National estimates of cigarette smoking-attributable morbidity were recently published in the Morbidity and Mortality Weekly Report . Researchers at the Roswell Park Cancer Institute, who led that publication, are now also releasing preliminary state-specific estimates of cigarette smoking-attributable morbidity.

To assess smoking-attributable morbidity, data were analyzed from three data sources: Behavioral Risk Factor Surveillance System (BRFSS), National Health and Nutrition Examination Survey III (NHANES), and the U.S. Census.

Estimates of the prevalence of smoking-related conditions were obtained from the NHANES III survey for 1988-1994 for current, former, and never smokers by demographic groups and state to estimate attributable fractions for smoking related diseases. The smoking-related conditions for which data were collected are those categorized by the Surgeon General as caused by smoking and addressed in NHANES III. Respondents reported whether a "doctor ever told" them if they had any of the following conditions: stroke, heart attack, emphysema, chronic bronchitis, and specific cancer types reported by respondents including lung cancer, bladder cancer, mouth/pharynx cancer, esophageal cancer, cervical cancer, kidney cancer, laryngeal cancer, or pancreatic cancer. Smoking-attributable morbidity estimates were obtained in two ways. For one estimate, each person was considered the unit of analysis, and persons with at least one smoking-related condition were counted as having a smoking-related condition. For the second estimate, the condition was treated as the unit of analysis so persons with multiple conditions were counted more than once. Estimates were derived separately for each condition, and the total of all conditions was summed.

The number of persons with a smoking-attributable morbid condition was estimated by state and demographic subgroups from the following five steps: 1) BRFSS smoking status estimates by demographic group were applied to census data to estimate the number of current, former, and never smokers in each demographic group in each state; 2) NHANES III smoking-related disease frequency data were applied to the population count estimates from the first step to estimate the number of adults with a smoking-related condition; 3) attributable fractions for current and former smokers in each demographic group were multiplied by the number of persons with a smoking-related disease to yield an estimate of the number of persons with a disease that is attributable to smoking (attributable fraction = [disease prevalence rateexposed - disease prevalence rateunexposed] / disease prevalence rateexposed); 4) the numbers obtained from the third step were summed across all demographic categories in each state to yield an estimate of persons with smoking-attributable conditions in each state; and 5) the numbers of smoking-attributable morbid conditions obtained in each state from step four were summed to yield an overall U.S. estimate.

In 2000 in the United States, an estimated 8.6 million persons had an estimated 12.7 million smoking-attributable diseases (Table). State-specific estimates are also presented in the Table. Data on the distribution of disease type overall and for current and former smokers is presented elsewhere1.

The findings indicate that more persons are harmed by tobacco use than is indicated by mortality estimates. For every tobacco-attributable death that occurs, there are approximately 20 people alive who are suffering from a serious, chronic disease that is attributable to cigarette smoking.

The findings in this report are subject to at least four limitations. First, the estimates do not adjust for all potential confounders; however, the impact of confounding on cigarette-attributable mortality was examined in a prospective cohort study of approximately one million persons and the findings indicated that adjustment for multiple factors reduced the smoking-attributable mortality estimate by 2.5% . Second, disease data are self-reported and might not represent the true rate or type of disease. Research shows that for chronic disease like cancer, stroke, hypertension, and lung disease, self-reported rates of disease largely underestimate the true rate of disease , . Therefore, these self-reported data are probably substantial underestimates of a true disease burden. Third, national NHANES III data are used to estimate disease frequency data for each state; states may vary in other characteristics (e.g., rates of early detection of disease or access to health care) that affect smoking-related disease prevalence. Finally, the scope of diseases considered in this report was limited to those diseases for which the Surgeon General has implicated smoking as a cause and for which survey data were available; therefore, the estimates presented are conservative.

Researchers at the Roswell Park Cancer Institute are currently working with CDC researchers to further refine these state-specific estimates and plan to publish these data in the future. In the meantime, the state-specific estimates presented in this report are useful for those in states to more fully characterize the disease and cost burden tobacco places on each state. This information can be used to justify continued and additional support for proven tobacco control cessation and prevention efforts including increasing the cost of cigarettes, increasing clean indoor air regulations, and implementing comprehensive tobacco control programs.

Acknowledgement: Funding for this project was provided by the Roswell Park Cancer Institute NCI-funded Cancer Center Support Grant, CA16056-26. We are also grateful for the past contributions to this work by Chris Vena, Roswell Park Cancer Institute and Paul Mowery, MS, Research Triangle Institute.