> From: philippe Ameline [mailto:[EMAIL PROTECTED]]
> I don't understand your point ; In my own project, I use very fine
> granularity in order to enable decision support ; and it seems to me it is
> the only reason why it is worth doing it.
I agree entirely with this. There is a distinction to drawn between systems
designed more than 6 years ago, when most designers did not understand the
importance of context and the problems of multi-purpose coding schemes, and
cutting edge systems of the past 6 years, wher this is all taken for
granted. Perhaps I am still fighting some of the wars of the mid 1990s.
> As soon as you *classify* the information, you loose its semantic meaning.
Exactly. And the old error was to assume that hierarchical classifications
(which can be very fine at the leaf level) are useful (by that I mean easy
to use safely) for more than the one job they were designed to do.
> > I am not quite sure where the boundary lies between systems such as
SNOMED CT, ontologies and terminology
> > servers such as Open Galen.
>
> I can try to point out the main differences :
> - A system with no semantic network can only be used for epidemiology
(terms have *no meanings*)
> - A semantic network can be *action oriented* (Odyssee for example) or
*langage analysis oriented* (Galen)
>
> People are spending lots of time elaborating multi-axis coding systems, or
> multi-classified terminologies ; I must confess I don't understand the use
> of it.
This is a point which I would like to understand a lot better than I do.
> Actually, epidemiologic studies can be made for two reasons :
> - for *individual improvement*. And the main concept is classification
> principle, SOAP+Problems for example. The classification itself - ICPC in
> this case - only has to be acurate for the choosen principle.
> - for *global studies*. And usually low granularity concepts are
sufficient,
> and a case mix approach is good enough.
ICPC is a nice simple 2-axis classification. It does not pretend to be
clinically useful. The problem lies with the BIG schemes such as Read Codes,
SNOMED, ICD9/10 (CM) which have tens and hundreds of thousands of terms.
>
> Your feedback is precious to me.
>
> Regards
> Philippe AMELINE
> Odyssee project
> www.nautilus-info.com
>
