Abstract:
|
When monitoring the health of subjects, often times multiple risk factors are measured over time. Modeling these longitudinal risk factors simultaneously, where the correlation between the risk factors are taken into account, can be beneficial, specially when there exists differential measuring density in the collected risk factors. Further, the association between the collected risk factors and the survival outcomes is often the practitioners' primary interest. We proposed a joint longitudinal-survival modeling framework with a longitudinal component capable of modeling multiple longitudinal processes simultaneously where the correlation between those processes are taken into account. Our proposed modeling framework is robust to common distributional assumptions in that we avoid simple distributional assumptions on longitudinal measures and allow for subject-specific baseline hazard in the survival component. Fully joint estimation is performed to account for uncertainty in the estimated time-dependent biomarker covariate in the survival model. Simulation studies are presented as well as results of applying our proposed model on real data on end-stage renal disease patients.
|