Association of Low Hospital Birth Volume and Adverse Short-Term Outcomes for Neonates Treated with Therapeutic Hypothermia in Rural States

Abstract

Objective:

We hypothesized that outborn neonates in three rural states would have more frequent adverse short-term outcomes following therapeutic hypothermia (TH).

Study Design:

Multicenter retrospective study comparing outcomes for low (<500 births/year), medium (501–1500 births/year), and high (>1500 births/year) birth volume hospitals in Northern New England. Multivariable logistic regression assessed the combined outcome of death/severe gray matter injury on MRI, controlling for encephalopathy severity and time to initiation of TH.

Results:

Death occurred for 35/531 neonates: 15/120 (12%) low, 7/193 (4%) medium and 13/218 (6%) for high birth volume hospitals (p=0.008). Severe gray matter injury occurred in 8%, 6% and 7% of low, medium and high birth volume hospitals, respectively (p=0.7). Odds of the combined outcome were 4.3-fold higher in low versus high volume hospitals (95% CI = 1.6, 12.1, p=0.004).

Conclusion:

Neonates born in low volume birth hospitals had significantly higher odds of death following treatment with TH.

Introduction:

Timely identification and treatment of neonates who meet criteria for therapeutic hypothermia (TH) poses significant challenges in rural settings, where access to intensive care, neonatal specialists, and diagnostic resources may be limited. The literature is variable regarding whether outborn neonates treated with TH have more frequent adverse outcomes compared to inborn neonates. Assessing this association is challenging due various factors that are linked with being born in a non-tertiary care setting. Outborn neonates are often not resuscitated by clinicians with the same skill set and experience as members of the multidisciplinary resuscitation team in a tertiary care center, suggesting that resuscitation practices may play a role. Outborn neonates also must be transferred from the birth hospital to the NICU, which can delay initiation of TH and make early temperature management during transport challenging(1–3). Time to reach target temperature for effective therapy can also vary and occur beyond the six-hour goal(4). Lastly, there may be differences in the overall health and pregnancy care of the mothers delivering in one setting versus the other.

Many studies have commented on the presence or absence of differences in outcomes for neonates born outside a hypothermia center, even if the study was not expressly designed to assess this question. Findings have ranged from studies that report that the majority of deaths occur in outborns(5), to a re-analysis of clinical trial data showing no difference in the benefit of therapeutic hypothermia for those who were inborn versus outborn(6, 7). However, other evidence suggests worse outcomes among outborn encephalopathic neonates with lower rates of seizure free survival(8, 9). Sabsabi et. al concluded that the level of care of the birth hospital in a metropolitan region of Canada significantly impacts outcomes, with neonates born at low-acuity centers experiencing worse outcomes compared to neonates born at intermediate and high-acuity medical centers(10).

Given these conflicting findings, we sought to evaluate data from the Northeast Regional Hypothermia Consortium, a data repository we created, that represents three rural states in Northern New England, each with only one or two NICUs. We aimed to reconsider outborn neonates not as a single homogeneous group, but rather as a population with varying levels of risk for adverse outcomes based on the annual birth volume of their hospital of origin. We hypothesized that the risk of short-term adverse outcomes would be higher for outborn neonates transported from low (up to 500 births per year) and medium birth volume hospitals (501–1500 births per year), compared to those born in high volume hospitals (more than 1500 births per year), where TH is performed on-site. The goal of this investigation was to achieve a more nuanced understanding of the risks associated with being born in a rural setting and outcome after TH.

Methods:

Study Design:

A retrospective cohort study of neonates treated with TH from 2010–2024 at four tertiary care centers that participate in the Northeast Regional Hypothermia Consortium (see Supplement 1 for details about each institution).

Primary Outcome:

The primary outcome was defined a priori as a combined outcome of in-hospital mortality or severe gray matter injury on brain MRI. This composite outcome was created to capture both mortality and survival with catastrophic brain injury, as in many cases, parental decision making is the factor determining which of these two outcomes occurs (11, 12). Severe gray matter injury was defined by gray matter sub-score ≥9.5 per the Weeke scoring system(13). This score assigns points for injury appreciated on diffusion weighted or conventional MRI in the following structures: thalamus, basal ganglia, posterior limb of the internal capsule, brainstem, peri-rolandic cortex, and hippocampus (maximum score 23). Each area is assessed as either 0 points (no injury), focal injury defined as affecting less than 50% of the structure (1 point), extensive injury affecting more than 50% of the structure (2 points) and additional 2 points if the injury is bilateral. Gray matter sub-score ≥9.5 is 92% sensitive and 95% specific for an abnormal developmental outcome at 2 years defined as death, cerebral palsy (Gross Motor Function Classification System(14) ≥ II), or Bayley Scales of Infant and Toddler Development(15) with scores <85. The gray matter injury score was assigned by a team including a neuroradiologist and a pediatric neurologist who reviewed images from all sites.

Classification of hospitals:

Hospitals were grouped according to average annual birth volume. Low birth volume hospitals were defined as a birth rate of 1–500 neonates per year, medium defined as a birth rate of 501–1500 neonates per year, and high birth volume hospitals (tertiary care centers) all had more than 1500 births per year and were also the centers where TH was implemented.

Environment:

At all four tertiary care centers, TH is employed for neonates who are less than six hours old at the time of initiation, are greater than or equal to 35 weeks gestation, and who have evidence of perinatal asphyxia with moderate or severe encephalopathy categorized by exam during the first six-hours of life. All four institutions initiate either passive or active cooling on transport depending on the availability of equipment, maintain the 33–34°C temperature for 72 hours and rewarm over 12 hours with continuous EEG monitoring.

Study Inclusion/Exclusion Criteria:

Neonates were included if they were treated with TH for HIE at any of the four centers. We excluded neonates who were less than 35 weeks’ gestation, completed less than 72 hours of TH, did not undergo brain MRI per family preference, or were transferred to a different institution due to the potential need for extracorporeal membrane oxygenation (ECMO) prior to obtaining a brain MRI. Neonates who were delivered at home or in small birthing centers were also excluded (Figure 1). Neonates with mild encephalopathy were not excluded from analysis.

Figure 1:

Flow diagram

Ethics approval and consent to participate:

Data were manually extracted from the electronic medical record by trained research coordinators at each of the tertiary care centers. Local Institutional Review Board approvals were obtained for data collection at each tertiary care center and approved as exempt. Additionally, the MaineHealth Institutional Review Board reviewed the registry that houses the data from the multi-site consortium as exempt (Northeast regional therapeutic hypothermia registry-1864993). All methods were performed in accordance with institutional policies for data privacy and confidentiality, the STROBE guidelines for observational studies, and relevant ethical regulations.

Statistical Analysis:

Statistical analysis was conducted using R v.4.2.1. Baseline differences between neonates grouped according to birth volume of hospital in which they were born (low, medium and high) were compared using chi square tests or Fisher’s Exact Tests for categorical variables and t-tests or Kruskal-Wallis rank sum tests for continuous variables. Logistic regression analysis was performed to investigate the effect of birth hospital size on the odds of the composite outcome. We created a directed acyclic graph based on statistical associations and clinical knowledge to determine possible confounders, mediators, and effect modifiers (Supplemental Figure 1). We did not identify any confounders, defined as having a statistical association with both birth volume and composite outcome and not being on the causal pathway. We controlled for encephalopathy severity and time to initiation of therapeutic hypothermia, which are possible mediators on the causal pathway, to determine the direct effect of birth hospital volume on composite outcome. We additionally controlled for risk factors at birth (shoulder dystocia, cesarean section, and prolapsed cord) that had an association with the composite outcome (p <0.1) but not birth volume to improve precision of the model estimates. We also controlled for study site to account for correlation in clinical practices within sites where TH is performed. While APGAR scores had a statistical association with both birth hospital volume and composite outcome, we felt it was on the causal pathway upstream of encephalopathy severity and would have therefore been inappropriate to control for. Stepwise selection was performed on the full model using the stepAIC function from the MASS package in R to achieve the most parsimonious model. A mediation analysis with 1,000 bootstrapped samples was also performed using the mediation package in R to assess the contribution of time to initiation of TH on the difference between outcomes for low versus high birth volume hospitals and medium versus high birth volume hospitals.

Results:

A total of 531 neonates were treated with TH across the four participating centers after 38 were excluded (Figure 1). Two hundred and eighteen neonates (41%) were inborn and 313 (59%) neonates were outborn. There were 120 outborn from low birth volume hospitals and 193 from medium birth volume hospitals.

Mothers in low and medium birth volume hospitals were younger (29 and 30 years versus 31 years for large birth volume, p=0.016) and had fewer medical comorbidities (Table 1). Tobacco use was highest for mothers in low birth volume hospitals (17%), and mothers from both low (17%) and medium (14%) birth volume hospitals had higher rates of marijuana use compared to 6.8% in high birth volume volume hospitals. There were no significant differences in the rate of Cesarean section or delivery related complications, including placental abruption, cord prolapse, shoulder dystocia, or late fetal heart rate decelerations, however chorioamnionitis (15% versus roughly 4% for both low and medium) and uterine rupture occurred significantly more frequently in the high birth volume hospitals (5% in high volume versus 2% in medium and 0% in low volume). Neonates in high birth volume hospitals skewed to a younger gestational age (Table 2). The umbilical cord gases demonstrated a greater degree of acidemia for neonates born in medium (median pH 7.08) and high (median pH 7.07) compared to low birth volume hospitals (median pH 7.15, p=0.005). There were significantly fewer cord gases collected in both low (49% not collected) and medium (30% not collected) compared to high (18%) volume birth volume hospitals, p<0.0001. Low birth volume hospitals had the greatest proportions of neonates with mild encephalopathy (22% versus 13% and 11% in medium and high volume hospitals respectively, p=0.009) (Table 2).

Table 1:

Characteristics of Birth Parent and Delivery by Hospital Birth Volume

Characteristic	0–500, N = 1201	501–1500, N = 1931	1501+, N = 2181	p-value2
Maternal Age	29.0 (23.0, 32.2)	30.0 (24.0, 33.0)	31.0 (26.0, 34.0)	0.016
Gestational Diabetes	11 (9.2%)	20 (10%)	28 (13%)	0.6
Preeclampsia/Eclampsia	11 (9.2%)	12 (6.2%)	29 (13%)	0.053
Multiple Gestation	1 (0.8%)	1 (0.5%)	12 (5.5%)	0.003
Tobacco	20 (17%)	16 (8.3%)	17 (7.8%)	0.021
Opioids	11 (9.2%)	10 (5.2%)	16 (7.3%)	0.4
SSRIs	9 (7.5%)	26 (13%)	24 (11%)	0.3
Benzodiazepines	1 (0.8%)	4 (2.1%)	2 (0.9%)	0.6
Marijuana	20 (17%)	27 (14%)	15 (6.8%)	0.013
Maternal fever	9 (7.5%)	8 (4.1%)	14 (6.4%)	0.4
GBS positive	29 (24%)	32 (17%)	56 (26%)	0.083
Chorioamnionitis	5 (4.2%)	7 (3.6%)	32 (15%)	<0.001
Prolonged Rupture of Membranes	18 (15%)	24 (12%)	29 (13%)	0.8
Late Decelerations	17 (14%)	33 (17%)	26 (12%)	0.3
C-Section	63 (52%)	98 (51%)	120 (55%)	0.7
Shoulder Dystocia	11 (9.2%)	19 (9.8%)	27 (12%)	0.6
Prolapsed Cord	3 (2.5%)	7 (3.6%)	6 (2.7%)	0.9
Placental Abruption	9 (7.5%)	24 (12%)	18 (8.2%)	0.2
Uterine Rupture	0 (0%)	4 (2.1%)	10 (4.6%)	0.03

¹Median (IQR); n (%)

²Kruskal-Wallis rank sum test; Pearson's Chi-squared test; Fisher's exact test

Table 2:

Characteristics of Newborns by Hospital Birth Volume

Characteristic	0–500, N = 1201	501–1500, N = 1931	1501+, N = 2181	p-value2
Gestational Age	39.00 (38.00, 40.00)	39.00 (38.00, 40.00)	39.00 (37.00, 40.00)	0.002
Gestational Age <37 wks	13 (11%)	18 (9.3%)	33 (15%)	0.2
Birth Weight (kg)	3.40 (3.00, 3.75)	3.35 (2.96, 3.73)	3.27 (2.89, 3.67)	0.2
Male Sex	65 (54%)	109 (56%)	125 (57%)	0.9
APGAR 1 min	2 (1, 3)	2 (1, 3)	2 (1, 3)	0.2
Not documented	1	0	0
APGAR 5 min	4. (3, 5)	5 (3, 7)	5 (3, 6)	<0.001
Not documented	1	0	0
APGAR 10 min	6 (4, 7)	6 (5, 8)	6 (5, 7)	0.008
Not documented	15 (13%)	22 (11%)	36 (17%)	0.3
Arterial Cord pH	7.15 (7.02, 7.27)	7.08 (6.96, 7.20)	7.07 (6.94, 7.20)	0.005
Not collected	59 (49%)	58 (30%)	39 (18%)	<0.001
Venous Cord pH	7.23 (7.07, 7.28)	7.19 (7.07, 7.29)	7.22 (7.08, 7.29)	>0.9
Not collected	70 (58%)	64 (33%)	36 (17%)	<0.001
Severity of HIE				0.009
Mild	27 (22%)	34 (18%)	24 (11%)
Moderate	76(63%)	146 (76%)	169 (78%)
Severe	17 (14%)	13 (6.7%)	25 (11%)

¹Median (IQR); n (%)

²Kruskal-Wallis rank sum test; Pearson's Chi-squared test; Fisher's exact test

The frequency of seizures (25%, 27% and 21% for low, medium and high birth volume hospitals, p=0.3) and severe brain injury on MRI (8%, 6% and 7% for low, medium and high birth volume hospitals p=0.7) was not statistically significantly different by hospital birth volume (Table 3). Mortality prior to hospital discharge was highest among neonates born at low birth volume hospitals (12%) compared to 3.6% and 6% for medium and high birth volume hospitals, respectively (p=0.008). The composite adverse outcome of in-hospital mortality or severe gray matter injury on MRI occurred for significantly more neonates from low birth volume hospitals (16%) compared to 6% from medium volume and 7% from high volume hospitals (p=0.002). Time to initiation of TH was delayed for both low and medium hospitals, a median of 4.5 and 4 hours, respectively, compared to a median of 2 hours for high volume hospitals (p<0.001).

Table 3:

Short-term Outcomes of Newborns by Hospital Birth Volume

Characteristic	0–500, N = 1201	501–1500, N = 1931	1501+, N = 2181	p-value2
Death	15 (12%)	7 (4%)	13 (6.0%)	0.008
Severe grey matter injury	8 (8%)	7 (6%)	8 (7%)	0.7
Death or severe grey matter injury	19 (16%)	11 (6%)	15 (7%)	0.004
Seizure	30 (25%)	53 (27%)	46 (21%)	0.3
Time to TH Initiation in  hours (IQR)	4.50 (3.00, 5.67)	4.00 (3.00, 5.00)	2.00 (1.00, 3.50)	<0.001

¹Median (IQR); n (%)

²Kruskal-Wallis rank sum test; Pearson's Chi-squared test; Fisher's exact test

The odds of gray matter injury or death was 4.3 times higher in low birth volume compared to high birth volume hospitals (95% CI = 1.6, 12.1, p=0.004), controlling for encephalopathy severity as the single variable that remained after stepwise selection. There was no difference in odds of gray matter injury or death between medium and high birth volume hospitals (p = 0.498). Time to initiation of TH did not mediate the association between low versus high birth volume hospital and the composite outcome (p=0.946).

Discussion:

In this analysis of neonates treated with TH in rural Northern New England, we assessed outcomes for outborns according to the annual birth volume of their hospital of origin. We found that neonates born in low birth volume hospitals compared to high birth volume hospitals have significantly higher odds of death or severe grey matter injury on brain MRI and that the delayed initiation of TH did not contribute to this difference in outcome. Contrary to our hypothesis, the odds of death or severe gray matter injury on brain MRI were not higher for neonates born in medium birth volume hospitals compared to high birth volume hospitals.

Our analysis is the second to consider outborn neonates as subgroups rather than as a homogeneous entity. This type of analysis was first performed by Sabsabi et al who investigated the impact of birthplace (birth in low, intermediate or high-acuity center) on outcome for asphyxiated neonates(10). They found a higher incidence of asphyxia for neonates born in low acuity settings compared to high acuity settings. In multivariate analysis, birth in a low acuity center was associated with a 2-fold higher risk of either death or brain injury. Our results were similar, although the magnitude was higher with 4-fold increased odds of death or brain injury on MRI. One possible reason for the difference in magnitude could be that our population is primarily rural, whereas the Sabsabi study was done in an urban environment(16–18). Another possible reason could relate to the fact that the Sabsabi study was done in a country with nationalized healthcare, a system that has a known association with better outcomes(19).

Low birth volume hospitals are intrinsically different from medium and high birth volume hospitals for several reasons. First, for low birth volume hospitals, the ability to perform an emergency Cesarean section within the standard 30-minute window may not be feasible(20). Low birth volume hospitals are also less likely to have in-house coverage to attend deliveries. This may leave the immediate resuscitation of an asphyxiated neonate in the hands of labor and delivery staff members who may have less experience in neonatal resuscitation(21, 22). This is not the case for medium and high birth volume hospitals, which usually have around the clock coverage provided by members of a multidisciplinary team who are more experienced in neonatal resuscitation.

In a propensity score-matched analysis of neonates born in the United Kingdom, Shipley et al reported that delivery at hospitals with TH capability was associated with a greater proportion of seizure-free survival (35%), compared to delivery at centers without the ability to provide TH (31% seizure-free survival)(8). Unlike our study, the Shipley et. al. study did not appreciate a significant difference in mortality according to location of birth. This may be attributable to the fact that in the UK, TH can be performed in intermediate and high acuity centers and in this study only 14% of the study population came from low acuity hospitals, whereas in our study nearly one-quarter of our patients are from low acuity hospitals. In another large retrospective propensity matched analysis from the Pediatrix group, the authors also did not find a statistically significant higher rate of mortality for outborn neonates(9). However, surviving outborn neonates had significantly higher rates of seizures and gastrostomy tube placement suggesting a difference in morbidity associated with perinatal hypoxia-ischemia. The outcome measures used in these large studies were chosen as available short-term outcomes since longer term outcome measures such as formal neurodevelopmental assessments such as the Bayley Scales of Infant Development (BSID) were not available. Unfortunately, our study also did not have long term data results to analyze neurodevelopmental outcomes after hospital discharge.

Time to initiation of therapeutic hypothermia has been linked to the question of outcome differences for inborn versus outborn neonates. Neonates born outside of a TH center typically have later initiation of TH, which we observed in both low and medium birth volume hospitals in this study. The relationship of earlier time to initiation of TH with improved outcome was first highlighted in preclinical experiments(23–25). In an observational study in humans, Thoresen et. al. demonstrated that surviving neonates with initiation of TH before three hours of life had better motor outcomes at 18–20 months compared to those with initiation after three hours(26). These data, combined with the knowledge that initiation of TH after 6 hours of life is associated with unclear benefit(27) resulted in a sense of urgency to initiate TH as quickly as possible.

More recent work has re-assessed this urgency. A retrospective cohort of 91 Canadian born neonates assessed outcome by early (median of 1.4 hours) versus late initiation (median of 4.4 hours) of TH, no differences in the severity of brain injury on MRI or in neurodevelopmental outcomes at 18 months were appreciated(28). A recent secondary analysis of the 500 neonates enrolled in the HEAL Study(29) also did not demonstrate differences in short-term outcomes or 2-year neurodevelopmental outcomes for early (<4 hours) versus late (>4 hours) attainment of the target temperature(30). Time to initiation is confounded by the fact that neonates with severe encephalopathy tend to have TH initiated the soonest, but also have the highest rate of adverse outcome, which was why we chose to use severity of encephalopathy by time to initiation as an interaction term in our regression analysis. Additionally, the timing of the hypoxic ischemic injury onset is often not known with certainty. Our results echo these other clinical studies(28, 30) indicating that time to initiation does not affect outcome.

Our study has several strengths, including a unique consortium, the Northeast Regional Hypothermia Consortium, formed by investigators from four NICUs in three of the most rural states in the United States: Maine, New Hampshire and Vermont. This novel consortium has allowed us to collect and analyze data that is not routinely available outside of major academic children’s hospitals. With the combination of four NICUs, our sample size is large for data sets emerging out of rural areas and may be generalizable to other more rural locations. We used advanced statistical analyses to assess the relationship between outcome and birth volume taking into consideration both severity of encephalopathy and time to initiation of TH, as well as other potential confounders such as site. Additionally, we collected the degree of encephalopathy for all neonates in the study, and our combined outcome captures both mortality and the severe impairment expected in children with high gray matter injury scores on MRI(31). This combined outcome is key in our opinion as the consequence of either death or severe gray matter injury is most often the result of parental decision making rather than the inability of a neonate to survive(32). Another strength is that we chose to exclude those born in birthing centers and home births which confer a higher risk for adverse outcome(33).

There are several weaknesses in this study, which include the lack of comprehensive data on longer term neurodevelopmental outcomes, and no data on time from birth to target temperature. Additionally, some data on outborn neonates was missing. For example, there were many more missing cord gases for neonates born in both low and medium birth volume hospitals, a challenge reported by others studying TH and outcomes in outborn neonates(3). There is also likely to be some residual confounding from sociodemographic variable differences for deliveries that occur in low and medium birth volume hospitals (e.g. incomplete data on drug exposure, socioeconomic status and prenatal care). This may have resulted in an overestimate of the difference in odds of adverse outcome between low and high birth volume hospitals. Finally, although each institution follows a standardized hypothermia protocol that has strong similarities between institutions, there are likely nuanced differences between institutions that are not captured by the data. We attempted to address this by controlling for site in the stepwise regression model. An additional limitation is the potential for changes in practice across time from 2010–2024. Use of a structured neonatal encephalopathy exam came into more widespread use during the epoch studied and it is very possible that intra-site differences in quality and conclusions of the exam changed during that time, particularly in differentiating mild from moderate encephalopathy. Lastly, we chose to include neonates who were treated with TH who had mild encephalopathy, which may limit generalizability to institutions that do not treat mild NE.

Conclusions:

This analysis evaluated neonates treated with therapeutic hypothermia from three rural states and separated the neonates that were treated with TH into subgroups based on the birth volume of the hospital of origin. We found that births in low compared to high birth volume hospitals have significantly higher odds of death or severe gray matter injury on MRI. This was not the case for medium volume hospitals, and our results did not show an effect of time to initiate TH on outcome. Further analysis is needed to delineate the obstetric and neonatal resuscitation factors that may be involved in our finding of increased morbidity and mortality associated with low birth volume hospitals in these three rural states. Potential strategies to support neonatal resuscitation and stabilization may be needed in these small, often critical access rural community hospitals. The use of telehealth modalities may facilitate interactions between expert subspecialists and primary care providers in these efforts to improve outcomes.

Abbreviations:

EEG

Electroencephalogram

HIE

Hypoxic Ischemic Encephalopathy

MMC

Maine Medical Center

MRI

Magnetic resonance imaging

NE

Neonatal Encephalopathy

NICU

Neonatal Intensive Care Unit

NLEMMC

Northern Light Eastern Maine Medical Center

TH

Therapeutic hypothermia

UVMMC

University of Vermont Medical Center

Data Availability Statement:

The datasets generated and analyzed for the current study are available from the corresponding author on reasonable request.