gnumed-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Gnumed-devel] Introduction / growth standards


From: Christof Meigen
Subject: [Gnumed-devel] Introduction / growth standards
Date: 10 Jun 2002 13:24:53 +0200
User-agent: Gnus/5.0808 (Gnus v5.8.8) Emacs/21.1

Hi,

i'm a computer science student and working for more than 3 years now
in the medical area.
I've been programming a lot in C and Perl and some dead or already
funny-smelling languages, but also in Java and C++.

I did some work at the children's hospital in Leipzig, where we have
a project called "CrescNet" where pediatrians send in length
and weight data of their children an get an automatic response
about their normality. At the moment 120'000 children with
more than 250'000 length and weight measurements are in the database.

Furthermore i'm doing work together with Micheal Hermanussen, 
concentrating on Synthetic normcurves (= constructing normcurves
from little information) and growth patterns, giving an answer
to the question: What growth development is normal? that
goes far beyond the classical growth velocity centiles.

I would like to see the entire question of growth standards 
cooperated into gnumed, and would be happy to contribute in that
area. 

In Leipzig, and in a proprietary Program I co-wrote and finally
in an website I'm working on (willi-will-wachsen.de, where
you can already get nice PDF-graphics for growth development
of children according to different (well, 2 right now) standards),
we do it the following way:

- We have different normcurves, containing:
  - mean and standard deviation of height for each sex and
    several age points, assuming a normal distribution (which
    is not quite true during puberty)
  - L, M, and S according to the LMS-method by Tim Cole
    (Box-Cox transformation for skew normal distributions)    
    for weight and BMI, again at several age points for each sex

- Furthermore we have common growth patterns, calculated via
    cluster analysis from about 800 children with complete
    longitudinal data (the well known children from the
    Prader, Sempe, Berkeley etc. studies), published earlier
    by Micheal Hermanussen and now refined by me and probably
    presented at a confrerence on auxology in november in
    Gl├╝cksburg, Germany

It is fairly easy to calculate L, M and S from centile data,
and i've done that already, becauso many -- esp. older --
studies don't contain them: Centile data alone is not too
useful, since you can only say: "well, this is above 
the 97. centile" or so, but it might be more interesting 
wether it is 2 or 2.5 SDS, especially if you look at the
development.

It shouln't be hard to get centile data, since there are
numerous collections avaliable, for example in 
proprietary but free (like free beer this time :-)
programs like GrowthAnalyzer or KSN, and I do 
also have some important standards here. Furthermore
it should be possible for users to enter their own standards
and while gnumed spreads we will soon have all
important standards entered into it, right?

Gee, this is getting long already, ok, I think the questions
are:

1)How are such standards treated in gnumed (put in a database / 
  config file), what other concepts like standards for parameter
  xy (head circumference, other medical stuff i have no idea
  of (he, what standards do you actually use for things like
  blood pressure. Does a physician just say: Well, that's high
  or are the kind of centile tables)) should be trated within the
  same framework

2)How should the checks be done
3)How should the results be presented to the user

I feel very at home in points 2 and 3, since I wrote
a C-program for that for the Leipzig childrens hospital,
a Delphi-program for Micheal Hermanussen and a perl-program
for the above-mentioned website. Python would fill a gap here :-)

What are your opinions esp. on point 1) and do you think that
whole subject is of any importance for gnumed? And is anybody
already working on that?

Kind regards,
        Christof



reply via email to

[Prev in Thread] Current Thread [Next in Thread]