|
Abstract:
|
Multivariate longitudinal outcomes are collected in many clinical studies on Parkinson’s disease (PD) to fully explore the impairment caused by the disease. If the outcome deteriorates non-linearly, it may be unrealistic parametric assumption in the underlying disease severity. For an accurate prediction of the time to functional disability, we have assumed functional coefficients in the latent variable of each covariate to make informative medical decisions for clinicians. In this article, we first propose a joint model that consists of a functional latent trait model for the multivariate outcomes and a survival model for event time by linking two submodels with a latent variable. Hypothesis testing procedures are provided to calculate the p-values of functional coefficients whether they depart from zero. We also develop a Bayesian approach for parameter estimation and a dynamic prediction framework for predicting target patients’ future risk of a survival event based on their current outcome and covariate histories. The proposed model is evaluated by simulation studies and applied to the PPMI, an observational study assessing the progression of clinical features of PD.
|
