TITLE:
Construction of Volume Overload Risk Prediction Model for Heart Failure Patients Based on Machine Learning
AUTHORS:
Juan Zhou, Jing Zhang
KEYWORDS:
Heart Failure, Volume Overload, Machine Learning, Predictive Model
JOURNAL NAME:
Journal of Biosciences and Medicines,
Vol.14 No.7,
July
15,
2026
ABSTRACT: Objective: To construct six machine learning models for predicting volume overload in patients with chronic heart failure (CHF) and identify the optimal predictive model to provide scientific evidence for early and accurate identification of high-risk populations with volume overload in CHF patients. Methods: A questionnaire on factors contributing to volume overload in CHF patients was developed through literature review and expert consultation. Using convenience sampling, retrospective data from 2477 hospitalized CHF patients admitted to the Department of Cardiology at a tertiary Grade A hospital in Jingzhou City from January 2023 to December 2024 were collected and randomly divided into a training set and a validation set in a 7:3 ratio. Variable screening was first performed on the training set to obtain a feature subset, followed by model construction using six machine learning algorithms: logistic regression, random forest, decision tree, eXtreme Gradient Boosting (XGBoost), support vector machine, and light gradient boosting machine (LightGBM). The area under the receiver operating characteristic curve (AUC) was compared to select the optimal model. Results: Among all samples, 1411 cases were diagnosed with volume overload, yielding a positive rate of 57%. Comparative analysis revealed that random forest and LightGBM demonstrated the highest AUC value (0.835) and optimal stability, indicating their strongest discriminative capacity. The XGBoost model (AUC = 0.828) followed closely, serving as a high-performance alternative model. A visual column-line chart was constructed based on a Logistic regression model to transform prediction results into an easily usable bedside scoring tool for clinicians. The model demonstrated performance comparable to the optimal machine learning model (AUC = 0.835), combining predictive accuracy with clinical interpretability. Conclusion: This single-center retrospective study confirms that machine learning algorithms, particularly random forest and LightGBM, can identify key features of volume overload in CHF patients, demonstrating preliminary clinical translation potential. The findings provide a practical intelligent tool for early identification and intervention of high-risk populations with CHF volume overload.