ABSTRACT

In medical diagnostic studies, the Youden index is a summary measure widely used in the evaluation of the diagnostic accuracy of a medical test. When covariates are not considered, the diagnostic accuracy of the test can be biased or misleading. By incorporating information from covariates using linear regression models, we propose generalized confidence intervals for the covariate-adjusted Youden index and its optimal cut-off point. Furthermore, under heteroscedastic regression models, we propose various confidence intervals for the covariate-adjusted Youden index and its optimal cut-off point. Extensive simulation studies are conducted to evaluate the finite sample performance of various confidence intervals for the Youden index and its optimal cut-off point in the presence of covariates. To illustrate the application of our recommended methods, we apply the methods to a dataset on postprandial blood glucose measurements.

摘要

在医学诊断研究中，约登指数是广泛用于评估医学测试诊断准确性的综合度量。当不考虑协变量时，测试的诊断准确性可能存在偏差或误导。通过使用线性回归模型合并来自协变量的信息，我们提出了协变量调整的约登指数及其最佳截止点的广义置信区间。此外，在异方差回归模型下，我们为协变量调整的约登指数及其最佳截止点提出了各种置信区间。进行了广泛的模拟研究，以评估约登指数的各种置信区间的有限样本性能及其在存在协变量的情况下的最佳截止点。为了说明我们推荐的方法的应用，