[Statistical principles for the validation of diagnostic procedures. A contribution to quality assurance exemplified by Doppler sonography].
This article presents an introduction to some basic considerations for the evaluation of diagnostic tests and procedures. In order to determine the characteristics of a diagnostic procedure sensitivity and specificity are estimated from results of an evaluation study. The statement of confidence limits gives an idea of the safety of the estimates. Different scales of measurement and their respective distributions must be taken into consideration. Quite often sensitivity and specificity are dependent on various threshold values. ROC-curves can be used to study this dependence. In practice the interpretation of test results relies on the predictive value. The Bayes formula permits calculation of the predictive value depending on sensitivity, specificity and prevalence. This paper illustrates the application of statistical methods by analysing data of Doppler ultrasound studies.