OHDSI 2022 Symposium

At this year’s Observational Health Data Science and Informatics (OHDSI) conference in Maryland, Yen Low of Atropos Health is presenting new research: ​​Towards Lexical, Semantic and Similarity Search in Phenotype Libraries (Tekumalla et al 2022). The abstract describes the team’s creation of a phenotype library with lexical, semantic, and similarity-based search.

The authors aggregated more than 600 concept sets, or clinical definitions collecting medical concepts, from OHDSI’s Phenotype Library2 and translated them to concepts in RxNorm, ATC, CPT,  HCPCS, LOINC, ICD9 and ICD 10 using OMOP 5.4. Further methods included extraction of keywords and metadata annotation (lexical and semantic search) and cosine similarity (similarity search). 

The largest number of similar concepts was generated by similarity search. Cosine similarity captured more clinically-related concepts, and concepts with erroneous naming or tagging, than either lexical or semantic search.

For more information, view the full abstract on OHDSI’s website here or meet with the authors at OHDSI 2022. You may also contact info@atroposhealth.com to learn more about Atropos Health’s methods and current to future directions for research.