Comparing SNOMED and ICPC retrieval accuracies using relational database models.

Abstract

While SNOMED International has been generally accepted by the international community of pathologists, its use for primary and secondary care remains limited. This can probably be attributed to the coding complexity of clinical concepts into this multiaxial postcoordinated nomenclature. The SNOMED editors propose the use of multiple codes (aggregates) for any nuanced clinical concept, thus allowing alternative rigorous representations of the concept with SNOMED codes. Some classification critics argue whether such redundant coding precludes precise retrieval of data. This research was initiated to compare the retrieval accuracies of a relational database using a simplified model of SNOMED against a classification-based model. SNOMED-based queries showed improvement over ICPC-based queries, regardless of the use of SNOMED cross-references. The addition of the latter significantly improved the queries sensitivity and false negative rate. In conclusion, the authors recommend using aggregates of SNOMED codes in relational database designs over classification-based designs in order to improve retrieval accuracy.