Transdiagnostic behavioral and sociodemographic influences on the cognitive-adaptive functioning gap in neurodivergent children.

Abstract

Adaptive functioning typically aligns with cognitive ability in the general population, but this relationship appears more complex in neurodivergent populations. Individuals with autism, attention-deficit/hyperactivity disorder (ADHD), and obsessive-compulsive disorder (OCD) frequently show a discrepancy between cognitive ability and adaptive functioning. However, the behavioral and sociodemographic factors associated with this cognitive-adaptive functioning gap remain unclear. Data were collected from the Province of Ontario Neurodevelopmental Disorders Network (POND) and included 902 participants aged 6-21 years (autism = 409, ADHD = 210, OCD = 36, neurotypical = 214, other = 33). Cognitive functioning was assessed using full-scale IQ from the Wechsler tests, and adaptive functioning using the Adaptive Behavior Assessment System-II (General Adaptive Composite). The cognitive-adaptive functioning gap was calculated as the difference between FSIQ and ABAS-II GAC scores. This gap was modeled transdiagnostically with phenotypic (social communication, ADHD and OCD traits, internalizing and externalizing symptoms) and sociodemographic (age, sex, race, household income, caregiver education) features. Three classifiers (i.e. logistic regression, CatBoost, and Explainable Boosting Machine) were trained to distinguish children with above- versus below-median cognitive-adaptive functioning gap scores. Feature importance was assessed for the selected model using SHAP values. Sensitivity analyses repeated the performance analysis and feature-importance assessment using a 15-point FSIQ-ABAS GAC gap threshold and a continuous-outcome model of the numerical FSIQ-ABAS GAC gap. Logistic regression, CatBoost, and Explainable Boosting Machine models showed comparable performance (ROC-AUC = 0.698-0.700). CatBoost was selected for SHAP-based interpretation because it achieved competitive performance and allowed efficient tree-based SHAP estimation. SHAP analysis identified internalizing problems, social communication challenges, and OCD symptoms as the strongest predictors. Behavioral factors contributed more to the gap than sociodemographic variables. Sensitivity analyses using a 15-point gap threshold and a continuous gap outcome retained the same core behavioral contributors: internalizing problems, social communication challenges, and OCD symptoms. Internalizing symptoms and social communication difficulties were most strongly associated with the cognitive-adaptive functioning gap.