Enhancing Clinical Decision-Making: Predicting Readmission Risk in Heart Failure Patients with Machine Learning (Preprint)

BACKGROUND Heart failure sufferers frequently face the likelihood of rehospitalization following an initial stay. Although several risk assessment tools have been developed to aid in clinical decision-making, clinicians and patients' families continue to encounter significant challenges in determining shared decision-making regarding maintenance versus hospice transition. To address these challenges, it is crucial to devise more precise predictive tools, which can assist medical professionals and relatives in the care management of individuals with heart failure. Currently, effective predictive models for the Chinese heart failure patient population are still insufficient, which limits our ability to accurately assess patient prognosis and implement personalized treatment strategies. OBJECTIVE The present study has formulated a predictive model for assessing the likelihood of readmission among patients diagnosed with heart failure. METHODS In this study, we analyzed data from 1,948 heart failure patients in a hospital in Sichuan Province between 2016 and 2019. By applying three variable selection strategies, 29 relevant variables were identified. Subsequently, we constructed six predictive models using different algorithms: logistic regression (LR), support vector machine (SVM), gradient boosting machine (GBM), extreme gradient boosting (XGBoost), multilayer perceptron (MLP), and graph convolutional network (GCN). RESULTS The graph convolutional network model showed the highest prediction accuracy with an area under the curve (AUC) of 0.866, accuracy of 77.6%, sensitivity of 58.54%, and specificity of 90.42%. CONCLUSIONS The model crafted in this study proves its effectiveness in forecasting the likelihood of readmission among heart failure patients, thus serving as a crucial reference for clinical decision-making.