Stream: Orders and Observation WG
Topic: Observation: How to track used reagents/disposables etc.
Ralf Herzog (Jun 26 2018 at 10:18):
After digging a bit into the observation resource I wonder how to model used reagents, disposables, etc. pp. into the observation resource. Is there a profile/extension for this or should it be included into the resource itself?
Jose Costa Teixeira (Jun 26 2018 at 10:20):
device? closest woudld be Procedure.usedReference.
Jose Costa Teixeira (Jun 26 2018 at 10:21):
(if Device seems not adequate, i've been thinking of a healthProduct, but that would be a pattern, not an implementable resource)
Ralf Herzog (Jun 26 2018 at 10:26):
For disposables: yes, but there is only a 0..1 relationship in observation to device.
But where to place reagent(s) (remember there might be multiple for one observation)...
Eric Haas (Jun 26 2018 at 14:04):
can you give a concrete example of your use case? I am leaning on Device/DeviceComponent as well.
Jose Costa Teixeira (Jun 26 2018 at 16:13):
The device that produced the information is on observation.device
Jose Costa Teixeira (Jun 26 2018 at 16:14):
if you consumed items, this is not part of the observation, but it is actually part of the procedure that prodced this observation.
Jose Costa Teixeira (Jun 26 2018 at 16:14):
hence the link being on Procedure.usedReference
Ralf Herzog (Jun 26 2018 at 16:14):
Let's try with an example Measurement of Glucose (but please do not nail me that Glucose is not measured that way).
Used reagents to measure Glucose are:
- Glucose Reagent
- Glucose Dilutent
Used Consumables
- Glucose Pipette
- Glucose Container
Jose Costa Teixeira (Jun 26 2018 at 16:15):
So, you have
Jose Costa Teixeira (Jun 26 2018 at 16:15):
observation1 (your observation).
Jose Costa Teixeira (Jun 26 2018 at 16:16):
observation-PartOf= Procedure1 (the acquisition procedure)
Jose Costa Teixeira (Jun 26 2018 at 16:18):
then on Procedure1.usedReference you have your consumables and reagents
Jose Costa Teixeira (Jun 26 2018 at 16:19):
the observation.Devicethat is if you need to indicate which machine was used in producing that result
Jose Costa Teixeira (Jun 26 2018 at 16:25):
makes sense?
Ralf Herzog (Jun 26 2018 at 16:34):
So you say that the observation is part of the Procedure where the reagents and consumables are linked too? Hmm...
I would argue there, that an observation is something where reagents and consumables were used to produce whatever value. And the observation would belong to a task/order.
I try to explain it in v2 kind of style.
You get an order via an OML^O33 message where the order: for Sample XYZ (SPM-2) in Container (SAC-3) Glucose (OBR-4) should be measured.
The result upload would be in a OUL^R22 message where the result would be in OBX-5 and the Substance information would be in subsequent INV Segments where the first would include the Glucose reagent (INV-1 something like GlucoseReagent^^99WhatEver) and the second the Glucose dilutent (INV-1 something like GlucoseDilutend^^ 99WhatEver). The Third INV segment would be the used disposable (INV-1 something like Disposable1^^99WhatEver) and so on.
I think the message structure of OUL^R22 as in the IHE LAB-29 Transaction makes more sense here.
To mention: this is in a centralized lab environment .
Jose Costa Teixeira (Jun 26 2018 at 16:35):
the v2 model does anchor us to some conceptual ways of expressing things.
Jose Costa Teixeira (Jun 26 2018 at 16:37):
"the observation is something where reagents..." - let me dissect that:
for my proposed suggestion, I assume that observation is the result of a process. Typically you don't capture the details of the process, just the outcome, so you don't need procedure. But if you need to capture consumables, time it took to process that, etc then you need something.
you are correct, it could be a task as well. but if you consider observation as the outcome of an activity, then the consumables are not part of the outcome, but they are part of the activity.
Ralf Herzog (Jun 26 2018 at 16:38):
yes, but at least it - for this use case- makes the modeling quiet easy. When you mention procedure I always think on procedures performed on patients (like the procedure resource is mentioning in the Description & Constraints "An action that is being or was performed on a patient"
Jose Costa Teixeira (Jun 26 2018 at 16:38):
on the other hand, if you consider observation as being the procedure, then indeed you have a point and something would be missing.
Jose Costa Teixeira (Jun 26 2018 at 16:39):
ah there you have a point - what is the way in FHIR to convey automated activities like a machine processing a specimen and producing a result - i don't know if that is clear
Ralf Herzog (Jun 26 2018 at 16:44):
Yes, I think we might need there a bit of a more lengthy discussion which I think this in the appropriate WG.
But we should involve some more people with an analyzer, IHE PaML (LDA and LAW) background.
Jose Costa Teixeira (Jun 26 2018 at 16:49):
@François Macary @Lorraine Constable @Andrea Pitkus, PhD, MLS(ASCP)CM @Eric Haas can help there?
Jose Costa Teixeira (Jun 26 2018 at 16:51):
@Riki Merrick (i was not forgetting you, but I could not find you under your full name :)
Riki Merrick (Jun 26 2018 at 20:30):
I agree that this needs a bit more thought before we add something to observation (that's normative). I think Eric might have mentioned device, becasue that is what I was thinking is used for procedures NOT on patient. I just reviewed the definition and it says on or for a patient, so lab tests on specimen would fall into the latter category, so procedure seems to be the right place.
Grahame Grieve (Jun 27 2018 at 04:28):
I don't think this is about observation at all. There's nothing inherent in observation about using anything to make it. If we want to do something about the process of making an observation, that's something different
François Macary (Jun 27 2018 at 14:32):
Let me give a try on this topic.
Reagents, like analyzers, and pipettes are medical devices. Their identification and representation is guided by the HL7 UDI Implementation Guidance DAM. In terms of use cases, the requirement to identify which analyzer produced the glucose observation, and what was the reagent lot number used for that, is not uncommon. I don't know precisely about the details of CLIA in the US, but in Europe, that would be part of the requirements of clinical laboratory accreditation based on the ISO 15189 standard.
In V2.9 (adopted for this purpose by the IHE LAW profile) the reagent properties (identifier, expiration date, lot number, ...) are conveyed by an instance of INV "Inventory Detail" segment, which is part of the RESULT segment group wrapping an OBX (observation) segment.
The logic of such quality requirements is to link this set of information to the observation itself. Just like for the method used, we are still characterizing the observation produced, rather than describing a procedure.
So the transposition to FHIR seems straightforward. We have Observation.method for the method, and we have Observation.device referencing an instance of Device or of DeviceComponent, which enables to describe all medical devices used to produce the observation: analyzer, reagent, and more.
Ralf Herzog (Jun 27 2018 at 15:13):
Hi François,
your answer sounds reasonable, but I still have the problem how to put into a 0..1 device (where Description says "(Measurement) Device") 0..* medical devices used to produce the observation (the used reagents (if more than one) are not in a part of relationship to each other but exist on the same level (same for used disposables if more than one is used).
Or am I completely wrong here?
Eric Haas (Jun 27 2018 at 16:20):
Devices have components use DeviceComponents or create an extension to Nest Device
François Macary (Jun 27 2018 at 16:23):
Yes, that would work, considering the reagent(s) used as components of the analyzer, even if they are not components stricto sensu.
Eric Haas (Jun 27 2018 at 16:29):
they are like the ink in an inkjet printer and if they came a little cartridge it would be more obvious that it is a component device.
Grahame Grieve (Jun 27 2018 at 19:55):
I'm confused how this information characterises the observation itself. what's an example?
Eric Haas (Jun 27 2018 at 20:34):
I think it characterizes the methodology/instrumentation
Andrea Pitkus, PhD, MLS(ASCP)CM, CSM (Jun 27 2018 at 22:57):
Jose, fyi, consumables and reagents are typically not tracked/messaged in HL7 for the laboratory like it is for pharmacy or supply chain areas.
Jose Costa Teixeira (Jun 28 2018 at 09:18):
Devices have components use DeviceComponents or create an extension to Nest Device
We are redesigning Device to support nested devices, so that could be an option.
We could see consumables as "components" of the performing device... this may well support(not sure how, but i'm confident) the downtream requirements including resupply, inventory and consumption tracking, etc.
Ralf Herzog (Jun 28 2018 at 17:35):
I think I might think here either to complex or I am simply not getting it.
So, how to get device with the "part-of" relation to model two reagents plus one disposable to be in observation?
Eric Haas (Jun 28 2018 at 19:17):
Screen-Shot-2018-06-28-at-12.16.10-PM.png
Ralf Herzog (Jun 29 2018 at 12:58):
Hi Eric,
your solution is not working for my problem, because
- A specific device has multiple reagents
- A specific device could perform multiple different codes/orders/tests where specific different reagents are used (i.e. two out of 100).
=> so the relation is between the reagent and the code/order/test not between the device and the reagent.
With your model I would need multiple devices with a limited validity where each device could perform one specific code/order/test with one valid combination of reagents. I do not think we should have a code and reagent specific device for a small set of observation for a test.
It would be possible to model that, but then the device would exist only for a short amount of time (as long as all the used reagents are valid and available) for one code/order/test. From there I could reference to the subdevice from where I could go to the device, ...
But: there is no elegant and easy way to trace from the observation on a specific (sub-)device with observation specific reagents to the observation specific reagents/disposables with your solution.
I think we need a "substance" reference (0..1) inside the observation - or a laboratoryObservation which would include this.
Ralf Herzog (Jun 29 2018 at 13:12):
or as diagram:
pasted image
Eric Haas (Jun 29 2018 at 13:41):
Have you asked anybody in the Devices group @Hans Buitendijk is this a livd thing?
Daniel Rutz (Jul 06 2018 at 14:03):
Not directly a LIVD thing, but related. Devices WG should probably be included too, monitoring systems probably don't have quite the same issue with consumables most of the time so I don't know if they'll have any specific solutions. I agree with Ralf that consumables/reagents as device components doesn't instinctively feel like it's quite correct; Substance seems plausible but might be too generic as-is (we'd like production/lot info in the reagent metadata) and there's still the question of how to reference it. I think it reagents need to be separate from the device, not nested in it (an observation/test result could be generated by a human doing some chemistry too, and then there's no automation or encompassing device but we'd still want to be able to document the reagents used).
Ralf Herzog (Jul 10 2018 at 13:50):
@Hans Buitendijk like to asked for a Resource Model (in an entity relationship kind of notation):
ResourceModel.png
Last updated: Apr 12 2022 at 19:14 UTC
