I think there may be some misconceptions about the difficulty of making a
terminology readily available for routine work. Some comments interspersed
below.
Date: Sat, 11 Feb 2006 09:27:29 +1100
From: Tim Churches <[EMAIL PROTECTED]>
Ian Cheong wrote:
> At 7:44 am +1100 11/2/06, David More wrote:
>> Ian,
>>
>> I think it is important to remember the lack of terminology was a key
>> reason for the failure of a number of the HealthConnect trials.
>>
>> Spending on terminology capability and development may not be a bad
>> investment at all in my view - it is required if any form of real
>> inter-operation between systems is to be achieved. Communication 'by
>> blob' helps - communication of understanding and context is way better.
>>
>> Cheers
>>
>> David
>
> Yes, but terminology is mainly for machine processing.
>
> Detailed comprehensive terminology costs a bomb and leads to enormous
> downstream costs.
>
> A limited terminology with classification is probably all one needs to
> do most decision support - something closer to 2000 terms, according to
> various experts around the traps.
>
> It is likely that yet another tiny little bureaucratic decision will
> point us in a less than optimal direction for decades.
A few observations:
0) Yes, encoding information using a clinical terminology is indeed
mainly for machine processing. But that's the point - it better enables
the machines to do the information processing drudge work, to allow us
humans to concentrate on more interesting things. Health informatics is
not only about speeding up human-to-human communication.
Automatic and semi-automatic methods of data capture at the point of care
should add little if any time cost if the systems are well designed.
While clinicians need data retrieval of patient records for their everyday
work and effective interoperability is fundamental to achievng that we also
have an eye on the prospect of radically improving the analytics of health
information systems. Effective data capture will enable us to deliver much
superior data analytics so that people can more effectively study what is
systemic about the nature of populations of health within their bailiwick.
1) The US Dept of Health and Human Services paid teh College of American
Pathologists (CAP) a once-off US $35m fee for a perpetual license for
all of SNOMED CT for all of US (available to everyone, public and
private sectors) with updates for 5 years and an option to renew for
updates after that.
2) On a population prorata basis that equates to about AUD$3m for a
similar five years of updates for all sectors of all of Australia, or
about $600k per annum. That's probably less than the annual fancy
sandwich meeting catering bill for the DoHA...
3) Just because SNOMED CT has several hundred thousands concepts in it
doesn't mean that you need to use them all. You can easily pick subsets
of SNOMED CT for particular purposes. But if you need more deatil, with
SNOMED CT, it is already there (in most cases - there are still some
gaps in its detailed coverage of concepts, but these can be filled in
due course, especially now that CAP is more open to shared governance
and ownership of SNOMED CT).
In the prototype systems we are trialling here we can search for terms across
all of SNOMED in about 1 second. If we restrict the search space to a range
suited to a particular speciality we can add in extra intelligence to use up
that 1 second.
4) The challenge is to develop smarter technologies for automatically
encoding medical concepts expressed or chosen via structured pick lists
or look-ups, in free text notes, and via natural language speech into
SNOMED CT codes. It's doable, and Jon Patrick's group at Sydney Uni has
already made a start.
We expect to demonstrate in our public showcase in 2 weeks a prototype of
real-time data entry onto an exact replica of an existing hospital form data
entry of categorical information (tick boxes), text notes and drawings,
including conversion to SNOMED codes. This is the product of a few students
over the summer.
There is a big opportunity for the development of
home-grown technologies to do this which don't cost a bomb, and which
can be incorporated in next-generation clinical information systems or
retro-fitted to existing ones.
here, here!! This is what we are working towards.
Tim C
_______________________________________________
Gpcg_talk mailing list
[email protected]
http://ozdocit.org/cgi-bin/mailman/listinfo/gpcg_talk
Jon
---------------------------------------------------------
Prof. Jon Patrick BH +61-2-9351 3524
Chair of Language Technology FX +61-2-9351 3838
School of Information Technologies
University of Sydney
Sydney, 2006
NSW
Australia
--------------
Scamseek Project http://www.cs.usyd.edu.au/~lkmrl/scamseek.htm
Personal WEB Page: http://www.it.usyd.edu.au/~jonpat
Language Technology Research Group http://www.cs.usyd.edu.au/~rcdmnl/
Information Systems WEB PAge: http://www.infosys.usyd.edu.au
School WEB Page: http://www.cs.usyd.edu.au
----------------------------------------------------------
_______________________________________________
Gpcg_talk mailing list
[email protected]
http://ozdocit.org/cgi-bin/mailman/listinfo/gpcg_talk
