Differences in Estimation Between the Longitudinal Model and the Longitudinal Portion of the Joint Model

• Activity Number: 67 - Longitudinal Biometrics Data
• Type: Contributed
• Date/Time: Sunday, July 30, 2017 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Kendra Plourde* and Yorghos Tripodis and Elizabeth Rose Mayeda
• Companies: Boston University and Boston University and University of California
• Keywords: joint models ; longitudinal models ; mixed effects

Abstract

We investigate the effect of joint survival and longitudinal models on the precision and accuracy of the longitudinal estimates. Mixed effects analysis has allowed investigators to incorporate more information by including repeated measures. Recently, joint models consisting of a cox proportional hazards model and a longitudinal mixed effects model have allowed investigators to also incorporate time-to-event data. By incorporating more information, we expect the estimates of the longitudinal portion of the joint model to be less biased and more precise. Research has shown improvement in estimation of the hazard function using joint models, but not much research has been done to investigate the differences in estimation of the longitudinal model. In this study, we compared the longitudinal model with the longitudinal portion of the joint model in terms of coverage, bias, and precision using the same simulation structure used previously (Mayeda, 2015). Our results show that the estimate of the longitudinal portion of the joint model is more accurate and less prone to type I error when an unmeasured continuous variable is correlated with both the exposure of interest and the event.