Explainable AI in Decision Support Systems : A Case Study: Predicting Hospital Readmission Within 30 Days of Discharge

Abstract

Explainable models are a critical requirement for predictive analytics applications in the healthcare domain. In this work we develop a hypothetical clinical decision support system for the classification task of predicting hospital readmission within 30 days of discharge. We compare a baseline logistic regression model with an implementation of the coordinate descent algorithm known as lasso. We choose lasso because it inherently performs variable selection during optimization which leads to an explainable model. Using model evaluation data we achieve an area under the ROC curve score of 0.795 improving on the baseline score of 0.683 without inflating the feature space.