Joint modelling multivariate longitudinal and time-to-event data: An application to ROI dataset

Modern health datasets increasingly include multivariate longitudinal data for addressing research questions about their association with time-to-event outcomes. However, high-dimensional longitudinal outcomes make standard shared-parameter joint models computationally infeasible. To overcome this limitation, a recently proposed joint modelling formulation assumes a common distribution for the outcome-specific and time-to-event random effects. This approach also enables estimation via pairwise-fitting to reduce computational complexity. We illustrate the application of this methodology to the IRCCS Rizzoli Orthopedic Institute dataset, addressing the research questions of interest and providing dynamic predictions for new patients.