ABSTRACT
A potential solution is to use cluster trials, where the allocation to intervention is performed at the group level. The authors can designate clusters that would naturally occur together, for example, school classes, patients under the care of a particular medical practice or a specific therapist, or perhaps geographic location. It avoids the practical difficulties of randomizing to intervention within established natural clusters, e.g., if children within a classroom are in different intervention arms, or a therapist is required to keep track of which cases are allocated to one intervention or another. In addition, if it is necessary to train administering the intervention, clustering means that fewer individuals need to be trained overall, as those working in each cluster would require training in just one method. The clustered approach minimizes the risk of contamination, as intervention clusters are less likely to have direct exposure to each other, and practitioners will consistently do the same intervention with everyone in the study.
