A retrospective review of vital signs and clinical outcomes of febrile infants younger than 3 months old presenting to the emergency department

Objectives Febrile infants younger than 3 months old present a diagnostic dilemma to the emergency physician. We aim to describe a large population of febrile infants less than 3 months old presenting to a pediatric emergency department (ED) and to assess the performance of current heart rate guidelines in the prediction of serious infections (SI). Materials and methods We performed a retrospective review of febrile infants younger than 3 months old, between March 2015 and Feb 2016, in a large tertiary pediatric ED. We documented the primary outcome of SI for each infant, as well as the clinical findings, vital signs, and Severity Index Score (SIS). We assessed the performance of the Paediatric Canadian Triage and Acuity Scale (PaedCTAS), Advanced Pediatric Life Support (APLS) guidelines and Fleming normal reference values, using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and area under receiver operating characteristics curve (AUC). Results 1057 infants were analyzed, with 326 (30.6%) infants diagnosed with SI. High temperature, tachycardia, and low SIS score were significantly associated with SI. Item analysis showed that the SIS performance was driven by the presence of mottling (p = 0.003) and high temperature (p<0.001). The APLS guideline had the highest sensitivity (66.0%, 95% CI 60.5–71.1%), NPV (73.3%, 95% CI 69.7–76.5%) and AUC (0.538), while the PaedCTAS (2 standard deviation from normal) had the highest specificity (98.5%, 95% CI 97.3–99.3%) and PPV (55.2%, 95% CI 32.7–71.0%). Conclusions Current guidelines on infantile heart rates have a variable performance. In our study, the APLS heart rate guidelines performed with the highest sensitivity, but no individual guideline predicted for SIs satisfactorily.


Introduction
Febrile young infants younger than 3 months old present a diagnostic dilemma to the pediatric emergency department (ED) physician. The potential for a missed serious infection (SI) poses the threat of premature death and long-term disability among these infants. [1,2] Despite decreasing early-onset neonatal sepsis rates due to obstetric prevention strategies [3], high rates of hospitalization and administration of parenteral antibiotics occur in this age group. Continual tension remains between the need for early and aggressive intervention among patients suspected with sepsis [4] and the global phenomena of increasing antibiotic resistance. [5,6] Research networks have attempted to build diagnostic algorithms to guide the identification of these ill infants. [7,8] These are often useful as adjuncts to the clinician's gestalt, but generalizability remains questionable.
Vital signs are of paramount importance in recognizing ill children and have been used in pediatric early warning system scores (PEWS) [9] and various triage systems [10]. Vital signs have resurfaced as the focus of research in recent years, with various groups purposing to update evidence-based normal heart rate ranges among children. [11][12][13][14][15][16] Normative heart rate ranges are infamously difficult to define due to the hemodynamic lability in these young infants, multiple confounders for abnormal heart rate, and the variable physiological response during acute stress states.
In this study, we aim to (1) describe the presentation and outcomes of a large population of febrile young infants less than 3 months old presenting to a pediatric ED in an Asian city, and (2) assess the performance of the Paediatric Canadian Triage and Acuity Scale (PaedCTAS) [10], the Advanced Pediatric Life Support (APLS) guidelines [11] and the Fleming normal reference values [12], in the prediction of SIs.

Study design and setting
This is a retrospective chart review from 1st March 2015 -29th Feb 2016 in KK Women's and Children's Hospital (KKH). Singapore is a small urban nation of 5.5 million people, with 16% of its population less than 15 years' old. [17] Among Singapore residents, about 194,432 are under 5 years' old.
[18] KKH is the larger of two pediatric specialty hospitals in the country and patients who attend the ED are charged a nominal fee. The annual ED attendance in KKH is about 174,000 children.

Participants
We screened all children less than 3 months old with an initial triage axillary temperature of > 37.5˚C. Temperature is routinely measured at the triage by a nurse, using an axillary thermometer (Terumo1 digital clinical thermometer), for infants < 6 months. Subsequently, if the repeat temperature was less than 38˚C (e.g. in cases of overwrapping) and the child remained clinically well with no subsequent investigations performed, this child was excluded from the analysis. Young infants admitted for jaundice or poor feeding but who did not mount a fever response were excluded.

Variables
Triage. Our triage system utilizes the Severity Index Score (SIS) [19] which is interpreted in the following categories: SIS 10 (Not very sick), SIS 8 or 9 (moderately sick), SIS 7 or less (Very sick). The SIS comprises the domains of respiratory effort, color, activity, temperature and play. Heart rate, pulse oximetry and blood pressure (BP) are automated measurements using the Dinamap GE ProCare 300 Vital Signs Monitor, while respiratory rate is manually measured by the triage nurse. BP is measured using the neonatal or infant cuff, as appropriate. On occasions where the neonate appeared well but repeated attempts to obtain the BP failed due to movement artifacts, the triage was based on heart rate and respiratory rate. Our ED triage uses the PaedCTAS guideline for normal heart rate and respiratory rate ranges. [10] For this study, the SIS and vital signs were extracted from the electronic health record.
History and physical examination. We included the following perinatal details: gestation and maternal group B streptococcus (GBS) status. Physical examination findings of respiratory distress, lethargy, and capillary refill time were collected.
Investigations and management. Biochemical results included total white blood cell count, absolute neutrophil count, hemoglobin, platelets, C-reactive protein, urinalysis, and cerebrospinal fluid (CSF) analysis. Microbiological culture results from urine, blood, CSF, stool and other sites (abscess, wounds), as well as nasopharyngeal aspirate for common viruses were recorded. We documented if the infant received intravenous antibiotics, fluid boluses, inotropes, ventilator support, required intensive care or if they died.
Ethics approval for this study was granted by the Singapore Singhealth Institutional Review Board, with waiver of informed consent.

Statistical methods
Categorical data were summarized by frequencies (and percentages) and continuous data by means (and SD) or medians (and interquartile ranges (IQRs)). Where data approached normality, we presented the data with means (and SDs) based on the central limit theorem. Fisher's exact test or chi square test was performed to assess for association between SI and categorical predictors of interest, while t-test or Wilcoxon rank sum was used for continuous variables, depending on normality of data. We presented the performance of the PaedCTAS [10], APLS guidelines [11] and Fleming normal reference values [12] using sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). We also described the performance of heart rate and each guideline using the area under receiver operating characteristics (ROC) curve (AUC). Significance was taken at a p value of less than 0.05. The analysis was performed using IBM SPSS Statistics v23.0.

Results
Out of 2093 young infants < 3 months who presented with an initial axillary temperature of > 37.5˚C, 99 infants (4.7%) were triaged as most urgent and were required to be seen immediately, while 1161 (55.5%) were triaged as urgent and required to be seen within 15 minutes of triage. Vital signs, triage status and ED disposition are described in Table 1. Among the 2093 infants, 1546 were hospitalized and 1057 were analyzed for the outcome of SIs. (Fig 1) Of these, 247 infants (23.4%) were younger than 1 month old and 326 infants (30.8%) were diagnosed with SI. (S1 Dataset) In the univariate analysis (Table 2), the triage parameters of temperature (p<0.001), heart rate (p = 0.006), SIS score category (p = 0.005) and triage category (p = 0.019) had a statistically significant relationship with the outcome of SI. Item analysis showed that the SIS performance was driven by color interpreted as the presence of mottling (p = 0.003), and the temperature of the child (p<0.001). (Table 3) Table 1

. Demographics and initial presentation of all infants (n = 2093).
Age in months, mean (SD Heart rate per minute, mean (SD) 159 (21) Respiratory Rate per minute, mean (SD) 42 (7) Pulse oximetry %, mean (SD) 98. 6  Investigations done showed that the total white blood cell count, absolute neutrophil count, platelets, and C-reactive protein were statistically significant in predicting for SI in our population. Only 19 (5.8%) infants with SI had a positive nasopharyngeal aspirate for common viruses, compared to 139 (19.0%) infants with no SI (p<0.001), with an overall positivity rate of 14.9%. A significantly larger proportion of infants with SI received intravenous antibiotics and had a longer duration of hospitalization, compared to infants without SI. (Table 2) Among the SIs, urinary tract infections were the most common, with E Coli and Klebsiella as the most likely pathogens. (Table 4) This was followed by meningitis and septicemia. The breakdown between viral and bacterial causes of meningitis (Table 5) showed that there was no statistically significant difference in vital signs or blood investigation results. Table 6 shows the performance of the PaedCTAS [10], APLS guidelines [11] and the Fleming normal reference values [12] in our study population. The APLS and Fleming (<10 th or >90 th centile) performed with the highest sensitivity (66.0% and 62.6%, respectively) and the highest NPV (73.3% and 71.4%, respectively). No single guideline reached a sensitivity of greater than 70%. Guidelines with more stringent criteria that define tachycardia (PaedCTAS with 1 or 2SD from normal range, and Fleming {<1 st or >99 th centile) performed with better specificity (78.1%, 98.5% and 73.9%, respectively). The AUC for heart rate was 0.548. The

Discussion
In this study, we described the triage, diagnostic parameters and clinical outcomes of a large Asian cohort younger than 3 months' old presenting to a pediatric ED with fever. We demonstrated that a low SIS, together with abnormal heart rate and high temperature, significantly predicted for SI. We also showed the performance of current published heart rate normal ranges in our population.
Despite heart rate being a universally recognized triage tool, heart rate normal ranges vary between guidelines [10][11][12] and remain difficult to validate in different pediatric populations. In a healthy cohort of Chinese children in Hong Kong, investigators found that a significant number of children had vital signs that fell outside the APLS age-based reference ranges. [21] In our study on infants younger than 3 months old, when compared to other heart rate guidelines and reference values, the APLS [11] guideline appeared to perform with the highest sensitivity and AUC. However, none of the guidelines reached satisfactory performance to be used singly as a predictive tool.
There was significant overlap in the heart rate distributions between infants with and without SIs, suggesting that heart rate alone is unlikely be discriminatory between the two groups. This was highlighted by the AUC of 0.548 for heart rate. We do recognize that single read-outs of a child's heart rate are easily confounded and may be inferior to an analysis of heart rate variability. [17] Research in neonatal intensive care units report that heart rate characteristic monitoring is useful in reducing mortality rate among very low birth weight infants. [22,23] These results have not been demonstrated in the ED setting, where heart rate variability may have an increasing role to play in assisting the ED physician's assessment of ill infants.
Clinical decision-making is often complemented by judicious use of biomarkers, including white blood cell count, absolute neutrophil count, C-reactive protein and interleukin-6. [24,25] More recent work has focused on the association of RNA biosignatures with bacterial infections in young infants 60 days or younger. [26] In our study, we showed that a high total white blood cell count, absolute neutrophil count and C-reactive protein significantly predicted for SIs. We do not perform tests on interleukins or RNA markers among febrile infants in our institution.
In this study, we addressed the important question of how current heart rate guidelines perform when predicting which febrile young infant has a serious infection. We highlight that despite robust attempts at quantifying normal heart rate ranges [10][11][12], the utility of these ranges in providing guidance to ED physicians who manage febrile young infants at risk of serious infections remains limited. ED physicians continue to rely on a constellation of vital signs, symptoms and signs to make clinical decisions, while awaiting a vital signs tool that can provide greater discriminatory power.
We recognize the limitations in our study. Firstly, being a retrospective study, the reviewers of health records were at risk for misclassification bias and were not blinded to the aims of the study. This however, would not have affected objective measurements like heart rate, temperature, or investigation results. Secondly, respiratory rate was manually measured at triage. Inaccuracies in this single measurement could have resulted in the non-significant association between respiratory rate and SIs in our study population. Thirdly, in choosing the primary outcomes of SIs, we included viral meningitis because of concerns of the long-term sequelae of such infections. We did demonstrate that there was no statistically significant difference in vital signs or blood investigation results between the viral and bacterial meningitis groups in our study population. Fourthly, in our study, we examined original heart rates without adjustment. This was to reflect the need for rapid dichotomous decisions at the ED setting in this young population. Finally, this is a single center study and our findings require validation in a new population.
In conclusion, we described a large population of febrile infants younger than 3 months' old presenting to the ED. In our population, the APLS heart rate guidelines performed with the highest sensitivity, but no individual guideline predicted for SIs satisfactorily.