External testing measured mortality-model performance in sepsis-associated liver injury
Source check: Source checked. Not independent scientific confirmation.
Finding: Models using first-24-hour features predicted 7-, 14- and 28-day in-hospital mortality. In an external cohort of 229 patients, AUROCs were 0.720, 0.717 and 0.763; the 28-day model exceeded modified SOFA and MELD-Na in this cohort, but transportability was moderate and 14-day calibration imperfect. A correction clarified that the analysis used first-24-hour data. Retrospective prognostic performance does not establish treatment benefit; the authors call for prospective multicenter validation.
Technology: Interpretable supervised machine-learning models with SHAP analysis
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Supervised models estimated mortality from first-day clinical features; SHAP was used to interpret predictions.
- Test and result
- Models using first-24-hour features predicted 7-, 14- and 28-day in-hospital mortality. In an external cohort of 229 patients, AUROCs were 0.720, 0.717 and 0.763; the 28-day model exceeded modified SOFA and MELD-Na in this cohort, but transportability was moderate and 14-day calibration imperfect. A correction clarified that the analysis used first-24-hour data.
- Limitation / still unknown
- Retrospective prognostic performance does not establish treatment benefit; the authors call for prospective multicenter validation.
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/PMC13719580/
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: 22 Nov 2026. Review scope: Original full-text Methods and Results reviewed from the primary paper or preprint; first result-bearing version checked from the supplied source review; author affiliations checked against exact-DOI metadata where available..
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 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.