Variation in Care of Well-Appearing Hypothermic Young Infants: A Multisite Study

We conducted a retrospective multisite study of well-appearing infants ≤90 days of age who were evaluated in the ED or hospital for hypothermia from September 1, 2016 to May 5, 2021. A total of 9 sites were included in this study; all sites are academic children’s hospitals, have primarily pediatric emergency medicine providers in the ED, and with estimated annual ED volumes ranging from 17 000 to 95 000 (Supplemental Table 4). Among all sites, only 1 has a clinical pathway in place for infants presenting with hypothermia, established in 2020.

Eligible patients were identified via 2 methods: ICD 9 and 10 codes and through manual chart review of an electronic health record (EHR) query. Similar to previous studies, patients with hypothermia were identified by the following ICD-10 codes (R68, P80, P808, P818, P819, T68, T69.8XXA) and ICD-9 codes (778.3, 778.4, 991.6, 780.65).^2,3,7  An additional search was performed to identify eligible patients through the EHR by searching for all patients ≤90 days old seen in the ED or admitted to the hospital with a documented temperature ≤36.0°C within 24 hours of the initial patient encounter. The temperature documented on the initial set of vitals was used to determine eligibility. The hypothermia cutoff of 36.0°C was selected as this definition is supported by groups such as the American College of Surgeons Advanced Trauma Life Support and International Pediatric Sepsis Consensus Conference.^2,6,8,9 

Inclusion criteria included young infants ≤90 days of age with historical or documented hypothermia. Historical hypothermia was defined as a low temperature recorded at home by a care provider. Documented hypothermia was defined as hypothermia found at a health care facility (primary care physician office, or on initial vital signs at ED or hospital). Hospital transfers were included except those with inaccessible culture results. Infants identified as ill-appearing were excluded, given any noted variation in care was more likely clinically indicated. Ill or toxic appearance was defined as the following terms documented on physical examination: ill appearance, toxic, limp, unresponsive, gray, cyanotic, apnea, weak cry, poorly perfused, grunting, listless, lethargic, or irritable.^10,11  Additional exclusion criteria included infants with known medical complexity (defined as a severe medical condition expected to last ≥12 months and requiring subspecialty care or involving multiple organ systems¹² ), infants with a central line, known trauma or cardiac arrest on presentation, fever, and/or hypothermia in the NICU or nursery during birth hospitalization.

Patient charts were manually reviewed for data extraction at each site. Site investigators extracted the following deidentified data from the EHR into a central Research Electronic Data Capture at Prisma Children’s Hospital: demographics, clinical characteristics, laboratory evaluation, imaging, treatment provided, hospital course, and outcomes. Demographic data included age at presentation, sex, and gestational age. Clinical characteristics included vital signs on presentation, presenting symptoms, and physical examination findings.

SBI was defined as bacteremia, bacterial meningitis, and/or urinary tract infection (UTI). Meningitis was defined as the presence of a bacterial organism from a cerebrospinal fluid (CSF) culture or polymerase chain reaction test that was treated as a pathogenic organism with antibiotics. Similarly, bacteremia was defined as growth of an organism from a blood culture that was treated as a pathogenic organism with antibiotics.^13,14  UTI was defined as ≥50 000 colony forming units/mL of a pathogen from a catheterized sample obtained with an abnormal urinalysis (presence of leukocyte esterase, nitrite, or >5 white blood cells per high power field), or a positive urine culture treated as a pathogenic organism with antibiotics.¹⁵  Invasive bacterial infection (IBI) was defined as bacteremia and/or bacterial meningitis. Herpes simplex virus (HSV) disease in young infants was defined as HSV detected by polymerase chain reaction from surface sources, blood, or CSF specimen and treated with acyclovir.^16,17  Partial SBI workup was defined as obtaining 1 to 2 sites of cultures, whereas a complete SBI evaluation was defined as obtaining 3 sites of cultures (urine, blood, and CSF).

Diagnostic evaluation consisted of laboratory studies and/or imaging obtained within 24 hours of the start of the patient encounter. Patient management consisted of patient disposition, empirical antimicrobial therapies, and LOS. Patient disposition was admission to the hospital or discharge from the ED. Empirical antimicrobial agents were the administration of antibiotics and/or acyclovir during the initial patient encounter. LOS was calculated as the number of hours from hospital admission (time admission order placed) to hospital discharge (time discharge order placed). Hospital readmissions were reviewed for any readmission within 30 days of the initial patient encounter with specific review for any potentially missed SBI or HSV from initial encounter.

Descriptive statistics were used to characterize the demographics, clinical characteristics, and outcomes of the cohort. For the comparison by site, a P value < .05 was considered statistically significant. Continuous variables were analyzed using ANOVA or Kruskal-Wallis test, whereas discrete variables were analyzed using χ-square and Fisher’s exact tests. All analyses are conducted using R statistical software (version 4.0.4).¹⁸ 

Of the 14 278 young infants initially identified, we included 739 infants in our final cohort (Fig 1). We removed 300 and 59 patients because of ill appearance and medical complexity, respectively. Nearly half (48.8%, n = 361) of included patients were not identified by ICD-9 or ICD-10 codes. The majority of patients in our cohort were ≤7 days old (57.2%, n = 423) and born at >37 weeks’ gestation (59.1%, n = 437). Only 2% (n = 18) of patients in this cohort were transferred from the referring facility. Known maternal HSV was documented in 5.9% (n = 44) of patients; 66.4% (n = 491) had unknown maternal HSV exposure. A total of 67.9% (n = 502) presented with hypothermia on their initial vital signs and 48% (n = 355) had repeated episodes of temperature instability. Overall, 23 (3.1%) infants had a SBI and 4 (0.5%) had HSV infection. Among the 23 infants with SBI, 10 had IBI, whereas 13 infants had only UTI. One infant with IBI had bacterial meningitis, bacteremia, and UTI all from the same organism, whereas another infant with IBI had bacterial meningitis and bacteremia from the same organism. Additional patient demographics are included in Table 1.

Overall, 496 (67.1%) well-appearing infants had a full or partial SBI evaluation. Across sites, 12% to 76% of infants had a full SBI evaluation and 11% to 36% of infants had a partial SBI evaluation. No SBI evaluation was done in 243 (32.9%) infants, with differences across sites ranging from 7% to 69% of infants having no cultures obtained on presentation. HSV testing was completed in 253 (34.2%) infants with differences across sites ranging from 7% to 84%. Complete blood count was obtained in 515 (69.7%) infants, C-reactive protein in 296 (40.1%) infants, and procalcitonin in 145 infants (19.6%). Respiratory pathogen panels were obtained in 216 (29.2%) infants with differences across sites ranging between 12% to 54%. Table 2 provides a full list of the diagnostic evaluations.

Empirical antibiotics were administered in 411 (55.6%) infants with differences across sites ranging from 20% to 87% (Table 3). Acyclovir was started on 279 (37.8%) patients with variation between sites of 8% to 82%. The majority of infants (n = 614, 83.1%) were evaluated in the ED. A total of 547 (74.0%) infants were admitted to the hospital with ranges between 45% to 100% between sites. Median hospital LOS was 56 hours (interquartile range 43–101). Between sites, 0% to 7% of infants were readmitted within 30 days of the initial encounter for hypothermia.

Among all infants evaluated for hypothermia, none were readmitted for bacterial meningitis, bacteremia, or HSV. Three infants presented within 30 days from the initial patient encounter for hypothermia and were treated for UTIs. One of these patients presented 11 days after the initial encounter and had a negative blood and urine culture during the first evaluation. The second patient had no cultures performed on the initial evaluation and 18 days later was treated for a UTI while presenting with bronchiolitis. The third patient had blood and urine obtained and was initially observed in the ED, then returned the same day after urine culture was positive. One patient died of dilated cardiomyopathy.

To our knowledge, this is the first study to describe variation in care for well-appearing hypothermic young infants accurately identified by EHR review. We observed significant variability across hospitals in diagnostic evaluation and patient management with no difference in hospital readmissions or missed invasive bacterial infections and infant HSV. Our study addressed limitations inherent to the existing literature by incorporating EHR review to verify hypothermia (rather than relying solely on billing codes for cohort identification) and to allow exclusion of ill-appearing infants, for whom variation in care is expected and appropriate.^3,7 

By reviewing clinical characteristics, we were able to focus specifically on the evaluation and management of well-appearing hypothermic infants. A recent study using Pediatric Health Information System data found significant variation in SBI evaluation, inflammatory marker testing, and use of empirical antimicrobial agents; however, this work was limited by reliance on diagnosis codes and lacked access to clinical characteristics.⁷  Hypothermic infants who are ill-appearing likely receive different evaluation and management from well-appearing infants. Additionally, the American Academy of Pediatrics febrile infant guidelines only focus on well-appearing infants presenting with fever.¹⁹  We also found markers of inflammation were inconsistently used in well-appearing hypothermic infants. Similar to febrile young infants, understanding care provided to well-appearing hypothermic infants is a necessary step to measuring variability in care and informing the development of clinical practice guidelines.^13,19 

Our study showed the majority (57.2%) of well-appearing hypothermic infants were ≤7 days old and 27.0% were born between 34 0/7 and 36 6/7 weeks’ gestation, both representing patient populations not included in the American Academy of Pediatrics clinical practice guideline for febrile infants.¹⁹  Temperature instability is a main cause for readmission in late preterm infants when compared with term infants.²⁰  These patient characteristics should be taken into consideration for risk prediction modeling and guideline development.

Marked hospital-level variability in diagnostic and management approaches was noted for all primary outcomes, despite narrowing our focus to only well-appearing patients without known medical complexity. Although significant research efforts have been made for standardizing care for febrile infants, data for hypothermic infants are limited and no universally accepted clinical practice guidelines exist for managing these patients. The level of variability noted in our study may reflect the lack of evidence available in this population, as clinical care tends to be more varied in areas where less definitive evidence exists.²¹  Our study was not powered to determine if the variation noted could be attributed to the provider training level or other historical elements about the infant such as gestational age, but these are also possible contributing factors.

Hypothermia has been identified as a risk factor for HSV.^16,17  We observed significant variation in HSV testing and empirical use of acyclovir, which has also been demonstrated in febrile infants.²²  Despite the wide variability in HSV testing in our study (7% to 84%), there were no repeat ED visits or readmissions for HSV within 1 month of initial presentation.

Nearly half of the patients identified on EHR query did not have a hypothermia-related diagnosis code. Our findings support the recommendations by Lo et al⁷  to not rely on ICD-9 and 10 diagnosis codes for the identification of hypothermic infants. Potential reasons for the inaccuracy of billing codes in hypothermic infants include the various conditions that can present with hypothermia²³  and the ambiguous diagnosis wording, such as “disturbance of temperature regulation,” which could be used to refer to both a low or elevated temperature on presentation. Similar to febrile infants, identifying diagnosis codes that correlate specifically with hypothermia inyoung infants is an important step when unable to access clinical characteristics.^24,25 

Amid rising concerns about the quality and cost of health care delivered in the United States, identifying variation in care is the first step toward change in clinical practice. Strategies to understand and address the modifiable sources of variation are needed and may reflect the lack of evidence on the optimal evaluation and management of hypothermic infants. Infants with SBI can present with hypothermia, and better identifying at-risk infants can improve resource utilization in this population. A value-driven care model can help to decrease unnecessary testing, antimicrobial exposure, and even LOS to limit patient harm and costs. One strategy to address variation in care is to develop guidance for best practices in caring for and risk-stratifying these infants. In our study, most hospitals did not have a clinical pathway in place to help guide physicians managing infants presenting with hypothermia. Among other pediatric illnesses, developing clinical care guidelines has been shown to safely reduce unnecessary utilization,^26–29  and therefore remains an important next step for this vulnerable population.

Our study has some important limitations. Despite the largest number of well-appearing infants included in a study, there were still significant differences in age at presentation and gestational age between sites, which may contribute to the variation in diagnostic evaluation and patient management. Our cohort size did not permit meaningful subgroup analyses by age or gestational age, so larger studies of well-appearing hypothermic infants are necessary to see if variability in care persists when patient age on presentation and gestational age are similar. Additionally, the exact temperature cutoff to identify hypothermia is inconsistently defined^9,30  and recent attempts to determine a reliable threshold were unsuccessful.⁸  As a result, we excluded infants between 36.1°C and 36.4°C. Our search strategy may not have reliably captured infants with hypothermia reported at home if that infant did not subsequently have hypothermia in the hospital and if no ICD code for hypothermia was applied to that patient. Although we had access to full EHR data for each infant, there may have been characteristics that were not well-documented during the encounter that influenced the care provided, and there may be subjective differences in the definition of ill-appearance between clinicians and sites. Finally, though our study sites were geographically diverse, all were tertiary academic sites; therefore, our results may not be generalizable to smaller, community-based hospitals.

Our findings highlight the substantial variation in care provided for well-appearing young infants presenting with hypothermia without differences in patient outcomes. An improved understanding of hypothermic young infants and their risk of infection may lead to the development of clinical decision tools to guide appropriate and value-driven evaluation and management.

We wish to acknowledge the efforts of the additional members of our Hypothermic Young Infant Research Collaborative, including Melissa Burns, Amber Domako, Meghan Gray, Noah Hellerman, Elizabeth Rinaldi, Ali Sawani, Emma Schwendeman, and Sanford Williams.