Application of the Pattern-Mixture Latent Trajectory Model in an Epidemiological Study with Non-Ignorable Missingness

Area of Science:

• Statistics
• Epidemiology
• Biostatistics

Background:

• Longitudinal studies are susceptible to bias from non-ignorable missing data.
• Traditional methods struggle with complex missing data patterns.
• Pattern mixture models offer a potential solution for handling such data.

Purpose of the Study:

• To combine pattern mixture models with latent trajectory analysis for robust handling of non-ignorable missing data.
• To provide a practical and implementable statistical approach for longitudinal data analysis.
• To estimate longitudinal trajectories in the presence of complex missing data patterns.

Main Methods:

• Utilized a pattern mixture model integrated with latent trajectory analysis (SAS TRAJ procedure).
• Employed a stochastic process to categorize missing data patterns into latent groups with distinct outcome trajectories.
• Estimated memory test trajectories over 12 years using data from a prospective dementia study, conditioning missing data on survival.

Main Results:

• Successfully categorized complex missing data patterns into meaningful latent groups.
• Demonstrated the ability to share information between patterns with varying amounts of data.
• Provided reliable estimation of longitudinal memory test trajectories over a 12-year period.

Conclusions:

• The combined pattern mixture and latent trajectory model offers a practical solution for non-ignorable missing data in longitudinal studies.
• This approach enhances the analysis of complex datasets, such as those in epidemiological research.
• The method is easily implementable using common statistical software, promoting wider application.