The growth and evolution of clinical terminologies: Comparison of SNOMED, LOINC and RxNorm: Have they come of age?
Thursday, September 15, 2016 — Poster Session II
- V Huser
The research use of healthcare Big Data resources, similar to Center for Medicare and Medicaid’s Virtual Research Data Center, Truven MarketScan Commercial Claims database, Optum Clinformatics database, or Food and Drug Administration Sentinel Platform, is rapidly increasing. An important component of such large databases is a terminology layer that allows semantic data integration (for example, the terminology infrastructure embedded in the Common Data Model (CDM) defined by the Observational Health Data Sciences and Informatics (OHDSI) initiative. We analyzed three terminologies: SNOMED (diagnoses), RxNorm (drugs), and LOINC (laboratory observations). With longitudinal clinical data spanning decades (1994 till 2015), evolution of terminologies is a significant challenge for any Big Data resource (addition of new concepts and retiring erroneous, ambiguous or obsolete concepts). Our poster will present data on the (1) growth of terms in the analyzed terminologies (eg, SNOMED added 139,893 terms between 2010 and 2015; a 23% increase); (2) rate and temporal evolution of concept inactivation (eg, RxNorm has relatively high inactivation ratio; in 2015 it had one obsolete concept per every 3.78 active concepts); and (3) analysis of where terminologies evolved the most (additions and inactivation by terminology domain). All studied terminologies had 10+ years to evolve (SNOMED since 1999, RxNorm since 2005, LOINC since 1995) and concept clarifications/additions should have theoretically improve their usefulness for data capture and analysis (is there decreasing trend in concept inactivation/addition?). Our poster will show graphical temporal and other visualizations that answer the question whether terminologies have come of age.