Machine Learning Predicts Stroke Mortality Risk
A systematic review and meta-analysis published April 2, 2026, evaluated 68 studies (75 models, 43 external validations) up to June 23, 2025 on machine learning models predicting stroke mortality. Pooled external-validation C-indexes were 0.727 for in-hospital mortality and 0.847 for out-of-hospital mortality, with sensitivities ~0.64–0.71 and specificities ~0.74–0.76. Substantial heterogeneity and elevated risk of bias limit immediate clinical deployment; external validation is recommended.
Key Points
- 1Pooled data from 68 studies (75 models, 43 validations) show 0.727 in-hospital C-index.
- 2Out-of-hospital models had higher discrimination (C-index 0.847), indicating stronger long-term prediction.
- 3Recommend external validation and cautious deployment due to heterogeneity, bias, and variable time performance.
Scoring Rationale
This peer-reviewed systematic meta-analysis presents novel, credible evidence on ML performance for stroke mortality (high novelty and credibility). Score reflects strong relevance and useful insights, but scope is stroke-specific and practical actionability is limited by substantial heterogeneity and risk of bias, so deployment requires external validation.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems