Neonatal Admission Temperature in Middle- and High-Income Countries

VON is a voluntary worldwide community of practice dedicated to improving the quality, safety, and value of newborn care through a coordinated program of data-driven quality improvement, education, and research. The University of Vermont’s Institutional Review Board determined that, because the use of data from the VON Research Repository for this analysis was completely deidentified and posed no risk to human subjects, it is not human subject research.

VON member hospitals contributed data on infants 401 to 1500 g or 22 0/7 to 29 6/7 weeks’ gestational age who were inborn or transferred to the reporting hospital within 28 days of birth. This analysis included inborn infants delivered from January 1, 2018 to December 31, 2021, in 12 MICs and 22 HICs that contributed data during this time period. All data were collected by local staff by using standardized definitions,¹⁰  which facilitated comparisons across countries. Mortality of infants who were transferred was tracked until infants’ ultimate disposition during the birth hospitalization.

Hospital characteristics were obtained from the VON annual member survey. Hospitals were divided into 4 groups on the basis of whether the hospital was required to transfer infants for assisted ventilation or whether 1 of 8 surgeries was performed at the hospital (omphalocele repair, ventriculoperitoneal shunt, tracheoesophageal fistula and/or esophageal atresia repair, bowel resection and/or reanastomosis, meningomyelocele repair, patent ductus arteriosus ligation, cardiac catheterization, or cardiac surgery requiring bypass). The 4 groups were (1) ventilation restrictions, no surgery, (2) no ventilation restrictions, no surgery, (3) no ventilation restrictions, surgery except cardiac requiring bypass, (4) and no ventilation restrictions, surgery including cardiac surgery requiring bypass.

The exposure variable was temperature within the first hour after admission to the NICU in °C. If multiple temperatures were taken, abstractors were instructed to record the first temperature. The order of preference for the type of temperature was rectal, esophageal, tympanic, or axillary. Temperature was categorized as hypothermic (<36.5°C), normothermic (36.5–37.5°C), or hyperthermic (>37.5°C).¹¹ 

Outcomes included mortality before initial discharge and key morbidities of prematurity. Chronic lung disease (CLD) was defined as oxygen requirement at 36 weeks postmenstrual age (PMA) or at discharge for infants discharged at 34 or 35 weeks PMA. Early bacterial infections occurred on or before day 3 from birth and included pathogens on a specified list recovered from blood or cerebrospinal fluid. Late infections occurred on day 4 from birth or later and included bacterial pathogens on a specified list recovered from blood or cerebrospinal fluid, coagulase-negative Staphylococcus recovered from blood or cerebrospinal fluid plus 1 or more sign(s) of infection and treatment with at least 5 days of intravenous antibiotics, or fungus recovered from a blood culture. Infants could have >1 type of infection. Severe intraventricular hemorrhage (IVH) included grades 3 or 4 occurring on or before day 28 from birth. Periventricular leukomalacia (PVL) could occur at any time. Definitions for variables describing birth, patient information, procedures, and interventions are in the VON Manual of Operations.¹⁰ 

All analyses were stratified by receipt of care at a hospital in an HIC (Austria, Canada, Chile, Denmark, Finland, Germany, Hungary, Ireland, Italy, Kuwait, Poland, Portugal, Puerto Rico, Qatar, Saudi Arabia, Singapore, Slovenia, Spain, Switzerland, Taiwan, United Arab Emirates, United Kingdom, and the United States) compared with MIC (Argentina, Botswana, Brazil, China, Columbia, India, Malaysia, Mexico, Namibia, Pakistan, South Africa, Turkey) as defined by the World Bank in 2023.¹²  Generalized estimating equation regression models were used to predict risk ratios (RR; using a Poisson distribution and log link)¹³  between neonatal hypothermia versus euthermia and outcomes, controlling for gestational age, small for gestational age (SGA; <10% according to Fenton growth charts),¹⁴  congenital anomaly, 1-minute Apgar score, multiple births, vaginal delivery, infant sex, and clustering of infants within hospitals and hospitals within countries.

After excluding 44 582 patients because of delivery room mortality, missing temperature data, or outborn status, 201 046 newborns remained, of whom 32 423 were reported by 12 MICs in 184 hospitals and 168 623 were reported by 22 HICs in 902 hospitals. Characteristics of hospitals in this sample varied by income status (Table 1).

The median altitude of hospitals in MICs was higher than for HICs (2363 compared with 381 feet above sea level). Median winter weather was warmer at hospitals in MICs compared with HICs (median of 9.4°C compared with 1.7°C).

The distribution of infants by admission temperature is shown in Fig 1. Overall, MIC hospitals had a substantially higher proportion of newborns admitted with hypothermia (64.0% vs 28.6%) and a lower proportion of those admitted with hyperthermia (2.0% vs 7.6%). The percentage of SGA infants who were hypothermic compared with euthermic was higher in both MIC (27.5% vs 22.4%) and HIC (29.8% vs 17.6%) hospitals (Table 2). The use of epinephrine or chest compressions was double in hypothermic newborns in both MIC (6.1% vs 3.2%) and HIC (4.6% vs 2.3%) hospitals. The percentage of newborns delivered by Cesarean section who were hypothermic versus euthermic were similar in MIC (25.3% vs 21.2%) compared with HIC (21.7% vs. 23.2%) hospitals. Chorioamnionitis was diagnosed less frequently in MIC compared with HIC hospitals in both hypothermic (6.6% vs 11.5%) and euthermic (7.3% vs 13.9%) newborns.

At discharge, infants who had been hypothermic on admission to both MIC and HIC hospitals had a discharge weight <10th percentile more often (75.4% and 57%) than those admitted euthermic (68.8% and 42.7%) respectively. Similarly, infants who had been hypothermic on admission to both MIC and HIC hospitals had a discharge head circumference <10th percentile more often (42.4% and 38.5%) than those admitted euthermic (37% and 29.9%) respectively.

Overall, 84.6% of newborns with birth weight (BW) <500 g in MIC hospitals were hypothermic (and 14% were extremely hypothermic [<34°C]) compared with 52.0% of newborns with BW <500 g with hypothermia (and 2.6% with extreme hypothermia) in HIC hospitals (Fig 2). Inversely, with increasing BW of newborns, the proportion of those with admission hypothermia declined to 55.8% and 17.9% in infants >1500 gm in MIC and HIC hospitals, respectively.

Mortality was approximately double for hypothermic compared with euthermic newborns in both MIC (24.7% and 15.4%) and HIC (12.7% and 7.6%) hospitals (Table 2). After adjustment with regression modeling, the relative risk of death among hypothermic newborns was 1.21 (95% confidence interval [CI] 1.09–1.33) in MIC and 1.26 (95% CI 1.21–1.31) in HIC hospitals (Table 3). Every 1°C increase in admission temperature was associated with a 9% (RR 0.91; 0.88–0.95) and 10% (RR 0.90; 0.89–0.92) decrease in the risk of mortality for MIC and HIC respectively. CLD among survivors was 10% lower in MIC than in HIC hospitals but did not differ by admission temperature before or after adjustment (Tables 2 and 3). The risk of infection and neurologic complications were higher in hypothermic newborns in MIC and HIC hospitals but after adjustment, this risk remained significant in HIC hospitals only.

Extensive literature reveals the increased risk of both morbidity and mortality for newborns who are hypothermic on admission to a neonatal ward.^4,5  Yet, this literature draws from relatively limited sample sizes and typically from a single income level, whether low, middle, or high. The VON database offers data on neonatal hypothermia in a sample size that is orders of magnitude larger than that of published research to date on this topic, enabling us to assess admission hypothermia, associations, and effect sizes from a global perspective. Previous research has revealed a wide range in the incidence of admission hypothermia in VLBW infants on the basis of definitions of hypothermia and level of resources,^5,14–21  with a trend toward decreasing incidence over time.^4,5  Although our data only spans 4 years, we found that hypothermia on admission is a common problem, even in HICs.

Multiple mechanisms may contribute to the effects of hypothermia on morbidity and mortality. It may alter neonatal transition to extrauterine physiology,⁸  causing abnormal respiratory and metabolic function, which leads to hypoxia,^1,22  respiratory distress, pulmonary hypertension,¹⁷  acidosis,^7,17,23  impaired fluid balance,¹  hypoglycemia,^1,17,18  hyperkalemia,¹  and the accumulation of toxic metabolic products.¹  Low birth weight newborns are at particular risk of hypothermia given their higher surface area to weight ratio,¹  immature skin,¹  decreased vernix caseosa,⁵  decreased subcutaneous fat,¹  inability to shiver, and inefficient vascular control needed for thermoregulation.¹⁷  Therefore, analysis of the VON database, with its extensive information on a large sample size of VLBW or preterm infants (22 0/7 to 29 6/7 weeks’ gestation) across MICs and HICs lends itself well to a deeper understanding of hypothermia.

Our data focuses on hypothermia on admission to a neonatal ward among inborn infants, reflecting thermal management in the delivery room, intra-facility transport, and admission practices. This immediate postnatal period is particularly high-risk for the development of hypothermia, reported to result in 2 to 4°C of heat loss in the first 10 minutes after birth unless measures are taken to keep newborns warm.¹  Interfacility transport introduces multiple additional risk factors for hypothermia.²⁴  Although hypothermia can be an issue throughout the hospitalization, focusing on admission hypothermia highlights the incidence of this preventable, dangerous condition and offers insight into gaps in care during this particularly high-risk window.

Previous literature revealed hypothermia to be more common after Cesarean deliveries, attributed to both colder operating room than delivery room temperatures, and the use of maternal neuraxial anesthetics that lower the body’s sympathetic response leading to a redistribution of blood flow from the core to the periphery.²⁵  Our data did not confirm this differential risk of hypothermia based on the mode of delivery.

We confirmed a strong relationship between hypothermia and BW, SGA status, weight and head circumference <10th percentile, and head circumference <10th percentile at discharge from hospital, as previously described.²⁶  The increased risk of hypothermia associated with the receipt of chest compressions or epinephrine is likely explained by both severity of illness and suboptimal use of exogenous heat sources during a prolonged resuscitation. It may also be confounded by the effect of resources, including human resources, on the prevention of hypothermia and the ability to provide effective early resuscitative measures.

Hypothermia has been associated with both respiratory distress^6,7,27  and CLD.^6,8  These respiratory complications have been attributed to abnormal surfactant distribution^1,7  and changes in pulmonary mechanics, including pulmonary venous constriction, increased pulmonary vascular resistance, and decreased left atrial pressure. However, our data did not confirm previously reported associations between hypothermia and CLD.

The authors of previous literature have reported associations between hypothermia and infections,⁶  including early onset,⁴  late onset,⁵  and nosocomial sepsis.⁷  After adjusting for confounders, the association between hypothermia and infection persisted only in HICs. Although the immature immune system of VLBW infants cannot be directly compared with older patients, it may be relevant that perioperative hypothermia in adults is associated with an increased risk of postoperative infection, attributed to a temperature-mediated immune impairment.²⁸  Encouragingly, decreasing perioperative hypothermia decreased these postoperative infections,²⁹  suggesting that temperature correction can improve outcomes.

Hypothermia has also been associated with IVH,^6,7  PVL,⁶  and poor neurodevelopmental outcome at 18 to 21 months.⁸  Again, after adjusting for confounders, we found that the association of hypothermia and neurologic complications persisted only in HICs. This finding may be explained by the higher mortality in MICs, with the population at greatest risk of neurologic complications not surviving.

Although perhaps not mediated via respiratory, infectious, or neurologic complications, hypothermia is consistently associated with higher in-hospital mortality.⁴  Our data confirmed this association in both MICs and HICs, even after adjusting for confounders, with a slightly higher RR in HICs. The authors of previous literature estimated that every 1°C decrease in admission temperature led to a 28% increase in mortality.⁵  We found that, for every 1°C increase in admission temperature, there was an ∼10% decreased mortality risk across all settings.

Although not a focus of this manuscript, neonatal admission hyperthermia, although rare, has become more common over time. This has been attributed to the overeager use of multiple modes of heat loss reduction and heat provision, against a background of maternal pyrexia.⁷  Previous studies revealed the proportion of infants with hyperthermia on admission ranged from 2.8% to 7.1%.^4,7,30,31  Not unexpectedly we found that newborns born to mothers with chorioamnionitis, both in MICs and HICs, were more likely to experience hyperthermia. The proportion of infants with hyperthermia in our MIC cohort was lower than those previously reported (2%) and consistent with the higher end in our HIC cohort (7.6%), supporting the theory that more readily available thermoregulatory options increase the risk of iatrogenic hyperthermia.

There are inherent limitations in our analysis of admission hypothermia based on the population and data included in the VON database used for this analysis. The VON database is composed of infants with BW <1501 g or gestational age <30 weeks, although newborns with higher BW and older gestational ages are also at risk for hypothermia.³⁰  We do not know delivery room temperatures, found to be a high-ranking contributor to hypothermia.⁷  Country-wide categorization by the World Bank income classification has inherent limitations due to the variability in resources within the same country that is well-described, especially for low- and middle-income settings.³²  The predominance of private hospitals in the MIC cohort could underestimate the country-wide proportion of infants with admission hypothermia because they may have funding to provide more timely and effective care than public hospitals in the same region.³¹  Although the higher mean altitude in MICs might be expected to track with lower temperatures because MICs are also closer to the equator, their median temperature was actually higher than HICs.³³ 

We also have no data on efforts to reduce heat loss, such as wraps or exothermic mattresses. Additionally, we do not know maternal temperatures proximate to delivery, more relevant for neonatal hyperthermia. We do not know the time of birth and, therefore, cannot report the potential effects on hypothermia because of less staffing and the potential for colder overnight temperatures. This risk may be augmented in MIC hospitals because of higher elevation, at which temperatures tend to fall more precipitously, and indoor heating may be suboptimal. The temperature data combines axillary, esophageal, tympanic, and rectal sources, with axillary temperatures estimated to be 0.17°C lower than rectal.^34–36  Pregnancies in MICs may have less accurate gestational age dating that could affect some of our outcomes such as the diagnosis of SGA and the determination of CLD at 36 weeks PMA.³⁷ 

We only have data on admission hypothermia and not hypothermia during the remainder of the hospitalization. The short exposure to hypothermia captured in this population may explain why we did not see more detectable effects on respiratory, infectious, neurologic, or other outcomes. It is also all the more remarkable that we found such a pronounced effect on mortality risk. Although we would ideally include LICs, VON only had 2 LICs in its database during the time of this study. As LIC membership grows, further global landscaping and analyses will be feasible.

We describe admission hypothermia and the associated maternal and neonatal characteristics across MICs and HICs. Lower BW, SGA status, and prolonged resuscitation stood out as perinatal risk factors. Exposure to admission hypothermia increased mortality risk, even after adjusting for confounders. Overall, hypothermia remains a common problem in both MICs and HICs. New technologies should focus on decreasing hypothermia without increasing hyperthermia.¹  Health care workers could benefit from education regarding the importance of euthermia, as well as the efficacy of available interventions,¹  including the promotion of Kangaroo Mother Care.^38,39  Health care administrators and leaders should recognize the importance of hypothermia and facilitate structural changes within the health system to optimize thermoregulation. Urging the medical community caring for newborns around the globe to renew their focus on this often-neglected vital sign is a key strategy to improve outcomes and achieve the neonatal 2030 Sustainable Development Goal.

BW

birth weight

CI

confidence interval

CLD

chronic lung disease

HIC

high-income countries

IVH

intraventricular hemorrhage

LIC

low-income countries

MIC

middle-income countries

PMA

postmenstrual age

PVL

periventricular leukomalacia

RR

risk ratio

SGA

small for gestational age

VLBW

very low birth weight

VON

Vermont Oxford Network