Joint Analysis of Survival and Longitudinal Data Subject to Left-Censoring and Non-Ignorable Missing

• Activity Number: 630
• Type: Topic Contributed
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Abdus Sattar*+ and Sanjoy Sinha and Xiaofeng Wang and Yehua Li
• Companies: and Carleton University and Cleveland Clinic Lerner Research Institute and Iowa State University
• Keywords: Joint modeling ; Frailty models ; left-censoring ; non-ignorable missing ; nonparametric smoothing

Abstract

Joint analysis of a longitudinal response variable and survival data that are collected from same individual is a standard statistical approach. This joint analysis can be complicated for a number of reasons. For instance, there could be non-ignorable missing and left-censoring in longitudinal data. In addition, there could be heterogeneity in the study population for a variety of reasons. Ignoring these problems in either longitudinal or survival sub-models, standard joint modeling approach likely to provide inconsistent parameter estimates and hence misleading inference. In this study we propose to show that the problem of left-censoring in longitudinal response sub-model can be overcome by applying a nonparametric smoothing technique and a numerical integration method in the likelihood framework. Issues of missing data in longitudinal sub-model and heterogeneity in survival data can be addressed using the method of maximum likelihood and frailty model, respectively. Simulation results will be presented to evaluate the performance of the proposed methodology. The proposed method also be applied to a clinical study, Genetic and Inflammatory markers of Sepsis Study.