The predictive power of diagnostic tests and the effect of prevalence of illness

There is a great interest in the potential usefulness of biomedical tests in psychiatry, but basic statistical principles required for critical evaluation of their value remain poorly integrated into clinical thinking. We developed a useful scheme for organizing clinical test data so as to permit simple calculations of the sensitivity, specificity, and predictive power of medical tests. Some loss of predictive power is inherent in moving from artificial populations used for test development into more typical clinical settings, where prevalence is usually substantially lower.