Mapping the categories of the Swedish primary health care version of ICD-10 to SNOMED CT concepts: rule development and intercoder reliability in a mapping trial

Abstract

Background: Terminologies and classifications are used for different purposes and have different structures and content. Linking or mapping terminologies and classifications has been pointed out as a possible way to achieve various aims as well as to attain additional advantages in describing and documenting health care data. The objectives of this study were: to explore and develop rules to be used in a mapping process, to evaluate intercoder reliability and the assessed degree of concordance when the 'Swedish primary health care version of the International Classification of Diseases version 10' (ICD-10) is matched to the Systematized Nomenclature of Medicine, Clinical Terms (SNOMED CT), to describe characteristics in the coding systems that are related to obstacles to high quality mapping.

Methods: Mapping (interpretation, matching, assessment and rule development) was done by two coders. The Swedish primary health care version of ICD-10 with 972 codes was randomly divided into an allotment of three sets of categories, used in three mapping sequences, A, B and C. Mapping was done independently by the coders and new rules were developed between the sequences. Intercoder reliability was measured by comparing the results after each set. The extent of matching was assessed as either 'partly' or 'completely concordant'

Results: General principles for mapping were outlined before the first sequence, A. New mapping rules had significant impact on the results between sequences A-B (p < 0.01) and A-C (p < 0.001). The intercoder reliability in our study reached 83%. Obstacles to high quality mapping were mainly a lack of agreement by the coders due to structural and content factors in SNOMED CT and in the current ICD-10 version. The predominant reasons for this were difficulties in interpreting the meaning of the categories in the current ICD-10 version, and the presence of many related concepts in SNOMED CT.

Conclusion: Mapping from ICD-10-categories to SNOMED CT needs clear and extensive rules. It is possible to reach high intercoder reliability in mapping from ICD-10-categories to SNOMED CT. However, several obstacles to high quality mapping remain due to structure and content characteristics in both coding systems.

Figure 1

Equally and not equally chosen concepts matched to categories in different chapters of ICD-10. Chapter 21 with the Z-codes had a 23% rate of equally chosen concepts. An equally chosen category = when the two coders matched the same SNOMED CT concept to one KSH97-P category.