Comparing accuracies of two screening tests in a two-phase study for dementia

摘要: Summary. A two-phase design has been widely used in epidemiological studies of dementia. The first phase assesses a large sample with screening tests. second, based on the test results and possibly other observed patient's factors, selects subset study for more definitive disease verification assessment. In comparing accuracies two tests dementia, inferences are commonly made from verified cases. omission non-verified cases can seriously bias comparison results. To correct this bias, we derive maximum likelihood (ML) estimators their corresponding correlation. p-values confidence intervals computed using asymptotic normality ML estimators. Our method is to compare We found that, although sensitivities new standard detecting diseased subject not different, performs better non-diseased subject.