Assessing the Potential Cumulative Effects of Longitudinal Interventions for Depression and Deviant Peer Association on Academic Achievement: a Causal Inference Approach.

In various studies, depressive symptoms, deviant peer association, and academic achievement appear to be reciprocally and cumulatively related across the transition from childhood to adolescence. This study asks a hypothetical question: how much would grade 12 academic achievement improve if a preventive-intervention reduced depressive symptoms and/or deviant peer association across grades 5, 7, and 10? To answer this question, we used longitudinal datasets from 361 students in the LIFT project and employed the Longitudinal Modified Treatment Policies approach, which uses causal inference and machine learning analyses to estimate the total effect of shifting the trajectory(ies) of an exposure(s) while properly handling time-varying confounding. We estimated the effects of reducing depressive symptoms and/or deviant peer association across the grades by a hypothetical magnitude ranging from 10 to 50%. A 50% reduction in depressive symptoms, deviant peer association, or both was associated with improvements in academic achievement of approximately 7.8%, 14.0%, and 17.1%, respectively; the effects remained positive and statistically significant at every reduction magnitude examined. Implications of the findings for identifying key factors across critical developmental windows are discussed.