Studies of diagnostic test accuracy: Partial verification bias and test result-based sampling.

BACKGROUND: Partial verification bias occurs in studies assessing the accuracy of existing diagnostic tests when a positive index test makes application of the reference standard more likely.

METHODS: This article first describes alternative sampling frames for a diagnostic test accuracy study then discusses partial verification bias.

RESULTS: One approach to avoiding partial verification bias is to apply the reference standard to all individuals who are positive on the index test and a random sample of those who are negative. Estimating sensitivity and specificity then requires adjusting for the sampling fraction in the test-negative group.

CONCLUSION: If a finding was already used to determine who should get the definitive test, one can't look at those who got the definitive test to determine the usefulness of the finding.