Abstract
Introduction:
Prior statewide research has documented the positive associations between Black residential segregation and severe maternal morbidity (SMM). However, little is known about the relationship of composite social determinants of health (SDoH) and multiple Black residential segregation measures with SMM in the national scope. We estimate the associations of SDoH Index and Black residential segregation with SMM.
Methods:
This retrospective cohort study of 84,421 nationwide childbirths from March 1, 2020, through May 31, 2022, linked county-level SDoH and Census residential segregation data with the National COVID Cohort Collaborative Data Enclave. SMM during pregnancy through 42 days postpartum was defined by any of the 20 Centers for Disease Control and Prevention-documented SMM indicators (except blood transfusion), identified using the Observational Medical Outcomes Partnership Clinical Data Model.
Results:
SMM rates per 10,000 childbirths showed geographic disparities ranging from 297.1 to 309.5 in the high versus low SDoH counties and from 268.2 to 395.1 in the low versus high Black segregation counties. In the multilevel log-binomial regression model, higher SDoH were significantly associated with lower risks of SMM, with the lowest risks in the highest SDoH tercile (relative risk [RR] = 0.81; 95% confidence interval [CI]: 0.69–0.94; p = 0.008). Black residential segregation was associated with increased SMM risk (RR = 1.30; 95% CI: 1.12–1.50; p < 0.001).
Conclusion:
SMM rates were disproportionately higher among birthing people in low SDoH (socially disadvantaged) and Black segregation communities, highlighting the importance of addressing multifaceted SDoH and residential segregation in the efforts to overcome maternal health disparities in the United States.
Introduction
Severe maternal morbidity (SMM)—potentially life-threatening conditions that manifest during pregnancy, at delivery, or postpartum—results in significant short- or long-term consequences for a woman’s health.1,2 These complications result in extended hospital stays, high medical costs, and could be fatal if not adequately treated.1,3 In the United States, SMM rates tripled from 49.5 in 1993 to 146.6 cases per 10,000 childbirths in 2014, 4 affecting over 60,000 women annually; in 2021, these outcomes have reached a new high. 1 These rates are unevenly distributed across sociodemographic groups, disproportionately affecting women of color. 5 Black women are two to three times more likely to experience SMM and mortality than their White counterparts, a disparity that has persisted over a decade,5,6 and the COVID-19 pandemic has exacerbated these Black–White disparities. 7
During the COVID-19 pandemic, racial and ethnic minority and socioeconomically disadvantaged communities in the United States have experienced increased SARS-CoV-2 infection rates, worse COVID-19 outcomes,8–10 and worse maternal outcomes.8,11,12 Birthing individuals in these communities may be underserved by services that support social distancing or healthy lifestyles and help address the social determinants of health (SDoH), such as health care access, education, housing, transportation, nutrition assistance, income supports, technology, employment, and social services (e.g., childcare). 13 The systemic SDoH inequities hinder marginalized people from anticipating, confronting, advocating, repairing, and recovering from the public health emergencies, leading to the risk of maternal morbidity and mortality, as well as severe illness from COVID-19.14,15 Indeed, the overall maternal mortality increased substantially from 658 deaths in 2018 to 1,178 deaths in 2021, with substantial variations by race and ethnicity. 14 Among Black birthing people, maternal mortality increased from 37.3 per 100,000 live births in 2018 to 68.9 in 2021, while White population saw an increase from 14.9 to 26.1 per 100,000 live births. 14 Many of these inequities stemmed from historical structural racism, perpetuating inequality. For example, residential segregation, the physical separation of groups based on the social construct of race, was fundamentally designed to prevent social interaction between White and Black individuals. 16 Residential segregation has been widely documented to associate with higher rates of SMM and adverse birth outcomes in statewide or city-level studies.7,17,18 The associations of racial residential segregation and overall SDoH level with SMM in the national scope have not been explicitly examined during the COVID-19 pandemic.
To ensure equitable policy response and recovery efforts for COVID-19 and other public health emergencies, it is crucial to understand county-level SDoH levels and identify maternal groups most at risk. This study leveraged national electronic health record (EHR) data from the National COVID Cohort Collaborative (N3C) Data Enclave, with two objectives as follows: (1) to examine maternal populations living in counties with low SDoH and with high Black residential segregation and (2) to evaluate the associations of social contextual factors (county-level SDoH Index, Black residential segregation) with SMM.
Methods
Data sources and study design
This retrospective cohort study integrated the 2020–2022 N3C Data Enclave with the 2019 SDoH data and the 2015–2019 Census population estimates data. The N3C data were used to construct a retrospective cohort of 84,421 childbirths to birthing people with residential location information across 1,371 counties in 50 U.S. states (DC did not provide maternal county information) from March 1, 2020, through May 31, 2021 (during the COVID-19 pandemic). The N3C data include a multicenter EHR data repository, contributed by 92 health systems and freestanding institutes in all 50 U.S. states. N3C enables data sharing, computable phenotypes, and collaborative data mining by harmonizing EHR data of diverse standards using the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM). Full details about the N3C data have been published previously. 19 The county-level SDoH Index ranging from 0 to 1 (higher indicates greater access to resources) was captured from the N3C, which mirrors the Centers for Disease Control and Prevention (CDC) Healthy People 2020 framework on the following five domains: neighborhood and built environment, health and health care, social and community context, education, and economic stability (Supplementary Appendix A1).20,21 Using iterative structural equation techniques, Boston University Sharecare’s team conducted principal component analysis to reduce the number of items retained while maximizing variance of county-level well-being scores explained, forcing five components.20,21 This iterative process resulted in the SDoH Index that included 59 items. 21 The 2015–2019 American Community Survey (ACS) U.S. Census tract-level racial and ethnic compositions were used to calculate the county-level residential segregation measure.
Both University of South Carolina Institutional Review Board and the N3C Data Access Committee approved this study. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline for cohort studies.
Study sample
Figure 1 illustrates the sample selection from the N3C data. The study cohort includes 84,421 women who were aged 15–49 years and delivered a live birth from March 1, 2020, through May 31, 2021, with residential county or ZIP code information in the EHR data. In the case when a residential county was missing but the ZIP code was not, ZIP codes that cross multiple counties were assigned to a county with the largest number of residences to merge with county-level social contextual factors. Birthing records were captured from several components in the EHR given the nature of labor and delivery services: condition occurrence (for diagnosis records such as labor), measurement (clinical laboratory orders and results, e.g., isoimmunization screening), procedure (surgical operations, e.g., cesarean, episiotomy, induction), and observation (clinical notes, e.g., intrapartum complications). All records were captured using the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM). 22 Procedures for constructing the cohort and evaluating outcomes have been published elsewhere. 23
Study sample selection diagram.
Measures
This study tested two key social contextual exposures. (1) County-level SDoH Index ranging from 0 to 1 (higher is better) was captured from the N3C. This composite county-level SDoH Index, established by Sharecare and Boston University School of Public Health funded by CDC, has been documented in prior literature.20,21 The SDoH Index mirrors the CDC Healthy People 2030 framework on the five following domains: neighborhood and built environment, health and health care, social and community context, education, and economic stability.20,21 The overall SDoH Index was categorized into tercile groups: low (0.33–0.55), medium (0.56–0.60), and high (≥0.60). (2) Black residential segregation was calculated by isolation index using non-Hispanic Black and White as population groups and then categorizing U.S. Census tracts in a county as low (<40%), medium (40 − 59%), and high (≥60%).7,24 Higher scores of Black versus White isolation index indicate more extensive isolation of Black residents across U.S. Census tracts in the county, and lower scores suggest less isolation or more Black versus White exposure, compared with a county where all other covariates are held at their mean values in terms of interracial isolation.
The primary outcome is SMM from pregnancy conception through 42 days postpartum. Birthing people experiencing SMM during this time frame were defined as having any diagnoses or procedures of the 20 CDC-documented SMM indicators (except blood transfusion)—in alignment with the Agency for Healthcare Research and Quality guidance 25 —identified using the OMOP CDM from the N3C. The pregnancy conception date was calculated from the date of childbirth to gestational weeks at childbirth × 7. Each SMM concept set and its description in the OMOP CDM were validated by three coauthors (C.L., Y.S., and P.H.) and one obstetrician (B.A.C.).
Covariates were chosen based on known or plausible associations with community SDoH and SMM.7,26 These covariates include SARS-Cov-2 infection, maternal race and ethnicity, age group, clinical conditions (prepregnancy overweight or obesity vs. normal weight, preexisting or gestational diabetes, preexisting or pregnancy-induced hypertension, depression and anxiety during pregnancy, plurality, gestational age at childbirth), smoking—including e-cigarette or other tobacco use—during pregnancy, mode of delivery, urban/rural residence, and residence census region.
Statistical analysis
Descriptive statistics and bivariate analyses of categorical variables were conducted using Pearson’s chi-square tests. To assess the associations of county-level social context (i.e., SDoH level and Black residential segregation) with SMM, we modeled multilevel log-binomial regressions for each exposure separately, controlling for the aforementioned covariates. To further examine whether each social context exposure moderated racial/ethnic disparities in SMM, as a sensitivity analysis, we also included two-way interaction terms in the models (i.e., social context × race/ethnicity). The SMM disparities by each social context were assessed through relative risk (RR) and its 95% confidence interval (CI).
All statistical analyses were undertaken within the N3C Enclave using R v.4.0.2 (R Foundation for Statistical Computing), with statistical significance set at p < 0.05 (two-tailed). Data were constructed and analyzed from November 2021 to January 2023.
Results
Study cohort
Of 84,421 childbirths, 43,768 (51.8%) were to birthing people in low-segregated Black communities (mean [standard deviation or SD] age, 28.1 [5.7] years; 20,523 [21.6%] Black, 5,126 [5.4%] Hispanic, 62,690 [65.9%] White), 48,172 (28.9%) to those in medium-segregated communities (mean [SD] age, 28.1 [5.8] years; 17,863 [37.1%] Black, 1,899 [3.9%] Hispanic, 25,129 [52.2%] White), and 23,521 (14.1%) to women in high-segregated communities (12,880 [54.8%] Black, 782 [3.3%] Hispanic, 7,988 [34.0%] White; Supplementary Appendix A2).
Sociodemographic and clinical characteristics by social contexts
Compared with birthing people in high SDoH counties, those in low SDoH areas were more likely to be younger than 25 years, to be diagnosed with obesity before pregnancy, to use tobacco during pregnancy, to have diabetes, hypertension, and/or depression and anxiety, to give preterm birth, to undergo cesarean delivery, and to reside in rural ZIP codes (Table 1). These maternal populations living in low SDoH areas were similar to those in high-segregation Black counties, except for race, depression and anxiety during pregnancy, and rural residence. Compared with those in low-segregated counties, birthing people living in high-segregated communities were less likely to have depression and anxiety during pregnancy and to live in urban America (Supplementary Appendix A3).
Maternal Characteristics by Residential County Social Determinants of Health Index, March 2020–May 2021
p-Values were calculated from Pearson’s chi-square tests for categorical variables.
By maternal race and ethnicity, non-Hispanic Black people were more likely to live in counties of high SDoH (high: 21.5% vs. low: 14.8%; Table 1) and high segregation (high: 34.0% vs. 10.1%; Supplementary Appendix A3) than non-Hispanic White people. Hispanic, Asian, or other Pacific Islanders, and other non-Hispanic groups were also more likely to reside in high-segregated counties, compared with White people.
SMM by social contexts
During the COVID-19 pandemic, unadjusted rates of SMM measures were highest among mothers who reside in low SDoH and/or high-segregation areas (Fig. 2). SMM rates per 10,000 childbirths showed geographic disparities ranging from 297.1 in the highest SDoH counties to 309.5 in the lowest SDoH counties and from 268.2 in the low-segregation counties to 395.1 in the high-segregation counties.
Severe maternal morbidity rates per 10,000 childbirths by social determinants of health and Black residential segregation categories across maternal race and ethnicity groups relative to the non-Hispanic White group. Blue markers represent each non-White group; severe maternal morbidity rates per 10,000 childbirths among the non-Hispanic White group in a social determinant of health and Black residential segregation category are illustrated in orange triangles. All race groups were among non-Hispanic individuals, including White, Black, Asian/non-Hispanic other Pacific Islander (NHOPI), and other. Detailed data on the number of cases and rates of severe maternal morbidity are available in the Supplementary Appendix A4.
Black birthing people regardless of residential segregation had consistently higher SMM rates (734 cases [440.6 per 10,000 childbirths]) than their White counterparts (953 cases [230.4 per 10,000 childbirths]; p < 0.001; Fig. 2). Black and Hispanic birthing people living in high-segregated communities had higher probabilities of SMM than their counterparts living in low-segregated communities. Among Black birthing people, those in low SDoH counties had significantly higher SMM rates than those in high SDoH (528.6 vs. 392.1 per 10,000; p < 0.001). Black people in high-segregation versus low-segregation counties had higher SMM rates (521.8 vs. 352.7 per 10,000; p < 0.001). Similarly, SMM rates for Hispanic birthing people were higher in high-segregation (380.8 per 10,000) versus low-segregation communities (300.1 per 10,000; p < 0.001).
On average, higher SDoH were significantly associated with lower risks of SMM, with the lowest risks in the highest SDoH tercile. When comparing the highest tercile of SDoH with the lowest tercile of SDoH, the adjusted RR [aRR] was 0.81 (95% CI: 0.69–0.94; p = 0.008; Table 2). Black residential segregation was associated with increased SMM risk (aRR = 1.30; 95% CI: 1.12–1.50; p < 0.001).
Log-Binomial Models on Maternal and County Characteristics Associated with Severe Maternal Morbidity During the COVID-19 Pandemic
Significant estimates are bold. Relative risks and 95% CIs from mixed-effects log-binomial regressions with state as a random effect. Multilevel log-binomial regression was conducted for Social Determinants of Health Index and for Black residential segregation separately.
CI, confidence interval; Ref., reference; SMM, severe maternal morbidity.
Racial and ethnic disparities in SMM by social contexts
In the sensitivity analysis, racial and ethnic disparities in SMM were similar by SDoH tercile levels and segregation level. The non-Hispanic other group in high SDoH or low-segregated communities also had no significant different SMM risks from non-Hispanic White people; however, birthing people of the non-Hispanic other group in low SDoH communities and high-segregation communities had increased risks, yielding higher disparities in SMM when compared with their non-Hispanic White counterparts (low vs. high SDoH: aRR = 1.64; 95% CI: 1.11–2.17; high vs. low isolation: aRR = 1.85; 95% CI: 1.11–3.09).
Discussion
This national cohort study using the EHR data from the N3C Data Enclave of birthing records during the COVID-19 pandemic found that SMM risk was highest among birthing individuals in communities with low SDoH and with high Black segregation. Individuals residing in low versus high SDoH communities were more likely to be overweight or obese before pregnancy, have tobacco use during pregnancy, and with preexisting or gestational diabetes, hypertensive disorder, and/or depression and anxiety during pregnancy. In addition, individuals from low SDoH communities had a disproportionately higher share of rural residents, thereby more non-Hispanic White people in low SDoH than in high SDoH counties. Yet, after controlling for these clinical factors, race, and ethnicity, both residential community SDoH and Black residential segregation were independently associated with elevated SMM risk during the pandemic.
Our findings are consistent with prior statewide or city-level research, which emphasizes the association between SDoH and maternal morbidity.7,17,18 In the United States, systemic socioeconomic disparities, such as poverty, inadequate housing, and limited access to quality health care, result in poorer health outcomes for racial and ethnic minority populations.27–30 These socioeconomic inequities further disadvantage these populations during disasters or public health emergencies. Risk factors for SMM include mental health disorders, prepregnancy obesity, cesarean delivery, and other recognized factors.2,3,31 Unfortunately, Black women and those residing in low-income communities face a higher prevalence of SMM, with insufficient access to high-acuity obstetrical care. 32 Structural racism7,33 and SDoH risk factors, such as housing instability, food insecurity, transportation problems, and interpersonal safety, are additional obstacles that minority populations face, potentially exacerbating their risk of SMM and underlying health conditions. Addressing these SDoH is critical to improving maternal health and reducing the occurrence of life-threatening pregnancy complications, with subsequent reductions in maternal mortality.
Our study findings reveal that individuals in low SDoH or high-segregated communities were more likely to have clinical risk factors such as obesity, tobacco use, and chronic diseases. Their higher SMM risks are therefore concerning. We found that even in communities with higher proportions of White birthing individuals, who historically experienced lower incidence of SMM, those in low SDoH counties still had significantly higher odds of SMM. These findings confirmed that maternal race and ethnicity is a social construct but not a surrogate for considering broader social determinants. Even when individual race/ethnicity was not associated with community SDoH levels, community SDoH or Black segregation plays essential roles in SMM disparities. Remediating these SDoH factors that may be contributing to the increasing incidence of SMM during the COVID-19 pandemic and eliminating the disproportionate burden of loss among socially vulnerable families are imperative for maternal health equity.7,14 However, structural determinants might take time to achieve equity. Our results suggest the need to focus on communities that were both with low SDoH and highly Black segregated.
The consistent community-level disparities in SMM and the potential exacerbating racial disparities for Black and Hispanic birthing people living in low SDoH and highly segregated Black communities have important implications. First, in the current study, people in high-segregated Black communities are disproportionately more likely to smoke during pregnancy, to have diabetes, hypertension, depression, and anxiety during pregnancy, which can increase their risk of complications during pregnancy and childbirth. 34 Second, systemic racism is increasingly recognized in maternal health disparities.9,10,12,35 While the associations between the quality of perinatal care and racial disparities in SMM for Black people remain exploratory, Black birthing people in this study were more likely to undergo cesarean deliveries and give preterm birth. These cesarean procedures and adverse birth outcomes are often more pronounced in communities with high levels of racial segregation, where preexisting conditions (e.g., preeclampsia) are common and health care resources might be scarce.16–18
As federal and state governments continue to invest in improving equitable access to financial stability, education, employment, transportation, health care, and health literacy training resource, 14 our study suggests that these efforts should be specifically targeted toward Black segregated communities. It is crucial to reverse the inequities created by historical policy and contemporary practices in outcomes based on Black residential segregation to put an end to these disparities. The recent Government Accountability Office report has documented the exacerbated racial and ethnic disparities in maternal morbidity during the COVID-19 pandemic. 14 To mitigate such gaps, this study highlighted the importance of addressing these disparities through targeted community interventions, addressing SDoH, and mitigating residential segregation-associated structural racism for communities of color.
Interpreting the study findings should consider the following limitations. First, N3C is an observational registry that collects data from nationwide participating sites. Therefore, the study population might not be nationally representative due to nonrandom contributions from participating health systems. Second, over half of the study sample did not have residential ZIP codes or county information in the national EHR data; these birthing individuals were yet meaningfully similar in social, behavioral, and clinical characteristics as those included in the final analysis (Supplementary Appendix A2). Third, while we did not consider cases with blood transfusion alone as SMM, as transfusion may not always indicate a severe maternal event, this approach may underestimate the associations observed, as it can reflect ongoing, untreated anemia and serve as a proxy for access to and receipt of prenatal care, which is often limited in these areas. Blood transfusion is both a common SMM indicator and an important patient-reported outcome, especially given its association with postpartum hemorrhage, a leading contributor to preventable maternal mortality. However, its inclusion has been debated, as some argue it may also overestimate the burden of SMM. Our exclusion of transfusion-only cases, therefore, reflects a trade-off and should be considered when interpreting our findings. In addition, this study only includes childbirths during the COVID-19 pandemic (March 2020 or later) due to substantially different health systems contributing to pre- and peri-pandemic data to the N3C Data Enclave. Therefore, we were unable to consider the changes in SMM or its disparities between the pre- and peri-pandemic periods. Fourth, our study measured SMM from pregnancy to the first 42 days postpartum, potentially underestimating the association between SDoH and SMM, as non-Hispanic Black populations experience higher rates of late postpartum SMM. 36 Future research should incorporate extended postpartum follow-up to better capture the full scope of SMM incidence and its social determinants. Moreover, this study includes only birthing individuals with live births, excluding those with stillbirth; findings may not be generalizable to those delivering stillborn infants. Future research is needed to explore how SDoH shape the experience of stillbirth and the associated SMM risks. In addition, this study relied on race and ethnicity data contributed by each health system and might not reflect the accurate categories. 37 However, our county- and state-level race and ethnicity mix in the data were similar to those reported in the ACS data, suggesting that the measurement bias, if any, is minimal. Finally, we used ACS 2015–2019 5-year estimates to calculate the residential segregation index; the role of this lagged exposure on SMM during COVID-19 might be underestimated.
Health equity implications
This national cohort study has revealed that SMM rates were disproportionately higher among birthing people in low SDoH (i.e., socially disadvantaged) and Black segregation communities. The most high-risk birthing people (those with obesity/overweight, diabetes, hypertension, smoking, and suffering from depression/anxiety during pregnancy) were more likely to live in the lowest SDoH communities where they might encounter unstable housing, health care access barriers, low income, unsafe neighborhoods, and/or substandard education to support their pregnancy or postpartum life. These results highlight the importance of addressing multifaceted SDoH and residential segregation in the efforts to overcome maternal health disparities in the United States.
Footnotes
Authors’ Contributions
P.H.: Conceptualization, data curation, funding acquisition, investigation, methodology, project administration, and writing—original draft. J.L. and B.A.C.: Conceptualization, funding acquisition, and writing—review and editing. C.L.: Data curation, funding acquisition, investigation, methodology, validation, and writing—review and editing. T.L.: Data curation, formal analysis, and validation. J.Z.: Conceptualization, funding acquisition, investigation, methodology, validation, and writing—review and editing. A.B.P. and K.S.: Data curation, formal analysis, and methodology. Y.S.: Data curation, investigation, visualization, and writing—original draft. N.H.: Software, resources, funding acquisition, and writing—review and editing. B.O.: Conceptualization, funding acquisition, data acquisition, and investigation. X.L.: Conceptualization, funding acquisition, investigation, resources, supervision, and writing—review and editing.
Author Disclosure Statement
All authors have read and approved the article; all affirm that they meet authorship requirements and declare no conflicts of interest.
Funding Information
The project is supported by the National Institutes of Health, the National Institute of Allergy and Infectious Diseases (3R01AI127203-05S2).
Abbreviations Used
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
