Stream: terminology
Topic: Disjointness Between Classes in an Ontology
Lin Zhang (Dec 09 2020 at 23:52):
Is it possible to represent disjointness between sibiling codes in an ontology/classification?
Michael Lawley (Dec 11 2020 at 13:05):
You certainly can in description logic -- the conjunction is subsumedBy bottom (i.e., A ⋀ B ⊑ ⊥)
Lin Zhang (Dec 11 2020 at 13:44):
How to express that in FHIR language?
Rob Hausam (Dec 13 2020 at 18:59):
In ConceptMap you can set group.element.target.equivalence = 'disjoint' in R4, or group.element.target.relationship = 'not-related-to' in the proposed R5. That's not necessarily defined strictly in accordance with OWL, but I think that's how you would express disjointness between concepts in FHIR. You might wonder why the R5 version seems less "precise" for declaring this? There's some considerable history on that, but I don't think that the semantic itself should be substantially different.
Lin Zhang (Dec 13 2020 at 23:15):
Sorry, the context info was missed in my question. I mean disjointness within a CodeSystem. And for now, it could be implicitly expressed with the hierarchyMeaning=classified-with. But explicitly in other ontology languages such as owl, obo. Thanks anyway.
Lin Zhang (Dec 13 2020 at 23:32):
It might be sensible to use CodeSystem.concept.properties of coding type for this purpose.
Michael Lawley (Dec 14 2020 at 01:23):
There's no way to say it within FHIR (other than in the description of the concepts or, as Rob suggests, in a ConceptMap).
But, the real question is _why_ do you want to be able to specify this in FHIR? It doesn't affect behaviour of any of the operations.
Lin Zhang (Dec 14 2020 at 02:55):
The reason is the value of disjointness. And this may be related to implementation of the hierarchyMeaning classified-with beyond just as a flag.
Michael Lawley (Jan 01 2021 at 05:55):
You could add a 0..* property to a concept that lists the concepts it is disjoint with.
Alternatively, you might add a flag to a parent concept to indicate that all its children are disjoint.
I'm sure there are many other options too. But they all rely on a client doing some kind of interpretation; there's nothing that the terminology server itself will do with this information.
Lin Zhang (Jan 01 2021 at 07:10):
@Michael Lawley Thanks. The first option looks more like OWL.
Last updated: Apr 12 2022 at 19:14 UTC
