Machine-learning prediction of suboptimal weight-loss response after sleeve gastrectomy
Source check: Source checked. Not independent scientific confirmation.
Finding: An XGBoost model using eight routine preoperative predictors estimated the risk of less than 25% total weight loss one year after sleeve gastrectomy. In retrospective cohorts from three hospitals (1,094 derivation patients and external cohorts of 84 and 134), external AUC was 0.761; the paper also reports better performance than ten clinicians in its comparison. This supports risk stratification in the studied cohorts, not a causal benefit from model-guided care.
Technology: Six candidate algorithms were compared; an XGBoost model using eight predictors was selected and interpreted with SHAP.
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Six candidate algorithms were compared; an XGBoost model using eight predictors was selected and interpreted with SHAP.
- Test and result
- Retrospective derivation and two independent external hospital cohorts; internal AUC 0.821 and external AUC 0.761. The study compared model predictions with ten clinicians.
- Limitation / still unknown
- All cohorts were retrospective and limited to the participating hospitals; no prospective impact study established that using the model improves weight-loss outcomes.
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/PMC13732101/
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 full text and PubMed bibliographic record, including cohort design, model results and publication date..
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 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.