Stream: implementers
Topic: remove binding from base profile
Jose Costa Teixeira (Jun 23 2020 at 05:47):
MedicationDispense.medicationCodeableConcept has an example binding that I find not helpful, and I want to remove it because it is confusing.
Can I profile to remove the binding?
Lloyd McKenzie (Jun 23 2020 at 05:50):
No
Jose Costa Teixeira (Jun 23 2020 at 05:51):
ok that is concerning
Jose Costa Teixeira (Jun 23 2020 at 05:51):
Are the any systems that use Snomed for the dispensed medication? @Jean Duteau (update - I don't think this is a good example binding but the issue is not as concerning as I thought - I can just profile it out)
Jose Costa Teixeira (Jun 23 2020 at 05:52):
I don't think 8% of the systems would put a snomed-ct code in there.
Jose Costa Teixeira (Jun 23 2020 at 05:53):
Can I override the binding in my profile?
Richard Townley-O'Neill (Jun 23 2020 at 06:13):
Jose Costa Teixeira said:
Are the any systems that use Snomed for the dispensed medication? Jean Duteau
In Australia SNOMED CT is one of the options, but not the ValueSet in the core resource. See http://build.fhir.org/ig/hl7au/au-fhir-base/StructureDefinition-au-medicationdispense.html
Jose Costa Teixeira (Jun 23 2020 at 06:16):
as medication type SCT makes sense, but not as a the medication code
Jose Costa Teixeira (Jun 23 2020 at 06:17):
I think i will do the same as you did @Richard Townley-O'Neill
Jose Costa Teixeira (Jun 23 2020 at 06:19):
and I added J#27863 to check this - I don't expect this to be a default binding in Dispense
Jim Steel (Jun 23 2020 at 06:21):
Australia would
Jim Steel (Jun 23 2020 at 06:22):
(not claiming that means 80%! But from a data gathering point of view, we use SNOMED codes for medications fairly commonly)
Grahame Grieve (Jun 23 2020 at 06:22):
Can I profile to remove the binding?
No, but you can replace it. Why wouldn't you do that?
Jose Costa Teixeira (Jun 23 2020 at 06:22):
oh and btw, @Richard Townley-O'Neill how did you remove the "Text Summary" tab from your profiles? I think that is very useful
Jose Costa Teixeira (Jun 23 2020 at 06:24):
Can I profile to remove the binding?
No, but you can replace it. Why wouldn't you do that?
That is what I meant with overriding it.
If I make a different binding, don't I need to define a new value set? At this moment we don't yet have a value set - we can use the national codes or gtins or something else.
Jose Costa Teixeira (Jun 23 2020 at 06:29):
seems my differential is not what it should be. I'm checking one thing
Jose Costa Teixeira (Jun 23 2020 at 06:32):
Yep, I figured out what my mistake was:
I wanted to override the binding, so I added the only thing I needed - the description.
But I did this in sushi, and sushi does not remove the entire binding object, only replaces the attributes I mention.
Jose Costa Teixeira (Jun 23 2020 at 06:33):
I should have just changed the differential json by hand instead of sushi.
Kevin Mayfield (Jun 23 2020 at 06:37):
UK (England) would expect to use SNOMED CT but I expect the MedicationDispense to look like:
Both have SNOMED CT slices. The example binding isn't necessary?
Jose Costa Teixeira (Jun 23 2020 at 06:40):
I think/hope I have the same point: The binding here will normally be for a product. SCT indicates more of a "substance" or "type"
Jose Costa Teixeira (Jun 23 2020 at 06:41):
conformance-wise, can we do this in a profile?
image.png
The binding object is still in the differential, but only contains a description
Jean Duteau (Jun 23 2020 at 15:28):
so a) we need to have a binding in the core specification, b) there is no global source of medication codes, therefore c) we made an example binding of SNOMED CT.
Since it is example, it is made to be overridden in profiles.
Melva Peters (Jun 23 2020 at 16:34):
I added a similar comment to the JIRA issue that was created.
Jose Costa Teixeira (Jun 23 2020 at 17:37):
Ok, I was not aware the core spec needed a binding. Given that dispense is not on substance, I wonder if SCT was the least evil solution..
Jose Costa Teixeira (Jun 23 2020 at 17:37):
can we "unbind" something in a profile? @Lloyd McKenzie ?
Jean Duteau (Jun 23 2020 at 17:40):
why would you "unbind"? it is only example so it just provides an example of the types of codes. and your profile should have a more specific binding - like a national registry of product codes.
Jose Costa Teixeira (Jun 23 2020 at 17:50):
we want to publish this without the terminology for now; we don't know which national product codes will be used.
Jose Costa Teixeira (Jun 23 2020 at 17:51):
we do know it's not snomed ct (which is why i think it would be confusing to keep it)
Jean Duteau (Jun 23 2020 at 19:15):
If you gave me a profile with no binding and just a description, I'm going to ask you to tell me what kinds of codes you would think would go there. That is why the core spec requires bindings always and why we have example bindings - they are not necessarily meant to be used in the real world but provide a hint (an example) to implementers and IG creators of the types of codes that the committee intended to go in that element.
Jose Costa Teixeira (Jun 23 2020 at 19:37):
I understand, but at this moment the requirement is "we don't know the codes, we know it's not snomed ct". For implementers that have to validate this, I'd prefer to have a stub for valueset (or nothing) rather than imply SCT.
Lloyd McKenzie (Jun 23 2020 at 19:38):
You cannot unbind. You can only change the binding
Jose Costa Teixeira (Jun 23 2020 at 19:39):
Can we make a 'void' valueset saying " this currently has no codes, use what you have now until we come up with a valueset"?
Jean Duteau (Jun 23 2020 at 19:39):
but an example binding doesn't imply SCT.
Jean Duteau (Jun 23 2020 at 19:39):
an example binding doesn't imply anything
Lloyd McKenzie (Jun 23 2020 at 19:40):
You could specify a binding that just had a description
Jose Costa Teixeira (Jun 23 2020 at 19:40):
(and I know this is not how you should make a system, it is just that we are seeking some agreement on using fhir to replace the infrastructure, and the words snomed CT created confusion already even if it was example binding)
Jose Costa Teixeira (Jun 23 2020 at 19:41):
You could specify a binding that just had a description
Yes, that is what I did - I created a binding on that element and just put a description in there.
Jose Costa Teixeira (Jun 23 2020 at 19:42):
It shows as my screenshot above
Jean Duteau (Jun 23 2020 at 19:42):
instead of whether what you did, I would make a value set that had a description and no codes.
Jean Duteau (Jun 23 2020 at 19:42):
so it had an actual binding to a clickable value set and the value set would only have a description.
Jean Duteau (Jun 23 2020 at 19:42):
that allows you to add codes to that value set in the future
Jose Costa Teixeira (Jun 23 2020 at 19:43):
That is useful, I can also do that, and may help push for the terminology people to catch up
Jose Costa Teixeira (Jun 23 2020 at 19:43):
Thanks
Richard Townley-O'Neill (Jun 24 2020 at 01:15):
Jose Costa Teixeira said:
oh and btw, Richard Townley-O'Neill how did you remove the "Text Summary" tab from your profiles? I think that is very useful
I think the change is in sd.html. See
https://github.com/hl7au/au-fhir-base/commit/afb77cb40a1c1dc909640c4a4b9131559fb7d270#diff-08b0229c202c28a36b45b7f6a28a66f8
Jose Costa Teixeira (Jun 24 2020 at 04:50):
thank you
Robert McClure (Jun 26 2020 at 13:14):
Sounds like you need a concept domain, if we only knew what that was... But @Jose Costa Teixeira , instead of an empty value set with a description, (FHIR's hack to represent the idea of a concept domain), why not make an example binding to one of the possible code systems you think could be chosen?
Jose Costa Teixeira (Jun 26 2020 at 13:22):
I cannot give examples of what may be, because that will be available later. Still, i find the idea of a concept domain interesting
Jose Costa Teixeira (Jun 26 2020 at 13:24):
For example sct 763158003 seems the concept i'm referring to
Robert McClure (Jun 26 2020 at 18:10):
@Jose Costa Teixeira That is an interesting approach to represent the idea of a concept domain within the constraints of FHIR (must bind a value set). I'd support the idea of example binding a value set with that single general SCT concept as a way of conveying what the actual value set should be.
Last updated: Apr 12 2022 at 19:14 UTC
