Cluster Randomized Trials

Abstract

In a randomized clinical or field trial, when randomization units are comprised of groups of individuals, many aspects of design and analysis differ greatly from those of an individually randomized trial. In this chapter, we highlight those features which differ the most, explaining the nature of the differences and delineating approaches to accommodate them. The focus is on design, as many readers will be familiar with the correlated data analysis techniques that are appropriate for many (although not all) cluster randomized trials (CRTs). Thus, the chapter begins by covering motivations for using a CRT design, basic correlation parameters, the variety of potential estimands, delineation and randomization of clusters, and sample size calculation. This is followed by sections on the analysis and reporting of results, which highlight ways to handle the multilevel nature of the data. Finally, ethical and monitoring considerations unique to CRTs are discussed.