Abstract

Often motivated by concerns about contamination of control subjects, stepped wedge cluster-randomized trials assign interventions to distinct clusters (e.g., hospitals) and protect against confounding by randomizing the timing of intervention delivery. However, when trial units are embedded in professional networks that span clusters, such trial designs are vulnerable to spillover from intervention subjects to control subjects. We first develop potential outcomes based definitions of the causal average direct effect and the causal average indirect effect of the intervention. We then use a longitudinal cluster-randomized trial and contemporaneous physician professional networks for a national provider organization in the United States to model and estimate the direct, indirect and direct-indirect interaction effects of an Advanced Care Planning (ACP) intervention. We show that the direct effect measures the intervention’s impact in a counterfactual world absent contamination, the indirect effect quantifies the spillover effect from intervention to control subjects, and the interaction term assesses whether spillover modifies the intervention’s effect, a phenomenon known as contamination. In addition, we exploit the staggered nature of the stepped-wedge cluster-randomized trial design to test whether intervened-on peers impact the sustainability of the intervention over follow-up and whether spillover reinforces or attenuates the intervention. In our empirical analyses, we find that without adjusting for network structure the intervention showed no significant effect on ACP billing. However, in the network adjusted analyses we detected a large spillover effect with strong evidence of contamination prior to a subject being intervened on and a significant intervention reinforcement effect post intervention. These findings suggest that the diffusion of intervention behaviors through the physician network were primarily driven by a large physician spillover effect that modified the direct effect at the time of intervention and subsequently over follow-up.