Combining Longitudinal Data From Different Cohorts to Examine the Life-Course Trajectory

Area of Science:

• Epidemiology
• Biostatistics
• Developmental Science

Background:

• Longitudinal data are crucial for understanding individual aging processes.
• Single cohort studies rarely cover an entire human lifespan.
• Combining data from multiple cohorts presents harmonization and data management challenges.

Purpose of the Study:

• To detail methods for constructing life-course trajectories from diverse, independent cohort studies.
• To address challenges including data harmonization, systematically missing data, and model selection across heterogeneous populations.
• To illustrate the approach by examining factors influencing children's weight trajectories.

Main Methods:

• Developed a framework for data harmonization across independent prospective cohort studies.
• Employed statistical modeling to handle systematically missing data and differing age ranges/measurement schedules.
• Utilized imputation methods for individual-level covariates within a multilevel nonlinear growth trajectory model.

Main Results:

• Successfully generated life-course trajectories by integrating data from five prospective cohort studies across Belarus and the UK.
• Demonstrated the ability to model trajectories over extensive age ranges (birth to early adulthood).
• Enabled direct comparisons of life-course segments across different regions and time periods.

Conclusions:

• The described methodology enables robust life-course trajectory analysis using multiple, independent cohort studies.
• This approach facilitates cross-study information sharing and comparative analyses of developmental trajectories.
• The methods provide a powerful tool for understanding long-term health and developmental patterns.