Joint modelling of longitudinal ordinal and multi-state data.

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Survival Analysis

Background:

• Joint modeling of longitudinal and survival data is common in biomedical research.
• Existing models struggle with longitudinal ordinal responses and non-ignorable missing data in multi-state processes.

Purpose of the Study:

• To introduce a new joint model for longitudinal ordinal measurements and multi-state data.
• To address non-ignorable missing values in longitudinal ordinal responses within a multi-state framework.

Main Methods:

• Developed a joint model with two sub-models: proportional odds for ordinal data and multi-state for transitions.
• Linked sub-models using shared random effects.
• Employed maximum likelihood estimation with a piecewise constant baseline hazard function.

Main Results:

• The proposed joint model is suitable for longitudinal ordinal responses with non-ignorable missing values in multi-state processes.
• Simulation studies demonstrated the model's effectiveness.
• The model was successfully applied to human immunodeficiency virus (HIV) patient data.

Conclusions:

• The novel joint model provides a robust framework for analyzing complex longitudinal ordinal and multi-state data.
• This approach enhances the understanding of disease progression and treatment effects in biomedical studies.
• The model is particularly valuable for datasets with event-dependent missingness.