I have a time series of binary outcomes (case yes/no) and several covariates. The goal is to estimate incidence and prevalence of the outcome. I would like to perform multiple imputation for both outcomes and covariates across the time series. Some of the subjects appear more than once in the time series. How does one perform the imputation to ensure sensible imputed values for each subject id? For example, if in year 2010 the imputation algorithm assigns subject id a "case", shouldn't this subject also be defined a case for subsequent years? I thought of perhaps sequentially doing the imputation process for each year, but it seems convoluted. Another thought was that one can consider subject id as just a random sample in each year and not be overly concerned with tracking the outcome status across time during the imputation process.
