You got me confused so i decided to look this up. [ in Medical Decision
Making by Harold Sox Jr]. PPV is not the same as "post test probability"
PoTP = All test positives/ Total no. of patients tested.
PPV = True positives/ All testing positive.
* PoTP is also = PreTP x Likelihood ratio of a positive test(or LR +) where
LR+ is the number of true positives / false positives.
The pretest probability or PreTP is also the prevalence of the disease which
is say for UTI, usually estabilished from previous data to be 3-5 % in the
setting of FWLS. When you have an example like I mentioned, of a 4 month
old female infant with a T = 104 and no source with a WBC of 20K the
likelihood of whatever screening test you choose being positive for UTI is
higher than the other male circumcised infant with FWLS. So, if your subset
of patients that undergoe enhanced UA happen to selectively more like the
first case, the true positivity (which is also the sensitivity BTW) of the
enhanced UA is higher and the no. of false positives are going to be lower.
This means the LR+ is greater and using equation * from above PoTP is
This referral bias of the sicker patients is also called "spectrum bias"
from what I read, which is what jay Fischer referred to. He may know more
about it. Coming back to his point, the predictive value of the various
screening tests for UTI will hence vary depending on the prevalence of UTI
in the individual subsets subject to this bias.
Hope I am not sounding confusing. Overall, i thought the article was
enjoyable and good and may be there is no referral bias. However, this would
have been more explicit if as JF said they had either done all tests on
everybody or given more details on the individual patients in each group.
For more information, send mail to [log in to unmask] with the message: info PED-EM-L
The URL for the PED-EM-L Web Page is: