Estimating the causal effect of time-dependent antidepressant usage on survival Lead Investigator: Ran Duan Institution : Department of Biostatistics,University of Kentucky E-Mail : ran.duan@uky.edu Proposal ID : 933 Proposal Description: The primary goal of this study is to estimate the causal effect of time-dependent antidepressant usage on survival of MCI patients using marginal structural models as a tool of correcting biased estimate using conventional modeling techniques. The hypothesis is that marginal structural models would generate less biased estimates in comparison to conventional modeling techniques. We propose to perform a retrospective cohort study among MCI patients using the Uniform Data Set. Survival time from first diagnosis of MCI would be the primary outcome of interest. Longitudinal antidepressant usage would be the primary exposure of interest. Other covariates of interest include age in years, sex, race, education in years, smoking in years, alcohol consumption, health conditions and medications that may influence the patients??? survival, global CDR, GDS score and MMSE score. Findings from this study may help inform the causal effect of long-term antidepressant usage on survival of MCI patients.