Intraclass Correlation Coefficient and Concordance　Correlation Coefficient for Evaluating Agreement among Observers

Abstract

In clinical studies, assessing agreement among observers plays an important role in evaluating whether different machines, raters, or methods give similar results of the measured responses. The intraclass correlation coefficient (ICC) and the concordance correlation coefficient (CCC) are two popular scaled indices for assessing the closeness among observers for quantitative measurements. In this talk, we will introduce the concept of agreement for ICC and CCC, and present the comparison of these two indices under a very general model. In addition, we will present our proposed new CCC-type index for data with repeated measurements. This index is used when the observer effect is treated as random (i.e. observers are randomly selected from the observer population). We use the method of moments approach for estimation and a modified bootstrap approach for inference. Vertebral body data and image data are used to illustrate our proposed methodology. This is a joint work with Dr. Huiman Barnhart.