I am working with the national inpatient sample database. It is a complex survey design with one stage power sampling unit - hospitals - among stratas, with national estimates weights on hospitals.
I want to compare the prevalence of a condition in a case cohort that is 0.5% of that sample carrying another baseline condition, essentially giving me a control cohort consisting of millions observations vs a few thousands from the case cohort. What are the implications of this?
What is the difference if i subset the sample first to the condition in question and then subdivide in groups with and without my baseline condition? How would that influence my odd ratio calculation?