Meeting: HCLS General Forum Convener: Michel Dumontier Date: May 7, 2013 @ 11am EDT Frequency: 1st Tuesday of the month Past Meetings: http://goo.gl/NFqPx Dial-In #: +1.617.761.6200 (Cambridge, MA) VoIP address: sip:email@example.com Participant Access Code: 4257 ("HCLS") IRC Channel: irc.w3.org port 6665 channel #HCLS IRC Web: http://irc.w3.org
- clinical pharmagenomics update
- CDISC2RDF moved to FDA/PhUSE - Kerstin Forsberg
- clinical data representations - ericP
[11:03] <@ericP> Topic: pharmacogenomics
[11:04] <@ericP> scribenick: bobP
[11:04] <bobP> Michel: PGx, we are taking inventory of priorties, planning
[11:05] <bobP> ... Bob Friemuth also has objectives, not yet posted here
[11:05] <bobP> ... next week we talk about which agendas w interdependencies
[11:06] <bobP> ... this is part of re-organization; not waiting for people, but sharing
[11:06] <bobP> ... conversations about what everybody wants to do and how to get there
[11:07] <bobP> ericP: You got objectives and plan
[11:08] <bobP> Michel: Yes; more fitting projects than having just one big project
[11:08] <bobP> ... synergy among projects
[11:08] <bobP> ... to complement each others priorities
[11:11] <bobP> CDISC from Roche
[11:12] <michel> kerstin: working on CDISC
[11:13] <michel> ... proposal from Roche to move emerging technologies working group; semantic web
[11:13] <michel> ... 40 people engaged in this
[11:14] <michel> ... formation of specific teams over the next few weeks. includes documentation and use cases
[11:14] <michel> charlie: representation of existing standards. participation by FDA is positive
[11:15] <dbooth> Topic: SemTech workshop on "RDF as a Universal Healthcare Exchange Language"
[11:16] <dbooth> There will be a free workshop on "RDF as a Universal Healthcare Exchange Language" at the SemTech Conference in San Francisco on Monday June 3: http://semtechbizsf2013.semanticweb.com/sessionPop.cfm?confid=70&proposalid=5296
[11:16] <dbooth> There will also be an associated panel discussion as part of the SemTech conference, on Wednesday June 5:
[11:16] <dbooth> The workshop is "by invitation only", but the main reason is just to ensure that we do not exceed the room capacity. There is a form (linked from the workshop page) that allows you to apply to attend, and we expect most to be accepted. I would encourage all in the HCLS to come and participate if you can. Participants in the workshop (which is free) will also receive a 15% conference discount if they choose to also attend the rest of the conference. Let me
[11:16] <dbooth> know if you have questions: firstname.lastname@example.org
[11:17] <dbooth> topic: clinical informatics
[11:17] <dbooth> http://www.w3.org/2013/Talks/0507-HCLS-egp/
[11:20] <michel> slide 3 - typical clinical data - encounter, allergies, immunizations, history, demographics, vitals, labs, diagnosis, prescription, etc
[11:20] <michel> slide 4 - example procedure
[11:23] <michel> slide 5 - simple RDF graph; encounter, location, patient, physician, clinic
[11:24] <michel> slide 6 - enable coding systems; replace "codeine" with "influenza vaccination"
[11:26] <michel> http://codingsystems.org/2.16.840.1.113883.6.88/2670
[11:28] <michel> ont:substance <http://codingsystems.org/2.16.840.1.113883.6.88/2670>
[11:29] <Kerstin> More info on URIs for SNOMED: http://informatics.mayo.edu/cts2/index.php/ExternalURI
[11:29] <michel> michel: are we heading to URIs?
[11:29] <michel> eric: that's the idea :)
[11:31] <michel> slide 7 - rdf description
[11:33] <michel> eric: measure what they observe against medical knowledge, in coded terms.
[11:34] <michel> ... where is the most energy being put in these systems?
[11:34] <michel> ... patient encounter, procedures performed.
[11:35] <michel> ... 23andme, patientlikeme - people putting in their data; bulk of data will be in CDA as part of meaningful use act
[11:36] <michel> ... hl7 v3 has not had the expected uptake; very complex in v3 messaging; little motivation to upgrade
[11:36] <michel> ... between CDA and BRIDG (clinical trials), defined in terms of the RIM. bulk of data is in the RIM.
[11:37] <rafael> http://ccda-scorecard.smartplatforms.org/static/ccdaScorecard/#/
[11:37] <michel> slide 8 - immunization record in CCDA; complex
[11:37] <michel> ... xml structure; who performed it, responsible organization
[11:38] <michel> slide 10 - a reflection of that complexity into RDF.
[11:40] <michel> slide 11 - e.g. manufactured material acted in the consumption role of a substance administration
[11:41] <dbooth> q+ To ask why not turn those codes into RDF properties?
[11:42] <michel> michel: easy simplification is turning those code references into URIs, and provide human readable descriptions for them
[11:44] <michel> eric: ids are provided for the type; some are specific instances
[11:44] <michel> ... some rules to infer the identity of an instance
[11:44] <michel> ... so that all the statements use that id as the subject
[11:45] <michel> ... so that records can be transferred
[11:46] <michel> time is usually fairly distinguishing... temporal tagging
[11:47] <michel> charlie: data is being distributed; things can get separated. microbiology test - issues around incompleteness, uncertainty
[11:47] <dbooth> The work of de-duping is pushed into the heuristic rules that generate URIs from a collection of (presumably) uniquely identify that thing, whether it is an event or whatever.
[11:49] <dbooth> s/collection of/collection of properties that/
[11:49] <michel> eric: timing information will provide heuristics to debug issues. when did an id fail to merge?
[11:50] <dbooth> eric: want to lean on institutions to mint new UUIDs for new events.
[11:51] <dbooth> But for de-duping, you want them also NOT mint UUIDs for non-new events.
[11:51] <michel> eric: uuids should be made for events under their jurisdication
[11:51] * michel thanks david, your help is invaluable here :)
[11:52] <michel> eric: having it in RDF not only means we can query effectively, but easier to integrate
[11:54] <michel> rafael: problem with all our systems is that we don't have a good method to cross-validate other assertions. provenance is key.
[11:56] <michel> eric: adding PROV to assertions would be the way forward; can figure out what information was present at that time
[11:56] <Kerstin> Beside the identifiers/de-duplificatation of the health records/documents we also have the identifiers of what they document. Check out: "Referent tracking" http://www.referent-tracking.com/RTU/?page=index
[11:56] <bobP> Prov Ontology http://www.w3.org/TR/prov-o/
[11:57] <michel> eric: this sparql at this time, we got these results; doing it later, gives different results, and compare
[11:57] <michel> cool :)
[11:58] <michel> rich: working on de-duplication; transfer of information for supportive care. use cases for supporting clinical research?
[11:59] <michel> eric: CDISC semantic web representation of clinical trials. If used BRIDG (based on RIM), reasonable to transform/merge data with some html document (?)
[12:00] <michel> eric: lot of information in a clinical trial; merging clinical care/clinical trial data offers much opportunities
[12:01] <michel> charlie: are there standards? yes for human subjects. problem is educating clinical trial community. increasing amount of effort into semantically interoperate between trials and research
[12:02] <michel> charlie: IMI funded EHR4CR ...
[12:02] <rafael> http://ccda-scorecard.smartplatforms.org/static/ccdaScorecard/#/
[12:02] <michel> eric: worth the time for all this work
[12:25] <Zakim> SW_HCLS()11:00AM has ended
[12:25] <Zakim> Attendees were +1.412.624.aaaa, Bob_Powers, +1.650.804.aabb, michel, Kerstin_Forsberg, ericP, DBooth, +18.104.22.168.aacc, +1.410.905.aadd, +1.510.421.aaee, Tony, +1.919.681.aaff