Considerations When Accounting for Race and Ethnicity in Studies of Poverty and Neurodevelopment.

Objective: Poverty and systemic racism are intertwined. Children of marginalized racial and ethnic identities experience higher levels of poverty and adverse psychiatric outcomes. Thus, in models of poverty and neurodevelopment, race and ethnicity, as proxies for exposure to systemic disadvantage, are regularly considered confounders. Recently, however, some researchers have claimed that using race and ethnicity as confounders is statistically dubious, and potentially socially damaging. Instead, they argue for the use of variables measuring other social determinants of health (SDoH). We explore this approach herein.

Method: Data are from 7,836 children 10 years of age in the Adolescent Brain Cognitive Development Study (ABCD Study). We fit mixed regression models for the association of household poverty measures with psychiatric symptoms, magnetic resonance imaging (MRI)-derived cortical measures, and cognition with and without (1) race and ethnicity adjustment, (2) poverty-by-race and ethnicity interaction terms, and (3) alternative SDoH variables. Propensity-based weights were used to calibrate the sample to key US demographics.

Results: For psychiatric and cognitive outcomes, poverty-outcome relationships differed across racial and ethnic groups (interaction of poverty by race and ethnicity, p < .05). For MRI-derived outcomes, adjusting for race and ethnicity changed the estimate of the impact of poverty. Alternative SDoH adjustment could not fully account for the impact of race and ethnicity on the associations explored.

Conclusion: Poverty and both race and ethnicity combine to influence neurodevelopment. Results suggest that the effects of poverty are generally inconsistent across race and ethnicity, which supports prior research demonstrating the nonequivalence of SDoH indicators by race and ethnicity. Studies exploring these relationships should assess the interaction between poverty and race and ethnicity and/or should stratify when appropriate. Replacing race and ethnicity with alternative SDoH may induce bias.