Identifying Adolescent Depression and Anxiety Through Real-World Data and Social Determinants of Health: Machine...

Area of Science:

• Adolescent mental health research
• Machine learning applications in healthcare
• Public health informatics

Background:

• Rising prevalence of adolescent depression and anxiety.
• Limited machine learning models utilizing real-world data (RWD) for early detection.
• Need for enhanced intervention strategies in youth mental health.

Purpose of the Study:

• To develop and evaluate machine learning models for identifying depression and anxiety in adolescents.
• To utilize real-world data (RWD) and social determinants of health (SDoH) for prediction.
• To assess the impact of SDoH on model performance.

Main Methods:

• Analysis of RWD for adolescents aged 10-17 years.
• Development of Extreme Gradient Boosting (XGBoost) models for anxiety, depression, and combined conditions.
• Nested cross-validation for performance evaluation and Shapley additive explanation for interpretation.

Main Results:

• Models achieved high predictive performance (AUC 0.80 for anxiety, 0.81 for depression, 0.78 for both).
• Exclusion of SDoH data had minimal impact on model accuracy.
• Key predictors included gender, race, education, and medical history.

Conclusions:

• Machine learning holds significant potential for early identification of adolescent mental health conditions using RWD.
• RWD-driven ML tools can empower healthcare providers for timely interventions.
• Improved identification can lead to better mental health outcomes for adolescents.