We have a long way
to go, we're working
on an infrastructure to
be able to generate pre-
dictive algorithms and
do a better job of iden-
tifying these subpopu-
lations. Overall, we
could be looking at
hundreds or even thou-
sands of variables, but
it's likely that only a
few dozen or less have
major contributions to
the outcomes we're measuring. We don't have to nail down every one,
just the ones that really matter.
Better guidance
Our goal is to continue collecting data about demographics, about
patients' goals and fears and emotional states as well as other patient
and treatment factors that matter to outcomes that measure value for
the entire care process. If we keep collecting data and use the appro-
priate analytical tools, our algorithms will get better and better, even if
we realize they'll never be perfect.
You can do one thing to one subpopulation and have it benefit them,
but that if you do the exact same thing to another subpopulation, it
can contribute to harm. And doing the same thing to yet another sub-
population might result in neither benefit nor harm, so it would sim-
ply be wasteful.
A P R I L 2 0 1 8 • O U T PA T I E N TS U R G E R Y. N E T • 6 5
The same mesh and the same technique
can produce dramatically different
outcomes in different patients.