Estimating smoking-attributable mortality in the United States: geographic variation

Andrew Fenelon, University of Pennsylvania
Sam Preston, University of Pennsylvania

Tobacco is the largest single cause of premature death in the United States. While evidence linking cigarette smoking to mortality is abundant, researchers have also developed methods for estimating the total number of deaths attributable to smoking in a population. We employ an indirect method that uses the statistical relationship between lung cancer and other causes of death across populations to estimate the impact of smoking on mortality. We apply our method to the United States and the nine Census divisions between 1990 and 2004. In 1990, smoking accounted for 29.6 percent of deaths among men aged 50+ compared with 14.3 among women. By 2004, these were 22.1 and 19.1 respectively. Across divisions, attributable fractions ranged from 11.6 percent in the West North Central to 36 percent in the East South Central. Our method has significant advantages in simplicity and robustness while producing results consistent with those of other techniques.