gms | German Medical Science

Introduction: Coding medical diagnoses and procedures in SNOMED CT allows an unparalleled diversity and granularity [1] in describing medical data and can improve data quality [2]. However mapping a SNOMED CT code to each diagnosis within a large research hospital is an extremely time consuming task requiring input and expertise from clinicians and informaticians. While directly attaching SNOMED CT codes to diagnoses can remain an optimal long-term goal, an intermediate solution is to translate the already used ICD-10-GM codes to SNOMED CT. A limitation on any translation is that ICD-10-GM has generally a lower level of granularity than SNOMED CT, therefore the complete scope of SNOMED CT cannot be utilized and additionally a difference in granularity can cause problems for translation [3]. Mapping between the related code system ICD-10 and SNOMED CT has already been tested, [4] however translation or mapping of ICD-10-GM to SNOMED CT has received less attention. Here we test a method to translate ICD-10-GM to SNOMED CT to evaluate its accuracy.

Methods: The ICD-10-GM codes are sent in a GET REST call to a ID Berlin terminology server using the translate function. The translate function uses the Wingert code system [5], [6] as a bridge between the ICD-10-GM and the SNOMED CT codes and finds the associated SNOMED CT code and sends back a FHIR parameter resource containing the corresponding SNOMED CT code in a valueCoding element.

Results: As test cases, we sent two sets of ICD-10-GM codes to the terminology server. The first set contains 27 ICD-10-GM codes from a value set used in a recent use case that describe sepsis and endocarditis diagnoses. The second set contains nine ICD-10-GM codes involving pneumonia caused by bacteria. The majority of the codes were correctly translated to SNOMED CT, but a sizeable minority was incorrectly translated. We next examined the granularity in a subset of these two sets between ICD-10-GM and SNOMED CT. In all cases, as expected, the SNOMED CT code had additional child codes underneath the translated code indicating that the full scope of SNOMED CT will not be exploited.

Discussion: We tested the ability of a terminology server algorithm to translate ICD-10-GM codes into SNOMED CT codes. From the 38 translations, there were enough incorrect results to indicate that the algorithm needs further improvements. The errors occurred in ways suggesting that some errors are due to structural differences between the code systems. The use of SNOMED CT would enable more expansive analysis of medical data due to its specificity [7] and hierarchies that in turn allow additional more specific filtering of data. This benefit however will be moderated if the data is translated from ICD-10-GM and not directly coded into SNOMED CT at the point of data entry.

Conclusion: Translating already existing ICD-10-GM codes into SNOMED CT can be a viable intermediate-term solution to take advantage of SNOMED CT if the translations accuracy can be improved.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.