Reliability of diagnostic tests.

Abstract

The reliability of a diagnostic test depends on the accuracy and reproducibility of the test results. The accuracy is defined by comparing the test results with a final true diagnosis. The predictive values are here the most important clinical measures. Since it may be impossible to establish a final true diagnosis the reliability must in some cases be measured by a determination of reproducibility. The reproducibility is measured by comparing results of repeated examinations of the same patient. The reproducibility is measured by the use of the kappa coefficient which adjusts the observed agreement for expected chance agreement. A study of reliability of a diagnostic test should fulfill the same methodological requirements as other clinical studies. Both the predictive values and the kappa coefficient are supposed to depend on the prevalence and this should be noticed when results of different studies are compared. Reliability of diagnostic tests is often poor and scientific development of how to improve clinicians' diagnostic practice is much needed.