gms | German Medical Science

A key phase in the evaluation of a diagnostic test is determining its diagnostic accuracy, that is, the ability of a test to distinguish between patients with and without the disease or health condition of interest (i.e. target condition). In diagnostic accuracy studies, the presence or absence of the target condition is ideally determined by a gold standard which provides an error-free classification of the target condition status. Test accuracy measures, such as test sensitivity, specificity, likelihood ratios, predictive values, or diagnostic odds ratio, express how well the results of the test under evaluation agree with the outcome of the gold standard. However, for most if not all diseases, a gold standard that is without error is simply not existing. In these circumstances, researchers use the best available practicable method to determine the presence or absence of the target condition, commonly called a reference standard instead of a gold standard. When studies ignore the imperfection of their reference standard, test accuracy estimators tend to be strongly biased in a direction that is often difficult to anticipate. Suggested statistical solutions that aim to alleviate the bias in estimators of test accuracy have had limited successes; they remain rarely applied.

In this talk I will argue that the lack of a gold standard could be seen as a measurement error problem. I will discuss how generic measurement error correction methods could be used to alleviate the problems in diagnostic accuracy studies in the absence of a gold standard.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.