Implications of mappings between International Classification of Diseases clinical diagnosis codes and Human Phenotype Ontology terms

We present a thorough empirical analysis of the compatibility between international classification of diseases (ICD) codes and human phenotype ontology (HPO) terms based on the unified medical language system (UMLS) Metathesaurus. ICD is used to annotate clinical diagnoses in EHR data, while HPO is used to annotate phenotypes in research databases. Bridging between the 2 artifacts is essential for health data integration and analysis. UMLS is a widely used source of cross-ontology mappings, and so it is important to quantitatively assess the extent to which ICD is mapped to HPO in the UMLS. The primary results from the paper include that a mere 2.2% of ICD codes in UMLS are directly linked to HPO. Furthermore, an analysis of our EHR dataset shows that less than half of the commonly used ICD codes can be mapped to HPO terms. Notably, commonly used ICD codes in EHR data tend to have corresponding mappings to HPO. In contrast, ICD codes representing rarer medical conditions are infrequently associated with HPO terms.