AI Science Tracker
EN

A random forest ranks mortality-associated markers in the Italian CORIST cohort

Mediterranea Cardiocentro; IRCCS Neuromed; CORIST collaboration · 31 Jul 2020 · medicine

Discovery

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.

Open in the full catalog for filters, charts and related discoveries. Browse medicine discoveries.