Comparing clinical vocabularies using coding system fidelity

Much effort has been directed toward the development of an ideal multipurpose controlled medical vocabulary for use in human and veterinary medicine. SNOMED International is one effort that has resulted in a larger and more complex nomenclature system. Although it was able to code more concepts, SNOMED International failed to statistically improve vocabulary fidelity when compared with the 30+ year old SNVDO vocabulary. We found that SNOMED has a lower intercoder consistency than SNVDO and that a greater number of codes were necessary to represent an individual concept. Our study shows a significant Coder-Vocabulary interaction which suggests that more emphasis should be placed on coding guidelines and coder training. Clinician data entry and coding may be necessary for maximum vocabulary fidelity.