Kidney-injury models show changing laboratory contributions before recorded onset
Source check: Source checked. Not independent scientific confirmation.
Finding: A reweighted XGBoost analysis of hospital records compared feature contributions at24,48 and72 hours before kidney-injury onset. Laboratory measurements accounted for54.5% of the model-score contrast increase from72 to24 hours. This describes learned-model attributions in one hospital, not the causal effect of changing those measurements.
Technology: Reweighted XGBoost; SHAP
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Modified feature-sampling weights emphasize prespecified modifiable feature views; fitted models and SHAP-derived score contrasts quantify how learned associations differ by retrospective prediction window.
- Test and result
- The24-hour cohort contained144084 encounters; models trained on2010–2016 and tested on2017–2018. Cohort sizes differed for longer windows. Fifty random-seed repetitions informed mean attributions.
- Limitation / still unknown
- Feature views were deliberately reweighted, so their importance is conditional on that modeling choice. AKI cases were anchored to first onset, controls to last creatinine measurement; repeated encounters and different window cohorts limit comparisons. Correlations among modifiable factors were not modeled explicitly. The source’s predicted intervention benefits are not measured treatment effects, patient-specific causal risks or prevention evidence. Supplementary algorithm/hyperparameter details were not independently read.
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/PMC12516820/ · https://doi.org/10.1016/j.ijmedinf.2022.104785
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: 21 Nov 2026. Review scope: Original primary clinical/computational Methods, model evaluation and high-level Results read; no operational biological protocols.
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.