Title: Prediction via clusters of CPT codes for improving surgical outcomes

Abstract: We develop a novel algorithm, Predictive Hierarchical Clustering (PHC), for agglomerative hierarchical clustering of current procedural terminology (CPT) codes, with the goal of finding clusters that improve the performance of a sparse logistic regression model for predicting surgical outcomes. The clustering scheme mimics traditional Hierarchical Clustering; however, our merge criterion is not based on a distance function and does not initialize with $n$ clusters. Our predictive hierarchical clustering aims to cluster subgroups, not individual observations, found within our data, such that the clusters discovered result in an improved performance of a classification model. Therefore, merges are chosen based on which pairings of the subgroups result in the largest improvement in prediction, as measured by the area under an ROC curve. The motivation is to predict patient-specific surgical outcomes using data from ACS NSQIP (American College of Surgeon's National Surgical Quality Improvement Program). An important predictor of surgical outcomes is the actual surgical procedure performed as described by a CPT code. We use PHC to cluster these subgroups together in a way that enables us to better predict patient-specific outcomes, instead of currently used clinically decided clusters. We present two different configurations of our algorithm, one incorporating the clusters of CPT codes as random slopes and the second as random intercepts in the classification model.