Bayesian semi-parametric inference for clustered recurrent events with zero inflation and a terminal event

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Survival Analysis

Background:

• Recurrent events are prevalent in clinical research and often influenced by terminal events.
• Pragmatic trials frequently involve clustered participant data (e.g., within clinics), introducing complexities in analyzing susceptibility to recurrent events.
• Existing statistical models may not fully capture the intricate hierarchical structures and shared influences common in such data.

Purpose of the Study:

• To develop a flexible Bayesian shared random effects model for analyzing recurrent events in clustered data.
• To enhance robustness by employing Dirichlet processes for modeling survival residuals and cluster-specific frailty distributions.
• To provide an efficient computational method for posterior inference in complex survival settings.

Main Methods:

• Development of a Bayesian shared random effects model incorporating Dirichlet process priors.
• Application of the accelerated failure time model for the survival process.
• Modeling of cluster-specific shared frailty distributions using Dirichlet processes.
• Implementation of an efficient sampling algorithm for posterior inference.

Main Results:

• The proposed Bayesian model effectively accommodates the complex data structure of recurrent events in clustered pragmatic trials.
• The use of Dirichlet processes enhances the model's robustness in capturing heterogeneity in survival processes and cluster effects.
• The developed sampling algorithm allows for efficient posterior inference, facilitating practical application.

Conclusions:

• The novel Bayesian shared random effects model offers a powerful and robust approach for analyzing recurrent events in clustered clinical trials.
• This methodology provides valuable insights into factors influencing recurrent events within hierarchical structures, applicable to various health research areas.
• The findings demonstrate the utility of advanced Bayesian techniques, including Dirichlet processes, for addressing complex data challenges in clinical epidemiology.