Skip to content

Abstract Search

Aging

Applying a Causal Decomposition Approach to Quantify the Role of Education in Cohort Differences in Dementia Prevalence in the US Ruijia Chen* Ruijia Chen Rafeya Raquib Jingxuan Wang Sarah Ackley Jennifer Weuve warre046@umn.edu Maria Glymour Andrew C. Stokes

Introduction: While existing population health research suggests a decline in dementia prevalence in the United States, the explanations behind such trends remain unclear. This study applied a causal decomposition approach to quantify the role of education in birth cohort differences in dementia prevalence. 

Method: We included participants from the 2000 to 2016 Health and Retirement Study (HRS) who were 70 years and older (N=17,710). We defined dementia status using an existing algorithm with similar sensitivity and specificity across racial/ethnic groups. We categorized participants into four birth cohorts: Asset and Health Dynamics Among the Oldest Old (AHEAD) <1924; Children of the Depression (CODA) 1924-1930; HRS 1931-1941; and War Babies (WB) 1942-1947. We applied a counterfactual causal decomposition approach that incorporates the age-period-cohort (APC) model with g-computation. We compared the observed prevalence of dementia in the natural course to a counterfactual scenario where all birth cohorts had the same distribution of education of the birth cohort with the lowest prevalence. Models adjusted for age, sex/gender, and race and ethnicity. 

Result: Across the four cohorts, the WB 1942-47 cohort had the lowest prevalence of dementia compared with other cohorts, independent of age and period effects. In the counterfactual of all cohorts having the same distribution of education as the WB 1942-47 cohort, dementia prevalence would be 20.0% (95% CI: 15.3-25.3) lower for AHEAD, 15.0% (95% CI: 9.7-20.6) lower for CODA, and 7.5% (95% CI: 2.3-13.0) lower for HRS (see Figure). 

Conclusion: We found marked declines in dementia prevalence across U.S. birth cohorts. Increases in educational attainment across cohorts may have substantially contributed to these declines.