Variability of Care Practices for Extremely Early Deliveries

In this multicenter retrospective cohort study, we screened all BPs admitted with live fetuses who delivered at 13 centers between 22 0/7 and 24 6/7 weeks’ gestation. Those whose fetuses had conditions that would significantly impact survival or neurodevelopment and, therefore, decision-making, such as known severe genetic or anatomic anomalies, twin-twin transfusion syndrome, or fetal surgery, were excluded. BPs with multiple gestations were included provided at least 1 fetus met inclusion criteria (live fetus without significant condition noted above). BPs with intrauterine fetal demise (IUFD) diagnosed within 6 hours of admission were excluded because of lack of opportunity for the perinatal interventions we aimed to study. Data on BP sociodemographic factors, obstetric care, inpatient consultations, and neonatal clinical course were collected at each center. Race and ethnicity data were obtained from the medical record as prior studies have reported associations with neonatal resuscitation at extremely early gestations.^20–22  Federal census-related geocoding data (2015 American Community Survey 5-year Estimates) were extracted using BPs’ residential zip codes.²³  Data were entered into a standardized form in a centralized REDCap database²⁴  housed at NorthShore University HealthSystem. Each center’s team worked under its institutional review board’s approval for waived consent to access patient records and contractual permissions to share data.

Birthing persons and newborns were divided into 2 epoch-defined groups by year of delivery to analyze changes over time. Epoch 1 (2011–2015) and epoch 2 (2016–2020) coincided with the periods before and after the year 2015, when the AAP antenatal counseling guidelines were updated to include shared decision-making at 22 weeks, and Rysavy et al demonstrated survival variability was most influenced by intercenter variability.¹⁸  We tabulated antenatal corticosteroids and neonatology consultation for the denominator of birthing people with fetuses alive at hospital admission. Cesarean delivery was tabulated for BPs without confirmed IUFD after admission; attempted resuscitation was tabulated for infants excluding IUFD; and live births and survival were tabulated for all infants (including IUFD and stillborn) as a measure of the full spectrum of perinatal and neonatal interventions. Corticosteroid use was defined as receiving at least 1 dose of antenatal corticosteroids. Survival was defined as survival to discharge or hospital transfer. χ² and Fisher’s exact tests compared rates of each intervention between epochs. Center-level rates were also calculated; variability of practices between centers at each gestational week was calculated using interquartile ranges (IQR). Levene’s test was used to analyze whether intercenter variability differed significantly between the 2 epochs.

BPs whose newborns received AR were compared with those whose newborns did not receive AR. Those who delivered after known IUFD and those whose multiple-gestation newborns received discordant delivery room treatment were excluded from this analysis. Given that newborns not receiving AR are sometimes recorded as “stillborn,” BPs delivering such newborns remained in parts of the analysis as noted in the results. AR was defined as receipt of assisted ventilation (continuous positive airway pressure, other forms of noninvasive positive pressure ventilation, or endotracheal intubation) in the delivery room. χ² and Fisher’s exact tests compared ordinal variables; Mann Whitney U tests compared continuous nonevenly distributed variables. Individual characteristics found to be significantly different (P < .05) were included in a multivariate logistic regression analysis to calculate probability of the infant receiving AR, as reported by odds ratios (OR) and 95% confidence intervals (95% CI). Delivery center was included in the regression but ORs are not reported individually for confidentiality. Statistics were performed using R v. 4.1.3 (Vienna, Austria, 2022).²⁵ 

Our cohort included 2028 BPs delivering 2327 newborns. Details of inclusion and distribution of cohort across centers are shown in Fig 1.

Figure 2 shows rates of interventions by gestational age and epoch for the study cohort as a whole. Rates increased over time at 22 weeks for the following: neonatology consultation (37.6 vs 64.3%, OR 2.99, 95% CI 2.01–4.47, P < .001), antenatal corticosteroids (11.4 vs 29.5%, OR 3.26, 95% CI 1.94–5.61, P < .001), live birth (66.2 vs 78.6%, OR 1.88, 95% CI 1.28–2.86, P < .001), AR (20.1 vs 36.9%, OR 2.32, 95% CI 1.53–3.53, P < .001), and survival (3.0 vs 8.9%, OR 3.11, 95% CI 1.31–8.21, P = .005). Rates increased at 23 weeks for the following: neonatology consultation (73.0 vs 80.5%, OR 1.52, 95% CI 1.05–2.23, P = .020), antenatal corticosteroids (63.7 vs 83.7%, OR 2.93, 95% CI 2.01–4.32, P < .001), cesarean delivery (28.0 vs 44.7%, OR 2.08, 95% CI 1.48–2.93, P < .001), live birth (88.1 vs 95.1%, OR 2.6, 95% CI 1.46–4.78, P < .001), AR (67.7 vs 85.2%, OR 2.74, 95% CI 1.89–3.99, P < .001), and survival (28.8 vs 41.6%, OR 1.76, 95% CI 1.29–2.40). Among newborns receiving AR, rates of survival to discharge or transfer did not significantly change between epochs; 15.4 vs 24.5% at 22 weeks, 44.1 vs 49.5% at 23 weeks, and 67.7 vs 67.1% at 24 weeks. Transfers accounted for 11.3% (107 of 948) of infants who survived.

In Fig 3, boxplots show the distribution of centers’ rates of interventions and outcomes by gestational week for each epoch. For deliveries at 22 weeks’ gestation, in epoch 2, variability of antenatal corticosteroid administration rates between centers increased (IQR 18.0 vs 42, P = .014). For deliveries at 23 weeks’ gestation, in epoch 2, intercenter variability decreased significantly for rates of neonatology consultation (IQR 23.0 vs 5.2, P = .045). Variability in antenatal corticosteroids and attempted resuscitation decreased, but these were not statistically significant. Variability between individual centers for live births and survival to NICU discharge were similar between epochs at each gestational week.

In the BP-level analysis of characteristics of those whose newborns received or did not receive AR (Table 1), 1971 BPs were included. Newborns of 75% of BPs (n = 1484) received AR in the delivery room, whereas newborns of 25% of BPs (n = 487) did not. Several factors were significantly associated with AR when analyzed individually (Table 1, univariate analysis significant P values indicated). However, when controlling for these independently significant factors in multivariate analysis, only the following remained significant: BP Black versus white race (OR 1.58, 95% CI 1.05–2.40, P = .031), previous delivery of a premature newborn (OR 1.54, 95% CI 1.01–2.38, P = .049), more advanced gestational age at delivery (OR 8.57, 95% CI 5.72–13.20, P < .001), and higher neonatal birth weight (OR 1.01, 95% CI 1.01–1.10, P < .001). AR also varied significantly by delivery center (P < .001); center-specific odds ratios for AR ranged from 0.13 to 14 compared with the reference center.

In this study, we examined the perinatal courses of 2327 newborns born to 2028 birthing persons between 22 0/7 and 24 6/7 weeks’ gestation at 13 centers across the United States over a 10-year period. We investigated factors associated with perinatal interventions. Would a given infant receive AR? The answer depended on where, when, and of which birthing person the question was asked.

In our study, we found significant intercenter variability in intervention rates. Delivery center was significantly associated with rates of attempted resuscitation in the delivery room when controlling for BP characteristics.

Surveys of US neonatologists have demonstrated variability in willingness to resuscitate in the gray zone, particularly at 22 to 23 weeks. In 2014, Arzuaga et al reported geographic variability within the United States in neonatologists’ willingness to resuscitate a 22-week gestation newborn, with rates of 0% to 5% in some census regions and 19% to 23% in others.¹⁰  A 2020 survey of US neonatologists found variability in willingness to offer resuscitation at 22 and 23 weeks’ gestation and in the resources reported to be available at their institutions, including equipment and interdepartmental support.²  A 2021 survey found that physicians’ resuscitation preferences at extremely early gestations varied by training site, training level, and projected quality of life.²⁶  A 2020 survey from our group found that NICU professionals’ views of patient suffering and future quality of life differed when grouped by center rates of resuscitation.¹⁴ 

These resuscitation preferences are borne out in analyses of actual interventions in the gray zone. A 2013 analysis of Neonatal Research Network data reported variability in intervention scores for newborns born at <25 weeks’ gestation, which accounted for a large part of intercenter variability in mortality.¹⁵  A 2020 analysis of Vermont Oxford Network data similarly demonstrated regional variability in interventions and survival, particularly at 22 and 23 weeks.⁹  Another 2020 analysis found that hospital of birth contributed as much as gestational age to survival in newborns born between 22 0/7 and 25 6/7 weeks.⁴  Although variability in parental values and decision-making should and will contribute to variability in resuscitation practices, these parental preferences cannot be incorporated into decision-making if resuscitation is not offered. Therefore, this regional and intercenter variability is critical to identify and address.

We found increased rates of interventions and outcomes over time in the cohort as a whole, including increased rates of neonatology consultation, antenatal corticosteroid administration, cesarean delivery, live birth, attempted delivery room resuscitation, and survival. Increased interventions were more common at 23 weeks’ gestation, perhaps reflecting increasing provider acceptance of resuscitation at this age.^2,3  A large study using vital statistics data also recently reported increased delivery room resuscitation at these gestational ages over time.²⁷ 

In addition, we found decreased intercenter variability in interventions over time for newborns born at 23 weeks. It is notable, however, that practice variability at 22 weeks remained wide. We posit that as the lower limit of the gray zone is undergoing transition and now appears to be moving toward 22 weeks,²  a new standard has not been fully established. Whether this variability decreases in the future will require study.

Temporal changes may reflect changes in practices influenced by evolving professional organization guidelines. New guidelines have been shown to influence practice changes and impact survival.^28,29  For example, 2011 guidance³⁰  from the American College of Obstetricians and Gynecologists (ACOG) recommended antenatal corticosteroids between 24 and 34 weeks and did not recommend them “before viability”; this guidance was updated in 2016³¹  to explicitly allow consideration of corticosteroids at 23 weeks. This change in practice guidance may have contributed to the increase in corticosteroid use at 23 weeks in the second epoch of our study. Conversely, the most recent update from ACOG³²  offering consideration of corticosteroids at 22 weeks was issued in 2021, after the time frame of our study; however, our data indicate that corticosteroid use at 22 weeks was already increasing before ACOG recommendations; this practice change seems to precede guideline changes, as previously reported.⁵ 

Similarly, we report increasing rates of neonatology consultation at 22 and 23 weeks, aligning with the AAP clinical guideline adjustment in 2015 toward individualized counseling for parents at 22 weeks’ gestation.¹  Despite an increasing body of literature on neonatology consultations,^33–36  data on the frequency of antenatal consults are limited and we are aware of no prior study assessing changes over time. Rising rates of interventions along with increased neonatology consultations suggest that the substance of these discussions may also be evolving.

Although overall survival at 22 weeks remains low, we demonstrated increased survival between epochs at 22 and 23 weeks. Studies with larger data sets have also reported increased survival of extremely early newborns over the last 15 years.^4,37–41  In relevant position statements, rates of survival at 22 weeks have been reported as 6%⁴²  or “extremely rare.”¹  However, in a 2021 meta-analysis, the pooled survival rate for actively treated 22 week newborns worldwide was reported as 29%.⁴³  A 2022 Neonatal Research Network analysis reported 30% survival among actively treated newborns at this age in the United States.³⁸  Worldwide, specific centers practicing universal attempted resuscitation at 22 weeks’ gestation have reported survival rates in excess of 50%⁴⁴  and as high as 70% to 83%.^45–47  In the most recent epoch, we also found that survival for actively treated 22-week newborns was 24.5% with study centers ranging from 0% to 67%, consistent with the worldwide and US data.

When we compared BPs whose newborns did and did not receive AR in multivariate analysis, we found that AR remained significantly associated with increased gestational age at delivery, higher birth weight, some delivery centers, previous premature deliveries, and Black BP race. Previous studies show conflicting associations between interventions for this gestational age group and BP race, with some reporting higher rates of neonatal but not prenatal interventions for Black BPs.^12,21,22  Another study reported a negative correlation with neonatal active treatment of BPs identified as Black, Asian/Pacific Islander, or Hispanic.²⁷  Although these studies, derived from vital statistics, had the strength of large sample sizes, center effects could not be analyzed. In our study, controlling for center effects, Black BP race remained significantly associated with AR.

Some studies have suggested that increased resuscitation for newborns of Black BPs may represent differences in values and preferences regarding life-sustaining treatment.^20,48  Further analysis of our data with respect to delivery center, BP race, and socioeconomic markers using geocoding is underway. More research is needed to understand this especially value-laden topic.

We found that AR occurred more often in BPs who had previously delivered premature newborns, although we cannot report if these premature deliveries were also at 22 to 24 weeks’ gestation, or if these children survived. One recent survey study has reported an association between previous NICU experience and parental request to resuscitate a hypothetical 22-week newborn⁴⁹ ; however, to our knowledge, ours is the first study to report the association of prior premature delivery with an actual future child’s resuscitation. Although concerns about the neurodevelopmental outcomes of premature newborns are common,^1,42,50  studies indicate that parents of extremely preterm newborns generally rate their children’s health highly, regardless of neurodevelopmental trajectory.^51–53  Perhaps parents of preterm newborns weigh their family’s unique experience more heavily in their decision-making than clinicians’ concerns about projected disability; additional research on this topic is needed.

Our study has several strengths. Collecting data at the level of individual BP and newborn records allowed us to examine factors not assessed in other studies, such as previous pregnancy data and occurrence of neonatology consultations. Unlike other large US studies of similar newborns, our sample was not limited to liveborn deliveries, allowing a more comprehensive look at perinatal practices and outcomes in this gray zone. Inclusion of multiple centers across the United States over a 10-year period allowed examination of temporal changes in center variability. Despite these strengths, we recognize limitations of our study. Use of individual chart-level data at participating centers resulted in a smaller sample size than studies using administrative or national registry data. Participating centers trained postgraduate trainees in their NICUs and obstetric departments; although 10 states were represented, about half of the centers were in the Midwest. Together these factors may limit generalizability to nonacademic, non-Midwestern centers. Additionally, our data collection had limitations: retrospective data collection did not allow for examination of individual parental choices. BP race and ethnicity were gleaned from medical charts’ demographics, which are not always self-reported.⁵⁴  Also, characteristics of nonbirthing parents who may have been involved in decision-making were not available. Clinical information was confined to the NICU course at the delivery center, thereby limiting completeness of data on survival to home.

Rates of active perinatal and delivery room interventions for newborns born in the gray zone of viability increased over time in our cohort. Although considerable intercenter variability existed in intervention rates, this variability appeared to decrease in the more recent epoch for 23-week births. This is reassuring, given justice concerns raised by antenatal and postnatal variability in practices. If a BP with threatened preterm delivery at 23 weeks is not offered antenatal corticosteroids or delivery room resuscitation at the hospital of initial presentation but would be offered these interventions in a different hospital, are we providing equal-opportunity counseling and treatment? Attributes of birthing persons, including previous preterm delivery and Black race, also impact delivery room decision-making, warranting continued empirical study with innovative approaches capturing the complexity of this care, to better understand whether these differences represent appropriately personalized decision-making or inappropriate systemic bias.

We thank our collaborating researchers in the INDEED study group and their study centers, listed alphabetically: Advocate Christ Hospital, Oak Lawn, IL—Richard Kampanatskosol, DO, Mina Sohrabi, DO, MS; Brandon Regional Hospital, Brandon Florida—Sanela Andelija, DO; Indiana University—Shelley M. Hoffman, MPH, Jasmine Soo, MD, Vanessa Villalobos, BA; Northwestern University, Chicago, IL—Jessica T. Fry, MD, Patrick Myers, MD; University of Chicago—Christine Carlos, MD; University of Michigan, Ann Arbor –Aunum P. Akhter, MD; University of Hawaii, Honolulu—Megan Stevenson, MD; University of Mississippi Medical Center, Jackson—Drew Hayslett, MD; Kristy Patel, MD; Vanderbilt University, Nashville, TN—Annica Alwine, MD, Olivia Kammegne-Simo, MD. Thank you also to Chi-Hsiung Wang, PhD at NorthShore University HealthSystem, and the statistical consulting team at University of Chicago (Minjun Cho, Sulagna Ghosh, Junming Guan, Theodora Ko, Sean O’Hagan, Minji Park) for statistical consultation.

AAP

American Academy of Pediatrics

ACOG

American College of Obstetricians and Gynecologists

AR

attempted resuscitation

BP

birthing person

CI

confidence interval

IUFD

intrauterine fetal demise

OR

odds ratio