Racial and Ethnic Disparities in Medical Complexity and In-Hospital Death Among US-Born VLBW Infants Free

We performed a retrospective, cross-sectional analysis of hospital discharge data (2009 and 2012) from the Healthcare Cost and Utilization Project’s Kids’ Inpatient Database. The Kids’ Inpatient Database is the largest multistate, nationally representative database for pediatric hospitalizations. It includes randomly sampled discharges from acute care hospitals in 44 states and is released every 3 years. The dataset is weighted for each discharge to produce national estimates of pediatric inpatient resources.¹²  The dataset includes variables regarding infants’ sociodemographic information, diagnoses, procedures, length of stay, and hospital characteristics, outlined below. Survey data are weighted to account for the complex survey design, nonresponse, and noncoverage. This project was reviewed by our university’s institutions review board and deemed exempt.

Our cohort included infants with the International Classification of Diseases, Ninth Revision (ICD-9) code indicating very low birth weight (<1500 g) and age at admission of 0 years.¹³  We included infants with complete demographic information, including age, birth weight, gestational age (completed weeks), race and ethnicity, transfer status, sex, insurance status, income quartile, hospital information, and length of stay. The cohort included infants in their birth hospitalization or those transferred in with a final discharge status to home, home health care, or died. Detailed selection criteria were previously published.⁴  The cohort selection is summarized in Fig 1. To ensure no significant biases were introduced by the exclusion of infants with missing data, a brief analysis of excluded patients was conducted (Supplemental Table 4).

The main exposure used in this analysis was race and ethnicity. Race and ethnicity were captured through documentation in the medical record and categorized as Hispanic, NHB, non-Hispanic White (NHW), or other. Race and ethnicity were reported in the data as Asian or Pacific Islander, Black, Hispanic, Native American, White or other (none of the previously listed choices). Asian or Pacific Islander, Native American, and other were combined into a single category, “Other,” for purposes of our analysis.

The primary outcome of interest was the presence of medical complexity or in-hospital death, given that some infants would die before being formally diagnosed with conditions indicating medical complexity. Medical complexity was defined by using the CCC scheme, a published set of diagnoses identified using ICD-9 codes that indicate conditions expected to last >12 months, involving either several organ systems or 1 organ system severely enough to require specialty pediatric care and, likely, a period of hospitalization,^14,15  and has been used extensively to describe children with medical complexity.^3,16,17 

Covariates included demographic, clinical, and hospital factors. Demographic variables included infant sex, insurance type (self-pay or no insurance, Medicaid or Medicare, or private insurance), patient zip code location (large metro/fringe, medium/small metro, or very small metro/rural), and household income quartile. Hospital characteristics included hospital region (defined by US Census region: Northeast, Midwest, South, or West), size (small, medium, or large based on the number of short-term acute care beds), type (government, private and nonprofit, or private and investor-owned), and teaching status stratified by rural/urban location (rural nonteaching, urban teaching, or urban nonteaching). Clinical characteristics included whether the infant was outborn, the median number of procedures (captured by ICD-9 procedure codes), having a procedure that occurred in the operating room (OR), gestational age, birth weight, final disposition (routine/home, died in hospital, home health care, other), length of stay, and total hospital charges.

Data were weighted to account for the complex sampling design as provided by the Healthcare Cost and Utilization Project.¹⁸  Prevalence based on a weighted linear model with the covariates of birth weight and race and ethnicity, along with corresponding 95% confidence intervals, were calculated for demographic variables and outcomes. Multiple logistic regression was used to model the association between relevant patient and hospital characteristics and the outcome of CCC or death, both to calculate adjusted odds ratios for each racial and ethnic group and using interaction terms to account for effect modification by racial and ethnic group. Covariates used in the model included birth weight, race and ethnicity, sex, insurance type, teaching hospital status, hospital region, hospital type, OR procedure, and transfer status. All variables were considered for inclusion and in a stepwise fashion, some variables were excluded in the modeling process because of nonsignificant relative to other patient characteristics. P values <.05 were considered statistically significant. All analyses were conducted by using SAS V9.4 (Cary, NC) and SUDAAN 11.03 (Research Triangle Park, NC).

The cohort included a total of 54 825 VLBW infants, representing a weighted total of 73 693 VLBW infants. Table 1 summarizes the weighted prevalence of demographic, hospital, and clinical characteristics for the overall study population and by race and ethnicity. The majority of VLBW infants weighed 1000 g to 1499 g at birth. NHB accounted for 34.3% of VLBW infants with a birth weight of <500 g, and among all NHB VLBW infants, 11.4% weighed <500 g. Patients excluded from the analysis are summarized in Supplemental Table 4.

Table 2 reveals the prevalence of CCC or death by birth weight category (<500 g, 500 g–999 g, or 1000 g–1499 g), in which we demonstrated a difference in the prevalence by race and ethnicity for all birth weight groups. For infants weighing <500 g, there was a significant difference in the prevalence of CCC or death with NHB infants having a weighted prevalence of 96.1% (95% confidence interval [CI], 95.3%–97.0%) and NHW infants having a weighted prevalence of 92.2% (95% CI, 91.4%–93.1%). This disparity did not persist in the birth weight categories of 500 g to 999 g and 1000 g to 1499 g, however, there continued to be a significantly different prevalence across racial and ethnic groups (Table 2).

Multivariable results for patient characteristics for each race and ethnicity category are shown in Table 3. We found that females of all races and ethnicities had a lower adjusted odds of CCC or death compared with males with no significant difference based on racial and ethnic group (P = .17). There was a variable association of public versus private insurance depending on race and ethnicity, whereas NHW, Hispanic, and other races were all equally likely to have CCC or death if they had public versus private insurance. NHB infants with public insurance, however, were less likely than NHB infants with private insurance to have CCC or death (P < .001). When examining teaching status, NHW, Hispanic, and infants of other races and ethnicities were all less likely to have CCC or death in a nonteaching versus teaching hospital, but NHB infants were equally likely with a significant difference in odds across racial and ethnic groups (P = .01).

For hospital factors, we found heterogeneity in the association of hospital region and CCC or death by race and ethnicity. NHW, NHB, and infants of other races and ethnicities were equally likely to have CCC or death in the Midwest compared with the Northeast; however, Hispanic infants were more likely to have CCC or death (Table 3, AOR 1.68, 95% CI, 1.37–2.05). When comparing the South to the Northeast for each race and ethnicity group, NHW and NHB infants were equally likely to have CCC or death, but Hispanic infants and infants of other races and ethnicities were more likely to have CCC or death in the South compared with the Northeast, again with a significant difference across racial and ethnic groups (P < .001).

For clinical characteristics, infants who had a procedure in the OR, were transferred in, and who had lower birth weights were more likely to experience CCC or death across all race and ethnicity categories. However, the magnitude of the odds for each racial and ethnic group varied significantly for those who had an OR procedure (P < .001) and those with a birth weight of <500 g compared with 1000 g to 1499 g (P < .001). Notably, although infants of all races who weighed <500 g compared with infants weighing 1000 g to 1499 g were more likely to have CCC or death, the magnitude of the odds varied significantly across races and ethnicities with NHB infants weighing <500 g having 63.6 times the adjusted odds of CCC or death compared with NHB infants weighing 1000 g to 1499 g. For infants weighing 500 g to 999 g compared with 1000 g to 1499 g, the odds were similar across all racial and ethnic groups (Table 3, P = .91).

The overall adjusted odds for CCC or death among NHB infants were 0.94 (95% CI, 0.89–0.98, Supplemental Table 6) compared with NHW infants. When further assessing the odds of CCC or death for clinical characteristics compared between specific racial and ethnic groups, we found that, among infants who had an OR procedure compared with those who did not, Hispanic infants were more likely to experience CCC or death than NHW infants (AOR 1.38, 95% CI, 1.20–1.59). NHB and infants of other races and ethnicities were equally likely (Supplemental Table 4. When comparing infants weighing <500 g to those weighing 1000 g to 1499 g at birth, NHB infants had 2.3 times higher odds of experiencing CCC or death compared with NHW infants (AOR 2.32, 95% CI, 1.57–3.42). There was no difference in odds of CCC or death in Hispanic and other race and ethnicity groups compared with NHW infants across birth weight categories (Supplemental Table 4).

In this population-based study of US-born VLBW infants, we demonstrate that CCC or in-hospital death is common across all races and ethnicities. Particular differences were demonstrated for certain racial and ethnic groups in the odds of CCC or death by sex, insurance type, hospital region, and birth weights. In particular, despite overall lower odds of CCC or death for NHB infants compared with NHW infants, for infants born weighing <500 g relative to 1000 g to 1499 g, NHB infants have disproportionately higher odds of CCC or death than NHW infants.

The results of this study reveal the importance of stratifying or disaggregating data to understand trends in perinatal health outcomes by specific maternal and infant characteristics. Analyzing perinatal health data by race and ethnicity is a critical first step to identifying and then addressing disparities and inequities. A large body of literature has revealed the inequitable burden of morbidity and mortality experienced by certain racialized minority groups. The rate of preterm birth among Black mothers has been shown repeatedly to be higher than among White women, even after controlling for socioeconomic and maternal health factors and despite acknowledgment of these disparities and ongoing public health efforts including targeting modifiable risk factors and populations at the highest risk.^6,19–22  Additionally, Black infants have a higher infant mortality rate which is explained only in part by disparities in rates of prematurity and low birth weight, and, subsequently, those who survive experience an increase in certain neonatal morbidities.^23,24  Our study is the first, to our knowledge, to also reveal racial and ethnic differences in overall medical complexity among VLBW infants.

Conversely to these well-established disparities affecting Black mothers and infants, some recent studies reveal conflicting findings of lower odds of certain morbidities among NHB infants. For example, findings of decreased rates of bronchopulmonary dysplasia have been demonstrated in NHB infants,²⁵  and this is postulated to be due to a variety of factors related to the accuracy of data elements, namely that gestational age dating for infants born to NHB mothers may be more inconsistent.²⁶  Particular methods have been proposed to account for these discrepancies, including the fetus-at-risk model, which assesses for outcomes at a given gestational age among both delivered and ongoing (not delivered) pregnancies, demonstrating a significantly greater risk for adverse outcomes for NHB infants.²⁷  In the setting of the methodological considerations with these otherwise conflicting reports of neonatal outcomes, our data are important to further expand the repository of data investigating perinatal racial and ethnic disparities.

Our findings of overall lower adjusted odds of CCC or death among NHB infants compared with NHW infants, yet disproportionately higher odds of CCC or death for NHB infants with birth weights of <500 g, highlight the need for even greater in-depth analyses of outcomes by race and ethnicity. Egbe et al recently demonstrated the importance of stratifying rates of preterm birth not only by race and ethnicity but also by nativity, to provide more specific analyses that can inform targeted interventions.²⁸  Our work addressing a specific area of perinatal health outcomes adds to this growing body of literature necessitating more granular analyses when reporting outcomes by race and ethnicity.²⁹ 

In addition to our findings regarding significant racial and ethnic disparities by birth weight, we also demonstrate differences in the prevalence of CCC or death by region based on race and ethnicity, highlighting that maternal–infant dyads from certain racial and ethnic groups may be disproportionately affected in certain regions. Higher maternal mortality and preterm birth rates have been demonstrated in the South,^19,30  but racial and ethnic variation by region is less well-described. Our findings of varying odds by race and ethnicity in certain regions, but not others, bring to light several possible explanations. For instance, studies have examined the effect of segregation of racial and ethnic groups on adverse health outcomes such as limited access to high-quality medical care, which disproportionately impact Black patients.³¹  In addition, Black mothers and infants are more likely to be cared for in lower-quality hospitals.³²  Our study is the first, to our knowledge, that has revealed regional-level differences in infant outcomes comparing racial and ethnic groups. To address region-specific inequity in perinatal health outcomes, some statewide perinatal quality collaboratives have developed initiatives focused specifically on understanding and reducing racial and ethnic disparities, particularly in the birthing hospital care of mothers and newborns. For instance, Parker et al implemented interventions in Massachusetts NICUs specifically aimed at reducing racial and ethnic disparities in the provision of human milk to very low birth weight infants.³³  State perinatal quality collaboratives offer a potentially powerful opportunity at the regional level to understand and address barriers to reducing racial and ethnic disparities in perinatal care processes and outcomes.^34–36 

Our study has several limitations. First, given the retrospective study design, we were limited by the availability of already-collected data elements. Moreover, as with all studies based on administrative billing data, reliability and missing data may not occur at random across comparison groups. We excluded infants who had missing data for race and ethnicity; characteristics of those excluded are provided in Supplemental Table 5. Although the included cohort and those infants excluded because of missing race and ethnicity had similar birth weights, gestational ages, sex, insurance status, and median household income, there were differences in the distribution of hospital region, patient residence, and transfer status. Second, race and ethnicity data were collected from medical records and details about how these data were collected and entered are not available. Although parental self-report of infant race and ethnicity would be the gold standard, we hypothesize that this likely did not occur at the majority of hospitals that contribute data to the Kids’ Inpatient Database. Additionally, the classification scheme of CCC was also not specifically developed for the neonatal population, and, despite being updated to include neonatal diagnoses,¹⁴  may exclude some patients whom clinicians would otherwise likely deem to have medical complexity given that functional limitations and psychosocial characteristics that can contribute to medical complexity would not be captured.³⁷  Thus, we believe that the definition we used is more likely a conservative estimate of medical complexity. Lastly, it is possible that some degree of the demonstrated differences by race and ethnicity in the rates of CCC or death is driven by differences in types and frequency of specific neonatal conditions experienced by each racial and ethnic group. Additional studies are needed that investigate the contribution of the incidence of specific conditions within racial and ethnic groups to disparities in neonatal outcomes.

In conclusion, we demonstrate that significant racial and ethnic disparities exist in CCC or in-hospital death among VLBW infants at the national level. Given that children with medical complexity represent the highest ongoing utilizers of pediatric healthcare,³⁸  understanding racial and ethnic disparities that begin in the perinatal period is essential to address the factors that may disadvantage some minority populations of newborns. Additional investigation is needed into the drivers of these disparities so that efforts at hospital, community, regional, and national levels can be developed to achieve perinatal health equity in care processes and outcomes.