[Gnumed-devel] Re: Gnumed-devel Digest, Vol 5, Issue 7

[sjtan]

On Thu, 2003-04-03 at 21:54, address@hidden wrote:

> 
> Today's Topics:
> 
>    1. Re: interoperable gehr (Karsten Hilbert)
> 
> 

> 
> > I was looking at some of the gehr stuff and at Fowler: has any authority
> > enumerated all the observation types, so that different systems share
> > the same identity of an observation type : e.g. observation type sex,
> > observation type age , observation type cholesterol level  etc...
> > sex=1, age=2, chol_level=2832 , serum_sodium=1234 ? Or what are other
> > ways to achieve inteorpeable data ?
> A severely limited answer would be SNOMED-CT/LOINC and
> friends. The conceptually better answer is http://www.openehr.org
> as was mentioned before here...
> 
> Karsten
> -- 
I looked at the openehr site, and it is in construction. It doesn't have
any non-generalizing , non-push-my cart documents. I looked at snomed-ct
but there's something about a license with this one, but it looks
promising: ? has anyone obtained the snomed-ct software to look at it?

I tried my icd-9 based search thing put on the past history panel ( in
testarea/sjtan/handler_devel/client - hey , everyone else is pushing
their carts, why not join in?)  , which Ian pointed out is a little slow
(average 1-3 seconds , up to 10 seconds if too many words are included) 
( but it sort of works !! ) : it finds conditions ok, and was able to
find most symptoms,  things like :

headache
temporal arteritis

but not ,  temporal headache

shortness of breath
asthma 


but not , wheeze , wheezy or rhonchi 


I know some of the exceptional data in the icd9 text was excluded , as
they were in tables and needed to be parsed e.g. the hypertension table,
and the diabetes table  ( um, only minor clinical conditions missing, 
uh?) 

It might be possible to convert the icd-9 into a coding-labelled  object
tree of Observation objects, but I'm not sure if it's worth the effort.

Has anyone got a publically available ( with data installed) terminology
database (object or otherwise)? The gehr hasn't got data in it, but 

Snowmed is supposed to have 3 * 10^5  terms, with over 1 * 10^6 words
etc.. I gather it's free for NHS related work , but royalties apply for 
commercial developers.
  

How about writing a tool to parse the icd9 file into relations , anyone
want me to have a go ? (don't care?) . 
Q. what is the optimal  structure of the table(s), as the icd9 text is
like a wide, short tree. 

because of Snomed-CT and the like, this effort is really prototypical ,
because it's just re-invention on a crude, backyard level.