FHIR Chat · Searching for postcoordinated code(fragements) · terminology

Stream: terminology

Topic: Searching for postcoordinated code(fragements)


view this post on Zulip Patrick Werner (Nov 29 2021 at 15:07):

As the fhir search api doesn't allow for searching parts of a code - dealing with postcoordinated codes is hard. As a result we forbid post-coordination in ICD10-GM in the latest german base IG version.
Implementers complaint, that the couldn't search for one of the concepts in the postcoordinated codes.

view this post on Zulip Patrick Werner (Nov 29 2021 at 15:08):

But there are some implementers who want to use/need to use postcoordinated codes (e.g.SnomedCT ones). Is there a best practice or experience on how you still could enable searchability, or search for parts of a postkoordinated code?

view this post on Zulip Grahame Grieve (Nov 29 2021 at 18:50):

the fhir search api doesn't allow for searching parts of a code

Doesn't it? where does it say that?

view this post on Zulip Daniel Venton (Nov 29 2021 at 19:24):

I suspect @Patrick Werner is referring to the idea that most codes are represented via a Token search. Token searches by default are "exact match" style. There is no modifier on a token search that is a "contains". So when the code value is "A21.1 B32.2 C21.5" how do you search for all items with "B32.2"?

view this post on Zulip Patrick Werner (Nov 29 2021 at 19:25):

correct @Daniel Venton Thats exactly what i meant

view this post on Zulip Grahame Grieve (Nov 29 2021 at 19:27):

well, there is a mechanism for it. It's indirect, but it has to be to deal with the complexity required here.

  • define a value set "all codes qualified by B32.2"
  • search for all concepts in the value set

Sure, that requires work to set up - but any server is going to have to work on something like that

view this post on Zulip Grahame Grieve (Nov 29 2021 at 19:31):

on a more ad-hoc level, there's :below.

  • search for :below "anything B32.2"

This is... obscure... and depends on code system semantics.

view this post on Zulip Grahame Grieve (Nov 29 2021 at 19:32):

I think that my server doesn't support either of these anyway, and I very much doubt others do. (well, my server supports value set searches, but I don't think that it supports the ICD-10 logic implied by the question)

view this post on Zulip Grahame Grieve (Nov 29 2021 at 19:32):

I did think we had some partial search arrangement on tokens, but it doesn't look like we do

view this post on Zulip Patrick Werner (Nov 30 2021 at 08:21):

and now it is too late to add i'm afraid? It would be very helpful to have :contains on token searches.

view this post on Zulip Grahame Grieve (Nov 30 2021 at 10:18):

no it's not too late to add.

view this post on Zulip Grahame Grieve (Nov 30 2021 at 10:26):

at least, not procedurally. It might be a hard to get up though

view this post on Zulip Michael Lawley (Nov 30 2021 at 10:30):

:text as a search modifier could allow this to work without changes?

view this post on Zulip Grahame Grieve (Nov 30 2021 at 12:08):

no because that searches the text, not the code

view this post on Zulip Carl Leitner (Dec 09 2021 at 17:26):

Grahame Grieve said:

well, there is a mechanism for it. It's indirect, but it has to be to deal with the complexity required here.

  • define a value set "all codes qualified by B32.2"
  • search for all concepts in the value set

Sure, that requires work to set up - but any server is going to have to work on something like that

For reference, ICD-11 puts post-coordinated codes in a valueset (and working on getting this exposed as FHIR now). Flagging for @Jonathan Payne and @Joe Amlung


Last updated: Apr 12 2022 at 19:14 UTC