Two-stage recurrent events random effects models.

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Recurrent events and terminal events often occur together in longitudinal studies.
• Existing models may not fully capture the complex dependency structures between these event types.

Purpose of the Study:

• To develop and evaluate semiparametric random-effects models for recurrent events in the presence of a terminal event.
• To model the dependency between recurrent and terminal events using shared random effects.

Main Methods:

• Utilized proportional marginal rate or mean models for recurrent events and a proportional model for the terminal event.
• Formulated two-stage models allowing for full or partial sharing of random effects.
• Employed a parameter estimation procedure that avoids tuning parameters and numerical integration.
• Standard errors were calculated using bootstrapping.

Main Results:

• The proposed estimation procedure is numerically stable and effective.
• The models successfully capture the dependency between recurrent and terminal events.
• The methods were validated using data from the Taichung Peritoneal Dialysis Study.

Conclusions:

• The developed semiparametric random-effects models provide a flexible and efficient framework for analyzing recurrent events with a terminal event.
• The two-stage estimation approach offers a practical alternative to methods requiring numerical integration.
• The application to dialysis patient data demonstrates the utility of the models in real-world health research.