AbstractBackgroundIdentifying mediators between education and later‐life cognition is an important step in developing prevention measures. We evaluate everyday and major lifetime discrimination as potential mediators of the impact of education on later‐life cognition. To address bias that might occur due to either time‐varying confounding or unreliable measurement of discrimination, we use a novel parametric g‐estimation approach.MethodParticipants in the Study of Healthy Aging in African Americans (STAR, n = 764, ages 52‐90) self‐reported education (high school [HS] degree, general educational development [GED], or less [ref]; bachelor’s degree; or graduate degree). Validated measures of everyday discrimination (9‐items assessing frequency with 6‐point Likert responses) and major lifetime discrimination (9‐items, dichotomous yes‐no), were conceptualized as latent variables (Table 1). The observed cognitive outcome was the average of the verbal episodic memory, semantic memory, and executive function Spanish and English Neuropsychological Assessment Scales (SENAS). We used linear regression to estimate the total effect of education on later‐life cognition and parametric g‐estimation with latent discrimination measures to estimate the controlled direct effect (CDE) of education on cognition if the discrimination measures were set to their sample averages. Models were adjusted for measurement error in the discrimination measures using Fuller’s Method. Standard errors were bootstrapped.ResultMost participants were female (68.32%) and the average age was 69.0 (Table 2). Both bachelor’s (b=0.41[0.21,0.60]) and graduate (b=0.43[0.22,0.64]) degrees were associated with better cognition compared to HS/GED or less (Table 3). More education was associated with higher levels of major lifetime discrimination. The CDEs of a bachelor’s and graduate degree, if everyone had the average level of everyday and major life discrimination, were 0.32[0.06,0.61] and 0.55[0.24,0.82], respectively (Table 3 and as a function of different fixed averages in Figure 2).ConclusionThis study provides a novel demonstration of parametric g‐estimation methods that correct for measurement error with latent variables and allow for possible feedback between education and early and late‐life confounders. Accounting for exposure to discrimination slightly increased the magnitude of the association between graduate education and cognition in this older sample of Black American adults, because education was associated with greater exposure to major life discrimination.