Improved Designs for Cluster Randomized Trials.

Studies in which clusters of individuals are randomized to conditions are increasingly common in public health research. However, the designs utilized for such studies are often suboptimal and inefficient. We review strategies to improve the design of cluster randomized trials. We discuss both older but effective design concepts that are underutilized, such as stratification and factorial designs, as well as emergent ideas including fractional factorial designs and cluster randomized crossover studies. We draw examples from the recent literature and provide resources for sample size and power planning. Given the inherent inefficiencies of cluster randomized trials, these design strategies merit wider consideration and can lead to studies that are more cost-effective and potentially more rigorous than traditional approaches.