The effect of depression on cognitive decline: correcting for the effect of
attrition due to illness and death related to depression and to cognitive
 
   Lead Investigator:    Kyle Steenland
   Institution      :    Emory
   E-Mail           :    nsteenl@sph.emory.edu
   Proposal ID      :    428



   Proposal Description:
   Studies of the effect of depression on cognitive decline are potentially biased toward the null by the loss of subjects who  are depressed and who may show marked cognitive decline, because both of these factors are associated with increased morbidity and mortality, and hence with increased likelihood to drop out of longitudinal studies over time.  Such a bias, and methods to avoid the bias, have been noted before in relation to smoking and cognitive decline (Weuve et al. 2012). Weuve et al. (2012) used marginal structural models (inverse probability weighting) to try to account for attrition in a study of cognitive decline and smoking in a Chicago cohort (Chicago Health and Aging Project participants,  who were current smokers or never-smokers, and  underwent cognitive assessments up to 5 times at 3-year intervals.).   The authors found that the negative effects of smoking on cognitive decline doubled in magnitude after accounting for attrition.   Chou et al. (2015) has recently used both MSMs and newer proposed methods to account for bias due to death during follow-up, in a study of the incidence of dementia in NACC data.  These authors noted that adjustment for such bias increased the estimate effect of depression, and other risk factors,  on the incidence of dementia.   We will continue this work using MSMS to adjust for bias due to drop-out in estimating the effect of depression on  cognitive decline over time,  with subjects stratified by normal or MCI status at baseline, and considering drop-out due to either death or other reasons.  We will contrast results using a standard approach with the approach using marginal structural models.  We will used the Milestone Reports to estimate who has dropped out of follow-up for either death or other reasons.  Using marginal structural modeling, we will estimate the effect of depression  on cognitive decline after weighting the data such that it mimics the data that would have been observed had the drop-outs stayed unde