10 Jan 2019 agenda Allergy CRG meeting

Hi all,

Was really an interesting discussion earlier. Would just add one point here as food for thoughts.

Russell mentioned that allergy alerts are overridden in 90% of the time. Perhaps it could be due to inaccurate (more likely imprecise) documentation of the allergy in terms of the reactions.

In clinical practice, I’ve seen my fair share of allergy documentations listed the reaction as non-immunological. Most common example, diarrhoea. I noticed in Jim’s list of reaction list, probably in top 20 or so, has “stomach pain”.

So, maybe apart from having a refset of “true” allergy list, what is needed is a way to distinguish between true allergy and adverse reactions. I'm not sure if SNOMET CT would be the best place to do so or it is local implementation in software.

Nevertheless, there should be a placeholder to document the reactions to drugs (or other substances) regardless if allergic or adverse effects. Clinicians can then make clinical decision on the treatment plans.

Allergy alerts would only be triggered if the allergy is indicated as immunological while those listed as adverse reactions could be more of a passive reminder. Maybe this can minimise alert fatigue.

I believe others have similar ideas too.

Cheers,

Nikki

Nikki, as Jim pointed out, in the EPIC system, there is a list of reaction types which include e.g. in my implementation, allergy, intolerance, contraindication and unspecified as well as a longer list of specific reactions which bridge the spectrum from true allergy to non-immunologic symptoms and disorders. Thus, the reaction is linked in the record to a reaction type such as allergy or intolerance. 

Furthermore there is also a top level hierarchy of Propensity to adverse reactions which subsumes intolerance and hypersensitivity. Hypersensitivity includes allergy (the hierarchy we are trying to clean up) as well as non-allergic hypersensitivity (pseudoallergy).

Bruce

Hi

In the system we are building with our EHR provider we are turning our (old flat locally coded dictionnary-linked)  "allergy box" more into an "adverse effect box" where people can capture the substance and the manifestation, those two being the only mandatory fields, with the possibility to set the manifestation to vague levels or even "unknown" for the old things coming from the patient's childhood he's not sure about. There will be, linked to that substance in the model, a field with the degree of certainty of the information, one with who reported it (physician or the patient or a family member for ex). Linked to the manifestation there will be a severity field but to be treated with caution as it's subjective what light, mild or moderate can mean for one persone and for another. The idea is that only severe and life-threathening are likely to be the really reliable degrees of certainty. Then there will be a field for the mechanism uinvolved, default set to adverse reaction in general and where one can switch to "allergy" if they deems the process to be of allergic nature, or intolerence, or non allergic hypersensitivity, etc. That's for capture. And people will be expected to capture what is said, observerd and not to interpret if they have not tested it. So an "allergy" will be reported to the product as closely as possible to what the patient actually took, brand and form included not to the active ingredient of this product, till more info is available because he saw the allergologist.

Then the question of what triggers an alert is to be solved in the decision support system, depending on the care context and the care provider. Some things like life threatening reaction will always trigger alerts as to the safety of the product or its components untill the specific ingredient involved has been identified. Some things will trigger care pathways to rule out or confirm that it's an allergy (latex, penicillin for ex), intolerence (lactose test), etc. And a big bunch will need to be defined by the hospital policy who will choose where he sets the alert line. Likely since we have almost no alerts right now, poeple will at start a lot of things to pop-up "to be sure it's safe". Then they'll get bored by the pop ups but in the mean time we'll have gotten data of what gets overridden and why and progessively we'll be able to cull out what's not needed, what one needs do be aware of even though it will be overridden (administred with added surveilance, and that can be turned into a suggestin of care pathway not an alert pop-up) and what can go out forever into oblivion as alerts.

That's our plan. Now I'll tell you in a year or two how it turned out in real life.