A random forest ranks mortality-associated markers in the Italian CORIST cohort
Source check: Source checked. Not independent scientific confirmation.
Finding: An exploratory random forest analysis of 3,894 hospitalized COVID-19 patients ranked renal function, C-reactive protein and age as influential mortality-classification features. Conventional survival analyses in the same cohort supported associated risks. These retrospective findings are neither causal effects nor an independently validated clinical risk tool.
Technology: Random forest, permutation feature importance and k-nearest-neighbor imputation
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- The learned forest and permutation analysis provided an exploratory nonlinear feature ranking; conventional survival modeling supplied complementary association estimates. Their agreement is not independent confirmation.
- Test and result
- The retrospective 30-center Italian cohort included 3,894 patients admitted from February 19 to May 23, 2020; 712 died and 532 were still hospitalized at the May 29 cutoff. RF development used 2,725 patients and a random 1,169-patient test subset, with training hyperparameter search. Renal function and C-reactive protein ranked above age in the reported feature-importance analysis; conventional Cox models analyzed associations with time to in-hospital death.
- Limitation / still unknown
- Imputation and normalization are described before the train/test split, without clear fold-specific fitting. Although Methods discuss test-set importance, the primary figure caption places the permutation-importance analysis in the training sample; this record does not claim independently reproduced feature rankings. The test subset came from the same pooled cohort, not a new hospital or external population. Ongoing hospitalization at cutoff complicates a binary death/no-death target. High reported survival sensitivity coexisted with low specificity; overall accuracy is not validated mortality sensitivity. Missing smoking/obesity data, absent vital signs, treatment differences and residual confounding limit interpretation. Cox hazard ratios are conventional same-cohort associations, not AI-derived causal findings.
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/PMC7833278/
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 primary computational Methods, cohort/reference definitions, represented Results and bounded limitations actually read; no code executed or independent reanalysis..
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.