The objective of this study was to study the epidemiology of neonatal hypothermia in preterm infants using World Health Organization (WHO) temperature criteria.
A population-based cohort of 8782 very low birth weight (VLBW) infants born in California neonatal intensive care units in 2006 and 2007. Associations between admission hypothermia and maternal and neonatal characteristics and outcomes were determined using logistic regression.
In all, 56.2% of infants were hypothermic. Low birth weight, cesarean delivery and a low Apgar score were associated with hypothermia. Spontaneous labor, prolonged rupture of membranes and antenatal steroid administration were associated with decreased risk of hypothermia. Moderate hypothermia was associated with higher risk of intraventricular hemorrhage (IVH). Moderate and severe hypothermic conditions were associated with risk of death.
Hypothermia by WHO criteria is prevalent in VLBW infants and is associated with IVH and mortality. Use of WHO criteria could guide the need for quality improvement projects targeted toward the most vulnerable infants.
In both developed and developing countries, hypothermia is an important risk factor for morbidity and mortality in newborns.1, 2, 3, 4, 5 The World Health Organization (WHO) established criteria for assessing hypothermia and published a guidebook on the thermal protection of newborns in 1997, indicating that this issue is a worldwide problem even for term infants, particularly in developing nations.6 WHO classifications of hypothermia are (1) cold stress or mild hypothermia: 36.0 to 36.4 °C (96.8 to 97.5° F); (2) moderate hypothermia: 32.0 to 35.9 °C (89.6 to 96.6° F); and (3) severe hypothermia: below 32 °C (<89.6° F). The purpose of this study was to assess the validity of the WHO classifications in the context of a contemporary neonatal intensive care setting. Specifically, we wanted to (1) determine the distribution of neonatal hypothermia as defined by WHO in a population-based cohort of VLBW infants; (2) identify the clinical associations and risk factors for these classifications; and (3) assess the degree to which neonatal morbidities and mortalities were independently associated with the WHO hypothermia classifications. A better understanding of the risk factors for and the mortality and morbidities associated with cold stress, moderate and severe hypothermia in this fragile population is critical to designing appropriate advocacy and neonatal practice interventions to promote normothermia.
Patients and methods
The California Perinatal Quality Care Collaborative (CPQCC) collects clinical data in a prospective manner for infants born in 127 member hospitals. Membership is offered to any hospital in California that provides neonatal intensive care. In the study period from 1 January 2006 to 31 December 2007, greater than 90% of California's very low birth weight (VLBW) infants were cared for in CPQCC hospitals, creating a unique data set. This study was approved by the Stanford University Institutional Review Board.
The CPQCC conducts yearly data abstractor trainings at locations throughout California. Each record has a variety of range and logic checks both at the time of data collection and before data closeout, and records with excessive missing data are audited.
During the study period, there were 11 664 VLBW infants born with birth weight less than 1500 g. Of them, infants with a gestational age (GA) less than 23 weeks (n=492) or unknown GA (n=13) or born with a birth weight less than 400 g (n=37) were excluded. Patients born at non-CPQCC hospitals and subsequently transferred to a CPQCC member hospital were excluded (n=1522). Infants with neurological or gastrointestinal congenital anomalies such as anencephaly, meningomyelocele, encephalocele, gastroschisis or omphalocele (n=34) were excluded because of the risk for increased evaporative losses. Also excluded were infants who had no temperature recorded in the first hour (n=482) or who died in the delivery room and had no temperature recorded (n=302). There were 8782 infants in the final analytical cohort.
The first temperature on admission was standardized by the CPQCC guidelines. The infant's core body temperature was measured and recorded within the first hour after admission to the neonatal intensive care unit. The temperature in degrees Centigrade to the nearest tenth of a degree was recorded. If the infant's temperature was measured several times within the first hour after admission, the first temperature measurement value was entered into the database. For centers that measure temperature in degrees Fahrenheit, a Fahrenheit-to-Centigrade conversion table was provided. Rectal temperature was the preferred value; however, if unavailable, esophageal, tympanic or axillary temperature was recorded.
The admission temperature and outcomes described below were recorded at the hospital of birth. If an infant was subsequently transferred within the first 28 days after birth, the outcomes that occurred at the transfer hospital were recorded and included in our analyses. If an infant was transported after 28 days, the majority of whom would have been transported for convalescent care, any outcomes that occurred during the period at this second hospital were not recorded, with the exception of the outcome of death, which was reported regardless of transport status. If an infant transported after 28 days developed morbidity that was serious enough to warrant readmission, these outcomes were recorded and included in the analysis.
A cross-sectional data analysis was conducted to examine the distribution, predictors and outcomes associated with neonatal hypothermia. Temperature criteria were based on the WHO classifications of hypothermia and are defined as follows: (1) cold stress or mild hypothermia, 36.0 to 36.4 °C (96.8 to 97.5° F); (2) moderate hypothermia, 32.0 to 35.9 °C (89.6 to 96.6° F); and (3) severe hypothermia, below 32 °C (<89.6° F). Temperatures 38.0 °C were considered as fever. Because of the low number of infants with temperature <32 °C (n=9), severe and moderate hypothermia were combined into one category. Predictor variables were as follows: maternal and intrapartum variables (age, race/ethnicity, prenatal care, antenatal steroids, spontaneous labor, multiple births, delivery mode, maternal or uterine infection, prolonged rupture of membranes (PROM) and hypertension); infant characteristics (gender, birth weight, gestational age) and delivery room variables (Apgar scores, resuscitation including oxygen administration, CPAP, bag-mask ventilation, chest compressions and epinephrine administration). Low Apgar scores were defined as an assigned value <7. Neonatal outcome measures included necrotizing enterocolitis, severe retinopathy of prematurity (ROP stage 3 or 4), oxygen dependence at 36-week-corrected gestational age, early (on or before day 3) and late (after day 3) culture-positive bacterial sepsis, severe intraventricular hemorrhage (IVH grade 3 or 4 determined by any modality used to image the brain on or before day 28 of life) and ultimate disposition (discharge home or death). Necrotizing enterocolitis was defined as one or more clinical signs (bilious gastric aspirate or emesis, abdominal distention or occult or gross blood in stool with no apparent rectal fissure) and one or more radiographic findings (pneumatosis intestinalis, hepatobiliary gas or pneumoperitoneum). ROP stage 3 was defined as the presence of a ridge with extraretinal fibrovascular proliferation and stage 4 was defined as partial retinal detachment. Grade 3 IVH was defined as intraventricular blood with ventricular dilation and Grade 4 was defined as intraparenchymal hemorrhage.
Associations between predictor variables and hypothermia were tested in a bivariate analysis, followed by stepwise logistic regression. Separate models were constructed for combined severe and moderate, mild and severe hypothermia. We then studied the potential contribution of hypothermia to neonatal outcomes. Separate models were constructed for neonatal outcomes in which severe, moderate or mild hypothermia was the independent variable. Relevant factors were entered into forward stepwise multivariable logistic regression models with an entry criteria α-level of 0.05 to give odds ratios and 95% confidence limits for those variables associated with hypothermia. Statistical analyses were computed using SAS 9.1 (SAS, Cary, NC, USA).
The cohort of 8782 infants had a mean (±s.d.) birth weight and GA of 1072±278 g and 28.4±2.8 weeks, respectively. The mean admission temperature was 36.3 °C (s.d. 0.8), with a range of 26.1 to 39.6 °C; 30.5% were mildly hypothermic, 25.6% were moderately hypothermic, 0.1% were severely hypothermic, 43.0% were normothermic and 0.8% were noted to have a temperature 38.0 °C (Figure 1).
The incidence of hypothermia was greatest and the mean admission temperature lowest with decreasing GA and birth weight. In all, 78.7% of infants born at 23 weeks and 82.8% of infants with a birth weight of 400 to 499 g were hypothermic. As GA and birth weight increased, there was a decrease in the percentage of moderate hypothermia and an increase in the percentage of cold stress. The mean admission temperature also increased with birth weight and gestational age, with stabilization occurring after 26 weeks and between 800 and 899 g gestation (Appendix Tables A1 and A2).
Descriptive characteristics for the total study cohort, with division into normothermic, mild and moderate/severe hypothermic infants, are listed in Table 1.
Risk factors for hypothermia
On bivariate analysis, black (P=0.0001) Hispanic (P=0.0025) and other/unknown (P=0.04) races were associated with hypothermia, as were maternal hypertension, Cesarean section and low 1- and 5-min Apgar scores (P<0.0001). Antenatal steroid use (P<0.001), PROM (P<0.001), spontaneous labor (P<0.001), multiple births (P=0.006) and uterine infection (P<0.001) were associated with normothermia. There were no associations found with Asian race, maternal age, gender, prenatal care or maternal infection. Similar risk factors were seen for cold stress (details on request).
Using forward stepwise logistic regression, models were created for the combined category of moderate/severe hypothermia, for cold stress and for any hypothermia as compared with normothermia (Table 2). For the outcome of moderate/severe hypothermia, the most notable effect came from birth weight, with increasing odds of hypothermia associated with decreasing birth weight between 400 and 999 g. Cesarean mode of delivery, low 1- and 5-min Apgar scores, maternal hypertension and Black race carried higher odds of moderate/severe hypothermia. Spontaneous labor, PROM and antenatal steroid administration were associated with lower odds of moderate-to-severe hypothermia, independent of birth weight.
Factors associated with the development of cold stress were similar to those of moderate/severe hypothermia. The model for severe/moderate hypothermia showed decreasing odds with increasing birth weight, whereas the cold stress model had statistically stable odds across birth weight categories.
We evaluated the relationship between hypothermia and stage 3 or 4 ROP, necrotizing enterocolitis, early-onset sepsis, late-onset sepsis, oxygen use at 36-week-corrected gestational age, IVH, death and the combined outcome of grades 3 to 4 IVH and death. Table 3 shows the frequency of these morbidities across WHO temperature criteria. Although there were significant bivariate associations between hypothermia and ROP, late-onset sepsis, oxygen at 36 weeks, IVH and death, many of these relationships were not statistically significant after risk adjustment. After risk adjustment, we found no association between mild hypothermia and any of the morbidities or death. Moderate hypothermia was associated with higher odds of IVH (odds ratio 1.3, 95% confidence limit 1.1 to 1.6) and death (odds ratio 1.5, 95% confidence limit 1.3 to 1.9). Severe hypothermia, although seen in only nine infants, was also associated with higher odds of death (odds ratio 5.6, 95% confidence limit 1.1 to 28.1).
This study evaluated transitional hypothermia in a population-based cohort of VLBW infants born in California neonatal intensive care units using the WHO classifications of hypothermia: cold stress or mild hypothermia, 36.0 to 36.4 °C (96.8 to 97.5 °F); moderate hypothermia, 32.0 to 35.9 °C (89.6 to 96.6 °F); and severe hypothermia, below 32 °C (<89.6 °F). We found that cold stress and moderate hypothermia on admission were very common (30.5 and 25.6%, respectively), whereas severe hypothermia was seen in only 0.1% of VLBWs. Both cold stress and moderate/severe hypothermia were associated with various maternal and infant characteristics. Cold stress did not increase the odds of morbidity or death. Moderate hypothermia was associated with increased odds of IVH and death, and severe hypothermia was associated with increased odds of death. Severe hypothermia may in fact have a greater impact on morbidity; however, given the small number of infants with severe hypothermia (n=9), these associations may not have been detected.
Maintaining infant normothermia has been shown to improve survival and outcomes for decades.7, 8, 9 In previous studies, hypothermia has been associated with increased neonatal morbidity and mortality, such as IVH and death, and to have an increased association with lower GA in premature infants at the threshold of viability.3, 10 Following an initial period of stabilization, LBW infants born in the United States are routinely maintained in a variety of thermoneutral environments. Several studies and reviews have been carried out to evaluate interventions that may be valuable in decreasing post-birth heat loss in very preterm infants,11, 12, 13 and currently the latest guidelines for newborn resuscitation from the American Academy of Pediatrics emphasize the goal of preventing hypothermia in premature infants.14 Despite the emphasis placed on maintaining normothermia, no population-based investigation of the distribution and predictors of initial temperatures using WHO criteria in VLBW infants born in contemporary neonatal intensive care units and their relationship to morbidity and mortality has been published.
The WHO divides hypothermia into three major categories: severe (<32 °C), moderate (32.0 to 35.9 °C) and mild/cold stress (36.0 to 36.4 °C).6 We used the WHO criteria as it was considered to provide more clinically relevant benchmarks than defining hypothermia as a dichotomous variable or in single-degree increments. For example, if one defines hypothermia as a single cutoff point of <36.5 °C, one combines cold-stressed infants who are not at risk for short-term morbidity or death with moderately and severely hypothermic infants whose risk for IVH and death is significantly increased. In addition, the WHO categories are used worldwide.
Our study using WHO categories also suggests that there is a maturational difference between severe/moderate hypothermia and cold stress. As an infant matures in size and gestation, the odds of being severely/moderately cold and at high risk for IVH and death decreased, whereas the risk for being cold stressed without an increased risk for death or short-term morbidity increased. At around 26 weeks of gestation and/or at 900 to 1000 g, the odds for hypothermia stabilized, indicating that perhaps a critical point in the ability to thermoregulate had been reached (Appendix Figures A1 and A2).
Maternal factors such as Cesarean delivery and hypertension carried higher odds of moderate/severe hypothermia, whereas spontaneous labor, PROM and antenatal steroid administration showed decreased odds of developing moderate/severe hypothermia, independent of birth weight. Cesarean delivery may contribute to hypothermia as operating rooms are often kept at cooler temperatures for the physician's comfort. The WHO has recommended that delivery or resuscitation room temperatures be set at a minimum of 25 °C (77° F), with a suggested range of 25 to 28 °C (77 to 82.4 °F),6 which anecdotally is often not the case. Infant characteristics that were associated with hypothermia most notably included decreasing birth weight and low Apgar scores. Prematurity and LBW are associated with a large surface area-to-body mass ratio, decreased subcutaneous fat, greater body water content, immature skin leading to increased evaporative water and heat losses, a poorly developed metabolic mechanism for responding to thermal stress and delayed development of skin blood-flow control reducing the ability to maintain heat by vasoconstriction,15, 16, 17 all of which may lead to hypothermia. Low Apgar scores may be associated with increased resuscitative efforts or increased inherent illness in the infant at birth, of which hypothermia may be a marker or result.
Cold stress was also associated with LBW, Cesarean section and a low 1-min Apgar score. Decreased odds of hypothermia were found for spontaneous labor, PROM and antenatal steroid use, as well as for no resuscitative efforts in the delivery room and multiple gestation. We speculate that the reason for no resuscitative effort being associated with normothermia may be due to the likelihood that a lack of resuscitation may indicate that the infant was relatively healthier and therefore was able to better maintain core body temperature. In addition, in the midst of resuscitation, measures to keep infants warm may be difficult to maintain or neglected altogether. Although cold stress was not found to be associated with short-term morbidities, the long-term consequences of mild hypothermia remain unknown.
Neonatal outcomes were evaluated and support previous studies that showed an association between hypothermia and IVH and death.1, 2, 3 Our study revealed higher odds of IVH with moderate hypothermia and higher odds of death with both severe and moderate hypothermia. Despite the association of ROP, oxygen use at 36-week-corrected age and late sepsis with hypothermia on bivariate analyses, in risk adjusted logistic regression models, hypothermia was not a significant risk factor, likely representing the confounding effect of factors such as birth weight.
A study by Laptook et al.18 in 2007 evaluated admission temperature of LBW infants to determine predictors and associated morbidities. The distribution of temperatures was similar to that found in our study. Using multivariate analysis, they concluded that birth weight alone had a relatively small effect on hypothermia. This was in contrast to our results. Their outcome analyses revealed an association of admission temperature with death and late-onset sepsis, but not with severe IVH.
There were several limitations to our study. The exact timing that the temperature was obtained within the first hour of admission is unknown. A longer time between birth and first recorded temperature may result in lower initial temperatures; however, a study by Loughead et al.19 found no significant difference in the time passed between delivery and initial temperature with a range of less than 10 min to greater than 31 min. The method by which temperature was obtained was also unknown (rectal, axillary or skin), which may lead to differences in reported temperatures.
Insights gained from this study have the potential to facilitate quality improvement efforts both in California and nationwide. This study revealed that using the WHO criteria for hypothermia is a valuable method for evaluating infants at the greatest risk for developing low temperatures and for those at greatest risk for poor outcomes because of hypothermia. In our California cohort, 25.8% of VLBW infants were at increased risk of mortality because of moderate/severe hypothermia following birth. A review of the rates of moderate/severe hypothermia by the CPQCC center shows a wide variation in the percentage of VLBW infants found to be hypothermic, ranging from less than 5% to greater than 70% (details available on request). We suspect that this high variability may be widespread in developed countries, making reduction in the percentage of infants with moderate-to-severe hypothermia a high priority for quality improvement initiatives. A variety of interventions, such as polyethylene bags, heating mattresses or increasing the delivery room temperature, have been shown to reduce hypothermia in premature infants, both in randomized trials and in quality improvement projects.20, 21, 22 Because of its relationship to IVH and neonatal mortality, we believe that change in the percentage of moderate/severe hypothermia as defined by WHO may serve as an important indicator of the effectiveness of one's interventions. Future studies could then focus on comparison data before and after any recommended practice changes are implemented to evaluate for effectiveness in improving neonatal hypothermia and outcomes.
This project was supported in part by NIH/NCRR/OD UCSF-CTSI Grant Number KL2 RR024130. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.