One challenge that may occur in cluster-randomized designs is that although a cluster’s units are “traditionally” assumed to be of equivalent size, this may not be true of cluster trials performed in healthcare settings. Such clusters (for example, physician practices or clinics) may be of substantially different sizes, which can affect the statistical power of the study and decisions about sample size (Cook et al. 2016). In order to address these issues, study statisticians need to have an estimate of the range in potential sample sizes. For example: in a trial that randomizes clinics to some strategy that will be applied to newly diagnosed diabetic patients, the statistician will need information on the number of such patients coming into each clinic over the past several months. The range in sample sizes will then be considered in calculating the number of clinics and patients needed for the study.
Cook AJ, Delong E, Murray DM, Vollmer WM, Heagerty PJ. 2016. Statistical lessons learned for designing cluster randomized pragmatic clinical trials from the NIH Health Care Systems Collaboratory Biostatistics and Design Core. Clin Trials. 13:504-512. doi:10.1177/1740774516646578. PMID: 27179253.