NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Hui C, Neto G, Tsertsvadze A, et al. Diagnosis and Management of Febrile Infants (0–3 Months). Rockville (MD): Agency for Healthcare Research and Quality (US); 2012 Mar. (Evidence Report/Technology Assessments, No. 205.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Diagnosis and Management of Febrile Infants (0–3 Months)

Diagnosis and Management of Febrile Infants (0–3 Months).

Show details


The synthesis of the literature on the diagnosis of serious bacterial illnesses (SBI) and invasive herpes simplex virus infection in infants less than 3 months of life has been challenging. In general, there is a lack of standard definitions in this field. Even simple issues such as what constitutes a fever or what should be included in the definition of SBI vary widely. The increase in precision of testing over time (i.e., aseptic meningitis vs. enteroviral meningitis), and emergence of different types of testing make it difficult to standardize the above mentioned definitions.

The evaluation of a patient is not always a one-time event and experienced clinicians value the ability to followup a patient over time for serial reassessment. Only the minority of studies report on reassessment and reassignment of the clinical status 26,57 and variable or no followup durations are reported. Additionally, only a fraction of studies reported to have employed lumbar puncture to diagnose bacterial meningitis. Similarly, majority of studies reporting the use of lumbar puncture did not employ this test on all included infants, thereby raising the possibility of incorrect test interpretations. Also, the vast majority of the studies did not report on long-term followup where partially treated meningitis might have been identified.

The heterogeneity of studies has precluded meta-analysis; therefore, simple summary statistics were not available except for the Rochester criteria and Philadelphia protocol. There was no clear difference in the study quality (QUADAS scores) between the studies reporting combined clinical and laboratory criteria such as Rochester, Boston, Philadelphia criteria/protocol and those reporting clinical or laboratory criteria alone. Moreover, the diagnostic test accuracy results for infants older than 3 months of age reported in some studies were not considered in this review (KQ1a). Such studies were included in the review only if they reported other relevant data (e.g., prevalence of SBI, outcomes related to management of febrile infants).

The clinical conundrum is how to balance the risk of missing an SBI (with potentially a devastating outcome) with the risks associated with diagnostic and management strategies. To date, a tremendous amount of resources and effort has been focused on developing tests, protocols, and criteria to attempt to minimize the first while almost ignoring the latter. The literature has revealed that the field of febrile infants less than 3 months is not homogenous and there are factors that either increase (i.e., neonatal age group) or decrease (i.e., viral syndrome) the risk and also the testing strategy accuracy. Equally heterogeneous are the risks associated with the specific types of SBI (e.g., urinary tract infections, bacteremia, and meningitis). A clinician fears the consequences of missing a case of meningitis much greater than missing a urinary tract infection; however the data are lacking to determine the accuracy of detecting the rarer and more devastating meningitis.

This systematic review has several strengths. We were able to calculate the test accuracy characteristics from raw data when possible; we provided test accuracy characteristics on the different types of SBI (UTI, bacteremia, meningitis) and for the neonatal period. To our knowledge, this is the first systematic review to seek the evidence on harms in the evaluation and management of febrile infants 0–3 months of age, to evaluate the role of viral infections or clinical bronchiolitis in prevalence of SBI, and to identify the factors that influence compliance in febrile infants or other infants with serious medical problems in infants 0–6 months of age.

KQ 1a and KQ 1b. In infants < 3 months old who present with a fever, what are the sensitivity, specificity and predictive values of individual or combinations of clinical features (history including information on the mother's history and previous testing, risk factors, findings on clinical exam, laboratory tests, and formal scoring instruments based on clinical features) for identifying those with serious bacterial illness (SBI)?
How do these findings vary by age within the age range 0–3 months?

The formal scoring instruments were the most evaluated standard approach used across multiple sites, with the Rochester criteria being tested the most. These criteria reported a higher sensitivity and negative predictive value compared with clinical criteria or only laboratory criteria. The use of other combined clinical and laboratory criteria also yielded high sensitivity and negative predictive values. Since these criteria were developed for having a high sensitivity, specificity across these criteria tended to be fairly low. There was a consistent trend of similar test characteristics between total SBI and bacteremia and meningitis across the criteria and various other tests, however reflecting the small numbers, the confidence intervals were large.

Generally, the studies evaluating clinical criteria alone revealed test characteristics (i.e., tendency towards higher specificity and lower sensitivity) that were not appreciably different from laboratory testing. Both types of tests demonstrated lower sensitivity and higher specificity values compared to combined criteria. Therefore, clinical or laboratory criteria alone may have limited ability to rule out the presence of SBI.

The pooled sensitivity estimates for the Rochester criteria and Philadelphia protocol were similar. Given the similar test accuracy between the two criteria, attention should be paid to the differences. Where the Philadelphia protocol requires the evaluation of the cerebrospinal fluid by lumbar puncture and a chest x-ray in the 1-2 month old group to define the low risk group, the Rochester criteria identifies the low risk group in 0 to 2 month old infants without using LP and CXR.

The neonatal period (0–28 days of life) was shown to have a higher prevalence of SBI compared with older children. When separately evaluated, neonates did not have the same test characteristics as the older children or whole group of less than 3 months of age. In only one study evaluating the Rochester criteria in neonates the testing in the neonatal age group showed better numerical accuracy than in the older age group. The rest of the combined, laboratory or clinical criteria demonstrated lower sensitivity in correctly identifying the presence of SBI in the neonates than in older groups of infants.

There remains controversy about the need for lumbar puncture in infants with fever. In our review, six studies reported to have misclassified eight (out of 42) cases of meningitis into low-risk for SBI (total number of meningitis were reported only in fives studies). Using the Rochester criteria (four missed cases), a data-derived model of combined clinical and laboratory (one missed case), clinical only (one missed case), and a laboratory test (two missed cases). None of these criteria included a lumbar puncture and CSF analysis. Our review does not answer the question of whether a lumbar puncture is required in all febrile infants or what parameter can predict for the need for a lumbar puncture.

KQ 1c. In infants < 3 months old who present with a fever, what are the sensitivity, specificity and predictive values of individual or combinations of clinical features (history including information on the mother's history and previous testing, risk factors, findings on clinical exam, laboratory tests, and formal scoring instruments based on clinical features) for identifying those with herpes simplex virus infection (HSV)? How do these findings vary by age within the age range 0 to 3 months?

Little evidence on invasive herpes simplex virus infection in febrile infants included in this systematic review does not indicate the lack of clinical cases. The literature mainly focused on the end diagnosis of HSV,129 rather than the clinical syndrome of a febrile infant. When invasive herpes simplex virus infection is reported in large series of febrile infants, the numbers are very small. In a recent study, Caviness et al. reported that during the season, enterovirus infection was 20 times more likely and a serious bacterial illness was 23 times more likely to occur in hospitalized febrile neonates as compared with HSV in febrile infants.63 This lack of evidence is likely due to the fact that HSV does not routinely present with fever (3/10 in Caviness 2008 study63) and the fact that herpes simplex infection due to skin eyes and mouth infection were likely excluded from other studies as they represent a focus of infection. Given the lack of evidence of diagnostic accuracy and the inability to target and adequately screen mothers,130,131 we are left only with expert opinion.132

KQ 2a. What is the evidence that clinical features alone, basic laboratory tests alone or the combination are sufficient to identify febrile infants < 3 months who are at low risk of having a serious bacterial illness (i.e., have a high negative predictive value)?

There were several studies that used clinical and laboratory criteria to identify infants at low risk for SBI. The first study to use this approach used the Rochester criteria and showed a high negative predictive value of 99.3 percent in infants < 3 months of age.27 Other studies that used the Rochester criteria showed similar negative predictive values (93.7–99.2 percent). The other low risk criteria (Philadelphia, Boston, and Milwaukee) also had high NPV for SBI indicating relatively low proportion of missed SBI cases in these studies (10.0 percent or less).

The prevalence of SBI in the low risk group is about 1.0–2.0 percent compared to the prevalence of ∼10.0 percent overall. Low risk criteria can identify infants unlikely to have SBI and who can be managed less aggressively. We found no information on variability among clinicians in terms of competence in assessing risk.

Infants < 1 month of age have been treated differently based on a higher baseline risk of SBI and the difficulty of clinical assessment. Several studies have shown that although the overall risk of SBI is higher, the Rochester criteria may be able to identify low risk infants in this age group. The negative predictive values for the Rochester criteria in this younger age group were 93.0–97.0 percent. The prevalence of SBI in the low risk group of neonates is 3.0–5.0 percent. The low risk criteria can identify infants < 1 month old who are unlikely to have SBI but a small number of infants with SBI will be missed. Although many studies had high negative predictive values, these should be interpreted with caution as predictive values vary based on prevalence.

KQ 2b. What is the evidence for the potential risks resulting from a delay in the diagnosis and treatment of patients who appear low risk but have a serious bacterial illness?

There is very little evidence on the risks of delay in diagnosis and management in low risk infants who were later found to have SBI. Several studies reported the outcomes of infants in which the diagnosis of SBI was initially missed. Most infants were subsequently hospitalized and treated with antibiotics. Although somewhat reassuring, the fact that there were no adverse outcomes in these infants may have been due to underreporting and/or lack of followup in these studies. Of note, several of the studies reported on contaminated urine, blood and cerebrospinal fluid cultures. The added management and harms associated with these false positive results were not reported.

Indirect evidence comes from the PROS research network febrile infant study. In the office based study by Pantell,5 many practitioners did not adhere to the most conservative approach for management of febrile infants (i.e., full sepsis workup on each febrile infant <3 months of age presenting to physician's office). In this study, only 54.0 percent of the infants had a urinalysis and 24.0 percent had no testing of blood, urine or cerebrospinal fluid. The prevalence of UTI and bacteremia in the overall group were 5.4 percent and 1.8 percent, respectively; in the infants that actually had testing, these rates were 9.7 percent and 2.4 percent, respectively. It is possible that cases of SBI were missed because many infants had no investigations. There were no adverse outcomes observed. It should be noted that many infants had either office or telephone followup which may enable the practitioner to have a less aggressive management approach. Note that most of the results of this study are based on suburban setting and may not accurately reflect the febrile infant risks present in the primary care urban settings.

Additionally, Newman reported results on UTIs from the same study.106 It was modeled so that in the 807 infants not initially tested or treated with antibiotics, there should have been 61 UTIs based on predictors of UTIs, whereas only two cases were diagnosed at followup. No adverse outcomes were reported with office and telephone followup, suggesting that some acute UTIs may have spontaneously resolved. The study was not designed to look at the long-term renal function of these patients; the findings of this study do not support the concern that all untreated UTIs lead to bacteremia.

The low risk criteria have been used in practice for over 10 years and yet there is minimal data on the morbidity and mortality of infants with SBI who are missed by the low risk criteria. As the literature and field has been focused exclusively on avoiding missed SBI, the consequences of iatrogenic harms have not been evaluated. To truly balance the risks and benefits of management strategies, the risks need to be fully delineated.

KQ3a. What is the evidence that clinical features alone, basic laboratory tests alone or the combination are sufficient to identify febrile infants < 3 months who are at high risk of having a serious bacterial illness (i.e., have a high positive predictive value)?

A confusing aspect of the literature on SBI in febrile infants is the focus on either identifying high risk patients or identifying low risk patients. It is important that studies reporting on indentifying low risk infants emphasize that that infants not meeting the low risk criteria are not necessarily high risk, and therefore are more appropriate to be labeled as not low risk.

In general, most studies demonstrated higher sensitivity and lower specificity. The low PPV values reported for the selected combined criteria (e.g., Boston, Milwaukee, Philadelphia, Rochester) are indicative of high false positive rates for SBI (i.e., high proportion of febrile infants without SBI classified as high risk). Lower PPVs for bacteremia and meningitis compared to PPVs for SBI are reflective of lower prevalence of the former among febrile infants 0–3 months of age.

There is little reported evidence on what factors are associated with variations in practice patterns among different individual providers. In the absence of better data on harms and the costs of diagnostics and therapeutics or improved positive predictive values, many clinicians will continue to opt to “over–treat” a large group of SBI negative patients.

KQ 3b. What are the benefits and harms of immediate antibacterial, antiviral therapy, and/or hospitalization (vs. delaying until diagnostic workup is complete) in patients at high risk of serious bacterial illness?

The realm of this question should encompass medical harms as well as cost associated with immediate hospitalization and treatment with antibiotics. Additionally, the psychological harms of the testing have not been explicitly stated in the studies. Unnecessary testing may have had the unexpected consequence of the parents viewing the infant as more fragile or have more anxiety around the chance of a serious bacterial infection although the literature has not delineated the presence or absence of such factors.

Byington and Paxton reported on a survey of parents of infants undergoing a ‘rule-out sepsis’ (ROS) evaluation months after admission. The majority of the sixty parents who interviewed reported finding the ROS evaluation very stressful with 28.0 percent believing their infant was to die. Additionally, 36.0 percent of mothers reported breastfeeding problems with 18.0 percent stopping breastfeeding, 35.0 percent perceived their child to be less healthy on followup, 43.0 percent reported financial stress, and 33.0 percent reported perceived iatrogenic problems.133

Overall, the reporting of harms for this area has been very poor.

KQ 4. What is the evidence that the presence of an identified viral infection predicts against a serious bacterial infection?

There seems to be some confusion surrounding SBI evaluation in a child with a recognized viral syndrome. 134-136 Advent of rapid testing for viral pathogens has given many clinicians the ability to diagnose viral infections in children less than 3 months of age. This review has shown a consistent statistically significant inverse relationship between viral testing positive or clinical bronchiolitis and the presence of SBI among infants with fever. Most of the SBI were UTI, although there were some cases of bacteremia, but no meningitis.

For the clinician in an office or with no access to rapid viral testing a clinical diagnosis is more applicable. Some of the studies enrolled patients with positive viral culture results due to their retrospective nature. As rapid antigen testing is not 100.0 percent sensitive (87.0 percent),86 some patients with rapid testing negative, who subsequently have viral culture that is positive would not benefit from this information. More concerning is the issue of false positives, where 3/135 were rapid testing positive, but viral culture negative, thereby providing the clinician with a false sense of security and potentially mislabeling the patient as being low-risk for SBI.86 This area is further confused by the development of PCR testing which is more sensitive than the previous “gold standard” of viral culture.

However, even in the absence of rapid testing, the clinician is able to obtain significant information that decreases the chance of SBI. The three studies by Lubinghl, Bilavsky and Kuppermann demonstrate a similar inverse relationship between clinically diagnosed bronchiolitis and SBI.

The sample sizes of the studies did not answer the clinical dilemma regarding the need for lumbar puncture in infants with clinical bronchiolitis or a positive viral test. The study by Levine et al. reported zero cases of meningitis in the RSV positive group. Although bacteremia and meningitis were lower in the RSV positive group, the difference did not reach the statistical significance. Luginbuhl's publication for the PROS study also could not answer this question due to the sample size (only 35 patients [16.0 percent] of the bronchiolitis group had a lumbar puncture). Bilavsky et al. reported no cases of meningitis or bacteremia in the bronchiolitis group compared with one and four in the no-bronchiolitis group respectively. The rarity of the entity of bacterial meningitis in RSV or bronchiolitis positive patients likely means that this question will not be answered without an enormous effort. Indeed, only a few cases of meningitis have been described in the literature in febrile infants with a viral infection.136,137

The lack of reporting of the age-specific sub-groups does not provide information on whether the group of infants 0-3 months of age is homogenous in terms of risk of SBI in the bronchiolitis or virus positive patient. However, given our understanding that the prevalence of SBI is significantly higher in the 0-28 day group and that the diagnostic tests differ in their accuracy in this age group compared with the entire 0-3 month group, it seems logical that the neonatal time period should be viewed differently. This is supported by the only study by Levine et al. that provided the prevalence of specific SBI in neonates and demonstrated no significant difference in prevalence of SBI between patients with and without proven RSV.

Overall, evidence in this review indicates that bronchiolitis or a positive result for a virus significantly predicts against SBI. The majority of cases of SBI were UTI. Caution should be used when evaluating neonates with these findings as the presence of bronchiolitis or virus in this sub-population may not be as predictive against an SBI as in older groups of infants.

KQ 5. What is the evidence that the prevalence of SBI varies among febrile infants presenting to primary care and emergency practice? What is the evidence that prevalence affects the predictive value of clinical and laboratory findings?

The majority of studies were conducted in an emergency department setting. The reported prevalence of SBI in North American emergency department settings varied from 4.0 percent to 25.0 percent. The prevalence of SBI in the primary care studies varied about 9.0 percent-10.0 percent. There appears to be a somewhat higher prevalence of SBI in the emergency department population. The difference in prevalence may reflect a difference in the patient population that seeks care in the emergency department. The patients seen in the emergency department may be a sicker group than those who those who wait to see their primary care provider.

There is considerable practice variation between emergency department and office settings. These differences bring up the following questions: in the emergency department, are infants being over-investigated or does this reflect a difference in their level of acuity? In office practice, are infants with bacterial infections being missed and is there any associated morbidity and long-term consequences?

Given the low prevalence of serious bacterial illness and very low prevalence of bacteremia and meningitis, many clinicians, especially physicians who evaluate low volumes of febrile infants less than 3 months of age, may never see a significant adverse outcome regardless of what their practice of diagnostics and management is. This may provide a “false” sense of security that the clinician is correctly managing these infants. Conversely, habitually adhering to more rigorous protocols for diagnosis and treatment may instill a “false” belief that they are necessary. The small numbers of bacteremia and meningitis in all the cited studies do not allow an accurate experienced-based understanding of the accuracy of current testing strategies for these more serious outcomes.

KQ6. Clinicians base decisions about initial diagnostic work-up and treatment of febrile infants not solely on the infants' medical status but also on their assessments of non-clinical factors (e.g., parental understanding, parents' ability to monitor the patient, access to care). A strategy of initial observation without extensive diagnostic tests or hospitalization depends on confidence that parents will reliably bring the baby back for a timely follow-up appointment if conditions warrant. How likely are parents whose infants are less than six months of age and have fever or other potentially serious medical condition to comply with a provider's recommendation that the parent bring the infant back (to that provider or another) for a return appointment to re-assess the condition(s) of concern?

KQ6a and KQ6b. What is the evidence that identifiable parental factors (e.g., education, insurance status, living situation, history of previous visits with the provider, time/distance required to travel to an appointment, etc.) allow a provider to judge the likelihood that a parent will adhere to treatment recommendations such as returning for followup if circumstances warrant?
What is the evidence that the clinical setting (community practice vs. emergency department and/or hospital outpatient clinic) in which care is sought independently influences the likelihood of compliance with a return appointment?

The dearth of studies in this area led us to expand our inclusion criteria to up to 6 months of age and to include infants with fever or other potentially serious medical condition. The lack of focus in this area is evidenced by the identification of only four studies with this expanded inclusion criteria. Although the followup was reported in these studies, they were not the primary focus. The high rate of success for outpatient therapy and telephone followup in these studies could in part be explained by the increased motivation of parents whose infants were enrolled in the studies.

Follow up and reassessment of the febrile infant is an important component of their care. A clinician's decision making can be highly influenced by his/her assessment that the patient's caregivers are likely to comply with followup or further testing. Very little is known about the factors that affect compliance for follow up in this population and it is an area where more research is needed.

There was a lack of evidence regarding the influence of parental factors (e.g., age, education, distance/time to travel to an appointment, living situation) or clinical settings (emergency department vs. primary care office) on rates of parental compliance to telephone or return visit followups.

Although there were no included studies in this review on parental factors or clinical setting influencing followup, a review of the broader literature reveals some potential factors that need to be further studied in the 0-3 month febrile infant population. In some studies Hispanic patients were less likely to comply with followup. The other identified parental factors such as lack of parental ability to speak English, having to make their own appointment, self-pay, lack of a primary care provider and followup greater than 24 hours seem self-evident but require further study.


Overall, the focus of the literature has been on ruling out SBI. Harms associated with testing or management strategies have been poorly reported. Attempts to identify high risk groups, as described in the minority of reports, were not accurate. The Boston, Philadelphia, Rochester, and Milwaukee were fairly accurate in identifying a low risk group for SBI in infants younger than 3 months of age. The diagnosis of a viral infection or clinical bronchiolitis significantly decreased the chances of a serious bacterial illness. Invasive herpes simplex virus infection is a significant differential diagnosis in the febrile infant, yet the relevant literature is presented from the diagnosis rather than from the syndrome point of view, making it difficult to draw conclusions of test accuracy or management efficacy in an undifferentiated febrile infant. Although crucial to the management strategies in the low risk group, there is sparse literature on factors associated with compliance in this population. Future studies should focus on identifying the risks associated with testing and observation strategies and on factors that influence compliance to followup care.

Research Needs and Future Directions

To move the field further, there is a need to delineate the risks associated with testing and also management options in this group. Rigorous studies need to be done with separate reporting for infants 0-28 days, 1-2 months, 2-3 months of age (see QUADAS in Appendix G).

Consideration should be given not to include the 0-6 day old group as these infants likely represent another clinical syndrome of early onset sepsis related to perinatal factors. Most clinicians when faced with a febrile infant 3 days of age would perform a full evaluation including lumbar puncture and admit the infant for intravenous antibiotics. The focus should be on the clinical conundrum of febrile infants with no apparent source of infection.

The group of low-risk patients needs to be determined by incorporating risks associated with age group and viral or clinical syndrome status and observed as outpatient or inpatient at followup. Detailed reporting of the harms associated with the diagnosis and in/outpatient observation of this low-risk group would be crucial. This should include management changes associated with contaminated specimens, parental anxiety, breastfeeding cessation, a long-term concern over “vulnerable child syndrome,” and financial costs. The outcomes should not only be numbers of SBI, but followup should be done to determine the long-term consequences of ‘missed’ or ‘delayed’ diagnosis of SBI such as decreased renal function with urinary tract infection, progression from UTI to bacteremia, and complications of meningitis.

Efforts of future research should also be directed towards the development of new constellations of those clinical and laboratory criteria that were shown in previous research to have better sensitivity for various sub-types of SBI (e.g., bacteremia, meningitis).

Integrated into these studies should be evaluations on the factors or interventions that increase parental compliance and/or clinicians' ability to predict compliance. Optimally, these studies should be multi-centered and evaluate both outpatient and emergency department settings. Once there are better data on harms of diagnostic and observation protocols a consensus expert panel could be struck to define the risk balance.

Furthermore, a registry or surveillance network should be developed to describe the changing pathogens and resistance over time as there is increasing concern over these shifts and their potential clinical significance.81,93,138 Effects of vaccination and other interventions may be appropriately studied on a population basis. Although the majority of the SBI cases are due to gram negative UTIs, bacteremia occurred predominantly due to Streptococcus pneumoniae, Group B Streptococcus and Haemophilus influenzae type b – all pathogens that have changed significantly over time due to vaccinations or interventions. Additionally, there is some evidence of resistance and different pathogens in more recent studies.93,138,139


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (2.2M)

In this Page

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...