Assessment of agreement of a quantitative variable: a new graphical approach

In clinical or epidemiologic research, the measurement of variables always implies some degree of error. Because it is impossible to control the various sources of variation, the assessment of the reliability of a measurement is essential. Otherwise, concordance analysis must take into account the "clinical" interpretation of the measurement under study, because its practical usefulness is of central importance. In this article, we propose a new approach to assess the reliability of a quantitative measurement. We use a graphical approach familiar to statisticians and data analysts of the biomedical area, associating to it the useful feature of interpretation based on the proportion of concordant cases. We believe that the proposed graphical approach can serve as a complement, or as a alternative, to the Altman-Bland method for agreement analysis. It allows a simple interpretation of agreement that takes into account the "clinical" importance of the differences between observers or methods. In addition, it allows the analysis of reliability or agreement, by means of survival analysis techniques.