Comparing Rasch analyses probability estimates to sensitivity, specificity and likelihood ratios when examining the utility of medical diagnostic tests.

INTRODUCTION Medical diagnostic tests are evaluated based on measures of sensitivity (Sn), specificity (Sp), and likelihood ratios (LR). These procedures are limited in the event of a biased gold standard or missing data. Interpretations of these measures are frequently inappropriate. PURPOSE The Rasch measurement model (RMM) was examined as a method to provide evidence of diagnostic test utility in order to overcome the limitations of Sn, Sp, and LR. METHODS Patients suspected of a knee ligament tear (n = 825) were studied, by evaluating four diagnostic tests. The RMM probability estimates for each test were compared to estimates of Sn, Sp, and LR. RESULTS The RMM provided probability estimates for the diagnosis that were comparable to likelihood ratios. These probability estimates correlated with the estimates of Sn, Sp, and LR. The RMM estimates were not affected by missing data. DISCUSSION The RMM may provide an alternative means to study the utility of medical diagnostic tests to estimate the probability of disease presence/absence.