Subpopulation analysis: Multiple imputation of survey data

Hi there,

I have a dataset with survey data elements (i.e., psu, strata) and am doing a multiple logistic regression with 12 variables (mostly binary or categorical). Three of which predict whether a variable is missing. Given that the variables are "missing at random" (MAR), I know complete cases analysis would yield biased estimates. Multiple imputation (MICE) is ideal.

However, I haven't found any documentation on multiple imputation of a survey dataset, where analysis is for a subpopulation. Has anyone done something similar in STATA? Or, do you have any leads on articles/books on the subject?