Comprehensive Analysis of Asynchronous Binary Variable Associations in Longitudinal End-of-Life Studies

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Epidemiology

Background:

• Understanding dynamic relationships between binary variables over time is vital in biomedical research.
• Existing methods face challenges with variables measured at different times and missing data.

Purpose of the Study:

• To develop and validate novel statistical measures for longitudinal bivariate associations.
• To address complexities of time-varying effects and missing data in longitudinal studies.
• To apply the methodology to real-world data for insights into long-term health trends.

Main Methods:

• Introduced bivariate time-varying odds ratio and relative risk measures.
• Developed a nonparametric approach for longitudinal samples with varying measurement timelines.
• Implemented a missing data model with inverse-probability weighting validated by simulations.

Main Results:

• The nonparametric approach effectively handles concurrent and nonconcurrent sampling.
• Inverse-probability weighting successfully corrected biases caused by missing data.
• Analysis of the Framingham Heart Study revealed temporal changes in mother-daughter hypertension association over 45 years.

Conclusions:

• The novel methodology provides robust tools for analyzing dynamic bivariate associations in longitudinal data.
• The approach is versatile, applicable to various sampling schemes and effectively handles missing data.
• The Framingham Heart Study application highlights the method's utility in understanding familial health trends over extended periods.