A seven-feature model estimates complications after congenital heart reoperation
Source check: Source checked. Not independent scientific confirmation.
Finding: Researchers trained a retrospective seven-feature risk model in 1,011 adults undergoing repeat congenital heart surgery. The selected gradient-boosted model reported test AUC 0.763 and validation AUC 0.782 for a composite of death and major complications. External validation and evidence of clinical benefit remain absent.
Technology: XGBoost gradient boosting and SHAP feature analysis
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Gradient boosting fitted clinical variables to a composite postoperative outcome; SHAP ranked model contributions and supported a reduced seven-variable model.
- Test and result
- The 2010–2022 single-center cohort was randomly split 70%/20%/10% for training/testing/validation. Eighteen preselected clinical variables were reduced to seven using SHAP contribution. Systemic ventricular ejection fraction was most influential in the fitted models. Reduced-model test/validation AUCs were .763/.782, with reported validation CI .568–.804 and five-fold range .631–.764.
- Limitation / still unknown
- Authors selected a model at the 90th percentile AUC from bootstrapped model distributions; this is not an unbiased locked-model estimate. Rare composite events, single-center surgical selection and feature/model selection constrain generalization. Logistic-regression AUC .727 had overlapping confidence intervals, so superiority is not established. Conventional survival analysis of heart failure is separate from the machine-learning result, and SHAP influence is not causation.
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/PMC10972775/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: linked separately.
Review date: 21 Nov 2026. Review scope: Original full-text bibliographic date, relevant computational/statistical Methods and Results, bounded cohort and selection caveats; no operational clinical instructions..
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 1/5 · 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.