Uploaded image for project: 'FHIR Specification Feedback'
  1. FHIR Specification Feedback
  2. FHIR-31621

Inconsistent usage of OIDs/URIs overview

    XMLWordPrintableJSON

Details

    • Icon: Change Request Change Request
    • Resolution: Unresolved
    • Icon: Medium Medium
    • FHIR Core (FHIR)
    • R5
    • Terminology Services Management Group
    • Terminologies - Systems

    Description

      HL7 has an OID Registry which is absolutely vital to V3 (CDA). As of FHIR, HL7 also has URIs, in many cases for the same things that OIDs already existed for. I've argued many times against readable alternative identification for the same thing: that is not what this issue is about. This issue about on par, in tandem governance for uris and oids.

      I'm noticing that terminology is sort of centralized into THO/UTG. Identifier maintenance also needs a proper home.

      While the OID Registry has been meticulously maintained and OIDs have a tendency to be stable, the URIs more or less have a tendency to be created in an ad hoc and decentralized way, and also to be changed over time.

      This means that many URIs exist for the same purpose. That in itself would be ok-ish, but it is in many cases impossible tell which is leading for a given version of FHIR, unless you have a reliable NamingSystem setup that marks one and only uri/oid as the preferred one for a given version of FHIR, and use that consistently in the FHIR Core specs and examples.

      It is not possible to switch preferred uri/oid mid-FHIR-version as that constitutes a breaking change in implementations. It is also not possible to mix and match different uris for the same thing within the same version of FHIR. It should also not be possible to change the preferred oid/uri in a patch or minor update as - again - it is a breaking change.

      HL7 has said to scale down V3 support to only supporting IG creation. The only way that holds is if OIDs and URIs bear a meaningful and computable relationship. That also means that new code system uris also SHALL have an OID assigned, e.g. ICD-11 and IDMP. They otherwise are unavailable for CDA and the world is not ready for full scale FHIR conversion.

      So the administration of OIDs and URIs should be done centrally/in tandem and not in ad hoc/per working group fashion. This is a governance issue. To support why that is necessary I've done some analysis on what the relationship currently looks like.

      You'll notice that in many cases the OID for the ValueSet is the same as for CodeSystem, and the same OID has been assigned to both a V2 and V3 ValueSet, i.e. different objects.

       

      Attachments

        Activity

          People

            Unassigned Unassigned
            ahenket Alexander Henket
            Watchers:
            5 Start watching this issue

            Dates

              Created:
              Updated: