Joint longitudinal and time-to-event models: development, implementation and applications in health research

It is common in epidemiology and clinical research to take repeat measurements of a marker (for example blood pressure) while observing the time course to an event (for example heart attack). This thesis focuses on novel statistical models for joint analysis of these two sources of information on a person's health. The research contributes to the uptake of these models in three ways. First, by extending the underlying methodology. Second, through the development of new software for fitting the models. Third, by demonstrating how the models can be used to answer important health research questions in separate studies of weight trajectories in patients with end stage kidney disease and impacts of natural disasters on multiple health conditions.