Interpretable machine learning prediction models for 28-day mortality in critically ill patients with atrial fibrillation and acute kidney injury - PubMed
2 days ago
- #mortality prediction
- #critical care
- #machine learning
- Study aims to develop interpretable ML models for predicting 28-day mortality in ICU patients with AF and AKI.
- Retrospective analysis used MIMIC-IV and eICU-CRD databases, including 11,510 and 2,565 patients respectively.
- GBM model performed best with AUC of 0.856 (internal) and 0.761 (external validation).
- SHAP analysis identified anion gap, heart rate, and age as top mortality predictors.
- Online risk calculator developed for clinical application and individualized risk stratification.