Admission Temperature of Low Birth Weight Infants: Predictors and Associated Morbidities

METHODS

The study was conducted among 15 centers of the National Institute of Child Health and Human Development Neonatal Research Network. Data were retrieved on all of the neonates with the following inclusion criteria: born at a network center between January 1, 2002, and December 31, 2003; born with a birth weight of 401 to 1499 g; admitted directly to a NICU from a delivery room; and born without a major congenital anomaly. The first temperature obtained on admission of each infant to the NICU from the labor and delivery department was recorded as the admission temperature along with the date and time charted. Temperatures recorded in the delivery room or during transport to the NICU were not recorded. The site of temperature measurement (axilla, skin, or rectal) was noted. Exclusion criteria were missing admission temperature, missing time of admission temperature, or temperature recorded after 2 hours of age.

Data on each mother and infant were prospectively collected as part of an ongoing survey of neonatal morbidity and mortality initiated in 1987.¹⁰ Trained research nurses reviewed the medical charts of mother and infant and entered predefined data items into an institutional review board-approved computerized database. Neonatal outcome data were assessed at discharge from the hospital, 120 days after birth, or at the time of death, using which ever came first.

Variables explored for associations with admission temperature included the following: (a) maternal variables, including exposure to antibiotics (any use during the hospitalization for delivery), tocolytics, antenatal steroids (partial or complete course), and the presence of multiple births; (b) intrapartum variables, including the presence of labor, ruptured membranes >18 hours, and the mode of delivery; (c) infant characteristics, including birth weight, gestational age (obstetric criteria), and gender; (d) delivery room variables, including intubation and/or chest compression with or without resuscitative medications, Apgar scores, and umbilical artery pH and base excess; (e) site (axilla, rectum, or skin) and age of temperature measurements; and (f) network center of birth. Neonatal outcomes included days of conventional ventilation, late-onset sepsis (positive blood culture after 72 hours of age), necrotizing enterocolitis (NEC; modified Bell's stage IIa or above),¹¹ grade III or IV intraventricular hemorrhage (IVH), and death after 12 hours of age and before hospital discharge. The assigned cause of death reflects the purported underlying, proximate disease process contributing to death and is based on autopsy and clinical findings using predefined causes in the manual of operations for the database.

Data analysis for associations with admission temperature were initially explored with bivariate analyses between admission temperature and maternal and intrapartum variables, infant characteristics, and delivery room events. Variables significant at a .10 level of significance in bivariate analyses were entered into multivariable linear regressions. Umbilical artery pH and base deficit were not included in the multivariable analysis, because values were available only for a subset of the cohort. Gestational age and chest compressions/resuscitative medications were not included because of collinearity with birth weight and intubation, respectively. One center was designated the reference center on the basis of the highest percentage of admission temperatures between 36 and 36.9°C (center 10). Center results in multivariable analyses were expressed relative to center 10.

In a similar fashion, analyses for associations between admission temperature and outcomes were initially explored with bivariate analyses, and variables significant at a .10 level of significance were entered into multivariable linear regressions for continuous outcomes and logistic regressions for categorical outcomes. These analyses were controlled for antenatal steroids, gender, race, birth weight, intubation, Apgar at 5 minutes, and center. Results of logistic regressions were expressed using odds ratios (ORs) and 95% confidence intervals (CIs). Results of multivariable linear regressions were expressed using the parameter estimate to indicate the magnitude of independent associations.

RESULTS

Between January 1, 2002, and December 31, 2003, there were 7498 infants entered into the database. Applying the inclusion criteria resulted in the exclusion of 1649 infants, of which 1450 were excluded for 1 criteria only (749 outborn, 462 not directly admitted to an NICU, 207 with an anomaly, and 32 out of the weight range). Applying the admission temperature criteria to the remaining 5849 infants resulted in the exclusion of 570 infants. Of the latter infants, 264 were considered viable (given delivery room interventions, ventilator support, intravenous fluids, etc), and 306 were nonviable (no care provided). All but 59 of the 570 infants excluded for the temperature requirement had missing temperature or time of temperature. Two additional infants fulfilled the temperature requirement but were nonviable and were excluded. The study cohort was composed of 5277 infants.

Descriptive characteristics of selected maternal and intrapartum variables, infant characteristics, and variables from the delivery room are listed in Table 1. Infants excluded for not meeting the admission temperature criteria (n = 570) were of a lower birth weight (880 ± 318 and 545 ± 151 g, mean ± SD for viable and nonviable, respectively) and gestational age (26.3 ± 3.2 and 22.7 ± 2.3 weeks for viable and nonviable, respectively). Umbilical artery blood gas results were available for 48% (pH data) and 46% (base deficit data) of the cohort. The mean admission temperature was 35.9 ± 1.0°C (range: 28–39.6°C). The distribution of admission temperatures among the cohort (Fig 1) demonstrates that 46.9% of the temperatures were <36°C. In contrast, the frequency of admission temperatures ≥37.0°C was 10.8% and ≥38°C was 1.3%. The frequency of admission temperatures <35 and <36°C increased with decreasing gestational age and birth weight (Table 2). The measurement site of admission temperatures varied on the basis of center practice and was recorded from the axilla (77.6%), rectum (15.5%), and skin (7.0%); 9 patients had missing data for this item. The mean age at the admission temperature was 23 ± 14 minutes with a median value of 20 minutes (25th and 75th percentiles of 14 and 27 minutes, respectively).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

Results are presented for the distribution of admission temperatures among 5277 low birth weight infants irrespective of measurement site (axilla, rectum, or skin).

Variables present before or at birth and associated with admission temperature on bivariate analyses were multiple births, labor, use of antenatal steroids, maternal antibiotics, rupture of membranes, mode of delivery, birth weight, gestational age, center, intubation, 5-minute Apgar, site and age of temperature measurement, and center. Only the variables listed in Table 3 were significantly associated with the admission temperature on multivariable analyses. Multiple births, use of antenatal steroids, and prolonged rupture of membranes were associated with a statistically significant but small change in admission temperature, each <0.2°C compared with the absence of the variable. A similar change in temperature was associated with age of temperature measurement. In contrast, birth weight and intubation, center of birth, and the Apgar score at 5 minutes were associated with the largest change in admission temperature (Table 3). There was a significant interaction between birth weight and intubation in the delivery room. The admission temperature was 0.04°C higher with each 100-g increase in birth weight; however, for infants requiring intubation in the delivery room, the admission temperature was 0.13°C higher with each 100-g increase in birth weight. The admission temperature was 0.05°C higher for each point increase in the Apgar score at 5 minutes. The site of temperature measurement was associated with the admission temperature in that rectal and axilla temperatures were 0.40 and 0.22°C higher than skin temperature, respectively. Finally, there was a prominent association between center of birth and the admission temperature. The average admission temperature of each of the 14 centers ranged from 0.3°C above to as much as 1.5°C below the average admission temperature of the reference center. The variability in the distribution of admission temperatures among the 15 centers is plotted in Fig 2.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2

Admission temperatures (axilla, rectum, or skin) of each center are plotted as the percentage of measurements in the following temperature strata: black, <35°C; gray, 35 to 35.9°C; white, 36 to 36.9°C; and striated white, ≥37°C. Center 10 is the reference center on the basis of the highest percentage of temperatures between 36 and 36.9°C (74%).

The frequencies of selected neonatal morbidities for this cohort were 6.3% for NEC, 10.3% for IVH grades III and IV, 23.3% for late-onset sepsis, and 10 ± 18 days of conventional ventilation. In-hospital mortality was 12.2% with 45.2% of the deaths occurring at <7 days of age. Major categories of assigned causes of death are listed in Table 4. In multivariable analyses, there was no association between the admission temperature and NEC (OR: 1.0; CI: 0.90–1.16), grade III/IV IVH (OR: 0.96; CI: 0.86–1.07), or duration of conventional ventilation (0.4 days per 1°C decrease in the admission temperature; P = .1). In contrast, for every 1°C decrease in admission temperature, the odds of late-onset sepsis were increased by 11% (OR: 1.11; CI: 1.02–1.20), and the odds of dying were increased by 28% (OR: 1.28; CI: 1.16–1.41).

DISCUSSION

This report documents the temperature on admission of a recent large cohort of very low birth weight infants born within 15 academic centers and transferred directly to the NICU from the labor and delivery department. The principal findings of this study are as follows: (a) low temperatures on admission are common; (b) there are important associations between the admission temperature and variables antecedent to admission that may be amenable to change; and (c) there is a prominent association between the extent of reduced temperature on admission and both late-onset sepsis and in-hospital mortality.

Efforts to limit heat loss are important initial steps in the stabilization of newborns immediately after birth and are incorporated in the Neonatal Resuscitation Program¹² and the World Health Organization's guide to thermal control of the newborn.¹³ Minimizing heat loss in low birth weight and premature infants is difficult because of high evaporative heat loss exacerbated by a large temperature gradient from the skin to the ambient air and physical characteristics of the premature infant (increased surface area/weight ratio, immature epidermal barrier, limited vernix caseosa, and subcutaneous fat).^14–18 There are relatively few reports on the frequency of low temperatures at birth among premature infants. Among hospitals in third-world nations, the frequency of temperatures <36°C at 2 hours after birth can be as high as 60% for cohorts that include both term and low birth weight infants, and these observations are linked to a high incidence of hypothermia at 24 hours of age.^9,19,20 Even in developed countries, the frequency of admission temperature <35°C for all infants born before 26 weeks' gestation within the United Kingdom and the Republic of Ireland during 1995 was 40% (EPICure Study).⁸ The cohort of infants in the current report represents a broader gestational age (21–42 weeks on the basis of obstetric criteria; mean of 28 weeks) defined by birth weight (<1499 g) and may account for the lower frequency of admission temperatures <35°C (14.3%). However, in this cohort, ∼47% of the admission temperatures were <36°C. An important limitation of this study is the observational design without a standard practice regarding site, time, device, and technique used for temperature measurements. In addition, no data are available on the maternal temperature at the time of or immediately proximate to delivery, the temperature of the delivery room, or the qualifications of the pediatric providers in attendance.

Interventions to minimize the extent of heat loss have been studied in small groups of infants. The most effective seems to be occlusive wraps for which there has been interest over the past 30 years.²¹ More recently, 3 randomized clinical trials demonstrated that the use of polyethylene wraps or polyurethane bags compared with drying in the delivery room prevented heat loss and better maintained rectal admission temperatures for infants <29 weeks' gestation.^22–24 Additional measures include the use of caps, which have been demonstrated to reduce the exchange of heat between the head and the ambient air.^25,26 The use of caps conveniently complements the application of occlusive wraps where heat loss from the exposed head is still a concern. Important questions regarding the use of occlusive wraps are whether there are low-frequency adverse effects on the skin, alterations in skin flora, or potential overheating of the body. Systematic reviews and formal meta-analysis suggest that elevated temperatures recorded on NICU admission among wrapped infants may reflect factors such as maternal temperature and infection rather than occlusive wraps and indicate the need for additional studies.^27,28 Information regarding the use of occlusive wraps or other means to reduce heat loss from the current cohort was not collected as part of this study. Given the frequency of low admission temperatures among infants in this report, surveillance of temperatures in the delivery room and on admission would seem to be an appropriate, worthwhile, quality improvement initiative.

Variables antecedent to and independently associated with prominent changes in admission temperature were birth weight and intubation, Apgar score at 5 minutes, and center of birth. The association with birth weight was expected, because the physical characteristics of low birth weight infants predispose to a mismatch between heat production and heat exchange with the ambient environment when high-risk infants are stabilized at birth.¹⁴ This is consistent with the EPICure Study, where an admission temperature <35°C occurred in 30%, 43%, and 58% of infants of 25, 24, and 23 weeks' gestation (median birth weights of 760, 680, and 600 g, respectively).⁸ Multivariate analysis in the present study indicates that the effect of birth weight alone is relatively small, with an average of a 0.4°C difference in admission temperature between infants with birth weights of 401 and 1499 g. In contrast, if intubation is performed, the effect of birth weight is an average of 1.4°C different between infants of the same 2 birth weights. The association of the Apgar score at 5 minutes may parallel the birth weight-intubation interaction, because lower Apgar scores may represent a proxy for infants having more resuscitative measures (of which the most common procedure is intubation) or the response to such interventions. Center of birth was associated with a prominent change in admission temperature relative to the reference center. Five of the 14 centers had an average admission temperature >0.75°C lower than the reference center, and 3 centers were >1°C lower. Delineation of the specific practices for maintenance of temperature in the delivery room at each center was not part of this study, but the results suggest that a quality improvement approach using benchmark initiatives may be helpful for some centers.

The admission temperature was not associated with NEC, severe IVH, or duration of conventional ventilation. Associations were present between admission temperature and both late-onset sepsis and in-hospital mortality. Thermal management has been labeled a cornerstone of neonatology.²⁹ The latter is based on the pioneering work of Silverman et al³⁰ that maintenance of body temperature through control of the thermal environment during the first 5 days of life (isolette temperature of 29 vs 32°C with resultant axillary temperatures of 31.1 vs 33.7°C, respectively) reduced mortality in low birth weight infants. Other clinical trials of low birth weight infants yielded similar observations.^31–33 These therapeutic trials outlined the effects of a thermal management scheme on mortality rather than an association between admission temperature and outcome.

Additional temperatures beyond admission to the NICU were not collected on infants in the present cohort. Whether prevention of low temperatures at birth decreases mortality or whether the low admission temperature is part of the casual path or simply a marker for an increase in the odds of mortality cannot be determined from this observational analysis. Previous investigations that have reported associations between admission temperature and mortality have insufficient sample size,²² were not adjusted for covariates,³⁴ and were not reproducible.²³ The association in this report between admission temperature and late-onset sepsis provides a potential path to link the admission temperature and mortality. In adults, a self-limited interval of perioperative hypothermia may promote postoperative infections via temperature-mediated impaired immune function³⁵; perioperative normothermia decreased the postoperative infectious complication.³⁶ Whether late-onset sepsis remote from birth is causally linked to admission temperature is unknown. In addition, the causes of death (Table 4) seem to reflect expected complications of prematurity.

The results of this observational cohort demonstrate that minimizing the extent of temperature reduction at birth for the low birth weight and premature infant remains challenging. The birth weight-intubation interaction with the admission temperature and the variability among the various participating centers suggest that thermal control for newborns requiring respiratory support at birth requires a reassessment of practice. The time, effort, and resources to determine whether avoidance of temperature reductions at birth reduces mortality seem to be well justified in view of a potential casual path via late-onset sepsis.