Missing data is a common problem in survey based research. There are many packages that compensate for missing data but few can easily compensate for missing longitudinal data. WinBUGS compensates for missing data using multiple imputation, and is able to incorporate longitudinal structure using random effects. We demonstrate the superiority of longitudinal imputation over cross-sectional imputation using WinBUGS. We use example data from the Australian Longitudinal Study on Women's Health. We give a SAS macro that uses WinBUGS to analyze longitudinal models with missing covariate date, and demonstrate its use in a longitudinal study of terminal cancer patients and their carers.