Donker D K, Hasman A, van Geijn H P
Department of Medical Informatics and Statistics, University of Limburg, Maastricht, The Netherlands.
Int J Biomed Comput. 1993 Jul;33(1):55-64. doi: 10.1016/0020-7101(93)90059-f.
The use of the kappa statistic is commonly accepted as a measure for interobserver variability. However, in some situations, the interpretation of kappa should be handled with care. In this study 21 obstetricians were asked to segment and classify 13 cardiotocographic recordings for the major fetal heart rate (FHR) patterns acceleration, baseline FHR level, deceleration and undefined segments. In two cases the kappa statistic showed a poor group agreement. These low kappa values, however, were mainly due to the high proportion of baseline segments indicated by the referees. This finding will be exemplified by a discussion of one of the cases.
kappa统计量的使用通常被认为是一种衡量观察者间变异性的方法。然而，在某些情况下，对kappa的解释应谨慎处理。在本研究中，21名产科医生被要求对13份胎心监护记录进行分段并分类，以确定主要的胎儿心率（FHR）模式，即加速、FHR基线水平、减速和不确定段。在两个案例中，kappa统计量显示出较差的组内一致性。然而，这些低kappa值主要是由于裁判员指出的基线段比例较高。这一发现将通过对其中一个案例的讨论加以说明。
