HCLSIG/PharmaOntology/Meetings/2010-04-01 Conference Call
Conference Details
* Date of Call: Thursday April 1 2010 * Time of Call: 12:00pm - 1:00pm ET * Dial-In #: +1.617.761.6200 (Cambridge, MA) * Dial-In #: +33.4.89.06.34.99 (Nice, France) * Dial-In #: +44.117.370.6152 (Bristol, UK) * Participant Access Code: 42572 ("HCLS2") * IRC Channel: irc.w3.org port 6665 channel #HCLS2 (see W3C IRC page for details, or see Web IRC) * Mibbit quick start: Click on mibbit for instant IRC access * Duration: 1h * Convener: Susie
Agenda
* VA VistA: a real EHR on the Semantic Web - Conor Dowling - Presentation * TMO - Data - Michel - Ontology - Elgar, Colin - Interface - Bosse - eMerge/ IO Informatics - Chris, Bosse, Elgar - Outreach (BioOntologies, NRDD, IHI - Michel, Colin, Chris * Drug Ontology Update - Susie * Inclusion/Exclusion Criterion - Michel, Bosse * Other Outreach - UPenn Translational Medicine (ITMIT) - Chris - Upcoming F2F - Susie * AOB
Minutes
Attendees: Chris, Michel, Shankar, EricM, Trish, Bosse, Scott, David, Matthias, Susie
Apologies: Julia, Elgar
<Susie> Conor presents on Vista
<Susie> Vista is very scalable
<Susie> Used widely in the US
<Susie> Also used in Finland
<Susie> Biggest deployed eHR in the US
<Susie> Kaisers system is big, but not as integrated
<Susie> Used over 200 systems
<Susie> They federate, so can look at records across locations
<Susie> Most are build on RDB with custom code, HL7, some terminology
<Susie> Vista has lots of VA specific terminology
<Susie> Vista has linkage to lab systems, radiology, etc
<Susie> Vista has lots of HL7 systems
<Susie> You wouldn't see this in most open source systems
<Susie> And Vista is free
<Susie> Arbitary release progress which isn't great
<Susie> Don't have much transparency as to what is coming
<Susie> They don't need to help you to understand it
<Susie> It's a fat client in Delphia
<Susie> It's showing its age
<Susie> FileMan is the core store
<Susie> But it isn't relational
<Susie> Dates back to the 70s
<Susie> It doesn't have a remote query capability like SQL
<Susie> But do have SQL like extensions
<Susie> Have 5K RPC calls
<Susie> Neither query application is ideal
<Susie> Provides opportunities for semweb
<Susie> Fileman would likely be called graphman today
<Susie> They have lots of nodes which can be literal values or pointers
<Susie> There are key nodes, e.g. patients
<Susie> And sub nodes such as problems, vitals, allergies, etc
<Susie> It's just a directed graph
<Susie> It's really strongly typed
<Susie> Down to the pointer level
<Susie> If map to OWL, will get a very detailed schema that you can extrapolate from
<Susie> Has extras like similie project at MIT
<Susie> Has concept of internal and external forms
<Susie> Can walk through system using linked data
<Susie> Want to preserve this as converting system to semweb
<Susie> They think about XML/SOAP/SOA for connectors
<Susie> That's the same with CDISC, and other pharma companies
<Susie> Need to put the terminology into their terms
<Susie> So call SPARQL a fileman query, etc.
<Susie> You don't need to tell them its sparql
<Susie> First release is focused on linked data
<Susie> Use stepping stones
<Susie> Let them walk through all data in vista
<Susie> Need to select all
<Susie> Need to describe a node - all predicates
<Susie> Need to see all nodes that refer to the patient
<Susie> Then display this in a web page
<Susie> Let them download what we have
<Susie> Make it clear that it isn't a smoke and mirrors demo
<Susie> So utility then terminology
<Susie> Christopher Jones is the sample patient
<Susie> Get all information that describes him
<Susie> Show that you can remove extraneous data
<Susie> When proposed in Jan
<Susie> A number of people didn't like the idea of going directly to fileman
<Susie> Concerned that he'd miss important information that is in the business logic
<Susie> RPC calls are very similar
<Susie> Just need a filter
<Susie> Business logic is usally re-formatting, and some filtering
<Susie> So they shouldn't be concerned about going to a relational store
<Susie> Have their own code sets equivalent to snomed
<Susie> 2 gaps - going through general ontology to va ontology
<Susie> Need to go through va code sets, and the ones we want to use in general
<Susie> Big payoff with semweb is there isn't a whole bunch of middleware
<Susie> Link one schema hopefully using equivalence to another, and one code set to another, then you can query it
<Susie> So can go from vista system to more general one
<Susie> Minimizes amount of mapping code
<Susie> Take va terminology into separate system and map to world in general
<Susie> Would then query the separate system that maps to va
<Susie> Think this is the optimal way to go
<Susie> Va gives its drug files to RxNORM
<Susie> And the connections are included
<Susie> Haven't done anything with the schema gap
<Susie> Want to talk about the optimal schema to map va to
<Susie> How can we connect our demos together
<Susie> Can we use TMO
<Susie> VistA is free
<Susie> The VistA world, clinical world, physicians need to be educated about Linked Data
<Susie> Need to show them things in their terms
<Susie> Like to see how our demos can connect to others
<Susie> Caregraf is a start up
<Susie> Want to sell solutions that fill healthcare gaps that arise from healthcare reform
<Susie> Focus on using semweb soluations
<Susie> Next year will be big roll of eHR to smaller practices
<Susie> Not publicly selling anything yet
<Susie> Trying to promote technologies on the web site
<Susie> Chris: How do you deal with confidentiality in the demo
<Susie> The VA system ships without data
<Susie> Medsphere have created some pretend patients
<Susie> We did a version of SPARUL to load data
<Susie> These are semi-realistic clinical scenarios
<Susie> A web server would provide security
<Susie> Chris: Does it make patient care better?
<Susie> Once we have linked data for everything
<Susie> We can begin to show how we can make things better for physicians who are treating patients
<Susie> The closet think so far is wha't under the patient record for Christopher Jones
<dbooth> Thank you Conor! Excellent presentation! Sorry I need to drop off.
<Susie> Need good example data
<Susie> And would need to fill in lots of people to demonstrate scalability
<Susie> Do we focus on filters on server side
<Susie> Or make sure you can put a whole graph for a patient into the system
<Susie> Identify disease, make better prognostics, or make better decisions regarding drug disc/dev
<Susie> Have been working on interface with Eric Miller
<Susie> And have created a flow diagram
<Susie> But have also spoken to Marshfield Clinic about gaining access to data
<Susie> CDISC very focused on XML standards
<Susie> Are getting support and funding from FDA