FHIR Chat · How to represent SNOMED CT < conceptid · terminology

Stream: terminology

Topic: How to represent SNOMED CT < conceptid

Alexander Henket (Feb 14 2020 at 10:40):

I got this error from the IG Publisher:

Error from server: The filter "concept descendent-of 442083009" was not understood in the context of http://snomed.info/sct

In looking at the SNOMED CT page in the spec I indeed see no mention of descendent-of. But the semantics of descendent-of match exactly what I want to express with:

< 442083009 |Anatomical or acquired body structure|

So question is: what is the right way to express the above statement in FHIR? Currently my ValueSet says:

<compose>
    <include>
        <system value="http://snomed.info/sct"/>
        <filter>
            <property value="concept"/>
            <op value="descendent-of"/>
            <value value="442083009"/>
        </filter>
    </include>
</compose>

Richard Kavanagh (Feb 14 2020 at 13:40):

hi @Alexander Henket take a look at http://hl7.org/fhir/STU3/valueset-administration-method-codes.xml.html I think this is what you need.

Alexander Henket (Feb 14 2020 at 13:42):

You mean "is-a" instead of "descendent-of" I think? But "is-a" is defined as including the head concept. While descendent-of is defined as excluding the head concept.

SNOMED CT < is supposed to exclusive while << is inclusive, is it not?

Richard Kavanagh (Feb 14 2020 at 13:45):

Hmm - not ideal but use the "is-a" and then add an "exclude" for the head concept.

Alexander Henket (Feb 14 2020 at 13:46):

http://hl7.org/fhir/STU3/valueset-filter-operator.html
and
https://confluence.ihtsdotools.org/pages/viewpage.action?pageId=28739405

The SNOMED CT Expression Syntax says:

descendantOf = "<"
descendantOrSelfOf = "<<"

I think descendent-of is exactly matching <, while is-a is exactly matching <<. That must mean that the FHIR SNOMED CT page is lacking

Yunwei Wang (Feb 14 2020 at 15:57):

I remember that each code system decides what filter FHIR should support. http://hl7.org/fhir/STU3/snomedct.html#filters lists three filters supported by SNOMED CT in FHIR.

Michael Lawley (Feb 15 2020 at 00:58):

You can use ECL with the constraint filter:

<compose>
    <include>
        <system value="http://snomed.info/sct"/>
        <filter>
            <property value="constraint"/>
            <op value="="/>
            <value value="<442083009"/>
        </filter>
    </include>
</compose>

Agreed, missing descendant-of and related ops from http://hl7.org/fhir/STU3/snomedct.html#filters is a bug.

Alexander Henket (Feb 15 2020 at 05:11):

It is still missing that notion in R5, so I took the liberty of filing a tracker for it: J#26305

Grahame Grieve (Feb 15 2020 at 05:12):

feel free to create a task. though it's possible the deficiency is in tx.fhir.org not the specification

Grahame Grieve (Feb 15 2020 at 05:12):

ah you did.

Alexander Henket (Feb 15 2020 at 05:14):

I can see how constraint could work, but that seems like a much harder solution for a fairly common expression, so support for descendent-of would be appreciated

Grahame Grieve (Feb 15 2020 at 05:17):

I think it would be good to add it too

Michael Lawley (Feb 16 2020 at 22:46):

It would kind of make sense that all the subsumption-related filter operations were defined by default for CodeSystems where hierarchyMeaning = is-a

Alexander Henket (Nov 11 2020 at 08:52):

Is there a call I could participate in where this could be discussed? (J#26305)

Rob Hausam (Nov 11 2020 at 20:30):

@Alexander Henket I think this should be a pretty straightforward discussion and resolution. I think we can include it as one of the FHIR items on the Vocab Main call tomorrow at 3:30 PM Eastern (we will break away from 4:00-5:00 PM for the co-chair meeting and then will resume again at 5:00). So if you can join I assume we would need to take this up during the first 30 minutes. @Ted Klein?

Rob Hausam (Nov 12 2020 at 03:52):

@Alexander Henket It's been added to the Vocab call agenda for tomorrow (should be in the first 30 minutes).

Alexander Henket (Nov 13 2020 at 08:01):

Hi all. I'm sorry I missed that call, but I'm glad with the outcome. [how do you guys keep track of all conversations in Zulip?]

Alexander Henket (Nov 13 2020 at 08:10):

Follow up question: would this SNOMED filter guidance, slated for R5 be limited to implementations on R5 and up or would this also benefit STU3/R4? In other words: is terminology system guidance tied into a version of FHIR specifically or do reference frameworks generally follow the latest insights on how to deal with SNOMED/LOINC/... ?

Michael Lawley (Nov 13 2020 at 08:14):

I see no reason why this should be necessarily tied to a FHIR version

Grahame Grieve (Nov 13 2020 at 08:49):

when I implement it, it will be for all versions

Last updated: Apr 12 2022 at 19:14 UTC