Bayesian hierarchical piecewise regression models: a tool to detect trajectory divergence between groups in long-term...

Area of Science:

• Longitudinal data analysis
• Biostatistics
• Developmental trajectories

Background:

• Bayesian hierarchical piecewise regression (BHPR) is a novel method for analyzing longitudinal data with distinct developmental phases.
• Understanding life-course risk factor development is key for early interventions, especially for health outcomes like type 2 diabetes mellitus (T2DM).
• Previous methods may bias the estimated age of trajectory divergence between groups.

Purpose of the Study:

• To demonstrate BHPR for estimating the age of trajectory divergence in longitudinal data.
• To model divergence in body mass index (BMI) trajectories between individuals with and without T2DM.

Main Methods:

• Utilized BHPR to estimate the point and credible interval for the age of trajectory divergence.
• Applied the method to body mass index (BMI) data from the Cardiovascular Risk in Young Finns Study (YFS).
• Compared BHPR with traditional methods using simulations to assess bias and sample size effects.

Main Results:

• Estimated BMI trajectories diverged at age 16 years for males and 21 years for females with T2DM.
• Identified a critical window for weight management interventions before natural decreases in BMI growth rate.
• BHPR provided unbiased divergence time estimates, unlike traditional methods sensitive to sample size.

Conclusions:

• BHPR is effective for modeling non-linear longitudinal outcomes and identifying divergence points in risk factor trajectories.
• The method is suitable for unbalanced longitudinal data with distinct developmental phases.
• BHPR aids in understanding life-course trajectories for targeted interventions.