SNOMED representation of explanatory knowledge in veterinary clinical pathology

• Published in: Veterinary clinical pathology 2005-01, Vol.34 (1), p.7-16
• Main Authors: Zimmerman, K.L, Wilcke, J.R, Robertson, J.L, Feldman, B.F, Kaur, T, Rees, L.R, Spackman, K.A
• Format: Article
• Language: English
• Subjects: animal pathology; Clinical pathology; communication (human); communications technology; controlled terminology; expert systems; guidelines; Information Storage and Retrieval; knowledge representation; nomenclature; Pathology, Veterinary; SNOMED; SNOMED nomenclature; Systematized Nomenclature of Medicine; Systemized Nomenclature of Medicine; terminology
• ISSN: 0275-6382; 1939-165X
• DOI: 10.1111/j.1939-165X.2005.tb00002.x

Background: The Systematized Nomenclature of Medicine (SNOMED) is an established standard nomenclature for the expression of human and veterinary medical concepts. Nomenclature standards ease sharing of medical information, create common points of understanding, and improve data aggregation and analysis. Objectives: The objective of this study was to determine whether SNOMED adequately represented concepts relevant to veterinary clinical pathology. Methods: Concepts were isolated from 3 different types of clinical pathology documents: 1) a textbook (Textbook), 2) the Results sections of industry pathology reports (Findings), and Discussion sections from industry pathology reports (Discussion). Concepts were matched (mapped) by 2 reviewers to semantically-equivalent SNOMED concepts. A quality score of 3 (good match), 2 (problem match), or 1 (no match) was recorded along with the SNOMED hierarchical location of each mapped concept. Results were analyzed using Cohen's Kappa statistic to assess reviewer agreement and chi-square tests to evaluate association between document type and quality score. Results: The percentage of good matches was 48.3% for the Textbook, 45.4% for Findings, and 47.5% for Discussion documents, with no significant difference among documents. Of remaining concepts, 40% were partially expressed by SNOMED and 14% did not match. Mean reviewer agreement on quality score assignments was 76.8%. Conclusions: Although SNOMED representation of veterinary clinical pathology content was limited, missing and problem concepts were confined to a relatively small area of terminology. This limitation should be addressed in revisions of SNOMED to optimize SNOMED for veterinary clinical pathology applications.