Stream: argonaut
Topic: Lab AMR structure
Jay Lyle (May 08 2018 at 16:48):
How are people implementing antimicrobial resistance tests (e.g., Clindamycin 7005-2)? They seem to fit the Prid=Observation + AMR=component pattern, though it's not clear whether the resistance should be a 'member' or 'sequel' to the organism. Or if there is some other tactic that would be better, e.g., diagnostic report.
Lloyd McKenzie (May 08 2018 at 16:50):
The resistance and the microbe would need to be conveyed together in a single Observation using Observation.component. They wouldn't be useable or interpretable if sent as distinct Observation instances.
Eric Haas (May 08 2018 at 19:14):
See these diagnostic report examples here in the FHIR specification. They are based on how lab does this in V2 today. Argonaut has not addressed the specifics for different kinds of lab reports. Although they may do so in the future.
Eric Haas (May 08 2018 at 19:15):
If you use components you limit your ability to treat each result independently.
Jay Lyle (May 08 2018 at 21:01):
I assume we'd want to limit the ability to treat results independently - you'd never want an "R" without a bug. But perhaps I'm missing something. Or perhaps structural alignment with V2 is a critical factor.
Lloyd McKenzie (May 08 2018 at 23:44):
Right. Capturing "resistant" or "amoxicillin" in separate observations is useless. Both have to be together to be meaningful. Thus components are required if you're dealing with resistance tests.
Eric Haas (May 09 2018 at 00:14):
I actually read the reference I gave and I do mention component as an option in a footnote
Eric Haas (May 09 2018 at 00:25):
for MICs btw your values are quantities and the interps are R,S,etc. There may be reason you want the resistances separate and can still revinclude to find out what the bug was
Eric Haas (May 09 2018 at 00:26):
but I would lean towards components - deferring to the lab folks ( if they ever implement fhir )
Lloyd McKenzie (May 09 2018 at 00:31):
The Observation needs to be completely meaningful on its own. If you have to revinclude to understand what it means or do anything useful with the information, then it's been structured incorrectly.
Eric Haas (May 09 2018 at 14:32):
OK, lloyd I agree. just replace the related with components.
Jay Lyle (May 09 2018 at 14:56):
Thanks. Another issue: component has value but not interpretation. Most of our source values are identical to their interpretations (S, I, R); those that are not are nonstandard strings without units (S <=2, BLac +, etc.). So we're planning to put the source "interpretation" in component.value.
Eric Haas (May 09 2018 at 14:59):
components have interpretations
Eric Haas (May 09 2018 at 15:00):
Screen-Shot-2018-05-09-at-8.00.29-AM.png
Eric Haas (May 09 2018 at 15:01):
Thanks to Jenni or Michelle for that one :-)
Jay Lyle (May 09 2018 at 17:13):
We'll look forward to that. Argonaut observation components don't, at least per this
Eric Haas (May 09 2018 at 20:45):
thats DSTU2 is in STU3 and USCore
Jay Lyle (Jan 19 2021 at 14:37):
Sorry; ID-10-t error.
David Pyke (Jan 19 2021 at 14:40):
Holy Resurrection of Old Threads, Batman!
Jay Lyle (Feb 03 2021 at 22:21):
OK, now I am resurrecting. Two questions.
-
R4 has Interpretation. Our feed gives us only an interpretation (S, I, R), not a quantified result. Should that be reported in Observation.component.interpretation, Observation.component.valueString, Observation.component.valueCodeableConcept? And if code, is it kosher to presume equivalence and assert that those codes are from http://terminology.hl7.org/CodeSystem/v3-ObservationInterpretation?
-
The example diagrams at http://build.fhir.org/diagnosticreport-examples.html use the term "related." I think this means "has member," contrary to Lloyd's (sensible, to me) direction to use Component. Eric, if that page mentioned "component" in the past it doesn't now. This example (http://build.fhir.org/diagnosticreport-micro1.json.html) has orphan susceptibility tests: they're the same bundle, but there's no association to the bug.
Eric Haas (Feb 04 2021 at 19:28):
I think it would be to better to propose an actual example in FHIR format that we all can throw rocks at until we get at least a partial consensus ( or jurisdictional consensus ). BTW this level of detail is not covered in US Core but has been proposed in the past as an Argonaut project but never got enough interest to pursue as a project. You may want to bug (pun intended) the CIMI folks to do some useful work on this. Also for broader context I would post on the #Orders and Observation WG
Eric Haas (Feb 04 2021 at 19:32):
Also I think that susceptibiltiies are resulted as MIC with interpretations or directly as results S,I,R so in the first the MIC is result value and in the second the susceptibity is. (though, I'm sure there variations on this theme in the wild )
Jay Lyle (Feb 04 2021 at 23:08):
Yeh, I assume the lab system has a MIC or a radius or some quant result, but the EHR just has SIR.
Last updated: Apr 12 2022 at 19:14 UTC
