Bayesian Joint Modelling

The selected primary model used cumulative creatinine exposure as its association structure because it had the best WAIC and LPML. Joint models simultaneously model repeated biomarker trajectories and event risk while accounting for measurement error, correlation, censoring, late entry, and the biomarker-event link. Model estimation was performed with JMbayes2 in R using weakly informative priors and MCMC. Separate longitudinal and survival models can be biased when biomarker trajectories are related to event status. The study used Bayesian joint modelling to connect serum creatinine trajectories with time to first AKI, CKD, or death in a UK paediatric autoimmune cohort. Joint models connect longitudinal biomarker trajectories with time-to-event outcomes while accounting for noisy and intermittent marker observations. The proposed framework lets researchers fit nested linear, quadratic, and spline association structures in joint models. The article presents joint modelling as suitable for paediatric nephrology because it can link biomarker history and event risk while updating predictions as new measurements arrive.