Assessing Risk Indicators in Clinical Practice with Joint Models of Longitudinal and Time-to-Event Data

Abstract

Studies in life course epidemiology often involve different types of outcomes and exposures being collected on individuals, who are followed over time. The data include longitudinally measured responses (e.g., biomarkers), and the time until an event of interest occurs (e.g., death, intervention). In many epidemiologic studies, these outcomes are separately analysed, although it may be of public health interest to study their association while including key exposures. To that end, it is desirable to employ methods that examine the associations of exposures with longitudinal measurement outcomes simultaneously. This method is referred in the statistical literature as joint modelling of longitudinal and survival data. The idea behind joint modelling of longitudinal and survival data is usually to couple linear mixed effects models for longitudinal measurement outcomes and Cox models for censored survival outcomes. Recent extensions of these models motivated by real-life applications, including shrinkage approaches, time-varying effects and latent classes, will be presented. Furthermore, subject-specific dynamic risk predictions based on the joint modelling framework will be illustrated.