Background: Various methods are used to estimate smoking-attributable mortality, but these methods often produce differing results. Most do not provide estimates for specific age groups or directly control for confounding factors.
Methods: US smoking-attributable mortality by age is estimated here based on relative mortality hazards for smokers and observed death rates for never-smokers. Hazard ratios for smokers are estimated through survival analysis of data from adults in the 1997-2004 National Health Interview Survey, followed for mortality through the end of 2006. A variety of possible confounding risk factors are controlled for in the analysis.
Results: There were an estimated 291,000 smoking-attributable deaths among men and 229,000 such deaths among women annually in the United States from 2002 through 2006. These figures are generally consistent with, but somewhat higher than, published estimates from the Centers for Disease Control and Prevention, particularly for women.
Conclusion: This method provides reasonable estimates of total and age-specific smoking-attributable mortality in the United States that include controls for potential confounding factors.