Shared frailty models for recurrent events and a terminal event

Area of Science:

• Biostatistics
• Survival Analysis
• Medical Statistics

Background:

• Recurrent event data analysis is increasingly important in medical research.
• Terminating events (e.g., death) can interrupt recurrent events and may depend on event history.
• Existing models often struggle to capture the dependence between recurrent and terminal events.

Purpose of the Study:

• To propose and illustrate frailty proportional hazards models for analyzing recurrent and terminal event data.
• To model the dependence between recurrent event history and time to a terminal event.
• To provide a method for estimating the degree of dependence between these events.

Main Methods:

• Utilized shared frailty proportional hazards models for both recurrent and terminal event processes.
• Incorporated covariate effects into the hazard functions.
• Employed a Monte Carlo EM algorithm with a Metropolis-Hastings sampler for maximum likelihood estimation and inference.
• Applied the methods to hospitalization and death data for waitlisted dialysis patients.

Main Results:

• The proposed model effectively analyzes recurrent and terminal event data, accounting for their dependence.
• Demonstrated the application of the model using real-world patient data.
• Showcased methods for validating the model's assumptions.
• The model successfully estimates the degree of dependence between recurrent and terminal events.

Conclusions:

• The developed frailty proportional hazards model offers a robust approach to analyzing complex event data where terminal events influence recurrent events.
• This method overcomes limitations of alternative approaches by directly modeling dependence.
• The analysis of dialysis patient data highlights the practical utility of the proposed statistical framework.