Clinical Practice Guideline: Evaluation and Management of Well-Appearing Febrile Infants 8 to 60 Days Old

Efforts to develop an evidence-based approach to the evaluation and management of young febrile infants have spanned more than 4 decades.¹ In the 1970s, concerns arose about the emergence and rapid progression of group B Streptococcus (GBS) infection in neonates, whose clinical appearance and preliminary laboratory evaluations did not always reflect the presence of serious disease.² Such concerns led to extensive evaluations, hospitalizations, and antimicrobial treatment of all febrile infants younger than 60 days,³ with many institutions extending complete sepsis workups to 90 days. However, the seminal 1983 study by De Angelis et al⁴ highlighted the iatrogenic complications that accompany hospitalizing young, febrile infants and provided an impetus for developing clinical strategies that would be more selective for hospitalizations. Today, the consequences of medical errors during hospitalizations are well known.^5,–7 

In the 1980s and 1990s, there were numerous efforts to develop and validate clinical prediction models for detecting serious bacterial illness (SBI).^8,–15 Efforts were hampered by the heterogeneity of the definition of SBI. Some studies included clinically obvious infections such as cellulitis. Others included pneumonia, which may be viral or bacterial; many included bacterial gastroenteritis in infants with diarrhea. All included urinary tract infection (UTI), bacteremia, and bacterial meningitis, but UTI is so much more common than the other infections that it distorts models attempting to identify all causes.

These prediction models involved a combination of clinical and laboratory test parameters that were based on a priori criteria and were not derived from the primary data. Each variable was defined arbitrarily, such as age groupings in weeks or months and integers ending in zero, for which there is no real physiologic or biological basis. For example, the variable that defined an abnormal white blood cell (WBC) count as <5000 per mm³ or >15 000 per mm³ was not statistically derived but established in advance as an indicator and tested in combination with other predictor variables.

Recommendations emerged that generally relied on clinical appearance, age, urinalysis, WBC count (and/or absolute neutrophil count [ANC], band count, and/or immature to total neutrophil ratio), and cerebrospinal fluid (CSF) analysis (except for the Rochester criteria, which did not require CSF).¹⁰ All had somewhat similar sensitivities and specificities as well as predictive values. The models were promulgated because of moderately high sensitivities (90% to 95%) and high negative predictive values (NPVs ) (97%–99%). The high NPVs were expected because of the uncommon occurrence of the most serious infections, which, along with modest specificities (20% to 40%), also explained the relatively low positive predictive values.

A major shift occurred in the mid-1980s when Powell et al in Rochester accepted the inability to predict who was at high risk and attempted instead to predict who was at low risk, even in the first month of life.^10,14 A pattern emerged in which it was recommended that all infants in the youngest group (<29 days of age) should receive extensive evaluations, hospitalization, and empirical antimicrobial treatment, and infants 29 to 90 days of age could be managed with presumptive intramuscular ceftriaxone as outpatients with pending blood, urine, and CSF culture results.¹⁵ 

In time, other groups used techniques to develop clinical prediction rules that rely on gathered data to derive and define the best, most precise, and parsimonious set of variables that predict a defined outcome that can be translated into recommendations.^16,–18 Still another approach was the sequential approach of established clinical and laboratory criteria.^19,20 Despite these substantial efforts, there has been ongoing evidence that community and emergency physicians do not routinely follow these recommendations in real-world settings.^17,21,–27 Clinical outcomes have not been shown to suffer despite nonadherence to contemporaneous standards of care.

Differing approaches to the management of very young febrile infants indicated the need for a guideline that is current, evidence-based, and developed by a national professional society or organization with broad representation. This led the American Academy of Pediatrics (AAP) to embark on developing this guideline with the assistance of an evidence review commissioned by the Agency for Healthcare Research and Quality (AHRQ).²⁶ 

Attention has been given to the following present-day considerations:

Since the 1980s, the epidemiology of bacterial infections in neonates and infants has changed as a result of many factors, including prenatal GBS screening and incorporation of immunization against Streptococcus pneumoniae. Furthermore, improvements in food safety may have resulted in a decrease in the incidence of disease caused by Listeria monocytogenes in this age group. Recent studies demonstrate that Esherichia coli is now the most common organism to cause bacteremia, while GBS remains the most common cause of meningitis in most studies.^25,27,–31 Infections with L monocytogenes are now rare in the United States.^32,33 The shift from Gram-positive to Gram-negative predominance has implications for the choice of tests, interpretation of values for decision-making, and the selection of antimicrobial drugs. Using the decision models of the 1980s today can lead to misclassification of bacterial meningitis in 23.3% to 32.1% of cases.³⁴ 

Studies indicate significant variation in care and consequently considerable differences in costs.^17,22,–24 Differential access, delays, language barriers, and fragmented care can also be costly to infants, families, and the health care system. A substantial basis for practice variability among clinicians is attributable to differences in infants’ clinical presentations and severities of illness. However, more than 50% of the variability has been unexplained.³⁵ Beyond unnecessary hospitalizations, and financial and social costs, there are also potential harms from hospital-acquired infections and iatrogenesis in prolonged hospitalizations.

Costs are justified on the basis of the magnitude of the benefit and/or reduction of potential harms. In studies of prediction models, instances of missed invasive bacterial infections (IBI) in well-appearing low-risk infants are uncommon. For infants not managed according to existing clinical prediction models, there are also uncommon misses reported in the literature. These factors suggested there is an opportunity to “safely do less.”³⁶ 

The WBC, ANC, and band count, combined with clinical appearance and urinalysis, have been the foundation of earlier clinical prediction models. With E coli replacing GBS as the most common bacterial pathogen in this age group, these markers are no longer as useful. C-reactive protein (CRP), an inflammatory marker (IM) produced by the liver in response to infections and numerous other conditions, is now available for point-of-care testing.³⁷ Procalcitonin, expressed mainly by thyroid C cells, is produced rapidly in response to infection and other tissue injuries. It is more specific for bacterial infections than other IMs and rises more quickly to abnormal values. Procalcitonin has emerged as the most accurate IM for risk stratification available, although not currently available at many sites in the United States with timely results on a 24/7 basis.^38,39 (See additional discussion in KAS 10)

There have been improvements allowing more accurate screening for invasive infections and more rapid and precise identification of bacterial, viral, and fungal pathogens. Automated blood culture systems can now identify most bacterial pathogens in <24 hours. Most recently, nested multiplex polymerase chain reaction (PCR) testing of positive blood cultures can identify bacterial pathogens and antimicrobial resistance genes in approximately 1 hour.^40,–42 Similarly, multiplex meningoencephalitis panels can provide results on CSF for 14 potential CSF pathogens in 1 hour.⁴³ 

The development of rapid viral PCR and multiplex respiratory viral testing has led to identifying emerging agents, such as parechovirus, and prompted analyses of their effect on risk stratification of young febrile infants.^44,–53 Although the presence of documented respiratory viral infections decreases the risk of IBIs in febrile infants (see Inclusion Criteria 5, Positive viral test), it remains unclear how a positive viral test result should influence further laboratory evaluation and management, especially in the first month of life. In addition, it is unclear whether a positive viral test result will either obviate or shorten hospitalization. Researchers in a study analyzing data before the widespread availability of multiplex viral testing (2000–2012) did not find a difference in length of stay between febrile infants with or without positive viral test results.⁵⁴ More work is needed, and this is included as an important question in Future Research.

The area of genomic diagnostics for IBIs is still in its relative infancy, including both genomic identification of viral and bacterial genetic material as well as identifying host genomic responses to viral or bacterial infections. Both need further work to see how these technologies compare in accuracy and timing to routine diagnostic techniques. But progress is being made for RNA transcriptional profiling⁵⁵ and next-generation sequencing of microbial cell-free DNA.⁵⁶ 

Advances in testing and clinical strategies can speed discharge. Data indicate that including evidence-based strategies in care process models can improve infant outcomes.⁵⁷ Hospital environments can be stressful for parents but can be restructured to support maternal/child bonding and breastfeeding.⁵⁸ See further discussion in KAS 6.

Although early studies largely emanated from single-site inner-city emergency departments (EDs), recent investigations conducted by large, geographically widespread research networks and integrated regional health care systems have developed more generalizable evidence.^{17,–20,22,25,57} Advances in data storage and analysis as well as adoption of statistical procedures⁵⁹ for developing clinical prediction rules offer advantages compared with earlier efforts. Collaborative efforts of primary care practices, EDs, hospitals, and integrated health systems are creating larger and more refined data sets. With personalized medicine, enabled by these large data sets and evolving modeling techniques capable of analyzing infants on dozens of variables, the committee anticipates that in the future we will see “one child, one guideline.”

This guideline, grounded in continually expanding evidence and including new technologies, should, for today’s clinicians, form the foundation on which a more nuanced and precise approach can be used to develop an optimal strategy for evaluating and managing each febrile infant. The committee encourages use of the 3 age-based algorithms in Figs 1–3 as a guide to arriving at the best approach. Approaches may differ somewhat depending on many perinatal or neonatal factors, clinician’s experience, parents’ abilities and values, nature of relationship with the infant’s family, characteristics of the clinical setting, and ability to obtain timely laboratory results, among others.

Ongoing research has challenged classifying all infants younger than 29 days as high risk. The Pediatric Research in Office Settings (PROS) study indicated that when combined with other variables, infants >25 days of age were at low risk for IBIs, 0.4%.¹⁷ Subsequently, the European Collaborative Group developed and validated the step-by-step approach with a combination of clinical and laboratory variables that included 22- to 28-day-old infants, capable of identifying infants at low risk for IBIs, ranging from 0.2% to 0.7%.^19,20 A recent scoring system methodologically derived by Aronson et al identified age >21 days to be useful in identifying low-risk infants.⁶⁰ In a prospective study of 4778 infants from the Pediatric Emergency Care Applied Research Network (PECARN), there was a significantly lower rate of bacteremia in the fourth week (1.6%) compared with weeks 2 (5.3%) and 3 (3.3%) and no difference from weeks 5 and 6 (P = .76).⁶¹ A prospective national surveillance study in England analyzed 22 075 episodes of IBI from 2010–2017.⁶² This population-based analysis documented a dramatic decrease in IBI after the first week of life, followed by a continuous stepwise decrease in population incidence over the next 8 weeks. The decline in bacteremia prevalence by age for regional and national studies is portrayed in Fig 4.

Because risk of IBI has extensively been documented to steadily decline over the first few months, any day or week cutoff is arbitrary and subject to interpretation depending on a clinician’s or a parent’s risk aversion or tolerance. These data form the basis for the committee developing a separate algorithm for infants 22 to 28 days of age.

A number of unique challenges confronted the development of an evidence-based approach to the febrile infant.

1. The initial challenge was to decide whether to include infants in the first week of life. The committee decided early on that infants in the first week of life are sufficiently different in rates and types of illness, including early-onset bacterial infection, that they should be excluded from this guideline.

2. Many published studies used SBI as an outcome measure. Because SBI is not a single clinical entity, analyses fell short of identifying the risks for specific infections. UTI is so much more common than the other bacterial infections that it can distort the accuracy of a prediction model to detect bacteremia or bacterial meningitis. This guideline addresses evidence for bacterial meningitis and bacteremia separately from UTIs; the committee strongly discourages further use of the term “SBI.”

3. Meningitis, the most serious bacterial infection responsible for infants’ fevers, is uncommon. Accumulating a large enough sample size to be able to accurately predict infrequent infections is a major research challenge; an even larger sample size is required to address the morbidity and long-term consequences accompanying meningitis.

4. As the epidemiology of bacterial species responsible for infections is continually changing, a prediction model or rule developed today will not necessarily be valid in the future. Species types and resistance patterns also vary geographically.

5. Existing clinical prediction models as well as prediction rules often rely on “clinical appearance,” well versus ill, a subjective assessment.^{8,–17,19,20} Despite an elegant process of development, the Yale Observational Score,⁸ a formal scoring system for illness appearance, has not proven to be useful in this age group.^63,64 The accuracy of clinician assessment is likely related to experience. Unfortunately, there is no measure or adequate definition for what constitutes “experienced,” or of “well appearing.” Researchers in large studies have often treated clinical appearance as binary: well appearing or not, or ill appearing or not. When offered 3 categories, however, both senior residents⁶⁵ and experienced pediatricians¹⁷ classified a quarter of the young febrile infants they encountered in an intermediate category, acknowledgment that the distinction between “well” and “ill” is not always clear-cut. The distinction is likely to be most difficult before the emergence of the social smile, which enables the infant to “respond to social overtures,” a key element in the Yale Observational Score.⁸ Clinicians differ in a variety of ways including knowledge, clinical experience with febrile infants, and in the time available to evaluate and monitor infants. The committee acknowledges that some clinicians may have different levels of experience and confidence in determining well appearance compared with experienced pediatricians.

6. Clinicians work in different settings with a range of familiarity with their patients and families, access to medical records, and abilities to follow-up with patients in a timely fashion.

7. Clinicians have variable access to newer diagnostic tests and timely results, particularly procalcitonin.

8. Families possess a spectrum of knowledge and skills to continuously observe and assess infants discharged from the hospital. Multiple factors may affect a timely return visit. There has been considerable interest focused on shared decision-making for young febrile infants,^66,–70 including a recent mobile device app to help clinicians communicate with parents.⁷¹ 

For purposes of this guideline, the committee believes that at a minimum, families should be provided with information about the risks and benefits of all procedures, including invasive procedures such as a lumbar puncture (LP) and a bladder catheterization. An opportunity for questions and dialogue between the family and care team should be provided. Families’ decisions about their infant will be made in the context of their previous experiences with the health system, their personal beliefs and values, and knowledge and understanding of their child’s condition and diagnostic and treatment options and outcomes.

The decision to actively monitor an infant at home or in the hospital requires a collaborative discussion between the family and the care team. The discussion should be centered on the best interest of the child, taking into account the family’s and the care team’s assessment of the multiple factors of risk and risk tolerance, experience and comfort of monitoring an ill infant, and ease and accessibility of transportation. Academic medical centers and children’s hospitals generally provide high-quality observation for ill infants, as do many community hospitals with dedicated pediatrics units. Many hospitals do not have nurses and staff with experience and skills caring for young infants, however. In the current health care system, insurance status and coverage may further affect the family’s and care team’s decision on location of monitoring.

Even with the availability of valid and reliable data, thoughtful investigators and clinicians will have different thresholds for recommending diagnostic tests and therapeutic interventions. The committee believes understanding risk tolerance is of fundamental importance to guideline interpretation. In a straightforward case of a febrile infant having CSF pleocytosis with a predominance of polymorphonuclear leukocytes and a positive Gram stain result, the committee would expect clinicians to unanimously agree the infant be hospitalized and receive immediate antimicrobial treatment. Similarly, on the basis of prevalences cited in KAS 1, 8, and 15, a risk for UTI can be estimated at 10%, which translates to a recommendation to perform 10 urinalyses to detect a single UTI, or a number needed to test of 10. This is an example in which agreement to perform a urinalysis is expected. However, challenges frequently occur. For example, if clinical and laboratory evaluations suggest the likelihood of bacteremia is 1:100 (number needed to treat = 100) or a risk of bacterial meningitis at 1:1000 (number needed to test = 1000), is it worth 100 doses of antibiotics to treat a single case of bacteremia while awaiting blood culture results? Should the committee recommend performing the number of LPs required to obtain 1000 samples of interpretable CSF to prevent a delay in recognizing and treating a single case of bacterial meningitis? Responses to these questions depend on how much risk is considered tolerable. The challenge in guideline development was succinctly stated as, “Thus, evidence alone never speaks for itself or conveys the truth because it always requires interpretation.”⁷² In the committee’s discussions, responses to the above questions and similar issues varied among and within the specialty groups constituting the committee and reviewers.

Differences in risk tolerance also exist between parents and physicians and may exist among family members. A clinician may estimate that an infant’s risk of meningitis is 1% and an LP is indicated, whereas a parent may have a higher threshold for consenting to the procedure. These differences, along with other parent beliefs and values, provide further challenges in an effort to share decision-making in an acute setting.

The recommendations in this guideline reflect universal agreement or a strong consensus among committee members. In the one situation when there was majority but not consensus agreement, additional committee members were appointed and added; subsequently, consensus was achieved. The major reason for disagreement was varying levels of risk tolerance among committee members. For these recommendations, a more detailed explanation of the uncertainties involved and attempts to derive numbers needed to test and numbers needed to treat are provided in the specific Key Action Statements.

The working group consisted of representatives from epidemiology; general pediatrics; pediatric subspecialties, including emergency medicine, infectious diseases, and hospital medicine; and family medicine. Individuals with expertise in guideline development, algorithm creation, and quality improvement were also included. During the development of this guideline, all members had access to the AHRQ evidence review,²⁶ the additional analyses by the committee epidemiologist (C.R.W. Jr.) as well as others, copies of all published literature cited in these reports, and the opportunity to participate in 4 meetings convened at the AAP and on conference calls. The authoring group relied on data and analyses from the following: (1) a formal analysis and systematic review of published articles from the United States and selected international countries that was conducted by an Evidence-Based Practice Center under contract to AHRQ; (2) a supplemental review and analyses were performed by the epidemiologist assigned to the committee; (3) consistent with a previous AAP guideline⁷³ if literature gaps existed, data were solicited and received from authors of previously published, peer-reviewed articles who performed additional analyses from their investigations: Kaiser Permanente Northern California; Intermountain Healthcare; the AAP PROS network; the Febrile Young Infant Research Collaborative (FYIRC); Boston Children’s Hospital; The European Collaborative Group; Cruces University Hospital, Barakaldo, Spain; and the PECARN; and (4) committee members with active research and data collection projects provided ongoing study reports. Ongoing data analyses from these works in progress are consistent with cited references and support the recommendations.

Finally, after the formulation of a set of recommendations, there was further consideration by AAP Sections and Committees, external organizations, physician reviewers, and parents, as well as focus groups of pediatricians from general pediatrics, pediatric hospital medicine, pediatric emergency medicine, pediatric critical care, and pediatric infectious diseases (see Acknowledgments for review groups).

The committee’s focus was to develop a guideline to improve the diagnosis and treatment of UTIs, bacteremia, and meningitis. Sometimes the term “SBI” is used because it was the only outcome measure reported in many investigations. In some analyses, bacteremia and bacterial meningitis are combined as IBIs because of the nature of those infections compared with UTIs.

Recommendations are contained in the algorithms for infants 8 to 21 days of age, 22 to 28 days of age, and 29 to 60 days of age and are expounded in the accompanying Key Action Statements. For each recommendation, the quality of available evidence on which the recommendation is based is rated, and the strength of each recommendation is provided (Fig 5). Risks and benefits also are indicated, and assessments of their balance are provided.

In accordance with recent suggestions by the National Academy of Medicine, the committee attempted transparency by occasionally commenting on value judgments.⁷⁴ A clinical decision involves more than just knowing a specific risk. The decision about what action is appropriate with a given risk depends on the experience, value judgments, and risk tolerance and aversion of the interpreting clinician. To the extent possible, it is appropriate to incorporate parents’ values and preferences in shared decision-making.

As noted above and consistent with all AAP clinical practice guidelines, each recommendation represents a consensus of the committee, although not necessarily universal agreement.

This guideline addresses febrile infants who are well appearing. Infants appearing moderately or severely ill are at higher risk for IBIs and are NOT addressed in the guideline. Because of the difficulties assessing well appearance discussed previously in Challenges, we recommend that when clinicians are uncertain as to whether an infant is well appearing, this guideline should not be applied.

For eligibility, this guideline addresses febrile infants who (1) are well appearing, (2) have documented rectal temperatures of ≥38.0°C or 100.4°F at home in the past 24 hours or determined in a clinical setting, (3) had a gestation between ≥37 and <42 weeks, and (4) are 8 to 60 days of age and at home after discharge from a newborn nursery or born at home.

The following merit additional consideration specific to their condition and are intended to be excluded from the algorithms:

1. Preterm infants (<37 weeks’ gestation).

2. Infants younger than 2 weeks of age whose perinatal courses were complicated by maternal fever, infection, and/or antimicrobial use.

3. Febrile infants with high suspicion of herpes simplex virus (HSV) infection (eg, vesicles).

4. Infants with a focal bacterial infection (eg, cellulitis, omphalitis, septic arthritis, osteomyelitis). These infections should be managed according to accepted standards.

5. Infants with clinical bronchiolitis, with or without positive test results for respiratory syncytial virus (RSV). A review by Ralston et al of 11 studies of bronchiolitis found no cases of meningitis, and researchers in 8 studies reported no cases of bacteremia.⁵¹ 

6. Infants with documented or suspected immune compromise.

7. Infants whose neonatal course was complicated by surgery or infection.

8. Infants with congenital or chromosomal abnormalities.

9. Medically fragile infants requiring some form of technology or ongoing therapeutic intervention to sustain life.

10. Infants who have received immu-nizations within the last 48 hours. The incidence of postimmunization fevers ≥38.0°C is estimated to be >40% within the first 48 hours.⁷⁵ 

Infants with the following may be included:

1. Respiratory symptoms: the presence of upper respiratory tract infection symptoms or other respiratory symptoms not diagnostic of bronchiolitis should not exclude infants from inclusion in the pathway.

2. Diarrhea: infants suspected of having diarrhea caused by treatable bacterial pathogens should have stool specimens tested. If studies for bacteria are negative, infants may then enter the decision tree pathway. Loose stools do not exclude infants from the pathway.

3. Otitis media: diagnosing infants with presumed otitis media does not preclude their entry into the pathway.

4. Current or recent use of anti-microbial agents in infants older than 2 weeks of age requires individualized interpretation for febrile infants who enter the pathway.

5. Positive viral test results: the availability of rapid respiratory molecular testing for a variety of viruses is increasing, outpacing the availability of evidence for how such testing should be used.

The 2014 Cochrane review that included older infants and children did not recommend respiratory viral testing in the ED.⁵² In evaluating the implications of a positive viral respiratory test result, numerous studies have documented lowering of IBI risk in subsets of patients. However, no prospective study has yet provided convincing data on whether a positive viral test result sufficiently reduces the IBI risk to change decision-making, after considering other historical, clinical, and available markers of inflammation.

In a 2004 study, Byington et al evaluated whether a positive respiratory viral test result lowered the risk of IBI in 1385 infants 1 to 90 days of age.⁷⁶ Viruses were detected in 35%, and the bacteremia risk in the viral-positive infants was 1%, significantly lower than the 2.7% in viral-negative infants. When positive viral test results were combined with the Rochester classification, there was no reduction in risk for infants already classified as low risk. Rochester high-risk group infants with positive viral test results had a similar prevalence of bacteremia as low-risk infants.

Emerging data from several large studies address viral testing in young febrile infants stratified by age. Infants <28 days of age with a positive viral test result have a risk of IBI from 1.1% to 2.1%.^44,–50,76 One study found statistically significant reductions in the prevalence of IBI when compared with viral-negative infants.⁵⁰ Other studies revealed lower rates of IBI but not statistically significantly lower.^44,47,48 In a prospective PECARN study for infants <28 days of age, bacteremia was detected in 1.1% and meningitis in 0.8% of infants with detected viral infections.⁴⁸ The risks of IBI in viral-positive infants <28 days of age are sufficiently high to warrant similar testing and treatment as viral-negative infants.

For infants 29 to 60 days of age, the bacteremia rate was significantly lower in viral-positive infants compared with viral-negative infants (0.6% vs 1.8%).⁴⁸ Another recent study of 29- to 90-day old infants detected bacteremia in 3.7% of viral-negative infants, whereas those with rhinovirus infections had a prevalence of 1.4% and a reduced relative risk of 0.52 (95% confidence interval [CI], 0.34–0.81).⁵⁰ There are situations in which viral testing may augment the recommended evaluation and management of febrile infants 29 days and older, such as during RSV, bronchiolitis,⁵¹ or influenza seasonal outbreaks. In these situations, individual tests for RSV or influenza can each be obtained at <3% the cost of a multiplex respiratory viral panel, according to the latest charges listed in Current Procedural Terminology; the cost of multiplex testing in other countries has been reported to be substantially lower. In summary, although viral testing should not affect entrance into the recommended pathway, for infants >28 days of age, it can be considered in individualizing evaluation and management decisions.

The following recommendations and options are for febrile (temperature ≥38.0°C), well-appearing, term infants 8 to 21 days of age without risk factors identified in the exclusion criteria.

A positive urinalysis result for purposes of this guideline is defined as the presence of any leukocyte esterase (LE) on dipstick, >5 WBCs per high-powered field (hpf) in centrifuged urine, or >10 WBCs/mm³ in uncentrifuged urine on microscopic urinalysis using a hemocytometer.

Urinalysis: Of the estimated 10% of febrile infants with UTIs, 94% have urinalysis positive for leukocyte esterase (LE) (95% CI, 91%– to 97%).⁸⁰ The sensitivity is even higher for UTI associated with bacteremia (97.6% and 100% in 2 studies).^80,86 Therefore, for 1000 infants, ∼approximately 94 to 98 infants with UTIs will be detected by a positive urinalysis result, and 2 to 6 may be “missed.” It is unclear whether a “miss” represents a UTI, asymptomatic bacteriuria, or contamination. Consequently, if a urinalysis result is negative, an estimated 200 to 500 catheterizations or suprapubic aspirations (SPAs) followed by cultures would be required to detect 1 additional infant with bacteriuria, and that infant might have asymptomatic bacteriuria or contamination rather than a true UTI.

Culture: In the AAP clinical practice guideline on UTI from 2011, reaffirmed in 2016, addressing infants 2 to 24 months of age, the diagnosis of UTI was made on the basis of pyuria and at least 50  000 colony-forming units (cfu) per mL of a single uropathogenic organism in an appropriately collected specimen of urine.⁷³ Recent studies indicate it is reasonable to extend the recommendation of the AAP UTI guideline to infants addressed here,^{77,–80,85,–89} although 10  000 colony-forming units/mL is now an acceptable threshold for diagnosing UTI from catheterized urine specimens when pyuria and fever are also present.^80,89 This new level also circumvents the problem of interpreting data from laboratories not reporting gradations from 10  000 to 100  000. Positive urine culture results obtained in the absence of an abnormal urinalysis indicating inflammation are likely to represent asymptomatic bacteriuria or contamination.

Culture of urine specimens not collected by catheterization or SPA is not recommended because of an unacceptable rate of false-positive results attributable to contamination of such specimens.^77,78 An initial urine specimen obtained by catheter or SPA obviates the delay and need for a second specimen by catheter or SPA following after a positive result from a bag urine. The sensitivity and specificity of urinalysis parameters for UTI from bagged specimens are somewhat less than those of catheterized specimens.^77,78 

For physicians with experience, SPA is effective, provides the “cleanest” specimen, and saves time; complications are rare.⁸¹ In some situations, such as phimosis or labial adhesions, SPA may be required⁷³; a training video is available online.⁸² 

Because it is recommended that all 8- to 21-day-old infants be hospitalized and treated, IMs are not required for these initial decisions. However, some clinicians consider them useful in decision-making about later management, such as whether to discontinue antimicrobial agents at 24 or 36 hours while awaiting final results of bacterial cultures.

CSF with pleocytosis or from infants with HSV risk factors should be evaluated for HSV.^116,117 Population-based rates of HSV in neonates range from 2 to 5 per 100  000, with 15% having fever as the only symptom.^108,–116 Although rare in well-appearing infants, prompt recognition and treatment of HSV in infants, especially those younger than 21 days with other risk factors, is essential. In addition to the presence of vesicles and/or seizures, infants should be considered at increased risk of HSV if any of the following are present: CSF pleocytosis with a negative Gram stain, leukopenia, thrombocytopenia, hypothermia, mucous membrane ulcers, or maternal history of genital HSV lesions or fever from 48 hours before to 48 hours after delivery. If liver function tests were obtained, an elevated alanine aminotransferase (ALT) also indicates a higher risk of HSV. For further details of evaluation and management of HSV, see the AAP Red Book.¹¹¹ 

Enterovirus (EV) PCR testing should be performed on CSF with pleocytosis and during months when there is a seasonal increase in enterovirus, regardless of pleocytosis. Rapid detection of enterovirus, along with HSV and an emerging viral cause of meningitis, human parechovirus (HPeV), can be accomplished with meningoencephalitis multiplex PCR panels identifying 14 pathogens.^43,118,119 When available in a timely fashion, multiplex PCR testing can enhance clinical decision-making.

Pleocytosis is detected overall in 8.8% of CSF analyses; the rate is higher in summer (17%) because of enterovirus.¹¹⁷ The likelihood of bacterial meningitis in the presence of enterovirus in the CSF is low.¹²⁰ Therefore, the detection of CSF enterovirus can eliminate the need for further interventions.^121,122 Newer tests provide rapid identification of enterovirus.^123,124 CSF pleocytosis is often detected in febrile infants with UTIs who do not have bacterial, enterovirus or HSV meningitis.^126,–128 These panels can give rapid results but should only be used as an addition to bacterial cultures. There are still relatively limited data on young infants so precise test accuracy is still uncertain, and there have been reports of both false-positive and false-negative results; Listeria is not in the panel.^118,119 

An LP is not always successful. The rate of failure and/or traumatic LP in infants younger than 90 days is 20% to 50%; the rate of unsuccessful or dry LP is 25% to 40%; the rate of bloody LP is 10% to 30%.^{106,130,–132} Ultrasonography may assist in obtaining CSF.¹³³ When using a bedside ultrasound landmark-guided technique, success in obtaining CSF on the first LP attempt was 58% compared with 31% without ultrasonography. Using ultrasonography resulted in a 75% success rate after 3 attempts.¹³⁵ 

There is also a significant rate of nonpathogenic bacteria cultured from CSF. In a multisite study with 410 positive CSF bacterial culture results in infants <90 days of age, researchers found only 13% were pathogens and the rest were contaminants.¹⁰⁷ Authors of another study from Kaiser Permanente Northern California found only 22% of CSF isolates from infants <90 days to be pathogens.²⁷ Authors in a study of febrile infants in the second month of life found that 40 of 41 positive culture results were caused by contaminants.¹⁰⁶ 

The CSF from a traumatic LP should be cultured and can be tested for HSV if indicated. In general, correction (or ratios) for red blood cells (RBCs) in CSF is discouraged because of lack of validating studies. It is reasonable to interpret CSF WBC counts at face value in CSF specimens with up to 10 000 RBCs per mm³ (Table 2).¹³³ 

The antimicrobial agents in Table 3 are recommended for initial empirical therapy and should be modified following results of cultures and sensitivities.

The recommendation to treat all infants 8 to 21 d of age is based on the prevalence of IBIs being highest in this age category (Fig 4) and ∼2% (number needed to treat 50) even in infants with negative urinalysis and or IMs. The preponderance of evidence indicates that infants with viral infections have a risk of IBI of ∼1% or a number needed to treat of 100. See above discussion.

Overall, for studies since the year 2000 in infants <90 days of age, Gram-negative organisms have been responsible for the majority of infections (60% to 80%). E coli has been the most common pathogen detected,

Enteroviral testing of CSF has been shown to shorten length of stay and duration of antimicrobial use.^120,137 It is helpful if available within a time period that will assist clinical decision-making. In general, if CSF is positive for enterovirus, antimicrobial agents should be discontinued (or withheld), because concomitant enteroviral and bacterial meningitis is rare. However, in some cases of enterovirus meningitis or meningoencephalitis, CSF may reveal a significant pleocytosis with a neutrophil predominance. In such cases, or in cases in which there is otherwise reason to suspect a concomitant bacterial infection, such as abnormal IMs, it is reasonable to continue antimicrobial agents until CSF and blood cultures are negative for 24 to 36 hours.

In communities with circulation of E coli strains that produce extended-spectrum β-lactamases, gentamicin should be used instead of ceftazidime for treatment of suspected bacteremia or sepsis, and meropenem should be used instead of ceftazidime when bacterial meningitis is suspected. Use of fourth- and fifth-generation cephalosporins may also be considered with expert consultation.

Cephalosporins do not provide adequate coverage for Listeria or enterococci. Ampicillin generally should be used as part of empirical therapy when these microbes are suspected.

The committee recommends that, to improve the care of hospitalized infants, efforts should be directed at optimizing the environment to support maternal/child bonding and breastfeeding. This can be accomplished through the following effective measures: allow parents to room-in with the infant; encourage the continuation of breastfeeding and provide lactation support including access to breast pumps for nursing mothers; provide timely communication with families about the results and interpretation of testing and expected consequences of having a diagnosis of UTI, bacteremia, and/or bacterial meningitis on the basis of ongoing results; provide timely communication with the infant’s primary care provider.

1. culture results are negative for 24 to 36 hours or only positive for contaminants;

2. the infant continues to appear clinically well or is improving (eg, fever, feeding); and

3. there are no other reasons for hospitalization.

Evidence Quality: B; Strong Recommendation

Although infants whose CSF is positive for enterovirus may be observed without antimicrobial agents, they should remain in a hospital setting for a minimum of 24 h because of the small risk of progression to enteroviral sepsis, which generally only occurs in infants <21 d of age.

Discontinuation of antimicrobial agents and discharge at 36 hours can potentially result in a lapse of treatment of a slow-growing pathogen and readmission, but this has seldom been reported. Automated blood culture techniques and multiplex PCR detection have reduced the time to identify pathogens.^40,–42 Time to positivity of blood culture is dependent on the type and concentration of bacterial organism. Between 4% and 17.6% of pathogens take >24 hours to grow; less than 5% take >36 hours.^138,–144 Compared with ill-appearing infants, infants not appearing ill are less likely to have pathogens identified in <24 hours (85.0% vs 92.9%). Pathogens vary in median times to positivity: GBS takes 9.3–14.3 hours^{138,–140,143}; E coli takes 11.3–13.6 hours^138,140,143; and S aureus takes 18.5–19.9 hours.^{138,–140,143} For E coli, the most common organism identified, 24% take longer than 24 hours to grow, whereas only 5.9% of GBS grow after 24 hours.¹³⁸ 

Nonpathogens generally take longer than 24 hours to grow in culture media. Approximately 25% of nonpathogens grow in the first 24 hours.¹³⁸ Antimicrobials can be stopped at 24 hours if a pure growth of a nonpathogen is identified. When available, multiplex PCR is capable of detecting many bacterial pathogens and antimicrobial resistance from a positive culture medium in an hour.^40,–43 

The following recommendations and options are for febrile (temperature >38.0°C), well-appearing, term infants 22 to 28 days old without risk factors identified in the exclusion criteria.

The evidence indicates the risk of bacteremia and bacterial meningitis is lower in infants 22 to 28 days of age than in infants 8 to 21 days of age. However, they continue to be at higher risk than older infants, leading us to separate this group as discussed above in the section on “Evidence for Age-based Risk Stratification.”

IMs have been included in every strategy proposed to address febrile infants. No single IM, in isolation, is reliable for risk stratification. Further study will allow ongoing accumulation of evidence and more precise values for these markers. The committee anticipates modification and refinement as efforts to improve the care of febrile infants continue.

• Temperature >38.5°C: A sign of inflammation, fever is the most readily available marker of infection. Surprisingly, it was not included in early studies of decision models,⁰ but there has been ongoing and recent work on the value of fever elevation in predicting IBI.^{16,17,48,57,60,95,96,147} It emerged as an important predictor in studies using recursive partitioning analysis to derive threshold fever values for prediction rules.^16,17 In the PROS Network study of 3066 infants with 63 cases of IBI, a temperature >38.5°C, when combined with ill appearance and age <25 days, had a sensitivity of 93.7% and NPV of 99.6%.¹⁷ A temperature ≥38.5°C at any point during the ED stay placed infants at higher risk in a study of 207 cases of IBI in well-appearing febrile infants ≤60 days seen in the EDs of 11 children’s hospitals in the Febrile Young Infant Research Collaborative.⁶⁰ Researchers in a PECARN analysis addressing SBI documented an increased in adjusted odds ratio of 1.8 for each 1°C increase >38.0.⁴⁸ Also, a temperature <38.5°C is used in Intermountain Healthcare’s Care Process Model to distinguish whether there is a need for further testing in infants older than 28 days who test positive for RSV.⁵⁷ Recently, by adding a temperature >38.5°C as an additional high-risk criterion to the Rochester criteria in 7- to 28-day-old infants, the Roseville Protocol documented a sensitivity of 96.7%.¹⁴⁷ Therefore, moderately elevated temperatures are useful in predicting IBI and can immediately suggest how extensive an evaluation may be appropriate. However, as an independent predictor, 30% of febrile infants with IBI have maximum documented fevers of ≤38.5.⁹⁶ Temperature elevation is a useful predictor of IBI when combined with other clinical features, and laboratory-based IMs can improve the sensitivity for detecting IBI.

• Elevated WBC count and its components: These tests are widely available, but with an evolving epidemiology of IBI and availability of newer tests, their usefulness in predicting IBIs is changing. The arbitrary thresholds (WBC count >15 000 per mm³, ANC >10 000 per mm³, band count >1500 per mm³, immature to total neutrophil ratio >0.2) that define “abnormal” have been used in numerous studies of predictive models.^0,19,20 These studies all used WBC count components in combination with other infant characteristics such as well appearance, or urinalysis results, to identify low-risk infants. Researchers who analyzed WBC count and/or ANC as independent predictors of IBI^{16,39,103,104} have documented that as a stand-alone screen, neither is sufficiently sensitive nor specific, although ANC is substantially better than the WBC count. Researchers in an ED study of 5279 infants <90 days of age identified 68 infants with IBIs.¹⁶ Using a derived multivariable prediction rule with recursive partitioning analysis, they found that there were 14 misclassified cases of bacteremia and 1 case of bacterial meningitis. Of these 15 infants, 9 had “normal” WBC counts (5000–15 000/mm³). This study indicates that a normal WBC count is not reassuring.¹⁶ In a French study of 2047 febrile infants seen in 15 pediatric EDs, the area under the curve (AUC) for WBC count was 0.48 compared with 0.61 for ANC.³⁹ In the PROS study, an abnormal WBC count (<5000/mm³, >15 000/mm³) was significant in a multivariate analysis with an adjusted odds ratio of 3.62 (95% CI, 2.13–6.15) and slightly increased the AUC of a non–laboratory-based model from 0.767 to 0.803. The committee does not recommend use of abnormal WBC count for risk stratification.

• ANC: >4000,¹⁸ >5200⁶⁰ cells per mm³. Although arbitrary values of ANC continue to be included in decision models, researchers in 2 studies methodologically derived optimal cutoffs. The subcommittee presents both values (>4000, >5200), reflecting the current state of the evidence.

1. In a prospective study of 1821 febrile infants with 30 cases of IBI younger than 60 days, the PECARN group used recursive partitioning to derive optimal thresholds for detecting IBI. This study found that an ANC of >4090 per mm³, when combined with an abnormal urinalysis and a procalcitonin of greater than 1.7 ng/mL, detected 29 of 30 cases, 96.7% (95% CI, 83.3%–99.4%) with a specificity of 61.5%.¹⁸ No case of meningitis was missed.

2. The Febrile Young Infant Research Collaborative study did not include procalcitonin but methodologically derived an ANC ≥5185 per mm³ as part of a scoring system to identify IBIs retrospectively. The sensitivity of its scoring system for 207 cases of IBIs was 98.8% (95% CI, 95.7%–99.9%) but had a specificity of 31.3%; none of the 26 cases of bacterial meningitis was missed.⁶⁰ 

The step-by-step method proposed by the European Collaborative of 11 EDs^19,20 selected a higher ANC threshold (10 000) for its model and detected 81 of 87 infants with IBIs. No cases of bacterial meningitis were missed; the sensitivity for IBIs was 92% (95% CI, 85.0%–97.2%), lower than the 2 American studies. The only prospective office-based study, using recursive partitioning, did not identify ANC as a predictor for the 63 cases of IBIs.¹⁷ 

ANC is helpful but not as accurate as newer IMs.¹⁶ In a subset analysis of 46 infants 8 to 60 days of age with bacterial meningitis, blood ANC ranged from 600 to 24 500, with a median of 4700; 39% had ANCs <4000 and 80% had ANCs <10 000.^17,20 As used in a PECARN analysis, an ANC of <4090 combined with a negative urinalysis result had a sensitivity of 76.6% (95% CI, 0.59%–0.88%); addition of procalcitonin was required to achieve the high sensitivity of its decision rule for IBI.¹⁸ Because of availability, timeliness, and these data, an elevated ANC is a useful IM when combined with other clinical and laboratory predictors.

Although several studies have identified ANC cutoffs for infants at low risk of IBI,^18,–20,60 counts <1000 should raise concerns for sepsis in the youngest infants.

• CRP (≥20 mg/L): In studies addressing laboratory markers, CRP has been shown to be more accurate than WBC count or ANC in detecting bacteremia and meningitis.^39,101,102 As independent predictors of IBIs, the AUC for CRP was documented as 0.77 compared with 0.61 for ANC,³⁹ with another study producing values of 0.75 and 0.65, respectively.¹⁴⁶ In the absence of procalcitonin and in combination with other clinical predictors, a CRP ≥20 mg/L has identified infants at higher risk.^19,20,101 It generally can be determined in a timely fashion and has recently become available as a point-of-care test.³⁷ 

• Procalcitonin (>0.5 ng/mL): Serum procalcitonin, as an independent predictor of bacterial infections, has better test characteristics than other laboratory markers of inflammation. In a prospective study of 15 French EDs, Milcent et al³⁹ identified 21 infants 7 to 90 days of age with IBIs. The AUC for procalcitonin, CRP, ANC, and WBC count were documented to be 0.91, 0.77, 0.61, and 0.48, respectively. In this study, a procalcitonin value of 0.3 ng/mL best demarcated low- and high-risk infants and in multivariate analysis was the only independent predictor of IBIs. These findings were replicated in a recent ED study from Spain¹⁴⁶ with 38 infants <60 days of age with IBIs. The AUC for procalcitonin, CRP, and ANC was 0.82, 0.75, and 0.65, respectively. The value of procalcitonin when used in combination with other clinical and laboratory findings is becoming clear.^{18,–20,38,0–105} Using a procalcitonin level of >0.5 ng/mL, along with other clinical variables, was useful in identifying a low-risk group (0.7%) for IBIs in infants >21 days but an unacceptably low sensitivity of 44% for younger infants.¹⁰⁰ The PECARN study, described above, demonstrated a sensitivity of 96.7% by adding an elevated procalcitonin (1.7 ng/mL) to leukocyturia and ANC >4090 mm³. Changing the procalcitonin level to 0.5 ng/mL (and the ANC to 4000 mm³) only minimally decreased rule specificity, so it is advocated by the PECARN investigators as a safer and easier-to-apply cutoff. Procalcitonin is the earliest IM to increase but may still be negative in febrile infants,¹⁸ including those evaluated in the first hours after onset of fever.¹⁴⁶ Although it is currently the best IM available, it should not be used alone for decision-making; 20% of febrile infants with bacterial meningitis had procalcitonin <0.5 ng/mL.²⁰ 

The committee recommends procalcitonin in all age groups. Procalcitonin testing is not yet routinely available in many institutions in the United States. If procalcitonin is unavailable or results are not reported in a timely fashion, the committee recommends using a fever of >38.5°C in combination with other IMs for purposes of risk stratification.

1. urinalysis result is negative or positive;

2. no IM obtained is abnormal;

3. blood and urine cultures have been obtained; and

4. infant is hospitalized.

Evidence Quality: B; Moderate Recommendation

There are insufficient data to estimate the probability of meningitis in this age group if only 1 IM is abnormal or if only a urinalysis result is positive. Almost all current decision rules and models rely on a combination of at least 2 IMs and a urinalysis to define risk.

Recent studies from primary care and EDs document LPs in infants <28 days of age being performed in 60% to 82% of evaluations. There is wide regional variation ranging from 10.7% to 31.3% of infants going without an LP.^23,24,148 With recent data, Kaiser Northern California documents 39% of 7- to 28-day-old infants with fever did not undergo LP. Infants evaluated in the ED were 5 times more likely to have an LP than those evaluated in the office.²² There were no reported cases of delayed recognition of bacterial meningitis in settings in which LPs were not universally performed.

In infants <28 days of age, none of the 21 cases of bacterial meningitis in the PROS, PECARN, and step-by-step studies were missed (sensitivity 100%; CI, 84%–100%). Using a bacterial meningitis prevalence in 22- to 28-day-old infants of 0.39²² or 0.46⁹⁴ or ∼1 in 200 to 250 and the lower end of the sensitivity CI (84%) suggests 1250 to 1560 interpretable CSF samples would be required to detect each additional case of bacterial meningitis (number needed to test = 1250–1560). Without procalcitonin, these studies detected 14 of 14 cases of bacterial meningitis (95% CI, 80%–100%), indicating a number needed to test of 1000 to 1250.

Researchers in a few studies have addressed a positive urinalysis result or UTI as a risk factor for meningitis. Data for 22- to 28-day-old infants are limited, as are data for UTI without abnormal IMs. For infants 7 to 30 days of age in the Reducing Variability in the Infant Sepsis Evaluation study of 1281 infants with positive urinalysis results who had an LP performed, 0.8% were treated for bacterial meningitis.¹⁴⁹ This was similar to the 1.0% of the 4644 infants with negative results on the urinalysis. The data also indicated that none of the 98 infants with positive urinalysis results did not have an LP ultimately had meningitis detected. Similarly, in an outpatient study of 100 infants with UTI <30 days of age, researchers found no cases of meningitis.¹⁵⁰ However, in both of these studies, the lower limits of the CI indicates up to 4% could be missed.

See note on KAS11a.

The antimicrobial agents in Table 3 are recommended for initial empirical therapy and should be modified following results of cultures and sensitivities.

1. CSF analysis suggests bacterial meningitis; or

2. urinalysis result is positive.

Evidence Quality: A; Strong Recommendation

1. CSF analysis is normal;

2. urinalysis is normal; and

3. any IM obtained is abnormal.

Evidence Quality: B; Moderate Recommendation

1. urinalysis is normal;

2. no IM obtained is abnormal; and

3. CSF analysis is normal or entero-virus-positive.

Evidence Quality: B; Weak Recommendation

Recent evidence documents the sensitivity of LE for UTI of 94% (95% CI, 91%–97%),⁷⁹ even higher in UTI associated with bacteremia (97.6% and 100% in 2 studies)^80,86; an NPV of 99% also supports a low likelihood of UTI.^78,85,–89 There are insufficient data to estimate precisely the risk of bacterial meningitis with normal CSF analysis, but, based on the scarcity of cases in the literature, the risk appears to be quite low. However, as current prediction rules fail to detect about 3% to 8% of bacteremia cases, antimicrobial agents may be administered.^18,20 

1. urinalysis is normal;

2. no IM obtained is abnormal; and

3. CSF analysis is normal.

Evidence Quality: C; Moderate Recommendation

If all IMs are normal and urinalysis and CSF analysis do not suggest infection, the risk of bacteremia is between 1% and 2% (number needed to treat 50–100).

1. urinalysis is normal;

2. no IM obtained is abnormal;

3. CSF analysis is normal or enterovirus-positive;

4. verbal teaching and written instructions have been provided for monitoring throughout the period of time at home for the following:

   • change in general appearance, particularly a dusky color, or respiratory or other distress;

   • behavior change, including lethargy, irritability, inconsolable crying, difficulty in consoling/comforting, or other evidence of distress;

   • difficulty feeding;

   • vomiting; and

   • decreased urine output;

5. follow-up plans for reevaluation in 24 hours have been developed and are in place; and

6. plans have been developed and are in place in case of change in clinical status, including means of communication between family and providers and access to emergency medical care.

Evidence Quality: B; Moderate Recommendation

Value judgments: The committee values careful infant monitoring provided by hospital staff skilled in the care of neonates and young infants. In some situations, infants may not be hospitalized because of lack of access to a local hospital unit able to care for young infants (in which case referral to a regional hospital is an acceptable alternative) or other circumstances. In primary care settings, in which close follow-up is possible, more than 30% of low-risk infants are managed at home after initial evaluation.^17,22 For infants seen in EDs, 15% to 30% are not hospitalized.^23,24 In these studies, the subsequent admission rate is 1% to 2%; delays in treating bacterial infections have been rare. Several recent studies suggest otherwise low-risk infants in the absence of CSF data may be of sufficiently low risk to safely be managed at home after initial evaluation.^18,20 

For infants discharged from the hospital after initial evaluation, phone or other telecommunication contact should be attempted and documented at appropriate intervals after returning home. Infants should be scheduled for repeat clinical evaluation within the next 24 hours or sooner, if deemed appropriate. If at 24 hours, the parents report no clinical worsening and all culture results are negative, a phone conversation may be sufficient for follow-up. Transportation difficulty is a contributor to health inequity. Given the importance of the ability to return for changes in clinical status and further evaluations we recommend institutions consider travel vouchers (taxi or ride-share) for families with transportation insecurity. Telemedicine is increasingly being used for follow-up visits and may be appropriate in some situations.

If the reevaluation will be performed at another location or by a different clinical evaluator, it is recommended that the site for medical reevaluation be arranged in advance and clinician-to-clinician communication be direct. Clear written and documented instructions should be given to parents as to the time and place of the return visit.

1. the infant is clinically well or improving (eg, fever, feeding);

2. there are no other reasons for hospitalization; and

3. there is no other infection requiring treatment (eg, otitis media).

Evidence Quality: B; Strong Recommendation

In the most recent large studies, bacterial pathogens were not detected by 24 h in 15% to 18% and longer than 36 h in 5% to 7%; for CSF, the respective times were 11% to 18% and 6% to 15%.^138,139 Growth by 24 h occurred in a lower proportion of well-appearing infants with bacteremia (85%) compared with ill-appearing infants (93%).¹³⁸ 

1. infant is clinically well or improving (eg, fever, feeding) at time of reassessment;

2. all cultures are negative at 24 to 36 hours; and

3. there is no other infection requiring treatment (eg, otitis media).

Evidence Quality: B; Strong Recommendation

The following recommendations and options are for febrile (temperature >38.0°C), well-appearing, term infants 29 to 60 days of age without risk factors identified in the exclusion criteria.

Circumcised boys have a likelihood of UTI <1% and may be exempted from this recommendation.

Although the sensitivity of LE is not 100%, the rate of positive urine culture results without an abnormal urinalysis is roughly the same as the rate of asymptomatic bacteriuria and contamination. Moreover, renal scarring appears to be mediated by host WBCs rather than the presence of bacteria.

In one high-volume ED, limiting catheterizations to children with positive urine screen results from bag specimens reduced catheterization rates by more than half (63%–<30%) without increasing length of time in the facility or missing any UTIs.⁸⁵ Use of bladder-stimulation techniques⁸⁴ is more time-efficient than urine bag collection.⁸³ In newborn infants, bladder and lumbar stimulation was highly successful in facilitating midstream urine collection in a median time of 45 seconds.⁹⁰ Specimens obtained by methods other than catheterization or SPA are not suitable for culture because of a high contamination rate.^77,78 

The prevalence of bacteremia is lower than in the younger groups of infants but still high enough to warrant a blood culture (see Fig 4).

For detailed discussion of IMs, see KAS 10.

There is substantial evidence IMs are predictive of IBI including bacterial meningitis.^{10,–14,16,18,–20} For this age group, the number of meningitis cases in published studies is still relatively small, 64 cases in 25 917 febrile infants (0.25%). Data are unavailable comparing prevalence in IM-positive versus IM-negative infants, but decision rules and models that include IMs have sensitivities greater than 90%. In KAS 10, the committee provided data indicating that individual IMs are seldom sensitive or specific for detecting bacteremia or meningitis. However, individual values that are exceedingly high or low or finding several abnormal IMs should be considered in decision-making, because they, in all likelihood, increase the risk of bacterial meningitis.

The committee supports not performing an LP in well-appearing infants meeting the specified criteria. For an estimated prevalence of meningitis in 29- to 60-d-old infants of 0.25% and using a prediction rule or model with a sensitivity of 90%, the chance of missing a case of meningitis would be 0.025%. Therefore, 4000 successful LPs would be required to avoid a delay in the detection of 1 case of bacterial meningitis.

If no IM is abnormal, the committee does not include a positive urinalysis result as an indicator for performing an LP.

The antimicrobial agents in Table 3 are recommended for initial empirical therapy and should be modified following results of cultures and sensitivities.

If CSF is not available or is uninterpretable, clinicians should use parenteral antimicrobial agents.

1. CSF analysis (if CSF obtained) is normal; and

2. any IM obtained is abnormal.

Evidence Quality: B; Moderate Recommendation

If CSF is positive for enterovirus, clinicians may discontinue (or withhold) antimicrobial agents as long as there are no other factors suggesting a bacterial infection, including abnormal IMs.

1. CSF analysis (if CSF obtained) is normal;

2. urinalysis result is positive; and

3. no IM obtained is abnormal.

Evidence Quality: B; Strong Recommendation

1. CSF analysis, if CSF obtained, is normal or enterovirus-positive;

2. urinalysis is negative; and

3. no IM obtained is abnormal.

Evidence Quality: B; Moderate Recommendation

The risk for well-appearing infants with these negative findings having bacteremia is 0.1% for infants 29 to 60 days of age,¹⁸ with a CI upper limit that indicates the number needed to test is >300. Recent evidence documents the sensitivity of LE for UTI of 94% (95% CI, 91%–97%),⁸⁰ even higher in UTI associated with bacteremia (97.6% and 100%) in 2 studies^80,86; an NPV of 99% also supports a low likelihood of UTI.^38,–40 There are insufficient data to estimate precisely the risk of bacterial meningitis with normal CSF analysis, but, based on the scarcity of cases in the literature, the risk appears to be quite low.

Value Judgments: There were different thresholds, within the committee, for treating with antimicrobial agents. The potential benefits are highlighted above. The overall sense of the committee was to administer antimicrobial agents if the number needed to test for bacteremia is 100 or less: that is, willing to treat as many as 100 infants with parenteral antimicrobial agents to avoid delaying treatment in 1 infant with bacteremia. The committee recognizes that parents and practitioners have different levels of risk aversion and thresholds for treatment that should be incorporated into decision-making.

In a PECARN substudy of 29- to 60-d-old infants, an ANC > 4000 per mm³ and/or procalcitonin >0.5 ng/mL had a bacteremia prevalence of 3.2%; the prevalence if these IMs were negative was 0.2%.¹⁸ 

1. CSF analysis, if CSF obtained, is normal;

2. urinalysis is negative;

3. all IMs obtained are normal;

4. appropriate parental education has been provided;

5. follow-up plans for reevaluation in 24 hours have been developed and are in place; and

6. plans have been developed and are in place in case of change in clinical status, including means of communication between family and providers and access to emergency medical care.

Evidence Quality: B; Moderate Recommendation

Value judgments: The low risk of bacteremia and meningitis in infants without positive IMs can potentially reduce hospitalizations without compromising infant safety.

1. urinalysis is negative;

2. all IMs obtained are normal; and

3. parents can return promptly if there is a change in infant condition and agree to follow-up in 24 to 36 hours. Infants monitored at home should be reassessed in the following 24 hours.

Evidence Quality: B; Moderate Recommendation

Value judgments: The low risk of bacteremia and meningitis in infants without positive IMs can potentially reduce hospitalizations without compromising infant safety.

1. all bacterial cultures are negative at 24 to 36 hours;

2. infant is clinically well or improving (eg, fever, feeding); and

3. there is no other infection requiring treatment (eg, otitis media).

Evidence Quality: B; Strong Recommendation

1. blood culture is negative;

2. CSF culture, if CSF obtained, is negative;

3. infant is clinically well or improving (eg, fever, feeding); and

4. there are no other reasons for hospitalization.

Evidence Quality: B; Strong Recommendation

1. urine culture result is positive;

2. all other bacterial culture results are negative at 24 to 36 hours; and

3. infant is clinically well or improving (eg, fever, feeding).

Evidence Quality: B; Strong Recommendation

Many of the unanswered questions faced in the committee’s review emanated from the challenges of conducting prospective research in clinical settings with a relatively uncommon symptom. Fever in this age group has an incidence rate of 14 per 1000 term, previously healthy births per year.²² Although >10% of febrile infants will have UTIs, the likelihood of more IBIs is much less, with bacteremia detected in <2% of febrile infants and bacterial meningitis in <0.5%. Negative outcomes, such as permanent renal damage and organ damage or death, from sepsis are rare. Permanent neurologic sequelae from bacterial meningitis occur in variable rates depending on the severity of the infection, onset of treatment, and organism. Therefore, although use of administrative databases has recently provided important information, large, prospective studies will be required to answer a number of the following questions to further refine clinical recommendations for preventing negative outcomes.

All of the following pertain to well appearing febrile infants 8 to 60 days of age.

1. Because analyzing data for SBI has obscured understanding of optimal approaches to detect and manage individual infections, the term “SBI” should be retired and the incidence of the following infections determined separately: a. bacterial meningitis; b. bacteremia; and c. UTI.

2. The incidence of each individual infection can then be used to identify the most appropriate age groupings expressed in days rather than the arbitrary ones currently in use (weeks, months). The age groupings used in this guideline are primarily based on data gathered by week of age, as set a priori; although expressed here in days corresponding to those weeks, age groupings in the future should be derived from day-by-day data, which may generate different age groupings from the ones used here.

3. What is the morbidity and mortality of each infection for each age group?

4. What is the current epidemiology of each infection for each age group?

5. What is the best predictive rule for each infection?

6. What is the optimal initial choice and route of antimicrobial agents?

7. What is the optimal duration of therapy?

8. What are the predictors for bacteremia and for bacterial meningitis in a patient with a positive urinalysis result?

9. When does bacteremia matter in an infant with a UTI? Should bacteremia affect treatment duration?

10. In what ways do patients referred to EDs differ from patients initially seeking care in EDs and from patients seen in community practices, and should management differ accordingly?

11. What will be the impact of newer biomarkers and of genomic and other “omic” testing?

12. How should results of multiplex viral testing be incorporated into prediction models for IBI?

13. What is the best way to individualize care? Most guidelines seek to maximize care for the vast majority of patients while allowing for individualized judgments to incorporate certain circumstances. However, most guidelines sort on a small number of variables while most patients present with a vast number of relevant factors. Collaborative efforts that generate consistently acquired patient characteristics have an opportunity, using newer statistical techniques, to match a patient with a presenting symptom to others who most closely resemble the patient’s own background and clinical features. In this way, it would be possible to create an individualized guideline for each patient or “one patient, one guideline.”

14. Research to individualize care must include patient factors, including better understanding of the role of patient preferences, decision-making, perceptions of risk and vulnerability, satisfaction, and understanding of care.

15. What is the most effective way to provide ongoing monitoring and follow-up? The role of telehealth and differing systems of care approaches should be explored.

16. For low-risk infants, what impact will this guideline have on reducing the use of antimicrobial agents, decreasing invasive diagnostic testing, decreasing hospitalizations, and shortening hospital lengths of stay?

17. What is the impact of individual social determinants of health on risk of IBI, diagnostic testing, management, morbidity and mortality, discharge planning, and follow-up?

As a first step, questions 1, 2, and 5 could be partially answered by an effort to combine existing data sets from the large clinical and research groups publishing in this area. There are also international networks with similar foci on febrile infants. Although this would be challenging, it would still provide the shortest time to obtain the most accurate current assessment of risks.

It is clear that both the bacteriology and the technology involved in risk stratification and organism identification are evolving. Future research would benefit from a collaborative effort among researchers to define a common data set, with uniform definitions of elements and agreements to combine data for specific analyses. This effort could also lead to a model to answer question 10. As for question 12, it is now both methodologically and technologically feasible for a clinician to be able to enter a number of demographic, clinical, and laboratory data for a febrile infant and get the best estimate of risk for that patient.

Robert H. Pantell, MD, FAAP

Robert H. Pantell, MD, FAAP, Chair

Kenneth B. Roberts, MD, FAAP, Vice Chair

Charles R. Woods Jr, MD, MS, FAAP, Epidemiologist

William G. Adams, MD, FAAP

Carrie L. Byington, MD, FAAP

Benard P. Dreyer, MD, FAAP

Nathan Kuppermann, MD, MPH, FAAP, FACEP

Jane M. Lavelle, MD

Patricia S. Lye, MD, FAAP

Michelle L. Macy, MD, MS, FAAP

Flor M. Munoz, MD, MSc, FAAP

Carrie E. Nelson, MD, MS

Sean T. O’Leary, MD, MPH, FAAP

Stephen J. Pearson, MD, FAAP

Keith R. Powell, MD, FAAP

Jeb S. Teichman, MD, FAAP

Robert H. Pantell, MD, FAAP, Lead

Kenneth B. Roberts, MD, FAAP, Co-lead

William G. Adams, MD, FAAP

Benard P. Dreyer, MD, FAAP, Ex Officio

Nathan Kuppermann, MD, MPH, FAAP, FACEP

Sean T. O’Leary, MD, MPH, FAAP

Kymika Okechukwu, MPA

Kristin Ingstrup

Jeremiah Salmon, MPH

Vanessa Shorte, MPH

Caryn Davidson, MA

The committee acknowledges the generosity of individuals who graciously performed additional analyses from their published data sets and wisdom for this endeavor: Paul Aronson, MD, for the Febrile Young Infant Research Collaborative; Richard Bachur, MD (Division of Emergency Medicine, Boston Children’s Hospital); Carrie Byington, MD (University of Utah and Intermountain Healthcare); Borja Gomez, MD (Pediatric Emergency Department, Cruces University Hospital); Tara Greenhow, MD (Kaiser Permanente Northern California); Nate Kuppermann, MD, MPH (PECARN); and Matthew Pantell, MD, MS (PROS). We also thank Eric Biondi, MD, for leading a series of focus groups of primary care and subspecialty pediatricians who scrutinized the guideline and provided feedback on implementation. We especially recognize Borja Gomez, MD (Pediatric Emergency Department, Cruces University Hospital). In a truly collegial fashion, he regularly ran subanalyses for us on his previously published data that helped us fill in many gaps and provide a more refined set of recommendations. The following groups provided feedback and suggestions that were incorporated during the process of development: AAP committees: Committee on Fetus and Newborn, Committee on Hospital Care, Committee on Infectious Diseases, Committee on Medical Liability and Risk Management, Committee on Pediatric Emergency Medicine, and Committee on Practice and Ambulatory Medicine; AAP council(s): Council on Quality Improvement and Patient Safety; AAP sections: Section on Administration and Practice Management, Section on Critical Care, Section on Emergency Medicine, Section on Epidemiology, Public Health, and Evidence, Section on Hospital Medicine, and Section on Infectious Diseases; other AAP groups: Family Partnerships Network, PROS, Quality Improvement Innovation Networks; and external groups: American Academy of Family Physicians, American College of Emergency Physicians, and Pediatric Infectious Diseases Society.

Abbreviations

AAP

American Academy of Pediatrics

AHRQ

Agency for Healthcare Research and Quality

ANC

absolute neutrophil count

AUC

area under the curve

CI

confidence interval

CRP

C-reactive protein

CSF

cerebrospinal fluid

ED

emergency department

GBS

group B Streptococcus

HSV

herpes simplex virus

IBI

invasive bacterial infection

IM

inflammatory marker

KAS

key action statement

LE

leukocyte esterase

LP

lumbar puncture

NPV

negative predictive value

PCR

polymerase chain reaction

PECARN

Pediatric Emergency Care Applied Research Network

PROS

Pediatric Research in Office Settings

RBC

red blood cell

RSV

respiratory syncytial virus

SBI

serious bacterial illness

SPA

suprapubic aspiration

UTI

urinary tract infection

WBC

white blood cell