An open discussion on the Use of “+” and “&” in LOINC Components and their representation in SNOMED CT as part of the LOINC/IHTSDO alignment
An open discussion on the Use of “+” and “&” in LOINC Components and their representation in SNOMED CT as part of the LOINC/IHTSDO alignment
7 Comments
Jim Case
Daniel Karlsson asked that the Editorial Advisory Group review and comment on the issue surround the use of disjunctive component in LOINC and how they might be addressed as part of the LOINC project. An issue brief is available from
https://confluence.ihtsdotools.org/download/attachments/27593317/LOINC%20-%20SNOMED%20CT%20cooperation%20project%20-%20Union-Disjunction.pptx?api=v2
Please post your comments here and we will add this to our next conference call
Bruce Goldberg
In my opinion, I think given the limited expressivity of SNOMED’s DL at this time, the current approach of creating disjunctive primitives and manually asserting the individual components underneath although resulting in some degree of inconsistency seems most reasonable. Another suggestion alluded to in Daniel’s presentation was to model a subset in Protégé using OWL and export the classified results back to SNOMED as a primitive hierarchy. This should be investigated further.
Daniel Karlsson
Thanks Bruce,
there's still a decision to be made if the primitives are to be added at the level of the substances/organisms/etc. or the observables.
/Daniel
Bruce Goldberg
Adding disjunctive substances will help with the issue we are seeing with allergy/adverse reactions to drug combinations such as amoxicillin+clavulanate - so I would favor that approach.
Bruce
Daniel Karlsson
And that is a bit special, while in most cases, e.g. concentration of Alanine+Beta Alanine+Sarcosine, it means any or all of the alternatives, i.e. concentration of (Alanine OR Beta Alanine OR Sarcosine), while in the case of susceptibility towards amoxicillin+clavulanate case and other cases where there is a synergistic effect all constituents must be present, i.e. SOME Amoxicillin AND SOME Clavulanate. I guess there are (at least) two distinct meanings of the +.
/Daniel
Keith Campbell
I wonder if we can encourage LOINC to take an alternative approach for and/or, &or. I don't think modeling in Protege and bypassing the standard description logic is a good idea. I think we should move toward Owl EL as we've stated in the past, but I think that the &or will require going beyond EL. I believe EL allows for disjointness, but with &or, you need more than disjointness, and it gets us into intractability land.
One option is to create "parent classes", and use them in the expressions: Have a primitive concept for Salmonella sp &or Shigella sp. that you can use in the definition of:
If there was a natural parent that would be ideal. And then separately have Salmonella sp identified, and Shigella sp identified as child tests so we can have more specificity if available.
And ideally work to retire LP39687-6 Salmonella sp &or Shigella sp identified and the primitive concept for Salmonella sp &or Shigella sp
I think this is essentially equivalent to Bruce's proposal.
Jim Case
This discussion was completed at the Wellington meeting. It was decided that the LOINC part disjunctive concepts would not be added to the International release. Concepts that would have been dependent on these concepts would be modeled as primitives.
Methods to handle disjunction were deemed to not be the purview of the editorial advisory group and will be communicated to the Modeling AG for future resolution.