A machine learning approach to predicting high-risk irritability trajectories across the transition to adolescence.

Abstract

Youth irritability, characterized as proneness to anger/frustration, is associated with negative mental health outcomes. This study identified irritability trajectories across the transition to early adolescence and applied machine learning to identify preadolescent predictors of irritability trajectories. Data are from the Adolescent Brain Cognitive Development (ABCD) study, a large longitudinal study that tracks youth development from preadolescence (n = 11,862, ages 8.92-11.08 years, M = 9.91, SD = 0.62). The baseline sample was 47.83% female (n = 5,680), 63.35% White, 15.74% Black, 12.07% Biracial or Multiracial, 2.32% Asian, and 20.30% Hispanic. Using parent-report youth irritability scores across four years, latent class growth analysis (LCGA) identified four distinct irritability trajectories: low stable (n = 8,692, 73.28%), high decreasing (n = 1,083, 9.13%), low increasing (n = 1,449, 12.22%), and high chronic (n = 638, 5.38%). Machine learning approaches generated predictive models for irritability trajectories, including baseline youth psychopathology, physical health, neurocognitive abilities, psychosocial environment, demographics, parent psychopathology, and brain structure metrics. The models demonstrated strong performance in detecting probability of being in the high chronic and low stable trajectories but poorer performance in classifying the irritability trajectories characterized by change. The most prominent predictors included youth externalizing and internalizing problems, parent psychopathology, and sleep problems. Findings provide insight into preadolescent predictors of irritability across a critical developmental period.