A beta-binomial mixed-effects model approach for analysing longitudinal discrete and bounded outcomes

Area of Science:

• Biostatistics
• Health Outcomes Research
• Longitudinal Data Analysis

Background:

• Patient-reported outcomes (PROs) are crucial for understanding patient health status but often exhibit complex distributions unsuitable for standard statistical models.
• Longitudinal PRO data presents challenges due to repeated measures and inherent correlation structures.
• Existing statistical methods have limitations in accurately modeling the discrete and bounded nature of PROs.

Purpose of the Study:

• To develop and propose an advanced estimation procedure for analyzing correlated discrete and bounded outcomes, specifically PROs.
• To implement this novel methodology within a user-friendly R package (PROreg).
• To compare the performance of the proposed method against existing approaches.

Main Methods:

• Development of a beta-binomial mixed-effects model tailored for longitudinal PRO data.
• Implementation of the statistical methodology in the R programming language.
• Comparative analysis with alternative statistical models commonly used in R.

Main Results:

• The proposed beta-binomial mixed-effects model demonstrates superior performance compared to existing methodologies.
• The new estimation procedure effectively handles the unique distributional properties of PROs.
• The PROreg package provides a robust tool for analyzing longitudinal health status data.

Conclusions:

• The beta-binomial mixed-effects model offers a statistically sound and effective approach for analyzing longitudinal PROs.
• This methodology facilitates a deeper understanding of disease progression and risk factors.
• The findings have significant implications for improving patient care and clinical research, particularly in chronic diseases like COPD.