Stream: conformance
Topic: Define an 'OR' for coded elements.
Marten Smits (Mar 16 2021 at 10:04):
We are creating profiles for data to be send to a quality registry. For registries it is common to need data like:
"The patient took Picoprep/Citrafleet”. So either X or Y. And the patient has a histrory of either A or B or C.
Currently the registry entry is filled out with a form by the practitioner, which just defines true/false to these question, but we translate those to the appropriate Medication/Condition/Observation resources. However, as far as I know, there is no way of expressing X or Y in FHIR in for example MedicationStatements is there? So I'm leaning towards creating registry specific codes for there cases, and in the future (when these items can be prepopulated by the EMR) these can be replaced by specific codes for X and Y.
However, this pattern is reoccuring, and it seems a bit of a shame that when we have 2 standardized codes that could be used, we cannot, and have to create and use something totally non-standardized, non-reusable. So I wanted to make sure there isn't any other option to express something like this.
Any suggestions?
Vadim Peretokin (Mar 16 2021 at 11:05):
It's a tight one. Can't use CodeableConcept for this either as it's just about one concept. At the end of the day your problem boils down to uncertainty about the data, and allergyintolerance solves this with an extension.
I'd go for the following: record both of the drugs (so 2 resources), tie them together with a common identifier, and add a modifier (important) extension for the certainty.
This'll make it more interoperable than custom codes - you'd be able to search on either drug using the standard interface with standard codes and you'll get a result. If you want to get more semantic information about it, you can, it'll be there thanks to the common identifier and the certain extension.
Marten Smits (Mar 19 2021 at 11:13):
Thanks for your input Vadim! But I'm afraid this over complicates stuff as well. Custom extensions and identifiers don't seem more interoperable than custom codes in my opinion. This is rather tricky isn't it?
Frank Oemig (Mar 22 2021 at 10:08):
Right. Interoperability is about the same understanding. So, every self defined profile, code or whatever complicates making progress in that regard.
Lloyd McKenzie (Mar 22 2021 at 13:11):
You're unlikely to get full interoperability in this case because a lot of systems won't understand what's going on. So your objective should be maximum safety - i.e. trigger the right decision support, show up in the right searches. That means a separate resource for each of the 'or' clauses and flagging each of them as uncertain. You can use extensions or common business identifiers in an attempt to link the collection as a single 'statement' but lots of systems won't recognize them. So put in the code.text and ideally the narrative the full or-clause so a human can see it.
Chris Grenz (May 17 2021 at 14:30):
If we wanted to come up with a better solution for this, what committee might be interested? This has be a very, very common problem in real-world usage of FHIR over the last couple years for me....
Grahame Grieve (May 17 2021 at 19:41):
which particular problem?
Chris Grenz (Nov 01 2021 at 16:12):
@Grahame Grieve Asserting an observation of a list of OR'd conditions, e.g. a Questionnaire response to "Have you had any of nausea, headache, or sore throat (yes/no)?"
Grahame Grieve (Nov 01 2021 at 19:52):
can you be more specific?
Lloyd McKenzie (Nov 01 2021 at 20:16):
Questionnaire makes it easy to say "Have you had any of nausea, headache or sore throat (yes/no)" - but how do you express that in Condition if you only have codes for "nausea", "headache", and "sore throat" but not the combination of them.
Grahame Grieve (Nov 01 2021 at 20:20):
I'm not understanding the question - is it 'how do you say things that your terminology doesn't allow you to say'?
Lloyd McKenzie (Nov 01 2021 at 21:34):
I think it's "Can the data model allow me to say things the terminology doesn't allow me to say."
Grahame Grieve (Nov 02 2021 at 19:34):
umm no? how could it?
Chris Grenz (Nov 08 2021 at 16:23):
Well...the terminology has codes for each of the concepts, but there's no conceptual (or syntactic) construct for "one of". The model could introduce the construct. With where we are today, I'd suggest a standard extension on Observation with that effect, probably a modifier extension at the root or on Observation.code.
Rob Hausam (Nov 08 2021 at 16:33):
There's always a tension between the expressivity and expressibility of the information and terminology models. I suggest giving careful consideration to both sides of that before deciding what's the "right" way to do this.
Last updated: Apr 12 2022 at 19:14 UTC
