Semantic interoperability and Snomed CT: A Case Study in clinical problem lists.

CAMPBELL, JAMES R; LÓPEZ OSORNIO, A; GONZÁLEZ, B. DE QUIRÓS F.; LUNA, DANIEL R.; REYNOSO, GUILLERMO

Congreso; 12th World Congress on Health Informatics MEDINFO 2007; 2007

Sharing clinical data between institutions requires a shared semantic model and reproducible metrics for comparing post-coordinated clinical record data. We evaluated the interoperable features of SNOMED CT by merging and comparing SNOMED problem list dictionaries between primary care sites in Nebraska and Argentina. The two problem subsets had modest differences in semantics, but differed substantially in frequency of postcoordinated content. 64.3% of the final merged problem vocabulary was post-coordinated. 0.9% of the merged concept set was found to have semantic duplication. Classification employing SNOMED Normal Forms was found to effectively identify semantic equivalency in 65.2% of cases studied. The most frequent reasons for failure of the classification procedures included: post-coordination not compliant with SNOMED guidelines (28.8% of cases), conflicting guidance presented in SNOMED precoordinated content (3.6%), unclear definition of source clinical concepts (1.2%), and limited expressiveness of the SNOMED definitional model (1.2%). We conclude that interoperable post-coordination employing SNOMED CT is feasible, but that better dissemination of SNOMED editorial guidelines, refinement of the concept model and training of the user community will be required for reproducible and reliable implementation.