Organism & Infectious Disease Model Project

Page tree

Matt Cordell

Owner
878 View 2 Comment In discussion Comments enabled In the category: Undefined

I just want to ask about concept like 702938004 |Klebsiella pneumoniae or Klebsiella oxytoca (finding)|

I notice that since 2014 these sort of things have been modeled as findings (I vaguely recall the discussions at the time).
However, this does complicate things for some microbial reports where a more specific identification isn't available/possible. And these sort of groupers are as clinically useful, and as specific as is required.

(To me at least) It seems on par with saying 127535004|Staphylococcus, coagulase positive| which is a more elegant way of saying "Staph. aureus or Staph. intermedius" (Clinically).

I also see there's a concept 720607001|Klebsiella species, not Klebsiella pneumoniae and not Klebsiella oxytoca (organism)| (created last year). Which would describe all the Klebsiella not subsumed by my earlier example.

Should these "OR" concepts exist as organisms? Or should we advice implementers to use a mix of "Organisms AND Findings" ? (which makes me a little uncomfortable...)

Contributors (2)

2 Comments

  1. Jeff Wilcke

    Matt,

    I am more than a little concerned about these concept classes as well.  That they are currently placed as findings is the result, I fear, of my having argued successfully that they could not be correctly placed in the (all primitive) organism hierarchy combined with my failure to propose a credible alternative.   I'm not enough of an expert in description logics (even SNOMED's) to provide a "deep" answer to this dilemma.

    The problem is related to a desire/need to create an exclusive or construction:

    If I use the concept 702938004 |Klebsiella pneumoniae or Klebsiella oxytoca (finding) in a laboratory report, I am indicating that the isolate is EITHER Klebsiella pneumoniae OR it is Klebsiella oxytoca BUT IT IS NOT BOTH.  (Therefore an exclusive or).   Placing it as a finding does not actually solve the problem, it likely masks it.  I agree with you that 702938004 is an organism class, not a finding class.   Perhaps if we consult the logic gurus, we may understand that there's a work-around that allows them to be placed correctly as findings (but I really don't think so).  All that were created in the findings hierarchy are primitive which conveniently allows logic engines to essentially ignore them.   In the end though, I still don't know how it should be placed in the organism hierarchy.

    In the context of the SCT organism hierarchy, 127535004|Staphylococcus, coagulase positive stands in for an inclusive or (and/or).  Both Staphylococcus aureus and Staphylococcus intermedius are correct subtypes.  The class actually has several descendants in SCT.  Staphylococcus aureus, Staphylococcus intermedius Group (organism group concepts are a similar problem we haven't solved) and Staphylococcus lutrae.  719237009 | Staphylococcus intermedius group (organism)  has 3 subtypes: Staphylococcus delphini, Staphylococcus intermedius and Staphylococcus pseudintermedius.  BECAUSE inclusive or is supported (at least the work-around works), 127535004|Staphylococcus, coagulase remains in the organism hierarchy.   So in SNOMED, Coagulase positive Staphylococcus means "ALL Staphylococci taxons that are coagulase positive" not "a single instance of a Staphylococcus that is coagulase positive."

    That said, I agree with you that in the context of a laboratory report, 127535004|Staphylococcus, coagulase positive would seem to be (at least a synonym for) an exclusive or construction.  It means "a single instance of a Staphylococcus that is coagulase positive."  I can know that because the laboratory report is the context through which I assert that the value is an isolate (an instance).   In a way, the meaning in the lab report conflicts with the meaning in SNOMED.

    In the end, I believe that the truth of these exclusive or constructions is that, even when the specific identity is not made clear, the value on the laboratory report is still an organism.  I do not believe that anything about them changes the subhierarchy to which they belong.  Unfortunately, SNOMED doesn't deploy logic complicated enough to accurately portray exclusive or AND NEITHER DO LABORATORY REPORTS.   I don't agree that the decision to create the exclusive or classes as findings is the correct decision, but I don't know what is correct. A practical answer may be a subset with the hierarchy designators removed (wink)...

    PS.  The "not" constructions enjoy all the same problems but make "calculating" the logic even worse.

  2. Matt Cordell

    Thanks Jeff.

    I think my view was clouded by assuming it was "obviously" exclusive or that what intended (since an "organism" can't simultaneously be two species). But the description logic doesn't show that.

    It seems like exclusive or might be applicable to primitive the "primitive hierarchies".. (organisms, body structures ...).  61685007 |Limb structure (body structure)|  effectively means "arm structure or leg structure". However, Lower limb structure and Upper limb structure are disjoint. As a specific structure won't simultaneously be subtypes of both.**

    Similarly, the Linnean classes are always disjoint.

    inclusive or (and/or) makes sense for findings, where conjunction can exist. (mixed infections, multiple fractures etc).

    Maybe the new DL profile will help?

    ** "Combined sites" might complicate things, I'm not sure what's happening there. The more I think about it, body structures might be a poor example :/