Consideration of Challenges to Longitudinal Sample Weights in a Community-Based Epidemiological Study

• Activity Number: 688
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Melissa Danielson* and Susanna Visser and Joseph Holbrook and Robert McKeown and Marc Elliott
• Companies: CDC and CDC and CDC and University of South Carolina and RAND Corporation
• Keywords: longitudinal study ; sample weights

Abstract

Longitudinal community-based epidemiological studies provide an important opportunity to develop prevalence estimates and monitor changes over time; however, they present methodological challenges regarding appropriate sample weights for population estimates. One such example is the Project to Learn about ADHD in Youth, which followed children with and without ADHD in 6 school districts in 2 states. Due to differences in school district size, initial sampling fractions, follow-up duration and attrition, the development of conceptually appropriate longitudinal follow-up weights has been challenging. Specifically, the larger districts experienced more attrition and shorter total follow-up, resulting in high variability in weights and districts with fewer observations having a large influence on estimates. In order to address these concerns, we propose to describe a precision weighting strategy for these data and compare outcome results to those obtained from a traditional weighting approach. Our goal is the presentation of weighted outcome estimates that reflect the underlying school populations with improved precision provided by the subpopulations who had higher response rates.