Journal of Clinical and Translational Science (Apr 2024)

394 A Machine Learning Approach to Predicting High-Risk Irritability Trajectories Across the Transition to Adolescence

  • Leslie S. Jordan,
  • Alyssa J. Parker,
  • Jillian Lee Wiggins,
  • Lea R. Dougherty

DOI
https://doi.org/10.1017/cts.2024.344
Journal volume & issue
Vol. 8
pp. 117 – 117

Abstract

Read online

OBJECTIVES/GOALS: Irritability, a proneness to anger and frustration, is a transdiagnostic symptom associated with poor mental health outcomes. Levels of irritability vary across development and high-risk trajectories have been observed. This study aims to use machine learning to predict irritability trajectories across the transition to adolescence. METHODS/STUDY POPULATION: Data were from the Adolescent Brain Cognitive Development (ABCD) Study, which is a 10-year longitudinal study that tracks the brain development, cognitive skills, physical health, and psychosocial functioning of a large, national sample starting from preadolescence. The baseline sample consisted of 11,861 9-10-year-old preadolescent youth. Irritability was parent-rated at baseline, 1-year, 2-year, 3-year, and 4-year follow-ups on the Child Behavior Checklist (CBCL) irritability index. Latent class growth analysis (LCGA) was used to determine developmental trajectories of irritability. Two machine learning approaches were applied to develop predictive models of youth irritability developmental trajectories. We used baseline (preadolescent) variables that spanned a wide range of domains. RESULTS/ANTICIPATED RESULTS: Preliminary results fromthe LCGA indicated best support for a four-class model that differentiated growth trajectories in irritability across the transition to adolescence: 1) persistent low irritability (n = 8691, 73.27%), 2) moderate irritability and decreasing (n = 1257, 10.60%), 3) low to moderate irritability and increasing (n = 1295, 10.92%), and 4) chronic high irritability (n = 618, 5.21%). We expect the machine learning analyses to generate predictive models with acceptable accuracy. We hypothesize that the most important predictors in the models will originate from the youth mental health domain, including baseline youth irritability, externalizing symptoms, internalizing symptoms, and oppositional behaviors, and the parent psychopathology domain, particularly parent irritability. DISCUSSION/SIGNIFICANCE: The present study elucidates unique developmental trajectories of irritability and generates predictive models to classify high-risk irritability trajectories using machine learning approaches. Clinicians can use these predictive models to identify at-risk youth and provide early intervention to preadolescents at high risk.