This is a question of a statistical nature about what multiple imputation can accomplish. I haven't used MI earlier. I'm working with colleagues to do a prospective cohort analysis of a public health evaluation project, where loss to follow up sample is