Background
Maternal mortality is higher among Black than White people in the United States. Whether Black-White disparities in maternal in-hospital mortality during the delivery hospitalization vary across hospital types (Black-serving vs nonBlack-serving and teaching vs nonteaching) and whether overall maternal mortality differs across hospital types is not known.
Objective
The aims of this study were to determine whether risk-adjusted Black-White disparities in maternal mortality during the delivery hospitalization vary by hospital types (this is analysis of disparities in mortality within hospital types) and compare risk-adjusted in-hospital maternal mortality among Black-serving and nonBlack-serving teaching and nonteaching hospitals regardless of race (this is an analysis of overall mortality across hospital types).
Study Design
We performed a population-based, retrospective cohort study of 5,679,044 deliveries among Black (14.2%) and White patients (85.8%) in 3 states (California, Missouri, and Pennsylvania) from 1995 to 2009. A hospital discharge disposition of “death” defined maternal in-hospital mortality. Black-serving hospitals had at least 7% Black obstetrical patients (top quartile). We performed risk adjustment by calculating expected death rates using predictions from logistic regression models incorporating sociodemographics, rurality, comorbidities, multiple gestations, gestational age at delivery, year, state, and mode of delivery. We calculated risk-adjusted risk ratios of mortality by comparing observed-to-expected ratios among Black and White patients within hospital types and then examined mortality across hospital types, regardless of patient race. We quantified the proportion of Black-White disparities in mortality attributable to delivering in Black-serving hospitals using causal mediation analysis.
Results
There were 330 maternal deaths among 5,679,044 patients (5.8 per 100,000). Black patients died more often (11.5 per 100,000) than White patients (4.8 per 100,000) (relative risk, 2.38; 95% confidence interval, 1.89–2.98). Examination of Black-White disparities revealed that after risk adjustment, Black patients had significantly greater risk of death (adjusted relative risk, 1.44; 95% confidence interval, 1.17–1.79) and that the disparity was similar within each of the hospital types. Comparison of mortality, regardless of race, across hospital types revealed that among teaching hospitals, mortality was similar in Black-serving and nonBlack-serving hospitals. However, among nonteaching hospitals, mortality was significantly higher in Black-serving vs nonBlack-serving hospitals (adjusted relative risk, 1.47; 95% confidence interval, 1.15–1.87). Notably, 53% of Black patients delivered in nonteaching, Black-serving hospitals compared with just 19% of White patients. Among nonteaching hospitals, 47% of Black-White disparities in maternal in-hospital mortality were attributable to delivering at Black-serving hospitals.
Conclusion
Maternal in-hospital mortality during the delivery hospitalization among Black patients is more than double that of White patients. Our data suggest this disparity is caused by excess mortality among Black patients within each hospital type, in addition to excess mortality in nonteaching, Black-serving hospitals where most Black patients deliver. Addressing downstream effects of racism to achieve equity in maternal in-hospital mortality will require transparent reporting of quality metrics by race to reduce differential care and outcomes within hospital types, improvements in care delivery at Black-serving hospitals, overcoming barriers to accessing high-quality care among Black patients, and eventually desegregation of healthcare.
Introduction
After decades of decline, maternal mortality in the United States has steadily risen from 7.2 per 100,000 live births in 1987 to 17.4 in the most recent data from 2018. Of major concern are ongoing racial disparities in maternal mortality; Black patients are more than twice as likely to die than White patients (37.3 vs 14.9 per 100,000 live births). , The reasons for racial disparities in maternal mortality are multifactorial. Although lifelong exposures to social and environmental inequities from structural racism affect risks of contributing comorbid conditions such as obesity and hypertension, racism in hospitals and communities may lead to worse medical care and adverse outcomes among Black patients. ,
Why was this study conducted?
Black women die from pregnancy-related causes at 4 times the rate of White women. We sought to determine whether disparities and overall maternal in-hospital mortality differed by hospital type (Black- and nonBlack-serving, teaching and nonteaching).
Key findings
Risk-adjusted maternal in-hospital mortality was 44% higher among Black vs White patients, regardless of hospital type. Risk-adjusted mortality in nonteaching hospitals, regardless of patient race, was significantly higher in Black-serving vs nonBlack-serving hospitals; 53% of Black and 19% of White patients deliver in these hospitals ( P <.0001).
What does this add to what is known?
The contribution of healthcare to in-hospital mortality, which may be preventable, is less well-studied than overall maternal mortality. To address the repercussions of racism in healthcare, our findings call for reporting of maternal outcomes within hospitals by race, tackling drivers of lower quality care in Black-serving hospitals, and desegregating healthcare.
Previous studies have suggested that approximately half of maternal mortality is preventable and that 90% of preventable deaths may be attributable to provider factors such as delayed diagnoses. Furthermore, although there is rigorous work on hospital factors associated with severe maternal morbidity, , less is known about maternal mortality during the delivery hospitalization because it is a rarer outcome. Thus, it is important to identify hospital factors that may be drivers of disparities in in-hospital mortality. Patients choose delivery hospitals for various reasons that may include geography, hospital reputation or quality, and insurance coverage. Hospitals differ with respect to the demographic composition of their patients. Previous studies have shown that low-income and patients of color often receive care at lower quality hospitals, including obstetric care. Hospital factors that have been implicated in maternal in-hospital severe morbidity and mortality include lack of a 24-hour on-site anesthesiologist, hospital volume (low and high), time of day, and staffing. ,
Another potential driver of differential maternal in-hospital mortality may be teaching status. Teaching hospitals often care for more complex patients. Although care quality is often higher in teaching hospitals, patient satisfaction scores are often lower. Because patient satisfaction can correlate with interpersonal interactions and communication that can vary based on implicit bias, , our hypothesis was that racial disparities in in-hospital maternal mortality may differ based on teaching status. Due to potential differences in cultural competence that may flow from exposure to diverse patients (emersion), our second hypothesis was that disparities would be larger in hospitals that care for a lower proportion of Black patients ( Videos 1 and 2 ).
Materials and Methods
We used linked birth certificate and hospital discharge data from California, Missouri, and Pennsylvania from 1995 to 2009 to compare maternal in-hospital mortality ratios for non-Hispanic Black and non-Hispanic White patients (n=5,679,044), hereafter designated as Black and White for brevity. Race or ethnicity was self-designated on the birth certificate. All other race or ethnicities were excluded in this analysis.
The primary outcome was maternal death before discharge during the birth hospitalization. We used a hospital discharge disposition of “death” to determine whether a patient died during the hospital stay. We analyzed 2 primary exposures: teaching status and Black-serving status. Hospital teaching status was obtained from the American Hospital Association database and linked using the hospital identifier. We calculated the percentage of patients who delivered who were Black at each hospital to generate quartiles and compared the top Black-serving hospital quartile (“Black-serving”) with the bottom 3 quartiles (“nonBlack-serving”); hospitals with >7.08% of Black patients were considered “Black-serving.” This threshold was chosen to be consistent with a previous study of severe maternal morbidity by Howell et al in which low Black-serving hospitals were the lowest 3 quartiles of the proportion of Black patients served. This designation was specific to the obstetric population. Hospitals could move in and out of each quartile as their demographics changed annually.
We performed risk adjustment using observed-to-expected (O:E) ratios. To generate expected rates of death, we performed a logistic regression model with the dependent variable of death and the following independent variables: maternal age, insurance, gestational age at delivery, multiple gestations, mode of delivery, rurality, year, pregnancy comorbid conditions (chronic hypertension, chronic cardiac disease, diabetes mellitus, and chronic renal disease) from hospital diagnosis codes, and state of delivery (California, Missouri, and Pennsylvania). We obtained predicated probabilities of death for each variable from the regression models using the “predict” command in Stata (StataCorp LLC, College Station, TX) and summed the probabilities to obtained expected death rates within strata of interest.
There were 2 primary comparisons made. First, we compared mortality among Black patients with mortality among White patients overall and within each hospital type by calculating Black-to-White risk-adjusted risk ratios using O:E/O:E. We dichotomized hospital types in 2 ways: teaching vs nonteaching and then Black-serving vs nonBlack-serving. Then, we created 4 mutually exclusive hospital types (teaching Black-serving, teaching nonBlack-serving, nonteaching Black-serving, and nonteaching nonBlack-serving). We used the same technique (O:E/O:E) to compare risk-adjusted mortality among all patients (Black and White), regardless of race, in the 4 mutually exclusive hospital types; these were analyses across hospital types. In each analysis, we generated 95% confidence intervals (CIs) by bootstrapping with 1000 iterations. ,
We also quantified the proportion of the racial disparity in maternal in-hospital mortality explained by delivering at Black-serving vs nonBlack-serving hospitals in models stratified by teaching status using VanderWeele’s approach to causal mediation. We used adjusted log-binomial regression with death as the dependent variable, Black race as the independent variable, and delivering at a Black-serving hospital as the mediator. We included the same covariates in the model as we did in the model used for risk adjustment.
With respect to missing data, only rural or urban status was missing among nonteaching hospitals for 19,279 deliveries (0.3%), and these deliveries were included in the reference (large urban) group given it was the most common setting (95.5%). This study was approved by the Children’s Hospital of Philadelphia Institutional Review Board. Statistical analyses were performed using both SAS 9.4 (SAS Institute Inc, Cary, NC) and Stata 16 (StataCorp LLC, College Station, TX).
Results
Patient and hospital demographics are presented in Table 1 . Patients delivering at teaching hospitals had a higher prevalence of all comorbidities than patients delivering at nonteaching hospitals. In contrast, the prevalence of comorbidities was similar in Black-serving compared with nonBlack-serving hospitals. There were 330 deaths among the 5,679,044 live births (5.3 per 100,000) during the delivery hospitalization.
| Teaching, % | Nonteaching, % | |||
|---|---|---|---|---|
| NonBlack-serving (n=316,369) | Black-serving (n=600,406) | NonBlack-serving (n=3,392,721) | Black-serving (n=1,369,548) | |
| Characteristic (n=5,679,044) | ||||
| Maternal age | ||||
| <25 | 26.9 | 28.5 | 28.9 | 36.2 |
| 25–35 | 55.3 | 51.5 | 53.6 | 49.0 |
| ≥35 | 17.8 | 20.0 | 17.6 | 14.7 |
| Race | ||||
| Non-Hispanic Black | 8.6 | 33.3 | 4.8 | 32.5 |
| Non-Hispanic White | 91.4 | 66.7 | 95.2 | 67.5 |
| Insurance category | ||||
| Private | 69.0 | 63.7 | 68.3 | 56.4 |
| Medicaid | 27.8 | 33.8 | 27.6 | 38.5 |
| Other insurance | 3.2 | 2.4 | 4.1 | 5.1 |
| Preterm (<37 wk gestation) | 12.7 | 12.8 | 8.3 | 10.7 |
| Multiple gestations | 3.6 | 3.6 | 2.9 | 3.0 |
| Cesarean delivery | 26.7 | 29.2 | 27.6 | 26.5 |
| State | ||||
| California | 32.8 | 24.8 | 62.8 | 55.7 |
| Missouri | 16.9 | 16.8 | 17.0 | 19.5 |
| Pennsylvania | 50.3 | 58.4 | 20.2 | 24.8 |
| Rural or urban a | ||||
| Large metro | 94.4 | 98.5 | 77.6 | 95.5 |
| Small metro | 0.2 | 1.5 | 9.6 | 1.6 |
| Rural | 5.5 | 0.0 | 12.4 | 2.6 |
| Maternal morbidity | ||||
| Chronic cardiac disease | 1.8 | 1.6 | 0.8 | 0.8 |
| Gestational diabetes | 5.2 | 3.9 | 3.7 | 3.7 |
| Diabetes mellitus | 1.1 | 1.1 | 0.5 | 0.7 |
| Chronic hypertension | 1.3 | 1.6 | 0.8 | 1.1 |
| Gestational hypertension | 4.8 | 5.8 | 3.0 | 3.6 |
| Chronic renal disease | 0.2 | 0.2 | 0.1 | 0.1 |
| Acute renal failure | 0.03 | 0.03 | 0.01 | 0.02 |
| Epoch | ||||
| 1995–1999 | 39.5 | 25.2 | 33.9 | 38.4 |
| 2000–2004 | 33.5 | 32.2 | 33.8 | 31.3 |
| 2005–2009 | 27.1 | 42.7 | 32.3 | 30.3 |
a Missing data: rural or urban status was missing for n=19,279 deliveries (0.3%).
Black-White disparities in maternal in-hospital mortality within hospital types
Black patients (11.5 per 100,000) were more than twice as likely as White patients (4.8 per 100,000) to die in the hospital (relative risk [RR], 2.38; 95% CI, 1.88–3.02). The risk-adjusted mortality ratio was significantly higher among Black patients than White patients (O:E/O:E, adjusted RR, 1.44; 95% CI, 1.17–1.79) ( Table 2 ). In each hospital type, Black patients had a higher unadjusted risk of death than White patients ( Figure ). After risk adjustment, the point estimates comparing Black with White patients were similar, but only in nonteaching hospitals was the mortality ratio significantly higher among Black patients than White patients ( Table 2 ). Including teaching status and Black-serving status together revealed a similar increase in mortality risk among Black compared with White patients ( Table 3 ).
