RE: [COI] Clinical Observations Interoperability Telcon January 15 2008

 

	I think I need to digest your Brainstorming slides first.
	[VK] Please let me know if you have and questions. Would be great to get
your feedback. 
	

		I was responding to Helen's request for information about  "a
good ontology version of RIM/CDA"
		

			[VK] That is the one of the goals of the COI project.
The methodology is to use some use cases to drive the analysis and design work
required to come up with an ontology version of the RIM/CDA. 
			

	Can you explain what you mean by "ontology version" of the RIM/CDA? What
is the meaning of an ontology version of an information model? How would it
different from Marley's RIM OWL ITS?
	 
	[VK]  My response to the above pre-supposes a particular view of what an
information model is. In general, the goal is to incorporate various artifacts
that are likely to be useful in a wide variety of use cases, e.g., patient
	recruitment, etc. So as discussed in the earlier e-mail, one possibility
is to pull in information model elements (e.g., Observation), controlled
vocabulary concepts (e.g., Snomed, etc.) and Data Types in one specification
	which could perhaps be RDF(S)/OWL The difference with Tom Markey could
be that our approach would cover more than just HL7 data types. Am not clear
what Tom's approach is as there is nothing on the wiki.
	Our approach would be more along Alan Rector's work on binding Snomed to
HL7.   

			There are so many ways to model eligibility criteria.  I
am not sure I understand "filling this table with HL7/RIM/CDA constructs." The
RIM Acts have criterion mood. So an eligibility criterion involving "Serum
Creatine," for example, can be an Observation Act in criterion mood with code =
a terminology code for Serum Creatinine and value = an interval of Physical
Quantity (and maybe effectiveTime to express temporal constraint. Presumably Tom
Marley's OWL ITS of the HL7 RIM would give you such a RIM construct the
criterion in OWL syntax. (
			[VK] This is very similar to the discussion I had with
Dan Russler. In my mind you seem to be proposing a methodology to express an
eligibility criteria (Serum Creatinine value > X).
			Observation.code for Serum Creatinine
			Act.effectiveTime for getting the time of the
observation if required.
			Observation.value = Interval of Physical Quantitiy

	I was not proposing a methodology. I was indicating my understanding of
"HL7/RIM/CDA construct" that represents a criterion.
	[VK] Understood. But the approach you outlined above, which was along
the lines of my discussion with Dan Russler, could very well be the basis of a
methodology or "best practice". Of course, it needs to be fleshed out in greater
detail.
	Another clarification, which is also discussed in the brainstorming
slides as well, is that we are trying to model things in two steps: (A) model
the data/information items; and (B) use these to represent Eligibility
criterion. 
	Whereas RDF/OWL might be good for (A), we ny have to go to a rules based
approach for (B), which referts to RDF/OWL classes, very much along the lines of
SWRL.     
	 
	Those properties are "take together" as an instance of the Observation
in criterion mood. HL7 Observation in criterion mood is a data structure, which
can have a representation in OWL syntax, once you model the RIM in OWL, as Tom
Marley has done. If you want to model the criterion as an OWL class expression
that a DL reasoner can evaluate, that's a different matter entirely, and I am
not sure HL7 RIM helps you. So I don't know what kind of OWL expression you are
looking for. 
	 
	[VK]  HL7/RIM is one of the candidate standards we are looking at and as
you describe above, it seems to be rather high level with the "semantics" really
coming from Vocabulary Domains and/or Terminology Bindings. The goal is not
necessarily to represent the
	HL7/RIM data structure using OWL syntax, which is likely to be of
limited utility, but to be able to capture and represent the semantics of
various types of clinical observation such as Vital Signs (Blood Pressure,
Height, Weight, etc.);   Laboratory Observations (HbA1c,
	etc,). This is similar to the work being done by Stan Huff and Tom Oniki
on Detailed Clinical Models and as suggested by Helen, some of it has also been
captured in the CDA. Hope that clarifies where we are coming from.  
	
	 GELLO assumes a fairly minimal object-oriented model. I don't see
RDF/OWL as having particular advantage in representing data. My past experiences
suggest that criteria will involve temporal/data/terminological abstractions. At
the moment I don't know of any language that is completely satisfactory.
	 
	[VK] Would like to know more about why you think RDF/OWL do not have an
advantage in representing data/information models. Would presume that it would
be a better choice compared to XML Schema for sure. Also, the ability to define
a set of constraints
	and axioms; and in the case of OWL 1.1, role chains might bring in some
advantages.
	 
	Agree with you on the temporal constructs and abstractions which are
very prevalent in the eligibility criteria I have looked at on
http://www.clinicaltrials.gov whcih is why OWL may (or may not be a suitable
choice), but a rule language could.
	But these are issues/best practices we hope to develop based on concrete
examples of criteria and real patient data..
	 
	Hopefully, I have clarified some of the questions and issues you have
raised. Let me know a good time to have a conversation and we can probably have
a more detaile discussion at that point.
	 
	Thanks and Regards,
	 
	---Vipul   
	
	


The information transmitted in this electronic communication is intended only
for the person or entity to whom it is addressed and may contain confidential
and/or privileged material. Any review, retransmission, dissemination or other
use of or taking of any action in reliance upon this information by persons or
entities other than the intended recipient is prohibited. If you received this
information in error, please contact the Compliance HelpLine at 800-856-1983 and
properly dispose of this information.