Emergency department readmission risk prediction: A case study in Chile

Abstract

Short time readmission prediction in Emergency Depart-ments (ED) is a valuable tool to improve both the ED managementand the healthcare quality. It helps identifying patients requiring fur-ther post-discharge attention as well as reducing healthcare costs. As inmany other medical domains, patient readmission data is heavily imbal-anced, i.e. the minority class is very infrequent, which is a challenge forthe construction of accurate predictors using machine learning tools. Wehave carried computational experiments on a dataset composed of EDadmission records spanning more than 100000 patients in 3 years, witha highly imbalanced distribution. We employed various approaches fordealing with this highly imbalanced dataset in combination with differ-ent classification algorithms and compared their predictive power for theestimation of the ED readmission probability within 72 h after discharge.Results show that random undersampling and Bagging (RUSBagging) incombination with Random Forest achieves the best results in terms ofArea Under ROC Curve (AUC).