Kappa coefficients in epidemiology: an appraisal of a reappraisal.

While we agree with the general message of the Thompson and Walter paper on the use and value of kappa, the mathematical bases of these conclusions, their equations, tables, and figures are based on mathematical assumptions so limiting that their application to real data is questionable. As a result, several of their specific conclusions are misleading. Some of the same results are here obtained without the limiting restrictions. From these results, the value of kappa as a measure of reliability, validity and (possibly) of general 2 x 2 association is demonstrated.