Stream: terminology
Topic: ValueSet with SNOMED codes in post-coordination
Katarina (Feb 04 2022 at 11:46):
Hi,
i want define a ValueSet with SNOMED-Codes. Some of the SNOMED-Codes should be in post-coordination. Is this possible? How would the codes look like?
My idea:
<ValueSet xmlns="http://hl7.org/fhir">
<compose>
<include>
<system value="http://snomed.info/sct" />
<concept>
<code value="281666001|family history of disorder|:246090004|associated finding|=22298006|myocardial infarction|" />
<display value="myocardial infarction in family history" />
</concept>
....
Is this correct?
Thank you,
Katarina
Mareike Przysucha (Feb 04 2022 at 11:57):
Duplicate of this thread: https://chat.fhir.org/#narrow/stream/179166-implementers/topic/ValueSet.20with.20SNOMED.20code.20in.20post-coordination.
Where should we answer?
Katarina (Feb 04 2022 at 12:00):
I tried to delete the other thread, but was not possible.
Katarina (Feb 04 2022 at 15:13):
@Mareike Przysucha Sorry, I tried to delete one of the duplicate questions, but was not possible. I appreciate any helpful information.
Mareike Przysucha (Feb 04 2022 at 15:57):
@Katarina : No problem. I just wanted to avoid different answers on different streams.
I think it's impossible to delete a thread, but I saw people changing the thread name or deleting their question, pointing to the other stream. Happens to all of us.
Grahame Grieve (Feb 04 2022 at 19:05):
I think that above is correct
Grahame Grieve (Feb 04 2022 at 19:05):
though we encourage not to include the term, just the concept id
Peter Jordan (Feb 04 2022 at 19:40):
Question removed - mistakenly thought you were talking about the concept.display element rather than excluding terms in the expression. :upside_down:
Grahame Grieve (Feb 04 2022 at 19:51):
because in FHIR, the code is purely for machine processing. The terms should go in the display not the code. So adding the term is just wasted bytes. But it's not wrong to include the term.
Lloyd McKenzie (Feb 05 2022 at 05:31):
It's "non-preferred" to include the term because it makes it hard for systems to do code-matching if they don't have a terminology server. A lot of systems will just do string-match to determine code equivalence - and if you include display terms in your codes, that software won't recognize them.
Grahame Grieve (Feb 05 2022 at 09:15):
well.... given the various possibilities for syntactical variation without semantic variation in expressions, it's really not feasible to aspire to post-coordination support in the absence of a terminology server
Lloyd McKenzie (Feb 05 2022 at 15:24):
Is including the display name limited to post-coordination? And it's still possible to have extensional value sets with post-coordination and for some of the systems involved to treat those as simple strings. (We do it all the time with UCUM...)
Grahame Grieve (Feb 05 2022 at 20:39):
well, it's a problem for UCUM too - really need grammar support to do code comparison. e.g. m/s = m.s-1
You can work around it with strict control over the value sets. I guess that same applies to Snomed CT too, but it always seems too much like this to me:
Lloyd McKenzie (Feb 05 2022 at 21:32):
I'm pretty sure that the vast majority of systems that accept kg and lb_av for weight would choke on kg/1 or kg{kilograms} - despite the equivalency that a fully UCUM-aware terminology server would associate with them.
Lloyd McKenzie (Feb 05 2022 at 21:33):
Which is why embedding displays in codes is bad practice, even if technically legal. FHIR provides a place to put the display information where it won't interfere with computation - and that's where the information should go.
Peter Jordan (Feb 05 2022 at 21:46):
Grahame Grieve said:
well.... given the various possibilities for syntactical variation without semantic variation in expressions, it's really not feasible to aspire to post-coordination support in the absence of a terminology server
... which uses a Classifier to perform validation and equivalence testing.
Lloyd McKenzie (Feb 05 2022 at 23:56):
Right. But the reality is, most systems won't be using a terminology server, even when working with post-coordinated terminologies (UCUM, SNOMED, ICD, CPT, whatever). And so, where the terminology allows embedding display names in the code, it's very bad practice to do that in FHIR instances.
Last updated: Apr 12 2022 at 19:14 UTC
