Joint modeling of longitudinal zero-inflated count and time-to-event data: A Bayesian perspective

Area of Science:

• Biostatistics
• Epidemiology
• Substance-Use Research

Background:

• Longitudinal zero-inflated count data are common in substance-use research.
• These data often involve time-to-event outcomes (e.g., death, dropout) that may depend on longitudinal measurements.
• Existing models may not adequately handle the complexities of both zero-inflated counts and dependent survival data.

Purpose of the Study:

• To develop and validate a joint statistical model for simultaneously analyzing longitudinal zero-inflated count data and time-to-event data.
• To address situations where longitudinal outcomes influence the risk of a terminal event.

Main Methods:

• Utilized a joint model combining a Cox proportional hazards model with a piecewise constant baseline hazard for survival data.
• Incorporated zero-inflated count models for longitudinal outcomes.
• Employed a Bayesian framework using Markov chain Monte Carlo (MCMC) simulations via the BUGS programming language.
• Conducted an extensive simulation study to evaluate model performance.

Main Results:

• The proposed joint model yielded more accurate parameter estimates compared to traditional independence models.
• Demonstrated the model's ability to account for the dependence between longitudinal outcomes and time-to-event data.
• The simulation study confirmed the robustness and improved accuracy of the joint modeling approach.

Conclusions:

• The developed joint model effectively handles longitudinal zero-inflated count data and time-to-event data with dependent structures.
• This approach offers improved accuracy for analyzing complex data common in substance-use and health-related research.
• The method was successfully applied to a lung cancer prevention study involving alpha-tocopherol and beta-carotene.