Retrospective cohort study predicts in-hospital mortality in ICU patients with CHD and diabetes, suggesting machine learning enhances risk stratification accuracy.
Key Points
Gradient boosting classifier achieved the highest AUC of 0.8532, indicating superior predictive performance over traditional methods.
Analysis included 2,213 ICU patients, with a notable 15.6% in-hospital mortality rate observed.
Retrospective cohort study based on MIMIC-IV data employed the Boruta algorithm for significant feature selection.
Machine learning models, particularly with SHAP values, provide deeper insights into mortality risk prediction.