Let's say we have a subgroup S of the population (e.g. inmates), and we do a prevalence study for a certain condition D (e.g. major depression) in S for which we try to diagnose D somehow.
But we don't have a control group consisting of the general population in our study.
Instead, we use some already existing, large prevalence study for the general population regarding D.
We combine them and find out that prevalence is higher in subgroup S. So we postulate a correlation between belonging to the subgroup and being at risk for the disease.
I've seen this approach in many studies.
OK - if we have an infectious disease, with low variability over time, like HIV, it seems easily comparable, because those are easily done via testing after random sampling, a very standardized diagnostic method.
But if we e.g. take a mental health condition, we use diagnostic interviews like CIDI. And critics point out that results get easily distorted by subtly different methodological differences between studies.
So when are such comparisons using another study acceptable in the context of mental health conditions?
