Related Experiment Video
Updated: May 19, 2026
05:19
Closed-Loop Neurostimulation for Biomarker-Driven, Personalized Treatment of Major Depressive Disorder
Published on: July 7, 2023
Can machine learning predict non-suicidal self-injury? A systematic review and meta-analysis
Qianhui Wen1,2, Rong Luo1,2, Qian Wang1,2
1Department of Pediatrics, West China Second University Hospital, Sichuan University, Chengdu, Sichuan, China.
Frontiers in Public Health
|May 18, 2026
Summary
Machine learning (ML) models show promise for predicting non-suicidal self-injury (NSSI) risk in adolescents and young adults. However, current evidence is limited by bias and heterogeneity, requiring further prospective studies for clinical use.
Area of Science:
- Psychiatry and Mental Health
- Computational Science
- Public Health
Background:
- Non-suicidal self-injury (NSSI) is a significant concern in adolescents and young adults.
- Conventional methods struggle with early NSSI detection.
- Machine learning (ML) offers potential for developing predictive models for NSSI.
Purpose of the Study:
- To systematically review and meta-analyze existing studies on ML models for NSSI prediction.
- To evaluate the performance of various ML models in identifying NSSI risk.
- To identify limitations and future research needs in this field.
Main Methods:
- A systematic review and meta-analysis of studies developing ML models for NSSI prediction.
- Searched multiple databases from inception to June 28, 2025.
- Synthesized model performance metrics (AUC, sensitivity, specificity) using a bivariate random-effects model.
- Assessed risk of bias using PROBAST+AI.
Main Results:
- Twelve studies with 33,366 participants were included.
- Ensemble ML models demonstrated favorable pooled discrimination (AUC: 0.83).
- Single ML models showed lower performance (AUC: 0.68).
- Substantial heterogeneity and high risk of bias were observed across studies.
Conclusions:
- ML models show potential for NSSI risk identification.
- Current evidence for prospective prediction is limited due to bias and heterogeneity.
- Prospective, multicenter studies with external validation are necessary for clinical translation.