A class of joint models for multivariate longitudinal measurements and a binary event

Area of Science:

• Biostatistics
• Obstetrics and Gynecology
• Perinatal Medicine

Background:

• Predicting adverse pregnancy outcomes like large birthweight is vital for clinical decision-making.
• Longitudinal ultrasound measurements offer valuable data for predicting fetal growth and birth events.
• Existing prediction models may not fully capture the complex relationship between serial ultrasound data and binary birth outcomes.

Purpose of the Study:

• To develop a flexible class of joint models for multivariate longitudinal ultrasound measurements.
• To predict binary events at birth, specifically focusing on large birthweight.
• To establish a robust statistical framework for integrating serial ultrasound data in perinatal risk assessment.

Main Methods:

• A skewed multivariate random effects model for longitudinal ultrasound data.
• A skewed generalized t-link function to connect binary outcomes with longitudinal processes.
• Bayesian inference using Markov chain Monte Carlo (MCMC) sampling for parameter estimation.
• Model comparison using deviance information criterion (DIC) and log pseudomarginal likelihood (LPML).
• Validation through a training-test set prediction paradigm.

Main Results:

• The proposed joint modeling approach effectively integrates multivariate longitudinal ultrasound data for predicting binary birth events.
• Shared random effects successfully link the underlying longitudinal growth trajectories with the probability of large birthweight.
• Model variations were compared, demonstrating the utility of the proposed skewed multivariate framework.
• The methodology showed good predictive performance on a held-out test set.

Conclusions:

• The developed joint models provide a flexible and powerful tool for predicting large birthweight using serial ultrasound data.
• This approach enhances the ability of obstetricians to identify high-risk pregnancies, potentially improving perinatal outcomes.
• The study highlights the importance of advanced statistical modeling in leveraging complex longitudinal data for clinical prediction in obstetrics.