Machine Learning-Driven Prediction of One-Year Readmission in HFrEF Patients: The Key Role of Inflammation.

Abstract

Background: Heart failure with reduced ejection fraction (HFrEF) is a global health issue with high morbidity and frequent hospitalizations. Predicting one-year readmission risk is crucial for optimizing treatment and reducing costs.

Methods: We conducted a single-center retrospective study on adult HFrEF patients admitted to the Cardiovascular Department of the First Affiliated Hospital, Zhejiang University School of Medicine on January 2020 and March 2023. Feature selection was performed using LASSO regression, with inflammatory biomarkers (PLR, MLR, NLR, SII, SIRI) prioritized. Seven machine learning (ML) algorithms were trained and validated using a 7:3 dataset split; the metrics of the model included the area under the curve (AUC), accuracy, sensitivity, specificity, F1 score, and Brier score. SHapley Additive exPlanations (SHAP) analysis provided model interpretability. A network-based dynamic nomogram was developed to visualize predictive models.

Results: This study included 733 patients, of whom 231 (31.5%) were readmitted within one year. LASSO regression showed that the key predictors included age, BNP, New York Heart Association (NYHA) class, LVEF, PLR, MLR, AF history, and ACEI/ARB/ARNI usage. The Random Forest (RF) model performed best, with an AUC of 0.89 (95% confidence interval (CI): 0.86-0.93), an accuracy of 0.83, a sensitivity of 0.87, and a specificity of 0.80. SHAP analysis showed that BNP was the most influential feature, followed by NYHA class and LVEF, which were also important predictors. In addition, MLR and PLR also played an important role in prediction, once again confirming the important predictive role of MLR and PLR as inflammatory indicators for readmission within one year in HFrEF patients.

Conclusion: The ML-based RF model effectively predicted one-year readmission in HFrEF patients, with inflammation indicators playing an important role. Integrating such models into clinical practice could improve risk stratification, reduce readmissions, and enhancing patient outcomes.