HCLS -- 29 Apr 2014

FHIR ontology by Claude

claude: Got resources represented. Approach is not very different from how FHIR does things. But want Tony's feedback on terminology.
... FHIR models using codable concepts, which make sense in XML. But in OWL it may not be ideal.

tony: agree. a lot of languages don't support instances belonging to multiple classes.
... so codable concept is a shortcut instead of saying that the thing is a member of a class.
... e.g., an allergy instance is a member of the class representing the snomed class.
... An allergy is a response to a substance.
... will be at HL7 meeting.

claude: maybe do something informal there.
... want to talk to Graham and Lloyd.

tony: suggest arch review board.
... I made a presentation last time on using supplemental schemas. i'm very concerned about HL7 inventing languages all the time.
... I'm looking for the test tools for whether a payload is conformant, and they aren't there yet.
... How do you represent codable concept?

claude: code is an instance, with a unique URI, and a value

tony: that instance is not a member of a class?

claude: it's an instance of a code class.

tony: so you're modeling the fhir structure, like josh did. that's very different from how i did it.
... I say that URI is actually a class.

claude: There are two ways you can represent this relationship. one is if you want to encode the way FHIR does then you can use the type of the class.
... diabetes melitis concept in the SNOMED hierarchy.

tony: that's how ITSDU extracts classes out of snomed. there are not things in snomed -- it's purely type.

claude: then i can say this thing is an instance.

tony: that's how i did it.
... what i've just done a few days ago is a sparql query that can detect a drug allergy interaction potential.
... Want to see if an allergy to a substance, classified as snomed type, is higher or equal to the class of the drug you're going to take.
... penicillin is not a class, it's an anonymous instance.
... so in the same class you see a medication and an allergy in the same class and to detect the match -- the conflict.
... so you can say the allergy is very general, and penicillin is in the subclass of that class.
... Then i do a filter saying the class of this equals the class of that. that allows the snomed hierarchy to do these things.

claude: That's the problem of the division between model and terminlogy.

tony: I have fhir ont importing snomed ont, and then my instance ont imports the fhir type ont.

claude: would be great to meet at HL7.

tony: I have a presentation ready, but not sure when and how it will be given.
... The started off with 'could we do a profile in FHIR that would capture DL semantics?'
... Then I realized you don't want to mix the two formats. Do then I looked at a DL attachment, to hold all the FHIR information in the DL representation.

david: What time arriving on Sunday?

tony: about 5, staying at hampton inn

claude: i started where josh started, but took a slightly different approach, but intending to converge.
... but not as set-oriented as your approach.

tony: there may be a transformation possible.
... Maybe two onts: one of the FHIR record itself, and then one that is what you want to use.
... Would be interesting to explore.

david: End game: adopt a standard FHIR ont.

tony: national cancer already has an ont based on ISO 11179. Don't know what they're using yet, but made major investments in 11179. Some overlaps FHIR profile. being discussed: should the profile align with i?
... But some orgs have already make a big investment in RDF. I think ONC is also interested in this.

claude: maybe just propose it as an HL7 project.

<scribe> ACTION: Claude to find out how to propose FHIR ontology as an HL7 project [recorded in http://www.w3.org/2014/04/29-hcls-minutes.html#action01]

tony: I've just been reading a book by Azamat ___. He says there is a universal ont.
... Called UFO. Concerns about whether this may be too big of a waterfall.

claude: fhir is not consistent. Each resource is being developed by different groups, and their approaches are differnt.
... VMR and QDM

tony: it's declaring local variables that are different from each other but they look the same.
... Semantics may be different, and bindings to terminologies likely to be different. Do they inherit from something further up that makes them consistent?

Mike: No, there's no property conservation at all, and that's intentional. They were trying to avoid the RIM formalities.
... So it's a big problem semantically. I don't think you can endure that hodge-podge.

tony: NIEM (Natl Info Exchange Model) is XML based and goes the other way: all properties are globally declared.
... And that causes problems also.
... Some things should inherit from a general concept.
... So the flexibility of it is interesting, but now you need to get a well-formed ontology out of it.
... i was happy to get a drug allergy conflict out of it.
... There's a conf 'KR for HC' in Europe. It's part of a bigger conference. They're taking paper submissions now.
... In Vienna.

<Claude> http://banzai-deim.urv.net/events/KR4HC-2014/

tony: if you write sparql to detect something, the turtle tells you what you're doing to do in FHIR to get all the stuff. So turtle becomes a langauge for a set of FHIR orchestration.

Next week

<mscottm> thanks for the great notes David!

david: may not be able to do a call next week, due to HL7 conference
... Those of at HL7 meeting will meet up in person though. Exchange contact info by email.

ADJOURNED

- DRAFT -

HCLS

29 Apr 2014

Attendees

Contents

FHIR ontology by Claude

Next week

Summary of Action Items

Scribe.perl diagnostic output