Estimation of Treatment Effects in Cluster-Randomized Trials by Calibrating Covariate Imbalances Between Clusters

• Activity Number: 595
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Zhenke Wu*+ and Constantine E. Frangakis and Thomas A. Louis and Daniel Scharfstein
• Companies: Johns Hopkins University and Department of Biostatistics, Johns Hopkins University and Johns Hopkins Bloomberg School of Public Health and Johns Hopkins Bloomberg School of Public Health
• Keywords: Covariate-calibrated estimation ; Bias correction ; Guided Care Nurse program ; Meta-analysis ; Paired cluster randomized design ; Rubin causal model

Abstract

We address estimation of intervention effects in experimental designs in which (a) interventions are to be assigned in clusters, because individual-level randomization is not feasible due to practical constraints; and (b) clusters are often selected to form pairs, matched on observed characteristics, and treatments are to be assigned within pairs. In such designs, inference that ignores covariates can be imprecise because cluster-level assignment can still leave substantial imbalance in the covariate distribution between experimental arms. However, most existing methods that model covariates have estimands that are not of policy interest. We propose methodology that explicitly balances the observed covariates among clusters in a pair, and retains the original estimand of interest. We demonstrate our approach through the evaluation of a recent Guided Care Nurse program.