Hi,
The "question of the day" for the FreeMedForms project is:
What is the exact pertinence of the drug drug interaction (DDI)
engine/database of FreeMedForms?
I've computed the maximum DDI available. The script evaluates the number of
available interactions per level of risk using pairs of international
non-proprietary names (INN) extracted from the ATC classification. This script
does not use a real drug database, only the INN which are unique. All
interactions come with a full description of the risks and sometimes with a
possible management of the risk (in french and in english).
Here are the results:
Contre-indication................................. 2985 2% } Are
severe (level 1)
Déconseillée...................................... 5413 5% }
Interaction de la glycoprotéine P................. 9963 9% } Are
important (level 2)
Interaction du cytochrome P450.................... 6756 6% }
A prendre en compte............................... 64298 61% }
Précaution d'utilisation.......................... 13934 13% } Are
information (level 3)
Information....................................... 618 0% }
Total DDI per INN pair............................ 103967 100%
Now think that one drug can contain X interacting molecules...
The question is now:
What is the exact pertinence of the drug database script when it needs
to link drugs component to ATC?
And this question is answered in each database, using a "pertinence index",
which represent the percentage of all drug components linked to ATC (we believe
that 75% is acceptable, 85% is a near-to-reality, 100% is only-commercial). We
may improve this "pertinence index", because users don't want to miss a
*severe* DDI. And all of these severe DDI should be tested... by hand. Testers
are welcome and can have a full access to the drugs database with DDI. We have
to test drugs database from these countries: France, Belgium, USA, Canada,
South Africa (and very partially Portugal). Only MD and pharmaceutists can do
this job. Please contact us: [email protected].
Eric, freemedforms.com
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
Archive: http://lists.debian.org/[email protected]
