hi george; ok if i call you Fermat ;-)
> Subject: Re: Screening for lung cancer
>
>
> In article <000401bdecae$4c0fa400$f75395c1@default>, owen dempsey
> <[log in to unmask]> writes
> >
> >the chance of getting a positive result is 15.7% (which will be wrong
> >94.5% of the time.....)
George wrote:
> Does this mean that a random CXR on a smoker which is positive for Ca
> Lung will be wrong most of the time (I can't work the numbers)?
yup; if there is a low prevalence of lung cancer in the population
you're doing the random xr on;
there is something like a 1 in 7 chance of inflicting a positive result
on the patient, which will be wrong more than 9 times out of ten
even if you increase the specificity of the cxr to 90% there would still
be a 12% chance of a positive result; which would be wrong 92% of the
time
looked at another way, if you thought that the test would be acceptable
if there was only going to be a 1 in 20 chance of a positive result
(which would still be wrong 84% of the time) then the specificity of the
cxr (its ability to correctly identify the negatives) would have to be
95.2%; now i'm guessing here but i doubt if the cxr is that good.
> >'best hand held calculator in the north'
>I thought that was Ahmad's line, except, of course, he's limited to
base
>5 and he's in the south. ;-)
i don't think the adverb limited applies in that particular instance :-)
regards
owen
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|