Primary Submission Category: Multilevel Causal Inference
Assessing Informative Cluster Size in Cluster-Randomized Trials
Authors: Bryan Blette, Brennan Kahan, Michael Harhay, Fan Li,
Presenting Author: Bryan Blette*
In cluster-randomized trials, the average treatment effect among participants (p-ATE) may be different from the cluster average treatment effect (c-ATE) when informative cluster size is present, i.e., when treatment effects or participant outcomes depend on cluster size. In such scenarios, mixed-effects models and GEEs with an exchangeable correlation structure are biased for both the p-ATE and c-ATE estimands, and GEEs with an independence correlation structure or analyses of cluster-level summaries are recommended in practice. However, when the p-ATE and c-ATE are equivalent, mixed-effects models and GEEs with exchangeable correlation structure can provide unbiased estimation and notable efficiency gains over other methods. Thus, a hypothesis test of whether informative cluster size is present could be useful to formally assess the validity of this key assumption. In this study, we develop and compare model-assisted and randomization-based tests for informative cluster size in cluster-randomized trials. We construct simulation studies to examine the operating characteristics of these tests and contrast them to existing model-based tests for informative cluster size or “confounding by cluster” in the observational study setting. The proposed tests are applied to data from a recent cluster-randomized trial, and practical recommendations for using these tests are discussed.