Machine learning-based early warning model for adolescent mental health risk using the p factor.

Area of Science:

• Psychiatry
• Machine Learning
• Adolescent Health

Background:

• Accurate identification of high-risk adolescents is crucial for timely mental health interventions.
• The general factor of psychopathology (p factor) offers a transdiagnostic approach for risk assessment.
• Ecological systems theory provides a framework for multidimensional risk assessment in adolescents.

Purpose of the Study:

• To develop and validate a machine learning model for identifying adolescents at high mental health risk.
• To utilize the transdiagnostic p factor as an outcome for risk prediction.
• To identify key predictors of mental health risk in adolescents.

Main Methods:

• Trained and validated machine learning models on data from 5,283 Chinese adolescents, with external validation on 968 participants.
• Constructed a multidimensional early warning framework with 59 indicators across individual, school, family, and societal domains.
• Employed Shapley Additive exPlanations (SHAP) to rank predictor importance and identify an optimal feature subset for the XGBoost model.

Main Results:

• The XGBoost model achieved high performance, with macro F1 scores of 0.73 (internal) and 0.80 (external validation).
• The final predictive model incorporated 23 key predictors.
• Sleep quality was identified as the most influential predictor, followed by repetitive negative thinking, interpersonal stress, impulsivity, and emotional intensity.

Conclusions:

• A p-factor-based machine learning model demonstrates significant potential for effectively identifying adolescents at mental health risk.
• Sleep quality emerges as a critical, modifiable factor for early intervention and prevention strategies in adolescent mental health.
• The study underscores the utility of transdiagnostic approaches and machine learning in advancing adolescent mental healthcare.