Purpose: Local health statistics are increasingly requested for policy-making and programmatic purposes; however, population-based surveys are often inadequate to support direct estimation for small areas. Model-based estimation techniques can be used to create local estimates for public health outcomes. Using the 2014-2015 South Carolina (SC) Adult Tobacco Survey, we examined tobacco-related outcomes at the county level using a spatial multilevel, poststratification approach.
Methods: To create county-level tobacco estimates, we used a two-level model with a spatially intrinsic conditional autoregressive random intercept. Stratum-specific (race, age, and sex) estimates for each county were then created and averaged based on population data obtained from the U.S. Census.
Results: The estimated prevalence of current smoking in SC counties among adults ranged from 7.4% to 35.1%, and the percentage reporting ever trying an e-cigarette ranged from 4.2% to 30.2%. Model validation showed considerable agreement between direct and indirect estimates (Pearson and Spearman correlations all >0.75) that varied by the sample size of the outcome, as hypothesized.
Conclusions: Data from the SC Adult Tobacco Survey were used to develop county-level estimates of multiple tobacco-related outcomes using a spatial multilevel, poststratification approach. The results showed heterogeneity in smoking behaviors across the state along with marked spatial correlation.
Keywords: Adult; Small area estimation; Smoking; Surveys and Questionnaires; Tobacco.
Copyright © 2018 Elsevier Inc. All rights reserved.