Bayesian Joint Modeling for Longitudinal Magnitude Data With Informative Dropout: An Application to Critical Care...

Area of Science:

• Biostatistics
• Biomedical Data Analysis

Background:

• Biomedical studies often analyze data magnitudes where signs are irrelevant.
• Repeated measures studies with magnitude outcomes require random effects to address individual heterogeneity and improve precision.
• Existing regression methods lack specific approaches for magnitude data with random effects.

Purpose of the Study:

• To introduce Bayesian regression modeling for magnitude data incorporating random effects.
• To extend the method for informative dropout using joint modeling.
• To analyze diaphragm thickness changes in ICU patients, examining sex-based impacts.

Main Methods:

• Developed Bayesian regression models for magnitude data with random effects.
• Implemented a joint modeling strategy to handle informative dropout.
• Conducted two numerical simulation studies to validate the proposed methods.
• Applied the models to analyze diaphragm thickness data from ICU patients.

Main Results:

• The proposed method demonstrates good estimation accuracy for magnitude data.
• Joint models effectively reduce bias caused by missing data.
• Simulations confirm the validity and performance of the developed techniques.

Conclusions:

• The novel Bayesian approach effectively analyzes magnitude data with random effects and handles informative dropout.
• The joint modeling strategy successfully addresses missing data challenges in repeated measures.
• The models provide a robust framework for analyzing complex biomedical data, as shown in the ICU patient study.