SV_MEETING_TITLE -- 8 Jan 2008

<LeeF> thanks, zakim

hello!

<Vipul> Welcome jennier

<LeeF> Scribe: Jennifer

<Vipul> EricP can you come on Zakim

<Vipul> Agenda:

<Vipul> 1. Roll Call

<Vipul> 2. Admin Issues: Set up mailing List

who will do this?

<Vipul> EricP, please come on line?

Eric will set up mailing list for COI task force, but he is not on the call

outpatient data

<LeeF> ScribeNick: fostel

<Vipul> 3. Mock Patient Data for Diabetes and Hypertension

<Vipul> http://esw.w3.org/topic/HCLS/ClinicalObservationsInteroperability/MockedUpPatientData)

vipul: good meting with endocrinologist;
... should have mock data soon

Parsa: format of data? out / in patient?
... historic data?

vipul: example of each in and outpt; data items in example;
... example protocols from clinicaltrials.gov

narrative or structured?

vipul -- format should not be an issue

parsa -- searched reports for outpt data + history + contextual data; have several cases in narrative form; will upload

parsa -- not of the data reports have all of the elements listed by vipul; need a handful of pts showing different data elements

parsa: discussion with UTx clinic; collect outpt data; willing to share whole data set, lab reports in a few days. 2K pts data; need to strip out IDs

vipul: sample of pts from this database

parsa: focus on diabetes nad hypertension; use remaining to test our format / model

next agenda: Jyoti clarification

Joyti - plannignmeeting with div epidemiology planned for early Feb; will try to obtain data

action item - jyoti report back after this meeting

<Vipul> ACTION: Jyoti to report back after 1st Feb [recorded in http://www.w3.org/2008/01/08-hcls-minutes.html#action01]

vipul Peter...?

<Vipul> Peter Elkin

joyti: peter Elkin not aware / not interested

jyoti: harold in not in the loop at this point; need to involve him; will discuss this week; send e-mail to vipul

vipul: tom still working on data; set up a meeting
... useful to get data from different sources to stress the system

<LeeF> +4s/joyti/jyoti

rachel: no access to data; provide dummy data?

sorry about typos!

vipul: progress on dummy pt data; data requirements ... George?
... George will come up with data requirements for hypertensive study
... minimum dataset for clinical study; particular data items must be included

fostel: we should check CONSORT for minimum data
... http://www.consort-statement.org/

rachel?: not aware of minimal data in research trials

fostel: argee with rachel re minimal reqs

vipul: data from free research protocols; distributed to data people; will upload the spreadsheet soon
... work on mapping
... will identify key data items from the spreadsheet;
... aim to look at data requirements eg pt age; HL7 RIM - how is pt age modeled
... role = pt; attribute = birth time; data type = time stamp

<Parsa> URL for CONSORT: http://www.consort-statement.org/

vipul; yesterday distributed spreadsheet; tom-- how will DCM elements be found?

vipul: gender example -- HL7 has code; RIM has well defined domains
... RIM has terminology, eg male / female; also can bind user own data terms
... blood pressure -- observation; obs type = physical exam for instance
... RIM obs.code and obs value; code = obs type; this is what Tom was referring to
... obs value domain bind to ICD9, or bind to SNOMED; HL7-RIM is agnostic to vocab, can bind ot any set of terms

who?: in NL can use terminology, JP can use another teminology

vipul: drug = zenical; HL7-RIM would need route, dosage, frequency, time of delivery
... we need to model time

fostel: time is a critical thing to model, indeed!

vipul: model needs to be general; pt recruitment, trial, etc. not hard code it based on one use case

Tom: some folks may store these data, eg administration of drug, in which case route, frequency, etc would be stored
... :pt is taking fentinyl" is a different case; eg pt is taking the drug but we do not have administration details
... if the derived data are frequently-queried, then store them; they always store granular administration data

parsa: direct and indirect evidence -- pts may have assertion that he has fever (in vital signs) or may have medication for fever in chart
... model depends on goal; physician note/diagnosis regarding disease, med, symptom; or is it sufficient to infer this from other evidence?

vipul: example - document posted on wiki -

parsa: self-reported fever; family-reported fevel; practitioner-reported fever; objeective fever (based on measurement)

vipul: we are thinking of modeling the first three as inference
... these will be identified suring mapping across systems
... assessment not inference
... example, protocol 8, criteria 16: uncontrolled hypertension has value greater than N; in ER would only have the raw values, reasoner would need to be able to infer diagnosis
... will distribute the spreadsheet; can we fill out DCM elements?
... any standard; not just DCM
... action item for tom -- DCM values for these rows
... HL7 has std way to go to vocab; will we use this method or another?

tom: they have internal vocab; may be mappable to LOINC, etc, but they bind their model to their terms

<Vipul> ACTION: Tom Oniki to come up with DCM elements for the list of HL7/RIM elements [recorded in http://www.w3.org/2008/01/08-hcls-minutes.html#action02]

parsa: std way to map?

vipul: next agenda: clarification to some folks; will send further
... scope of project; something accomplished within 6 to 9 months, but not too narrow a scope
... good prospects - Deni (institution in Ireland?) has identified a student to help with the tasks
... lee (online but muted) cambridge informatics, implementation of mapping model
... need to answer questions raised before mapping

<LeeF> Cambridge Semantics, actually :)

thaks, lee!

vipul: much work in mapping to be done; RIM to use case

<Vipul> ACTION: Vipul to send out the brainstorming slides to the rest of the group [recorded in http://www.w3.org/2008/01/08-hcls-minutes.html#action03]

who? fine grained details about contents of fields; this is useful in order to line up groups to handle the tasks

vipul: high level functional specs; details (RDF, Owl, ..) up to implementer

who = Parsa

parsa: is there a sematic representation of HL7RIM that we can use as basis of mapping or do we have to create one?
... every narrative has to be converted ton RDF/OWL representation

vipul: no single structure now; we are not making one; we want to make a representation that will be useful to other use cases

<LeeF> suggestion: model the RIM to the extent we have instance data for it for the use case

parsa: yes, we are not modeling entier RIM; but part that is useful to our use case still needs to bemapped

<LeeF> (ah, I guess that's what Parsa just suggested also :-)

vipul: we are sending the specs around

<JyotiPathak> Apologies -- I am signing out; need to run for another meeting.

who: RIM Is Obj Oriented; they have UML; we need to translate this to class hierarchy in automated way

who?: propose RIM-OWL demonstration;

who? look at data description not RIM model; RIM is high level

<LeeF> who = Helen, I think

vipul: look at sematic binding

helen: then go to CDA
... RIM doesn't lend much to semantics

tom: RIM is high level; not as up to date as Dan as how they do constraints
... constrain HML representation of document

vipul: look at CDA into more detail

susie (aka helan): translate directly into RDF from CDA

keep us all inthe loop!

vipul: work with Dan on HL7 descriptions, Tom RIM, Helen CDA

<scribe> ACTION: email spreadsheet [recorded in http://www.w3.org/2008/01/08-hcls-minutes.html#action04]

<Vipul> ACTION: Helen Chen to investigtat the CDA [recorded in http://www.w3.org/2008/01/08-hcls-minutes.html#action05]

updates?

vipul: see you next week!

fostel: bye!

- DRAFT -

SV_MEETING_TITLE

8 Jan 2008

Attendees

Contents

Summary of Action Items

Scribe.perl diagnostic output