Confidence limits for logistic regression equation

I have generated a logistic model for a disease status Y (unaffected-affected) using several independent variables on a very large (600K+) dataset. I then want to apply that formula to a much smaller subset of that data set, having a particular characteristic not included in the model. I then sum the probabilities to get the expected number of affecteds, given that characteristic, and create a relative risk estimate by dividing the observed in that group by the expected. What I'm unable to do, it seems, is generate an upper and lower confidence interval around that expected value (and, thus, CI's for the RR). I've never seen this done, so I've found no guidance -- and maybe it's just a bad idea.

Several packages I've used give std. errors for the Beta's for each independent variable, but no std. error for the entire equation. (I suspect that this is because logistic regression is not closed form as OLS regression is.) In absence of that, could I create two alternative logistic regression equations by modifying all the beta's by their std. errors (+/-1.96*std. error), then calculate new probabilities for each person, and sum them? Is there an alternative? Any help would be much appreciated.