J. Gutiérrez Botella, C. Armero, T. Kneib, J. García-Seara
Patients with Heart Failure (HF) are treated with Cardiac Resynchronization Therapy (CRT), which is proved to work well in short-term periods. In this study, patients with HF undergoing CRT are monitored for more than 10 years. During the follow-up period, baseline clinical and demographic information is recorded, as well as longitudinal cardiac markers, in particular ejection fraction (EF) and mitral insufficiency (MI). Causes of death, cardiovascular and non-cardiovascular, and subsequent survival times are the main outcomes of the study. Bayesian joint competing risks models were developed, providing longitudinal information (mixed beta regression for EF, mixed ordinal regression for MI) for the cause-specific Cox hazards for the survival sub-model. Model performance was assessed via Bayesian metrics (WAIC) and predictive measures (CPOs and LPML). Posterior distribution was estimated via MCMC with JAGS, while derived posterior and predictive outputs were also obtained and discussed.
Keywords: Cardiovascular and non-cardiovascular death, Heart Failure, Mixed beta regression, Mixed ordinal regression
Scheduled
Métodos bayesianos aplicados a la medicina
June 12, 2025 7:00 PM
Sala VIP Jaume Morera i Galícia