Modeling multivariate survival data by a semiparametric random effects proportional odds model

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Multivariate survival data analysis requires modeling complex dependence structures.
• Existing models like proportional hazards may not capture all nuances of such data.
• Examples include clustered data and recurrent event times, common in medical research.

Purpose of the Study:

• To propose a flexible random effect semiparametric proportional odds model for multivariate survival data.
• To offer an alternative to traditional proportional hazards models.
• To develop a unified estimation procedure for regression and dependence parameters.

Main Methods:

• A random effect semiparametric proportional odds model with multivariate normal random effects.
• Additive effect of random effects on the baseline log-odds function.
• Marginal-likelihood approach with Monte Carlo approximation for estimation.

Main Results:

• The proposed model accommodates various dependence structures and includes special cases like shared random effects and additive variance components models.
• Regression parameter estimates are robust to the choice of correlation structure.
• Simulation studies demonstrate the method's performance.

Conclusions:

• The proposed model provides a flexible and robust framework for analyzing diverse multivariate survival data.
• The unified estimation procedure effectively handles regression and dependence parameters.
• The method is applicable to real-world data, including clustered and recurrent event data.