University of Ghent (1 of 2)

SNOClass project, commanded by Belgium National Release Center, aims at creating maps from SNOMED CT terminology to three classifications: International Classification of Functioning, Disability and Health (ICF), International Classification of Health Interventions (ICHI) and International Classification of Primary Care - 3rd Revision (ICPC-3).

The mapping process will be semi-automated, relying on lexical and ontological analysis of both the terminology and the classifications.

The suggested mappings will be evaluated by experts. This methodology step is resource and knowledge demanding and may not be compatible with the availability of experts.

We present a method based on SNOMED CT intrinsic features to reduce the very large number of concepts from SNOMED CT to consider, while avoiding missed mapping opportunities.

SNOMED CT has 19 high level concepts (synonym: domains) and 58 hierarchy tags. Except for (physical object) and (product), the hierarchy tags are associated only with one domain. For each of the 58 hierarchy tags we extracted the number of SNOMED CT codes (see table in Annex). Clinical domain experts chose for each target classification which SNOMED CT hierarchy tags are relevant.

The resulting number of concepts may still be too high, and additional reduction strategies are considered such as starting from existing collections of frequently used lexical entries, existing mappings, and existing core sets to identify a relevant domain-specific sample of SNOMED CT concepts that can be adequately handled (mapped to the domain classification) within the constraints of a one year project, but still be relevant for testing in specific use cases.