Stream: snomed
Topic: Snomed version
Grahame Grieve (Aug 12 2018 at 04:47):
One of the tasks for FHIR R4 was to stop using an sct code that is no not active, and to update the build tool to emit a warning for an use of inactive codes
Grahame Grieve (Aug 12 2018 at 04:48):
the first thing I did was to update to the latest sct intl release, and that caused me 10000s of issues (invalid sct codes). I'm not sure what and I think it's something infrastructural, and I don't have time to deal with that. So the version of FHIR that will be balloted in the sept ballot cycle will be based on SCT 20160731. I will update to 20180731 by the time the final version of FHIR is released
Grahame Grieve (Aug 12 2018 at 04:49):
However I'm going to need help from people here to find me replacement code for 10-20 SCT codes that appear to be a genuine issue (scanning through the 1000s of errors quickly).
Grahame Grieve (Aug 12 2018 at 04:50):
it might be that some of my issue relates to 1000s of old style snomed codes somewhere.... I'm not sure
Rob Hausam (Aug 12 2018 at 04:53):
I don't recall encountering "old style" codes very often, if at all - I wouldn't have have guessed there were thousands, but maybe could be (behind the scenes somewhere)?
Grahame Grieve (Aug 12 2018 at 04:56):
in the imaging stuff ? i don't know and I don't have time right now
Rob Hausam (Aug 12 2018 at 04:57):
maybe - most common in pathology, I expect, but I don't think we likely have a great deal of that?
agree there's not time to sort it out now
Grahame Grieve (Aug 12 2018 at 05:13):
no my main problem is not fixed by backing out the version change
Grahame Grieve (Aug 12 2018 at 05:13):
<concept> <extension url="http://hl7.org/fhir/StructureDefinition/codesystem-replacedby"> <valueCoding> <system value="http://snomed.info/sct"/> <code value="F-00E1F"/> <display value="Minimum diastolic blood pressure"/> </valueCoding> </extension> <code value="109021"/> <display value="Diastolic pressure nadir"/> <definition value="The lowest pressure value excluding any undershoot artifact"/> <property> <code value="status"/> <valueCode value="deprecated"/> </property> </concept>
Grahame Grieve (Aug 12 2018 at 05:13):
There's 90 of these in the DICOM code system
Grahame Grieve (Aug 12 2018 at 05:17):
@Elliot Silver @David Clunie my understanding of the SCT response on this issue (at https://confluence.ihtsdotools.org/display/FHIR/URIs+antecedent+SNOMED+versions) is that this usage above is not compliant with the SCT license, and you need to regenerate the code system (or I do? I'm not clear on it's origin right now) to use SCT Ids not RTs
Grahame Grieve (Aug 12 2018 at 05:32):
for now, I'm changing them all to an FHIR SID to work around the validation problem (my validation is very enthusiastic on this code system, it gets validated plenty of times, and I was getting showered with duplicate errors)
Michael Lawley (Aug 13 2018 at 06:59):
If you can send me a list of the problematic codes, then I should be able to build you a map of their replacements. Actually, those ConceptMaps already exists as implicit maps in Ontoserver. URIs are ReplacedBy: http://snomed.info.sct?fhir_cm=900000000000526001
and SameAs: http://snomed.info.sct?fhir_cm=900000000000527005
Public Ontoserver has International 20180131 but not 20180731 yet
Michael Lawley (Aug 13 2018 at 07:20):
SCT have not been at all helpful with the issue of these old codes. You're not licensed to create new records with them (they can't force people to removed them from existing records), but it doesn't help when you need to create artefacts whose purpose is to support migration from old to new
Grahame Grieve (Aug 13 2018 at 10:04):
ok I'll do that when I get to it again
Grahame Grieve (Aug 13 2018 at 10:04):
thanks
Elliot Silver (Aug 13 2018 at 18:11):
@Grahame Grieve, I'm not quite following. Are you finding SCT values in http://build.fhir.org/codesystem-dicom-dcim.html?
Rob Hausam (Aug 13 2018 at 20:13):
It looks like he's finding legacy SNOMED identifiers (not strictly from SNOMED RT) like "F-00E1F" - apparently 90 of them. And those identifiers should not be used for any current data or going forward.
Elliot Silver (Aug 13 2018 at 20:26):
Where are they being found?
Grahame Grieve (Aug 13 2018 at 20:58):
each of the deprecated codes - but they aren't rendered, so look in the source e.g. http://build.fhir.org/codesystem-dicom-dcim.xml.html
Elliot Silver (Aug 13 2018 at 22:09):
OK, so DICOM has retired codes that say go see SNOMED RT, but SNOMED has retired RT. So right now, they are marked as replaced with an invalid replacement. On the one hand, this doesn't impact any of the currently defined DICOM vocabulary. On the other hand, DICOM isn't giving good guidance for what to replace their retired terms with. Is that the issue? If so, I can file a DICOM change proposal (CP) to provide updated mappings, but that could easily take 6 months to get addressed. Beyond that, we'll have to get @David Clunie involved.
Michael Lawley (Aug 13 2018 at 22:38):
Seems like SNOMED International should also be approaching DICOM to get them to migrate properly - @Linda Bird @Peter Williams ?
Grahame Grieve (Aug 13 2018 at 23:31):
I think that
- DICOM should provide concept ids as well as the RT codes
- SNOMED international should allow Dicom & HL7 to publish the old codes so that people with those codes know what do with them
Rob Hausam (Aug 13 2018 at 23:31):
Maybe it's not really important, but I'll mention again that this isn't (at least mostly) about identifiers that were introduced in SNOMED RT. These are the legacy identifiers that were used in all of the versions of SNOMED prior to RT and then were also carried forward as additional (not the primary) identifiers in both SNOMED RT and the earlier (RF1) versions of SNOMED CT.
Grahame Grieve (Aug 13 2018 at 23:32):
I suppose SCT publishes a map of these somewhere, and I could find it if I spent the time...
Grahame Grieve (Aug 13 2018 at 23:32):
I will try to remember to call them 'legacy identifiers'
Rob Hausam (Aug 13 2018 at 23:32):
ok :)
Rob Hausam (Aug 13 2018 at 23:33):
and I'll try not to mention it again
Michael Lawley (Aug 14 2018 at 01:50):
right, they exist in a file called SnomedCT_IdentifierRefset_DEPRECATED_20170731T120000Z.zip
containing SnomedCT_IdentifierRefset_DEPRECATED_20170731T120000Z/Snapshot/Refset/Map/der2_sRefset_SimpleMapSnapshot_INT_20170731.txt
which says that F-00E1F
maps to 314451001
for example
Grahame Grieve (Aug 14 2018 at 07:22):
thanks. I guess I'll have to find it and process it
Peter Williams (Aug 14 2018 at 08:29):
@Michael Lawley Yes Jane Millar had a conversation with DICOM and I believe they are heading in that direction but there are technical reasons for them not having already made the move. I'll follow up with her, thanks.
Grahame Grieve (Aug 26 2018 at 22:17):
I'm not following Snomed Intl 's logic here. As things stand, we can't even publish a mapping table from SRT codes to SCT concept ids, because we can't define a system by which to represent SRT codes. So, SI's desire to get people to stop using SRT codes is preventing us (and/or DICOM) from publishing ConceptMap resources that would help people to stop using SRT codes)
Michael Lawley (Aug 26 2018 at 22:34):
Exactly. And as I see things, unless this position can be changed (e.g., via the SHOMED on FHIR working group) the only path forward will be for FHIR to establish its own CodeSystem URI for these old codes.
Grahame Grieve (Aug 26 2018 at 22:36):
well, perhaps it's time to do that.
Grahame Grieve (Aug 26 2018 at 22:36):
http://hl7.org/fhir/sid/srt would be our normal practice
Rob Hausam (Aug 26 2018 at 23:17):
I still say that these are not SNOMED RT id's - they are legacy SNOMED identifiers which began being assigned and used in the earlier SNOMED versions (long before SNOMED RT was ever imagined). And they didn't stop being available for use as alternate identifiers at first (in RF1) when SNOMED RT was replaced with SNOMED CT. I would propose that we use http://hl7.org/fhir/sid/snm (or something similar).
Michael Lawley (Aug 27 2018 at 00:09):
For the record, here is the discussion on this topic from before https://confluence.ihtsdotools.org/display/FHIR/URIs+antecedent+SNOMED+versions
And this is the documentation for the two relevant maps published by SNOMED International https://confluence.ihtsdotools.org/display/DOCTSG/4.3.1+Legacy+Code+Map+Reference+Sets
It looks like two URIs are needed - http://hl7.org/fhir/sid/s3x for SNOMED 3.x and (my proposal) http://hl7.org/fhir/sid/snid for the SNOMEDIDs - avoiding things that reference SNOMED RT is probably best since it included SCTIDs.
Rob Hausam (Aug 27 2018 at 11:56):
I like Michael's proposal, except that I'm not fond of 'snid' (at least to me it's maybe a bit too close to "snide"). I would suggest 'snmid', which is consistent with the prior V2 table 0396 codes ("SNM" and "SNM3") and with "Systematized Nomenclature of Medicine".
Michael Lawley (Aug 27 2018 at 11:58):
I almost suggested 'snmid' originally but arbitrarily went for a shorter option.
Grahame Grieve (Aug 27 2018 at 12:08):
snid - sounds like a Dr Suess name
Rob Hausam (Aug 27 2018 at 18:07):
yes, it does :)
Michael Lawley (Aug 29 2018 at 21:22):
which is almost reason enough to prefer it :)
Last updated: Apr 12 2022 at 19:14 UTC