Challenges in estimating effects of hypothetical interventions on resources patterned by structural racism: An example in a rural North Carolina Medicaid population.

Structural racism has likely shaped the geographic distribution and resource allocation of rural populations and marginalized racial/ethnic groups. We sought to 1) quantify disparities in severe maternal morbidity (SMM) and distributions of resources by race and racial composition of county, and 2) determine whether a hypothetical intervention on resources would reduce racial disparities in SMM, using linked birth certificates and claims from Medicaid beneficiaries giving birth from 2014-2019 in rural North Carolina (61 rural counties, 77,665 births). We used ratio of mediator probability weights to enact a hypothetical intervention that would equalize distributions of pregnancy care provider ratios and obstetric units across race and racial composition of county. Despite observed disparities in the distributions of resources and SMM, we were unable to demonstrate that the hypothetical interventions would reduce SMM. This may be due to a lack of common support - marginalized groups never experienced the more optimal extremes of the healthcare resources distributions that privileged groups did. Our findings may have implications for the use of causal inference methods for addressing health disparities more broadly: if distributions of resources among privileged groups are outside those that marginalized groups experience, hypothetical interventions on these distributions cannot be emulated with data.