Joint random-effects transition models for multivariate longitudinal data 

 
   Lead Investigator:    Bo Zhang
   Institution      :    Oregon State University
   E-Mail           :    bo.zhang@oregonstate.edu
   Proposal ID      :    230


   Publications:


1  Liu W,  Zhang B, Zhang Z, Zhou XH
   Joint Modeling of Transitional Patterns of Alzheimer's Disease
   Journal Article  PLoS One,8:e75487,2013,PMC3779177

   PUBLISHED





   Proposal Description:
   The ongoing project is a statistical methodology study, where we intend to develop multivariate transition models for longitudinal data with various types of outcomes and use NACC data of Alzheimer's disease as the analysis demonstration of the proposed model. The scientific focus of many longitudinal clinical trials and epidemiologic studies is on the disease process over time. Especially in the studies on chronic diseases, interest usually lies in understanding the longitudinal transitions of primary disease endpoints, as the transition patterns deliver important message on the disease evolution and prognosis. A common problem in analyzing the data from these studies is that nowadays the disease status is constantly represented by, not only one, but multiple endpoints. As a consequence, analyzing the study outcomes individually by univariate transitional models fails to account for the association between important disease characteristics and can also result in severe bias. This project proposes an approach for analyzing the transition patterns of multivariate longitudinal data. The expected data structure requested is stated as follows. The transitional models are multivariate models. Therefore, we need multiple primary outcomes in the data. The number of our primary outcomes is expected to be larger or equal to 4. In addition, to demonstrate the strength of our method, we need some variability in outcome types that is, it would be ideal if we have some continuous outcome, some binary outcome, and even better have some count outcome. Each outcome variable is one of the important characteristics of the Alzheimer's disease that can capture the evolution and development of the disease. Each outcome variable needs to be measured over time, say every half year or every one year. As a result, besides the primary outcomes, the time will be one covariate. Of course, we welcome the onsite epidemiologists to give suggestion and would love to get more variables as the addi