Abstract:
|
The current emphasis on including correlation when comparing diagnostic test performance is quite important, however, there are cases in which correlation effects may be negligible with respect to inference. This work examines the impact of correlation between continuous biomarkers on confidence intervals (CIs) comparing the optimal performance of two diagnostic tests with multiple outcomes. We define the optimal point using Bayes Cost, a metric that sums the weighted misclassifications within a diagnostic test using a cost/benefit structure. Current methods for CIs comparing continuous tests using Bayes Cost relies on delta and generalized methods with an assumption of independence between the biomarkers. Through simulation, we quantify the impact of correlation on standard errors comparing two tests and evaluate the resulting errors with respect to CI coverage and width under varying diagnostic test accuracy, sample size, cost/benefit structures and correlation levels. We provide guidance on the testing conditions that require formulas which include effects of correlation for CI calculations to compare optimal performance between two diagnostic tests using Bayes Cost.
|