连续变量诊断试验数据的ROC分析

Analysis of receiver operating characteristic(ROC) for continuous measurements in diagnostic test

目的介绍一种连续诊断变量的ROC回归模型,以及在独立和相关结构下的参数估计方法,给出参数误差估计的Bootstrap方法。方法应用SAS软件中的GENMOD过程和SAS语言编写的程序实现上述过程,并通过实例说明其应用效果。结果利用ROC曲线方程,可以扣除协变量对诊断试验结果评价的影响,并能够计算出在不同协变量取值下的ROC曲线下面积,提供更为丰富和可靠的信息。结论文中给出的ROC回归模型可以有效地用于连续变量诊断试验数据的ROC分析。

Objective We introduce a ROC regressive model for continuous measurements in diagnostic tests and the parameters estimate methods under the data structure of independence and correlation. The bootstrap method used for estimate errors is also presented. Methods Using SAS language and the GENMOD procedure in SAS software to achieve those goals and illustrate the effect via real cases. Results The impact of covariates on diagnostic test assessment can be deducted adopting the methods of ROC curve equation. Conclusion The ROC regression model can be effectively applied to the ROC analysis of continuous measurements in diagnostic tests.