Mixed-effects models for health care longitudinal data with an informative visiting process: A Monte Carlo simulation...

Area of Science:

• Biostatistics
• Health Informatics
• Clinical Research Methodology

Background:

• Electronic health records (EHRs) are increasingly vital for medical research, enabling detailed clinical questions.
• However, EHR data presents methodological challenges, notably informative observation times correlated with disease severity.
• Traditional longitudinal data analysis methods often assume independence, which is violated in healthcare data.

Purpose of the Study:

• To compare analytical approaches for informative observation processes in longitudinal data.
• To formalize a joint model for observation and longitudinal outcomes within an extended framework.
• To assess the unbiasedness of different analytical methods using Monte Carlo simulations.

Main Methods:

• Monte Carlo simulation to compare various analytical approaches for informative visiting processes.
• Formalization of a joint model integrating the observation process and longitudinal outcome.
• Application of the joint model to pragmatic trial data for enhanced chronic kidney disease care.

Main Results:

• The study evaluates the performance of different analytical methods in handling informative observation times.
• The proposed joint modeling framework is shown to provide unbiased results for longitudinal data analysis.
• User-friendly software is introduced for fitting the joint model.

Conclusions:

• Informative observation times in EHR data require specialized analytical methods beyond traditional approaches.
• Joint modeling offers a robust framework for analyzing longitudinal outcomes when observation processes are informative.
• The developed methods and software can improve the analysis of clinical research data, enhancing insights from EHRs.