Recurrent Neural Networks with Non-Sequential Data to Predict Hospital Readmission of Diabetic Patients

Abstract

Hospital readmissions are recognized as indicators of poor quality of care, such as inadequate discharge planning and care coordination. Moreover, most experts believe that many readmissions are unnecessary and avoidable. In the present paper, we design a Recurrent Neural Network model to predict whether a patient would be readmitted in the hospital and compared its accuracy with basic classifiers such as SVM, Random Forest and with Simple Neural Networks. RNN showed highest prediction power in all the models used and thus this can be used by hospitals to target high risk patients and prevent recurrent admissions.