A penalized algorithm for event-specific rate models for recurrent events

Area of Science:

• Biostatistics
• Survival Analysis
• Recurrent Event Data Analysis

Background:

• Recurrent event data analysis is crucial in many fields.
• Semiparametric models are widely used for recurrent events.
• Classical estimators may lack efficiency in small to moderate sample sizes.

Purpose of the Study:

• To develop a covariate-specific total variation penalty for recurrent event data.
• To enhance the performance of semiparametric models like the Cox and Aalen's additive models.
• To provide more accurate estimations for recurrent event processes.

Main Methods:

• Introduction of a covariate-specific total variation penalty.
• Application to stratified Cox and stratified Aalen's additive models.
• Demonstration of consistency and asymptotic normality for penalized estimators.

Main Results:

• Penalized estimators show consistency and asymptotic normality.
• Simulation studies indicate superior performance of penalized estimators over classical ones for small-to-moderate sample sizes.
• The method was applied to real-world bladder tumor recurrence data.

Conclusions:

• The proposed covariate-specific total variation penalty offers an improvement for recurrent event data analysis.
• This method enhances the reliability of semiparametric models, particularly when sample sizes are limited.
• The findings have practical implications for analyzing time-to-event data in various research areas.