutherland I . Mortality in relation to smoking: 40 years' observations on male British doctors. BMJ 1994;309:901-911
CrossRef | Web of Science | Medline
To the Editor:
. . . The study by Barendregt et al. does not account for the possibility that exsmokers, who are presumably included in the population of nonsmokers, have a greater proportion of morbidity and mortality than those who have never regularly smoked. In the United Kingdom, over half of persons over 60 years old previously smoked regularly.1
Sophisticated incidence-based data sets are necessary to establish the true health care costs incurred by smoking. Health care economists accept that prevalence-based cost-of-illness studies give little insight into expected lifetime costs and that incidence-based studies are theoretically superior.2 If the data in this study are recalculated with the usual discount rate for time preference of 5 percent and with the incorporation of additional smoking-related costs, such as the lifetime costs of care after a stroke,3 it is unlikely that the “break-even” year after smoking cessation would ever occur. It is reasonable to assume that “growing old gracefully” without smoking-related morbidity is likely to result in a lesser burden on health care systems.
3 ReferencesDominic Heaney, B.M., B.Ch.
National Hospital for Neurology and Neurosurgery, London WC1N 3BG, United Kingdom1
Office of Population Census Surveys. General household survey 1994. London: Her Majesty's Stationery Office, 1994.
2
Drummond MF, Stoddart GL, Torrance GW. Methods for the economic evaluation of health care programmes. Oxford, England: Oxford University Press, 1987.
3
Taylor TN ,Davis PH ,Torner JC ,Holmes J ,Meyer JW ,Jacobson MF . Lifetime cost of stroke in the United States. Stroke 1996;27:1459-1466
CrossRef | Web of Science | Medline
To the Editor:
Contrary to what Barendregt et al. suggest, any smoking-related disease that generates substantial health care costs in excess of health care cost savings from premature death could change their results, though perhaps not their overall conclusions.
For example, the results shown in Figure 2 of their article could change if the costs of pediatric disease from parental smoking were included. The $4.6 billion per year in U.S. health care costs for such disease is almost three times as much as the $1.7 billion in undiscounted health care cost savings from the deaths (6200 deaths of children [mostly infants] each year from parental smoking,1 times $3,600 per year in per capita health care costs, times 75 years of life lost [U.S. life expectancy]2 ). In addition, the authors' results could change if the health care costs among adult nonsmokers for generally nonfatal asthma exacerbations from secondhand smoke and burn and blast injuries from others' careless smoking 3 exceeded the health care cost savings from premature death. The authors' results could also change if the health care costs for expensive, nonfatal smoking-related injuries (e.g., back strains and orthopedic injuries)3,4 were included. Such injuries may account for a substantial portion of the 12 percent of U.S. health care costs that are due to injuries.2
The total health care costs associated with smoking include the costs of disease in children due to parental smoking, nonfatal smoking-related injuries and cataracts in smokers, and the often nonfatal secondhand effects of others' smoking in nonsmoking adults. These factors, especially in combination with the use of a discount rate, reduce the likelihood that excess health care costs due to smoking cessation will be seen in our lifetimes, or ever.
4 ReferencesBruce N. Leistikow, M.D.
University of California, Davis, CA 95616Ted R. Miller, Ph.D.
National Public Services Research Institute, Landover, MD 207851
Aligne CA ,Stoddard JJ . Tobacco and children: an economic evaluation of the medical effects of parental smoking. Arch Pediatr Adolesc Med 1997;151:648-653
CrossRef | Web of Science | Medline2
Fingerhut LA, Warner M. Injury chartbook, health, United States, 1996–1997. Hyattsville, Md.: National Center for Health Statistics, 1997.
3
Leistikow BN . Injury prevention. N Engl J Med 1998;338:133-133
Full Text | Web of Science | Medline4
Kwiatkowski TC ,Hanley EN Jr ,Ramp WK . Cigarette smoking and its orthopedic consequences. Am J Orthop 1996;25:590-597
Medline
To the Editor:
The calculations of the health care costs of smoking reported by Barendregt et al. may be interesting, but their conclusion that “eventually, smoking cessation would lead to increased health care costs” is misleading and dangerous. The authors infer that continued smoking would reduce the health care costs in the long run and nonsmokers are therefore, in the long run, health care–cost producers. The authors, however, do not take into consideration the years lost through smoking-related morbidity and mortality. The productivity — to stay strictly within the context of the authors' economic terms — of a human being during these lost years certainly outweighs by far the health care costs associated with later years in life. . . .
Christian Sauter, M.D.
University Hospital, CH-8091 Zurich, SwitzerlandAuthor/Editor Response
The authors reply:
To the Editor: Hodgson correctly points out that there are more smoking-related diseases than the major ones we included, but additional diseases would change our conclusions only if they posed a clear excess risk for smokers, accounted for large costs, and had a low associated risk of mortality. We can think of no such diseases. Four of the five diseases Hodgson mentions have high mortality rates, and we discussed cataracts.
Hodgson finds much higher differences in per capita costs. The cited 100 percent and 70 percent higher costs for male and female smokers, respectively, however, are in just one age group. The differences in the other age groups are more in line with our results, and the differences are even much smaller for the elderly.1 We do not reject discounting. Discounting involves the evaluation of future resources as part of a decision about which of two or more alternative choices is preferred. Hodgson ignores the distinction we make between the estimation of population costs for smokers and nonsmokers and the evaluation of an intervention. In the former case, we reject discounting because when estimating costs, we are not evaluating which of the alternatives we prefer. Moreover, in the stationary-population interpretation of the life table, there is no time — hence, no future. The application of discounting to the life-table results lowers costs incurred at higher ages, not at a future moment in time, and may very well explain the different results of other studies, because nonsmokers are older on average. When evaluating an intervention, we do discount and show that the intervention can be economically attractive because savings precede the eventually higher costs.
We can assure Fries that in our analysis competing risks are taken into account and that the differences in mortality between smokers and nonsmokers do narrow at a higher age: the age-independent rate ratios are based on disease incidence, not on mortality. Because of the nonlinear relation between incidence and mortality, the rate ratios for disease-specific mortality decline with age, and so do the rate ratios for total mortality: for men, from 2.5 at 65 years to 1.6 at 90 years, and for women, from 2.2 to 1.6; with the lower bounds, the rate ratios are 1.9 to 1.4 for men and 1.7 to 1.4 for women, with a difference in life expectancy of 5.2 and 4.1 years, respectively.
Heaney is correct that a large proportion of current nonsmokers are in fact former smokers. The problem plagues epidemiologic studies; rate ratios will be biased when not measured for smokers as compared with persons who have never smoked. We ignored this problem on the grounds that the bias will be well within the sensitivity range of the rate ratios.
Whether incidence-based studies are superior depends on the purpose of the study. When estimating population costs for smokers and nonsmokers, population-based prevalence data are perfectly adequate.
Leistikow and Miller rightly mention the costs due to smoking among nonsmokers — in particular, children. However, including these costs would not change our conclusions. The total costs for lung diseases among persons from infancy to 19 years of age, for example, account for less than 1 percent of all costs among men and just over 0.5 percent of costs among women.
Sauter points to the costs of lost productivity. Like some other financial differences between smokers and nonsmokers, such as pensions and tobacco taxes, these are not health care costs and are therefore outside the scope of our article.
1 ReferencesJan J. Barendregt, M.A.
Luc Bonneux, M.D.
Paul J. van der Maas, Ph.D.
Erasmus University, 3000 DR Rotterdam, the Netherlands1
Hodgson TA . Cigarette smoking and lifetime medical expenditures. Milbank Q 1992;70:81-125
CrossRef | Web of Science | Medline
- Citing Articles (1)
Citing Articles
1
Sajjad Ahmad, Gregor A. Franz. (2008) Raising taxes to reduce smoking prevalence in the US: A simulation of the anticipated health and economic impacts. Public Health 122:1, 3-10
CrossRef
