W3C home > Mailing lists > Public > public-semweb-lifesci@w3.org > August 2012

Re: seeks input on Study Data Exchange Standards An alternative approach

From: Prabhu Shankar <prv_shankar@yahoo.com>
Date: Wed, 22 Aug 2012 12:50:11 -0700 (PDT)
Message-ID: <1345665011.29463.YahooMailNeo@web39406.mail.mud.yahoo.com>
To: "Peter.Hendler@kp.org" <Peter.Hendler@kp.org>, "conor-dowling@caregraf.com" <conor-dowling@caregraf.com>
Cc: "david@dbooth.org" <david@dbooth.org>, "helena.deus@deri.org" <helena.deus@deri.org>, "kerstin.l.forsberg@gmail.com" <kerstin.l.forsberg@gmail.com>, "LINMD.SIMON@mcrf.mfldclin.edu" <LINMD.SIMON@mcrf.mfldclin.edu>, "meadch@mail.nih.gov" <meadch@mail.nih.gov>, "mscottmarshall@gmail.com" <mscottmarshall@gmail.com>, "public-semweb-lifesci@w3.org" <public-semweb-lifesci@w3.org>, "ratnesh.sahay@deri.org" <ratnesh.sahay@deri.org>
Peter,

I work with Epic and deal with data requests for research at Emory Children's. I am keen to learn about your CMT and would appreciate if we could have a Webex! 

Thanks!
 
Regards,
Prabhu




________________________________
 From: "Peter.Hendler@kp.org" <Peter.Hendler@kp.org>
To: conor-dowling@caregraf.com 
Cc: david@dbooth.org; helena.deus@deri.org; kerstin.l.forsberg@gmail.com; LINMD.SIMON@mcrf.mfldclin.edu; meadch@mail.nih.gov; mscottmarshall@gmail.com; public-semweb-lifesci@w3.org; ratnesh.sahay@deri.org 
Sent: Wednesday, August 22, 2012 2:43 PM
Subject: Re: seeks input on Study Data Exchange Standards An alternative approach
 

It is completely driven by the doctors
who use the system. Every month from our various hospitals and clinics,
in the various regions, we get "requests" to add terms to our
E.H.R. 

Some of the requests are from doctors
in the field who find something missing that they want.  Other requests,
from the look of them, seem to be from someone who is reading an ICD9 or
ICD10 book and says, "I see this in ICD10 but not in our EHR so I
will request it now". 

The requests come into CMT (we have
an organized work flow for this). It is reviewed.  Sometimes we can
get back to the requester and say, we already have this term, but we'll
add your phrase as a synonym for you.  If we don't have the term,
then we add it by adding a "display name" which is local to us,
and follows some naming conventions. 

Then it gets mapped to ICD9 and ICD10.
 Then we look for it in SNOMED. 

If there is a one to one match we map
it. 
If there is only a parent, then we map
it to the parent as an "Is A" match and we flag that for later
modeling. 
Eventually we get around to adding a
newly  modeled one-to-one term.  This is in our SNOMED extension. 
That eventually gets submitted to NLM
and IHTSDO where it has a few possible fates. 
It can get added to SNOMED international,
it can get added to the USA extension, or it can remain only in the Kaiser
extension. 

Meanwhile, we also map all terms to
a "patient facing name".  this name is what shows up on
the patient portal. 

We realize hardly anyone can maintain
such a large team just to deal with terminology.  So we donated all
of CMT to NLM.  We haven't given them it all at once because, we need
to check it for errors at least a little before we release it to the public.
 So we sequentially, with some schedule, release "bundles".
 For example, Cardiac conditions, or Neurology or Urology. 
Eventually NLM will have everything. 

I have some traveling coming up, but
we could do a webex to present CMT to this group in the future. 

We don't solve all problems related
to vocabulary, but we have a good system that does a lot, and even lets
us use SNOMED subsumption.


NOTICE TO RECIPIENT: If you are not the intended recipient of this e-mail, you are prohibited
from sharing, copying, or otherwise using or disclosing its contents.  If
you have received this e-mail in error, please notify the sender immediately
by reply e-mail and permanently delete this e-mail and any attachments
without reading, forwarding or saving them.  Thank you.
 



Conor Dowling <conor-dowling@caregraf.com>  
08/22/2012 11:17 AM  To Peter Hendler/CA/KAIPERM@KAIPERM  
cc david@dbooth.org, helena.deus@deri.org,
kerstin.l.forsberg@gmail.com, LINMD.SIMON@mcrf.mfldclin.edu, meadch@mail.nih.gov,
mscottmarshall@gmail.com, public-semweb-lifesci@w3.org, ratnesh.sahay@deri.org  
Subject Re: seeks input on Study Data Exchange
Standards An alternative approach 
  
 


thanks Peter. 

Just so I'm clear - and maybe getting too detailed for
this discussion but while I have you here - so the CMT (on
NLM here) defines selections and provides display
names for SNOMED but do you have a formal approach to the level of coordination?  

Is severity always or never coordinated etc or is the
approach "if it's in CMT, it's used" and there's no formal definition
of "this is for the model, this is for the terminology". Or is
Epic's model the determinant ie/ if Epic has two slots - "allergin:
Strawberry/ disorder: Allergy" - then you don't coordinate and use
separate terms from SNOMED but if Epic has one slot - "Disorder type:
Allergy to Strawberry" - then you coordinate? 

Conor
 
On Wed, Aug 22, 2012 at 10:28 AM, <Peter.Hendler@kp.org>
wrote: 
Well now you're going to make me admit
more.
Although we are hoping to move to a RIM like clinical statement like SMIRF
 (SMall Isolated RIM Fixed Model) as our enterprise clinical model,
what most of our data is really in, ins Epic.
Which of course is a proprietary E.H.R. system and is of course an "extensional"
model.

Now the drop down boxes where physicians pick diagnoses for example, are
populated with local codes and local terms.  This is because doctors
in general want to see the words they like, and not the "preferred
name" of SNOMED or ICD9.

We maintain via our CMT (Convergent Medical Terminology) system a back
end infrastructure.  CMT not only populates the Epic "master
files" that determine what the user sees on the GUI but we also maintain
all the mappings in a background database.  All of our local terms
for procedures and diagnosis are mapped to SNOMED.  Our drugs are
mapped to First Databank and I'm thinking probably also RxNorm.

So our whole system is at base, extensional based on Epic.  But, since
we have diagnosis and procedures mapped to SNOMED, it give us an ability
to use subsumption when ever we need to come up with a list of diagnosis
or procedures for populating work flows, preference lists or even quality
reporting or outcomes research.

We don't use any intensional ontology tricks in the "model".
 We hope to eventually mirror our data in an HL7 RIM based SMIRF model.

But as it is now, the only ontology we map to in SNOMED, and pretty much
the only reasoning we do is with the Clinical Findings and Procedures.

That said, it is a heck of a lot more useful and powerful than just using
lexical searches to find these result sets.

Your issues about all the different ways you can divide a CCD (which is
an HL7 RIM based structure) into it's RIM OO part and the SNOMED part are
true issues that are not resolved.

I may have opinions on how I'd do it, but there are no agreed ways to converge
on, and it would be hard to even state the rules if you were to come up
with a single way.

The choice might be influenced by the actual model the E.H.R. system uses.
If it has expected places for your "strawberry" then you would
put it there, but if it doesn't you'd push it into a vocabulary space or
even a post coordinated SNOMED expression.  But that's not an issue
I want to tackle now.

I'd be happy to have people agree in the simple cases of a simple Dx, to
use SNOMED in the "code" slot of the Observation, and label that
node as being safe to use a SNOMED reasoner over the Clinical Findings
hierarchy.  This doesn't answer all the questions or cover all the
things you need to do.  But it covers a heck of a lot of important
cases, and it's simple and confined and easy.  An it might get many
healthcare organisations started for the first time using any kind of subsumption
or reasoning at all, without overwhelming them.

 
NOTICE TO RECIPIENT: If you are not the intended recipient of this e-mail, you are prohibited
from sharing, copying, or otherwise using or disclosing its contents.  If
you have received this e-mail in error, please notify the sender immediately
by reply e-mail and permanently delete this e-mail and any attachments
without reading, forwarding or saving them.  Thank you.


 
Conor Dowling <conor-dowling@caregraf.com>  
08/22/2012 09:34 AM  

To Peter Hendler/CA/KAIPERM@KAIPERM  
cc david@dbooth.org, helena.deus@deri.org, kerstin.l.forsberg@gmail.com, LINMD.SIMON@mcrf.mfldclin.edu, meadch@mail.nih.gov, mscottmarshall@gmail.com, public-semweb-lifesci@w3.org, ratnesh.sahay@deri.org  
Subject Re: seeks input on Study Data Exchange
Standards An alternative approach 

  
 



Peter,  

how does Kaiser decide what level of coordination to use? Let me take RxNORM
- do you use that for drugs? - does your OO/model record dosage, route
etc or is that coordinated in the "what"? For SNOMED, do you
stay clear of coordinated allergies (Allergic to Strawberries") and
use two model fields 'Allergy' and (substance = strawberry) or do you use
as much SNOMED coordination as you can? I suppose I'm asking, for each
concept scheme, does Kaiser have a "profile" where you decide
on coordination vs what's in the "model"?  

I had to do a lot of CCD stuff last year (opinion: a big messed up, going
no-where hack) where the split of "model"/xml tags and terminologies/codes
was never set and so you can lean on a terminology or the "model"/xml
tags (see
post). This is where I see the value in the
split of concept scheme and clinical care vocabulary. It's not closed/open
per se but in guidance or at least explicit recognition of multiple choices
for expression beyond the obvious who-where-when. 

On Wed, Aug 22, 2012 at 9:12 AM, <Peter.Hendler@kp.org>
wrote: 
My answer is not that there is anything intrinsically special about "what"
vs the when why how etc.

It is just based on what we actually do, and how SNOMED is. SNOMED is very
good in the "what" of clinical models.

On the one hand we're talking about all the possible things you could do,
and on the other hand, the main real thing we really do is use SNOMED for
"what".

SNOMED is very strong in "Clinical Findings, Diseases, Observables,
Procedures"  so those pre existing SNOMED hierarchies are in
general "what" things.




NOTICE TO RECIPIENT: If you
are not the intended recipient of this e-mail, you are prohibited from
sharing, copying, or otherwise using or disclosing its contents.  If
you have received this e-mail in error, please notify the sender immediately
by reply e-mail and permanently delete this e-mail and any attachments
without reading, forwarding or saving them.  Thank you.

 
Conor Dowling <conor-dowling@caregraf.com>  
08/21/2012 11:54 PM  

To David Booth <david@dbooth.org>  
cc Peter Hendler/CA/KAIPERM@KAIPERM, helena.deus@deri.org, kerstin.l.forsberg@gmail.com, LINMD.SIMON@mcrf.mfldclin.edu, meadch@mail.nih.gov, mscottmarshall@gmail.com, public-semweb-lifesci@w3.org, ratnesh.sahay@deri.org  
Subject Re: seeks input on Study Data Exchange
Standards An alternative approach 

 
  
 




apologies if this is a tangent but why is what's identified as "extensional"
particular to health-care? "Who" said/observed/found/acted "What",
"When" and "Where" is surely a general notion. Does
it need a health-care model or ontology? Yes, particular findings or observations
or procedures can be of health but also of finance, indeed of most of life. "Constantine
boiled his wife in 326" talks of who, what and when.  

Does Health-care need its own (closed-world) model for "observation"
("kill method" == "boil") vs "finding" ("was
the boiler") or "procedure" ("dunk and hold down").
Just as we should avoid our own ontologies for demographics (a provider's
address is not different than the address of a supermarket), shouldn't
we avoid special handling for the likes of "observe", "find",
"act" and leave each domain concentrate on its particular observations,
findings et al? That is, SNOMED gets to be health-special but what here
is called the "extensional model", that is so general (again,
maybe I'm missing something), doesn't that belong with FOAF et al? 

On Tue, Aug 21, 2012 at 3:43 PM, David Booth <david@dbooth.org>
wrote: 
On Tue, 2012-08-21 at 15:11 -0700, Peter.Hendler@kp.orgwrote:
> [ . . . ]  Can you use RDF in a closed world way when ever you
want, 
> or is it only safe when the model you're dealing with, like FHIR,
> really is known to be closed world?
> 
I think so, provided that you understand that you are making the closed
world assumption, i.e., that your results reflect only what you
*currently* know.

Almost every application makes the closed world assumption at some
point. 


--
David Booth, Ph.D.
http://dbooth.org/

Opinions expressed herein are those of the author and do not necessarily
reflect those of his employer.
Received on Wednesday, 22 August 2012 19:50:41 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 18:01:13 GMT