Does race have confounders?

When estimating the effects of race/racism, many standard covariates like education, income and comorbidities play the role of mediator rather than confounder. Inappropriately adjusting for mediators blocks the causal pathway on which the mediator lies, yielding the controlled direct effect (CDE), which is biased for the total effect.1 In particular, if the indirect effect and CDE are in the same direction the CDE will underestimate the total effect.2 This raises the question of whether confounding of racial effect exists at all and what variables could be used to adjust for underlying differences.

The life-course framework may present a solution. Figure 1 provides a causal diagram for assessing the relationship between child’s race and health outcome, potentially in adulthood, including intergenerational variables. Figure 1 identifies the root confounding variable as parent’s race; however, adjusting for parental race may violate the positivity assumption and induce…

Read the full article ›