Subgrouping Ulcerative Colitis Patients using Administrative Claims Data

Abstract

Approximately 3 million patients in the US have been diagnosed with Ulcerative Colitis, a chronic inflammatory disease affecting the colon. Uncovering patient subgroups could improve treatment guidelines and help physicians choose an appropriate treatment plan for a patient. Here, we outline a Python implementation to generate a cohort from a dataset in the OMOP Common Data Model (CDM), propose a patient timeline visualization tool, create and analyze a cohort of Ulcerative Colitis patients using a claims dataset. We extract patient features and use dimensionality reduction techniques along with clustering to identify patient subgroups. We observe four patient subgroups consisting of distinct patient characteristics, most prominently age, insurance type, sex, and type of initial conventional therapy prescription.