Understanding Who Benefits the Most from Interventions: Implications for Baseline Target Moderated Mediation Analysis with Multiple Moderators.

Recently, Baseline Target Moderated Mediation (BTMM) has received a lot of attention in the field of prevention science. Prevention scientists are interested in BTMM because the model goes beyond whether an intervention achieves effects but also details how and for whom the intervention is most effective. In BTMM, baseline measures are used to investigate potential baseline-by-treatment interactions. However, BTMM has some important challenges including how to incorporate multiple moderator variables when identifying subgroups that benefit the most from the intervention and how to interpret subgroup effects in the presence of multiple moderator variables. Further, with the emergence of causal mediation analysis, it is important to investigate potential treatment-by-mediator interactions which allow the posttest mediator-outcome relation to vary in magnitude across intervention groups. Few methodological developments have addressed the challenges of assessing BTMM in the presence of multiple baseline-by-treatment interactions and the treatment-by-posttest mediator interaction. If the goal is to identify subgroups of individuals who respond better/worse to the intervention, it is important to use a method that can handle the many possible interactions while capturing the heterogeneity within the subgroups of interest. There are three aims of this paper. First, we describe the methodological challenges and substantive interpretation of mediation effects in the presence of multiple moderating variables. Second, we describe two statistical methods to estimate conditional mediation effects in the presence of multiple moderating variables. Third, the methods are applied to an empirical example from the ATLAS study. Implications for BTMM are discussed.