Semantic Web Health Care and Life Sciences Interest Group Teleconference -- 02 Jun 2015

<trackbot> Date: 02 June 2015

<inserted> Scribe: dbooth

<scribe> Scribe: Claude

Je suis le scribe

FHIR RDF and Validation/Translation Task Force

<dbooth> http://www.w3.org/2015/05/27-hcls-minutes.html

<dbooth> http://www.w3.org/2015/05/20-hcls-minutes.html

<dbooth> dbooth: Those of us on the call were convinced that we can go ahead with using blank nodes where needed.

<dbooth> Explanation of blank nodes working with OWL: http://dbooth.org/2015/fhir/bnodes/bnode-test.html

dbooth: Group resolved to going ahead with using blank nodes for things that don't have an obvious URI

<dbooth> RESOLVED: Okay to use blank nodes for things with no obvious URI

dboth: Eric updated generation software from DSTU1 to DSTU2.

Debriefing from HL7 Meeting - Paris May 10-15

<dbooth> http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20150511

<dbooth> http://www.w3.org/2015/05/11-hcls-minutes.html

<dbooth> claude: we met with a number of HL7 folks, and on the phone. started with an intro to the working group.

<dbooth> ... moved on to discuss potential collab with groups from Europe. Two groups potentially quite interesting.

<dbooth> ... One: SALUS, from Turkey.

<dbooth> claude: One action item would be for SALUS to present to us about what we're doing, and potential alignment.

<dbooth> eric: SALUS is able to go from terminology mapping to terminology mapping.

<Marc_Twagirumukiza> SALUS web:http://salusproject.eu/

<dbooth> ... Mappings from term1 to term2 to term3, transitively, found errors, identified pitfalls.

<dbooth> charlie: they have a full RDF rep of ISO 11179, so I assume so.

<dbooth> eric: They use bits of 11179 to complement RDF.

<dbooth> charlie: They found they could do semantics better in RDF.

<dbooth> ... But use a mixture.

<dbooth> charlie: Anil said he could do it any week in June

<dbooth> ACTION: Charlie to ask Anil about presenting next week or the following [recorded in http://www.w3.org/2015/06/02-hcls-minutes.html#action01]

<trackbot> Error finding 'Charlie'. You can review and register nicknames at <http://www.w3.org/2014/HCLS/track/users>.

<dbooth> Claude: Vidal doing interesting work also. Should we have him present also?

<dbooth> charlie: Vidal is moving slowly for internal political reasons, but good idea. Or I might.

<dbooth> ... But Jean-Francoise is assuming that you'll meet w him in paris, so do that first, and then schedule it.

<dbooth> ACTION: Claude to follow up with Jean-Francoise about getting Vidal present [recorded in http://www.w3.org/2015/06/02-hcls-minutes.html#action02]

<trackbot> Error finding 'Claude'. You can review and register nicknames at <http://www.w3.org/2014/HCLS/track/users>.

<dbooth> claude: lloyd also presented the FHIR ontology (so far) in Paris.

<dbooth> ... Also discussed publishing terminologies in RDF, in canonical form. Need to define this canonical form.

<dbooth> ... Or a more optimistic route would be to publish in a common representation.

<dbooth> ... A lot of use cases depend terminologies, so not having them in a common form is a problem.

<dbooth> ... Potentially define a standard RDF export format.

<dbooth> tony: It also applies to HL7 internal terminologies.

<dbooth> dbooth: Potentially spin off a task force for this.

<dbooth> claude: One problem is everyione reinventing the wheel. Another is looking at potential overlaps. Otherwise everyone has to separately figure out the alignment of them.

<dbooth> ... In short term, everyone has to do this through mappings. In longer term, it would be valuable to coordinate.

<dbooth> dbooth: yosemite project has been thinking about that. maybe we should talk about that.

<dbooth> rob: Terminology landscape tends to work against this goal. Hard to get people to work together. one exampel of working together is SNOMED and LOINC, but it's been 15 years and still not there yet.

<dbooth> ... If we can contribute something that would be good.

<dbooth> Tony: Maybe we can contribute the meta model.

<dbooth> dbooth: Maybe we should make this a focus of one of the next calls.

<dbooth> charlie: If this is going to be successful, it needs to be guided or initiated at HL7 board level. Board would need to go to LOINC, SNOMED, CDISK boards to give them the compelling reasons for canonical RDF, and the response will certainly be "we don't know how", and then we'd have to help.

<dbooth> ... Pat Vandyke thought it is definitely a board level agenda. If we can come up with a strategy and value prop, we can get it onto board agenda.

<dbooth> eric: late in the paris meeting, Josh and i worked out an arch in which FHIR could have a registry of clinical termonologies and that could claim to be authoritative or it could forward if an org wanted to be authoritative.

<dbooth> ... That allows us to come up with a design that doesn't prevent the orgs from deploying their own data.

<dbooth> dbooth: good idea! but would the URIs change?

<rhausam> +q

<dbooth> eric: Our intuition was that we wouldn't want to change identifiers, so orgs would be stuck with HL7 URIs even if they took over the hosting later.

<dbooth> rob: can we get a more full description of that architecture? How does it compare with what we already have?

<dbooth> dbooth: Sounds like a purl server.

<dbooth> eric: Yes, but not at the server level. At the RDF level.

<Tony> +q

<dbooth> ACTION: Eric to write up his and Josh's proposal for a registry of clinical terminologies [recorded in http://www.w3.org/2015/06/02-hcls-minutes.html#action03]

<trackbot> Created ACTION-26 - Write up his and josh's proposal for a registry of clinical terminologies [on Eric Prud'hommeaux - due 2015-06-09].

<dbooth> tony: If we plan a transition from HL7 hosted to something else, we might be able to use the same prefixed names even if the URIs change

<dbooth> eric: I'm inclined to say the URIs shouldn't change.

Approve Minutes of previous meetings

dbooth: Two things remaining to cover - approval of minutes and then Charlie's topic.

<dbooth> May 5: http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20150505

dbooth: May 5, 12, 19, 26 minute approvals

<dbooth> May 11 in Paris: http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20150511

<dbooth> May 12 informal: http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20150512

<dbooth> May 19: http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20150519

<dbooth> May 26: http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20150526

<dbooth> wiki page minutes for paris: http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20150511 http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20150511

dbooth moves to approve minutes

Tony seconds

Paul approves minutes. No abstentions. No negatives.

Computable semantics between FHIR and RIM

<dbooth> APPROVED: Minutes of May 5, 11, 12, 19, 26

Charlie: In pre-FHIR days, Charlie was chair of architecture board. Grahame was member of board. Having computable semantics between RIM and FHIR is good idea but did not know how to do this.

<dbooth> charlie: We talked a lot about the role of RIM in FHIR semantics 3 years ago. Grahame's comment: i agree. somebody else needs to do it.

<dbooth> ... Thursday I was at an ARB meeting, Ken Mckaslin and ___ I asked why ARB isn't leading a project on computable semantics.

Charlie: Why is ARB not leading project to develop computable semantics between RIM and FHIR as this is an architectural concern.

<dbooth> ... The consensus was yes, we should by how?

<dbooth> ... I said that there's a whole group of people who could do that work but need a little direction from ARB.

<dbooth> ... Then I went to a mtg in paris where FHIR specs and lack of clear RIM semantics came up.

<dbooth> ... I mentioned that there's a possibility of computable semantics from ARB, and they were enthusiastic.

<dbooth> ... If ARB were to take this on, they would to it only with the agreement that people with tech savvy are available and willing to do it.

<dbooth> tony: I'm digging in the FHIR build, and there's mapping for RIM in each resource in the documentation.

<dbooth> charlie: yes, but inconsistently and not computable.

<Zakim> ericP, you wanted to talk about https://www.w3.org/wiki/HCLS/ClinicalObservationsInteroperability/C-CDA#Mapping_to_FHIR

<dbooth> eric: I was stymied by the complexity of it all. i was learning RIM and lots of other things.

<dbooth> ... I have some sense of the expressivity of that FHIR mapping to RIM.

<dbooth> ... There's RIM ITS. RIM is a graph model, XML root element with the things that go with it, are RIM ITS.

<dbooth> ... The spec in the FHIR defs is doing xpath into RIM ITS, and having a library of functions like 'firstOf'.

<dbooth> ... when i tried to model 'firstOf' in OWL i couldn't. could have done it in sparql.

<ericP> https://www.w3.org/wiki/HCLS/ClinicalObservationsInteroperability/C-CDA#Mapping_to_FHIR

<dbooth> ... another way to do this is the above URL.

<dbooth> ... There's instance data and shape expressions to map.

<dbooth> charlie: clarification: lloyed didn't take RIM. He wrote xslt to turn the spec to RDF.

<ericP> http://www.w3.org/2013/C-CDA/IJ.xml

<dbooth> eric: The above link is a CCDA doc that has been turned into OWL by xslt.

<dbooth> ... The result is chatty in ways that RDF heads won't like, but it's reality, so it may be the right representation.

<dbooth> ... The ShEx on the previous page maps ORIM to FHIR RDF, but in a coarse way, like if you were doing a CDE approach.

<dbooth> ... The other possibility was to stick in a SPARQL expression in the next column from RIM ITS semantic expression, that says how you map in terms of ORIM.

<rhausam> I have to drop now

<dbooth> charlie: Is this group interested enough in this, for me to tell them tonight that this group will help?

<dbooth> paul: In paris grahame surfaced the notion that, given that so few people populated the RIM column, maybe RIM should not be a requirement at all.

<dbooth> charlie: John and Kim agreed we need to do something.

<dbooth> ... It's on the agenda tonight as a possible project.

<dbooth> ... But they need to know there is a tech resource availabel that they could call on.

<dbooth> dbooth: We're specifically chartered to assist other groups on RDF and OWL.

<dbooth> our charter: http://wiki.hl7.org/images/3/31/Rdf-semantic-interop-subgroup-v9.docx

<dbooth> paul: Suppose we decide that the best way to do this via SPARQL. The bulk of RIM was done a Thursday night a few years back.

<dbooth> eric: We can't be doign domain modeling, because FHIR has already done that.

<dbooth> charlie: The point of the ISO HL7 pharm is that there's increasing concern about the lack of semantics.

<dbooth> ... The pharm spec has done as good a job as any, and they're willing to work with ARB and RDF group to use that as a pilot project about how to make those links computable.

<dbooth> ... I would propose to ARB that they take the project on, and work with the pharm spec as first iteration./

<dbooth> Tony: We might be able to give them patterns.

<mscottm> sorry - have to run

<dbooth> eric: Can we ask for resources? Day job sponsorship? Maybe members could sponsor?

<dbooth> charlie: Won't come from ARB, but maybe pharmacy group.

<dbooth> Proposal: This group is interested in helping on computable links between FHIR specs and RIM semantics in RDF.

<dbooth> AGREED: group agrees

Task Force call tomorrow

<dbooth> eric: Claude to discuss CQL, for clinical decision support

<dbooth> ... whether our FHIR RDF with something on top would meet CQL use cases.

<dbooth> tony: Still need to finish side-by-side

<dbooth> ADJOURNED

- DRAFT -

Semantic Web Health Care and Life Sciences Interest Group Teleconference

02 Jun 2015

Attendees

Contents

FHIR RDF and Validation/Translation Task Force

Debriefing from HL7 Meeting - Paris May 10-15

Approve Minutes of previous meetings

Computable semantics between FHIR and RIM

Task Force call tomorrow

Summary of Action Items

Scribe.perl diagnostic output