Identifying Clusters from Multidimensional Symptom Trajectories among subjects
with mild cognitive impairment 
   Lead Investigator:    Sudeshna Paul
   Institution      :    Emory University
   E-Mail           :    spaul5@emory.edu
   Proposal ID      :    1213



   Proposal Description:
   Motivated by the National Alzheimer's Coordinating Center Uniform Data Set, we propose to jointly model the multivariate symptom trajectories of the subjects with mild cognitive impairment and classify them into multiple groups or clusters. By focusing on the joint distribution of the symptoms (cognitive, behavior, motor), we hope to better describe their association and evolution over time as well as identify groups of subjects with similar characteristics.    We will implement a multivariate generalized linear mixed effects model with a mixture of normally distributed random effects to model the joint trajectories. In addition, effects of time-varying predictors will also be accounted for to enhance our model. A Bayesian inferential approach based on the Markov chain Monte Carlo will be exploited to fit the model parameters. By using this methodology we can potentially identify groups of subjects that are at higher risk of developing severe symptoms and help clinicians develop targeted interventions.