Cluster-randomized trials are essential for evaluating prevention, health care delivery, and policy interventions in real-world settings, but their analyses often rely on working models for which assumptions are difficult to verify. When these models are misspecified, standard covariate-adjusted analyses can lose efficiency, blur the target estimand, or introduce bias, especially when cluster sizes vary after randomization. This presentation will introduce a model-robust framework for estimating cluster-average and individual-average treatment effects in cluster-randomized trials. The presenter will show how familiar tools, including generalized estimating equations and mixed models, can be adapted using weighted g-computation to improve interpretability and robustness. There will then be a discussion of efficient estimators that allow flexible covariate adjustment, including machine learning, while maintaining valid inference and addressing informative cluster-size variation. Simulations and real trial examples will illustrate how these methods can strengthen the reliability, precision, and transparency of evidence from prevention research.

Dr. Bingkai Wang is an Assistant Professor of Biostatistics at the University of Michigan School of Public Health. He specializes in randomized clinical and cluster-randomized trials and causal inference, with a focus on covariate adjustment, repeated-measures analysis, and pragmatic trial design. His research develops rigorous and practical statistical methods to improve the efficiency, robustness, and interpretability of clinical and translational studies. Dr. Wang’s work leverages modern analytical tools, including machine learning and large language models, to address key challenges in trial design and analysis, with an emphasis on methods that are both theoretically sound and implementable in real-world settings.