Explainable machine learning for postoperative respiratory failure prediction in open-heart surgery patients — a study based on the MIMIC-IV database

Key Points

• The study aims to enhance the predictability of postoperative respiratory failure using machine learning techniques.
• Utilized the MIMIC-IV database for data analysis
• Implemented a Gradient Boosting Machine (GBM) model for prediction
• Conducted SHAP analysis to interpret model outputs
• The GBM model achieved balanced performance in discrimination and calibration.
• SHAP analysis revealed critical hemodynamic and metabolic markers affecting PRF predictions.