Harvard School of Public Health - Department of Biostatistics
Peer influence and social interactions can give rise to spillover effects in which characteristics of one individual may affect outcomes of other individuals. This can occur even if the intervention under study occurs at the group or cluster level as in group-randomized trials. The issue arises because in many such studies, questions of mediation are of interest and the mediator may be at a lower level than the treatment. For example, in many school-based settings, often the treatment is at the school level, the mediator is at the classroom level and the outcome is at the individual level. If a mediator, say class quality, is assessed at the classroom level then it is possible that the class quality of one classroom may affect outcomes of children in another classroom because children interact not simply in their classrooms but also in the hallways and on the playground. Failure to account for such spillover effects of one classroom on the outcomes of children in other classrooms can result in bias and problems with interpretation. Using a counterfactual conceptualization of direct, indirect and spillover effects, a framework is provided that can accommodate issues of mediation and spillover effects in group randomized trials. We show that the total effect can be decomposed into what is defined as a natural direct effect, a within-classroom mediated effect and a spillover mediated effect. We give identification conditions for each of the causal effects of interest and we provide theoretical results on the consequences of ignoring "interference" or "spillover effects" when they are in fact present. The methodology is applied to a group-randomized trial of the 4R's educational intervention with classroom quality as the mediator and child depressive symptoms as the outcome.