Assessing the Role of Machine Learning in Predicting Hospital Readmissions: A Study Using Real-World EHR Data

Abstract

This article assesses the role of machine learning in predicting hospital readmissions. One of the tasks that concern not only healthcare but also are considered very necessary is hospital readmission prediction that should be utilized to recognize vulnerable patients and ultimately improve the clinical outcomes. Machine learning (ML) models exhibit the potential to automate this process by taking data in the form of Electronic Health Records (EHR). The purpose of the study is to learn what the purpose of ML is in predicting hospital readmission and how the usage of real world EHR data can be applied so as to generate the predictive model. The research exploits various ML algorithms including the Random Forest (RF), Support Vector Machine (SVM), and the Neural Networks (NN) in the quest to interpret attributes on the basis of the patient demographics, medical history, and hospitalization records. It is indicated in the results that the model such as Random Forest of the ensemble method was more accurate and interpretable than the other models and possessed an AUC of 0.82. The trends suggest that one can utilise the ML models in clinical practices to foretell the readmissions at the initial phase. Nevertheless, quality of data, lack of interpretability of a model and restriction of generalizability towards diverse populations still exist. Future research ought to be focusing within the determination of upsurge in model openness and gains in the production of hybrid models that include the domain learning as well as clinical knowledge. The paper would contribute to the body of literature on articles on the AI in healthcare as it would demonstrate the operation of the ML techniques in reality in the story to predict the hospital readmission.