The Behavioral Risk Factor Surveillance System (BRFSS) is designed to produce precise estimates for each state, but not for smaller areas such as counties. This paper summarizes our research on the small area estimation techniques to estimate the proportion of individuals having a severe work disability in all 3,116 U.S. counties from the BRFSS. To include multiple sources of variation, we apply the multi-level regression method to use data at both individual-level and county-level. A hierarchical model and a generalized linear mixed model are applied. We discuss several inference methods for both models in our analysis and provide estimates with corresponding estimated confidence intervals. Furthermore, cross-validation and several other validation methods are used to assess the quality of estimation. The analysis reveals that the multi-level regression method provide valid and reliable estimates. Because work disability is associated with some county-level environmental factors (i.e., unemployment rate) as well as respondent's demographic characteristics, our estimates are better than the commonly used synthetic estimates. We also introduce a Monte Carlo method to assess the power of the estimator, or the ability to detect differences when actual changes occurred. The results indicate that the power of the regression estimator is higher than the power of the synthetic estimator, and under certain conditions, is not lower than the power of the direct survey estimator. In conclusion, this regression-based procedure is computationally simple, fast, and provides a significant improvement over the commonly used synthetic estimates state health departments and CDC.
Learning Objectives: 1. summarize small area estimation techniques to obtain county-level data from the Behavioral Risk Factor Surveillance System 2. estimate county-level severe work disability rates from the BRFSS 3. validate the results
Keywords: Data/Surveillance, Disability
Presenting author's disclosure statement:
Organization/institution whose products or services will be discussed: None
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.