Stream: implementers
Topic: lab results resource
Lital Inghel (Oct 06 2021 at 18:41):
as part of lab results resource
where can we represent
Microbiology finding
Pathology finding
Microbiology finding for example has info about the microorganism and susceptibility
I wonder if 'has member' is the right place or 'component'
Lloyd McKenzie (Oct 06 2021 at 18:59):
I think it would be 'component' as the micro-organism needs to be known in terms of understanding the Observation. An Observation needs to be meaningful by itself without looking at any other Observations.
Lloyd McKenzie (Oct 06 2021 at 18:59):
@Rob Hausam @Hans Buitendijk
Hans Buitendijk (Oct 06 2021 at 20:32):
For micro and pathology the .hasMember is likely the more appropriate way to structure. See https://build.fhir.org/observation.html#gr-other for notes on micro and sensitivities. Components should only be used to when the components cannot be separated, which would not be applicable for either of these use cases. There is work in progress to further clarify that beyond what is already there. Notes and initial diagram can be found here (https://confluence.hl7.org/download/attachments/40731564/Parent-Child%20Structures.pptx?api=v2) and look at slides 11, 12.
Lloyd McKenzie (Oct 06 2021 at 21:41):
@Hans Buitendijk, the modeling here: https://build.fhir.org/diagnosticreport-micro1.xml.html seems to violate one of the fundamentals of FHIR - a resource instance needs to be useful by itself. Seeing an Observation that says "sensitive" that doesn't itself say "to what" breaks that rule. The organism should absolutely be conveyed as a component here. What was the rationale for doing it otherwise?
Hans Buitendijk (Oct 07 2021 at 00:16):
@Lloyd McKenzie : Have a look at the slides referenced. The chain is there through .hasMember. If we want to go in reverse if .hasMember is not considered clear enough to establish that relationship and need to use .focus perhaps as it is about the observation that establishes the organism, fair question. But each sensitivity can exist on its own without having to exist with the other sensitivities. That's how that is currently done in HL7 v2 as separate OBX segments, NOT additional OBX-5 values.
Lloyd McKenzie (Oct 07 2021 at 02:05):
@Hans Buitendijk I misread the instance. You are conveying susceptibility in a single Observation. However, you're not using component because you're conveying the drug in the code. What I want to confirm is that if you didn't have a specific code for "susceptibility to drug X" but instead had to use a generic "susceptibility to drug" code that you would convey the drug as an Observation.component, not an Observation.hasMember.
Rob Hausam (Oct 07 2021 at 02:06):
@Lloyd McKenzie @Hans Buitendijk I think this modeling does break one typical FHIR "rule" - but it's not the one that Lloyd mentions. The "rule" that is violated is that the resource that occurs later should point to the resource that came before - but using hasMember the linkages occur in the other direction. Maybe we should consider whether we should do something about that (though I don't think .focus is appropriate here, as the observation is still on the patient of record). But that's not Lloyd's issue. There are some oddities in this example (in particular I think it links the susceptibility panels in the wrong way). But overall it otherwise accurately reflects the structure of this data as it's typically presented in laboratory reports and, as Hans mentions, also reflects the way that this data is being structured and sent by labs in v2 messages. With all of the parts and multiple linkages involved (e.g. culture, one or more organism identifications/isolates, one or more susceptibility panels performed on each organism/isolate, and multiple drug susceptibility tests included in each susceptibility panel), there is absolutely no way that this can be done in a complete and useful way by representing the organisms/isolates as components of the culture observation. What we possibly could do is represent the individual drug susceptibility results in the susceptibility panel as components (because they are the "leaf" level) - but I don't think that would buy very much, if anything, at all (and since the linking would be different at that level only it might actually seem even a bit more confusing).
Rob Hausam (Oct 07 2021 at 02:09):
In response to the post Lloyd just made, I don't think that whether there is a specific code for "susceptibility to drug X" really would affect this very much. As I mentioned, even with the specific codes you still could represent the individual drug susceptibility results as components - but not the rest of it.
Last updated: Apr 12 2022 at 19:14 UTC
