The Effectiveness of Random Forest Models in Predicting Readmission Risk: An EHR-Based Approach

Abstract

Risk of hospital readmission is one of the crucial tasks in the sphere of healthcare management that directly impact patient outcomes and resource distribution. As EHR data is increasing, machine learning models, like Random Forests, have become particularly popular when it comes to readmission risk prediction. In the study, the authors discuss Random Forest models in the framework of the prediction of readmission risk among patients though EHR data. It aims to consider and compare the predictive capacity of the Random Forest model with a few other machine learning models, i.e., the common logistic regression and the decision tree models. The population size of data of the current study is formed by patient demographic, medical history and treatment data of various hospitals. Such measures as accuracy, precision, recall, and F1-score are used to analyze the performance of the model. The findings show that the Random Forest model is more effective than traditional methods and it has a 85 percent level of accurately predicting readmission risk. Results exhibit the possibility of Random Forests in healthcare and provide a good insight to hospitals trying to enhance the management of patients. Conclusions and suggestions are carried in the conclusion at the end of the research concerning the implication of the research and research in the future in the field of EHR-based predictive modeling.