On Wed, 28 Mar 2001 22:52:53 philippe Ameline wrote:
...
>FILS GUIDES :
>Fils guides are probably the major difference between GEHR and Odyssee.
>Usually, doctors have first to choose a *model tree* (made by an *expert*)
>and instanciate what he wants to describe in it (when I say a *model tree*,
>it can be a form as well, a form being a narrow tree).
Hi Philippe,
Very nice description. The most difficult part is the most facinating part. :-) I am
most interested in the FILS guides since it embodies some probabilistic and (soft)
constraint information. This is in sharp contrast with both GEHR and OIO where we use
hard constraints. Of couse, part of this is because your tree is much bigger than our
trees :-). You still have hard constraints which are the terms and links that make up
your tree.
In particular, your statement that a form is a special case of very narrow tree
(which I also read on your web-site) nicely ties in with OIO's forms-based
representation (that seeks to build a tree from forms).
More below.
>There are two severe problems with that : first a *model tree* made by an
>expert carry expert's habits as well as expert's knowledge (usually
>knowledge is close from the leaves), then as you can't mix the trees, you
>have to choose that one or that one, but patient can have a proctological
>problem and a headache !
Therefore, you either need to allow flexible traversal of a giant tree or allow users
to use more than one tree. Correct?
>It is not easy to describe what Fils guides are, but I will try :
>Imagine some expert's trees, then take every branchs and take it apart : you
>get a set of *crossings*, each of them with some proposals (say one being
>size/aspect/location and another one moderate/mean/severe and so on). Now,
>you ask the experts "take each crossing, and tell me what would be the
>typicall (and more general) path
You are encoding probabilistic information there. What if different prior-conditions
(i.e. higher up branch points) lead to different posterior probability? Do you
recalculate at every step?
Even with a simple scheme where you only encode the most likely branch, there could be
so many possible combinations of prior-conditions that you will have a very large
encoding task leading to very difficult traversal. How do you address this?
...
>That way you can mix experts knowledge, and have it come only when it is
>appropriate. You can make a set of Fils guides to a colleague, let him try
>it and keep it or put it apart.
So a Fils guide is a preferred path through a tree?
This is basically what a GEHR archetype and an OIO form is! The difference is that
Odyssee lets the user change the "form" (or archetype) during data entry.
...
>you
>cannot guess which next Fil guide will be elected by the system (since it
>depends on the actual user's path).
How well does it work in practice? It seems computationally intensive to use the
user's current path to pick the best fitting Fils guide to apply. How many Fils guides
have you tried in a given set?
Do you give users' a choice to turn off the Fils guide?
>Well, if you are near Paris, I can show.
Is this part of the downloadable package?
In summary, I am very glad that you published your very nice description of Odyssee in
your message! Please let me know if I am wrong in my understanding of your approach. I
look forward to reading your response to my new questions and I look forward to
reading Thomas' (from GEHR) comments for you!
Best regards,
Andrew
---
Andrew P. Ho, M.D.
OIO: Open Infrastructure for Outcomes
www.TxOutcome.Org
Assistant Clinical Professor
Department of Psychiatry, Harbor-UCLA Medical Center
University of California, Los Angeles
Join 18 million Eudora users by signing up for a free Eudora Web-Mail account at
http://www.eudoramail.com
