Medication models: Where does "route" belong?

Currently, we model the medication route in the INSTRUCTION.medication_order archetype. FHIR, however, places this element inside their Dosage resource. I’m struggling to understand why they would place it there, as most orders (I’d say 99% if not 100%) are for a single route or at least the same set of possible routes for all dosages/timings, while there may be a significant number of “Dosage” structures within one order.

Could we get some input from clinicians, implementers and others about this? Are there any orders where the route differs between dosages/timings? Is it a problem that we’ve structured this element differently between openEHR archetypes and FHIR resources?

I would suggest always to model the clinical normal case faithfully. Sub-optimal message standards can always be compensated for at a technical level, and will always change in the long term.

If there is no strict clinical imperative nor any implementer issues identified and if there were some other possible refinements for the models that required a new version, what are the pros & cons? Philosophically, should we consider aligning with the FHIR model?

Philosophically, I have no objection to aligning with FHIR resources, in fact I agree that we should do so where possible. However, I think this is one area where we have the structure correct, for the reasons that @siljelb gave - the route is associated with the whole order, and in fact the dosage structure would not be used in the context of a meds administration at all.

We have a set of models that are certainly not perfect but they are being used daily in real-world prescribing systems both in primary and secondary care settings, not just for information exchange. Always prepared to change things - break if needs be!! but I’m not convinced by this one.

1 Like

Super keen to get input from implementers who are everyday dealing with mappings/transformations. Are these questions important to consider re compromising our modelling, or should we just keep doing our openEHR ‘thing’.

1 Like

well the value of the models is the value of the data, and the querying, reporting and analytics that will be running over that data for years. Compromising the semantic integrity of models for the sake of aligning with today’s message standard (which will be replaced in a decade, either by new versions of itself or something else) doesn’t make sense to me. If the design of a FHIR resource is clinical ‘right’, sure, no reason for everyone not to use the same model. But otherwise… (imagine if we had followed this logic with CDA or HL7v3…)

1 Like

We have discussed this in our modeller meeting and we can’t think of any use cases where orders are given with different routes depending on dosage. The one exception is initial doses, for example misoprostol for induction of labour, where the initinal dose may be administered vaginally and subsequent doses may be given orally.
However this would usually be handled as a separate order, as part of an order set.

1 Like

There are other situations such as an order for antiemetic to be given prn, either orally or IM depending on whether the patient can tolerate an oral medication without throwing up. In the current modelling scenario, it would require two separate prn orders, one for each route.

The clinical purist in me, with a medicolegal bias, would prefer amount, timing and route all to be found in one place. Software could manage the apparent replication of the route, but in fact it would be more faithful in recording the deliberate clinical intent.

It would be interesting to investigate if separation of the route into a different part of the model could contribute to a clinical safety issue unless the software reminds the clinician to check the assumed route.

Always worth checking we’re in agreement. This is our modelling intent, but given the ubiquity of FHIR, working out where to draw the line where alignment turns to compromise can be grey sometimes and how to balance pragmatism vs purist approaches.

Anyone disagree?

I don’t think this would be necessary. The “Route” element in INSTRUCTION.medication_order has occurrences 0…*, so you could put in as many alternative routes as you wish.

As long as the assumption holds that 99% of orders have the same single route or same set of alternative routes for all different dosages and timings within that order, I don’t see how bundling route with dosage would be more faithful to the clinical intent.

It could be interesting to see the justifications for modelling it the way they have in FHIR. Anyone know where we can do that? Or who we could ask?

Different routes often require different doses - oral vs IM analgesics/antiemetics will require 2 orders.

Just as we record the actual term a clinician chooses from a value set, even if a preferred term is stored and used for persistence, I’m still curious about the investigating the medicolegal issues re ensuring that the complete order for how a drug should be administered needs to be explicitly recorded.

Also understanding the provenance of our model - where else is this design pattern used in parallel standards or are we isolated here?

I’m assuming we’re talking hospital settings here? When you’re into PRN injectable analgesics/antiemetics they would normally be titrated, not given according to a set dosage.

Sure. Do we have any legal frameworks to lean on in this level of detail?

Good question. The “Route” element was added to the INSTRUCTION.medication_order archetype by @ian.mcnicoll on 29 Oct 2015, with this set of references:

openEHR Foundation Medication archetypes

NEHTA's Therapeutic Good Use Data Group from the NEHTA Website

Intermountain Healthcare Medication order model, Personal Communication to Sam Heard by Dr Stan Huff.

Royal Australian College of General Practitioners. Fact Sheet: Medicines List. 2010.

At that time, FHIR was at DSTU2, and had Route as part of the “DosageInstruction” group within the MedicationOrder resource: I.e. similar to today’s pattern only they’ve divided it up into separate resources. I don’t know how to find the references for FHIR resources.

The ‘complete order’ from a legal /safety perspective has to include the medication name and form and anything else, so you could argue that the FHIr Dosage is more disconnected in that sense.

THe background to this work was all pre-FHIR and came out of a whole bunch of prior art UK GP2GP , the NEHTA work, C-CDA, som nice data models out of University of Dundee and a lot of practical feedback from implementers.

I can sort of maybe see an equal argument for doing it the FHIR way and some potential downsides as discussed above, but I can’t see any compelling reason for changing things any time soon. The differences are trivial from a tech perspective, in terms of mappings. Once the full set of models are constructed in a template the final templates/profiles would look almost the same, and our models have been thoroughly road-tested in terms of applicability to prescribing systems, not just data exchange. We have many, many patient records running against the current models.

and FHIR is still a bit of a movable feast (which I agree with BTW), so may change again. Just don’t think this question is worth much time or effort right now. If the mapping was awkward that might be different but it is really pretty trivial to adjust between the two sets of models.


Generally speaking, if it ain’t stated in the documentation of the specific resource page in FHIR, then it ain’t true :wink:

I found this discussion in the FHIR Jira:

I’m not sure I’m able to follow it, but to me the argument for moving ‘Route’ to ‘Dosage’ seems to be “it was this way in HL7v2”. :woman_shrugging:t2:

THe examples that are in that discussion we can actually support as we do allow multiple routes.

Examples provided by Scott:

An example of PO/IV

Ondansetron 8mg orally or IV twice a day as needed for nausea

Really 2 orders:

Ondansetron 8mg orally twice a day as needed for nausea


Ondansetron 8mg IV twice a day as needed for nausea (ignoring for the moment that it should be given over 30 min)

Here's a classic:

Compazine® (prochlorperazine) 5-10mg PO or 25mg PR bid prn n/v

And as 2 orders

Compazine® (prochlorperazine) 5-10mg PO bid prn n/v


Compazine® (prochlorperazine) 25mg PR bid prn n/v

What we do not support is differential dosage/site/method instructions for each route. There are definitely some use cases for that but I have also seen guidance that suggests that this is not good prescribing practice, and I’m pretty sure I would have got a b*****ing from the nursing and pharmacy staff if as a junior I had tried to combine them, at least in a written ward medication order!! THey would have said, create separate orders for each route.

It feel unsafe to me but I think we need to seek guidance from pharmacy colleagues on best practice.

1 Like

There is one (very common) medication in perioperative medicine which is frequently ordered with multiple routes: Paracetamol

A typical adult prescription is Paracetamol 1g PO/IV/PR QID, with potentially severe consequences if it is administered simultaneously via different routes.

1 Like

Thanks Nathan! Paracetamol is sometimes ordered with multiple routes even on regular wards, where the idea is that the nurse chooses the most appropriate route at the time of administration. But I’ve never seen it ordered with multiple routes in one order, with different dosages based on the route?

Ah yes - I think that I misunderstood the problem sorry. We are talking about one particular situation - low bioavailability medications, where the amount of absorption of the drug differs significantly depending on the route.

This is tricky clinically, and even trickier to cope with on the information tech side. The bioavailability (which is usually estimable) really needs to be included as part of the model, because what matters from the safety perspective is the amount absorbed by the patient.

For example, a patient receiving 20mg morphine PO along with 5mg morphine IV is much more likely to be in good shape than one receiving 20mg morphine IV and 5mg PO!

In clinical practice, we do tend to only prescribe one form at a time with these meds for that safety reason. But it sure would be nice to be able to confidently prescribe these drugs via different routes and differing doses. This is however a 1% thing, as you say.


Thanks Nathan, that makes sense. I’d be interested to see the current recommendations for handling this - two separate orders or one rather complex order? Btw, are these often regular orders, or usually PRN?