W3C

- DRAFT -

Semantic Web Health Care and Life Sciences Interest Group Teleconference

24 May 2016

See also: IRC log

Attendees

Present
Amol_Bhalla_(Infor), David_Booth, Sharam_Shahpouri, Harold_Solbrig, EricP, James_Anderson, Brian_Pech, Quoqian_Jiang, Rob_Hausam, Thomas_Lukasik
Regrets
Chair
David Booth
Scribe
dbooth

Contents


ShEx validation of FHIR RDF (Harold / Grahame)

harold: Pretty close to converged. There was more discussion between Grahame and Lloyd about where the FHIR URIs belong.
... There was fhir:CodedConcept that can have one or more Codes, and within each Coding it is okay to have a concept. But should additional code concepts be allowed?
... There was a hope that the aggregate of the multiple codings could be reasoned about in the outer coded concept.

<hsolbrig> fhir:Observation.code [ fhir:CodeableConcept.coding [ fhir:index 0; fhir:concept loinc:29463-7; fhir:Coding.system [ fhir:value "http://loinc.org" ]; fhir:Coding.code [ fhir:value "29463-7" ]; fhir:Coding.display [ fhir:value "Body Weight" ] ], [ fhir:index 1; fhir:concept loinc:3141-9; fhir:Coding.system [ fhir:value "http://loinc.org" ]; fhir:Coding.code [ fhir:value "3141-9" ]; [CUT]

<hsolbrig> http://hl7-fhir.github.io/observation-example.ttl.html

harold: Sometimes we can create a URI for the concept, such as fhir:concept loinc:29463-7;
... But sometimes we cannot, as with fhir:index 3
... The troubling point is that in some circumstances Lloyd thought it could be used as a compositional grammar, such as ARM and RIGHT.

dbooth: My previous understanding of multiple codings under an fhir:Observation.code was that it is a logical conjunction, i.e., AND, of the codings.

<hsolbrig> harold: a third interpretation would be that ANY of the codes completely represents the intent.

<ericP> ericP: it may be more accurate to describe it as a Union.

<hsolbrig> In that a receiver can proceed if it understands SNOMED or if it understands LOINC or if it understands something else

eric: Maybe it means more like a union.

sharam: I understand it, and we use it at infor, that all of these codes independently represent the concept.

dbooth: That sounds like logical AND -- all are independently true.

<ericP> i think it's a Union, not an Intersection

eric: AND is more like an intersection. If there is a coding that I don't understand, then the the other codings are still true.
... If it's a union, and all apply, even though it is treated as a union then you can still get use out of it if you treat it as an intersection.
... If you start modeling them as intersection then the modelers need to be concerned about whether then are the same kinds of things.

<ericP> ... e.g. x:FracturedTibia and y:SpiralFracture

harold: This affects both senders and receives. We need to nail this down. Also, the coding element itself includes capability of compositional grammar.
... If you are going to allow fractureOfArm above, then how does it relate to below?
... My understanding is that if the receiver did not understand one coding then it may understand another.

guoqian: Does multiple coding violate the cardinality constraint in the model?

harold: No. CodedConcept cardinality is 1, but it can contain any number of codings inside.

sharam: From data modeling perspective in Infor, if you look at these concepts they have system property. When you have concepts from different systems, it is semantically difficult to combine them from multiple systems, such as ARM from LOINC and LEFT from SNOMED.

harold: I thought we removed the fhir:Concept from the Observation.code level.

<hsolbrig> http://hl7-fhir.github.io/observation.shex.html

dbooth: Need to go back to FHIR group to nail down the semantics of multiple codings?

guoqian: Not all codings have fhir:concept in the example.

harold: Right. You create it if you can.

<hsolbrig> <CodeableConcept> {fhir:concept IRI*;

harold: I'm thinking of disallowing fhir:concept at the CodeableConcept level -- only have it at the Coding level.
... I also mentioned the type arc for references and said it is optional. Grahame said we had decided.

dbooth: We decided that *if* it is included then it must be a specific reference type, such as ObservationReference. But we postponed the decision about whether it should be required/optional/disallowed.

https://github.com/w3c/hcls-fhir-rdf/issues/26

https://www.w3.org/2016/04/19-hcls-minutes.html

eric: david didn't want extra type arcs; I wanted a type arc to enable the patient to be validated without dereferencing the URI.

https://github.com/w3c/hcls-fhir-rdf/issues/25

<scribe> ACTION: Harold to start discussion in zulip about the exact semantics of multiple codings [recorded in http://www.w3.org/2016/05/24-hcls-minutes.html#action01]

<trackbot> Created ACTION-56 - Start discussion in zulip about the exact semantics of multiple codings [on Harold Solbrig - due 2016-05-31].

ShEx validator service (EricP)

harold: Eric has made a shex-based validator service for FHIR RDF

<ericP> Git repo for ShEx server

eric: It's a small server. Uses a KOA. I'm adding better error messages to it.
... Right now if it doesn't match then you get null.

harold: Michael Van Der Zel is using it.

eric: If you give it a focus node identifier it interprets as a relative URL relative to the document. Relatively friendly
... Planning to make it a web-based service.

harold: Can it be on the yosemite site?

dbooth: I think so. i'll look into it.

<hsolbrig> http://fhir.fhir-schema-org.appspot.com/

FHIR on schema.org (Harold)

http://fhir.fhir-schema-org.appspot.com/

harold: It's relatively easy to build if you go to github instructions.

<hsolbrig> http://fhir.fhir-schema-org.appspot.com/docs/full.html

harold: If you go to that page --> schemas -> Full list --> Extension 'fhir'

http://fhir.fhir-schema-org.appspot.com/docs/full.html

<hsolbrig> http://fhir.fhir-schema-org.appspot.com/Observation

harold: scroll down and look under w5: http://w5.fhir-schema-org.appspot.com/w5
... scroll down to Observation below w5
... One point raised by Marc T (responsible for W3C life sci part of schema.org)

<hsolbrig> http://schema.org/docs/meddocs.html

harold: when dealing with schema.org you have 3 choices: 1. add to schema.org itself 2. create an extension (what was done) 3. conforms to schema.org but not under schema.org
... Marc also mentioned that schema.org people believe in local predicates. We have Observation.bodySite. They would want to have predicates of bodySite without the prefix, which goes back to discussions that we have had, that the meaning of the predicates changes based on the context.
... Part of the discussion: value of doing this at all; how to align this with medical stuff that is in there.

<hsolbrig> http://health-lifesci.schema.org/MedicalCondition

harold: If you dig down to the medical stuff they have already gone to great lenghts to align with established medical terminologies such as SNOMED.
... You cannot see the connections to LOINC and SNOMED, but they exist.

<hsolbrig> https://github.com/schemaorg/schemaorg

<hsolbrig> https://github.com/schemaorg/schemaorg/tree/sdo-deimos/data/ext/health-lifesci

harold: the above URL is the github that goes into schema.org
... and the second link is the health-lifesci part that goes into schema.org

<hsolbrig> https://github.com/schemaorg/schemaorg/blob/sdo-deimos/data/ext/health-lifesci/med-health-core.rdfa

harold: Line 66 shows a mapping to snomed-ct

<hsolbrig> http://my.clevelandclinic.org/health/diseases_conditions/hic_anemia

harold: If you view source of that Cleveland Clinic URL, you'll see it uses that medical stuff from schema.org
... Marc T wants to align with FHIR.

dbooth: We should absolutely align.

eric: Cleveland Clinic use case is talking about medical knowledge domain, rather than the medical records domain.
... My guess is that there will be more use cases for medical knowledge than records in schema.org.

harold: we learned that at the moment google's external search engine pays attention to this, but their internal appliance does not.
... It would help dramatically if it did.
... There is deidentified data at Mayo that could be exposed using FHIR.

<scribe> ACTION: DBooth to look into hosting ShEx-based FHIR RDF validator on yosemiteproject.org [recorded in http://www.w3.org/2016/05/24-hcls-minutes.html#action02]

<trackbot> Created ACTION-57 - Look into hosting shex-based fhir rdf validator on yosemiteproject.org [on David Booth - due 2016-05-31].

harold: Need to show that the FHIR RDF is round trippable.

dbooth: Should be checked in the build, for all examples.

ADJOURNED

Summary of Action Items

[NEW] ACTION: DBooth to look into hosting ShEx-based FHIR RDF validator on yosemiteproject.org [recorded in http://www.w3.org/2016/05/24-hcls-minutes.html#action02]
[NEW] ACTION: Harold to start discussion in zulip about the exact semantics of multiple codings [recorded in http://www.w3.org/2016/05/24-hcls-minutes.html#action01]
 

Summary of Resolutions

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.144 (CVS log)
$Date: 2016/05/24 16:11:51 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.144  of Date: 2015/11/17 08:39:34  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/Mayo/Cleveland Clinic/
No ScribeNick specified.  Guessing ScribeNick: dbooth
Inferring Scribes: dbooth
Present: Amol_Bhalla_(Infor) David_Booth Sharam_Shahpouri Harold_Solbrig EricP James_Anderson Brian_Pech Quoqian_Jiang Rob_Hausam Thomas_Lukasik
Found Date: 24 May 2016
Guessing minutes URL: http://www.w3.org/2016/05/24-hcls-minutes.html
People with action items: dbooth harold

[End of scribe.perl diagnostic output]