Using Repeated Measures Techniques To Analyze Cluster-correlated Survey Responses

• Activity Number: 73
• Type: Contributed
• Date/Time: Sunday, August 7, 2005 : 4:00 PM to 5:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Celia Eicheldinger*+ and G. Gordon Brown and James Chromy
• Companies: RTI International and RTI International and RTI International
• Keywords: Cluster Correlated Responses ; CAHPS ; Repeated Measures

Abstract

The researchers for the Consumer Assessment of Health Plans Study (CAHPS) of Medicare Fee-for-Service and Disenrollee/Assessment surveys have used a repeated measures approach in SUDAAN to create individual-level composite scores on a number of constructs (e.g., the CAHPS composite "Getting Needed Health Care" combines four survey questions that ask a beneficiary about experiences with getting needed care). This approach, applicable to any survey analysis with cluster-correlated responses, estimates models that assume each question in the composite is a repeated measure. Using our method allows researchers to fit models and conduct complex hypothesis tests that account for covariates, survey design, and intracluster correlation present in most surveys as well as the inherit clustering in the composites. We will present the analysis using the CAHPS composites as an example followed by an explanation of how this technique can be used in other cluster correlated analyses. Our goals are to disseminate this useful technique to researchers using the CAHPS methodology and generalize the technique to other cluster-correlated survey responses.