Details
-
Change Request
-
Resolution: Persuasive with Modification
-
Medium
-
FHIR Core (FHIR)
-
R5
-
Terminology Infrastructure
-
ConceptMap
-
-
Grahame Grieve/Michael Lawley: 6-0-0
-
Enhancement
-
Non-compatible
-
R5
Description
Task FHIR-39151 proposed to normalise ConceptMap property definitions to be like the way that the CodeSystem properties are defined. That was approved by TI, but I reopened it because they are not the same thing: the existing 'properties' elements which the task proposed to change are specifically defined to be devoted to wrangling between data elements of different scope:
"this is a product of the mapping that goes in the output data set somewhere"
But Michael Lawley has a different use case:
"something that helps advise which mapping to choose"
E.g. a property of the mapping itself, which is what task 39151 was trying to provide
This is a legitimate use case, but not one we'd considered outside task 39151 - at least, not formally to my knowledge. Michael's just been using the existing product property since that's what's was there, though it's not a proper use.
There's even got an example in the spec: https://github.com/HL7/fhir/blob/master/source/conceptmap/conceptmap-example-priority.xml. But that example has the same problem: it's using output products but actually they contain properties of the mapping.
So after talking about it, Michael and I propose that we:
1. rename the existing 'property' to 'mapAttribute' in both the root and on ConceptMap.group.element.target.dependsOn and ConceptMap.group.element.target.product
2. clone the CodeSystem property framework (as originally proposed in the task) and add ConceptMap.group.element.target.property for the things Michael has a need for. (and make it clear that these are properties of the mapping, not the underlying concepts)
3. Add new parameters to $translate for properties in and out and clarify the other parameters
4. while we're at it, adopt the disposition to FHIR-38171 for some $translate parameters too (38171 proposed the adoption of FHIR search semantics for value set operations and code system operations that deal with properties, so it's natural to apply the same considerations in ConceptMap)
Since this very definitely has to be committed to the spec by the end of this week, it appears that this is procedurally only possible if the cochairs approve it directly, though it would be good to get wider endorsement here
The obvious concern that this is a pretty big change to be making right now at this point in time. I think it's justified because:
- We have in effect adopted the requirements through the task and the example
- I did convince the committee to walk that adoption back but that was on internal consistency grounds, not an issue with the actual use case
- The properties framework in CodeSystem is well understood and normative, so adding it to ConceptMap isn't a big leap
- We're making other breaking changes to ConceptMap now anyway, so this is the time to do it
- it's evident that calling what we have now 'property' is confusing for people, and has caused confusion in the resolution of the task, and we really don't want to fix it later
Attachments
Issue Links
- relates to
-
FHIR-39151 Support for ConceptMap property definitions and standard properties
- Published
-
FHIR-40544 Update language around ConceptMap properties and additionalAttributes
- Published