LATEBREAKER
Causal Inference
Assessing the contribution of obesity to trends in diabetes across U.S. birth cohorts using a causal decomposition approach Rafeya Raquib* Andrew Stokes Meghan Podolsky
Obesity prevalence in the U.S. has increased since at least 1960. Environmental factors have been cited as drivers of this trend, but the relative effect of exposure in different birth cohorts has yet to be established. We used a counterfactual causal decomposition model combining the age-period-cohort approach with g-computation to examine how changing body mass index (BMI) distributions across birth cohorts impacted diabetes prevalence. Using data from the National Health and Nutrition Examination Survey (NHANES) III (1988-1994) and continuous waves (1999-2020) (N=33,386), we compared the observed prevalence of diabetes in the natural course to a counterfactual scenario where all one-year synthetic birth cohorts had the same distribution of BMI as the 1930 birth cohort, taking into account age, sex, race, ethnicity, and education. Diabetes status was established as having measured Hemoglobin A1c (HbA1c) ≥6.5% or HbA1c <6.5% with self-reported diabetes treatment. BMI was calculated as a lagged variable that combined weight 10-years prior to survey with height at survey to reduce the risk of reverse causality between the mediator and outcome variables. Holding age (50) and period (2008) constant, the prevalence of obesity (BMI 30+) increased from 13.4% (7.6-18.1) in the 1930 birth cohort to 34.9% (22.8-56.5) in the 1975 birth cohort. We found that diabetes prevalence would be on average 2.0 (1.0-3.9) percentage points lower for the 1960 birth cohort and 4.2 (2.7-6.7) percentage points lower for the 1970 birth cohort with the BMI distribution of the 1930 birth cohort. These results suggest that increasing obesity rates contributed significantly to rising diabetes prevalence across U.S. birth cohorts. Additional research is needed to elucidate the factors contributing to increases in obesity across birth cohorts, including early life social factors and accumulated exposure to obesogenic factors over the life course.
