Multivariate Generalized Linear Mixed-Effects Models for the Analysis of Clinical Trial-Based Cost-Effectiveness Data

Area of Science:

• Health Economics
• Biostatistics
• Clinical Trials

Background:

• Economic evaluations alongside randomized controlled trials (RCTs) are crucial for assessing healthcare intervention cost-effectiveness.
• These studies typically collect prospective data on resource use, economic costs, and clinical outcomes.
• Analyzing multiple, correlated outcomes of mixed data types presents a statistical challenge.

Purpose of the Study:

• To extend the generalized linear mixed-model framework for simultaneous modeling of multiple, mixed-type outcomes.
• To develop and implement new wrapper functions for statistical software (Stata and R).
• To compare the performance of these new methods against existing implementations.

Main Methods:

• Extension of the generalized linear mixed-model framework.
• Simultaneous modeling of multivariate hierarchical data structures.
• Implementation in Stata and R using maximum and restricted maximum quasi-likelihood estimation.
• Comparison with Stata's merlin/gsem and a Bayesian approach in WinBUGS.

Main Results:

• The new methods provide broadly similar results to existing approaches.
• Empirical applications using observed and simulated clinical trial data validate the methods.
• The developed wrapper functions facilitate the analysis of complex economic evaluations.

Conclusions:

• The extended generalized linear mixed-model framework effectively handles multivariate hierarchical data in health economic investigations.
• These methods are particularly applicable to economic evaluations conducted alongside randomized controlled trials.
• The new statistical tools enhance the ability to generate high-quality evidence on the cost-effectiveness of healthcare interventions.