The reference group is the 3302 infants born to mothers who presented and delivered at level III hospitals. The antenatal transfer group includes the 677 infants born to women who were transferred to a level III hospital. The non–level III hospital group includes the 838 infants born at non–level III hospitals. The reference, antenatal transfer, and non-level III birth groups represent the study groups. Together, the reference and antenatal transfer group represent inborn infants who were VPT.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Shah KP, deRegnier RO, Grobman WA, Bennett AC. Neonatal Mortality After Interhospital Transfer of Pregnant Women for Imminent Very Preterm Birth in Illinois. JAMA Pediatr. 2020;174(4):358–365. doi:10.1001/jamapediatrics.2019.6055
Is antenatal transfer to a level III hospital prior to very preterm delivery associated with neonatal mortality?
In this cross-sectional study, neonates born very preterm in level III hospitals had similar odds of mortality, regardless of whether women presented directly or were transferred to the level III hospital prior to delivery. Neonates born very preterm in non–level III hospitals had higher odds of mortality than when women presented for delivery at level III hospitals.
Increasing antenatal transfers to level III hospitals for women with threatened very preterm delivery may be one means to lessen neonatal mortality in infants who are very preterm.
Reducing neonatal mortality is a national health care priority. Understanding the association between neonatal mortality and antenatal transfer of pregnant women to a level III perinatal hospital for delivery of infants who are very preterm (VPT) may help identify opportunities for improvement.
To assess whether antenatal transfer to a level III hospital is associated with neonatal mortality in infants who are VPT.
Design, Setting, and Participants
This population-based cross-sectional study included infants who were born VPT to Illinois residents in Illinois perinatal-network hospitals between January 1, 2015, and December 31, 2016, and followed up for 28 days after birth. Data analysis was conducted from June 2017 to September 2018.
Delivery of an infant who was VPT at a (1) level III hospital after maternal presentation at that hospital (reference group), (2) a level III hospital after antenatal (in utero) transfer from another hospital, or (3) a non–level III hospital.
Main Outcomes and Measures
The study included 4817 infants who were VPT (gestational age, 22-31 completed weeks) and were born to Illinois residents in 2015 and 2016. Of those, 3302 infants (68.5%) were born at a level III hospital after maternal presentation at that hospital, 677 (14.1%) were born at a level III hospital after antenatal transfer, and 838 (17.4%) were born at a non–level III hospital. Neonatal mortality for all infants who were VPT included in this study was 573 of 4817 infants (11.9%). The neonatal mortality was 10.7% for the reference group (362 of 3302 infants), 9.8% for the antenatal transfer group (66 of 677 infants), and 17.3% for the non–level III birth group (145 of 838 infants). When adjusted for significant social and medical characteristics, infants born VPT at a level III hospital after antenatal transfer from another facility had a similar risk of neonatal mortality as infants born at a level III hospital (odds ratio, 0.79 [95% CI, 0.56-1.13]) after maternal presentation at the same hospital. Infants born at a non–level III hospital had an increased risk of neonatal mortality compared with infants born at a level III hospital after maternal presentation to the same hospital (odds ratio, 1.52 [95% CI, 1.14-2.02]).
Conclusions and Relevance
The risk of neonatal mortality was similar for infants who were VPT, whether women initially presented at a level III hospital or were transferred to a level III hospital before delivery. This suggests that the increased risk of mortality associated with delivery at a non–level III hospital may be mitigated by optimizing opportunities for early maternal transfer to a level III hospital.
The reduction of neonatal mortality has been and continues to be a public health goal.1-4 In the United States, approximately half of all deaths of children younger than 5 years and two-thirds of all deaths of children younger than 1 year occur within the first 28 days after birth. Prematurity is a major contributor to these deaths.5-9
One strategy to reduce infant mortality has been to improve risk-appropriate care through perinatal regionalization. Perinatal regionalization aims to improve risk-appropriate care by encouraging the delivery or transfer of infants to a level of care consistent with their medical needs.2,10,11 Previous studies3,12 have shown that infants who are very preterm (VPT; born at less than 32 weeks’ gestational age) and infants who have very low birth weight (VLBW; born at birth weights less than 1500 g) are less likely to die prior to hospital discharge or in the neonatal period if they are born at a hospital offering a higher level of perinatal care, and they may also face a reduced burden of morbidity. In line with these findings, there is a national goal to increase VLBW deliveries at level III and IV perinatal hospitals.3,12
Despite efforts to improve risk-appropriate care, the US rate for VLBW deliveries at level III hospitals has hovered around 75% for more than 10 years.1 Achieving delivery at a level III perinatal hospital (referred to as level III hospitals) for infants who are VPT requires that a pregnant woman either presents to a level III hospital for delivery or receives a transfer to a level III hospital from a lower level of care prior to the delivery of the infant. While previous studies have noted the association of VPT and VLBW delivery at non–level III hospitals with increased risk of mortality, there is scant recent, population-based, published literature examining the association of antenatal (in utero) transfer of pregnant women on VPT neonatal mortality.13-17 Examining antenatal transfer is specifically important and relevant, in that women who present to level III perinatal hospitals directly for care may have different social, demographic, and health profiles than women who present to non–level III hospitals.18-20 Thus the risk factors for neonatal mortality may be different. In addition, as more hospitals, particularly in the rural setting, lose obstetric resources, better understanding of the role of antenatal transfer of pregnant women is important in understanding how to improve neonatal outcomes.21,22
This study aims to better understand whether transferring pregnant women to level III hospitals prior to VPT delivery confers a reduced risk of neonatal death. Understanding the association of antenatal transfer with VPT neonatal outcomes can help guide public health resource allocation and clinician management decisions for pregnant women at risk of VPT delivery.
This is a retrospective cohort study examining the association between neonatal mortality and antenatal transfer to level III hospitals in a population of infants who are VPT, born at 22 to 31 completed weeks’ gestational age (or in other words, younger than 32 weeks’ gestational age) in Illinois. The study included all infants born VPT to Illinois residents in state perinatal network hospitals over the 2 years between January 1, 2015, and December 31, 2016. Prior to the study, institutional review board exemption was obtained from the Ann & Robert H. Lurie Children’s Hospital of Chicago. Informed consent was not required in this setting of minimal risk and no interaction with study participants.
The study population was stratified into 3 groups based on exposure to delivery hospital’s level of care: the reference group (infants born to women at a level III hospital after direct admission), antenatal transfer group (infants born to women who received antenatal transfer to a level III hospital from another hospital), and the non–level III birth group (infants born to women at non–level III hospitals). The reference and antenatal transfer groups are subsets of infants who were inborn. The group with VPT births at a level III hospital after direct maternal presentation to that hospital was chosen as the reference group because these infants were expected to have the lowest risk of neonatal mortality based on a review of the literature.
The Illinois perinatal network includes all hospitals within the state of Illinois, as well as 4 hospitals in St Louis, Missouri, which are included in Illinois’ perinatal network because of frequent use by Illinois residents living in the southwest region of the state. For the hospitals in St Louis, only infants born to Illinois residents are included in the state birth certificate registry and this analysis. Infants who were born to Illinois residents but delivered outside of the state perinatal network were excluded, because the level of care at delivery was not consistently available.
All data were obtained from the Illinois Department of Public Health birth and death certificate registries (vital records). These registries are described in detail elsewhere.23 Birth and infant death certificates were probabilistically linked using LinkPlus software.24 The outcomes of interest were neonatal mortality (death in <28 days), first-week mortality (in <7 days), and first-day mortality (at <24 hours).
In this study, level of care was based on the Illinois perinatal levels of care, which are described in the Illinois administrative code.25 At the time of this study, the highest designation of level of care in Illinois was level III. Level III hospitals manage neonates at high risk (including the full range of gestational ages and birth weights), provide invasive and noninvasive ventilation, and have access to pediatric subspecialty services. Some level III hospitals also offer surgical subspecialty services. Other perinatal levels in Illinois include level II hospitals with extended neonatal capabilities (designated to care for infants of ≥30 completed weeks’ gestational age and birth weights of ≥1250 g), level II hospitals (designated to care for infants of ≥32 completed weeks’ gestational age and birth weights of ≥1500 g), level I hospitals (care for full-term newborns), and level 0 hospitals (nonbirthing). Of note, pregnant women at less than 32 weeks’ gestation who presented to a level II hospital with extended neonatal capabilities were still expected to be transferred to a level III hospital prior to delivery when possible. The care for neonates of less than 32 weeks’ gestational age was considered a neonatal exception, not a maternal exception.25
Variables examined for eligibility as covariates for logistic regression modeling included gestational age (categories: 22-23, 24-26, 27-29, or 30-31 completed weeks’ gestational age), infant sex, presence of any congenital anomalies, plurality (singleton or multiple gestation), the receipt of antenatal steroids, maternal age (<20, 20-24, 25-29, 30-34, 35-39, or ≥40 years), maternal race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, or non-Hispanic other/multiple races/ethnicities), maternal education (less than high school, high school diploma, some college, or college degree), maternal hypertension before or during the pregnancy, maternal diabetes before or during the pregnancy, maternal birth history (primiparous, multiparous with no previous preterm births, or multiparous with 1 or more previous preterm births), the method of delivery (vaginal or cesarean), and geographic variables. Except when specified, variables were categorized dichotomously.
In Illinois, all level III hospitals are in urban counties with the greatest density of level III hospitals in Cook County (which includes Chicago) and its 5 adjacent counties (DuPage, Kane, Lake, McHenry, and Will counties; also referred to as collar counties). Because proximity to a level III hospital was expected to have an association with the exposure of delivery at a level III hospital, the outcomes associated with geography were examined in 2 ways: (1) distance from maternal residential zip code to a level III hospital (<10, 10-24, 25-49, or ≥50 miles) and (2) maternal residential county type (Cook County, collar counties, other urban counties, or rural counties). Designation of urban and rural counties was based on the Illinois Behavioral Risk Factor Surveillance System classifications.26
Of note, birth weight was not adjusted for in the multivariate model because of collinearity with gestational age.27 Prenatal care adequacy was assessed using the Adequacy of Prenatal Care Utilization Index28 but was ultimately removed from all analysis because of a high level of missing values (>5%).
In bivariate analysis, unpaired t tests were used to compare means of continuous variables by exposure group. For categorical variables, χ2 tests were used to analyze the associations between each variable and the exposure. The P values for all tests were 2-sided, and statistical significance was set at P < .05.
Multivariable logistic regression analysis was used to evaluate the association between the exposure group and neonatal mortality, first-week mortality, and first-day mortality. Backward, stepwise model building was used, so that variables that were found to be significant with P < .10 in the bivariate analysis were introduced to the multivariable logistic regression. If the variable did not have a significant association with the outcome, the variable was removed. Thus, variables were removed one by one, so that the final model retained only variables with P < .10. This strategy was chosen to develop a parsimonious model with a good fit to describe the association between neonatal mortality and delivery hospital level of care and maternal transport status. Odds ratios (ORs) and 95% CIs were estimated for the associations of interest. A sensitivity analysis was done for the final neonatal mortality model to determine if restricting to infants of 24 to 31 weeks’ gestational age created meaningful changes in the odds ratios. We used SAS version 9.4 (SAS Institute) for all statistical analyses.
There were 5067 infants who were VPT and born to Illinois residents from January 1, 2015, to December 31, 2016. There were 249 infants excluded because of delivery outside of the Illinois perinatal network; 1 more was excluded because of unknown maternal transfer status. Accordingly, 4817 infants who were VPT were born in Illinois perinatal hospitals and included in this study. Of the VPT births in this study, 3302 infants (68.5%) were born at a level III hospital after maternal presentation to that hospital (the reference group), 677 (14.1%) were born at a level III hospital after antenatal transfer (the antenatal transfer group), and 838 (17.4%) were born at a non–level III hospital (the non–level III birth group) (Figure). When pregnant women who went on to deliver an infant born VPT presented to a non–level III hospital, 44.7% (n = 677/[677 + 838]) were transferred to a level III hospital prior to delivery. Thus, of the analytic sample of infants who were VPT, 3979 of 4817 (82.6%) were delivered at a level III hospital.
Table 1 shows characteristics of the study population. Infants who were VPT had no significant differences in mean gestational age or sex among the 3 study groups. Compared with the reference group, infants in the antenatal transfer group were more likely to be singletons (2313 of 3302 [70.1%] vs 545 of 677 [80.5%]; P < .001); delivered vaginally (1121 [34.0%] vs 257 [38.0%]; P = .046); and be born to women who received antenatal steroids (1900 [57.5%] vs 446 [65.9%]; P < .001), were younger (mean [SD] age, 29.9 [6.3] years vs 28.0 [6.3] years; P < .001), lived farther from a level III hospital (mean [SD] distance, 7.0 [13.4] miles vs 18.4 [20.8] miles; P < .001), and had had a previous preterm birth (444 [13.5%] vs 110 [16.3%]; P = .008). Infants in the antenatal transfer group were less likely than the reference group to be born to women who had college degrees (1113 [34.7%] vs 140 [21.2%]; P < .001), lived in Cook County (1886 [57.2%] vs 258 [38.1%]; P < .001), and had private insurance (1781 [54.7%] vs 284 [42.7%]; P < .001).
Compared with the reference group, infants in the non–level III birth group were more likely to be vaginally delivered (1121 of 3302 [34.0%] vs 400 of 838 [47.7%]; P < .001) and be born to women who live in the collar counties (631 [19.1%] vs 234 [27.9%]; P < .001) and farther from a level III hospital (mean [SD] distance, 7.0 [13.4] miles vs 16.6 [22.8] miles; P < .001). Compared with the reference group, infants in the non–level III birth group were less likely to have congenital anomalies (185 [5.6%] vs 20 [2.4%]; P < .001) and be born to women who received antenatal steroids (1900 [57.5%] vs 275 [32.8%]; P < .001), had a bachelor’s degree (1113 [34.7%] vs 148 [18.4%]; P < .001), lived in Cook County (1886 [57.2%] vs 314 [37.4%]; P < .001), had private insurance (1781 [54.7%] vs 295 [35.7%]; P < .001), and had hypertension (797 [24.1%] vs 92 [11.0%]; P < .001) or diabetes (397 [12.0%] vs 62 [7.5%]; P < .001). With regards to race/ethnicity, the reference group was born to women who were more likely to identify as non-Hispanic white (1346 [41.5%] vs 310 [37.9%]) and less likely to identify as non-Hispanic black (1019 [31.4%] vs 291 [35.6%]) or Hispanic (657 [20.2%] vs 182 [22.2%]) than infants in the non–level III birth group (P = .003; Table 1).
Compared with the reference group (444 [13.5%]), the antenatal transfer group (110 [16.3%]; P = .008) and the non–level III birth group (126 [15.2%]; P < .001) were more likely to be born to women with a history of preterm birth. The antenatal transfer group differed from non–level III birth group in that infants in the non–level III birth were less likely to have congenital anomalies (37 [5.5%] vs 20 [2.4%]; P = .002) and be born to women who received antenatal steroids (446 [65.9%] vs 275 [32.8%]; P < .001), had private insurance (284 [42.7%] vs 295 [35.7%]; P = .009), and had hypertension (187 [27.6%] vs 92 [11.0%]; P < .001) or diabetes (73 [10.8%] vs 62 [7.5%]; P = .02) (Table 1).
Table 2 shows the neonatal, first-week, and first-day mortality for infants who were VPT in the 3 study groups. Neonatal mortality for all infants who were VPT included in this study was 573 of 4817 infants (11.9%). The neonatal mortality was 10.7% for the reference group (362 of 3302 infants), 9.8% for the antenatal transfer group (66 of 677 infants), and 17.3% for the non–level III birth group (145 of 838 infants). First-week mortality was 8.8% for the reference group (292 of 3302 infants), 7.7% for the antenatal transfer group (52 of 677 infants), and 15.9% for the non–level III birth group (133 of 838 infants). First-day mortality was 5.7% for the reference group (187 of 3302 infants), 5.0% for the antenatal transfer group (34 of 677 infants), and 11.3% for the non–level III birth group (95 of 838 infants). In unadjusted analysis, odds ratios at all 3 mortality points comparing the antenatal transfer group and reference group were not significant. Conversely, the non–level III birth group had odds ratios that were significantly higher than the reference group for all mortality points (unadjusted odds ratios: neonatal mortality, 1.70 [95% CI, 1.38-2.10]; first-week mortality, 1.95 [95% CI, 1.56-2.43]; first-day mortality, 2.13 [95% CI, 1.64-2.76]).
Results for the multivariable regression analyses examining mortality after adjustment for potential covariates (with <5% missing data and P < .10) are shown in Table 2. In the backward stepwise selection of variables with P < .10, the significant variables retained in the model were gestational age, infant sex, congenital anomalies, plurality, antenatal steroids, method of delivery, maternal race/ethnicity, maternal hypertension, and maternal diabetes. Adjusted analysis showed no association between mortality in the antenatal transfer group compared with that of the reference group. The odds of neonatal mortality, first-week, and first-day mortality remained significantly higher for non–level III birth group compared with the reference group (adjusted odds ratios: neonatal mortality, 1.52 [95% CI, 1.14-2.02]; first-week mortality, 1.76 [95% CI, 1.29-2.40]; first-day mortality, 1.73 [95% CI, 1.20-2.49]).
A sensitivity analysis with the final neonatal mortality model was done with restricting to infants of 24 to 31 weeks’ gestational age. Removing infants born at 22 to 23 weeks’ gestational age from the model resulted in a change of less than 10% in the final adjusted odds ratios.
When pregnant women were transferred prior to delivery of an infant who was VPT (the antenatal transfer group), neonatal, first-week, and first-day mortality were no different than outcomes for infants born VPT to women who presented directly to level III hospital (the reference group), even after adjustment for a multitude of medical and social characteristics. Consistent with previous literature, all categories of mortality examined were highest in infants born VPT at non–level III hospitals compared with those infants delivered VPT in a level III hospital after maternal presentation to that hospital (the reference group).3,10,29 As such, national, state, and organizational recommendations have supported antenatal transfer to a level III hospital in the setting of threatened VPT or VLBW delivery.1,4,15,30 This study adds to prior work by specifically analyzing the association of antenatal transfer with the mortality of infants who were VPT through a population-based analysis.
Our findings demonstrate that antenatal transfer with subsequent VPT delivery was associated with better outcomes, since infants born VPT in a level III hospital after antenatal transfer had similar odds of mortality as infants born to women who presented directly to a level III hospital for delivery. The lower odds of neonatal mortality for infants who were VPT when born at a level III hospital, regardless of initial maternal presentation or antenatal transfer to the delivering hospital, remained present even after removing infants born in the periviable gestational age range, in which discussions around comfort care could alter clinical management decisions.
Indeed, the absolute risk difference for neonatal mortality between non–level III birth and antenatal transfer group in this study was 7.6%. Although we did not study antenatal transfers that resulted in nondelivery, the absolute risk difference suggests that approximately 13 pregnant women would have needed to be transferred to a level III hospital prior to a VPT delivery to prevent 1 neonatal death. These findings support the focus of public health efforts on a specific event, antenatal transfer, as 1 systems-based effort that can be used to reduce neonatal mortality.
Important to the discussion on antenatal transfer are the many factors that may influence where a pregnant woman presents for care, including but not limited to means of travel (eg, emergency medical services, personal vehicle), proximity, referral by a clinician or personal contact, and insurance coverage. An understanding of these factors can allow for identification of opportunities to improve risk appropriate care. Additionally, while pregnant women may have different reasons for where they present for care, the differences in social, demographic, and medical characteristics of the women in these 3 groups may be a reflection of the disparities in health care facing pregnant women. Further research is needed to understand the root cause of these disparities and support appropriate interventions. Some studies have suggested that infants born VPT and at VLBW outside of a level III hospital have higher rates of prematurity-associated morbidities and chronic disease complications.15,17,31 This study did not evaluate the association between antenatal transfer and morbidity, but further exploration of any association may have implications toward the broader outcomes of antenatal transfer. Another unanswered question is what percentage of antenatal transfers lead to delivery and, more specifically, preterm delivery. It is anticipated that transfer practices may vary significantly by referring hospital level of care, group practices, and local policies.
More than half of all neonatal deaths occurred in the first day after birth in this study. While continuing to encourage a goal of increased antenatal transfer to level III hospital for VPT delivery, it is likely that there will also continue to be a small population of pregnant women who deliver at a non–level III hospital. Thus, continued efforts to improve neonatal resuscitation skills and develop alternative support resources for non–level III hospitals, such as telemedicine opportunities, will likely benefit neonates who are VPT born at non–level III hospitals. In addition, supporting emergency medical services teams with the resources and preparation for transport of neonates who are VPT to higher levels of care may also improve neonatal outcomes for the subset of infants born VPT at non–level III hospitals.
Additional opportunities to enhance care for infants who are VPT and optimize site of delivery may be found by examining the differences in social and health characteristics between pregnant women who present to a level III hospital vs a non–level III hospital with threatened VPT delivery. For example, previous preterm birth, a known significant risk factor of future preterm birth, was not associated with increased representation in the reference or antenatal transfer groups in this study. This could represent an instance for better risk stratification of pregnant women.
There are several strengths of this study. By using state-level data, we had a large sample size available for each study group. In addition, the chance of selection bias is small and the chance of generalizability greater, because the study included nearly all VPT births in Illinois (ie, there was minimal loss of study participants because of inadequate or incomplete data). Lastly, although the accuracy of some vital statistics data has been questioned,32 the Illinois Perinatal Quality Collaborative has demonstrated more than 95% accuracy on fields relevant to this analysis after they completed their Birth Certificate Accuracy Initiative in 2015.33
Limitations also should be noted. This analysis could not identify the degree of progression of labor at the time of maternal presentation to the hospital, and thus all women may not have been equally appropriate for transfer. Also, other relevant characteristics about care may not have been able to be ascertained. For example, the timing of antenatal steroid use is not recorded in the vital statistics records. In addition, when pregnant women were transferred to a level III hospital prior to delivery, the originating hospital was not identifiable in this study. Unpublished data from one Illinois perinatal network, however, showed that transfers from a level III hospital to a level III hospital were uncommon and accounted for only 2.6% of antenatal transfers.
Antenatal transfer of pregnant women to a level III hospital prior to delivery of infants who are VPT is associated with a similar risk of neonatal mortality as when pregnant women present directly to a level III hospital for delivery. The United States continues to fall short of the risk appropriate care goal for VPT neonatal delivery at level III hospitals, as does Illinois in this study.1,4 There are opportunities in antenatal transfers to improve risk appropriate care and growing efforts to identify methods to do so.34,35 One challenge to assuring risk-appropriate care for infants who are VPT and consequently reducing neonatal mortality is the low availability of level III hospitals in some geographic areas, particularly rural communities.36 Additionally, deregionalization, the proliferation of neonatal intensive care services that outpaces local need, has decreased delivery and neonatal intensive care units volumes within institutions, which may negatively affect neonatal mortality.12,29 Antenatal transfer prior to the delivery of infants who are VPT may act as a counterbalance on the tensions against efficient functioning of perinatal regional networks and thus may help reduce neonatal mortality.
Accepted for Publication: October 25, 2019.
Corresponding Author: Kshama P. Shah, MD, Ann & Robert H. Lurie Children’s Hospital, Division of Neonatology, Department of Pediatrics, Northwestern Medicine, 225 E Chicago Ave, PO Box 45, Chicago, IL 60611 (email@example.com).
Published Online: February 17, 2020. doi:10.1001/jamapediatrics.2019.6055
Author Contributions: Dr Bennett 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.
Concept and design: All authors.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Shah, deRegnier, Bennett.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Shah, Bennett.
Administrative, technical, or material support: deRegnier, Bennett.
Supervision: deRegnier, Grobman, Bennett.
Conflict of Interest Disclosures: None reported.
Create a personal account or sign in to: