A measurement model of disease severity in absence of a gold standard

In the absence of a gold standard, we propose the use of confirmatory factor analysis for the quantification of agreement between diagnostic measurements and the true disease severity. The essential assumption is conditional independence of diagnostic measurements adjusted for the severity of the disease. However, depending on the number of measurements available, the method works even if some of them are conditionally dependent. We illustrate the method using an example related to glaucoma eye disease.