Stream: implementers
Topic: probability of condition.verificationStatus
Erich Schulz (May 21 2016 at 01:54):
so... I'm looking at condition.verificationStatus
Erich Schulz (May 21 2016 at 01:56):
the options are: provisional | differential | confirmed | refuted | entered-in-error | unknown
Erich Schulz (May 21 2016 at 01:57):
refuted, i guess means p=0, confirmed means p=1
Erich Schulz (May 21 2016 at 01:57):
I would read "provisional" as meaning p > 0.5
Erich Schulz (May 21 2016 at 02:00):
the notion of "differential" is more complicated (and as stated means this item is one of mutually exclusive set of diagnosis) - but there isn't a mechanism to define the set
Erich Schulz (May 21 2016 at 02:02):
but my real question thought about how to put a bit more finnesse in the list especially in the expression of probability
Erich Schulz (May 21 2016 at 02:03):
a possible value would kinda close the gap but its not feeling tidy either
Erich Schulz (May 21 2016 at 02:05):
ie possible ~= p > 0 and < 0.5
Grahame Grieve (May 21 2016 at 02:06):
provisional and differential are more about the state of the assessment process than the chance of the it being correct, I think
Erich Schulz (May 21 2016 at 02:08):
my practical desire is to express notions like "potential sleep apnoea"
Erich Schulz (May 21 2016 at 02:08):
or "potential HIV"
Grahame Grieve (May 21 2016 at 02:08):
how would that be different to differential?
Erich Schulz (May 21 2016 at 02:09):
differential relates to the diagnostic process for a specific issue/condition
Erich Schulz (May 21 2016 at 02:10):
ie headache, dd = brain tumor, migraine, tension headache
Erich Schulz (May 21 2016 at 02:10):
the "potential sleep apnoea" doesnt fit in that list
Grahame Grieve (May 21 2016 at 02:11):
does it have that narrow meaning? How is 'potential x' different to a diagnostic process'?
Erich Schulz (May 21 2016 at 02:12):
because the sleep apnoea is an incidental problem unrelated to the condition getting attention
Grahame Grieve (May 21 2016 at 02:13):
so? ah, you want to differentiate it from the differential diagnosese for the problem at hand?
Erich Schulz (May 21 2016 at 02:13):
but its still important because if it exists and too much opioids given the patient can just die quietly in the night...
Erich Schulz (May 21 2016 at 02:14):
well the entire "provisional"/"differential" set is a bit different from the "problem list"
Erich Schulz (May 21 2016 at 02:15):
ie often a "presenting complaint" which then may generate a set of "differential diagnoses" and 0..1 "provisional dx"
Erich Schulz (May 21 2016 at 02:16):
but this really a notion that applies during a search for diagnosis
Erich Schulz (May 21 2016 at 02:16):
which is only a subset of all encounters
Erich Schulz (May 21 2016 at 02:17):
sitting beside a "PC", "DDx", and "PDx" will be the "problem list"
Grahame Grieve (May 21 2016 at 02:17):
well, first, what is your problem list? Have you looked at what the specification has to say about current problem list?
Erich Schulz (May 21 2016 at 02:18):
condition looked like the "problem list" resource to me..
Erich Schulz (May 21 2016 at 02:19):
i'm looking here http://hl7.org/fhir/2016May/condition.html
Erich Schulz (May 21 2016 at 02:20):
should i be looking at something else?
Erich Schulz (May 21 2016 at 02:23):
this is relevant http://hl7.org/fhir/2016May/riskassessment.html but is more about the risk of an outcome
Erich Schulz (May 21 2016 at 02:26):
this issue is that the "provisional" and "differential" ideas are tied to a single "presenting/primary complaint"
Erich Schulz (May 21 2016 at 02:28):
but there are often important conditions that need to go into a problem list where p is between 0 and 1
Erich Schulz (May 21 2016 at 02:30):
with some of them - the mere fact the possilbility has been raised will trigger an immediate step to clarify (ie by ordering a test) but some will just fester on the list because the tests are unclear or too costly
Erich Schulz (May 21 2016 at 02:31):
but in meantime this potential condition needs a mechanism for capture
Erich Schulz (May 21 2016 at 02:33):
i guess in clinical notes its often written "? sleep apnoea", or "???? HIV"
Erich Schulz (May 21 2016 at 02:36):
to get me going I'm thinking I might just extend verificationStatuswith some kind of soft scale - "unlikely | possible | likely | very likely" ??
Grahame Grieve (May 21 2016 at 02:38):
this bit: http://hl7-fhir.github.io/lifecycle.html#current
Grahame Grieve (May 21 2016 at 02:38):
you should think about this and how it intersects with your problem
Erich Schulz (May 21 2016 at 02:39):
cool will have a read
Erich Schulz (May 21 2016 at 02:39):
i was just looking at this http://hl7.org/fhir/2016May/detectedissue.html
Erich Schulz (May 21 2016 at 02:39):
but it doesnt capture probability either
Erich Schulz (May 21 2016 at 02:42):
mmm just read the link
Erich Schulz (May 21 2016 at 02:42):
I dont think it helps really
Erich Schulz (May 21 2016 at 02:43):
the issue is that we may "currently" think "this patient probably/may/could have mild/mod/severe X"
Grahame Grieve (May 21 2016 at 02:43):
point is, a differential diagnosis list is a list of conditions from a particular perspective
Erich Schulz (May 21 2016 at 02:44):
yes and not the perspective i need...
Erich Schulz (May 21 2016 at 02:44):
one is about diagnosis, the other is about "holistic care" (normally that word gives me hives)
Grahame Grieve (May 21 2016 at 02:45):
there are many such lists. I think that getting the lists right is geneerally a better to handle things than permutating status
Erich Schulz (May 21 2016 at 02:45):
not that the diagnostic perspective isn't important
Erich Schulz (May 21 2016 at 02:46):
i gotta go take the kids to the park :-)
Erich Schulz (May 21 2016 at 02:46):
so you suggest another field then to capture probability?
Erich Schulz (May 21 2016 at 02:46):
it seems a continume from confirmed...refuted to me...
Erich Schulz (May 21 2016 at 02:47):
gotta go but will check in later :-)
Erich Schulz (May 21 2016 at 04:04):
ok been to park with kids...
Erich Schulz (May 21 2016 at 04:04):
some reflections:
Erich Schulz (May 21 2016 at 04:05):
I think the notion of probability probably needs to be handles in the same way as onset - ie polymorphic
Erich Schulz (May 21 2016 at 04:05):
http://www.clipular.com/c/6222340951638016.png?k=JD1gAVCHEpRYR1HIIPMJt-pLXio
Erich Schulz (May 21 2016 at 04:07):
ie allow different representations
Erich Schulz (May 21 2016 at 04:08):
in some contexts a "highly likely - possible -highly unlikey" family of codes will work
Erich Schulz (May 21 2016 at 04:09):
sometime it is possible to quantify a probabliliy numerically eg p = 0.72
Erich Schulz (May 21 2016 at 04:10):
other times a probility range maybe appropriate p between 0.2 and 0.45
Erich Schulz (May 21 2016 at 04:12):
the CDS teams must have some thought about this?
Grahame Grieve (May 21 2016 at 04:15):
they might have. But it seems like something that's going to be exceedingly difficult to get agreement on. The clinicians I have talked to have freaked when we start discussing firming up probability or the definitions of those codes.
Grahame Grieve (May 21 2016 at 04:15):
you can certainly propose adding something (propose a change at the bottom of every page), but suspect it would have to be an extension.
Erich Schulz (May 21 2016 at 04:16):
its a pretty core requirement
Grahame Grieve (May 21 2016 at 04:17):
I think you might find that difficult to demonstrate. It might be that you may get a new status for 'suspected' or something, but something more solid - I'd be pessimistic. but it will be up to the Patient Care workgroup - that's where any task you create will go
Erich Schulz (May 21 2016 at 04:21):
well not that hard to demonstrate... open any clinical record, find the problem lists and count the number of question marks or use of words "likely", "probable", "possible", "unlikely" etc
Erich Schulz (May 21 2016 at 04:21):
but thanks for the guidance! I'll open a ticket
Erich Schulz (May 21 2016 at 04:50):
i guess another way of summarising this is that the current condition resource only copes with a single diagnostic uncertainty (via provisionsal and differential) as well as allowing no mechanism to express degree of (un)certainty.
Erich Schulz (May 21 2016 at 04:50):
(i'll dump these notes into the ticket)
Last updated: Apr 12 2022 at 19:14 UTC
