In message <001701bf691f$271a4940$32c94382@heidithornhil>, Atle Klovning
<[log in to unmask]> writes
>
>There was a debate in the Jounal of Clinical Epidemiology on whether the
>definition of the negative predictive value should be changed.
>
>Put simply, using probabilities instead of odds:
>the post test prob of a positive test = positive predictive value
>but the post test prob of a negative test = (1-negative predicitve value)
>One author thought this was inconsistent and meant that
>but the post test prob of a negative test should be redefined into negative
>predicitve value
>
Yes, this is confusing, like all double negatives. A NPV of 0.9 means a
post-test probability of 0.1 - the higher the MPV, the lower the post-
test probability.
It's not a problem if papers also give sensitivity and specificity, but
it can be very irritating if they only cite PPV and NPV, since these
vary with the pre-test probability. It's a bit better if LR+/LR- are
also given but I have to confess that I find the maths to get back to
sensitivity and specificity a pain and definitely not carried around in
my head! If I want to apply the findings to my patients I need to start
with their pre-test probability, which might be different from the
population in the paper. You could argue that LR with pretest
probability and Fagan nomogram is adequate, but I like to know the false
positive/false negative rate as well as post test probability.
Toby
--
Toby Lipman
General practitioner, Newcastle upon Tyne
Northern and Yorkshire research training fellow
Tel 0191-2811060 (home), 0191-2437000 (surgery)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|