Identifying multilevel predictors of trajectories of psychopathology and resilience among juvenile offenders: A machine learning approach

Abstract

Abstract Mental ill health is more common among juvenile offenders relative to adolescents in general. Little is known about individual differences in their long-term psychological adaptation and its predictors from multiple aspects of their life. This study aims to identify heterogeneous trajectories of probable psychiatric conditions and their predictors. Participants included 574 juvenile offenders who were first convicted for serious crimes and without detention history. The participants were assessed at 11 timepoints over seven years (2000–2010). Growth mixture modeling revealed the same three trajectories for both probable anxiety and probable depression: stable low trajectory (75.96%; 75.78%), stable high trajectory (15.16%; 10.98%), and recovery (8.89%, 13.24%). Least absolute shrinkage and selection operator (LASSO) logistic regression identified three multilevel predictors for memberships of different trajectories. Risk factors against stable low trajectory lay within personal (e.g., neuroticism), relationship (e.g., parental hostility), and contextual levels (e.g., chaotic neighborhood). Resilience factors for stable low trajectory included strong work orientation and low education level of father. Recovery was predicted by Black race, self-identity, high education level of father, and nonincarcerated sentencing. Our findings suggest that both psychopathology and psychological resilience could be predicted by multiple personal, relationship, and contextual factors in the social ecology of juvenile offenders.