A Machine Learning-Based Predictive Model for One-Year All-Cause Readmission in Chronic Heart Failure: Implications for Precision Nursing Interventions

This study aims to develop and validate a machine learning model for predicting 1-year all-cause readmission in patients with chronic heart failure (CHF), and to assess its clinical utility in early risk stratification and targeted nursing interventions.

This single-center retrospective study included 293 patients with chronic heart failure (CHF) who were consecutively admitted to Affiliated Hospital of Nantong University between January 2023 and May 2024. Participants were randomly allocated into a training cohort (n = 205) and a validation cohort (n = 88). Multidimensional clinical variables were extracted from the electronic medical record system of the hospital. Feature selection was performed using Least Absolute Shrinkage and Selection Operator (LASSO) regression (with 10-fold cross-validation and the minimum lambda plus one standard deviation criterion to select λ to prevent overfitting). Five machine learning models, including Logistic Regression, Support Vector Machine, ElasticNet, XGBoost, and Gradient Boosting Decision Trees (GBDT), were developed, and their predictive performance was compared. The hyperparameters of the model were optimized with grid search combined with 10-fold cross-validation. Model performance was assessed using the area under the Receiver Operating Characteristic (ROC) curve (AUC), sensitivity, specificity, precision, recall, and F1 score. Finally, SHapley Additive exPlanations (SHAP) were applied to interpret the optimal model.

The GBDT-based model demonstrated the strongest predictive performance, achieving an AUC of 0.969 (95% confidence interval (CI) 0.950-0.988), a sensitivity of 0.896, and a specificity of 0.922 in the internal validation set. SHAP analysis identified N-terminal pro-B-type natriuretic peptide (NT-proBNP), New York Heart Association (NYHA) class, respiratory infection, hyponatremia, and estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m² as the top five predictors associated with readmission risk.

The GBDT-based machine learning model effectively predicts 1-year all-cause readmission in CHF patients, demonstrating superior discriminative accuracy and interpretability via SHAP analysis, offering a clinically practical tool for management. Given the single-center retrospective design and limited sample size, future validation in larger multicenter studies is warranted before broader clinical application.