Stream: terminology
Topic: Terminology Capabilities and ordinal extension
Grahame Grieve (Aug 10 2020 at 21:33):
- if the ordinal value extension is found on either the underlying code system or in an include, it will be returned as part of any expansion
- value set ordinals override code system ordinals
- ordinal values will also be included in $lookup results, in the property 'ordinal-value'
Carol Macumber (Aug 11 2020 at 14:25):
I'm OK with this. But, @Rob Hausam should this be documented more formally in a FHIR ticket to add the guidance to https://www.hl7.org/fhir/terminologycapabilities.html, and therefore discussed at a FHIR Tracker call?
Rob Hausam (Aug 11 2020 at 16:02):
@Carol Macumber @Grahame Grieve Yes. I'm not sure that TerminologyCapabilities is the right (or only) place for this, but I agree that it needs a tracker. We could possibly discuss this in the initial 30 min. of the Tracker call this Thursday, before we need to end the call (60 min. early) to move to the Co-Chair Webinar.
Grahame Grieve (Aug 11 2020 at 17:24):
well, sure, we need a task, but I wanted to get a take on it before I created it. Why would TerminologyCapabilities not be the only or right place for this?
Alexander Henket (Mar 02 2021 at 08:41):
Was a task ever created for this? I'm now experiencing that $expand does not return ordinal defined in the CodeSystem using the extension, and that the ordinal in the ValueSet.compose goes poof as well.
The proposed guidance would be within my expectation and I'd love to vote Affirmative on it.
Michael Lawley (Mar 02 2021 at 08:44):
The ordinal extension issue came up in conversation for me today. I was left wondering why an extension is needed instead of a standard FHIR-defined property?
Alexander Henket (Mar 02 2021 at 08:45):
Well ... the extension exists and I used it comparably in the ValueSet and the CodeSystem.
Alexander Henket (Mar 02 2021 at 08:45):
Why use property in the one and extension in the other?
Alexander Henket (Mar 02 2021 at 08:50):
And as for property: I'd first need to define a code for it. Ordinal is something any assessment scale could have. I now seem to have total freedom in defining the property code, which could lead to "ordinal", "Ordinal", "ordinalValue", "gcs_ordinal" etc. etc. which is much less making use of standards than the predefined extension?
Michael Lawley (Mar 02 2021 at 08:54):
That's why I said "standard FHIR-defined property", and in fact we already have the requirement that $lookup return the value as the property named ordinal-value, so it would seem to make sense that we could just defined codes with that property and they could behave as per the extension.
As to why have two different ways -- the preferred source of ordinal value is from the code system definition, but the extension allows you to post-hoc add them in other places (even directly in a Coding or Questionnaire.item.answerOption) -- these do seem to be "extension" cases.
Alexander Henket (Mar 02 2021 at 08:56):
Sure, but ordinalValue extension explicitly has CodeSystem.concept as context: http://hl7.org/fhir/extension-ordinalvalue.html so should this extension then be deprecated in favor of property (with code ordinal-value)?
Michael Lawley (Mar 02 2021 at 08:58):
I would like to suggest so. In fact, I think I would suggest that it be treated as if such a property was defined.
I note the extension is draft status, maturity level 1, so on the surface this would seem to be reasonable.
Alexander Henket (Mar 02 2021 at 09:01):
Alright. I can update the CodeSystems like that. Is there any merit in keeping both extension and property? Would that help or hurt compatibility?
Michael Lawley (Mar 02 2021 at 09:02):
The thing that makes these special, though, is the fact that they should appear in expansions and can be overridden by a ValueSet-specified value.
I'm not entirely sure how they should manifest in the expansion since the extension doesn't list ValueSet.expansion.contains.code as context
Alexander Henket (Mar 02 2021 at 09:02):
It would still help if the core guidance on CodeSystem would be extended on ordinals and maybe more common scenarios, so pursuing the task would still help. Do I create one, or is there one already?
Michael Lawley (Mar 02 2021 at 09:03):
Agreed. And I suspect including both would help compatibility.
Alexander Henket (Mar 02 2021 at 09:05):
I'm not entirely sure how they should manifest in the expansion since the extension doesn't list ValueSet.expansion.contains.code as context
Yikes? I'll definitely have to create a task for that
Michael Lawley (Mar 02 2021 at 09:05):
I'm not sure where R5 is with $expand also returning properties, but that seems to me to be the natural approach going forward.
Michael Lawley (Mar 02 2021 at 09:07):
Lastly, I don't know what the use-case is for the ValueSet-defined ordinals, nor the overriding semantics. If it weren't for the overriding part, then I'd have suggested that a CodeSystem supplement would be a better way of providing ordinals from outside of the base CodeSystem.
Grahame Grieve (Mar 02 2021 at 09:56):
the extension was defined because people do define the ordinals in value sets or directly in questionnaires
Alexander Henket (Mar 02 2021 at 12:28):
FHIR#31393 Extension ordinalValue is missing context for ValueSet expansions
Colin E. (Mar 08 2022 at 12:32):
Back about a year ago (Mar 2021) I can see there was a fair bit of discussion around handling Ordinal value sets. I'm back looking at what we have referred to as Semi-quantitatlive results in the UK, which seem to be best represented as either Nominal values- where a standard FHIR Valueset or SNOMED Refset works well, or as members of Ordinal value sets.
Per @Michael Lawley 's comment below, I can provide a use case for valueset-defined ordinal values I think, but i'm guessing the conversation has moved on from a year ago.
So-
1) Have these discussions resulted in an agreed design for Ordinal valuesets?
2) Are the results applicable to older revisions of FHIR, or only relevant to a specific recent version (e.g. 4.6 or 5)?
I can still see references to the Ordinal-value extension for example in the v5 documentation, so it's not clear if this has been superseded by some more elegant solution built-in to core.
Michael Lawley said:
Lastly, I don't know what the use-case is for the ValueSet-defined ordinals, nor the overriding semantics. If it weren't for the overriding part, then I'd have suggested that a CodeSystem supplement would be a better way of providing ordinals from outside of the base CodeSystem.
Richard Townley-O'Neill (Mar 11 2022 at 04:33):
I'm also interested in this.
Last updated: Apr 12 2022 at 19:14 UTC
