Semiparametric multivariate joint model for skewed-longitudinal and survival data: A Bayesian approach

Area of Science:

• Statistical modeling
• Biostatistics
• Longitudinal data analysis

Background:

• Traditional joint models often assume normality for longitudinal data, which may not reflect real-world subject variations.
• Handling multiple correlated longitudinal outcomes of mixed types (continuous/categorical) and informative missing data presents significant challenges in statistical inference.
• Existing parametric models may lack flexibility for complex longitudinal patterns.

Purpose of the Study:

• To develop a flexible semiparametric joint model for analyzing mixed-type longitudinal data and survival outcomes.
• To incorporate the skew-t distribution to better accommodate non-normal error distributions in longitudinal models.
• To address challenges posed by correlated multiple outcomes and nonignorable missing data within a unified framework.

Main Methods:

• Development of an extended multilevel item response theory (MLIRT) model for mixed-type longitudinal data.
• Integration of the MLIRT model with a Cox proportional hazards model via shared random-effects.
• Application of a Bayesian approach for parameter estimation and inference in the joint model.
• Utilizing skew-t distribution to model longitudinal data errors, enhancing flexibility over normal assumptions.

Main Results:

• Simulation studies demonstrated the proposed model's performance in handling complex data features.
• The joint model effectively estimated parameters under various nonignorable missing data mechanisms.
• Analysis of a primary biliary cirrhosis study provided insights into disease progression and survival.

Conclusions:

• The proposed MLIRT-based semiparametric joint model offers a robust approach for analyzing longitudinal and survival data with mixed types and non-normal errors.
• The Bayesian framework facilitates effective inference, particularly when dealing with informative missing data.
• This methodology enhances the accuracy of statistical inference in complex clinical studies.