A precision health approach to medication management in neurodivergence: a model development and validation study using four international cohorts.

Abstract

Background: Psychotropic medications are commonly used for neurodivergent children, but their effectiveness varies, making prescribing challenging and potentially exposing individuals to multiple medication trials. We developed artificial intelligence (AI) models to predict medication success for stimulants, anti-depressants, and anti-psychotics. We first demonstrate feasibility using cross-sectional data from three research cohorts, then use a cohort of patients from a pharmacology clinic to predict medication choice by class, longitudinally, from electronic medical records (EMRs).

Methods: Models were built to predict cross-sectional medication usage from the Child Behaviour Checklist. Data from the Province of Ontario Neurodevelopmental (POND) network (N=598) trained and tested the models, while data from the Healthy Brain Network (HBN; N=1,764) and Adolescent Brain Cognitive Development (ABCD; N=2,396) studies were used for external validation. For the EMR cohort, data from the Psychopharmacology Program (PPP; N=312) at Holland Bloorview Kids Rehabilitation Hospital were used to predict longitudinal success. Stacked ensemble models were built separately for each medication class, and area under the receiving operating characteristic curve (AU-ROC) evaluated performance.

Findings: The research cohorts demonstrated feasibility, with internal testing (POND) achieving an AU-ROC (mean [95% CI]) of 0.72 [0.71,0.74] for stimulants, 0.83 [0.80,0.85] for anti-depressants, and 0.79 [0.76,0.82] for anti-psychotics. Performance in external testing sets (HBN and ABCD) confirmed generalizability. In the EMR cohort (PPP), AU-ROC were high: 0.90 [0.88,0.91] for anti-psychotics, 0.82 [0.92,0.83] for stimulants and 0.82 [0.80,0.84] for anti-depressants.

Interpretation: This study demonstrates the feasibility of using AI to enhance medication management for neurodivergent children, with expert clinician decisions learned with high accuracy. These findings support the potential for AI decision aids in community settings, promoting faster access to personalized care while highlighting the complexity of clinical and sociodemographic factors influencing medication decisions.