A Flexible Bayesian Semi-Parametric Multivariate Mixed-Effect Model Framework for Skewed Longitudinal Data

Abstract

While existing mixed-effects models address multivariate outcomes, nonlinearity, or non-normality separately, their joint integration remains limited in medical research. This study proposes a flexible semi-parametric multivariate mixed-effects model that simultaneously accounts for correlated outcomes, nonlinear covariate effects, and skewness, providing a more comprehensive framework for analyzing complex longitudinal data. The proposed approach is illustrated using longitudinal data on fasting blood sugar (FBS) and systolic blood pressure (SBP) among individuals with type 2 diabetes (T2D) and hypertension. A simulation study was also conducted to assess model performance. Results indicate that the proposed model outperforms conventional approaches by effectively capturing nonlinear patterns and asymmetry in the data. The findings further show a strong and direct relationship between FBS and SBP over time. Overall, the proposed model provides a robust and flexible framework for analyzing complex multivariate longitudinal data in medical research.