Wearable Signals for Diagnosing Attention-Deficit/Hyperactivity Disorder in Adolescents: A Feasibility Study.

Area of Science:

• Neurodevelopmental disorders
• Adolescent health
• Biomedical engineering

Background:

• Attention-deficit/hyperactivity disorder (ADHD) is a common neurodevelopmental disorder.
• Current ADHD diagnoses rely heavily on subjective parent and teacher ratings.
• Objective measures are needed to enhance ADHD assessment accuracy.

Purpose of the Study:

• To investigate the feasibility of using objective activity monitoring data for ADHD diagnosis in adolescents.
• To compare the diagnostic performance of objective measures versus subjective measures.
• To explore the utility of wearable activity trackers in monitoring ADHD in adolescents.

Main Methods:

• A longitudinal study involving Chinese adolescents aged 16-17 years.
• Collected objective data (actigraphy: movement, heart rate, sleep) and subjective data (questionnaires).
• Developed eXtreme Gradient Boosting (XGBoost) machine learning models for ADHD classification and analyzed feature importance using SHAP.

Main Results:

• Machine learning models using objective measures alone achieved high ADHD classification accuracy (AUC = 0.844).
• Models integrating both objective and subjective measures showed superior performance (AUC = 0.933).
• Key predictors for ADHD included irritability, sex, and quality of life; for medication status, heart rate and physical activity intensity were key.

Conclusions:

• Actigraphy offers noninvasive, granular objective vital sign monitoring for adolescents.
• Objective measures, if validated in larger samples, can enhance multidimensional assessment and diagnostic accuracy for ADHD.
• Wearable activity trackers show potential to supplement existing ADHD diagnostic methods.