Background: In diagnostic accuracy studies, index test results are verified by comparing them with the results of the reference standard. In practice, this reference standard may be too invasive, costly or otherwise burdensome to be used in all study participants. Partial verification is known to lead to bias. Complete but differential verification, using multiple reference standards, has been suggested as an alternative.

Objective and Methods: To explore the mechanisms that can lead to bias in diagnostic accuracy studies relying on partial and differential verification methods, to discuss the possible directions of these biases, and to provide practical guidance when faced with the situation where not all index test results can be verified by the same reference standard

Results: Partial verification may occur by design, but more commonly occurs due to the practitioner's, or patient's decision not to have the reference standard. In most cases, negative index test results are less likely to be verified, leading to lower estimates of specificity and higher values for sensitivity. When different reference standards are used, for different patients, the susceptibility of a research design to bias depends on the accuracy and reliability of the reference standards, the proportion of patients verified differently, and the verification scheme. For many feasible scenarios, the net result of differential verification is an increased estimate of accuracy.

Conclusion: In designing a diagnostic accuracy study, researchers can be faced with a choice about partial or differential verification. Researchers should provide data separately for each reference standard and index test combination. Whenever possible, test results and patient characteristics should be recorded and presented for all unverified or differentially verified patients. Furthermore, implications of non-verification or differential verification for the estimates of diagnostic accuracy should be evaluated.