A flexible joint modeling framework for longitudinal and time-to-event data with overdispersion.

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Survival Analysis

Background:

• Joint statistical models are crucial for analyzing multiple correlated outcomes.
• Existing models often impose restrictive assumptions on non-Gaussian outcomes.
• Incorporating survival data alongside non-Gaussian outcomes presents unique analytical challenges.

Purpose of the Study:

• To develop an extended joint statistical model incorporating both conjugate and normal random effects.
• To accommodate non-Gaussian outcomes, particularly survival data, within a unified framework.
• To improve model fit and the accuracy of significance tests in longitudinal studies.

Main Methods:

• Development of a joint model combining conjugate random effects for non-Gaussian outcomes and normal random effects for correlations.
• Utilizing conjugate random effects to relax restrictive mean-variance assumptions.
• Employing normal random effects to capture within-subject correlation and between-outcome associations.
• Maximum likelihood estimation facilitated by analytical integration over conjugate random effects.

Main Results:

• The proposed extended framework significantly improves model fit compared to standard approaches.
• Switching to the enhanced model can impact the significance of statistical tests.
• The model demonstrates applicability and improved performance in a chronic heart failure case study.

Conclusions:

• The novel joint model offers a more flexible and accurate approach for analyzing complex longitudinal data with non-Gaussian and survival outcomes.
• This framework enhances statistical power and reliability in health-related research.
• The method is computationally feasible, estimable using standard statistical software.