Case-based retrieval of similar diabetic patients

Patients suffering from diabetes often develop several comorbidities such as hypertension and dyslipidemia. The presence of the comorbidities leads to more complex patient profiles associated with specific patient treatments. In this paper we present a novel algorithm to help physicians, given a new case, in retrieving similar past patient cases. This novel algorithm is based on the bag-of-words (BoW) model to encode as features, the occurrence of each pre-computed cluster, for each patient, according to the approach of document classification. We then evaluate the algorithm on a real de-identified dataset of 3201 diabetic patients, demonstrating the advantage of our approach.