Diagnostic sensitivity: Newcombe paired proportion method

[Megan Moreton]

Hi,

I want to statistically compare the sensitivities of two methods of conducting a diagnostic test. The study design is paired (ie, method1 & method2 both tested on the same individual).

Currently, I estimate the sensitivity of each method with the binomial exact CI. I then test the difference between groups using McNemar's as follow:

ci proportions method1
ci proportions method2
mcc method1 method2

mcc gives me the difference between groups. But is it possible to determine the 95%CI for the difference using the Newcombe paired proportion method in Stata? (https://pubmed.ncbi.nlm.nih.gov/9839354/) I also very much welcome any advice if you think this method is not correct.

Is it also useful to present the interrater agreement (kappa-statistic) between the 2 methods?

kap method1 method2

Or would the difference in proportion suffice?

Thank you for any help.

Best wishes,
Megan