Alternating Logistic Regressions for Cluster Trials with Binary Outcomes

• Activity Number: 44
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): John S. Preisser*+ and Jamie Perin and Beth Reboussin
• Companies: The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill and Wake Forest University School of Medicine
• Keywords: bias ; generalized estimating equations ; odds ratio ; sample size ; sandwich estimators

Abstract

Potentially, alternating logistic regressions, a procedure for modeling within-cluster association among binary outcomes, provides an integrated design and analysis approach for cluster trials. In particular, it provides odds ratio estimates of association for within-cluster outcome pairs that can be used in sample size calculations for planning such studies. The procedure is especially useful when cluster sizes are large, a common feature of cluster trials. However, application of the method is limited because it requires a large number of clusters, and few cluster trials have this characteristic. Simulation study results suggest that bias corrections to the association parameter estimating equations and covariance estimators justify their use in scenarios with a smaller number of clusters. The proposed methods are applied to a national community trial to reduce underage drinking.