Racial and Ethnic Differences Over Time in Outcomes of Infants Born Less Than 30 Weeks’ Gestation

We conducted a retrospective study on infants born between 22 and 29 weeks’ gestation at 1 of the 789 NICUs located in the United States or Puerto Rico and participating in the VON Very Low Birth Weight Database between January 1, 2006, and December 31, 2017. We restricted our study sample to inborn, singleton infants without congenital malformations. The University of Vermont’s committee for human research approved the use of VON’s deidentified research repository for this analysis.

Self-reported maternal race and/or ethnicity was collected by personal interview with the mother; if unavailable, the information was obtained through birth certificate or medical record review. We combined race and ethnicity resulting in the following: non-Hispanic African American (hereafter labeled African American), non-Hispanic white (labeled white), Hispanic, Asian American, American Indian, and other race.

We examined changes over time in maternal morbidities, care practices, and newborn mortality and morbidities. Maternal morbidities included chorioamnionitis and hypertension (chronic or pregnancy induced). We divided care practices into obstetric practices (antenatal corticosteroids [ANSs], cesarean delivery), delivery room (DR) practices (tracheal intubation, epinephrine, cardiac compressions, face mask ventilation, surfactant), and NICU practices (steroids for chronic lung disease [CLD], surfactant therapy at any time, which also includes surfactants if received in the DR, conventional ventilation, ventilation after continuous positive airway pressure [CPAP], high-frequency ventilation, nasal ventilation, CPAP before ventilation, nasal CPAP, high-flow nasal cannula, inhaled nitric oxide, postnatal life support [surfactant therapy at any time, endotracheal tube ventilation, ventilator support at any time, epinephrine, or cardiac compressions]). All the above variables were available for the entire period (2006–2017) except for chorioamnionitis, hypertension, and inhaled nitric oxide, which were added in 2008.

Mortality was defined as death before hospital discharge. Hypothermia was defined as the infant’s body temperature <36.5°C measured within the first hour of NICU admission.¹⁶ Respiratory distress syndrome (RDS) was defined as room air Pao₂ <50 mm Hg, room air with central cyanosis, supplemental oxygen to maintain Pao₂ >50 mm Hg, or supplemental oxygen to maintain a pulse oximeter saturation >85% and a chest radiograph consistent with RDS within the first 24 hours of life.¹⁶ Necrotizing enterocolitis (NEC) was diagnosed at surgery or postmortem or required ≥1 clinical sign (eg, bilious gastric aspirate, abdominal distension, occult blood in stool) and ≥1 radiographic finding (eg, pneumatosis intestinalis, hepatobiliary gas, or pneumoperitoneum).¹⁶ Focal intestinal perforation, separate from NEC, was defined as a single focal perforation with the remainder of the bowel appearing normal, diagnosed at surgery time or postmortem examination.¹⁶ NEC and focal intestinal perforation were combined into 1 outcome labeled NEC. Early-onset sepsis (EOS; on or before day 3 of life) was defined as bacterial pathogen recovered from blood or cerebrospinal fluid.¹⁶ Late-onset sepsis (LOS; after day 3 of life) was defined as bacterial pathogen or coagulase-negative Staphylococcus recovered from blood or cerebrospinal fluid or fungus recovered from blood culture.¹⁶ Coagulase-negative Staphylococcus infection also required ≥1 sign of generalized infection and treatment with ≥5 days of intravenous antibiotics.¹⁶ Severe intraventricular hemorrhage (sIVH) was defined as grades 3 or 4 using Papile’s classification within 28 days of birth.¹⁷ Severe retinopathy of prematurity (sROP) was defined as stages 3 to 5 based on a retinal examination before hospital discharge.¹⁸ CLD was defined as any supplemental oxygen use at 36 weeks’ postmenstrual age or on oxygen at discharge at 34 to 35 weeks if transferred or discharged <36 weeks’ gestation.¹⁶ We also examined survival to discharge without a major neonatal morbidity defined as ≥1 of early or late bacterial or fungal infection, NEC/focal intestinal perforation, CLD, sIVH, cystic periventricular leukomalacia (multiple small periventricular cysts on a cranial ultrasound, computed tomography, or MRI), or sROP. Reporting hospitals tracked mortality before hospital discharge and after transfer until ultimate disposition.

We computed summary statistics of maternal and neonatal characteristics and care practices by maternal race and/or ethnicity. We subsequently calculated the following:

1. incidence rates of maternal morbidities, obstetric practices, DR practices, NICU practices, and other practices by race and/or ethnicity and birth year along with 95% confidence intervals (CIs);

2. incidence rates of infant mortality and morbidity by race and/or ethnicity and birth year along with 95% CIs; and

3. differences between rates for African American and Hispanic versus white infants by birth year, along with 95% CIs.

We tested differences in trends over time between white and African American or Hispanic infants for mortality and morbidity rates using likelihood ratio tests (LRTs) on nested binomial regression models. The null model included main effects for race and/or ethnicity and a third-degree polynomial on birth year. The alternative model included the same main effects as well as an interaction between them. No adjustments were made for other variables related to medical care to prevent masking of temporal effects.

Similar analyses were conducted separately to compare white and Asian American infants, provided as Supplemental Information. Statistical significance was set at 5%. Analyses were performed by using R¹⁹ (version 3.2.2.) and SAS software (version 9.4; SAS Institute, Inc, Cary, NC).

We assessed the sensitivity of the results when restricting the study sample to centers participating in VON throughout the entire 12-year period. The restricted sample had 414 centers and 75.5% (n = 165 432) of the infants.

Of 226 283 singleton infants born at VON centers, we excluded infants with unknown sex (n = 37), missing race and/or ethnicity (n = 935), and missing (n = 126) or implausible birth weight (n = 912). Implausible birth weights were defined as values below or above gestational age–specific cutoffs, determined by using a method based on data sparsity.^20,21 Overall, 1986 infants were excluded, resulting in 224 297 infants. Given the limited sample size, we further excluded infants of American Indian (n = 1628) or other (n = 3535) race. This resulted in 219 134 infants for our main analyses; 40.6% white, 34.8% African American, 20.4% Hispanic, and 4.2% Asian American. Table 1 shows maternal and newborn characteristics and care practices by race and/or ethnicity. African American mothers had the highest rates of hypertension and chorioamnionitis. African American and Hispanic mothers had the lowest rates of ANSs.

Chorioamnionitis rate slightly increased over time with African American mothers having the highest rate. The rate of maternal hypertension showed an increasing trend over time for all groups but was highest for African American mothers and lowest for Hispanic mothers (Supplemental Fig 4).

Use of ANSs increased over time for all groups, and although the increase was more prominent for African American and Hispanic than white mothers (% change 2006–2017 white: 13.0%, African American: 18.7%, Hispanic: 21.2%), the rate remained highest for white mothers. Cesarean delivery rate increased over time for all groups but remained highest for white mothers (Supplemental Fig 5).

DR tracheal intubation, epinephrine, cardiac compression, and surfactant therapy rates decreased, whereas face mask ventilation rate increased over time for all groups (Fig 1). Tracheal intubation rate was higher for African American than Hispanic and white infants, whereas DR surfactant rate was lowest among Hispanic infants. Supplemental oxygen use remained relatively constant over time with similar rates for all groups.

Nasal CPAP use revealed positive trends over time with slightly lower rates among African American infants (Fig 2). Nasal ventilation and high-flow nasal cannula use also increased over time with relatively similar rates for all groups. Conventional ventilation decreased over time for all groups of infants, but rates were higher among African American (2017 rate: 71.6%) than white (69.6%) and Hispanic (67.8%) infants. High-frequency ventilation rate was relatively constant over time with similar rates among all groups. CPAP before assisted ventilation increased over time but revealed lower rates for African American (2017 rate: 48.7%) than Hispanic (57.6%) and white (54.9%) infants. Ventilation after CPAP revealed a negative trend followed by a positive trend with slightly lower rates among African American infants. Inhaled nitric oxide use was relatively constant over time with lower rates among Hispanic infants. Although the rate of steroids for CLD increased over time for all infants, it remained lowest for Hispanic (2017 rate: 13.8%) infants (white: 17.0%, African American: 16.2%). Surfactants decreased for all infants, with the lowest rates observed for Hispanic (2017 rate: 69.1%) infants (white: 73.5%, African American: 72.2%). Postnatal life support rates increased for all groups.

Supplemental Table 2 shows the overall rates of outcomes by race and/or ethnicity. Mortality and morbidity rates across the years are shown in Supplemental Fig 6, separately for each group. Rate differences between African American and Hispanic versus white infants during the study period are shown in Fig 3.

For African American versus white infants, RDS, EOS, sIVH, sROP, CLD, pneumothorax, and morbidity-free survival revealed a constant rate difference over time. EOS rates revealed occasionally slightly higher rates for African American infants, whereas sIVH was higher among African American than white infants for most of the years. RDS and sROP were lower among African American infants for some but not all years. Hypothermia was higher, whereas CLD and pneumothorax were lower among African American infants throughout the study period. Morbidity-free survival was higher among African American infants for all years except one. The rate difference for mortality, hypothermia, NEC, and LOS remained higher for African American infants throughout the study period. However, the rates for these outcomes decreased faster over time for African American than white infants with a significant year-race interaction (LRT P < .0001, mortality; P < .0001, hypothermia; P = .028, NEC; P < .0001, LOS).

For Hispanic versus white infants, hypothermia, NEC, EOS, LOS, sIVH, sROP, CLD, and morbidity-free survival revealed a constant rate difference over time. NEC, EOS, and morbidity-free survival revealed occasionally slightly higher rates for Hispanic than white infants. Mortality, LOS, sIVH, and sROP revealed higher rates among Hispanic infants for some but not all years. RDS and pneumothorax were lower among Hispanic infants for some but not all years. CLD rate was lower among Hispanic infants for the whole study period. The rates for mortality, RDS, and pneumothorax decreased faster over time for Hispanic than white infants, with a significant year-race interaction (LRT P = .032, mortality; P < .0001, RDS; P = .0016, pneumothorax).

Similar results were obtained after restricting the analysis to VON centers that participated for all 12 years, although the results were strengthened for the year-race interaction among African American infants for NEC (P = .0049) and among Hispanic infants for mortality (P = .020) (Supplemental Fig 7).

For Asian American infants, maternal morbidities, obstetric, DR, and NICU, are reported in Supplemental Figs 8–11, respectively. Briefly, trends for Asian American infants were as those reported for white infants. Maternal hypertension and cesarean delivery were higher for white mothers. Asian American infants had lower rates of DR tracheal intubation and surfactants. For NICU practices, Asian American infants had lower rates of conventional ventilation, steroids for CLD, and surfactants and higher rates of CPAP use before ventilation.

Mortality and morbidity rates for Asian American and white infants during the study period are shown in Supplemental Fig 12. Rate differences between Asian American versus white infants are shown in Supplemental Fig 13. Few to no significant rate differences across the study period among Asian American versus white infants were observed for mortality, hypothermia, NEC, EOS, LOS, and sIVH. For some but not all years, Asian infants had higher rates for sROP and morbidity-free survival and lower rates for RDS, CLD, and pneumothorax. Only RDS revealed a significant race-year interaction in both the main and the sensitivity analysis restricted to VON centers participating for the whole study period (P = .013, main analysis; P = .028, sensitivity analysis).

In a study of >215 000 infants born at 22 to 29 weeks’ gestation between 2006 and 2017 in the United States, we report that the disparity for certain care practices and outcomes have narrowed over time for African American and Hispanic infants. Compared with white infants, African American infants had a faster decline for mortality, hypothermia, NEC, and LOS, whereas Hispanic infants had a faster decline for mortality, RDS, and pneumothorax. However, by 2017, mortality and several morbidities remained elevated, especially for African American infants.

Disparate patterns in the site of delivery by race and/or ethnicity might contribute to the observed disparities in newborn care practices and outcomes. Several studies in the adult literature report that minority racial and/or ethnic patients disproportionately receive care at the lowest quality hospitals and care providers.^22,–26 Recent studies also reveal that this is applicable to the perinatal health field. In New York City, the quality of the birth hospital explained 40% of the African American–white disparity and 30% of the Hispanic-white disparity in hospital adverse outcomes of very preterm births, but there was no evidence of within-hospital disparities.²⁷ In California NICUs, there was evidence of racial and/or ethnic disparities in quality of care between and within NICUs.²⁸ A national study from VON showed that African American, Hispanic, and Asian American preterm infants were segregated across NICUs, with large differences in NICU quality between geographic regions.²⁹ Compared with white infants, Asian American and Hispanic infants were concentrated at higher quality NICUs, whereas African American infants received care at lower quality NICUs.²⁹ Accounting for region of residence explained the association for Hispanic but not for Asian American or African American infants.²⁹ 

Despite the racial and/or ethnic disparities in which minority (especially African American infants) might receive care, our study suggests a positive outlook. The narrowing of the gap in the rate of evidence-based care practices such as ANSs and the decrease in rates of mortality and several morbidities such as hypothermia and LOS among minority infants suggests that care among these infants has been improving. In a recent study, Horbar et al⁶ examined the change in the performance of 756 VON NICUs admitting very low birth weight infants between 2005 and 2014 and reported that rates of in-hospital death and severe morbidities declined. By 2014, >98% of all NICUs achieved the risk-adjusted mortality rate of the best 10% of units in 2005, and 75% of all NICUs achieved performance rates equivalent to the best quartile of the 2005 benchmark for all severe morbidities (NEC, LOS, sIVH, sROP) except CLD.⁶ The application of evidence-based care practices through quality improvement (QI) methods has been cited as a potential explanation for this progress.⁶ Indeed, QI initiatives have been successful in improving care and reducing adverse newborn outcomes.^30,–36 LOS has been a major focus of several of these initiatives, with evidence suggesting that infection rates in NICUs can be reduced.^37,–41 The VON has been a pioneer in QI collaboratives with >550 NICUs participating in such collaboratives since 1995.⁶ Given that minority-serving NICUs have lower quality of care than non–minority-serving NICUs,^27,29,42,43 one would hypothesize on the basis of the above study⁶ and our findings that the improvement in performance has also occurred among minority-serving NICUs.

Although we depict in our study several encouraging findings, further progress is warranted. Human milk at discharge based on VON data remains suboptimal among African American infants,⁴⁴ potentially contributing to disparities in NEC and LOS rates.^45,–49 Collaborative QI initiatives in California and 4 Southern US states have been successful in demonstrating a significant and sustained increase in breast milk feeding rates, which can decrease racial inequities.^49,50 Similarly, targeted interventions can reduce hypothermia rates,^51,52 which were highest among African American infants by the end of the study period. Further improvements in ANS rates among African American and Hispanic infants can also contribute to decreasing sIVH.⁵³ 

We are uncertain how racial and/or ethnic differences in response to life-saving treatments⁵⁴ and usage of more invasive approaches to respiratory support might have contributed to outcomes such as sIVH and CLD. Our findings suggest that African American infants have higher rates of DR intubation and lower rates of CPAP use before ventilation than white infants. Previous studies have revealed that avoidance of DR intubation when possible and decreasing the number of intubation attempts can reduce sIVH.^53,55 For CLD, potentially better practices include avoiding intubation by using early CPAP, by using surfactants selectively, and minimizing duration of mechanical ventilation.⁵⁶ However, QI initiatives have not been as successful in reducing CLD rates contrary to the declines observed for other morbidities.^4,6 We show in our study, similar to previous studies, that African American compared with white infants have lower CLD and pneumothorax rates.^28,57 However, despite their lower CLD rates, African American infants develop worse longer-term respiratory outcomes than white infants.⁵⁸ 

Strengths of our study include a large sample size with data on care practices and outcomes of >215 000 infants over a 12-year period. These data represent 88% of all US births at 22 to 29 weeks’ gestation. Limitations of our study include lack of data on causes of death to probe whether deaths related to RDS, NEC, and LOS also decreased over time. We also realize that race and/or ethnicity are complex social constructs,⁵⁹ with substantial heterogeneity in health outcomes within Hispanic and Asian American subgroups.^12,15,60 

Between 2006 and 2017, the disparity gap for care practices and certain outcomes narrowed between minority and white infants. However, by 2017, disparities in care practices and several outcomes remained, especially for African American infants. That the quality deficit among minority infants occurred for several care practice measures and potentially modifiable⁶ outcomes suggests a critical role for QI initiatives tailored for minority-serving hospitals. In the absence of effective preterm birth prevention strategies, it is imperative that researchers of future studies continue to assess care and outcome trends to ensure that infants of all racial and ethnic backgrounds benefit equally from medical developments.

We thank our colleagues who submit data to the Vermont Oxford Network on behalf of infants and their families. Participating centers are listed in Supplemental Table 3.

ANS

antenatal corticosteroid

CI

confidence interval

CLD

chronic lung disease

CPAP

continuous positive airway pressure

DR

delivery room

EOS

early-onset sepsis

LOS

late-onset sepsis

LRT

likelihood ratio test

NEC

necrotizing enterocolitis

QI

quality improvement

RDS

respiratory distress syndrome

sIVH

severe intraventricular hemorrhage

sROP

severe retinopathy of prematurity

VON

Vermont Oxford Network