Scholarlink

Background <p>Machine learning (ML) has been widely used to predict suicidal attempts (SA) in adolescents and adults. However, there is a need for studies of accurate and efficient SA models specifically for use in adolescents with mood disorders due to a lack of existing research integrating risk variables when predicting clinical SA. As such, this study aimed to explore the potential of ML models to predict SA among adolescents with mood disorders.</p> Methods <p>Study participants were 134 Chinese adolescents with mood disorders (20.1% boys, <i>M</i><sub>age</sub> = 15.26, <i>SD</i> = 1.67) and their treating physicians. Of these adolescents, 53 had clinically documented SA. Four ML models – logistic regression (LR), random forest (RF), eXtreme gradient boosting machine (XGBM), and light gradient boosting machine (LGBM) were used to predict SA in participants, and their results were compared to identify key predictive variables as well as the best-performing model. Class imbalance was handled via Synthetic Minority Oversampling Technique applied only to each training fold, and every model was tuned and evaluated with stratified 10-fold cross-validation.</p> Results <p>The LGBM model achieved the highest discriminant performance, with an AUC of 0.79 and an accuracy of 0.75, with factors of the emotional expression functioning of non-suicidal self-injury, suicidal planning, peer relationship issues, physician-assessed depression levels, and patient self-reported anxiety and depression levels all demonstrating the highest values in predicting SA.</p> Conclusion <p>The findings of this study indicate that the LGBM model, which uses adolescents’ clinical characteristics and experiences, was the most effective in estimating SA in adolescents with mood disorders. Future research should validate this model further in other populations, as well as assess its utility in selecting suicide prevention interventions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prediction of suicidal attempts among Chinese adolescents with mood disorders: a clinical study using a machine learning approach

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月15日

查看全文
  • Jianbing Li,
  • Yinqiu Zhao,
  • Chi Yang,
  • Wenqing Li,
  • Zhihao Huang,
  • Changhe Fan

摘要

Background

Machine learning (ML) has been widely used to predict suicidal attempts (SA) in adolescents and adults. However, there is a need for studies of accurate and efficient SA models specifically for use in adolescents with mood disorders due to a lack of existing research integrating risk variables when predicting clinical SA. As such, this study aimed to explore the potential of ML models to predict SA among adolescents with mood disorders.

Methods

Study participants were 134 Chinese adolescents with mood disorders (20.1% boys, Mage = 15.26, SD = 1.67) and their treating physicians. Of these adolescents, 53 had clinically documented SA. Four ML models – logistic regression (LR), random forest (RF), eXtreme gradient boosting machine (XGBM), and light gradient boosting machine (LGBM) were used to predict SA in participants, and their results were compared to identify key predictive variables as well as the best-performing model. Class imbalance was handled via Synthetic Minority Oversampling Technique applied only to each training fold, and every model was tuned and evaluated with stratified 10-fold cross-validation.

Results

The LGBM model achieved the highest discriminant performance, with an AUC of 0.79 and an accuracy of 0.75, with factors of the emotional expression functioning of non-suicidal self-injury, suicidal planning, peer relationship issues, physician-assessed depression levels, and patient self-reported anxiety and depression levels all demonstrating the highest values in predicting SA.

Conclusion

The findings of this study indicate that the LGBM model, which uses adolescents’ clinical characteristics and experiences, was the most effective in estimating SA in adolescents with mood disorders. Future research should validate this model further in other populations, as well as assess its utility in selecting suicide prevention interventions.