See also: IRC log
<MarkW> Hey Swoosh! Grats on the paper, dude!!!
<Joanne_Luciano> Mark Dumontier - OSCAR health care records
<Joanne_Luciano> To provide OSCAR with controlled vocab and pharmacogenomic
please mute your microphone
<mscottm> Marc Dumontier
<Joanne_Luciano> (thanks for the correction! (on spelling of Marc's name)
<bobP> ok now?
<Joanne_Luciano> There is an echo
<mscottm> bobP - the echo seems to have originated from GVoice, not your phone
<bobP> k
<Joanne_Luciano> Physicans don't want to type in code - EVER, so will be in form of narrative in the patient chart
<Joanne_Luciano> so, we're stuck with getting the info out of the record narrative -
<Joanne_Luciano> Mark and Michel talked about developing a system that provided a form on the side of the screen that included the concept that they can edit (remove) if incorrect
<matthias_samwald> JBoss Drools
<Joanne_Luciano> would someone type in what matthias just said?
discussion about the feasibility of using OpenCDS, since it's based on JBoss Drools
1. NLP parsing of narrative
<matthias_samwald> matthias: OpenCDS = JBoss Drools + HL7 "Virtual Medical Record" interface for CDS + Apelon DTS for taxonomy management
NCBO annotator - Nigam Shah has used this to do concept recognition and large scale associations - noisy part doesn't really matter in this case
<Joanne_Luciano> I'm wondering if we can do 'training' on the entries, like what's done with voice recognition
each part of the clinical record contains different kinds of information;
specialist - cardiology, dermatology - all context that could be used to downweight suggestions
semantic drupal could be effective
marc: mostly interested in progress notes - iterative development to make it more intelligent - pull out from past notes
template driven - needs to be filtered out
<mscottm> michel - who is talking? :)
Rafael richards from baltimore
<mscottm> Raphael Richards (MD from Baltimore)
rafael: semantic vista
... problem oriented approach would be more effective to track
progress
marc: oscar does support this, but physicians don't enter the codes, so it can't be tracked properly
rafael: codes are typically used for billing and reimbursement
marc: being able to not enter codes is the goal
<BobF> I'm back - is there a webex or slides, or just voice today?
mscottm: snomed codes for
insurance; can't expect physician to pull out the code out of a
dictionary; in our terminology task force with john madden,
daniel rubin, you could have a report validation service
... comes back with questions to resolve ambiguities
... if typing free text in a conclusion section, they can be
assisted with autocomplete to access the preferred terminology
encoding system; end up with semantically annotated content -
"behind" the free text
... try to resolve ambiguity at the point of data entry
rafael: smart interactive
questions
... how to capture high quality data?
... are there open source tools to give you eclipse-like
autocompletion, etc?
mscottm: NCBO has widgets for autocompletion
here's the NCBO REST API docs: http://www.bioontology.org/wiki/index.php/BioPortal_REST_services
rafael: need a clean interface to compose a note - and it's going to load up phrases, vocabulary, etc; note about patient blood pressure; help you write high quality notes
mscottm: John Madden had an XML-based approach for reporting; anything that works in RDF?
michel: how does OSCAR architecture facilitate the incorporation of new tool and support
marc: no plugins architecture as
of now; AJAX like call or real time web service call; thinking
about having a second window that acts as a support;
... web service calls would be best
bobf: a couple of examples that i can provide of using a service call to enter content; protege plugin for semi-automated content markup
<BobF> Semantator (Protege plug-in for semi-automated markup of clinical notes using ontologies): http://informatics.mayo.edu/CNTRO/index.php/Semantator
marc: large beta-testing community with different patient demographics - should be able to get good feedback in a reasonable amount of time; measure success - cost to provider
2. CDS
marc: CDS for 2 projects a.
ontario renal network - screening patients for potential renal
disease
... to what degree should we be considering a lab test for
chronic kidney disease?
... work with diabetes clinics; monitor data; to be able to
organize the data would already be helpful; could be applied to
many diseases
... b. natal care - reminders based on prior lab tests; or
potential test suggestions
... currently hard coding specific things which could be more
generic ; sounds like opencds could be useful here
... not too sure how easy it is to author rules; but access to
it via service seems convenient
matthias: opencds makes use of
JBoss drools - rule framework - tightly integrated java api;
seemingly quite efficient over moderate size kb;
... rules are stored in a format defined by JBoss drools -
positive is that it is an industry grade rules framework - but
not sure how easy it is to translate to another format
... SOAP-based web service provides access; only send data (?)
useful for CDS.
... opencds mixes standards based approach with
pragmatism
... fairly complex; need lots of knowledge engineers to encode
this into the system;
... not sure if I would use opencds - still a bit new - not
much in terms of medical rules - system seems a bit heavyweight
and complex
marc: like the idea of having a
central webservice for this; oscar deployed at many clinics -
could imagine each building rules; being able to share these
would be important without having to redistribute them
... pretty complex; not too sure how to build rules; have to
create drool files
matthias: used the rules that
were provided, never created my own
... want to make it as easy as possible to create these rules;
don't want to necessarily interact with IT professionals
... from the W3C perspective - how could parts of the framework
be connected to SW technologies
... other kinds of terminology or CDS services
... drools developers are looking at adding RDF/OWL
mscottm: what are the features of
drools that attract you? cecil lynch mentioned jess for a rule
engine - but it has a commercial license
... rules are a form of knowledge that ought to be
shareable;
matthias: being able to integrate it into other workflows would be important
<BobF> (follow up from the first agenda topic...) Here are some toy examples of autocomplete using a terminology service on the back end: http://informatics.mayo.edu/cts2widgets/search.html (autocomplete on the 'Ontology' field), http://informatics.mayo.edu/cts2/rest/widgets/viewmap.html (pulls from an eXist XML database)
matthias: with semantic web technologies - OWL reasoning, SPARQL queries
<mscottm> BobF - interesting - there is an EHR system in NL based on eXist, XForms, XQuery, etc.
This is scribe.perl Revision: 1.136 of Date: 2011/05/12 12:01:43 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) No ScribeNick specified. Guessing ScribeNick: michel Inferring Scribes: michel WARNING: No "Topic:" lines found. WARNING: No "Present: ... " found! Possibly Present: Bob_Powers GVoice Joanne_Luciano MacTed MarkW P18 P20 P21 P26 P28 P7 Suajn Tony aabb aacc aadd aaee aaff aagg achille_z benvvalk bobP bobf ericP jun luke marc matthias matthias_samwald matthias_samwald1 michel mscottm rafael sujan swoosh You can indicate people for the Present list like this: <dbooth> Present: dbooth jonathan mary <dbooth> Present+ amy Got date from IRC log name: 15 May 2012 Guessing minutes URL: http://www.w3.org/2012/05/15-HCLS-minutes.html People with action items: WARNING: No "Topic: ..." lines found! Resulting HTML may have an empty (invalid) <ol>...</ol>. Explanation: "Topic: ..." lines are used to indicate the start of new discussion topics or agenda items, such as: <dbooth> Topic: Review of Amy's report[End of scribe.perl diagnostic output]