Machine learning-based prediction of suicide attempts among adolescents: a national study using explainable...

To develop and evaluate machine learning models for predicting adolescent suicide attempts and to examine predictor contributions using explainable artificial intelligence.

A repeated cross-sectional study used pooled data from the 2017-2024 Korea Youth Risk Behavior Web-Based Survey (n=448, 798). Models including logistic regression, random forest, and XGBoost were developed to classify suicide attempts. Performance was evaluated using AUC, F1 score, and sensitivity-oriented screening thresholds to reflect population-level screening purposes. SHapley Additive exPlanations (SHAP) quantified predictor contributions.

The models showed moderate predictive performance, with XGBoost achieving the highest F1 score and random forest showing the highest sensitivity under the screening condition. SHAP analysis indicated that hopelessness contributed most strongly to prediction, followed by school violence and perceived stress. Additional contributors included self-rated health, household economic status, sleep satisfaction, and behavioral indicators. Under the screening condition, sensitivity improved, although positive predictive values remained low.

Machine learning models demonstrated moderate performance in predicting adolescent suicide attempts. Psychological and social variables contributed most strongly to prediction, and SHAP improved the interpretability of model outputs, supporting their potential utility for population-level screening support.