Trend analysis

Dear Listers,

I plan to run a trend analysis to explore how original wellbeing score (bad vs. good vs. very good) is associated with recovery (yes/no) after surgery. The data were collected in 2 hospitals (and randomisation was conducted separately in each hospital).

I was originally planning to use mhodds - mhodds recovery wellbeing, by(hospital)

I can't find anywhere in the documentation whether there is a minimum of clusters but 2 seems a rather small number.

I am now considering running a trend analysis using logistic regression assuming linearity

logistic recovery i.wellbeing i.hostpital
contrast p.wellbeing

Any suggestion or resource would be greatly appreciated.

Thanks,
Laura