Background
Black-serving hospitals are associated with increased maternal risk. However, prior administrative data research on maternal disparities has generally included limited hospital factors. More detailed evaluation of hospital factors related to obstetric outcomes may be important in understanding disparities.
Objective
To examine detailed characteristics of Black-serving hospitals and how these characteristics are associated with risk for severe maternal morbidity (SMM).
Methods
This serial cross-sectional study linked the 2010-2011 Nationwide Inpatient Sample and the 2013 American Hospital Association Annual Survey Databases. Delivery hospitalizations occurring to women 15-54 years of age were identified. The proportions of non-Hispanic Black patients within a hospital was categorized into quartiles, and hospital factors such as specialized medical, surgical and safety-net services as well as payer mix were compared across these quartiles. A series of models was performed evaluating risk for SMM with Black-serving hospital quartile as the primary exposure. Log linear regression models with a Poisson distribution (and robust variance) were performed with unadjusted and adjusted risk ratios (aRR) with 95% confidence intervals (CIs) as measures of effect.
Results
Overall 965,202 deliveries from 430 hospitals met inclusion criteria and were included in the analysis. By quartile, non-Hispanic Black patients accounted for 1.3%, 5.4%, 13.4%, and 33.8% of patients. Many services were significantly less common in the lowest compared to the highest Black-serving hospital quartile including cardiac intensive care (48.9% versus 74.5%), neonatal intensive care (28.9% versus 64.9%), pediatric intensive care (20.0% versus 45.7%), pediatric cardiology (29.6% versus 44.7%), and HIV/AIDS services (36.3% versus 71.3%) (p≤0.01 for all). Indigent care clinics, crisis prevention, and enabling services (p≤0.01 for all) were more common at Black-serving hospitals as was Medicaid payer. Following adjustments for detailed hospital factors, the lowest Black serving hospital quartile carried the lowest risk for SMM. However, SMM risks were similar across the 2 nd (aRR 1.31, 95% CI 1.08, 1.59), 3 rd (aRR 1.27, 95% 1.05, 1.55), and 4 th (aRR 1.29, 95% CI 1.07, 1.55) quartiles.
Conclusion
Black-serving hospitals were more likely to provide a range of specialized medical, surgical, and safety-net services and to have a higher Medicaid burden. Payer mix and unmeasured confounding may account for some of the maternal risk associated with Black-serving hospitals.
Introduction
Racial disparities account for a substantial proportion of overall obstetric morbidity and mortality with non-Hispanic Black women at higher risk for adverse outcomes compared to other racial and ethnic groups. Prior studies across medical specialties have demonstrated that site of care may be an important factor in disparities; hospitals with large proportions of Black patients have higher risk for adverse outcomes including major morbidity and death. Observational research in obstetrics has demonstrated that delivering at a Black-serving hospital is associated with increased severe maternal morbidity (SMM) with Black women who deliver at such hospitals at highest risk.
Determining the cause of differential obstetric outcomes in Black-serving hospitals is an important goal in the overall effort to reduce disparities. Prior research has not clarified to what degree differentials in outcomes are due to incomplete case mix adjustment, hospital characteristics, or other structural factors. Prior administrative data research on maternal disparities has generally included limited hospital factors.
Why was this study conducted?
This study aimed to analyze detailed characteristics of Black-serving hospitals and how these characteristics are associated with risk for severe maternal morbidity.
Key findings
Black-serving hospitals were more likely to provide a range of specialized medical and surgical services and to have a less favorable payer mix.
What does this add to what is known?
Payer mix and unmeasured confounders may account for some of the maternal risk associated with Black-serving hospitals.
Given that more detailed evaluation of hospital factors related to obstetric outcomes may be important in understanding disparities, the purpose of this study was to analyze detailed characteristics of Black-serving hospitals and how these characteristics are associated with maternal outcomes.
Materials and Methods
Data source
The National (Nationwide) Inpatient Sample (NIS) from the Agency for Healthcare Research and Quality for the years 2010 to 2011 was used for this retrospective repeat cross sectional study. The NIS is one of the largest publicly available, all-payer inpatient databases in the United States, and is comprised of a sample of approximately 20% of hospitalizations in the US. The NIS includes both academic and community hospitals, as well as general and specialty-specific hospitals. Data after 2011 was not included in this analysis as the sampling approach for the NIS changed in 2012; full data from individual hospitals is not available from 2012 on because of changes in the sampling structure of the NIS.
Hospitals from the NIS were linked to the 2013 American Hospital Association (AHA) Annual Survey. The AHA Annual Survey contains detailed hospital information including facility characteristics, utilization, finance, hospital staffing, medical services offered, ownership, and management structure. These hospital characteristics are not reported in the NIS. The NIS includes limited data on characteristics including hospital bed volume (small, medium, and large), rurality (rural, urban), geographic location (Northeast, Midwest, South, or West), and teaching status (teaching, non-teaching). The AHA Annual Survey has been linked to other databases and used for health services and outcomes research in obstetrics and a wide range of other specialties.
Study population
Delivery hospitalizations in the NIS of women age 15 to 54 years were identified using the International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM) codes v27.x and 650 as these codes identify over 95% of delivery hospitalizations. Maternal race was analyzed by the following categories within the NIS: non-Hispanic White, non-Hispanic Black, Hispanic, Asian or Pacific Islander, other (including Native American), and unknown. Four quartiles with roughly equal numbers of deliveries were created based on the hospital-level proportion of non-Hispanic Black patients undergoing delivery hospitalization. For both 2010 and 2011, three states did not contribute data on race (Minnesota, Washington, and West Virginia in 2010 and Minnesota, North Dakota, and West Virginia in 2011). Hospitals in these states for these years were excluded. Hospitals in the NIS without identifiers allowing linkage to the AHA Annual Survey were excluded. Finally, hospitals performing <500 deliveries per year were also excluded.
Underlying comorbidity for individual patients was determined using an obstetric comorbidity index. This index provides weighted scores for comorbidity for individual patients based on the presence of specific diagnosis codes and demographic factors present in administrative data. Higher scores are associated with increased risk for severe morbidity. In the initial study validating the comorbidity index in a general obstetric population, patients with the lowest score of 0 had a 0.68% risk of severe morbidity whereas a score of >10 was associated with a risk of severe morbidity of 10.9%. This comorbidity index was subsequently validated in an external population. We categorized women based on comorbidity index scores: 0 (lowest risk), 1, 2, and ≥3 (highest).
Study objectives and outcomes
Primary objective
The primary objective of this study was to describe detailed characteristics of Black-serving hospitals in the United States. We aimed to determine how Black-serving hospitals differed from other hospitals in three ways: (i) whether Black-serving hospitals provided more specialized medical and surgical services, (ii) whether Black-serving hospitals provided more safety-net services, and (iii) whether Black-serving hospitals had a higher Medicaid burden. We evaluated these characteristics based on two hypotheses; we hypothesized that: (i) Black-serving hospitals were more likely perform a range of specialized, referral medical and surgical care, and (ii) Black-serving hospitals were more likely to care for patient populations enrolled in Medicaid and requiring safety-net services.
Based on data from the AHA Annual Survey, we determined whether hospitals provided the following specialized medical and surgical services: medical, surgical, cardiac, pediatric, and neonatal intensive care, psychiatric care including emergency services, cardiac services including interventional cardiology, hemodialysis, HIV and AIDS care, oncology services, transplant surgery, radiology services, organ transplant services, genetic testing and counseling, and other medical services. The AHA Annual Survey was also used to determine if hospitals provided the following safety-net services: crisis prevention services, enabling services, enrollment services, and indigent care clinics. Hospital ownership structure (government, not-for-profit non-government, for profit) was also determined based on AHA Annual Survey data.
Using data from the NIS, Medicaid burden was analyzed for each hospital. To determine Medicaid burden, the hospital-level proportion of delivery hospitalizations with Medicaid as the payer was determined. Hospitals were grouped into four approximately equal quartiles in terms of delivery volume based on the proportion of Medicaid deliveries. A second measure of Medicaid burden (not restricted to obstetric delivery hospitalizations) was additionally calculated evaluating all hospitalizations.
Secondary objective
The secondary objective of this study was to determine the relationship between Black-serving hospitals and risk for severe maternal morbidity (SMM) accounting for detailed hospital factors and patient-level comorbidities. We hypothesized that some of the maternal risk associated with Black-serving hospitals could be accounted for by Black-serving hospitals (i) being referral centers for specialized medical and surgical care, (ii) having higher Medicaid burdens, and (iii) having patient populations more likely to require safety-net services.
SMM was defined by criteria from the Centers for Disease Control and Prevention (CDC) that includes ICD-9-CM diagnosis and procedure codes for 21 indicators that represent potentially life-threatening illnesses and indicators of organ failure. For this analysis, we evaluated SMM excluding transfusion from the composite as transfusion, compared to other diagnoses in the composite, is much less likely to lead to long-term disability or death. The remaining 20 indicators were included in the analysis: acute myocardial infarction, aneurysm, acute renal failure, adult respiratory distress syndrome, amniotic fluid embolism, cardiac arrest, conversion of cardiac rhythm, disseminated intravascular coagulation, eclampsia, heart failure or arrest during surgery or procedure, hysterectomy, puerperal cerebrovascular disorders, pulmonary edema or acute heart failure, severe anesthesia complications, sepsis, shock, sickle cell disease with crisis, air and thrombotic embolism, temporary tracheostomy, and ventilation.
Statistical analysis
For the primary objective, hospital characteristics including specialized medical and surgical services, safety-net services, and Medicaid burden were compared between Black-serving hospital quartiles using the chi-squared tests or Fisher’s exact tests as appropriate with the Bonferroni correction applied for multiple comparisons.
For the secondary objective of determining the relationship between Black-serving hospitals and SMM, we first analyzed unadjusted risk between individual hospital and patient characteristics and risk for SMM. Then we performed three sequential adjusted models for risk for SMM with each model including additional data on hospital characteristics. For the first adjusted model for SMM, we included the following factors: (i) underlying patient comorbidity, (ii) Black-serving hospital quartile, (iii) obstetric Medicaid burden quartile, and (iv) maternal race and ethnicity.
For the second adjusted model for SMM, we then added hospital factors from the NIS (hospital bed volume, rurality, geographic location, and teaching status) and the AHA Annual Survey: (i) medical and surgical intensive care, (ii) neonatal intensive care, (iii) trauma center, (iv) genetic testing and counseling, (v) HIV/AIDS services, and (vi) transplant services. Hospital safety-net factors from the AHA Annual Survey included hospital ownership and the following: (i) crisis prevention services, (ii) enabling services, (iii) enrollment services, and (iv) indigent care clinics. These factors were chosen to reflect a broad range of services. However, not all hospital characteristics were included in this model because of concerns related to collinearity.
For the third adjusted model, principal component analysis was used to address potential concerns about collinearity. Principal component analysis is a method that reduces the dimensionality of the data while preserving most of the variation in the dataset. This approach identifies a list of principal components that maximizes the variation from variables of interest (37 hospital factors in the current analysis). Principal components are independent orthogonal linear combinations of the individual variables and are listed in decreasing order of proportion of explained variance. The first component accounts for as much variance as possible in the data. The next component will account for as much of the leftover variance as it can, given the assumption that it is uncorrelated with the previous components. The eigenvalue-one criterion was used to determine the number of components retained in the analysis where eigenvalues of components (the amount of variance that is accounted for by a given component) are 1.00 or greater. The hospital variables and component factors pattern are presented after varimax rotation that tend to maximize the variance of a column of the factor pattern matrix. The rotated factor pattern ( Supplemental Figure 1 ) identifies hospital variables that demonstrate high loading for a given component and determines what these variables have in common. The cut-off we used for loading a hospital variable to a factor is an absolute value of component loadings greater than 0.4. In subsequent analysis, identified principal components can be used as covariates in outcome regression models as was the case in this study in which principal components were analyzed together with patient-level comorbidity and health insurance status. For all four models for SMM, log linear regression models with a Poisson distribution based on the robust variance estimation method were performed with unadjusted (RR) and adjusted risk ratios (aRR) with 95% CIs as measures of effects. Results additionally accounting for hospital clustering with broader confidence intervals are presented. Given the de-identified nature of the data this study was deemed exempt by the Columbia University Institutional Review Board. All analyses were performed with SAS 9.4 (SAS Institute, Cary, NC).
Results
Of 1,559,523 delivery hospitalizations occurring at 699 hospitals to women age 15 to 54 in 2010-2011, 965,202 deliveries from 430 hospitals met inclusion criteria. Across the four quartiles of Black-serving hospitals, non-Hispanic Black patients accounted for 1.3%, 5.4%, 13.4%, and 33.8% of patients respectively ( Table 1 ).
| Demographics | Black-serving hospital quartile | P value | |||
|---|---|---|---|---|---|
| First (lowest), n (%) | Second, n (%) | Third, n (%) | Fourth (highest), n (%) | ||
| Race | |||||
| Non-Hispanic White | 140,549 (57.9) | 137,187 (56.9) | 112,767 (47.4) | 91,715 (37.6) | <.001 |
| Non-Hispanic Black | 3051 (1.3) | 12,923 (5.4) | 31,942 (13.4) | 82,415 (33.8) | |
| Hispanic | 60,888 (25.1) | 57,241 (23.8) | 60,661 (25.5) | 45,498 (18.7) | |
| Asian | 18,737 (7.7) | 13,611 (5.6) | 14,463 (6.1) | 8614 (3.5) | |
| Other | 9271 (3.8) | 13,930 (5.8) | 10,217 (4.3) | 12,998 (5.3) | |
| Unknown | 10,241 (4.2) | 6065 (2.5) | 7700 (3.2) | 2518 (1.0) | |
| Obstetrical comorbidity index | |||||
| 0 | 150,249 (61.9) | 145,783 (60.5) | 142,311 (59.9) | 141,396 (58.0) | <.001 |
| 1 | 62,293 (25.7) | 62,897 (26.1) | 63,098 (26.5) | 67,926 (27.9) | |
| 2 | 23,039 (9.5) | 24,167 (10.0) | 24,208 (10.2) | 25,262 (10.4) | |
| ≥3 | 7156 (2.9) | 8110 (3.4) | 8133 (3.4) | 9174 (3.8) | |
| Bed size | <.001 | ||||
| Small | 31,545 (13.0) | 23,982 (10.0) | 21,589 (9.1) | 20,001 (8.2) | |
| Medium | 71,552 (29.5) | 63,112 (26.2) | 42,060 (17.7) | 63,663 (26.1) | |
| Large | 139,640 (57.5) | 153,863 (63.9) | 174,101 (73.2) | 160,094 (65.7) | |
| Teaching hospitals | 50,618 (20.9) | 124,144 (51.5) | 131,434 (55.3) | 163,110 (66.9) | <.001 |
| Urban area | 215,760 (88.9) | 231,911 (96.2) | 230,476 (96.9) | 230,072 (94.4) | <.001 |
| Region | <.001 | ||||
| Northeast | 30,177 (12.4) | 59,306 (24.6) | 86,337 (36.3) | 52,877 (21.7) | |
| South | 5,607 (2.3) | 36,402 (15.1) | 77,372 (32.5) | 167,345 (68.7) | |
| West | 175,836 (72.4) | 113,130 (47.0) | 56,113 (23.6) | 5,967 (2.4) | |
| Midwest | 31,117 (12.8) | 32,119 (13.3) | 17,928 (7.5) | 17,569 (7.2) | |
| Obstetrical Medicaid burden | <.001 | ||||
| First quartile (lowest) | 86,296 (35.6) | 68,409 (28.4) | 60,642 (25.5) | 27,437 (11.3) | |
| Second quartile | 62,526 (25.8) | 60,574 (25.1) | 69,242 (29.1) | 47,563 (19.5) | |
| Third quartile | 43,332 (17.9) | 60,020 (24.9) | 53,744 (22.6) | 82,359 (33.8) | |
| Fourth quartile (highest) | 50,583 (20.8) | 51,954 (21.6) | 54,122 (22.8) | 86,399 (35.4) | |
Black-serving hospitals were more likely to provide a range of specialized medical and surgical services ( Table 2 ). Cardiac intensive care was present in 48.9% of hospitals in the lowest Black-serving quartile compared to 74.5% of hospitals in the highest Black serving quartile, neonatal intensive care in 28.9% versus 64.9%, pediatric intensive care in 20.0% versus 45.7%, psychiatric care in 49.6% versus 69.1%, pediatric cardiology in 29.6% versus 44.7%, genetic testing and counseling in 35.6% versus 58.5%, and HIV/AIDS services in 36.3% versus 71.3% (p≤0.01 for all). Black-serving hospitals were also more likely to perform higher risk deliveries, with patient comorbidity increasing significantly by Black-serving quartile (p<0.01). While statistical comparisons for some medical and surgical services were non-significant, there were no services that were more likely in the lowest compared to the highest Black-serving hospital quartile.
| Services | Black-serving hospital quartile | P value | |||
|---|---|---|---|---|---|
| First (lowest) (n=135) | Second (n=103) | Third (n=98) | Fourth (highest) (n=94) | ||
| Medical and surgical services (% of hospitals) | |||||
| Medical-surgical intensive care | 82.2 | 85.4 | 92.9 | 90.4 | .07 |
| Cardiac intensive care | 48.9 | 55.3 | 62.2 | 74.5 | <.01 a |
| Neonatal intermediate care | 50.4 | 60.2 | 66.3 | 70.2 | .01 |
| Neonatal intensive care | 28.9 | 39.8 | 46.9 | 64.9 | <.01 a |
| Pediatric intensive care | 20.0 | 31.1 | 39.8 | 45.7 | <.01 a |
| Psychiatric care | 49.6 | 59.2 | 68.4 | 69.1 | <.01 |
| Adult cardiology services | 79.3 | 82.5 | 89.8 | 86.2 | .16 |
| Pediatric cardiology services | 29.6 | 43.7 | 49.0 | 44.7 | .01 |
| Adult interventional cardiac catheterization | 62.2 | 76.7 | 84.7 | 79.8 | <.01 a |
| Adult cardiac surgery | 54.8 | 65.0 | 68.4 | 72.3 | .03 |
| Genetic testing and counseling | 35.6 | 49.5 | 57.1 | 58.5 | <.01 a |
| Hemodialysis | 74.8 | 86.4 | 87.8 | 86.2 | .02 |
| Certified trauma center | 48.2 | 47.6 | 50.0 | 61.7 | .15 |
| HIV and AIDS services | 36.3 | 53.4 | 63.3 | 71.3 | <.01 a |
| Oncology services | 80.7 | 82.5 | 86.7 | 86.2 | .56 |
| Neurology services | 73.3 | 82.5 | 85.7 | 83.0 | .08 |
| Magnetic resonance imaging | 88.9 | 88.3 | 91.8 | 89.4 | .86 |
| Ultrasound radiology services | 88.1 | 88.3 | 91.8 | 90.4 | .79 |
| Image-guided radiation therapy | 52.6 | 58.3 | 66.3 | 66.0 | .10 |
| Proton beam therapy | 48.9 | 61.2 | 58.2 | 63.8 | .10 |
| Transplant services—bone marrow | 17.8 | 25.2 | 26.5 | 33.0 | .07 |
| Transplant services—heart | 17.0 | 20.4 | 19.4 | 27.7 | .26 |
| Transplant services—kidney | 21.5 | 26.2 | 26.5 | 34.0 | .21 |
| Transplant services—liver | 15.6 | 23.3 | 16.3 | 29.8 | .04 |
| Transplant services—lung | 9.6 | 18.4 | 12.2 | 16.0 | .22 |
| Safety-net services (% of hospitals) | P value | ||||
| Indigent care clinic | 34.1 | 52.4 | 62.2 | 64.9 | <.01 a |
| Enrollment services | 77.0 | 79.6 | 86.7 | 80.9 | .32 |
| Crisis prevention | 43.0 | 50.5 | 62.2 | 61.7 | .01 |
| Enabling services | 38.5 | 50.5 | 54.1 | 61.7 | <.01 |
| Social work services | 87.4 | 89.3 | 87.8 | 89.4 | .95 |
| Ownership | P value | ||||
| Investor | 14.8 | 15.5 | 10.2 | 19.1 | <.01 |
| Nonprofit, nongovernment | 77.0 | 77.7 | 78.6 | 57.5 | |
| Government | 8.2 | 6.8 | 11.2 | 23.4 | |
| Obstetrical Medicaid burden (% of hospitals) | P value | ||||
| First quartile (lowest) | 25.2 | 22.3 | 16.3 | 7.4 | <.01 a |
| Second quartile | 28.1 | 28.2 | 26.5 | 16.0 | |
| Third quartile | 23.7 | 27.2 | 25.5 | 31.9 | |
| Fourth quartile (highest) | 23.0 | 22.3 | 31.6 | 44.7 | |
| All hospitalization Medicaid burden (% of hospitals) | P value | ||||
| First quartile (lowest) | 27.4 | 29.1 | 17.4 | 12.8 | <.01 a |
| Second quartile | 30.4 | 34.0 | 36.7 | 22.3 | |
| Third quartile | 23.0 | 20.4 | 23.5 | 27.7 | |
| Fourth quartile (highest) | 19.3 | 16.5 | 22.5 | 37.2 | |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
