A1432
Title: Joint modeling of zero-inflated longitudinal and survival data with a cure fraction: An application to AIDS data
Authors: Taban Baghfalaki - The University of Manchester (United Kingdom) [presenting]
Mojtaba Ganjali - Shahid Beheshti University (Iran)
Abstract: The purpose is to develop a Bayesian joint modeling framework for analyzing zero-inflated longitudinal count data and survival outcomes, with explicit incorporation of a cure fraction to account for individuals who will never experience the event of interest. The longitudinal trajectory is modeled using a flexible mixed-effects Hurdle specification to address excess zeros and overdispersion that commonly arise in biomedical count data. The survival process is represented through a Cox proportional hazards mixture cure model, enabling separation of cured from susceptible subjects. To capture the association between the two processes, the survival model incorporates a linear combination of current longitudinal values as time-dependent predictors. Bayesian inference is carried out using Hamiltonian Monte Carlo, which ensures efficient posterior sampling and reliable parameter estimation in complex settings. The framework supports dynamic prediction, enabling individualized risk assessment and personalized clinical decision-making. Extensive simulation studies are conducted to evaluate estimation accuracy, predictive performance, and robustness of the proposed model. Finally, the methodology is applied to a real AIDS cohort, illustrating its capacity to integrate longitudinal biomarkers with survival information for improved prediction of patient outcomes. Results underscore the clinical utility of joint modeling in advancing personalized medicine.
