In observational studies of, eg, hospital admission outcomes,
I have often seen (and occasionally done) the following: after
selecting some cohort of patients, in the event that some patients
have multiple admissions, only one random admission per patient
is retained. This seems reasonable when there are only a few
patients with multiple admissions (and the outcome is likely to
be correlated across admisssions), but I can't dig up anything
in the literature arguing for or against this technique, or even
simply suggesting or evaluating it. Does anyone out there know
of anything?
thanks,
Jeph
--
Jeph Herrin
[log in to unmask]
|