Trends in Resources for Neonatal Intensive Care at Delivery Hospitals for Infants Born Younger Than 30 Weeks’ Gestation, 2009-2020

Key Points Question Have neonatal intensive care resources at hospitals where infants born extremely preterm are delivered changed over the past decade? Findings In this cohort study including 357 181 infants born at 22 to 29 weeks’ gestation between 2009 and 2020, births at neonatal intensive care units (NICUs) with lower levels of care or lower birth volumes increased, while births at NICUs with higher levels of care or higher birth volumes decreased. Meaning The findings of this study suggest increasing deregionalization of extremely preterm birth.


Introduction
In 1976, the March of Dimes Committee on Perinatal Health published recommendations on perinatal care regionalization in the US that included the referral of mothers and infants with high risk of adverse perinatal outcomes to a hospital with a regional neonatal intensive care unit (NICU). 1 Regionalization was soon extensively implemented in the US, partly due to regulatory mechanisms organized by state certificate of need laws. 2 Starting in the late 1980s and continuing into the 2000s, various studies, mostly from single states, found that perinatal deregionalization was occurring, [3][4][5][6][7][8][9][10][11] with numbers of NICUs mainly increasing in urbanized areas that are within a reasonable distance to an existing tertiary NICU. 12However, to our knowledge, there are no recent data examining national perinatal regionalization trends.
[18][19][20][21][22][23][24][25] However, to our knowledge, there are no data examining national trends in the distribution of infants with high risk in low-vs high-volume NICUs, nor have variations in these trends been examined by US region.
To address these issues, we used data from Vermont Oxford Network (VON), a nonprofit, voluntary, worldwide collaboration of health care professionals dedicated to improving the quality, safety, and value of care for newborns.We examined regionalization trends in the birthplace NICU level, incorporating NICU volume from 2009 to 2020 overall and by US region among newborns born at 22 to 29 weeks' gestation.

Methods
For this cohort study, the use of VON's deidentified research repository was determined to be exempt from review and informed consent by the University of Vermont's committee for human research because it was not human participants research.This study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Study Sample
We included neonates born at 22 to 29 weeks' gestation who were born at or transferred to VON US centers within 28 days of life from January 1, 2009, through December 31, 2020.eTable 1 in Supplement 1 shows the number of centers participating in VON throughout the study period.

Statistical Analysis
We conducted descriptive analyses examining the distribution of maternal and newborn characteristics by the birthplace level of neonatal care.We estimated national and regional trends in proportions of births in hospitals with A-level, low-volume B-level, or high-volume B-or C-level NICUs between 2009 and 2020.Models were run for all newborns at 22 to 29 weeks' gestation and separately for newborns at 22 to 25 and 26 to 29 weeks' gestation to assess any difference in trends between gestational age intervals.We used multinomial regression (nonproportional-odds cumulative logit model) to estimate the probability of birth in each NICU level, along with 95% CIs.
We also calculated the absolute change in the percentage of births along with the 95% CI by NICU level between 2009 and 2020.
We conducted sensitivity analyses to determine the robustness of our findings.The first analysis was conducted restricting to hospitals that were in the sample the whole period.

Study Sample
From 2009 to 2020, 322 407 neonates were inborn at 822 VON centers and 39 652 were outborn and transferred to 714 VON centers.Among the inborn infants, 305 866 infants (94.9%) did not die in the delivery room and were admitted to the NICU.Among the outborn infants, 17 464 infants transferred from a birth center with a known NICU level.likely to be investor owned and less likely to be nonprofit owned, while hospitals with high-volume Bor C-level NICUs were less likely to be investor owned and more likely to be nonprofit owned (Table 1).Individuals with pregnancy complications and multiple gestations were more likely to deliver at hospitals with high-volume B-or C-level NICUs and less likely to deliver at hospitals with level A NICUs (Table 1).

National Trends Between 2009 and 2020
Figure 1 shows the national trend of births in A

Regional Trends Between 2009 and 2020
Among infants born at 22 to 29 weeks' gestation, most regions followed the nationwide trends (eFigure 3 in Supplement 1).The exceptions were the Pacific, which started with a much smaller share of these infants at high-volume B-or C-level centers and showed some small changes in the site of delivery during the study period, and the East South Central region, which had a lower regionalization rate at the start of the study and shifted toward the national mean by the end of the study (Table 3

Sensitivity Analyses
Restricting to hospitals that were in the sample the whole period showed similar national trends, although the changes between 2009 and 2020 were smaller (eTable 4 in Supplement 1).The changes comparing the whole to the restricted sample were less pronounced if the region did not experience the addition of several new hospitals (eTable 5 in Supplement 1).Between 2009 and 2020 among newborns at 22 to 29 weeks' gestation, births decreased at hospitals with high-volume B-or C-level NICUs, while births increased at hospitals with level A NICUs.Depending on the sensitivity analysis, births at low-volume B-level NICUs either decreased or increased (eTable 6 and eTable 7 in Supplement 1).

Discussion
In this cohort study using a national sample, we found a nationwide shift toward deregionalization of perinatal care among infants born at 22 to 29 weeks' gestation.Between 2009 and 2020, births across all regions increased at level A by 5.6% and at low-volume B-level NICUs by 3.6%, while births at high-volume B-or C-level NICUs decreased by 9.2%.15][16][17][18][19][20][21][22][23][24][25]27 Similar trends were observed among infants born at 22 to 25 and 26 to 29 weeks' gestation, with the shift away from the high-volume B-or C-level centers being more pronounced among infants born at 26 to 29 weeks' gestation.
4][15][16][17] In a meta-analysis of studies published through 2010, birth in a nonregional hospital, compared with a regional hospital, was associated with increased odds of mortality (adjusted odds ratio [aOR], 1.55; 95% CI, 1.21-1.98)among infants born at 32 weeks' gestation or less. 13This association was even stronger among infants weighing less than 1000 g (aOR, 1.80; 95% CI, 1.31-2.46). 13Another study showed that among newborns younger than 28 weeks' gestation in a tertiary hospital, birth in a nontertiary hospital was associated with increased odds of mortality (OR, 2.32; 95% CI, 1.78-3.06),while transfer from a nontertiary to a tertiary hospital in the first 48 hours was associated with increased odds of severe brain injury. 27][16][17][18][19][20][21][22][23][24][25]27 However, previous studies have underestimated the benefits associated with regionalization and with NICU level and NICU volume.Given that patients who are more severely ill are more likely to deliver at hospitals with high-level and high-volume NICUs, studies conducted using instrumental variables methods that control for this unobserved selection bias show even stronger associations with improved neonatal outcomes when infants with high risk of adverse outcomes are delivered at high-level and high-volume NICUs, compared with low-level and low-volume NICUs. 14,22,23For example, a study by Wehby et al 23  ) increased mortality odds using the instrumental variable approach. 23 speculate that the increasing share of infants delivered in hospitals with level A or low-volume B-level NICUs is accounted for by a decrease in the share of infants delivered in hospitals with high-volume B-or C-level NICUs.A study conducted in California examined how delivery location shifted in response to midlevel NICU openings (equivalent to our level A or low-volume B-level). 6Almost all (88%) of the VLBW deliveries at these new NICUs were shifted from hospitals with a high-volume B-or C-level NICU. 6Thus, this increase in midlevel NICUs was essentially all deregionalization, not improved NICU care access. 6e proliferation of midlevel NICUs is mainly happening in urban and suburban areas that are already being served by tertiary centers. 6,7In California, for example, approximately 80% of births that occurred in smaller, lower-level NICUs were located within 25 miles of a large tertiary NICU. 12is deregionalization in perinatal care has occurred due to the wide availability of neonatal intensive care technologies and neonatologists, economic factors and financial incentives derived from installing new NICUs, 12,28 and different state policies that influence regulation of regionalized systems. 2,29The effects of these different factors have also varied by US region, contributing to the regionalization trends we observed.Most births at 22 to 29 weeks' gestation (75%) 30 in the Pacific region occur in California.
California had already shown deregionalization trends starting in the 1990s. 5,20,31This trend continued and between 1990 and 2001: there were 48 new midlevel units established or upgraded, 6 and between 2005 and 2011, the overall percentage of infants with VLBW born at hospitals providing the highest degree of care (defined as centers providing major surgery with >100 infants with VLBW per year and centers providing major surgery, extracorporeal membrane oxygenation, and cardiopulmonary bypass) decreased from 42.5% to 26.5%. 19ven the negative outcomes associated with birth at nontertiary or low-volume hospitals, the fact that more than 50% of these infants with high risk were born in hospitals with level A and lowvolume B-level NICUs is unacceptable.Previous studies have shown that the delivery of these infants with high risk can be shifted to tertiary centers if there is a systemwide effort.In 1990, Portugal closed all the small maternity units and small NICUs and implemented a system to facilitate maternal transport of high-risk deliveries. 32This resulted in more than 90% of all very preterm deliveries occurring in a hospital with a high-volume NICU, which was associated with decreasing the neonatal mortality rate from 8.1 deaths to 2.7 deaths per 1000 live births. 32On a local level, in the greater Cincinnati, Ohio, region, the implementation of perinatal outreach programs stressing the importance of transfer of mothers with high risk to subspecialty perinatal centers decreased the percentage of infants with VLBW delivered at hospitals without tertiary perinatal care from 25% to 11.8%. 33

Limitations
This study has several limitations.One limitation is the missing data on the birth hospital level for a large proportion of outborn infants.We analyzed our data by region, which might mask individual state-level differences.To classify the NICU level for outborn infants, we used publicly available resources, which might have incomplete and inaccurate web descriptions that claim a higher level of care.

Figure 2 .
Figure 2. Regional Trend of Births by NICU Level Between 2009 and 2020 Among 22-to-29-Week Newborns Between 2009 and 2020, births at high-volume B-or C-level NICUs decreased between 3.7% and 21.2% and births at level A NICUs increased between 4.2% and 14.0%.Births at low-volume B-level NICUs increased across 6 regions, and the increase ranged between 2.3% and 8.1%.Only 2 regions stood out in their regionalization trends: the East South Central and Pacific regions.The East South Central region had a lower regionalization rate at the start of the study and shifted toward the national mean by the end, while the Pacific region had a minor change in the delivery site during the study period because this region had already experienced a decline in the level of NICU care regionalization.

JAMA Network Open | Pediatrics Resources
we further classified level B centers into low-volume (ie, <50 inborn infants at 22 to 29 weeks' gestation per year) and high-volume (ie, Ն50 inborn infants at 22 to 29 weeks' gestation per for Neonatal Intensive Care at Delivery Hospitals for Infants Born Younger than 30 Weeks' Gestation JAMA Network Open.2023;6(5):e2312107.doi:10.1001/jamanetworkopen.2023.12107(Reprinted) May 5, 2023 2/12 Downloaded From: https://jamanetwork.com/ on 09/17/2023 year).Since many studies have also found no outcome differences between high-volume B-level vs C-level centers (essentially all high volume), resulting in 3 distinct NICU categories.In January 2015, VON started collecting the names of the non-VON birth center for outborn infants transferring to a VON center.For these non-VON centers, we obtained the center's level of neonatal care from publicly available resources.Level I centers were coded as having a well-baby nursery; level II, having an equivalent of VON's classification of level A NICU; level III, having a level B NICU; and level IV, having a level C NICU.Centers with well-baby nursery and those with level A NICUs were combined into 1 group, labeled as level A NICUs.For outborn infants born before 2015, we did not know the NICU level of the non-VON center, so we only included infants transferred to VON centers within 3 days of life because we were more confident in assuming that these infants were being transferred from hospitals with level A NICUs or well-baby nurseries.NICU census regions and divisions were classified according to the US Census Bureau classifications as West (Mountain and Pacific), Midwest (East North Central and West North Central), South (South Atlantic, East South Central, and West South Central), and Northeast (New England and Mid-Atlantic) (eFigure 1 in Supplement 1).26MemberNICUs contribute data from medical records using standardized VON forms.Maternal race and ethnicity were obtained by personal interview with the mother, review of the birth certificate, or medical record, in that order.Race and ethnicity were classified as American Indian, Asian, Black, Hispanic, White, and other (for individuals who did not identify as any of the provided categories).Race and ethnicity were included in analysis as descriptive characteristics.

Downloaded From: https://jamanetwork.com/ on 09/17/2023 to
VON centers with NICU levels A (4.8%), B (22.6%), or C (72.6%).For the final sample size, we excluded 2100 infants with missing data on birthplace level of newborn care, 2399 infants missing data on race and ethnicity, 337 infants missing data on congenital anomalies, and 124 with missing data on sex, resulting in 357 181 infants (mean [SD] gestational age, 26.4 [2.1] weeks; 188 761 [52.9%] male), including 320 243 inborn infants and 36 938 outborn infants.Overall, 108 283 infants received care at level A NICUs, 62 061 infants received care at low-volume level B NICUs, and 186 837 infants received care at high-volume level B or level C NICUs (Table1).Among 74 957 inborn infants delivered at hospitals with level A NICUs, 7164 (9.6%) were born at centers with restricted ventilation.Table1 showsmaternal and newborn characteristics by birth hospital NICU level.Across regions, the Pacific (20 239 infants [38.3%]) had the lowest while the South Atlantic (48 348 infants [62.7%]) had the highest percentage of births at a hospital with a highvolume B-or C-level NICU.Among VON centers with level A NICUs, the median (IQR) volume of infants born at 22 to 29 weeks' gestation ranged between 22 (15-65) infants per year in New England and 48 (28-101) infants per year in the East North Central region, while among high-volume B-and C-level VON NICUs, it ranged between 70 (58-94) infants per year in the Pacific and 110 (81-134) infants per year in the East South Central region.Infants with congenital anomalies were more likely to be born at hospitals with high-volume B-or C-level NICUs (9801 infants[5.3%])andlesslikely to be born at hospitals with level A NICUs (3962 infants [3.7%]).Acute transfer among inborn infants was higher if newborns were delivered at hospitals with level A NICUs(20 112of 71 152 infants [28.3%]) and lower if newborns were delivered at hospitals with high-volume B-or C-level NICUs (10 362 of 176 785 infants [5.9%]).Hospitals with level A and low-volume B-level NICUs were more

Table 1 .
Maternal and Newborn Characteristics by Birth Hospital NICU Level in the Vermont Oxford Network Resources for Neonatal Intensive Care at Delivery Hospitals for Infants Born Younger than 30 Weeks' Gestation −24.0% to −18.2%), while the East South Central region had the largest increase (22.2%; 95% CI, 17.8% to 26.6%) in births at high-volume B-or C-level NICUs (Table 3).In 2020, the Pacific had the lowest number of births at high-volume B-or C-level NICUs (38.5%; 95% CI, 35.9% to 41.0%).The West South Central region had the largest increase (8.1%; 95% CI, 4.2% to 11.9%), while the West a Data were missing for 1179 inborn and outborn infants on antenatal corticosteroids, including 575 inborn infants; 221 infants on hospital ownership; 711 mothers on chorioamnionitis; 439 mothers on hypertension; 12 mothers on multiple gestation.Infants born at non-VON centers and transferred prior to 3 days of life coded as being born at a hospital with well-baby nursery/NICU level A. b Row percentages reported.cVolumevariable restricted to VON centers and includes inborn infants only.dAcutetransfer excludes delivery room deaths and includes both inborn infants and outborn infants who were transferred to VON centers within 3 days of life.Acute transfer defined as transfer for medical or diagnostic services or surgery for infants born at VON centers and as transfer within 3 days of life for outborn infants.eAcute transfer excludes delivery room deaths.Acute transfer defined as transfer for medical or diagnostic services or surgery for infants born at VON centers.f Data were available for 20 090 infants at level A i Data collection on hypertension and chorioamnionitis variables started in 2008.JAMA Network Open | Pediatrics JAMA Network Open.2023;6(5):e2312107.doi:10.1001/jamanetworkopen.2023.12107(Reprinted) May 5, 2023 5/12 Downloaded From: https://jamanetwork.com/ on 09/17/2023

Table 2 .
Percentage of Births by NICU Level in 2009 and 2020 Among Newborns Born at 22 to 29 Weeks' Gestation Resources for Neonatal Intensive Care at Delivery Hospitals for Infants Born Younger than 30 Weeks' Gestation 4%).The Middle Atlantic and West South Central regions had the largest increases while the West North Central region had the largest decrease in births at low-volume B-level NICUs.By 2020, the Middle Atlantic had the highest percentage of births at low-volume B-level NICUs (30.7%; 95% CI, Abbreviation: NICU, neonatal intensive care unit.aForoutborn newborns transferred within 3 days of life to VON centers, NICU level coded as level A.JAMA Network Open | PediatricsJAMA Network Open.2023;6(5):e2312107.doi:10.1001/jamanetworkopen.2023.12107(Reprinted) May 5, 2023 6/12 Downloaded From: https://jamanetwork.com/ on 09/17/2023 46.

Table 3 .
Percentage of Births by NICU Level and Region in 2009 and 2020 Among Newborns Born at 22 to 29 Weeks' Gestation Resources for Neonatal Intensive Care at Delivery Hospitals for Infants Born Younger than 30 Weeks' Gestation 1% to 36.3%).New England had the largest increase while the East South Central region had the largest decrease in the percentage of births at level A NICUs (eTable 2 in Supplement 1).Between 2009 and 2020, decreasing trends of births at high-volume B-or C-level NICUs for infants born at 26 to 29 weeks' gestation occurred across all regions except the East South Central region (eFigure 5 in Supplement 1).Between 2009 and 2020, increasing trends of births at low-volume B-level NICUs occurred across all regions except the West North Central region, East South Central, and Mountain regions (eTable 3 in Supplement 1).By 2020, the Middle Atlantic had the highest percentage of births at low-volume B-level NICUs (28.0%; 95% CI, 24.3% to 31.6%).New England and the West South Central region had the largest increase while the East South Central region had the largest decrease in the percentage of births at level A NICUs.
found that being delivered at hospitals with low-volume NICUs 34 US Vermont Oxford Network Participating Centers by Region Between 2009 and 2020 eFigure 1. US Census Regions and Divisions eFigure 2. Disposition Status of Inborn and Outborn Infants Born at 22-29 Weeks' Gestation Between 2009 and 2020 in the Vermont Oxford Network eFigure 3. Regional Trend of Births by NICU Level Between 2009 and 2020 Among Newborns Born at 22-29 Weeks' Gestation eFigure 4. Regional Trend of Births by NICU Level Between 2009 and 2020 Among Newborns Born at 22-25 Weeks' Gestation eTable 2. Percentage of Births by NICU Level by Region in 2009 and 2020 Restricting to Newborns Born at 22-25 Weeks' Gestation eFigure 5. Regional Trend of Births by NICU Level Between 2009 and 2020 Among Newborns Born at 26-29 Weeks' Gestation eTable 3. Percentage of Births by NICU Level by Region in 2009 and 2020 Restricting to Newborns Born at 26-29 Weeks' Gestation eTable 4. Percentage of Births by NICU Level in 2009 and 2020 Among Preterm Newborns Restricting to Centers Participating in VON Throughout the Whole Study Period eTable 5. Percentage of Births by NICU Level and Region in 2009 and 2020 Among Newborns Born at 22-29 Weeks' Gestation Restricting to Centers Participating in VON Throughout the Whole Study Period eTable 6. Sensitivity Analysis: Percentage of Births by NICU Level in 2009 and 2020 Among Newborns Born at 22-29 Weeks' Gestation eTable 7. Sensitivity Analysis Restricted to Newborns With Nonmissing Birth NICU Level Data: Percentage of Births by NICU Level in 2009 and 2020 Among Newborns Born at 22-29 Weeks' Gestation eTable 8. Vermont Oxford Network Participating Centers Between 2009 and 2020