Machine learning prediction of adverse events during minutes 15–30 of recovery in patients event-free at 15 min after sedated gastrointestinal endoscopy
Source check: Source checked. Not independent scientific confirmation.
Finding: In 1,070 patients undergoing sedated gastrointestinal endoscopy at West China Hospital, XGBoost predicted prespecified respiratory or hemodynamic events during minutes 15–30 of recovery among patients event-free at minute 15. The model achieved an AUC of 0.910 in the independent validation set (cross-validated training AUC 0.971). Prospective, temporally separated evaluation and external validation are still needed.
Technology: XGBoost with Boruta feature selection and SHAP interpretation
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Boruta selected candidate predictors; XGBoost estimated risk of the specified post-anesthesia events, and SHAP described contributions without implying causality.
- Test and result
- The study evaluated six supervised models in a retrospective cohort; 103 of 1,070 patients had the prespecified event. XGBoost had validation AUC 0.910, PR-AUC 0.815 and Brier score 0.076, using five final predictors.
- Limitation / still unknown
- The cohort came from one hospital and the study calls for prospective testing with predictor and outcome windows separated in time, plus external validation, before clinical implementation.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13558014/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 22 Nov 2026. Review scope: Primary result-bearing material for the stated method, result and evaluation, with bibliographic version information..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.