Introduce an attribution model for FHIR - 2018-Jan Core #52

XMLWordPrintableJSON

    • Type: Change Request
    • Resolution: Not Persuasive
    • Priority: Medium
    • FHIR Core (FHIR)
    • STU3
    • Modeling & Methodology
    • (methodology)
    • Hide

      FHIR has instituted patterns such as the FiveWs, Request, Event and Definition patterns to encourage consistency in naming. However, domain requirements and alignment with domain terminology trump the need for consistency. Sophisticated systems that want to do cross-resource analysis are better able to handle variations of resources than simple implmenters who know only their domain terminology can cope with the learning curve of model-driven naming. Consistency is managed through mapping to patterns and reference models, not through imposing consistency of names and models that do not align with implementer practice.

      Feedback on the patterns and how mappings are represented that would make them more useful to CIMI and others is welcome. We are also happy to cross map and work on alignment between the existing FHIR patterns and the CIMI attribution model.

      Show
      FHIR has instituted patterns such as the FiveWs, Request, Event and Definition patterns to encourage consistency in naming. However, domain requirements and alignment with domain terminology trump the need for consistency. Sophisticated systems that want to do cross-resource analysis are better able to handle variations of resources than simple implmenters who know only their domain terminology can cope with the learning curve of model-driven naming. Consistency is managed through mapping to patterns and reference models, not through imposing consistency of names and models that do not align with implementer practice. Feedback on the patterns and how mappings are represented that would make them more useful to CIMI and others is welcome. We are also happy to cross map and work on alignment between the existing FHIR patterns and the CIMI attribution model.
    • Grahame Grieve/Rob Hausam: 4-0-0
    • Clarification

      Comment:

      In FHIR, attribution information varies significantly across resources. For instance, in Procedure, when an action was performed is represented as performed[x]. For medication administration, the performance time is represented as effective[x]. In communication, it is represented using the attribute 'sent'. Information about the performer is also represented fairly heterogeneously. For instance, in Procedure it is captured with the structure Performer which contains role, actor, and on behalf of. In MedicationAdministration, there is also a performer structure but it lacks role. In Communication, there is a 'sender' even though it is also an action which could be performed by a role, onBehalfOf. In Procedure, there is a location field which is missing for MedicationAdministration and Communication. In order to ensure better cross-resource consistency, I would strongly recommend that FHIR consider an attribution model or discusses the attribution model currently considered by the CIMI group and align on a consistent approach for representing this information. This lack of consistency will impose a cost on implementers who will need to define different transformations for each resource when such variability does not confer significant benefits.

      Summary:

      Introduce an attribution model for FHIR

            Assignee:
            Unassigned
            Reporter:
            Claude Nanjo (Inactive)
            Claude Nanjo (Inactive)
            Watchers:
            2 Start watching this issue

              Created:
              Updated:
              Resolved: