Longitudinal models updated ICU mortality risk with incomplete records
Source check: Source checked. Not independent scientific confirmation.
Finding: Longitudinal ICU models using MIMIC-III data predicted death 12–24 hours after an observation window. Multiple imputation expanded the analysis from 4,276 complete cases to 16,532 patients with 70 predictors while retaining similar discrimination; reported AUCs were about 0.78–0.79. The first result-bearing version was the 2022-04-28 medRxiv preprint; the journal article followed in 2023. This retrospective proof of concept has no external validation or demonstrated clinical benefit.
Technology: Deep neural networks and penalized regression with multiple imputation
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Learned models updated short-term mortality estimates from longitudinal EHR measurements, and imputation addressed incomplete records.
- Test and result
- Retrospective MIMIC-III analysis with repeated five-fold cross-validation; authors describe it as proof of concept and call for evaluation in other ICUs.
- Limitation / still unknown
- No independent, prospective or clinical-impact evaluation is reported unless specified in the tested field.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://www.medrxiv.org/content/10.1101/2022.04.28.22274405v1.full
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: Read the original full text, including author affiliations, Methods and Results; checked the first result-bearing version and active/archive catalog for same-result records..
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.