Confounder related

data analysis causal inference
1

I am working with an Epidemiologist and she said

“Why deprivation index, it is only a confounder” ? I am trying to understand what this means and here’s the context behind this statement.

The study objective is to estimate the effects of Race/Ethnicity on Lead content in the mothers blood. So exposure here is Race/Ethnicity (A), outcome is Lead content in the blood (y).

The scientist involved in this study suggested including deprivation index as a confounder. Deprivation index is a quantitative measure indicating how deprived a neighborhood is ? For example if I am staying in Boston, MA then the deprivation index = 80 indicates that Boston, MA is good neighborhood, a deprivation index = 10 , Flint Michigan indicates that this is a bad neighborhood . So this is the deprivation index variable.

The Epidemiologist said, why deprivation index its only a confounder ? I have no idea what this means, I need some help understanding the meaning behind this statement. Arent we supposed to include confounders that are related to exposure in our model ?

Thanks.

-Sudhi

2

In your example, I don’t think it would be wise to consider deprivation index as a confounder of the ‘effect’ of ethnicity on lead exposure.

It seems more likely that deprivation acts as a mediator. I.e., people with certain ethnicities live in more deprived areas (usually due past or present racism and racist policies), which in turn leads them to have greater exposure to environmental hazards. Thus, for the primary analysis, you shouldn’t condition on deprivation as this would ‘knock out’ the part of the disparities caused by inequalities in deprivation between ethnicities.

VanderWeele and Robinson have a nice paper where they describe ways to interrogate such mediating factors: On the causal interpretation of race in regressions adjusting for confounding and mediating variables - PubMed

See also Kaufman and Cooper for some more discussion of these issues: https://academic.oup.com/aje/article/154/4/291/61865

4 Likes
3

@Harry_TB , Excellent suggestion. I will go through this paper.

Thanks.