64th ISI World Statistics Congress - Ottawa, Canada

64th ISI World Statistics Congress - Ottawa, Canada

A time-varying joint model for longitudinal and survival data

Conference

64th ISI World Statistics Congress - Ottawa, Canada

Format: IPS Abstract

Keywords: dynamic, joint models, longitudinal, survival

Session: IPS 161 - Modeling complex correlated data: new directions and innovations

Thursday 20 July 10 a.m. - noon (Canada/Eastern)

Abstract

Motivated by the Midlife in the United States (MIDUS) study, we propose a time-varying joint model (TV-JM) for longitudinal and time-to-event outcomes. In our approach, these two outcomes are modeled through a common set of subject-specific random effects. We depart from the literature on joint modeling of longitudinal and survival data by allowing dynamic response-predictor as well as response-response relationships in the model. For estimation of the model parameters, an approximate Expectation-Maximization algorithm is developed. In the E-step of the algorithm, Gauss-Hermite quadrature procedure is used to predict the underlying random effects, and in the M-step, local linear regression techniques are employed to fit the time-varying coefficients. The standard error formulas for the estimated parameters are also derived. The finite-sample performance of our proposed method is assessed using a Monte Carlo simulation study. We illustrate our method by applications to the MIDUS data to explore the effects of subject-level covariates on survival and socio-economic status during adulthood of older adults in the United States.