Re: [Gnumed-devel] Measurements (was: New developer searching for a task)

[Christof Meigen]

Dr Elizabeth Dodd <address@hidden> writes:
> Therefore 2.5% of 
> normal persons have 'hi' results and 2.5% of normal 
> persons have low results on a given test. That is the 
> result of the way the 'norms' are calculated.  

Yes, it's the nature of statistical tests. 

> Then add the confusion by labs with different refernece 
> ranges as just described. The doctor seeing the patient 
> has to make a decision about the result in the context of 
> the patient, (we treat patients not lab results) and must 
> be able to mark the result as satisfactory or 
> unsatisfactory regardless of lab norms. 

It's obvious that doctors should be able to add comments or judgements
"by hand" for each measurement. That doesn't raise much problems.

But such judgements can be supported by norms, which might give better
information that just low/normal/high (_how_ high, in terms of: What
percentage of healthy patients have such a high value. If the result is
like: Only 0.1% of the healthy population has such a high value,
it's a sure bet that there's something wrong)

In the only case I'm familiar with (IGF), the Lab doesn't tell you what
norm they use (it's not published), so, as a doctor you have no idea
how they come to their judgement. 

My point is: It remains undisputed that the doctor should be the one
who finally says: This result is ok/noticable/alarming etc. But the
Computer/GnuMed should give as much guidance as possible, and that
means in my opinion, offering a quite comprehensive set of of norms,
of which, partly automatic, partly by hand, the most appropriate is
chosen to say, i.e: According to this norm, only 1.6% of the
population have such a strange value of X. And when the doctor sees
that the child is a Turner-girl, which is strange by definition, it
may seem ok, but there are published norms for several parameters of
Turner-girls, so GnuMed may say: Based of the fact that this is a
Turner-girl, "I" apply a norm for Turner girls and it appears that
only 0.02% of the Turner girls have such a hight value of X, so the
situation is actually more alarming. The crucial point is knowing what
the automatic judgements are based on. Are these 30-year old norms
from somewhere far away? Then perhaps I don't have to pay too much
attention.  Or are these very actual, very specific norms? Then I
should take a close look.

Now a third party comes into play, which is the Lab, saying: This value
is (not) normal. Now, every combination is possible, like: The Lab says it's
not normal, the most actual, most appropriate norm in GnuMed's database
says it's normal and the doctor decides it's at least noticable.

Confusing, ehr? Especially when you imagine a chart with the not
normal (according to what?) measurements marked. 

        Christof