[ExI] good bexarotene article

[spike]

>... On Behalf Of Brian Manning Delaney
...

Spike wrote, earlier:
>> This goes back to a discussion we had here a dozen years ago.  There
should be a database somewhere, where people can dump medical information...
and maintain patient privacy.

>...This is a great idea. Get as much data out there as we can, and see what
correlations we find -- about bexarotene, or anything else that might be
useful!  ... But about the specific idea of looking for Alzheimer's rates in
the many people who've taken the drug for skin cancer, it occurred to me
after I sent my post that the researchers of the new paper have probably
already thought of that... Patient records would be spread out across
dozens, perhaps hundreds of physicians and clinics, unfortunately... (This
is why the absence of the database you hope one could pull together really
hurts....) Doable, just a huge amount of work. So much so that Landreth may
have thought of the idea annd rejected it as impractical.  -Brian
_______________________________________________

Brian, we would need to invent a system, or look for one in existence,
whereby doctors can estimate the degree of progressive diseases.  We already
have cancer described that way, four stages, so we could do the same with
Alzheimer's or anything.  It doesn't matter if different doctors evaluate
differently; statistics can handle that problem.  If we have any resident
expertise in this or the locals have friends who are hip to medical
statistics, they can surely help.  Dammit Jim, I am just a country rocket
scientist, not a doctor.

There may be a way to reduce diseases to a number or series of numbers.
This will be needed for sorting purposes.  For instance, all cancer could be
4.  Then perhaps colon could be 12, and hyposquamal scatorsic cell could be
16 with frumbic
slambonian harcomipsis syndrome is 5, so a patient with that type of disease
would be 4.12.16.5.3 once they hit stage 3.

The medications would also be numbered in some systematic fashion.  We would
encode somehow the age of the patient, their race, their bad habits, their
good habits, whatever we can put in a matrix.  Then we can apply our
mathematical tools that we know so very well, sparse matrix techniques used
in satellite control systems for instance.  It wouldn't even hurt much if
the patients lie: statistical methods can deal with that.

Now we would need a central website I suppose and some means of getting
people to drop data into it.  We could set it up in some sort of spreadsheet
perhaps, and let our volunteer background process team grind away, looking
for correlations or signals in the pile of data.

Have we any medics among us who can suggest a data structure?  Or point to
one that already exists?

Once a system starts somewhere, perhaps it would grow, or start competing
systems, kinda like dust bunnies.

spike