Re: fuzzy and medicine

RICHARD OLSON (olslogc@primenet.com)
Wed, 26 Jun 1996 11:58:06 +0200


On 18 Jun 1996 14:58:39 GMT, you wrote:

>
>The List:
>I'm working with the hospital administration theory (born in Yale U.) that
>the case-mix of all the illness for any patient could be related to one
>of 600 Diagnostic Related Group Coefficients (DRG), referred to a specific
>coefficient and days of stay.
>The coefficient is used to calculate the relative cost of helth care and for
>measuring the performance of the hospital administration.

>The base points of my theory is:
>1) The set of patients with some set of illness and coming from some social
> set is a "fuzzy set".
>2) The set of point 1) depends of the age and sex (crispy sets).
>3) All the illness are "fuzzy sets"
>4) All the social sets are "fuzzy sets".
>5) The DRG are uncertainly coefficients, not crispy sets.(Is the result of
> statistical probability, the principal questionable method for the evident
> sum of fuzzy factors).
>6) The DRG could be false sets for any performance ponderation.

>To ponderate the relative cost and health care performance I suggest
>to research the application of hospital software with fuzzy logic engine
>and expert system for data base.
>Anyhow, the input of any hospital software is the more noisy and distorted
>data treatment difficulty, because the fuzzy nature of the morbidity,
>co-morbidity and treatment definitions, at sight of the MD.

>I'll be grateful for any comment.
>

> Ernesto Landera
> rnt@mpdisi.gov.ar
__________________________________________________________________________________________________________________________________

Ernesto,

It certainly sounds feasible to develop a fuzzy logic system which
could give you an accurate DRG.

1. I believe you should use an fuzzy SQL data base.

2. Crisp sets and fuzzy sets can be used together very well in an
SQL fuzzy logic data base.

3. To handle the noisy data, a knowledge mining system (neural
networks, Wang-Mendell) could be used to set up the initial fuzzy

fuzzy sets and rules governing the system. At the very least
neural networks transformation methods could help in cleaning up
your noisy data.

4. Fuzzy system methodology has worked on far more complicated
problems than you seem to be dealing with.

Regards,

Richard Olson