Pediatric fever is a common complaint in children. The most common cause is self-limited viral infection. However, neonates and young infants are evaluated and treated differently than older, vaccinated, and clinically evaluable children. Neonates should be admitted to the hospital, young infants in the second month of life may be risk stratified, and those deemed low risk on testing may be sent home with close follow-up. Children older than 2 months may be evaluated clinically for signs of bacterial infection that require intervention. Urinary tract infections cause more than 90% of serious bacterial illness in children, and younger children have a higher incidence of infection.
Fever is common in children.
Viral infections are the most common cause of pediatric fever.
Fever in the young infant deserves special mention and a nuanced approach. The clinical management of the neonate/young infant remains controversial.
Urinary tract infections are currently the most common serious bacterial illness in children.
Herpes simplex infections in neonates may be initially subtle and carry significant morbidity and mortality rates.
Fever accounts for 10% to 20% of all pediatric emergency department visits. Most febrile illnesses are self-limited viral infections. However, fever in the young infant (<3 months of age) demands special attention and nuance, as young infants are at increased risk for serious bacterial illness (SBI). Approximately 8% to 12% of young febrile infants have an SBI. Most of these infections are urinary tract infections (UTIs), but 1% to 2% of infections are invasive bacterial infections (IBI) such as bacteremia and meningitis. In addition to the higher incidence of SBI, the clinical clues are often subtle or simply not present in young infants; this is most drastic in the case of a neonate (≤28 days) in which the risk of SBI/IBI is highest ( Figs. 1 and 2 ). Delayed or missed diagnosis and treatment of infections directly affects morbidity and mortality.
Controversy remains regarding the best practice in risk stratification of the young infant. There is also great variation in testing, treatment, and overall resource utilization in the management of this patient population. The goal is to be able to accurately risk stratify this group with both sensitivity and specificity and thereby reduce invasive testing, antibiotic administration, hospitalizations, and iatrogenic sequelae. The controversy lies in acceptable sensitivity and risk with the problem of lack of specificity of risk stratification. Most young infants stratified as “high risk” will not have an SBI. But increasing specificity typically results in more missed infections. Diagnostic testing and hospitalization both carry risks to young infants, and the risk versus benefit ratio must be carefully weighed in this vulnerable population. Risk of intervention include procedural complications, antibiotic resistance, hospital-acquired infection, disruption of neonate family bonding and breastfeeding, and increase in cost. Missed SBI/IBI may lead to death or permanent disability in survivors.
The normal body temperature varies with age, time of day, level of activity, and hormonal fluctuations. There is a morning nadir and a late afternoon peak. The commonly cited “normal temperature” of 98.6°F/37°C was based on a nineteenth century study. The mean temperature in young adults was shown in a more recent study to be 37.2°C (98.9°F) in the am and 37.7°C (99.9°F) overall. Infants have higher baseline temperatures than older children. A temperature of 38°C/100.4°F is two standard deviations higher than the mean of normal and considered a fever in an infant.
In most clinical scenarios, the appearance of the child is of much greater value than a temperature number. Clinical signs of serious illness such as alteration of consciousness, irritability, lethargy, or meningismus should be concerning regardless of the presence of fever.
The height of fever does not correlate with likelihood of bacterial infection. , However, this may not be the case in young infants, and this group may be more likely to have an SBI with a fever.
The temperature elevation with an infection is an orchestrated series of events that begins with peripheral cytokine release and results in anterior hypothalamus PGE2 release and an increased set point for body temperature. An elevated temperature retards the growth and replication of both bacteria and viruses and enhances immunologic function. Some animal studies have shown an enhanced survival in those who mount a fever with infection. However, a fever does cause discomfort (peripheral PGE2 induces myalgias/arthralgias) and increases oxygen consumption and cardiopulmonary demand (which is of little consequence in most otherwise healthy children but is an important consideration with significant comorbidities). Fever also induces slow wave sleep.
Despite widespread belief by caregivers and providers alike, there is no evidence that elevated temperatures result in long-term sequelae of neurologic damage. The upper limit of temperature due to a fever is 107.8°F/42°C, but it is uncommon to exceed 106°F/41°C without an element of concomitant hyperthermia.
Method of Temperature Measurement
A precise measurement of temperature is important when it will change clinical management. In most children, the clinical appearance of the child dictates management more than a number, particularly if the child is previously well and up to date on vaccinations. Accurate measurement of temperature is important in the neonate and young infant due to the increased risk of SBI with fever. Rectal temperature is the most accurate and closest to core temperature of commonly used methods. Rectal temperatures should be avoided in neutropenia, bleeding diathesis, or necrotizing enterocolitis. Oral temperature is the preferred method in older, cooperative patients. However, this method can be affected by mouth breathing and ingestion of hot/cold liquids. Oral temperatures are typically about 0.6°C or 1°F lower than the rectal temperature. Axillary temperature is the most common method used in neutropenic patients. Axillary temperatures are also lower than rectal temperature, but the absolute difference varies and there is not a standard conversion. Infrared thermometry is commonly used in research and by anesthesiologists with precision. However, the devices used in most hospitals and homes are not yet as accurate.
Fever in Young Infants
Neonates and young infants have a higher rate of SBI than any other time in child. The risk of SBI directly correlates with age (see Figs. 1 and 2 ). Approximately 20% of febrile neonates in the first week of life will have an SBI; this decreases to approximately 12% in the fourth week of life and 8% after 1 month of life. Overall, the prevalence of bacterial meningitis is 0.6% to 0.9% in infants aged 90 days or younger. However, IBIs occur in approximately 5% in the first month of life, approximately 4% in the second month of life, and less than 1% in 60 to 90 days.
Fever may be the only presenting sign of an SBI, as symptoms may be subtle and early infections may not be clinically apparent. The management of pediatric fever has classically placed young infants in high risk (neonate, ≤28 days), moderate risk (29–59 days), and lower, but still of mention, risk (60–90 days). Although these cut-offs represent a broader risk cut-off, there is not a drastic difference between infants aged28 days and 29 days. Rather, there is a gradual improvement of immune function, an exhibition of interaction, vaccine implementation, and decreasing overall risk of SBI, as the infant ages along the continuum from birth to older than 3 months. Of additional management consideration, younger infants have increased morbidity from viral infections such as an increased risk of dehydration, and they are prone to apnea with bronchiolitis.
The prediction rules developed in the 1980s/90s (ie, Boston, Philadelphia, Rochester criteria) have high sensitivity but low specificity to identify bacterial infections. These rules were also created before widespread vaccination implementation against Haemophilus influenzae type b (Hib) and Streptococcus pneumoniae . In addition, these were published before more sensitive biomarkers (C-reactive protein [CRP] and procalcitonin; see diagnostic test section), which more accurately predict SBI and IBI. However, these decision tools are still commonly used because they are well validated.
There is currently no consensus on the best clinical predication model for wide-spread implementation to help differentiate a febrile young infant with an SBI. There is also controversy regarding ideal antibiotic coverage, given changing epidemiology of infections. When the most widely used risk stratification criteria were developed, the most common SBI was bacteremia. Now UTI has surpassed as the most common infection. Current investigation is focused on excellent sensitivity but also specificity for IBI, which would reduce unnecessary testing, antibiotic administration, and required hospitalization.
Epidemiology of Infections
Current vaccine schedules have changed the epidemiology of bacterial infections in children. Before vaccination, Hib (20%) and S pneumoniae (80%) caused essentially all IBIs. Occult bacteremia occurred in 3% to 11% of children, and 5% of patients progressed to more serious invasive disease. After widespread vaccination, the incidence of occult bacteremia is less than 1%. Drawing “routine” (ie, without specific indication) blood cultures in children outside of young infancy is more likely to result in a false positive than a true positive result. Herd immunity decreases IBI risk in unvaccinated children, but the risk remains higher than in fully immunized children.
Currently the most common causes of SBIs and IBIs are Escherichia coli , group B streptococcus (GBS), Staphylococcus aureus , and Klebsiella . Herpes simplex virus (HSV) infections are increasing in incidence (correlating with increased incidence of maternal herpes infections) and can cause devastating sequelae (see Neonatal herpes below). Methicillin-resistant S aureus (MRSA) is relatively increasing in incidence. Listeria monocytogenes is rarely identified in febrile infants with an IBI. The highest incidence of bacterial meningitis remains in children younger than 2 months, but the median age has shifted from younger than 5 years to midadulthood in the current vaccination era. Currently, S pneumoniae and Neisseria meningitidis are the most common causes of bacterial meningitis in children older than 1 month. The most common causes of sepsis in older infants/children include Streptococcus pneumoniae , N meningitidis , S aureus (including MRSA), and gram-negative enteric bacteria. The impact of the current vaccination schedule on the diagnostic test performance for less virulent bacterial infections is not entirely known.
An estimated 1500 cases of neonatal herpes occur in the United states every year. These infections are frequently underrecognized and undertreated. Herpes infections in young infants cause significant morbidity and mortality. Survivors frequently have permanent neurologic sequelae.
Infections are most common in the first 3 weeks of life (peak incidence in second week of life) but may occur in young infants outside of the neonatal period. Most infections are acquired vertically during the peripartum period but may also be acquired before or after birth. The incubation period is 5 to 21 days. The greatest transmission risk occurs in infants delivered vaginally to women with primary active genital HSV infection. Primary HSV infections have the greatest risk of infection, as viral loads are higher and maternal antibodies are lower. Prematurity is associated with increased risk of transmission and worse prognosis. Prolonged rupture of membranes (>4–6 hours) also increases transmission risk. Caesarian section decreases but does not eliminate transmission rates. In addition, maternal prophylactic/suppressive treatment is not 100% effective. However, most neonates with HSV disease are born to mothers without a history of HSV infection or identifiable risk factors. This reality has important consideration when applying risk stratification criteria to neonates.
Clinical syndromes of herpes infections
There are 3 clinical syndromes of herpes infection that occur with some overlap: Localized skin, eye, mouth (SEM), central nervous system (CNS) infection, and disseminated disease. Both HSV-1 and HSV-2 can cause each syndrome, but HSV-2 carries a worse prognosis. Rash is typically present in only half of the cases overall.
SEM infections most commonly involve the face and scalp. Fetal scalp monitoring is a common risk factor and site of infection. More than half of the SEM infections can progress to CNS or disseminated infections if not treated. SEM may be subtle, particularly with conjunctival involvement, and may present with mild redness or tearing. Keratitis and chorioretinitis may occur with permanent visual sequelae.
CNS infection accounts for about one-third of neonatal HSV and typically presents in the second to third week of life. Clinical manifestations may be initially subtle, but seizures and lethargy are characteristic features. Cerebrospinal fluid (CSF) studies and imaging may be normal early in the disease course.
Disseminated disease typically presents within the first week of life and has the highest mortality rate. Approximately one-fourth to one-third present with disseminated disease, which manifests similarly to sepsis and frequently involves multiple organs (ie, transaminitis, hemorrhagic pneumonitis, meningoencephalitis, myocarditis/myocardial dysfunction, coagulation defects, bone marrow abnormalities, and acute kidney injury). Lethargy and significant transaminase elevation portend a poor prognosis. Hypothermia is often more common than hyperthermia in disseminated disease. Greater than 90% of neonates with disseminated HSV disease have HSV DNA in their CSF.
In SEM disease, swabs of the conjunctivae, mouth, nasopharynx, and rectum may be obtained. Viral culture is the best for detection of HSV, but HSV polymerase chain reaction (PCR) assays are acceptable alternatives. A lumbar puncture should be performed on all neonates with suspected HSV (even if only SEM is apparent). CSF PCR is more sensitive than viral culture (75%–100% sensitive and 71%–100% specific) but blood or high protein may produce false-negative results. Body fluid aspirates (tracheal, amniotic fluid, peritoneal drains) may show HSV DNA. Blood or plasma may also be tested for HSV PCR. In one study, plasma HSV PCR was positive in 78% of neonates with SEM, 64% with CNS disease, and 100% with disseminated disease. Higher levels of copies/mL were also associated with increased risk of death.
Brain MRI becomes abnormal days to weeks into the illness demonstrating parenchymal edema, hemorrhage, or destructive lesions. The temporal lobe is most commonly affected, but multifocal involvement (including brainstem or cerebellum) may occur.
Acyclovir is the antiviral agent of choice for all cases of neonatal herpes simplex virus infections. Administration improves survival in all manifestation of infections. Treatment improves outcomes with disseminated disease but does not impact the neurologic outcome of CNS disease. Dosing is 60 mg/kg/d divided over 8 hours. SEM is treated for a minimum of 14 days and disseminated and CNS infections for a minimal of 21 days (and until CSF HSV PCR is negative). Long-term suppression with oral acyclovir is often given to reduce risk of CNS recurrence.
Utility of Diagnostic Studies
Risk stratification criteria have previously used white blood cell (WBC) count, absolute neutrophil count (ANC), and band percentage to help risk stratify infants at risk for SBI. Each of these are imperfect in both sensitivity and specificity. Newer tests have better utility but have pitfalls preventing use as a stand-alone test. Blood cultures, although the gold standard for demonstrating bacteremia, are not immediately available. Now that they are continuously monitored, results are available at a median time of 13 hours, which remains significant diagnostic delay. Gram stain is also insensitive for detecting IBI.
CRP is an acute-phase reactant synthesized in the liver within 4 to 6 hours after tissue injury, which peaks at 36 hours. Although it is more specific than WBC count for the detection of bacterial infection, it is not sensitive and cannot be used as a sole decision-making data point in risk stratification.
Procalcitonin (PCT) is released by the liver and mononuclear cells 4 hours after tissue injury. It peaks 6 hours after tissue injury with a sustained peak level for 8 to 24 hours. The benefit of procalcitonin is that it increases more rapidly in bacterial infections compared with CRP and other biomarkers. PCT is both more sensitive and specific for SBI and IBI compared with WBC, ANC, and absolute band count. However, it is not accurate in fever less than 2 hours duration.
The biggest challenge in PCT implementation is risk stratification, as many studies use different thresholds to define positivity. The most sensitive cut-off used is 0.13 ng/mL but this also has low specificity. The most consistently used cut-off has been 0.5 ng/mL. This cut-off has a high specificity but inadequate sensitivity for SBI. PCT performs better for detection of IBI (meningitis, bacteremia/sepsis). Trippella and colleagues reported that at a threshold of 0.5 ng/mL PCT identified IBI with a sensitivity of 82% and a specificity of 86%. In studies considering IBI, the negative predictive value of PCT lower than 0.5 ng/mL ranged from 98.9% to 99.8%, whereas considering SBI it ranged from 79.5% to 96.7%.
There are several experimental assays under investigation that are promising to perform with a higher sensitivity than CRP and PCT. RNA biosignatures may be able to precisely differentiate viral infections from bacterial infections. Further study and a rapid turnaround time are both required before widespread implementation.
Newer Prediction Models
This algorithm (by Mintegi and colleagues in 2014) was validated by Gomez in 2016 for children 90 days or younger with fever without clear source. This algorithm uses age, clinical appearance, urine dipstick, PCT, CRP, and ANC in a stepwise fashion to determine which infants are high-, intermediate-, and low-risk for SBI and IBI (see www.mdcalc.com/step-step-approach-febrile-infants ). During validation, Gomez demonstrated that step by step was more sensitive for IBI (92%) than Rochester criteria. However, this algorithm has false negatives for neonates younger than 28 days and infants febrile for less than 2 hours. It is the investigator’s preference to use caution in these clinical scenarios.
Kuppermann and colleagues derived and validated an accurate prediction rule to identify infants aged 60 days and younger at low risk for IBIs. This rule, published in JAMA Pediatrics in 2019 uses urinalysis (UA), ANC, and procalcitonin levels to determine infants at risk of SBI. This rule still requires validation on an independent cohort. The authors of this study comment that this rule should not yet be used for neonates aged 28 days or younger, as this rule was designed for IBI, not for serious viral infections such as HSV. This rule requires larger validation on febrile infants in the first month of life with bacteremia and bacterial meningitis before implementation ( www.mdcalc.com/pecarn-rule-low-risk-febrile-infants-29-60-days-old ). This rule also uses a procalcitonin value of 1.71 ng/mL. The group is planning to implement the cut-offs of 4000/μL for ANC and the more commonly used cut-off of 0.5 ng/mL for PCT in future validation.
The REVISE is a national collaborative quality initiative of 124 sites and greater than 20,000 well-appearing febrile infants published in Pediatrics in 2019 (see pediatrics-aappublications-org.easyaccess1.lib.cuhk.edu.hk/content/144/3/e20182201 ). This initiative created a comprehensive toolkit for community-based sites to decrease variability of testing and treatment of febrile young infants. The goals were to standardize hospitalization and length of stay. Well-appearing infants aged 7 to 60 days were included, and those with obvious source of infection, comorbidities, or ill appearance were excluded. Appropriate workup included UA and a marker of inflammation or infection (such as WBC, CRP, or PCT). A lumbar puncture was not required. Appropriate hospitalization included infants who are non-low risk (<31 days of age, those with abnormal laboratory findings, significant past medical history, or social indications for admission). Appropriate length of stay was 42 hours for high-risk infants and 30 hours for low-risk infants. Those diagnosed with meningitis, bacteremia, or UTI were excluded.
The biggest controversy with this initiative is the lack of mandated lumbar puncture for febrile neonates. Many experts are concerned that this is too risky for the younger vulnerable population with significant incidence and associated morbidity and mortality of missed infections, particularly HSV infections.
Risk Stratification Application
Prior studies have repeatedly demonstrated that the neonatal population is difficult to accurately risk stratify. Approximately 5% of febrile neonates have an SBI even when stratified as low risk via laboratory values. Because of improved vaccinations and the associated herd immunity, IBI are less common overall, and UTI is the most common SBI in children. However, significant risk of IBI remains in the neonate, and these infections are frequently not initially clinically apparent. These newer decision tools show promise but do have associated controversy, particularly in the febrile neonate.
See Fig. 3 for a simplified algorithm of the management of the young febrile infant. Because of the significant increased risk of both IBI and SBI in neonates, a complete septic evaluation is recommended. This includes cultures of blood, urine, CSF, and serum/CSF laboratory studies including HSV PCR. Antibiotic administration and hospitalization are also recommended. Hospitalization without antibiotic administration (and without obtaining CSF) may be considered in a well-appearing neonate older than 21 days without HSV signs or risk factors with normal laboratory markers and urine studies and with easy access to medical follow-up. Blood and urine cultures should be obtained, and these patients should be serially examined for signs of decompensation. A chest radiograph (CXR) and stool studies should be obtained respectively with respiratory and gastrointestinal symptoms.