Rod Jackson wrote:
> Hi - unfortunately I think you have all got this wrong. PICO or the
> similar PECOT (Participants, Exposure, Comparison, Outcome, Time - which
> is just a more generic form of PICO(T)) describes the components of
> every epidemiological study. In the simplest case of a randomised trial
Thanks, this is interesting.
I thought of the so called "PICO question" as a synonym for the
"foreground clinical question", that is, more as a practical tool for
structuring a clinical query about an individual patient than a way to
describe epidemiological studies. The two ways are probably
complementary, because mapping the question on the characteristics of
epidemiological studies may help in retrieving relevant research.
>
> With diagnostic test accuracy studies, the Participants (P) are those
> people who you plan to test. E are those who are disease positive (ie
> reference standard positive) and C are those who are reference standard
> negative. The outcome in a diagnostic test study is the test result
> (positive or negative just as in a trial the outcome may be death - yes
> or no). As diagnostic test accuracy studies are cross-sectional T = 1.
Which comes back to my original question: an acronym like PICO or PECT
makes sense only as far as their letters are mnemonics for meaningful
concepts. Now, how do you link "E" to "positive" and "C" to "negative"?
Should then the acronym be PPNO? Or PCa(se)C(ontrol)O ?
This is going to more confusing than helpful.
>
> Many people get this wrong and think of the I and C from PICO or the E
> and C from PECOT as test positive and test negative because clinicians
> actually use diagnostic tests like prognostic markers - in someone with
> a positive test the diagnosis (or prognosis) is worse than in someone
> with a negative test. However in a diagnostic test accuracy study, you
Exactly. In the bottom line a clinician need the postest probabilities,
not sensitivity, specificity, LRs and the like (obviously, most of the
times these are needed in between to get the bottom figures).
Practically, the usefulness of a test in an individual patient is best
judged by its ability to the modify the probability of disease. Although
for an epidemiologist the relation between this and LRs may be obvious,
medical students tend to understand much better more natural figures
like pre- and postest probabilities.
I initially thought that a model with P(characteristics of patient), I
(positive test result), C (negative test result) and O (probability of
diagnosis according to a gold standard) could be useful. As you noticed
it is very similar to the one that you propose, but with reversed terms
(I don't think that you can call it "wrong" because of this, though).
However I could not find a way of pairing the "I" of PICO with "positive
test result". Since I need it for teaching, mnemonic consistency with
the acronym is important, or it would become useless.
regards,
Piersante Sestini
|