Measuring Association Between Screen Test Results Using Kappa K Statistic
Publication Date: 16/06/2020
Author(s): Oyeka I.C.A, Nwankwo Chike H., Onyiaoha C.A. Adaora, Efobi Chilota C.N..
Volume/Issue: Volume 3 , Issue 3 (2020)
Abstract:
We have in this paper developed and presented a more generalized and formatted method of estimating kappa k values often used for analyzing data obtained in comparative clinical trials and diagnostic screening tests of subjects for some condition such as disease in a population. The proposed method is generalized in the sense that it can be used to analyze data obtained in comparative, prospective studies and comparative retrospective studies with either two or more possible response classifications, categories or outcomes. A rule of thumb for determining the level or degree of concordance or agreement between test results as suggested or indicated by the estimated kappa k statistic is presented and used in the illustrative example. Test statistics are also presented for testing the significance of any desired hypothesized values of kappa k. Sensitivity and specificity dependent measure of association and odds ratios which are equivalently used as measures of association when there are only two possible response options or outcomes in comparative clinical trials and diagnostic screening tests are presented and discussed. These results are compared with the results obtained with kappa k statistics when there are only two response options or outcomes which are equally applicable in comparative studies with only two possible responses or classes. The results are shown to be quite similar leading to essentially the same conclusions.
