Hospital Variation and Risk Factors for Bronchopulmonary Dysplasia in a Population-Based Cohort

Importance  Bronchopulmonary dysplasia (BPD) remains a serious morbidity in very low-birth-weight (VLBW) infants (<1500 g). Deregionalization of neonatal care has resulted in an increasing number of VLBW infants treated in community hospitals with unknown impact on the development of BPD.

Objective  To identify individual risk factors for BPD development and hospital variation of BPD rates across all levels of neonatal intensive care units (NICUs) within the California Perinatal Quality Care Collaborative.

Design, Setting, and Participants  Retrospective cohort study (January 2007 to December 2011) from the California Perinatal Quality Care Collaborative including more than 90% of California’s NICUs. Eligible VLBW infants born between 22 to 29 weeks’ gestational age.

Exposures  Varying levels of intensive care.

Main Outcomes and Measures  Bronchopulmonary dysplasia was defined as continuous supplemental oxygen use at 36 weeks’ postmenstrual age. A combined outcome of BPD or mortality prior to 36 weeks was used. Multivariable logistic regression accounting for hospital as a random effect and gestational age as a risk factor was used to assess individual risk factors for BPD. This model was applied to determine risk-adjusted rates of BPD across hospitals and assess associations between levels of care and BPD rates.

Results  The study cohort included 15 779 infants, of which 1534 infants died prior to 36 weeks’ postmenstrual age. A total of 7081 infants, or 44.8%, met the primary outcome of BPD or death prior to 36 weeks. Combined BPD or death rates across 116 NICUs varied from 17.7% to 73.4% (interquartile range, 38.7%-54.1%). Compared with level IV NICUs, the risk for developing BPD was higher for level II NICUs (odds ratio, 1.23; 95% CI, 1.02-1.49) and similar for level III NICUs (odds ratio, 1.04; 95% CI, 0.95-1.14).

Conclusions and Relevance  Bronchopulmonary dysplasia or death prior to 36 weeks’ postmenstrual age affects approximately 45% of VLBW infants across California. The wide variability in BPD occurrence across hospitals could offer insights into potential risk or preventive factors. Additionally, our findings suggest that increased regionalization of NICU care may reduce BPD among VLBW infants.

Bronchopulmonary dysplasia (BPD) is the most prevalent serious morbidity in preterm infants.^1-4 First described by Northway and colleagues⁵ in 1967, BPD traditionally focused on severe lung injury from assisted ventilation and oxygen of premature noncompliant lungs. The pathophysiology of BPD was altered following the implementation of surfactant therapy; however, BPD remains a highly prevalent condition in very low-birth-weight (VLBW) infants (<1500 g) including those who require minimal to moderate support during the early postnatal days.^6-8 Because survivors with BPD have significant pulmonary and nonpulmonary morbidities, including cerebral palsy and growth, developmental, and academic difficulties, it is imperative to optimize care delivery systems, including regional referral of infants likely to be affected.^9-13 Identification of risk factors would possibly allow for directed therapies toward reducing the likelihood or the severity of BPD to improve long-term outcomes.

The etiology of BPD is multifactorial involving individual genetic predispositions in combination with prenatal and postnatal environmental influences.^14,15 Some of the most commonly known risk factors include lower birth weight (BW), younger gestational age (GA), male sex, fetal growth restriction, and prolonged ventilator–induced injury.¹⁶ Medical care practices, as part of the postnatal environment, are likely to influence the development of BPD. Prior studies supporting this hypothesis identified variations in BPD in both the pre- and post-surfactant era, focusing on regional perinatal centers.^17-21 With a growing number of neonatal intensive care units (NICUs), VLBW infants increasingly receive care in midlevel low-volume units, raising concerns as VLBW mortality is greater among those infants born outside of regional and tertiary care centers.^22-27 The impact of the deregionalization of neonatal care on the specific risk for developing BPD is presently unknown.

The purpose of this study was to identify independent risk factors for the development of BPD and the extent of hospital variation in BPD rates in a population-based cohort. This was a means to ultimately examine the effects of neonatal level of care on BPD rates among VLBW infants.

This study used data from the California Perinatal Quality Care Collaborative (CPQCC) from January 2007 to December 2011. The CPQCC prospectively collects clinical data in greater than 90% of VLBW infants receiving NICU care in California. Membership is offered to all hospitals in California providing neonatal intensive care. During the study, there were 132 member hospitals. This study was approved by the Stanford University institutional review board; parental consent was waived as all data in the CPQCC database are deidentified.

Eligible infants for the study were born between 22 to 29 weeks’ GA, with BW between 401 and 1500 g. The limits of GA and BW were similar to those enrolled in the Neonatal Research Network VLBW registry.⁴ To study the impact of initial hospital course on outcomes, only infants born at or transferred within the first 2 days of life to a participating CPQCC hospital were included. The 2-day window for transfer was selected to capture those infants born close to midnight and transferred the following calendar day in a timely manner. This would also allow for the necessary time to arrange for transport to a regional center hospital from a distant hospital. Infants with any major congenital anomalies were excluded from the study. Major congenital anomalies included, but were not limited to, congenital heart defects, chromosomal abnormalities, gastrointestinal atresias, brain malformations, congenital diaphragmatic hernia, renal agenesis or dysplastic kidneys, and inborn errors of metabolism.

The primary outcome of interest was moderate to severe BPD, with a goal of approximating the standard criteria developed by the Eunice Kennedy Shriver National Institute of Child Health and Human Development/National Institutes of Health.^7,15,28 The California Perinatal Quality Care Collaborative does not currently collect data on specific fraction of inspired oxygen or require institutions to perform a physiologic challenge assessing the fraction of inspired oxygen requirement. Therefore, moderate to severe BPD was defined as any infant requiring continuous supplemental oxygen at 36 weeks’ postmenstrual age (PMA) or at hospital discharge, whichever occurred earlier. To account for infants transferred back to a non-CPQCC hospital prior to 36 weeks’ PMA, the infants were further stratified into 3 categories as follows: (1) The infant was categorized as having moderate to severe BPD (hereafter called BPD) if 1 of the following criteria was met: Continuous oxygen requirement at 36 weeks’ PMA while inpatient; discharge home prior to 36 weeks’ PMA on continuous oxygen; transferred at 34 to 36 weeks’ PMA on continuous oxygen; or death prior to 36 weeks’ PMA. These infants were included in the BPD category as mortality is a competing risk outcome for the most critically ill infants. (2) The infant was categorized as having no BPD if either there was no supplemental oxygen at 36 weeks’ PMA or the infant was transferred between 34 to 36 weeks’ PMA to a non-CPQCC hospital not on supplemental oxygen. (3) The infant BPD category was unknown if they were transferred to a non-CPQCC hospital prior to 34 weeks’ PMA.

A combined measure of BPD or death prior to 36 weeks’ PMA was used in the statistical analyses. Infants with BPD or death and without BPD were compared with respect to maternal and neonatal characteristics. A multivariable logistic regression model accounting for hospital as a random effect was used to assess individual risk factors for the development of BPD. Hospital as a random effect was selected to account for clustering of patients at the hospital level. As BW and GA were highly correlated, only GA was included in the models to estimate BPD rates. Gestational age was selected because neonatal outcomes are often correlated with GA and this marker is often used for counseling of parents. This multivariable logistic regression model with all covariates allowed us to assess independent risk for the primary outcome including the impact of the level of care on an individual patient.

This model was then applied to determine risk-adjusted combined rates of BPD or death in participating NICUs. We excluded NICUs in which there were fewer than 10 VLBW infants cared for during the study to include only those units that routinely cared for these patients. We also specifically looked at the combined BPD or death rates as related to the NICU level of care.

Level of care was defined according to the Committee on Fetus and Newborn of the American Academy of Pediatrics (AAP) policy statement.^29,30 Previous 2004 policy stratified NICUs into 3 different levels with 5 subdivisions.³⁰ This policy was revised in 2012 into 4 different levels redefining a level IIIC NICU as IV.²⁹ Neonatal levels of care are currently classified as well newborn nursery (level I), special care nursery (level II), NICU (level III), and regional NICU (level IV). Level II nurseries have the ability to provide brief duration of mechanical ventilation (<24 hours) and care for those infants born at 32 weeks’ GA or later and weighing 1500 g or more. Level III nurseries can provide sustained life support with a broad range of pediatric subspecialists available. Level IV nurseries have the capabilities of a level III unit with the addition of being able to surgically repair complex congenital or acquired conditions, as well as facilitate transport and education throughout their catchment area. As the study years ranged from 2007 to 2011 under the previous AAP levels of care, hospitals with NICU level IIIC were recategorized as level IV. All statistical analyses were performed using SAS (SAS Institute Inc).

A total of 17 863 infants met inclusion criteria according to GA, BW, and birth location. The following patients were excluded: 2055 infants with major congenital anomalies and 29 infants with missing data. The study cohort included 15 779 infants in 116 NICUs, with 1534 infants who died prior to 36 weeks’ PMA and 1238 infants who died either in the delivery room or within the first 12 hours of life (Figure 1). A total of 7081 infants, or 44.8%, of infants in the cohort met the primary outcome of death or BPD. The adjusted rate of BPD alone among survivors to 36 weeks’ PMA was 33.1%.

Maternal and neonatal characteristics are listed in Table 1. Significant differences between mothers with infants who developed BPD included maternal race/ethnicity, chorioamnionitis, maternal uterine infection, obstetrical bleeding, malpresentation or breech, preterm premature rupture of membranes, lack of antenatal corticosteroids, previous and current cesarean delivery, and intrauterine growth restriction.

Bronchopulmonary dysplasia was associated with lower Apgar scores, younger GA, lower BW, and male sex. Bronchopulmonary dysplasia rates were 80.7% for BW less than 750 g, 49.3% for BW 750 to 999 g, 25.1% for BW 1000 to 1249 g, and 13.1% for BW 1250 to 1500 g. Similarly, BPD rates were 93.8% for those with GA fewer than 24 weeks, 71.4% for GA 24 to fewer than 26 weeks, 44.1% for those with GA 26 to fewer than 28 weeks, and 20.8% for those with GA 28 to fewer than 30 weeks.

Odds ratios (ORs) were estimated with 95% CIs in a single multivariable logistic regression model using all covariates listed in Table 2. Statistically significant (P < .05) individual risk factors for BPD development included maternal hypertension, obstetrical preterm premature rupture of membranes, lack of antenatal corticosteroids, lower Apgar scores, lower GA, multiple births, and male sex. Maternal diabetes mellitus, black race, lack of prenatal care, and spontaneous labor were associated with reduced risks for BPD.

Using parameter estimates from the multivariable logistic regression model, risk-adjusted BPD rates were calculated for CPQCC member NICUs. Adjustments to the combined BPD or death rates included all risk factors listed in Table 2 along with hospital as a random effect. During the 5-year study, there were 132 individual CPQCC member NICUs, with an average of 127 hospitals participating during each year of the study. Of these, 116 hospitals had at least 10 patients included in the study population. The combined rates of BPD or death across these hospitals varied from 17.7% to 73.4% (interquartile range, 38.7%-54.1%) (Figure 2).

Hospital BPD rates were also evaluated according to AAP neonatal level of care using parameters from the logistic regression model to adjust for individual risk factors. Unadjusted median BPD rates were 55.1%, 43.8%, and 48.7% for levels II, III, and IV, respectively. Adjusted BPD rates stratified according to AAP neonatal level of care had a median of 50.3%, 46.1%, and 47.7% for levels II, III, and IV, respectively (Figure 2). The least amount of variation was observed among level IV units. The estimated covariance parameter for hospital of care was 0.270, with a standard error of 0.0455, indicating that there was significant variability among hospital rates even after accounting for patient mix (P < .001).

We also calculated risk-adjusted ORs for BPD development for neonatal level of care at the individual patient level for infants identified at hospitals cared for at NICUs identified as level II to level IV units. With level IV as the reference unit, level II units had 419 of 823 infants (50.9%) develop BPD or die (risk-adjusted OR, 1.23; 95% CI, 1.02-1.49). Level III units had 4813 of 10 364 infants (46.4%) develop BPD or die (risk-adjusted OR, 1.04; 95% CI, 0.95-1.14). Level IV units had 1768 of 3701 infants (47.8%) develop BPD or die.

This population-based study demonstrated that moderate to severe BPD in VLBW infants remains very common, with a primary outcome of BPD or death in 44.8% of at-risk infants in California. Bronchopulmonary dysplasia rates widely differ between hospitals, with level III and level IV having lower BPD rates than level II NICUs.

Significant individual risk factors for the development of BPD identified included male sex, lower GA, and lower BW, consistent with previous reports. Maternal chorioamnionitis was associated with increased odds of developing BPD. Chorioamnionitis has been postulated to play a role in the development of BPD by inflammatory processes and disturbance of lung maturation.^31,32 The lack of antenatal corticosteroids was also significantly associated with increased odds of BPD. Previous meta-analyses demonstrated that while antenatal corticosteroids may decrease the incidence of respiratory distress syndrome, it does not reduce the incidence of BPD.^16,33 However, the Cochrane review included a heterogeneous population as studies included were from both the pre- and post-surfactant eras. With advances in neonatal care, the impact of interventions is continually evolving, and it is possible that in the modern era of less-aggressive ventilatory strategies, antenatal steroids may have longer-term positive influences.

On the other hand, spontaneous labor and lack of prenatal care were associated with reduced risks for BPD development, which has not been shown consistently in prior studies. We hypothesized that lack of prenatal care and spontaneous labor may lend to a stressful perinatal environment with associated cortisol response.

Previous studies in the post-surfactant era reported BPD ranging from 4% to 74% similar to the wide range observed here.^17,19,20,34 We hypothesized that specific hospital environments and varying clinical practices affect the development of BPD. Further investigation into potential reasons for the wide variation of occurrence of BPD across NICUs may lead to insights for future efforts to reduce the prevalence of BPD. Collaborative efforts in the past have translated potentially better practices into significant improvement.^34,35 However, not all efforts have been successful and may be hampered by factors not identified at the time as significant moderators of BPD.^15,36

A previously published study using CPQCC data showed BPD rate across California at 32%; however, this did not include a combined measure with mortality.³⁷ In our current study, the rate of BPD among survivors to 36 weeks’ PMA was similar at 33.1%. Studies using data from the Eunice Kennedy Shriver National Institute of Child Health and Human Development Neonatal Research Network and the Vermont Oxford Network showed lower rates of BPD at 22% and 26.3%, respectively, compared with those observed here.^1,38 In part, we were unable to use the physiologic definition of BPD in our analyses and thus likely have a higher number of infants identified with BPD. Another potential explanation is that CPQCC captures a broader range of hospitals such as community hospitals beyond regional perinatal centers. Although overall rates were similar by level of care (Figure 2), we found significant differences at the individual risk level for BPD development in level II units compared with level IV units (OR, 1.23; 95% CI, 1.02-1.49). The variation in combined BPD or death rates was lowest for the level IV units (Figure 2).

Deregionalization of neonatal care has prompted studies examining its impact on VLBW infants. The increased risk for BPD shown here is concerning because others have also demonstrated increased risks for morbidity and mortality among those VLBW infants born outside of regional centers.^22-27 By definition of AAP level of care, VLBW infants should be admitted to level III and level IV nurseries only. A small number of VLBW infants are cared for at level II units, however, with unknown long-term impact on the health of those infants.

The strength of this study was its unique and extensive population-based cohort capturing infants born at all levels of care. The limitations of this study were those inherent to working with a large database, which includes the inability to delve further into differences in clinical practice. While the CPQCC database contains a broad range of maternal and neonatal clinical variables, there may exist other unobserved key factors not included in the database that could contribute to variability. In addition, we did not have access to information regarding fetal deaths. The CPQCC also does not require documentation of specific fraction of inspired oxygen requirements or room air challenge; therefore, we were unable to use the physiologic definition of BPD. This allows for variations in oxygen prescriptive practices to dictate whether an infant met mild, moderate, or severe BPD criteria and likely results in higher BPD rates reported here than those studies using the physiologic definition of BPD.^10,39

Our study demonstrated that BPD remains a significant morbidity in VLBW infants, with new concerns that deregionalization might contribute to increased risk of BPD. On an individual patient level, our study supports the association between chorioamnionitis and the development of BPD. In contrast to previous studies, the lack of antenatal corticosteroids was a significant risk factor for BPD. A previous California study demonstrated that there exists large variability in the rates of antenatal corticosteroid administration, which may indicate an opportunity for improvement to enhance care for at-risk patients.⁴⁰ We observed large variation in BPD rates across individual hospitals in the CPQCC and differences between neonatal levels of care. Further studies to examine the impact of deregionalization on other significant morbidities and mortality may be insightful and necessary to further advocate regionalization of the perinatal system.

Corresponding Author: Wannasiri Lapcharoensap, MD, Division of Neonatal and Developmental Medicine, Stanford University School of Medicine, 750 Welch Rd, Ste 315, Palo Alto, CA 94304 (wlapchar@stanford.edu).

Accepted for Publication: December 17, 2014.

Published Online: February 2, 2015. doi:10.1001/jamapediatrics.2014.3676.

Author Contributions: Dr Lee had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Lapcharoensap, Gage, Profit, Shaw, Gould, Stevenson, O’Brodovich, Lee.

Acquisition, analysis, or interpretation of data: Lapcharoensap, Kan, Profit, Shaw, Gould, Stevenson, O’Brodovich, Lee.

Drafting of the manuscript: Lapcharoensap, Gage, Kan, Shaw, Stevenson, O’Brodovich, Lee.

Critical revision of the manuscript for important intellectual content: Profit, Shaw, Gould, Stevenson, O’Brodovich, Lee.

Statistical analysis: Lapcharoensap, Kan, Shaw, Lee.

Obtained funding: Stevenson.

Administrative, technical, or material support: Lapcharoensap, Shaw, Stevenson, O’Brodovich, Lee.

Study supervision: Gage, Profit, Gould, O’Brodovich, Lee.

Conflict of Interest Disclosures: None reported.

Funding/Support: This project was sponsored by grant K23HD068400 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development.

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the Eunice Kennedy Shriver National Institute of Child Health and Human Development or National Institutes of Health.