Comparison of predictive models for hospital readmission of heart failure patients with cost-sensitive approach
Junar Arciete Landicho, Vatcharaporn Esichaikul, Roy Magdugo Sasil
Abstract
Readmission rates for heart failure patients remain high but it is potentially preventable. Many predictive models have been developed over the years to identify heart failure patients who are at high risk of readmission but only a few of them incorporate cost considerations. The goal of this study is to compare the performance of four machine learning algorithms in predicting the readmission of heart failure patients with cost consideration. We also aim to identify the risk factors associated with a patient’s readmission within one year of a retrospective cohort study. The best model selection was found after four machine-learning methods were tested; these include logistic regression, support vector machine, random forest and neural network. The study found a support vector machine to have the best prediction performance with an AUC score of 0.602. The model showed twelve (12) predictors that are significantly associated with the identification of heart failure patients at high risk of readmission.