Survival prediction for heart failure complicated by sepsis: based on machine learning methods

Area of Science:

• Cardiology
• Critical Care Medicine
• Medical Informatics

Background:

• Heart failure and sepsis significantly increase mortality risk.
• Early identification of high-risk patients is crucial for resource allocation.

Purpose of the Study:

• To develop and validate a survival prediction model for patients with comorbid heart failure and sepsis.
• To identify key predictors of 28-day all-cause mortality in this patient population.

Main Methods:

• Utilized eICU-CRD database for model development (3891 patients) and MIMIC-IV database for external validation (2928 patients).
• Employed Boruta method for feature selection and XGBoost for feature ranking.
• Compared Logistic Regression (LR), XGBoost, AdaBoost, and Gaussian Naive Bayes (GNB) models.
• Assessed model performance using AUC, accuracy, sensitivity, and specificity; utilized SHAP for interpretability.

Main Results:

• The LR model demonstrated superior performance with a validation AUC of 0.746.
• Key predictors included age, ventilation, norepinephrine, white blood cell count, total bilirubin, temperature, phenylephrine, respiratory rate, neutrophil count, and systolic blood pressure.
• External validation using MIMIC-IV yielded an AUC of 0.699.

Conclusions:

• A robust LR-based model accurately predicts 28-day mortality in heart failure patients with sepsis.
• The model provides a valuable tool for clinicians to identify high-risk patients and guide clinical practice.