W3C home > Mailing lists > Public > public-hcls-coi@w3.org > April to June 2008

Re: An argument for bridging information models and ontologies at the syntactic level

From: <dirk.colaert@agfa.com>
Date: Tue, 15 Apr 2008 15:37:02 +0200
To: dan.russler@oracle.com
Cc: Dan Corwin <dan@lexikos.com>, "Ogbuji, Chimezie" <OGBUJIC@ccf.org>, public-hcls-coi@w3.org, public-semweb-lifesci@w3.org, public-semweb-lifesci-request@w3.org, rector@cs.man.ac.uk, Samson Tu <swt@stanford.edu>, "Oniki, Tom (GE Healthcare, consultant)" <Tom.Oniki@ge.com>, "Kashyap, Vipul" <VKASHYAP1@PARTNERS.ORG>
Message-ID: <OFCA452DC9.AEC7582C-ONC125742C.004981B5-C125742C.004ACBE5@agfa.com>
Sorry to drop in this discussion. I would argue that the difference 
between a diagnosis and a medical sign is the following:
a diagnosis is an assessment of the patient's medical condition by means 
of which the physician tries to categorize that condition into a known set 
of (artificial) conditions (called diseases). I call these conditions 
"artificial" because they are an abstraction of the physio pathology as we 
understand it. They may change in time. What we called HIV we know 
understand to be a conglomerate of related diseases and maybe HIV as such 
is not to be kept as an entity. Because this is an assessment the 
physician can be wrong and he (or someone else) may have to re-classify 
the patients condition.

a medical sign is a measurement and is obtained by a procedure. multiple 
procedures kan lead to the same sign and each procedure has its own 
sensitivity and specificity. Based on that decision support systems can 
even suggest one procedure above the other. Most of the medical signs are 
physical parameters (pulse rate, temperature, ...) but they don't have to 
be. For example an ADL score of 24 means an 'Acttivity of Daily Life' 
score obtained by some questionnaire




Kind Regards,

Dirk Colaert | Agfa HealthCare
Advanced Clinical Application Manager | HE/Advanced Clinical Applications 
Research
T  +32 3444 8408 | F  +32 3444 8401

Quadrat NV, Kortrijksesteenweg 157, 9830 Sint-Martens-Latem, Belgium
http://www.agfa.com/healthcare/
R.O.: Septestraat 27, B-2640 Mortsel, Belgium | RLE Ghent | VAT BE 
0426.312.921 | IBAN BE95570125275558 | Citibank International PLC, B-1050 
Brussels
Click on link to read important disclaimer: 
http://www.agfa.com/healthcare/maildisclaimer



Dan Russler <dan.russler@oracle.com> 
Sent by: public-semweb-lifesci-request@w3.org
11/04/2008 14:05
Please respond to
dan.russler@oracle.com


To
"Ogbuji, Chimezie" <OGBUJIC@ccf.org>
cc
Dan Corwin <dan@lexikos.com>, "Oniki, Tom (GE Healthcare, consultant)" 
<Tom.Oniki@ge.com>, Samson Tu <swt@stanford.edu>, "Kashyap, Vipul" 
<VKASHYAP1@PARTNERS.ORG>, rector@cs.man.ac.uk, 
public-semweb-lifesci@w3.org, public-hcls-coi@w3.org
Subject
Re: An argument for bridging information models and ontologies  at  the 
syntactic level







Hi Chimezie,

I've been watching this discussion. It parallels a discussion we had in 
HL7 in the 1990's.

Some comments below:

Ogbuji, Chimezie wrote:
Dan,

I've very familiar with the SOAP model.  The primary motivation for my 
questions about assessment had more to do with distinguishing an action 
from data that is derived from it.  This speaks directly to the problem of 
the 'anti-pattern' where ontologies for medical records are built 
*directly* from models that were designed with data-level concerns in mind 
and thus semantically inconsistent (so called "information models").

The sense of assessment as used in this paper suggests that an assessment 
is data (and thus appropriate to consider a diagnosis), but consider that 
there are other senses of the word and one in particular is "the act of 
judging or assessing a person or situation or event".  In the latter case, 
an assessment refers to the act.  I was simply trying to tease out which 
of these Tom had in mind.
<danR> It is true that in traditional lab department systems, the 'data 
from the assessment' was modeled separately from the 'assessment action.' 
This is not exactly "wrong." However, it was noted that one cannot deliver 
a "numeric result" without restating the action that generated the result, 
e.g. serum WBC is the action and serum WBC of 10,000 WBcells/ml is the 
result. In physical sciences, it is considered good practice to always 
include the methodology of the action when describing the data. 
Accordingly, it is best practice in the science of healthcare to also 
report on the nature of the action itself at the same time as reporting on 
the data derived from the action. Once the "best practice" is defined, 
then one can evaluate individual conditions. What if the result is ordinal 
rather than numeric? What if the result is categorical rather than ordinal 
or numeric? To a health scientist, best practice does not change depending 
on the datatype of the result.


On a separate note, I would argue, however, that a diagnosis is a 
hypothesis of a model of a health condition and thus should be separate - 
at least in an ontology - from a tangible (objective) health condition or 
phenomenon.  This is why (at least in my CPR ontology [1]) a diagnosis is 
'data' and a disease is a real / physical "pathological disposition." 
After all, it is my understanding that a diagnosis can later be found out 
to be incorrect whereas a medical sign rarely is (by virtue their being 
completely objective and determined physically). 
<danR> It is a common misconception that medical signs are rarely wrong. 
The whole science of laboratory and physical exam practice is based on how 
to reduce the incidence of incorrect findings. Many writings in the 
history of medicine and the history of science have come to the conclusion 
that any differences between "diagnosis" and "medical sign" cannot be 
established on the basis of "frequency of incorrectness." In fact, most 
scientists in this area will state that the only difference between 
medical sign and diagnosis is the subject of the observation. "Diagnosis" 
might be considered to have a subject of observation of data derived from 
the patient whereas medical sign has a subject of observation of the body 
of the patient. So now that the basic science and best practice are 
established, you can start to define an "ontology" that meets the needs of 
science and medical practice.


Actually, I think it would be even more foolish for us to claim to be 
interested in the use of ontology for medical record representation 
without the due diligence of being very explicit about our definitions and 
their distinctions from one another.  After all, this is the very 
advantage of using rigorous, expressive languages to represent medical 
records electronically.  Otherwise, we should simply stick to paper charts 
(which is the context from which the SOAP model was formulated, by the 
way).

> (1) pick SOAP as a de facto standard;

I'm not sure how this would work as SOAP represented (originally) 
categories of notes (not even structured content).

> (2) formalize it by citing any well-written spec on "Objective" vrs 
"Assessment" (like the ones found in the above paper);

You could certainly cite such a source, but I would be wary of doing a 
copy-and-paste of SOAP (my previous point).

> (3) build on top of that foundation, *separately* for each of SOAP's 
four disjoint aspects.

I would claim that in the CPR ontology the disjoint aspects are accounted 
for:

Subjective:  This would constitute data derived from clinical actions such 
as cpr:medical-history-screening-act ("Any diagnostic act which attempts 
to determine / realize anamnesis")

Objective: This would consist of data derived from clinical actions in the 
categories listed in the above paper: cpr:laboratory-test, 
cpr:clinical-examination.  More specifically, cpr:images and 
cpr:medical-signs would constitute most of this.

Plan: cpr:substance-administration, cpr:procedure, (progress notes are 
admittedly problematic since they are narrative by definition)

Assessment: cpr:clinical-diagnosis

So, I would think of SOAP as applicable to how the underlying structured 
(and ontologically-committed) medical record is presented to a clinician 
but should not be sole basis for how to go about modeling the parts of the 
record.  A much deeper analysis is needed to derive the distinctions 
needed for this.


[1] http://purl.org/cpr/owl

Chimezie (chee-meh) Ogbuji
Lead Systems Analyst
Thoracic and Cardiovascular Surgery
Cleveland Clinic Foundation
9500 Euclid Avenue/ W26
Cleveland, Ohio 44195
Office: (216)444-8593
ogbujic@ccf.org




-----Original Message-----
From: Dan Corwin [mailto:dan@lexikos.com]
Sent: Fri 4/11/2008 3:20 AM
To: Ogbuji, Chimezie
Cc: Oniki, Tom (GE Healthcare, consultant); Samson Tu; Kashyap, Vipul; 
rector@cs.man.ac.uk; public-semweb-lifesci@w3.org; public-hcls-coi@w3.org
Subject: Re: An argument for bridging information models and ontologies at 
the syntactic level

Chimezie -

The SOAP pattern is widely used for writing medical charts.  As
a well-known industry model, it can clarify related terminology.
Its parts describe the "Subjective", "Objective", "Assessment", and
"Plan" aspects of a typical medical care-giving encounter.

Each part uses its own *separate* professional terms and (often)
specialized syntax too - like codes, units of measurement, etc.
This is appropriate, for the four parts model *disjoint* things:

   http://hillside.net/plop/2005/proceedings/PLoP2005_tsorgente0_1.pdf

In this paper (and many like it), "Assessment" is equivalent to
"diagnosis" - a model of health conditions in the actual patient.
This chart aspect should (and usually does) give pathophysiology
views, hopefully accurate, in some local "Assessment" ontology.

Other SOAP parts model other things/statements that are "evidence"
to justify (and not preclude) the "Assessment"; or else are medical
interventions (if any) which that "Assessment" suggests may help.
Coding can occur or not in any part.  The parts differ because of
*what* they model, not because of *how* they attempt to do that.

It would foolish to ignore such a tested, widely-ingrained pattern,
or spend much time trying to rearrange its details.  We'll do far
better to just (1) pick SOAP as a de facto standard; (2) formalize
it by citing any well-written spec on "Objective" vrs "Assessment"
(like the ones found in the above paper); (3) build on top of that
foundation, *separately* for each of SOAP's four disjoint aspects.

regards,
Dan Corwin


Ogbuji, Chimezie wrote:
> Thanks for that drill down, Tom.  It gives a good backdrop to call out
> important distinctions
>
> [[[
>       * Clinical care - what we do to the patient based on our
> assessments of the pathophysology of the patient
>       *       I'm not sure if Alan is deliberately saying that this
> "clinical care" level addresses actions taken *based on*
> assessments/observations, but that assessments/observations themselves
> *do not* fall into this category, but belong elsewhere (in the "record"
> category below?)
> ]]]
>
> My understanding is that the distinction here is between the
> characteristics of the patient (pathophysiology) and the clinical
> actions that are part of the provision of care (diagnostic actions,
> therapeutic actions, etc..).  Following this distinction, I would
> consider an assessment to belong to this category, but I would not
> consider an observation or a finding to fall in this category.  The
> former is an action (or what BFO would call a Process), the latter are
> data points recorded as a result of a clinical action.
>
> [[[
>       *       Here we have a "hypertension"
> observation/finding/diagnosis.
>               *       A hypertension observation may be a subclass of
> a more generic "observation"
>                       *       It *makes reference to* the
> "hypertension" of the pathophysiologic realm above, but is not
> equivalent to that hypertension.
> ]]]
>
> Yes, this separation is critical IMHO.  Findings, diagnoses, measurement
> data, and observations are sections of a medical record
> (representational artifacts) that are *disjoint* from what they refer to
> (in the pathophysiologic realm).
>
> [[[
>       * The record of that care and those assessments
>       *       At this level, the hypertension record and the heart
> rate record speak of RIM- or Archetype- or Clinical Element
> Model-specific data structures.
>               *       It models from data capture/data exchange/data
> retrieval use cases, which differs from the above levels
>                       *       e.g., it allows a heart rate measurement
> without a measurement location, even though in reality every heart rate
> measurement is taken from some location.
>               *       Significantly, it is here that codes and code
> systems are introduced -- the records reference these codes and code
> systems, which as Alan says are "meta" to the pathophysiology realm
> above, i.e., they're data structures that attempt to represent that
> realm.
> ]]]
>
> I agree, however, with the caveat that I would group observations,
> measurement data, and findings here (as a record of care and in the
> realm of information models such as RIM, Archetypes, etc..).  The term
> measurements is a bit ambiguous, so I wonder if what you had in mind was
> the process (or action) of measurement or the resulting data. 
>
> Chimezie (chee-meh) Ogbuji
> Lead Systems Analyst
> Thoracic and Cardiovascular Surgery
> Cleveland Clinic Foundation
> 9500 Euclid Avenue/ W26
> Cleveland, Ohio 44195
> Office: (216)444-8593
> ogbujic@ccf.org
>
> 
>
>
> ________________________________
>
>       From: public-semweb-lifesci-request@w3.org
> [mailto:public-semweb-lifesci-request@w3.org] On Behalf Of Oniki, Tom
> (GE Healthcare, consultant)
>       Sent: Thursday, April 10, 2008 11:28 AM
>       To: Samson Tu; Kashyap, Vipul; rector@cs.man.ac.uk
>       Cc: public-semweb-lifesci@w3.org; public-hcls-coi@w3.org
>       Subject: RE: An argument for bridging information models and
> ontologies at the syntactic level
> 
> 
>
>       It seems like we're having difficulty getting to a point where
> we can see if we agree or not.
>
>       Let me attempt to further a convergence by drilling down on what
> Alan said earlier about the 3 levels of medicine, giving my
> thoughts/interpretation:
>
> 
>
>       * Pathophysiology - what we know about the patient
>
>       *       A patient may have hypertension
>
>               *       The patient actually has to have some certain
> "kind" or "subtype" of hypertension, e.g., "labile diastolic
> hypertension" - no patient simply has "hypertension"
>               *       The patient's hypertension has a severity, e.g.,
> "mild"
>
>                       *       In this realm, the severity is mandatory
> - *every* instance of hypertension has a severity
>
>               *       A patient has a heart rate, and the heart rate
> has a value, e.g., "60 bpm"
>
>               *       In this realm, every patient has a heart rate
> and every heart rate has a value, even if it's "0"
>
>       * Clinical care - what we do to the patient based on our
> assessments of the pathophysology of the patient
>
>       *       I'm not sure if Alan is deliberately saying that this
> "clinical care" level addresses actions taken *based on*
> assessments/observations, but that assessements/observations themselves
> *do not* fall into this category, but belong elsewhere (in the "record"
> category below?)
>       *       Regardless, I'm sure he's saying that
> assessments/observations must be regarded separately from the
> pathophysiology realm above.
>       *       For the sake of this drill-down, I'll assume
> observations/measurements/assessments fall into this "clinical care"
> category.
>       *       Here we have a "hypertension"
> observation/finding/diagnosis.
>
>               *       A hypertension observation may be a subclass of
> a more generic "observation"
>
>                       *       It *makes reference to* the
> "hypertension" of the pathophysiologic realm above, but is not
> equivalent to that hypertension.
>                       *       One may observe that a patient has
> "hypertension", even though, as noted above, no patient actually has
> "hypertension", but must have some particular subtype of hypertension.
> (Or maybe at this level we would say the observation still has to be of
> a certain type of hypertension, and it's only at the next level - the
> record of care - that we're permitted to say the patient has generic
> "hypertension"?)
>
>                       *       We may also observe the severity of the
> hypertension, i.e., a severity observation is an observation about the
> hypertension's severity described above in the pathophysiology realm.
>
>                       *       One can make a hypertension observation
> without making a severity observation although, as stated earlier, in
> the pathophysiologic realm, every hypertension has a severity.
>
>               *       Here we have a "heart rate measurement"
>
>               *       We measure the heart rate at a location, e.g.,
> the left wrist
>
>                       *       Every heart rate measurement has a
> location - you can't measure the heart rate without measuring it at a
> location
>
>                       *       The value of the heart rate measurement
> might not be the "true" value of the (pathophysiologic) heart rate, but
> we accept it as an approximation
>
>       * The record of that care and those assessments
>
>       *       At this level, the hypertension record and the heart
> rate record speak of RIM- or Archetype- or Clinical Element
> Model-specific data structures.
>
>               *       It models from data capture/data exchange/data
> retrieval use cases, which differs from the above levels
>
>                       *       e.g., it allows a heart rate measurement
> without a measurement location, even though in reality every heart rate
> measurement is taken from some location.
>
>               *       Significantly, it is here that codes and code
> systems are introduced -- the records reference these codes and code
> systems, which as Alan says are "meta" to the pathophysiology realm
> above, i.e., they're data structures that attempt to represent that
> realm.
>
> 
>
>       Look forward to your comments.
>
> 
>
>       Tom
>
> 
>
>       ________________________________
>
>               From: public-hcls-coi-request@w3.org
> [mailto:public-hcls-coi-request@w3.org] On Behalf Of Samson Tu
>       Sent: Wednesday, April 09, 2008 11:41 PM
>       To: Kashyap, Vipul
>       Cc: public-semweb-lifesci@w3.org; public-hcls-coi@w3.org
>       Subject: Re: An argument for bridging information models and
> ontologies at the syntactic level
>
> 
>
>       On Apr 8, 2008, at 6:50 PM, Kashyap, Vipul wrote:
>
> 
> 
> 
>
>       Fundamentally, the only interpretation that works is to regard
> codes as being "meta" to the ontology.  I.e. the individuals in the
> ontology are things in the conceptualisation of the world - cases of
> diabetes, people, livers, etc. - individual codes represent classes in
> the ontology. 
>       [VK] Agree. Codes represent classes in some ontology or
> information model.
>
> 
>
>       IMHO, codes don't represent classes in some information model.
> An information model has classes like Observation, whose instances are
> clinical statements made by some entity (person or machine). I think
> information model is "meta" in the sense that its instances are
> statements  (The observation that "John has diabetes") about something
> that happens in the real world (the person named John has an instance of
> Diabetes).  In BFO term, the observation is an instance of
> information-content-entity, as opposed to an assertion about the John
> instance of Person and an instance of Diabetes.
>
> 
> 
> 
>
> 
>
>        The entire information structure - HL7 or Archetypes  - in
> fact, is at a meta-level.
>       [VK] Agree with this. In particular, the HL7/RIM has a very
> confusing construction. It could be viewed as a meta-model but then it
> also has fields to store patient data,
>
>       For e.g., one may view a class of lab values, say HbA1c as an
> instance of the RIM Observation class (making it a meta-class), however,
> the RIM Observation class also
>
>       has the value field for the value of those labs and is in some
> sense a multi-layered representation, which is probably why it is so
> confusing.
>
> 
>
>       I don't understand how a class of HbA1c can be an instance of
> the RIM Observation class.  I don't see how the Observation class having
> the value field is the issue.
>
> 
>
>       Samson
>
>
> ===================================
>
> P Please consider the environment before printing this e-mail
>
> Cleveland Clinic is ranked one of the top hospitals
> in America by U.S. News & World Report (2007). 
> Visit us online at http://www.clevelandclinic.org for
> a complete listing of our services, staff and
> locations.
>
>
> Confidentiality Note:  This message is intended for use
> only by the individual or entity to which it is addressed
> and may contain information that is privileged,
> confidential, and exempt from disclosure under applicable
> law.  If the reader of this message is not the intended
> recipient or the employee or agent responsible for
> delivering the message to the intended recipient, you are
> hereby notified that any dissemination, distribution or
> copying of this communication is strictly prohibited.  If
> you have received this communication in error,  please
> contact the sender immediately and destroy the material in
> its entirety, whether electronic or hard copy.  Thank you.
>
>
>
>



===================================

P Please consider the environment before printing this e-mail


Cleveland Clinic is ranked one of the top hospitals
in America by U.S. News & World Report (2007). 
Visit us online at http://www.clevelandclinic.org for
a complete listing of our services, staff and
locations.


Confidentiality Note:  This message is intended for use
only by the individual or entity to which it is addressed
and may contain information that is privileged,
confidential, and exempt from disclosure under applicable
law.  If the reader of this message is not the intended
recipient or the employee or agent responsible for
delivering the message to the intended recipient, you are
hereby notified that any dissemination, distribution or
copying of this communication is strictly prohibited.  If
you have received this communication in error,  please
contact the sender immediately and destroy the material in
its entirety, whether electronic or hard copy.  Thank you.
 
Received on Tuesday, 15 April 2008 14:36:52 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 15 April 2008 14:36:52 GMT