Supplementary file 2_Can machine learning predict non-suicidal self-injury? A systematic r

Twelve studies involving 33,366 participants were included in this systematic review and meta-analysis of machine learning models for predicting non-suicidal self-injury. The analysis synthesized model performance, finding a pooled AUC of 0.83 for ensemble models and 0.75 across all 19 models. This supplementary document, authored by Qianhui Wen and licensed under CC-BY-4.0, was last updated on April 30, 2026.

Use Cases

• Benchmarking new machine learning models for NSSI prediction based on the reported performance metrics (AUC, sensitivity, specificity).
• Identifying methodological gaps and risk of bias in existing NSSI prediction studies based on the PROBAST+AI assessment.
• Informing the design of prospective multicenter studies based on the review's conclusions about heterogeneity and validation needs.

Strengths

• Includes a quantitative synthesis of 12 studies and 33,366 participants, providing a consolidated evidence base.
• Reports specific pooled performance metrics, including an AUC of 0.83 for ensemble models and 0.75 across all models.
• Uses a standardized risk-of-bias assessment tool (PROBAST+AI) for methodological evaluation.

Limitations

• The dataset is a 28.4 KB DOCX file, indicating it is a supplementary document summarizing findings rather than a primary data repository.
• Column-level documentation is absent; field semantics must be inferred after download.
• The evidence base is constrained by substantial heterogeneity and a high risk of bias, as noted in the conclusions.

Provenance

Source

figshare

Collection Method

Systematic review and meta-analysis of studies from multiple databases searched from inception to June 28, 2025.

Time Range

Studies from database inception to June 28, 2025.

Freshness

Last updated 2026-04-30 14:01:21; freshness should be verified.

Geography

null