Who thrives against the odds? Mapping adolescent trait heterogeneity of emotional maltreatment effects with causal forests DML

Abstract

Emotional maltreatment (EM) during childhood and adolescence exerts lasting effects on psychopathology, yet individuals exposed to similar adversity often show divergent outcomes. To examine heterogeneous effects of cumulative EM on young-adult psychopathology and identify adolescent strengths associated with resilient subgroup among adolescents with high EM. Data came from 738 participants in the Future of Families and Child Wellbeing Study. Caregivers reported EM at ages 3, 9, and 15 (CTSPC); adolescents self-reported five strengths at age 15 (EPOCH); and major depressive and anxiety disorders were assessed at age 22 (CIDI). Causal forest DML models estimated the average marginal effect (AME) and heterogeneous conditional average treatment effects (CATEs) of cumulative EM on adult psychopathology, controlling for early emotional symptoms and sociodemographic factors. Within the high EM subgroup ( 67th percentile), resilient and vulnerable subgroups were defined using pre-specified tails of the out-of-fold CATE distribution. Adolescent strengths were then examined as predictors of the resilient subgroup using gradient boosting. Greater cumulative EM predicted higher adult psychopathology risk (AME = 0.027, 95% CI [0.013-0.038]), though effects varied across individuals. Within the high-EM subgroup, approximately 20% were classified into the resilient subgroup. Associations between adolescent strengths and the resilient subgroup were nonlinear: engagement peaked at mid-range levels; happiness showed benefits mainly at higher levels; optimism and perseverance showed direction-changing associations across the range; and connectedness showed smaller, relatively stable associations. Cumulative EM exerts robust adverse effects with marked heterogeneity. Adolescent strengths operate non-linearly, underscoring resilience as a dynamic developmental capacity informing precision prevention.