Creating Consistent Diagnoses List for Developmental Disorders Using UMLS

Abstract

In the field of developmental disorders, there is no commonly accepted medical vocabulary. Vocabularies, such as ICD-10, are unsatisfying to clinicians, who try to create their own diagnostic lists. This results in inconsistency in the terms used in clinical practice. When attempting to apply automatic computational methods on patients’ data, the need for common consistent diagnoses list arises. To this end, we mapped a set of different diagnoses used in clinical practice to UMLS —a well-known Unified Medical Language System that organizes and unifies over 100 vocabularies. Diagnoses that were defined by different terms and were mapped to one concept at UMLS were joined as synonyms. Concepts that were not found at UMLS were mapped to the closest concept found. We found that SNOMED-CT is the most comprehensive vocabulary (85.7% term coverage) in UMLS. We propose a framework that, when applied on inconsistent manually constructed set of diagnoses, leads to minimal and consistent set of diagnostic terms.