W3C home > Mailing lists > Public > public-semweb-lifesci@w3.org > April 2008

Re: Multi-layered Knowledge Representations for Healthcare (was RE: An argument for bridging information models and ontologies at the syntactic level)

From: Dan Russler <dan.russler@oracle.com>
Date: Thu, 24 Apr 2008 15:41:45 -0400
Message-ID: <4810E279.9090000@oracle.com>
To: Samson Tu <swt@stanford.edu>
CC: "Elkin, Peter L., M.D." <Elkin.Peter@mayo.edu>, "Kashyap, Vipul" <VKASHYAP1@PARTNERS.ORG>, public-semweb-lifesci@w3.org, public-hcls-coi@w3.org
Hi Samson,

Here is some basic information on the use of  HL7 RIM:

You are correct that classes in HL7 may have sub-classes. To be more 
specific, by definition, once a class in HL7 is instantiated, the 
classCode and the moodCode can never be changed throughout the lifecycle 
of the instance. Therefore, operationally, the HL7 RIM ontology is 
definitively declared when the instance is created. Further granularity 
in the semantic meaning of the instance is declared in the "code" 
attribute, which contains a series of fields: Original Text; mapping of 
orginal text to an expression from a published vocabulary (e.g. SNOMED); 
and translations of the original text and original coding into other 
vocabularies. Note that although one can certainly add codes to the 
Original Text field in order to express its meaning in formal terms, it 
doesn't make sense to add or update "Original Text" for an instance once 
the instance is created. Therefore, the semantic meaning of an instance 
is fully understood only when coordinating the parts of the "expression" 
including class, mood, Original Text, and the various formal coding used 
is understood. The essential rule of Term Info in HL7 is that none of 
these parts of an "expression" may contradict the other, although each 
part may contribute to the total semantic meaning of the "expression." 
It is also important that the semantic meaning of the "class" within its 
hierarchy in the RIM and the meaning of the published code within its 
hierarchy in the published coding system not contradict each other. 
However, much work remains in order to remove contradictions in the 
hierarchies of all these ontologies when used together.

(As noted earlier, the RIM is a compromise between the very abstract, 
raw, models like ASN.1 or EAV and the more concrete models often found 
in database schemas for a narrow domain.)

What are called Archetypes in OpenEHR correspond to HL7 structures 
called Care Structures in HL7 Patient Care. These "Care Structures" 
represent aggregations of classes used to represent a medical record 
construct such as a problem list or care plan. Care Structures typical 
provide the "context" to very granular concepts. For example, by itself, 
the term "diabetes Type 2" is merely a concept. Once diabetes is placed 
within a problem list care structure for a specific patient, the "sense" 
of  what is meant by "diabetes Type 2" in a particular assertion of the 
term is more clear.

In HL7 templated CDA documents (like CCD), templates are used to bind to 
a schematron conformance test that validates that a certain XML Care 
Structures (again, aggreations of classes, attributes, and vocabulary) 
do not extend beyond a specific set of allowable constraints. Therefore, 
templates don't really add to semantic meaning. However, the do enforce 
semantic meaning, and therefore support improved interoperability.

I hope this long-winded description helps in this "multi-layered 
Knowledge Representation" discussion. How one classifies the concept of 
"context" for a given concept, or the concept of "conformance testing 
the constraints on an aggregation of structure and vocabulary" in a 
multi-layer Knowledge Representation is not clear to me.


Samson Tu wrote:

> It's good that we separate the conceptual "layers" (views) from the 
> logical/representation levels. My understanding of HL7 RIM classes is 
> that they (at least classes like Observation) are meant to be 
> *subclassed* into more specific domain classes. Templates/archetypes 
> are constraints on (and hence describe subclasses of) these 
> domain-specific classes. Instances of patient-specific HbA1c 
> observations are also instances of Observation logically. HL7 experts 
> among us can correct me if I am wrong.
> Samson
> -- 
> Samson Tu                                                     email: 
> swt@stanford.edu <mailto:swt@stanford.edu>
> Senior Research Scientist                               web: 
> www.stanford.edu/~swt/ <http://www.stanford.edu/%7Eswt/>
> Center for Biomedical Informatics Research  phone: 1-650-725-3391
> Stanford University                                          fax: 
> 1-650-725-7944
> On Apr 22, 2008, at 1:07 PM, Elkin, Peter L., M.D. wrote:
>> Dear Vipul and Dan,
>> In order to not confuse the Ontology classification with First Order 
>> / Second Order / Higher Order logics, we use Level 1 Ontologies to be 
>> domain independent (EAV just being a representational mechanism for a 
>> logical system), level 2 Ontologies are domain dependent (e.g. CDA), 
>> and level three contain defined instances as well as class based 
>> definitions.  We have been able to make these distinctions work 
>> across multiple projects.  If there is a level zero I would suggest 
>> that Metaphysics or perhaps value systems might be something that I 
>> have not seen well represented by the upper level Ontologies with 
>> which I am familiar.
>> I believe we need a final single formal representational schema where 
>> constructions defined across Information Models and Terminological 
>> Models can be validated.  This interlingua should be defined from 
>> transforms from all other valid logical languages and should empower 
>> all those SMEs familiar with any valid logical system to work as they 
>> are comfortable.  In the end, that work product must be validated 
>> through the common interlingua to ensure that meaning is preserved 
>> and therefore we are not creating unrecognized ambiguity.
>> With warm personal regards,
>> Peter
>> Peter L. Elkin, MD, FACP, FACMI
>> Professor of Medicine
>> Mayo Clinic College of Medicine
>> Baldwin 4B
>> Mayo Clinic
>> (507) 284-1551
>> fax:  (507) 284-5370
>> ------------------------------------------------------------------------
>> From: Kashyap, Vipul [mailto:VKASHYAP1@PARTNERS.ORG]
>> Sent: Tuesday, April 22, 2008 2:48 PM
>> To: dan.russler@oracle.com <mailto:dan.russler@oracle.com>
>> Cc: Samson Tu; public-semweb-lifesci@w3.org 
>> <mailto:public-semweb-lifesci@w3.org>; public-hcls-coi@w3.org 
>> <mailto:public-hcls-coi@w3.org>; Elkin, Peter L., M.D.
>> Subject: Multi-layered Knowledge Representations for Healthcare (was 
>> RE: An argument for bridging information models and ontologies at the 
>> syntactic level)
>> Dan and Peter,
>> Based on conversations on this topic, there appears to be consensus 
>> of the need for multi-layered knowledge representation schemes
>> for heatlhcare.  Will be great if we could brainstorm and come to 
>> some sort of consensus on these "layers". Would like to propose a
>> strawman as enumerated below.
>> Layer 0 = Entity - Attribute - Value or RDF triple based rerpesentations.
>> Layer 1 = MetaClasses, e.g., Observation as in HL7/RIM
>> Layer 2 = Classes in a Patient Model, Document Models, etc, e.g., the 
>> class of HbA1c results for a class of Patients.
>> Layer 3 = Data that are instances of Classes, e.g., a particular 
>> HbA1c result for a patient John...
>> As per your e-mail, you seem to be suggesting that there is something 
>> in between Layer 1 and Layer 2. However, please note that Layer 2 
>> consists
>> of classes of assertions in the patient record and not instances.
>> More reespnses are embedded in the e-mail below.
>> <dan> With apologies to Peter in case I misrepresented your SOA 
>> presentation...Last week, Peter Elkin of Mayo Clinic delivered a 
>> presentation where he called the HL7 RIM a "first order ontology" 
>> because of the abstraction level of the RIM. He called the models 
>> derived from the RIM, e.g. analytic models, patient care document 
>> models like CDA, etc, "second order ontology" because they add a 
>> layer of concreteness to the abstractions of the RIM, i.e. an object 
>> with classCode of observation and moodCode of order becomes an 
>> "observation order object" with neither a classCode nor a moodCode. 
>> [VK] Are there mathematical ways of describing these "derivations" 
>> for e.g., by using operations such as instantiations and 
>> generalizations/specializations.
>> Also, in the above, it's not clear what the semantics of an 
>> "observation order" object is?
>> For e.g., observations and orders are semantically distinct concepts, 
>> so in some sense an observation order class is likely to be 
>> unsatisfiable?
>> The semantics of "moodCode" is not clear in Knowledge Representation 
>> terms. For instance, do various mood codes partition the instances of 
>> a class
>> into subclasses that are possbily mutually disjoint?
>>  Finally, the coding systems themselves support the concreteness of a 
>> "third order ontology." For example, the SNOMED concept becomes an 
>> object itself without a code attribute, moodCode attribute, or 
>> classCode attribute, e.g. a WBC order. />
>> [VK] One way of looking at a Snomed code is that it defines a 
>> class (e.g., blood pressure) of all the instances of blood pressure 
>> readings which would imply that it belongs to Layer 2 as defined above? 
>>>         <dan> see above for the "first order to third order model."
>>>         Your metaclass looks like Peter's "first order ontology."
>>>         However, your "instances" get introduced too early...your
>>>         "instances" point to actual medical record assertions, and
>>>         Peter's model suggests that there is more "in between." In
>>>         Peter's model, the actual medical record assertion would be
>>>         an instance of his "third order ontology." />
>>>         [VK] Agree. As per the layering introduced above,  Layer 2
>>>         would correspond to classes of assetions and Layer 3 would
>>>         correspond to actual instances or assertions.
>>>          <dan> I completely agree that the HL7 RIM is one level more
>>>         "concrete" than the earlier EAV models. The EAV model
>>>         represents the ultimate in abstraction, similar to RDF
>>>         triples. Perhaps Peter would be more correct to say that EAV
>>>         is a "first order ontology" and that the HL7 RIM is a
>>>         "second order ontology." />
>>>         [VK]  Agree: As per layering introduced abiove,  The EAV/RDF
>>>         triples layer could be layer 0, and the HL7/RIM layer could
>>>         be layer 1
>>>         Look forward to further brainstorming and feedback on this.
>>>         Cheers,
>>>         ---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.
Received on Thursday, 24 April 2008 19:42:46 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:20:33 UTC