Abstract
Background:
In the United States, Black maternal mortality is 2–4 × higher than that of White maternal mortality, with differences also present in severe maternal morbidity and other measures. However, limited research has comprehensively studied multilevel social determinants of health, and their confounding and effect modification on obstetrical outcomes.
Materials and Methods:
We performed a retrospective multistate analysis of adult inpatient delivery hospitalizations (Florida, Kentucky, Maryland, New Jersey, New York, North Carolina, and Washington) between 2007 and 2020. Multilevel multivariable models were used to test the confounder-adjusted association for race/ethnicity and the binary outcomes (1) in-hospital mortality or maternal end-organ injury and (2) in-hospital mortality only. Stratified analyses were performed to test effect modification.
Results:
The confounder-adjusted odds ratio showed that Black (1.33, 95% confidence interval [CI]: 1.30–1.36) and Hispanic (1.14, 95% CI: 1.11–1.18) as compared with White patients were more likely to die in-hospital or experience maternal end-organ injury. For Black and Hispanic patients, stratified analysis showed that findings remained significant in almost all homogeneous strata. After statistical adjustment, Black as compared with White patients were more likely to die in-hospital (1.49, 95% CI: 1.21–1.82).
Conclusions:
Black and Hispanic patients had higher adjusted odds of in-patient mortality and end-organ damage after birth than White patients. Race and ethnicity serve as strong predictors of health care inequality, and differences in outcomes may reflect broader structural racism and individual implicit bias. Proposed solutions require immense and multifaceted active efforts to restructure how obstetrical care is provided on the societal, hospital, and patient level.
Introduction
The United States is suffering from a Black-White disparity–driven maternal health care crisis. 1 The 2021 maternal mortality ratio (MMR) was 32.9 deaths per 100,000 live births, with ratios for deliveries to Black mothers (69.9) being 2.60 times that of deliveries to White mothers (26.6). 1 Internationally, the overall U.S. MMR is more than three times higher than that of other developed nations, with most higher resourced countries having rates at <10. 1 –3
Research has shown that Black mothers have maternal mortality two to four times higher than White mothers, with disparities observed in severe maternal morbidity (SMM), failure to rescue, and other obstetrical outcomes. 1,4 –13 Findings for deliveries to Hispanic mothers have been less consistent, with the unadjusted Hispanic MMR (28.0) being equivalent to the non-Hispanic White MMR and some multivariate analysis showing higher adjusted mortality and morbidity. 1,6 –8,11,14
Drivers of obstetrical disparities include structural or systemic racism such as unequal medical care resource deployment and variation in care site quality; social determinants of health; differences in baseline comorbid conditions; and provider–patient interactions subject to (implicit) bias, prejudice, or racism. 4,8,11,14 –17 Structural racism is systemic and institutionalized perpetuation of racial group inequity through public policies, institutional practices, cultural representations, and other norms that are embedded in long-standing social policy. 5,15 –20 Examples of structural racism include racialized residential segregation (“redlining”), economic suppression, mass incarceration and police violence toward minoritized populations, and health care inequity 16,17 ; these policies and practices establish expansive social disinvestment in Black neighborhood infrastructure, education, services, and employment. 16,17
Social determinants of health are the conditions in which people are born, live, learn, work, play, worship, and age, which affect a wide range of health, functioning, and quality-of-life outcomes. Social determinants of health are affected by the imbalances of influence and wealth established through structural racism 21,22 and are grouped into five domains: health care access and quality, education access and quality, social and community context, economic stability, and neighborhood and built environment. 23
Differences in social determinants of health (such as safe affordable housing, living wage, quality education, transportation, availability of food, job security, social connections, and safety) lead to differences in exposure to psychosocial stress and unhealthy behaviors, which is associated with a disparity in distribution of disease, illness, and well-being. 24,25 Through these processes, Black women experience increased acute and chronic stressors over their life course, which are associated with an increased risk of chronic disease, comorbidities, and obstetrical-related mortality and morbidity. 26,27
Implicit bias is influences, attitudes, or stereotypes toward minoritized racial and ethnic groups that affect health care professionals' understanding, actions, and decisions in an unconscious manner; implicit bias has been linked to poorer patient care provision, negatively affects patient–physician interactions, and leads to disparities in health care delivery. 15,28 Few studies have comprehensively examined multilevel social determinants of health, their confounding, and effect modification on maternal disparities. 4,14,29
Herein, we aim to conduct a 14-year (2007–2020) multistate analysis to quantify the associations between racial and ethnic disparities, possible effect modifiers, and maternal end-organ injury or inpatient mortality (primary outcome), and inpatient mortality only (secondary outcome). 30 Our analytic plan will assess separately Black race and Hispanic ethnicity as compared with non-Hispanic White race independently and in stratified models; we will also conduct analysis for the heterogeneous categories of Other and Missing race.
We aim to evaluate findings by geographic-, hospital-, birthing type-, and patient-level factors, 31 with adjustment for a comorbidity index specific to the obstetric population. 30,32 Our previous research has shown that within individual insurance payer and median income subgroups, Black race alone predicted inpatient mortality, prolonged length of hospital stay, and routine use of cesarean delivery. 31,33
Additional research has demonstrated that Black race alone predicted SMM in a series of multilevel social determinants of health stratified models by geographic-, hospital-, birthing type-, and patient-level factors. Given these results and the extensive literature documenting the Black–White maternal disparity gap, we hypothesize that Black mothers will have worse outcomes for maternal mortality and morbidity among most stratified homogeneous subgroups.
Materials and Methods
Study database and population
We retrospectively selected inpatient delivery hospitalization data from Kentucky, Maryland, New Jersey, North Carolina, Washington (2007–2020); Florida (2007–2019); and New York (2007–2017), from the State Inpatient Databases (SID); Health care Cost and Utilization Project (HCUP), Agency for Health care Research and Quality (AHRQ). These seven states were purchasable from HCUP, and were chosen for analysis to capture racially/ethnically diverse and largely populated segments of the United States (per individual state availability for year listed in parentheses if not 2007–2020).
The SID is an all-payer inpatient administrative database consisting of nonfederal, nonpsychiatric hospitals. Data are coded so that each inpatient hospital admission corresponds to an individual record; pregnant patients who had multiple gestations could appear more than once in the database, but there is no definitive way to identify these patients in the data. Present-on-admission comorbidity diagnosis indicators facilitate differentiating pre-existing medical comorbidities and peripartum complications. 5,31
Delivery hospitalization records were abstracted using International Classification of Diseases, Ninth Revision and Tenth Revision, Clinical Modification codes (ICD), and Diagnosis Related Groups. 34,35 Inclusion criteria for our study included all records ≥18 years. Exclusion criteria included records with missing data on age or sex, or those lacking a hospital identification number. For the HCUP sex variable, all nonmale, nonfemale (e.g., “other”) values are set to missing. 36 Multivariable regression analyses used complete case analysis. All study activities were approved by the Weill Cornell Medicine Institutional Review Board (#1308014181). STROBE guidelines were followed. 37
Recorded patient race/ethnicity was our exposure of interest (White [reference category], Black, Hispanic, other [heterogeneous category consisting of Asian or Pacific Islander, Native American, Other], or missing); our comparisons of interest were between White and Black and White and Hispanic, independently. 31 White race was chosen as the reference category because it denotes the largest racial and ethnic category in our sample population, allowing for a sizable reference comparison group.
Other patient-level covariates included patient age; primary expected insurance payer (private [reference], Medicaid, Medicare, Uninsured, Other, coded as missing); median household income quartile by patient's state of residence ZIP code (referred to as median income; some values coded as missing); delivery type (vaginal [reference], operative vaginal [forceps or vacuum], or caesarean); and a validated present-on-admission flagged obstetric comorbidity index used to predict maternal end-organ injury or inpatient mortality. 30,32
Hospital-level covariates included state (reference: Florida; Florida was chosen as reference category because it contains the largest population by state in our sample population); year of hospital discharge (reference: 2007); hospital birthing volume quartile based on overall sample distribution (reference: first quartile); proportion of deliveries to Black mothers by hospital based on overall sample distribution (categorized as high: top 5th percentile, medium: 5th–25th percentile, low: all other hospitals [reference]) 5 ; and hospital safety-net burden (percentage of a hospital's total cases billed to Medicaid or uninsured, categorized in tertiles based on individual state sample distribution secondary to state Medicaid rule differences[reference: low safety net]). 5
Patient's geographic location to define rural and urban status was coded as unordered: central counties with metro areas of >1 million population (urban, reference), fringe counties of metro areas of >1 million population (suburban), counties with metro areas of 250,000–999,999 population (medium city), counties with at least one metro area of 50,000–249,999 population (small city), counties with population cores of 10,000–49,999 population (micropolitan rural), noncore counties, which contain no city or town of <10,000 population (noncore rural), and counties coded as missing. 38
Study outcomes
Our primary outcome was a composite binary measure of the presence of maternal end-organ injury or inpatient mortality during the delivery hospitalization; this outcome measure represents both maternal mortality and morbidity, and has been extensively used in obstetrical research using administrative data. 30,32 Our secondary outcome was the individual binary measure of inpatient mortality during the delivery hospitalization. The individual composite measure elements for maternal end-organ injury were selected as exploratory outcomes. 30,32
Maternal end-organ injury included the presence of acute heart failure, acute renal failure, acute liver disease, acute myocardial infarction, acute respiratory distress syndrome/respiratory failure, disseminated intravascular coagulation/coagulopathy, coma, delirium, puerperal cerebrovascular disorders, pulmonary edema, pulmonary embolism, sepsis, shock, status asthmaticus, or status epilepticus. 30,32 For inclusion, indicators of end-organ injury needed to be coded as “not present-on-admission (nPOA).” 5,31
Statistical analysis
Descriptive statistics were calculated across race/ethnicity categories. Standardized mean differences were calculated for our patient and hospital characteristics with a meaningful difference defined as ≥10%.
Generalized linear mixed models (GLMMs) were fit to the data, and adjusted odds ratios (aORs) for dichotomous outcomes with 95% confidence intervals (CIs) were reported; GLMMs account for potential violations of the conditional independence assumption by clustering patients within individual hospitals. A priori all variables listed above were included. 5,31 We developed separate models for our binary outcomes of interest. 30,32 For primary and secondary outcome models, we calculated an E-value and confidence limit to measure the minimum strength of association that an unmeasured confounder would need to have with both race/ethnicity and the outcome to fully explain away the found observed association of interest. 39 Sensitivity regression models using ICD-10 sample data only were fit for each outcome measure, restricting data to a start date of October 1, 2015.
To assess potential effect modification, we performed stratification analysis. We fit stratified multivariate models for primary and secondary outcomes by substratum geographic- (rural–urban status), hospital- (Black-serving birth unit designation, safety-net burden designation, birthing volume), procedure- (birthing type), and patient- (insurance payer, median income) level variables. The models contained all the main model variables, except for the stratified-by variable.
p-Values were two sided with statistical significance evaluated at <0.05 alpha level. Statistical tests and analysis were performed using SAS V.9.4 (Cary, NC) and Stata SE V.16 (College Station, TX).
Results
Patient and hospital characteristics
The total number of delivery hospitalization records with an identified singleton or multiple births was 11,115,847. After dropping cases with missing age, missing sex, not ≥18 and with missing hospital identifiers, the final analysis sample size was 10,874,289 (2.17% reduction; Fig. 1).
Patient flowchart. Adult patients who underwent inpatient birthing hospitalizations from 2007 to 2020 data (as available) from Florida (limited to 2007–2019), Kentucky, Maryland, New Jersey, New York (limited to 2007–2017), North Carolina, and Washington. After exclusion, the final analysis sample size was 10,874,289 (2.17% reduction).
Descriptive statistics and outcomes by race/ethnicity are found in Tables 1 and 2. A total of 60,005 (0.6%) patients experienced either in-hospital mortality or maternal end-organ injury (Black: 0.8%, Hispanic: 0.6%, White: 0.5%, Other: 0.6%, Missing: 0.2%), including 768 inpatient mortalities. Individual morbidity outcome measures (with individual prevalence >0.1%) by race/ethnicity are shown in Figure 2.
Bar graph of individual morbidity outcome measures (with individual prevalence >0.1%) by race/ethnicity. X-axis represents individual morbidity outcome measures by race/ethnicity category. Y-axis is prevalence percentage.
Patient Characteristics According to Race and Ethnicity Category
Data provided as n (%), mean (SD), or median (IQR), as appropriate.
Percentages may not sum to 100 due to rounding and missing values.
IQR, interquartile range; SD, standard deviation; SMD, standardized mean difference.
Outcome Measures According to Race and Ethnicity Category
Data provided as n (%), mean (SD), or median (IQR), as appropriate.
Percentages may not sum to 100 due to rounding and missing values.
Black and Hispanic, as compared with White patients, were more likely to be in the poorest median income quartile, be on Medicaid or self-pay/no-charge insurance, be from urban counties, and deliver at high safety-net burden hospitals with medium or high Black-serving delivery. The Bateman comorbidity index median scores were similar; individual pre-existing comorbidities with higher standardized mean differences among Black as compared with White patients included Human Immunodeficiency Virus, pre-existing hypertension, and sickle cell disease.
Primary outcome: in-hospital mortality or maternal end-organ injury
After statistical adjustment, Black (aOR: 1.33, 95% CI: 1.30–1.36; E-value: 1.99 [lower limit of 95% CI: 1.92]), Hispanic (aOR: 1.14, 95% CI: 1.11–1.18; E-value: 1.54 [lower limit of 95% CI: 1.46]), and Other (aOR: 1.23, 95% CI: 1.19–1.26) as compared with White patients were more likely to experience in-hospital mortality or maternal end-organ injury (Table 3 and Supplementary Table S1). Missing (aOR: 0.70, 95% CI: 0.66–0.74) as compared with White patients were less likely to experience in-hospital mortality or maternal end-organ injury. Sensitivity analysis of only ICD-10 codes corroborated our findings. In stratified analysis, these findings for Black and Hispanic mothers remained significant in almost all homogeneous strata (Figs. 3 and 4).
aORs for Black compared with White women (95% CIs) for in-hospital mortality or maternal end-organ injury, by subgroup. X-axis scale represents adjusted odds ratio and 95% CI for each corresponding category substratum. aOR, adjusted odds ratio; 95% CI, 95% confidence interval.
aORs for Hispanic compared with White women (95% CIs) for in-hospital mortality or maternal end-organ injury, by subgroup. X-axis scale represents aOR and 95% CI for each corresponding category substratum.
Results from Univariable and Multivariable Regression Models for Outcome Measures According to Race and Ethnicity Category
Reference category for all models: White patients.
Crude and adjusted ORs were produced from models clustering on hospital.
p < 0.05, ** p < 0.01, *** p < 0.001.
aOR, adjusted odds ratio.
Secondary outcome: in-hospital mortality
After statistical adjustment, Black (aOR: 1.49, 95% CI: 1.21–1.82; E-value 2.34 [lower level of the 95% CI: 1.71], Other (aOR: 1.40, 95% CI: 1.09–180), and Missing (aOR: 1.99, 95% CI: 1.43–2.79) patients, as compared with White patients, were more likely to die in-hospital (Table 3 and Supplementary Table S1). There was no significant difference between Hispanic and White patients. A sensitivity analysis of only ICD-10 data showed no significant difference between Black or Hispanic and White patients. Stratified analyses did not show any consistent pattern of disparities throughout homogeneous strata (Table 4).
Results from Stratified Generalized Linear Mixed Models for Outcome of Maternal In-Hospital Mortality for Black Race and Hispanic Ethnicity as Compared with White Race
Exponentiated coefficients; 95% CIs in parentheses.
Statistically significant results are bolded.
North Carolina model unable to be fit due to nonconvergence.
p < 0.05, ** p < 0.01, *** p < 0.001.
CI, confidence interval.
Exploratory multivariable regression models for individual maternal end-organ injury indicators showed that after statistical adjustment, Black, Hispanic, Other, and Missing, as compared with White patients were more likely to experience heart failure, renal failure, disseminated intravascular coagulation, and shock. Black patients were also more likely to experience myocardial infarction, respiratory distress, puerperal cerebrovascular disorders, pulmonary embolism, and sepsis (Table 3).
Discussion
Our analysis found that in ∼11 million delivery hospitalizations from 2007 to 2020, Black maternal race was associated with higher inpatient mortality and end-organ damage and inpatient mortality only. In post hoc exploratory models, Black race was significantly associated with higher adjusted incidence of 10/15 inpatient end-organ damage individual measures. The consistency of this association in most of a series of stratification models makes our findings robust, and highlights the effect modification found between Black race and other social determinants. Hispanic ethnicity was associated with higher inpatient mortality and end-organ damage, but not with inpatient mortality only.
To our knowledge, our study features the most up-to-date, longitudinal, multilevel analysis; prior studies utilized dated records, were limited in scope, did not extensively address confounding and effect modification, or did not take into account appropriate statistical adjustment for pre-existing obstetrical comorbidities. 4,14,31 Our analysis found mixed results for the secondary outcome of inpatient mortality only, likely reflecting the relatively rare outcome occurrence, and reinforces our a priori primary outcome decision to use the mortality and morbidity composite measure. 30,32
Our findings are consistent with recent population-based research. 4,7,8,31 Using 2007–2014 multistate data, Tangel et al. found that Black as compared with White deliveries experienced higher adjusted mortality, independent of differences in obstetric comorbidity index score, insurance payer, and median income category; Hispanic deliveries were not more likely to have higher adjusted mortality. 31 Other research has shown that Hispanic maternal ethnicity has both positive and negative associations with outcomes, possibly reflecting the “Hispanic Paradox,” issues with methodology, lack of adjustment for spoken language, or other unmeasured confounders. 14,40 –42
Admon et al., using 2012–2015 national data, showed that every racial and ethnic minority category as compared with non-Hispanic White had higher SMM incidence. 6 Furthermore, using 1999–2017 national data, Guglielminotti et al. found that the failure-to-rescue ratio from SMM to mortality was highest for Black deliveries, and was increased for Hispanic, other, and missing race and ethnicity. 7 A 1995–2008 multistate analysis by Burris et al. examining hospital characteristics showed that 47% of the Black–White excess mortality in nonteaching hospitals was found in Black-serving hospitals. 43
Luke et al., using 2012–2017 National Inpatient Sample data, found that Black women in urban and micropolitan rural counties had the highest adjusted odds of SMM and mortality. 38 Differences also exist in birthing methods used: Black and Hispanic women, as compared with White women, have increased rates of cesarean birth and the increased cesarean births among non-Hispanic Black and Hispanic women accounting for ∼15.8% and 16.5% increased maternal morbidity, respectively, as compared with non-Hispanic White women. 4,11 Herein, we showed that our study outcomes exist independent of these potential confounders.
Race represents a social construct and not a biological determinant, and it would be a misinterpretation of our study to conclude that any race or ethnicity is inferior secondary to biological or genetic factors. 44 We believe that our results are rooted in the racism, prejudice, and provider bias that Black mothers experience in societal and health care interactions. Black women suffer from systemic and institutionalized racism, which presents as inequalities in health care access. 44
Chronic exposure to psychosocial, economic, and environmental stress has a “weathering” effect that can cause cellular-level DNA changes, telomere shortening, and adverse health outcomes. 44 This disparity is so pervasive that 2007–2017 CDC data found that pregnancy-related mortality ratios for Black women with at least some college education were higher than those from all other racial/ethnic groups with less than a high school diploma. 45 This finding is shocking because higher levels of education are generally supposed to be protective of health, mitigating the Black–White MMRs through education's effects on cognitive abilities, economic resources, and increased autonomy. 45,46
This paradoxical finding is attributed to the role of life-course socioeconomic and contextual disadvantage, implicit bias and barriers faced, and the inequality stress processes that Black women experience relative to Whites. 45 There is equivalency of socioeconomic, contextual, psychosocial, and health disadvantages between college-educated Black women and White women with a high school degree or less. Black women who achieved a bachelor's degree or more lived in neighborhoods with similar median incomes, higher unemployment rates, and similar poverty rates to low-educated White women and report similar rates of stress, victimization, parental death, and parental incarceration. 45 Further, at a given education level, there are significant differences in earnings and wealth between Black and White people.
Our study found that patients of Other racial and ethnic categorization (a heterogeneous grouping composed of Asian or Pacific Islander, Native American, and Other patients) were more likely to experience in-hospital mortality or maternal end-organ injury or in-hospital mortality only, as compared with White patients. A recent systematic review examining social determinants of pregnancy-related mortality and morbidity in the United States found that Asian-Pacific Islanders had increased in-hospital mortality, individual SMM indicators, and overall composite SMM, as compared with White patients; Native American and American Indian patients were found to have increased individual SMM indicators and overall composite SMM, as compared with White patients. 6,14,47,48
Our study also found that patients with Missing racial and ethnic categorization (a heterogeneous category composed of hospital discharges where race and ethnicity are missing or invalid at the hospital data collection level) had divergence in primary (decreased) and secondary (increased) study findings, most likely secondary to category heterogeneity. A systematic review examining the accuracy of U.S. race and ethnicity data found that there were high levels of accurate data for White and Black patients, but that there were relatively high levels of misclassification and incomplete data for Hispanic, Asian, Pacific Islander, and American Indian/Alaska Native. 49 Patient populations with missing race or ethnicity data are more likely to be poorer and less likely to have commercial insurance, more likely to have more comorbidities, and to have worse health care experiences than patients with available or accurate race and ethnicity data. 50,51
Our models show a complex association between the social construct of race, differences in quality and access to care, and maternal outcomes. 4,5,14,31,38,43 These findings suggest that factors that contribute to increased maternal mortality and morbidity in the United States may be amplified in Black (and Hispanic) individuals, but interventions may benefit the population more widely. 52 Recommended actions include an overarching approach to simultaneously address and bring explicit visibility to structural racism and macrosocial factors, hospital structural characteristics, patient-specific risk factors, and solutions with a focus on patient safety imperatives 20 ; however, data on the outlined approaches effectiveness are limited. 4,7,8,21,53 –58
Legislation and care systems can also improve maternal health through efforts to increase pregnancy-related public health expenditures (expanded insurance coverage adoption, nutritional assistance), address access limitations rooted in structural racism (transportation availability, housing instability, hospital funding), promote growth and diversity of maternal health workforce, and advance data collection and research efforts. 4,8,21 Perinatal quality initiatives that promote changes in medical culture through practice recommendations and actionable prevention methods can further be expanded. 53,55
Further suggestions to mitigate health care disparities include a call to action at the health care system, clinician/provider, researcher, and patient level. 21,52,59,60 Health care systems should create a culture of equity with provider group diversity with appropriate patient racial and language concordance and health equity and implicit bias training; provide health equity initiatives in clinical fields through standardized care protocols; provide financial incentives for measures that are linked to health equity; and promote governance that supports health equity and patient safety imperatives (e.g., care coordination, quality assurance practices, evidence-based care, enhanced maternal mortality, and SMM reviews). 52,59 –61
This includes the utilization of clinical informatics tools that support appropriate patient-centered care rooted in shared medical decision making through cultural awareness, competency, and humility. 60,61 Individual clinicians or providers should recognize personal cognitive and implicit biases; encourage motivation to change biases through participation in health disparities continuing education; have patient communication training; engage in community outreach initiatives; and advocate for health equity at a national and legislative level. 52,59
Efforts should include all providers and not just those from historically marginalized communities. Physicians and scientists should perform race-conscious medicine that emphasizes structural racism, rather than race, as key determinants of illness and health, and to focus efforts to mitigate health inequities. This is in contrast to race-based medicine, which characterizes race as a biological variable leading to inappropriate and misguided health care bias, stereotyping, and tailored clinical practices that have the potential for harm to patients through promotion of inequitable care. 62
Our study has numerous strengths. HCUP-SID dataset validity and internal consistency are verified by quality control procedures. Our sample represents about a quarter of the total U.S. population and U.S. deliveries for 14 consecutive years. 63 This sample size and statistical power allowed for multivariable multilevel logistic regression and stratification as well as for models to be executed to examine confounding and effect modification between Black race and Hispanic ethnicity, structural characteristics of hospitals, birthing type, other socioeconomic variables, and the adverse maternal outcome measure. We acknowledge that our selection of states excludes much of “Middle America” (five of our states are Eastern/Atlantic and one is Western/Pacific), and that our findings may not be generalizable to the entire U.S. population, other nonanalyzed states, or other countries.
Like other retrospective observational studies, our study has several limitations, including the potential of misclassification bias and residual confounding. We believe that any misclassification would be “at random” and should bias toward the null hypothesis. As shown by the calculated E-value estimates, unmeasured confounders could potentially bias our estimates. We were unable to examine the impact of structural racism and implicit bias on our associations, including measures representing residential segregation (neighborhood racial and economic polarization, neighborhood population racial inequity), hospital resources, educational attainment, employment status, judicial treatment, demographics of medical team staffing, or its interaction with race/ethnicity. 8 –10,12,15 –17
Measures of structural racism have previously been identified as main drivers of racial and ethnic disparities in maternal mortality and severe morbidity. 9,10,12 We were also unable to account for variables such as spoken language, patient preferences, medications, disease severity, or clinical characteristics. 15 Our outcome of interest was based upon inpatient records, meaning adverse events occurring outside the hospital could not be analyzed; this may underestimate adverse outcome event rates. For instance, among pregnancy-related deaths, 31% occurred during pregnancy, 36% at or within 1 week after delivery, and 33% from 1 week to 1 year after delivery. 64
In the HCUP-SID coding, Hispanic ethnicity is prioritized over race, contributing to misclassification bias (a patient who identifies as Hispanic and Black would be classified as Hispanic). We were unable to account for multiple gestations in our analysis, which could bias our findings. We performed multiple statistical comparisons, which presents a risk of Type 1 error because of multiplicity. Given the cross-sectional nature of this study, our findings only suggest correlation and not causality. Some of our data are 15 years old and may no longer be applicable. However, our sensitivity analysis, which utilized only ICD-10 data starting from October 1, 2015, largely confirms the main findings of the study and mitigates concern for using older data.
Conclusion
In summary, we found that Black and Hispanic delivery hospitalizations had higher in-patient mortality and end-organ damage after birth than White delivery hospitalizations. Our study suggests that race and ethnicity serve as strong predictors of health care inequality, and indicate that differences in outcomes may reflect broader structural racism and implicit bias in the health care system. 4,21 Solutions to combat disparities require immense and multifaceted active efforts to restructure hospital- and patient-level care. 21 Future research should aim to explore additional factors associated with structural racism and implicit bias. Physicians have traditionally been patient safety advocates; we should embrace this challenge to support health equity, and act as a solution to race- and ethnicity-based health care disparities. 65,66
Footnotes
Authors' Contributions
R.W. was responsible for the overall content as guarantor and took full responsibility for the finished work, and/or the conduct of the study, had access to the data, and controlled the decision to publish. R.W., V.T., and K.P. conducted study design and planning. V.T. and S.J. performed data collection. V.T. and R.W. carried out data/statistical analysis. R.W., V.T., B.L., S.J., K.P., and S.A. contributed to article writing and drafting.
Author Disclosure Statement
No competing financial interests exist.
Funding Information
Dr. Robert White is the recipient of a Foundation for Anesthesia Education and Research (FAER) Mentored Research Training Grant ID: MRTG-08-15-2021-White (Robert).
Supplementary Material
Supplementary Table S1
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.
