Use of relative operating characteristic analysis in epidemiology. A method for dealing with subjective judgement.