Unveiling sex difference in factors associated with suicide attempt among Chinese adolescents with depression: a machine learning-based study.

Abstract

Background: Adolescents with depression are at heightened risk of suicide, with a distinct sex difference in suicidal behaviour observed. This study explores the sex-specific factors influencing suicide attempts among Chinese adolescents with depression.

Methods: Data were collected from 2343 depressed adolescents across 14 hospitals in 9 provinces through self-report questionnaires. The survey was conducted between December 2020 and December 2023. Thirty-six potential risk factors were selected from validated measures of psychological, sociodemographic, and social stress domains. The dataset was split by sex, and SMOTE was applied to address class imbalance. Logistic regression, elastic net regression, random forest, XGBoost, and neural networks were used to model the data, evaluated by accuracy, precision, recall, and F1 score. The optimal model was employed for SHapley Additive exPlanations (SHAP) analysis to identify key factors influencing suicide attempts.

Results: The Random Forest model exhibited the best performance for both sexes (AUC: females 0.720, males 0.736). Non-suicidal self-injury and depression were significant predictors for both sexes. Among females, factors like difficulty identifying emotions and physical abuse had a stronger impact, while resilience and hopelessness were more predictive for males.

Conclusions: The study highlights sex differences in suicide attempt predictors, emphasizing the need for sex-specific prevention strategies.