Interoperable and explainable machine learning models to predict morbidity and mortality in acute neurological injury in the pediatric intensive care unit: secondary analysis of the TOPICC study
Source check: Source checked. Not independent scientific confirmation.
Finding: In a TOPICC secondary analysis of 1,860 pediatric intensive-care patients flagged for possible acute neurological injury, an ensemble of machine-learning models improved mortality precision measures over the original logistic model. The work concerns retrospective risk prediction and does not establish that using the models improves care.
Technology: An ensemble meta-regressor combines random forest, gradient boosting, and SVM on admission physiology; SHAP ranks contributing predictors.
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- An ensemble meta-regressor combines random forest, gradient boosting, and SVM on admission physiology; SHAP ranks contributing predictors.
- Test and result
- Out-of-bag assessment reported mortality AUROC 0.91 and average precision 0.59 versus 0.49 for the original model; at specificity 0.995, PPV was 0.79 versus 0.60.
- Limitation / still unknown
- Secondary retrospective analysis with internal resampling; no prospective decision impact or patient benefit was tested.
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/PMC10338865/
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: Primary full text, including Methods and Results; primary publication record or first result-bearing version checked..
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 3/5 · 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.