Flexible age-period-cohort modelling illustrated using obesity prevalence data

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Generalized linear models (GLMs) with non-linear functions for age, period, and cohort effects enable interpretable and reliable estimation.
• These methods were applied to Australian women's obesity prevalence data from two distinct study designs.

Purpose of the Study:

• To demonstrate the application of GLMs for estimating and visualizing age, period, and cohort effects on obesity prevalence.
• To analyze obesity trends in Australian women using data from national surveys and a longitudinal study.

Main Methods:

• Utilized data from Australian National Health Surveys (1995-2017/18) and the Australian Longitudinal Study on Women's Health (starting 1996).
• Employed age-period-cohort analysis using generalized linear models with splines for non-linear effects.

Main Results:

• Both datasets showed similar obesity patterns: increasing with age until middle age, slight increase across surveys, and steady increase with birth year until the 1960s, followed by acceleration.
• Younger generations of Australian women exhibit higher obesity prevalence at younger ages.

Conclusions:

• The GLM approach provides accessible and interpretable estimation and visualization of age, period, and cohort effects.
• Incorporating cohort effects alongside age and period is crucial for effective obesity prevention strategies in Australian women.