Let’s see if we can get some more discussion going here! This is a great opportunity for our entire community to engage with this process. The goal is to begin to work towards not to a vague goal of “standards”, but what I think of a “standard use of standards”, i.e. we all use the standards in the same way so we can have true interoperability. Hashing things out together is a great way to do that.
I’m interested in what people think about the allergy/ADR domain of this step of the NZCDI:
Work is already going on to improve the current NZ Medical Warning System. This includes work on a spec for Adverse Drug Reactions. Although it’s gone on hibernation for the time being as well, here’s what we were up to:
Basically, in addition to the substance and reaction codes as per the NZCDI proposal, this resource also includes Clinical Status, Verification Status, Category, Onset Date, Recorded Data, Recorder, and the possibility of a free-text Note.
This was based somewhat on the Australian CDI process and what they set as the base for their Allergy/ADR standard. It’s still not everything that we might want in a standard, but keep in mind that this is just a first step. So the question is: should we beef up the NZCDI standard to match what’s likely to go into place with the MWS anyway?
Without the AllergyIntolerance criticality and type - plus the Reaction substance, onset and severity - I don’t see how this would enable anyone to distinguish between a genuine, and critical, clinical allergy and a mild intolerance.
I agree with Peter - yes, go for more beef where possible, but it is hard to see much utility with a reduced number of fields. However, I appreciate that this represents a ‘lowest common denominator’ and the aim is practical interoperability between variable quality systems and not the ideal EPR data.
As a coal-face clinician, from this at best I could hope for a ‘heads up’ that there might be an allergy / adverse reaction problem slightly more reliably than the current poorly connected reality.
What I’d much prefer was proper structured data in this domain, with a single source of truth. For an example of the ‘ideal’, the latest OpenEHR archetype is educational:
FWIW, the FHIR Resource AllergyIntolerance was co-designed by a collaborative group that included representatives from both openEHR and HL7 - plus clinical domain experts.
It is important to keep in mind that the NZCDI process will be iterative - as systems gain abilities and maturity, we will be able to enhance the CDI. This is the case across all the categories, not just adverse reactions.
@pkjordan, to respond to some of your specific comments. Issues arise when trying to balance clinical pragmatism/practicality and academic rigour/purity. Honestly, I think the high-end specifications tend towards the latter - done by allergists in systems where there is easy access to formal testing (or at least the specialists think there is). That is not close to the reality in most health care systems, and certainly not in NZ. Many of the elements not included in the spec I posted above were taken out because they are too subjective. How severe is a reaction? Would you rather have severe fatigue or mild Steven-Johnson syndrome? Type is also quite difficult to determine. Anaphylaxis is an allergic reaction and diarrhoea is (usually) an adverse reaction, but there are all kinds of rashes caused by substances that can be allergic or non-allergic. Thus what some other clinician thought a reaction type was years ago when it was entered, with no other information available, is simply just not very useful at the point of care. If we had a team of specialists who could descend on a patient every time a new reaction was recorded and use their specialist knowledge and tests to determine the exact nature of the reaction, so it could be fully recorded in an ideal data model, would be fantastic. But that’s obviously not going tot happen. Therefore, simply stating what the reaction was, allowing the clinician to have a conversation with the patient, is entirely sufficient to deliver good care.
I understand you position, Matt - but that just leaves us with data (i.e. an unquantified risk) that has little use beyond the immediate purposes for which it was captured.
Yeah I get that. But at this point I’d be happy just to reliably have data just for the purpose it was collected. From there we can expand the secondary use possibilities. We have ways to go with workflow and UI before we can expect clinicians to be capturing data beyond the immediate purposes without causing extra burdens.
As a layman, I would argue that information about severe reactions to a substance and a high risk if exposed to that substance in the future are critical entries in a patient record for life. NZCDI should be aspirational, not a representation of the lowest common denominator. Otherwise, how are we going to progress?
As a clinician, I greatly appreciate your faith that clinicians would be able to accurately and reliably both record and use information about severity and risk. I note that in the openEHR archetype and the FHIR model, the “criticality” element has an “indeterminant” option. I would posit that the overwhelming number of entries into a system would be labelled as such. And when it wasn’t, it would be obvious. A reaction of “anaphylaxis” or “Steven Johnson syndrome” (a reaction in which basically all your skin falls off and you have to be cared for in a burn unit) are by definition “severe” and “high risk”. There is no value in adding those labels. Meanwhile, what if someone has a reaction listed of “nausea” that is “severe” and “high risk” - what am I going to do if the overwhelmingly best drug to treat them with cross-reacts with their allergen? I’m going to have a shared decision making discussion with them - which is exactly what I would do anyway if those labels weren’t there. The point is that those labels don’t add much utility to clinicians in their decision-making but add complexity and difficulty to clinicians entering the information. The most useful information is a description of the reaction. The extremes of severe and mild are easy to identify, and everything in the middle would be subjective and unreliable, and thus require a conversation with the patient. Which is going to happen anyway.
I do appreciate there are other research and surveillance reasons to have more information. But personally I believe it is more useful to build a pragmatic solution first, then expand it once it’s established. And I would argue that that is the aspiration of NZCDI. We’re not trying to create a perfect, complete, crystalline model of health data on the first round. The process of refinement and improvement is the goal.
Interesting and insightful reply @MValentine . However, I’m not entirely sure that what is obvious to a clinician recording an event is necessarily obvious to another healthcare worker in a future encounter with that patient. Might it not be helpful to have the record of a patient who has severe reactions to a food type or medication have the related Allergy record marked as ‘critical’ so they are not exposed to the relevant substance in an emergency, unplanned or cross-border situation?
Echoing @MValentine the first release of NZCDI is deliberately a modest beginning, @pkjordan. We’re calling out the central data requirements rather than extensive data requirements, recognising that we’ll be making considerable headway to get everyone using SNOMED CT to code health conditions, for example.
