Assessment of Small-Area Estimates from a Complex Survey Cancer Surveillance Project

• Activity Number: 210
• Type: Contributed
• Date/Time: Monday, August 7, 2006 : 2:00 PM to 3:50 PM
• Sponsor: Section on Health Policy Statistics
• Author(s): Van Parsons*+ and Nathaniel Schenker and Trivellore E. Raghunathan and Dawei Xie and William Davis
• Companies: National Center for Health Statistics and National Center for Health Statistics and University of Michigan and University of Pennsylvania and National Cancer Institute
• Keywords: Bayesian ; NHIS ; BRFSS ; cancer risk factors

Abstract

Cancer surveillance research requires that estimators of prevalence for risk factors over small areas be accurate and precise. Many of these prevalences can be estimated from the National Health Interview Survey (NHIS) and from the Behavioral Risk Factors Surveillance System (BRFSS). While the NHIS is a nationally representative face-to-face survey, and the BRFSS is a system of RDD state surveys, direct estimation for small areas, e.g., counties, becomes problematic for each survey because of small or no samples over the small areas and coverage biases. We have developed a Bayesian hierarchical model on the combined NHIS+BRFSS surveys to produce county level estimates. To assess the model-based accuracy and precision, we evaluate both the model-based estimates and direct survey estimates, compare the results, and describe the merits and drawbacks of the model- based approach.