Bivariate Mixed Effects Analysis of Clustered Data with Large Cluster Sizes

Area of Science:

• Biostatistics
• Health Services Research
• Clinical Trials

Background:

• Linear mixed effects models are standard for clustered data.
• Comparing hospital length of stay (LOS) between percutaneous coronary intervention (PCI) and coronary artery bypass graft (CABG) is crucial.
• Existing statistical software struggles with large datasets in clinical trials.

Purpose of the Study:

• To propose a bivariate linear mixed effects model for joint modeling of clustered PCI and CABG LOS.
• To address computational challenges in maximum likelihood (ML) and restricted maximum likelihood (REML) inference for large-scale clinical trial data.
• To develop an efficient and stable REML EM algorithm for analyzing complex LOS data.

Main Methods:

• Development of a bivariate linear mixed effects model.
• Implementation of an expected and maximization (EM) algorithm for REML inference.
• Utilizing existing software for ML inference with modifications.

Main Results:

• The proposed REML EM algorithm is computationally stable and efficient.
• The algorithm successfully circumvented memory allocation issues in commercial software.
• Meaningful results were obtained for the analysis of clustered LOS data.

Conclusions:

• The novel REML EM algorithm provides a viable solution for analyzing large, clustered LOS data.
• This method facilitates robust comparisons of treatment outcomes like PCI vs. CABG.
• The approach enhances the feasibility of complex statistical modeling in large international clinical trials.