A formula for the probability of discordant classification in method comparison studies.

Method comparison studies are often reported in terms of the "limits of agreement' between the methods. Many clinical applications use the measurements simply to classify patients, for example, whether or not they are hypoglycaemic. In such circumstances it may be useful to present an estimate of the probability of discordant classification based on the bivariate Normal distribution. This paper provides as accurate and easily calculated approximation to this quantity.