Enhancing ADHD Prediction in Adolescents through Fitbit-Derived Wearable Data

Attention-deficit/hyperactivity disorder (ADHD) is a prevalent neurodevelopmental condition characterized by a complex etiology. The diagnostic process for ADHD is often time-consuming and subjective in nature. Recent advancements in machine learning offer promising avenues for improving ADHD diagnosis automatically using various data sources. In this study, we harness Fitbit-derived physical activity measurements to investigate potential associations with ADHD and evaluate machine learning classifiers for their predictive accuracy in ADHD diagnosis. Our analysis involves a sample of 450 participants from the Adolescent Brain Cognitive Development (ABCD) study data release 5.0. We conduct correlation analyses to explore the connections between ADHD diagnosis and Fitbit-derived measurements, including sedentary time, resting heart rate, and energy expenditure. Subsequently, we employ multivariable logistic regression models to assess the predictive capability of these measurements for ADHD diagnosis. Furthermore, we train machine learning classifiers to achieve a diagnosis by automatically categorizing individuals into ADHD+ and ADHD-groups. Our correlation analysis unveils statistically significant associations between ADHD diagnosis and Fitbit-derived measurements, suggesting a potential link between physical activity patterns and ADHD. Importantly, multivariable logistic regression models demonstrate that some of the Fitbit measurements significantly predict ADHD diagnosis. Notably, our Random Forest machine learning classifier outperforms other classifiers with cross-validation accuracy (0.89), AUC (0.95), precision (0.88), recall (0.90), f1-score (0.89) and test accuracy (0.88), surpassing the performance of previous ADHD classification studies. These findings not only lay the groundwork for further exploration but also offer insights into the clinical integration of wearable data for a deeper understanding and improved identification of ADHD.