Objective: To develop a multivariable approach to determine the added value of tests in routine practice where some diagnostic information is already available from clinical history.
Methods: Multivariable logistic regression models were built in a stepwise fashion, considering the clinical sequence used in the rapid access ambulatory diagnosis clinic (clinical history followed by transvaginal ultrasonography and hysteroscopy). The reference standard for confirmation of diagnosis was endometrial biopsy. The diagnostic accuracy of the models was determined by the area under the receiver operating characteristic curve.
Results: The area under the receiver operating characteristic curve for the model including historical features alone was 0.78. When hysteroscopy and ultrasonography were each added to the model, it increased to 0.81 (P =.008 for improvement) and 0.82 (P =.02 for improvement), respectively.
Conclusion: The type of stepwise analysis we have developed is crucial in facilitating meaningful clinical interpretation about the value of diagnostic technology. Our finding that hysteroscopy or ultrasonography marginally but significantly increased the prediction of serious endometrial pathology above that predicted from patient history alone needs validation in an independent data set. The use of this approach is recommended when evaluating strategies for diagnosis.