Predicting the Persistence of Mental Health Problems in Adolescents: A Longitudinal Study

AbstractBackgroundAdolescent mental health problems often follow varied courses, including persistence or remission. Early identification of these trajectories is essential for targeted intervention. MethodsUsing data from the Adolescent Brain Cognitive Development (ABCD) study, we identified participants (ages 9–10) with clinically relevant symptoms of six mental health problems (ADHD, anxiety, conduct, depressive, oppositional defiant, and somatic) based on the Child Behavior Checklist (CBCL). At 2-year follow-up, individuals were classified as persisters and remitters. We trained random forest models using CBCL scores to predict individual symptom persistence. In addition, structural MRI (sMRI), resting-state fMRI (rsfMRI), and diffusion tensor imaging (DTI) data were analyzed to identify baseline and developmental brain differences between persisters and remitters.ResultsPersistence rates ranged from 36% to 50% across problems at 2 year follow up. Models with CBCL scores showed acceptable prediction for anxiety (AUC = 0.72), while other disorders showed lower accuracy (AUC = 0.55–0.65). Higher CBCL total scores at baseline were consistently observed in persisters (p < .05). Imaging analyses revealed significant baseline brain differences for anxiety, somatic, and oppositional defiant problems. Longitudinal neurodevelopmental differences were only found for oppositional defiant problems.ConclusionsMental health problems in adolescents often persist over two years. Persistence of anxiety symptoms can be predicted using CBCL items, whereas predictions for other problems remain limited. Neuroimaging results indicate that symptom persistence is partly linked to preexistent brain differences. Predictive models may help guide early identification and personalized intervention for youth at risk of chronic mental health problems, in particular for anxiety.