John B. wrote:

But that isn't what happens. Every study I've
seen is based on visits to
a medical facility.

Yes, and why is that a problem?

Stats from the wild will never be perfect.
If two tendencies are compared, the
sample/compile method and thus the stats will
always favor one of the two, unfairly.

It won't be perfect, but the better the method,
the closer it gets. And if one tendency is much
stronger, the stats will favor it no matter
the imperfection.

--
underground experts united
http://user.it.uu.se/~embe8573