Stream: implementers
Topic: AllergyIntolerance - code and reaction.substance
Jose Costa Teixeira (Oct 21 2021 at 08:08):
Q1. I presume that AllergyIntolerance.code should/shall contain the code for the allergy and not the code for the substance i.e.
91937001 |Allergy to seafood (finding)
and not
44027008 |Seafood (substance).
Is this correct? And how do we handle things like the code for reaction
15920401000119102 |Allergic reaction caused by seafood (disorder)?
or different granularities e.g.
419972009 |Allergy to shrimp (finding)
Jose Costa Teixeira (Oct 21 2021 at 08:15):
Q2. What is the reason that reaction.substance is called "substance"? From a reporting perspective, we notice the reactions to products, not to substances. We actually may not know if it's a substance.
example: a child has a reaction when taking a specific amoxicillin product, which contains strawberry syrup. what would be the substance?
I'd say it's amoxicilin because that's as much as we know
Perhaps it is to the amoxicilin, perhaps it is to the strawberry syrup. Or perhaps it is to the maltodextrin in the syrup.
But that would be reflected on the AllergyIntolerance.code as the allergy is investigated, but not in the reaction.
Jose Costa Teixeira (Oct 21 2021 at 08:15):
Is the above analysis missing anything?
Peter Jordan (Oct 21 2021 at 09:04):
Q1: There is a comprehensive explanation in the specification
Q2: The clearest explanation that I've heard from clinicians is that an allergy is the risk of an adverse reaction if exposed to a substance based on one, or more, previous adverse reactions to that substance. However, I do fully understand that it's often not possible to be certain what the exact substance is if the adverse reaction resulted from exposure to a medication or foodstuff containing multiple ingredients. The seafood example is a favorite of @Russell Leftwich who regards it as a type of restaurant, rather than an allergen.
Jose Costa Teixeira (Oct 21 2021 at 11:06):
Q2: So should this be element really be called "Substance"?
Russell Leftwich (Oct 21 2021 at 12:30):
Scientifically it is certain that individuals react to a single substance, almost always a protein when it comes to foods, and usually a metabolite when it comes to drugs because drug molecules are too small for your immune system to "see" them until they combine with some tissue protein to create a larger molecule. Penicillin antibiotics, like amoxicillin, are reactive by design. That's how they kill bacteria. Inert ingredients, fillers, in oral medications and flavorings are by design non-reactive, non-allergenic substances. The FDA in the US will not approve inert ingredients, colorings or flavors if they can cause an adverse reaction. There is zero chance they are causing the reaction. Although there are urban myths about reactions to food colorings and flavorings, there are no documented cases of which can be reproduced (an anecdotal report is not scientific documentation).
@Peter Jordan is correct: seafood is a type of restaurant. People do have allergic reactions to organisms that live in the water, but it doesn't have to be saltwater or seawater, and it is specific for a particular species or genus. The most common is shrimp, and those individuals often react to other crustaceans. It is caused by a specific protein in crustaceans. Other people react to mollusks, clams and oysters, etc, due to sensitivity to a different and unrelated protein. And a few people react to a protein in the muscle of bony fish, salmon, trout, cod, etc. It is possible to be allergic to all 3 but not very common. Likely to react to different members of the same genus -- shrimp and crabs for example.
And I agree with @Jose Costa Teixeira that it makes little sense that we list the substance and not the reaction, but that is the way we do it. It should really be both, but no one listens to me except Peter.
Jose Costa Teixeira (Oct 21 2021 at 12:49):
I think this element name is misleading
Russell Leftwich (Oct 21 2021 at 13:02):
Agree it is not well defined and only distinguishes from physical causes of adverse reactions like radiation, thermal, and vibration.
Lloyd McKenzie (Oct 21 2021 at 13:20):
@Jose Costa Teixeira What do you think it should be called? The key thing is that the AlleryIntolerance.code tries to get as close as possible to the overall cause of reactions, while reaction.substance is "what were you exposed to on that particular occasion?"
Jose Costa Teixeira (Oct 21 2021 at 14:04):
I'd think "product" or "agent" would be more appropriate. "substance" seems to be indicating that it's the actual specific substance
Jose Costa Teixeira (Oct 21 2021 at 14:05):
causativeAgent?
Russell Leftwich (Oct 21 2021 at 14:13):
Product is not workable. People do react to a substance in the product, even if I don't particularly like that term. CDS can't deal with "product" and there are numerous examples of people who had serious reactions to 500 mg amoxicillin when 250 mg amoxicillin was on their allergy list, but that is a different product.
Lloyd McKenzie (Oct 21 2021 at 15:08):
CDS really shouldn't be triggering off the reaction.substance - it should be triggering off the AllergyIntolerance.code. reaction.substance is there to help understand the evidence for the assertion of AllergyIntolerance.code.
Jose Costa Teixeira (Oct 21 2021 at 15:48):
Right. In my understanding, this element should contain "whatever we know triggered the reaction" - more or less granular. If we know that the patient developed a rash after taking this brand of amoxicillin, one may capture that, and not "amoxicillin". Actually I don't think it is about amoxicillin. It can be the clavulanic acid, or the syrup
Russell Leftwich (Oct 21 2021 at 17:13):
@Jose Costa Teixeira products contain multiple substances (or ingredients or however you want to label the components). In 99 + percent of cases it is only one of those substances/ingrdeients that is responsible for an allergy'intolerance reaction. Other products that are ordered by a prescriber (or chosen by the patient/individual in the case of OTC or foods) may contain the RESPONSIBLE substance/ingredient/component.
You cannot protect patients/individuals, which is the ultimate intent of this data, by recording "product". If it is a generic product, often the case for commonly prescribed treatments, it is not reasonalble to expect CDS to determine what is in every "product" that may contain the responsible trigger of the adverse reaction. Not even true in a single jurisdiction, let alone globally
Russell Leftwich (Oct 21 2021 at 17:20):
@Jose Costa Teixeira if a product contains more than one active ingredient, then it is possible that more than one should be represented. But in terms of scientific knowledge, clavulinic acid has approximately zero chance of causing an adverse reaction and exactly zero chance of causing an immunology reaction, if that is the way you define 'allergy". The mechanism of adverse reactions is not commonly understood, but for for most such reactions it is established scientific fact. Irrefutable.
Jose Costa Teixeira (Oct 21 2021 at 17:45):
Russell Leftwich said:
You cannot protect patients/individuals, which is the ultimate intent of this data, by recording "product". If it is a generic product, often the case for commonly prescribed treatments, it is not reasonalble to expect CDS to determine what is in every "product" that may contain the responsible trigger of the adverse reaction. Not even true in a single jurisdiction, let alone globally
I can't conceive a reaction to a prescribed product :) I can imagine that a reaction happens when a patient takes a product, whether it was prescribed or not. In other words, a patient may be prescribed amoxic+ clav acid; they may be dispensed a specific brand; and they may even be allergic to amoxicillin, but what we know is that reaction happened to that product that was ingested.
Jose Costa Teixeira (Oct 21 2021 at 17:53):
I don't think this data element should be trying to solve the complexity of product composition and formulation. What we know is that the patient took something (which is physical, not an abstraction). There may be a reasoning process after that to assert what the suspected specific substance is. But the reaction is not containing that, I think.
Jose Costa Teixeira (Oct 21 2021 at 17:59):
My assumption is that AllergyIntolerance.reaction captures the reaction event, not the clinical reasoning process. I'm not dismissing that knowledge, quite the contrary - I am trying to see how to capture the data that triggers that process.
I do think that Product is not good enough, but not because people may have different understandings of what a "product" is. To me it is something that is produced (as in "brought into existence"). And that may not fit the bill.
Agent would be a better term, I think.
Jose Costa Teixeira (Oct 21 2021 at 18:00):
(I use the terms based on my latin understanding of these words and their etymology, but current English may have given them different meanings)
Russell Leftwich (Oct 21 2021 at 19:48):
an individual reacts to one specific molecular entity. Call it what you want. That entity may be in multiple products. That individual would be expected to have a reaction to any one of them. There is no global listing of product components. Knowing only the product name or code leaves that individual at risk for future exposure and reaction.
Jose Costa Teixeira (Oct 21 2021 at 19:59):
That is not the discussion and i don't see the point of bringing the global listing of products here. That's exactly what I am avoiding.
Jose Costa Teixeira (Oct 21 2021 at 20:01):
I presumed that the AllergyIntolerance.reaction.substance could be used to capture that on this date, the child had a reaction to that medicinal product (although we do not yet know which specific substance actually caused the problem).
Jose Costa Teixeira (Oct 21 2021 at 20:02):
If that is not the case, what is the reaction.substance for?
Jose Costa Teixeira (Oct 21 2021 at 20:03):
The reaction was also interpreted as "a description of what may happen if the patient is exposed to a substance'
Jose Costa Teixeira (Oct 21 2021 at 20:04):
So there seems to be some ambiguity in the information
Jose Costa Teixeira (Oct 21 2021 at 20:04):
As an implementer I find this resource could be more clear
Jose Costa Teixeira (Oct 21 2021 at 20:05):
And I suggest reaction.substance to be renamed to agent. I would be happy to join a call if that is considered useful
Jean Duteau (Oct 21 2021 at 20:22):
at the risk of stepping into a conversation for which I don't have any issue, @Jose Costa Teixeira you really need to go make a Jira issue with your concern and suggestion so that the Patient Care group can discuss.
Jose Costa Teixeira (Oct 21 2021 at 20:41):
@Jean Duteau sure, I intend to, if this discussion goes somewhere. I thought the resource attribute was a bit unclear, now I have more information but still lack clarity
Jose Costa Teixeira (Oct 21 2021 at 20:43):
Patient Care calls are at a very inconvenient time for me. And perhaps I would be missing something or perhaps some public discussion helps me see what could be the issue.
Jean Duteau (Oct 21 2021 at 20:43):
putting my incoming TSC vice-chair hat on, i know that your first inclination is to come here, but for issues like these, they really need to be Jira issues first and be discussed at the Workgroup. As you know, not every member of the various HL7 WGs reads Zulip faithfully so you are getting a very limited participation. Discussions should really start in the workgroup.
Jose Costa Teixeira (Oct 21 2021 at 20:44):
I disagree (discussions should really start anywhere IMHO), but that is also not the discussion I was looking for.
Jose Costa Teixeira (Oct 21 2021 at 20:53):
So, question from implementers:
Q2: Since AllergyIntolerance.reaction captures the adverse events, which can be reactions to several causative agents (e.g. pharmaceutical products), and assuming that at the time of capture, we do not know which actual substance caused the event (Amocixilin, Clavulanic acid, strawberry, maltodextrose), do we really want to use the word "substance"? Can I not capture the product?
If not, where do I capture that, and why does the text reference "pharmaceutical product" - and what is the meaning of pharmaceutical product in that sentence?
Jose Costa Teixeira (Oct 21 2021 at 20:57):
(For comparison, AdverseEvent.suspectEntity refers to Medication, which can be "Augmentin children" which is a product, not a substance)
Peter Jordan (Oct 21 2021 at 21:01):
Good discussion and typically sagacious contributions from @Russell Leftwich . We have to 'follow the science' if this resource is going to play a role in reducing harm. While I agree with @Jean Duteau that the Patient Care group should discuss this, I do believe that it's worthy of broader input from the community as a whole. As a layman, I see generic (and even active ingredient) prescribing as a necessary step away from linking adverse drug reactions to specific trade/branded products.
Jose Costa Teixeira (Oct 21 2021 at 21:03):
I agree the WG should discuss this too. I just don't think the discussion should start there. I'm also a layman and conveying the misinterpretations that the resource can cause (for example Q1 was unclear to me, but I learned that quickly. Q2 is less clear now)
Jose Costa Teixeira (Oct 21 2021 at 21:05):
I have a simple example (Augmentin Pediatric caused an allergic reation) and I don't know how to tell my stakeholders whether we need an extension in the AllergyIntolerance to capture that.
Jose Costa Teixeira (Oct 21 2021 at 21:38):
Peter Jordan said:
I see generic (and even active ingredient) prescribing as a necessary step away from linking adverse drug reactions to specific trade/branded products.
I don't understand this, esp the "prescribing as a step away" part. I don't see what would be a role for prescription in this. Do you mean something like "representing products on generic and even active ingredient levels, so that when capturing adverse reactions, we can actually point to the substance itself"?
Peter Jordan (Oct 21 2021 at 21:47):
It's about preventing harm and, of course, accurate recording of previous adverse reactions is an important part of that. Russ' practical example about recording at the atomic 'product' level makes this point far better than I can.
Jose Costa Teixeira (Oct 21 2021 at 21:49):
I admit I'm lost. We all want to prevnt harm. I want to prevent harm by capturing an allergic reaction to the strawberry syrup, which is not an atopic product level.
Jose Costa Teixeira (Oct 21 2021 at 21:50):
Perhaps there's a simple way of explaining this to implementers but I don't get it.
Russell Leftwich (Oct 21 2021 at 22:51):
At the point that something is an entry in an allergy/intolerance resource someone has made a clinical judgement that there is a future risk of a reaction to a molecular entity (whatever you want to call it) if exposed. The risk is exposure to that molecular entity, whatever product it is in. As far as a reaction, it is true that at some point in time there may only be a temporal relationship between administration of a product and some sudden physiologic change that appears to be caused by the administration. Clinical reasoning is involved in determining whether it is side effect, overdose, allergy/intolerance reaction, wrong route of administration, contaminated 'product, or something else. in the first 3 cases it is still the so called active ingredient, not the product.
"substance" has a long history of usage outside of FHIR and outside of data standards. I have never liked the term, but I don't think changing it in the FHIR spec seems appropriate.
Richard Townley-O'Neill (Oct 22 2021 at 05:09):
@Jose Costa Teixeira I think that some of the resistance to your suggestion of using "product" instead of "substance" is that for pharmacists "product" is as much a technical a term as "substance" is.
Maybe a word that has no technical meaning to clinicians is requited. Maybe "stuff". :smile:
Jose Costa Teixeira (Oct 22 2021 at 05:17):
I suggested "agent"
Jose Costa Teixeira (Oct 22 2021 at 05:47):
And I agree, "product" is a loaded term. Not only for pharmacists but also in supply, etc.
Jose Costa Teixeira (Oct 22 2021 at 05:59):
My suggestion is to first find the meaning and then find a term.
For me, reaction.substance should mean "whatever is understood to have triggered the reaction, either a specific substance or an actual product or item, to whatever specificity level it is known". That is my understanding.
Perhaps what @Russell Leftwich means is "the actual specific substance that has been judged to have caused the reaction"
I understand this to be actually more appropriate for AllergyIntolerance.code, not for the reaction event history.
Jose Costa Teixeira (Oct 22 2021 at 06:01):
Reaction is an event and as such I think it does not capture the clinical inferrence and evaluation that occurs from the event.
Jose Costa Teixeira (Oct 22 2021 at 06:04):
In any case, I think this discussion would be helped by a story board, to see the evolution from "Child reacts to Augmentin Pediatric" to figuring out that the child reacted to the strawberry syrup, not to amoxicilin. At least I'd value that.
Lloyd McKenzie (Oct 22 2021 at 17:21):
I don't understand you concern with 'substance' @Jose Costa Teixeira. It's just a generic term. It could be an ingredient, it could be a manufactured product. It could be "seafood chowder I had for lunch". I think you're reading in more to the name than is there. "substance" here in no way limits you to the Substance resource.
Jose Costa Teixeira (Oct 22 2021 at 19:08):
If substance means "whatever is understood to have triggered the reaction, either a specific substance or an actual product or item, to whatever specificity level it is known", I can handle that - although that should be perhaps clearer, i.e. we must make sure the WG agrees with that definition. I don't know if that is agreeable. If so, then yes, I am reading too much into the word "substance" (and I think "Agent" would be more accurate)
Jose Costa Teixeira (Oct 23 2021 at 10:04):
2 conclusions from my side:
- Q1 is clear, I'll update our documentation
- Q2 is less clear now. Should I create a ticket to discuss the use case for allergic reaction to augmentin pediatric? This is really impactful and I think if we need to make it better, then we should do it now. Some people spend many years thinking they have an allergy when they had an allergic reaction to something else and this is not documented clearly. We can't fix that for all of them, but perehaps we can make things right from now on.
- I think we should have broader example scenarios in our spec when we want to go normative, to check for functional silos.
Rob Hausam (Oct 24 2021 at 23:06):
Doesn't this definition text for AllergyIntolerance.reaction.substance make the intent (and the boundaries on that) rather clear (including how it applies in @Jose Costa Teixeira's Augmentin Pediatric scenario)?:
Identification of the specific substance (or pharmaceutical product) considered to be responsible for the Adverse Reaction event. Note: the substance for a specific reaction may be different from the substance identified as the cause of the risk, but it must be consistent with it. For instance, it may be a more specific substance (e.g. a brand medication) or a composite product that includes the identified substance. It must be clinically safe to only process the 'code' and ignore the 'reaction.substance'. If a receiving system is unable to confirm that AllergyIntolerance.reaction.substance falls within the semantic scope of AllergyIntolerance.code, then the receiving system should ignore AllergyIntolerance.reaction.substance.
Last updated: Apr 12 2022 at 19:14 UTC
