Compliance in an anti-hypertension trial: a latent process model for binary longitudinal data

Area of Science:

• Statistics
• Biostatistics
• Longitudinal Data Analysis

Background:

• Generalized Estimating Equations (GEE) are commonly used for binary longitudinal data.
• GEE can be challenging with long time series and non-common observation times per subject.
• A need exists for alternative methods suitable for complex longitudinal datasets.

Purpose of the Study:

• To propose a novel statistical method for inference on binary longitudinal data.
• To provide a practicable alternative to GEE for data with long time series and irregular observation schedules.
• To address limitations of existing methods in handling complex longitudinal data structures.

Main Methods:

• The proposed method models binary responses based on an unobserved stationary continuous process.
• Conditional independence of binary responses is assumed given the underlying process realization.
• An algorithm for parameter estimation was developed and validated through simulation.

Main Results:

• The new method is shown to be effective through simulation studies.
• The methodology was successfully applied to real-world clinical trial data.
• The approach offers a viable alternative for analyzing complex binary longitudinal data.

Conclusions:

• The proposed method provides a practical and effective alternative to GEE for binary longitudinal data analysis.
• This approach is particularly useful for datasets with long time series and non-uniform observation times.
• The study demonstrates the utility of modeling underlying continuous processes for longitudinal binary outcomes.