Screening and diagnostic tests
What a test really does in a population: how many positives are true, how many are false alarms, how many cases are missed. Prevalence changes everything.
Confirm positives with a second test?
More options
A positive result is right
27%
Of the time (positive predictive value). 3.7 people test positive for every true case found.
A negative result is right
100%
Of the time (negative predictive value).
True cases found
180
False alarms
490
Cases missed
20
Cost per true case found
$278
Everyone screened, at a glance
1 □ ≈ 1%
How prevalence changes the meaning of a positive
For your protocol
Screening 10,000 people with a test of 90% sensitivity and 95% specificity at a prevalence of 2.0% would find about 180 true cases, with 490 false positives and 20 missed cases. The positive predictive value is 27% and the negative predictive value 100%.
How this simulation works
Bayes' theorem with natural frequencies. Of the people screened, a share (the prevalence) have the condition; the test finds a share of them (sensitivity) and correctly clears a share of the rest (specificity). With a confirmatory test, only people positive on both count as positive, assuming the two tests err independently, which raises specificity and lowers sensitivity. Real tests may perform differently in different settings and populations.