Interpretable machine learning for prognostic prediction in critically ill patients with coronary artery disease: a multicenter study
Source check: Source checked. Not independent scientific confirmation.
Finding: Across 15,930 ICU patients with coronary artery disease in MIMIC-IV, a random-forest model predicted 28-day and 365-day mortality and was evaluated on MIMIC-III. AUCs were 0.858 and 0.851 in internal validation and 0.914 and 0.901 in the external database, respectively. Both cohorts are retrospective database evaluations.
Technology: Random-forest mortality models with LASSO feature selection and SHAP interpretation
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- LASSO and commonality analysis selected clinical features; random forest estimated 28-day and 365-day mortality risk, and SHAP summarized model feature contributions.
- Test and result
- The paper compared seven machine-learning algorithms after feature selection. Random forest performed best for 28-day and 365-day mortality in MIMIC-IV internal validation and MIMIC-III external validation, with AUCs 0.858/0.914 and 0.851/0.901.
- Limitation / still unknown
- The validation uses two retrospective MIMIC database cohorts; prospective clinical utility and effects on care or mortality were not tested.
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/PMC13070936/
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.