The 13th World Conference on Tobacco OR Health

Objective: In order to highlight differential smoking patterns, the Massachusetts Tobacco Control Program sought to develop smoking prevalence estimates for every town and city within the Commonwealth. By creating accurate prevalence estimates, the Commonwealth and its partners would be able to better target scarce resources and thus assist more smokers to make successful quit attempts.

Methods: A hybrid data set was developed with information about individual smokers and non-smokers. A second data set was prepared with town level data. There was a one to one mapping between the two sets. Using the hybrid data, a logistic equation was computed to estimate the likelihood that an individual would be a current smoker. This same equation was applied to the town level data. Since there was a one to one mapping of data elements, the resulting predictions can be interpreted as smoking prevalence estimates for all 351 towns in the Commonwealth.

Results: To assess the accuracy of the town level estimates, comparisons were made to those obtained from the Behavioral Risk Factor Surveillance System (BRFSS). By aggregating five years of BRFSS surveys, 81 towns were determined to have reliable smoking prevalence estimates based on the BRFSS alone. When town level logistic estimates were compared to those from the BRFSS, 77 of the 81 predictions fell within the 95% confidence intervals. Given the accuracy of the estimates, partnerships are now being explored with major insurance groups that will produce a more efficient targeting of smokers and their physicians in high smoking areas.