Annotating medical forms using UMLS

Abstract

Medical forms are frequently used to document patient data or to collect relevant data for clinical trials. It is crucial to harmonize medical forms in order to improve interoperability and data integration between medical applications. Here we propose a (semi-) automatic annotation of medical forms with concepts of the Unified Medical Language System (UMLS). Our annotation workflow encompasses a novel semantic blocking, sophisticated match techniques and post-processing steps to select reasonable annotations. We evaluate our methods based on reference mappings between medical forms and UMLS, and further manually validate the recommended annotations.