When pre-test probability can be deceptive