ADHDSymTracker: Predicting ADHD Symptoms using Apple HealthKit Data

Abstract

Attention Deficit Hyperactivity Disorder (ADHD) is a prevalent condition that impacts cognitive and behavioral functioning, posing significant challenges for individuals’ academic and daily lives, particularly among college students. The core symptoms are inattention, hyperactivity and impulsivity. Current diagnostic and symptom tracking methods, whether clinician-administered or self-reported, have several limitations, such as recall bias, high costs, and the necessity for manual intervention. This underscores the necessity for an objective, accurate, and cost-effective tool for ADHD diagnosis that requires minimal manual intervention. To address this issue, we propose a novel approach, ADHDSymTracker, which uses Apple HealthKit data to predict ADHD symptoms. We calculated behavioral features using data collected from 38 college-age students including some with ADHD and developed a suite of machine learning models for ADHD symptom prediction. Our results from ADHDSymTracker indicate that most symptoms can be predicted with reasonable accuracy, achieving an $F_1$ score as high as 0.72, rendering it a promising solution for automatic and continuous ADHD monitoring.