Combined measures of diagnostic model performance - misleading or helpful?
View Presentation View Presentation
*Megan Neely, Duke University 


: Sensitivity and specificity have traditionally been used as the metrics for assessing the diagnostic performance of a model. However, other metrics, like area under the receiver operating curve (AUC), net reclassification index (NRI), net benefit and relative utility, have gained traction in the last decade and have been widely applied in practice. With the pool of performance measures growing, it can be daunting for investigators to determine when they should use one of the newer metrics in place of the standard sensitivity and specificity and how to interpret values that may appear conflicting. In this talk we will explore the relationship between the newer performance metrics and sensitivity and specificity. Specifically, we will discuss the expected values of the newer metrics depending on desired levels of sensitivity and specificity.