Re: [Gnumed-devel] clin_medication

[Ian Haywood]

On Thu, Oct 14, 2004 at 09:54:00AM +0200, Karsten Hilbert wrote:
> storing "enough" drug information to identify a drug across
> various sources is the problem ?
Yes.

>  ufk_drug text
>  drug_db text
> 
> , eg. an unchecked foreign key into the drug database drug_db.
> This is an oqapue value as far as PostgreSQL is concerned.
> Applications need to know what to do about it. Nevertheless it
> is of great utility.
Certainly.

> In Germany we have the PZN (PharmaZentralNummer - central drug
> number) which is supposed to uniquely identify a drug over
> time. Writing that to a data soup field helps a lot.
You're very lucky!

> *Additionally* one must store sufficient data (name,
> packaging) to be able to identify the drug regardless.i
There's the rub.

I see 2 options: formatted free string, such as
"ticarcillin/clavulanic acid 3g/100mg", or a second table to
store the (ID, drug, amount, unit) tuple, with multiple entries 
for compounds. I'm not sure which is best.

dosage could be represented by two intervals: length of dose cycle, and
length of treatment, as Jim has said. The problem is "asymmetric"
dosing, such as 40mg frusemide in the morning, 20mg at midday, so
we need an array of up to four floats to describe the dosing pattern
throughout the interval (default 24 hours),
so "2-1-0-0" for above, this is the "German" idiom (or is it the
"Hilbert" idiom?? ;-) In Australia
the client would need to translate between this and the local
Latin-based idiom, viz. "Frusemide 20mg, II mane I midi"

Certainly we need a free string as a "backup", triphasic Pills are
an example where trying to represent the constituents as they vary in
the 28-day cycle in structured data is too much complexity for too 
little gain IMHO, 

Ian

pgpKK1tF3XInR.pgp
Description: PGP signature