Study Protocol: the Maternal Health Multilevel Intervention for Racial Equity (maternal Health MIRACLE) Project.

Purpose: To test the effectiveness and cost-effectiveness of a multilevel intervention for population-level African American (AA) severe maternal morbidity and mortality. Background: Severe maternal morbidity and mortality in the U.S. disproportionately affect AA women. Inequities occur at many levels, including community, provider, and health system levels. Design: Intervention. Throughout the two intervention counties, we will expand access to enhanced prenatal care services using telehealth and flexible scheduling (community level), provide actionable maternal health-focused anti-racism training (provider level), and implement equity-focused community care maternal safety bundles (health system level). Partnership. Interventions were developed/co-developed by intervention county partners, including AA women, enhanced prenatal care staff, and health providers. For equity, 46% of project direct cost dollars go to our partners. Most study investigators are female (75%) and/or AA (38%). Partners are overwhelmingly AA women. Sample, measures, analyses. We use a quasi-experimental difference-in-differences with propensity scores approach to compare pre (2016-2019) to post (2022-2025) changes in outcomes for Medicaid-insured women in intervention counties to similar women in the other Michigan, USA, counties. The sample includes all Medicaid-insured deliveries in Michigan during these years (n similar to 540,000), with women observed during pregnancy, at birth, and up to 1 year postpartum. Measures are taken from a linked dataset that includes Medicaid claims and vital records. Conclusion: This study is among the first to examine effects of any multilevel intervention on AA severe maternal morbidity and mortality. It features a rigorous quasi-experimental design, multilevel multi-partner county-wide interventions developed by community partners, and assessment of intervention effects using population-level data.