On understanding reliability for diagnostic tests.

For professional practice to be responsible, any diagnostic tests used must be reliable. Therefore, the reliability of any diagnostic test needs to have been measured. The classical statistic for quantifying reliability is Kappa. Although Kappa can be promptly determined using a programmed calculator, using an algorithm to derive Kappa provides greater insight into what it is actually measuring and why. Kappa scores can be graded, with verbal descriptor applied to different grades. However, those descriptors do not necessarily reflect the degree of skill required to achieve different grades of Kappa. High levels of skill attract high Kappa scores, but Kappa scores described as fair or moderate are not necessarily flattering because they can be achieved with questionable levels of skill. Various corrections can be applied to the calculation of Kappa scores in order to raise their value, and to improve the verbal descriptors of their grade, but these may not be legitimate or necessary. Low Kappa scores do not condemn tests but they serve to raise questions about their reliability.