On determining the most appropriate test cut-off value: the case of tests with continuous results