Educational Attainment and Cohort Differences in Dementia Prevalence: A Counterfactual Decomposition Analysis

BACKGROUND: It remains unclear why age-specific dementia prevalence has declined over recent decades. Improvements in educational attainment across cohorts are one potential explanation. We estimated the reduction in dementia prevalence expected if earlier cohorts had attained the educational distribution of the most recent cohort. METHODS: We analyzed 17,762 participants (82,856 person-waves) from the 2000-2020 Health and Retirement Study (HRS) and grouped participants into four cohorts: Assets and Health Dynamics Among the Oldest Old (AHEAD; born 1900-1923); Children of the Depression (CODA; 1924-1930); HRS (1931-1941); and War Babies (WB; 1942-1947). We incorporated g-computation with age-period-cohort analyses to compare natural-course dementia prevalence with counterfactual scenarios in which earlier cohorts were assigned WB's education distribution. RESULTS: If earlier cohorts had the same education distribution as the WB cohort, dementia prevalence would be lower across ages and periods. At age 70 in 2010, prevalence based on the Langa-Weir classification of dementia would be reduced by 14% (95% CI: 11% -16%) in HRS, 21% (95% CI: 19% - 24%) in CODA, and 26% (95% CI: 23% - 29%) in AHEAD. Equalizing educational attainment would close 0.7 percentage points (95% CI: 0.5, 0.8) of the HRS-WB difference, 1.2 (95% CI: 0.9, 1.4) of the CODA-WB difference, and 1.6 (95% CI: 1.3, 1.9) of the AHEAD-WB difference. Reductions in dementia prevalence based on the Gianattasio-Power dementia ascertainment algorithm were smaller. Large reductions were observed within sex/gender and race and ethnicity subgroups. CONCLUSIONS: Educational attainment contributes to cohort differences in dementia prevalence overall and across demographic groups.