Objective. The risk of serious bacterial infection (SBI) in febrile infants who are classified as low risk (LR) or high risk (HR) by the Rochester criteria has been established. LR infants average a 1.4% occurrence of SBI, whereas HR infants have an occurrence of 21%. The occurrence of SBI in Rochester LR or HR infants with confirmed viral infections is unknown. The objective of this study was to determine the occurrence of SBI in Rochester LR and HR infants with and without viral infections.
Methods. All febrile infants who were 90 days or younger and evaluated at Primary Children’s Medical Center between December 1996 and June 2002 were eligible. Infants were classified as Rochester LR or HR, and discharge diagnoses were collected. Viral testing for enteroviruses, respiratory viruses, rotavirus, and herpesvirus was performed as indicated by study protocol, clinical presentation, and season of the year. Results of all bacterial cultures were reviewed.
Results. Of 1779 infants enrolled, 1385 (78%) had some form of viral diagnostic testing and 491 (35%) had 1 or more viruses identified. By the Rochester criteria, 456 (33%) infants were classified as LR and 922 (67%) infants as HR. For infants with viral infections, the occurrence of SBI was significantly lower than in infants without a viral infection (4.2% vs 12.3%). Rochester HR virus-positive (HR+) infants had significantly fewer bacterial infections than HR virus-negative (HR−) infants (5.5% vs 16.7%). When compared with HR− infants, HR+ infants were less likely to have bacteremia, urinary tract infection, or soft tissue infections, and HR+ infants had a similar occurrence of bacteremia as LR infants (0.92% vs 1.97%).
Conclusions. Febrile infants with confirmed viral infections are at lower risk for SBI than those in whom a viral infection is not identified. Viral diagnostic data can positively contribute to the management of febrile infants, especially those who are classified as HR.
Approximately 8.5% of infants who are younger than 90 days and have a temperature ≥38°C will have a serious bacterial infection (SBI).1 The remaining infants are presumed to have a viral cause for fever. Identifying individual febrile infants who are most likely to have SBI continues to challenge clinicians. The most widely accepted approach has been to use a combination of historical, physical, and laboratory findings known as the Rochester criteria to identify infants who are at low risk (LR) or high risk (HR) for SBI.2–5 A 1992 meta-analysis of 14 studies of febrile infants performed between 1977 and 1990 found that infants who were classified as LR had an occurrence of SBI of 1.4%, whereas those who were classified as HR had an occurrence of SBI of 21%.6
The role of viral diagnostic studies in the management of febrile infants has not been investigated adequately. Only 2 of the studies in the above meta-analysis included any viral diagnostic testing.2,7 The ability to diagnose viral infections has improved substantially in recent years, and because most febrile infants are presumed to have viral infections, making a specific diagnosis could contribute significantly to their treatment.
An important consideration, however, is whether febrile infants with viral illnesses could have a concomitant SBI. Several studies of infants and children up to 3 years old with respiratory syncytial virus (RSV) infection have shown a concomitant SBI in 0% to 1.9%.8–12 Similarly, 2.5% to 6.7% of infants who were younger than 90 days and had enteroviral (EV) infections were found to have concomitant SBI.2,13 There is, however, no information in the medical literature regarding the risk of SBI in febrile infants in whom Rochester classification has been determined and in whom a viral infection has been definitively diagnosed.
We hypothesized that infants with viral illness, even those classified as HR, would have a lower occurrence of SBI than those without viral infection. To test this hypothesis, we designed a study to determine the occurrence of SBI in Rochester LR and HR infants with and without confirmed viral infections.
Approval to conduct this study was granted by the Institutional Review Boards of the University of Utah and Primary Children’s Medical Center (PCMC) in Salt Lake City, Utah.
Selection of Participants
Febrile infants who were 1 to 90 days old and evaluated for sepsis at PCMC were enrolled in a prospective study between December 1996 and June 2002. All patients had been discharged from the hospital after birth and presented to PCMC for the evaluation of fever of ≥38°C. All had an evaluation for sepsis including bacterial cultures of blood, urine, and cerebrospinal fluid (CSF). Infants with clinical findings of abscess, cellulitis, gastroenteritis, or pneumonia may have had additional bacterial cultures or chest radiographs obtained. Infants were excluded when they had received antibiotics in the 48 hours preceding the evaluation. Infants were also excluded when they had received the oral polio vaccine, a live EV vaccine. Patients who were evaluated from December 1996 to June 1999 all were admitted to PCMC, whereas patients who were evaluated after July 1999 include both inpatients and outpatients.
Identification of Viral Infections
Infants were tested for a variety of viral infections depending on study protocol, clinical findings, and season of the year. The original study protocol was to test all infants for EV infection by the polymerase chain reaction (PCR). Beginning in 1999, all infants who were enrolled during the winter months (November-April) were tested for respiratory viruses, including RSV; influenza A and B; parainfluenza 1, 2, and 3; and adenovirus if parents gave consent. In addition to viral testing performed as part of a study protocol, the results of all viral diagnostic testing performed as part of routine medical care were collected.
EV infections were diagnosed by the PCR (EV Amplicor; Roche Molecular Systems, Branchburg, NJ, for 1996–2000; ARUP EV-RT PCR, for 2000–200214) or by culture. Specimens for EV PCR included blood and CSF. Specimens for EV culture included CSF, stool, and nasopharyngeal and throat swabs. Respiratory viral infections were identified by enzyme-linked immunosorbent assay (VIDAS RSV ELISA, BioMerieux, Hazelwood, MO, for 1996–1998), by PCR (Hexaplex; Prodesse, Milwaukee, WI, for 1999), or by direct fluorescent assay detection (SimulFluor Respiratory Screen, Light Diagnostics, Chemicon International, Temecula, CA, for 2000–2002) performed on nasal wash specimens. Rotavirus infections were diagnosed by enzyme immunoassay performed on stool specimens (Premier Rotaclone, Meridian Diagnostics, Cincinnati, OH). Herpes simplex virus infection was diagnosed by culture of skin lesions or mucous membranes. The diagnosis of varicella infection in a single infant was made on the basis of exposure history (mother and sibling with varicella) and a vesicular skin rash consistent with varicella.
All infants were classified as either Rochester LR or HR on the basis of historical, clinical, and laboratory findings using a standard classification form. Infants were classified by trained research nurses (n = 2), and the same nurses classified infants throughout the study. Infants were classified at the time of enrollment, before the results of bacterial cultures were known. Infants with missing data were classified as indeterminate. To be classified as LR, an infant had to have all data available.
Determination of SBI and Discharge Diagnosis
The results of all bacterial cultures were obtained from computerized microbiology records. Infants with positive bacterial cultures were classified as having SBI or contaminated cultures. SBI was defined as bacteremia, bacterial meningitis, urinary tract infection (UTI), soft tissue or bone infection, bacterial pneumonia, or bacterial enteritis. This designation of SBI matches that published in the 1992 meta-analysis regarding the probability of bacterial infections in febrile infants.6 Bacterial isolates such as Staphylococcus epidermidis or Viridans group Streptococcus were considered contaminants unless they were isolated from 2 or more bacterial cultures. Discharge diagnoses were collected, and each chart was reviewed for physical examination, radiographic and surgical findings, and antibiotic treatment to confirm the presence of SBI. A single infant with disseminated Aspergillus infection was included in the infants with SBI.
Data were stored and sorted using Microsoft Access (Microsoft Seattle, WA.). Epi-Info Version 6.04 (Stone Mountain, GA) was used for statistical analysis. Infants were divided into groups determined by Rochester classification and viral status and compared using Mantel-Haenszel, χ2, or the Fisher 2-tail exact test. Confidence intervals for proportions were determined using the exact binomial method.
Of the 1779 infants who were enrolled in the study, 1385 (78%) had some form of viral diagnostic testing. The most important reasons that infants did not have viral testing performed were that parents did not give informed consent or insufficient specimen was available for testing. The infants who had viral testing performed compose the study group. Demographic information is given in Table 1. The mean age of infants with viral infection identified was not different from those without viral infection (36 vs 34 days).
All infants were tested for at least 1 virus. The number of infants tested for individual viruses is given in Table 2. Of the 1385, 491 (35%) were found to have 1 or more viral infections and 894 had negative results. The viruses identified are given in Table 2.
All 1385 infants with viral testing had a Rochester classification assigned. There were 922 (67%) HR infants, 456 LR (33%) infants, and 7 (<1%) indeterminate infants.
Determination of SBI on the Basis of Viral Status
Of the 1385 infants, 131 (9.5%) were identified with SBI. The most common pathogens identified were Gram-negative enteric organisms such as Escherichia coli and Klebsiella species.15 Three infants had SBI caused by coagulase-negative staphylococci or Viridans group Streptococcus (2 UTI and 1 bacteremia with meningitis).
The occurrence of SBI in infants with viral infections was 21 (4.2%) of 491 compared with 110 (12.3%) of 894 in infants without a confirmed viral infection (P = .0001). The occurrence of bacteremia in infants with viral infections was 5 (1%) of 491 compared with 24 (2.7%) of 894 in infants without an identified viral infection (P = .038). No cases of bacterial meningitis were detected in infants with viral infections; 6 (0.67%) cases were identified in infants without viral infections.
Determination of SBI on the Basis of Rochester Classification and Viral Status
The occurrence of SBI for Rochester LR and HR infants with and without viral infection is given in Table 3. The occurrence of SBI in LR infants was 2.6%. The occurrence of SBI in LR virus-positive (LR+) infants was not statistically different from LR virus-negative (LR−) infants (1.7% vs 3.1%; P = .4). The occurrence of SBI in HR infants was 12.8%. However, HR virus-positive (HR+) infants, were significantly less likely to have SBI than HR virus-negative (HR−) infants (5.5% vs 16.7%; P < .0001). HR− infants were 3.4 times more likely to have SBI than HR+ infants (95% confidence interval: 2.0–5.6). When compared with LR+ infants as a reference, there is an increase in the occurrence of SBI for each group, with HR− infants 13.67 times more likely to have SBI.
The occurrence of individual types of SBI in HR infants with and without viral infection is given in Table 4. HR+ infants were significantly less likely to have bacteremia, UTI, and soft tissue infection when compared with HR− infants. When compared with LR infants, HR+ infants had the same occurrence of bacteremia (0.92% vs 1.97%; P = .24) but higher occurrence of UTI (3.7% vs 1%; P = .002).
Our data suggest that the use of viral diagnostic studies in combination with the Rochester criteria can substantially improve the ability of clinicians to stratify infants according to their risk of SBI. Categorizing HR infants into those who are virus positive and virus negative revealed an occurrence of SBI of 5.5% in those who were HR+ compared with 16.7% in those who were HR−. In addition, the use of viral diagnostics allowed the identification of a subset of HR infants whose risk of bacteremia is ∼1%, significantly less than that reported in previous studies and similar to the risk in LR infants.6
Overall, 35% of the infants who were enrolled in the study had a viral illness identified. The most common viruses identified were enteroviruses and RSV. These 2 groups of viruses represent important pathogens that cause fever in infants younger than 90 days.2,13,16,17 Data from our institution and others indicate that enteroviruses are identified in ∼25% of infants who are admitted for an evaluation for fever and up to 50% in those who are evaluated during the summer and fall.13,16 RSV causes predictable seasonal epidemics, primarily in the winter months in temperate climates.18 Both infections can be diagnosed rapidly given available technology.19,20
The most important findings of this study relate to the occurrence of SBI in infants with and without viral illnesses. The occurrence of SBI in infants with a viral infection was significantly lower than in infants without identified viral infections (4.2% vs 12.3%). Infants with viral infection were also significantly less likely to have bacteremia (1% vs 2.7%). When the occurrence of SBI is analyzed further on the basis of Rochester classification and viral status, the results indicate that viral testing is especially helpful in infants who are classified as HR.
Infants who were classified as HR in the present study had an occurrence of SBI of 12.8%. This is significantly lower than the 21% reported in studies done between 1977 and 1990.6 During the 1990s, a dramatic decrease in the occurrence of bacteremia in infants has been documented as a result of the widespread immunization against Haemophilus influenzae type b and prophylaxis against group B Streptococcus.21,22 However, even in this era of decreasing SBI, striking differences in rates of occurrence of SBI could be demonstrated between HR+ and HR− infants, indicating that viral testing can contribute to patient treatment.
The addition of selective rapid viral diagnostics to the Rochester criteria could allow physicians to better categorize infants regarding individual risk for SBI. Infants who are HR+ have a lower risk of SBI in general and specifically a lower occurrence of bacteremia, UTI, and soft tissue infection. Conversely, those who are HR− are nearly 3.5 times more likely to have SBI.
Clinically, bacteremia is the most difficult SBI to predict because there are no physical or radiographic findings or screening laboratory tests that reliably identify infants with bacteremia at their initial evaluation. Infants with no focus for fever are often treated with antibiotics to protect those who might be bacteremic. On the basis of this study, the knowledge that an HR infant has a viral illness confers a risk for bacteremia comparable to that of LR infants in general, the majority of whom are now treated as outpatients, many without antibiotics.23
Two important limitations of this study are that not all febrile infants had complete viral studies performed and that testing for potentially important viral pathogens such as human herpesvirus 624 and rhinoviruses was not performed. Therefore, the true prevalence of viral infection in this population of febrile infants cannot be stated. It is likely, however, that our study underestimates the number of febrile infants who had viral infections. The differences in occurrence of SBI between the 2 groups likely would have been even greater if complete identification of viral infections had been possible. More data are needed regarding the risk of SBI with specific viral illnesses. The risk for bacteremia seems low (< 1%) in infants with RSV infection in this study and others9,10,25 but may be higher for infants with enterovirus or parainfluenza infections.
Our data suggest that HR+ infants have a moderate risk of SBI, approximating 5%. Viral testing, especially for enteroviruses, influenza A and B viruses, rotavirus, and RSV, should be performed in season as a routine part of the sepsis evaluation in all HR infants with fever and or clinical findings such as bronchiolitis or diarrhea. If the results of viral diagnostics are available in the first 24 hours of hospitalization, then they can be combined with 24-hour bacterial culture results to optimize treatment plans for individual patients.26 Recent data regarding bacterial culture results in febrile infants indicate that ∼90% of bacterial pathogens from patients who are evaluated in the outpatient setting are identified within the first 24 hours of incubation.27–29 If these data can be confirmed by others, then they should provide physicians with assurance that hospitalized virus-positive infants with negative bacterial cultures after 24 hours do not need additional antibiotic treatment, and many may no longer require hospitalization.
Dr Byington is supported by the Robert Wood Johnson Generalist Physician Faculty Scholar Program. This investigation was supported by Public Health Services research grant M01-RR00064 from the National Center for Research Resources.
We thank Karen Osborne, RN, Shawna Baker, RN, and Dixie Thompson, RN, for assistance with the study. We thank Paul C. Young for review of the manuscript.
- Received July 30, 2003.
- Accepted November 5, 2003.
- Reprint requests to (C.L.B.) Department of Pediatrics, University of Utah Health Sciences Center, 50 North Medical Dr, Salt Lake City, UT 84132. E-mail:
Portions of this study were presented at the annual meeting of the Pediatric Academic Societies; May 5, 2003; Seattle, WA.
Current addresses: Richard Tuohy, University of Minnesota, Minneapolis, MN; Karen C. Carroll, MD, Johns Hopkins University, Baltimore, MD; John C. Christenson, MD, University of Indiana, Indianapolis, IN.
- ↵Jaskiewicz JA, McCarthy CA, Richardson AC, et al. Febrile infants at low risk for serious bacterial infection—an appraisal of the Rochester criteria and implications for management. Febrile Infant Collaborative Study Group. Pediatrics.1994;94 :390– 396
- ↵Broner CW, Polk SA, Sherman JM. Febrile infants less than eight weeks old. Predictors of infection. Clin Pediatr.1990;29 :438– 443
- ↵Byington CL, Taggart EW, Carroll KC, Hillyard DR. A polymerase chain reaction-based epidemiologic investigation of the incidence of nonpolio enteroviral infections in febrile and afebrile infants 90 days and younger. Pediatrics.1999;103(3) . Available at: www.pediatrics.org/cgi/content/full/103/3/e27
- ↵Byington CL, Rittichier KK, Bassett KE, et al. Serious bacterial infections in febrile infants younger than 90 days of age: the importance of ampicillin-resistant pathogens. Pediatrics.2003;111 :964– 968
- ↵Verboon-Maciolek MA, Nijhuis M, van Loon AM, et al. Diagnosis of enterovirus infection in the first 2 months of life by real-time polymerase chain reaction. Clin Infect Dis.2003;37 :1– 6
- ↵Landry ML, Ferguson D. SimulFluor respiratory screen for rapid detection of multiple respiratory viruses in clinical specimens by immunofluorescence staining. J Clin Microbiol.2000;38 :708– 711
- ↵Byington CL, Bryan P, Castillo H, et al. Evaluation of a protocol for early discharge of hospitalized febrile infants 1–90 days of age. Pediatr Res.2002;51 :213A (abstr 1238)
- ↵McGowan KL, Foster JA, Coffin SE. Outpatient pediatric blood cultures: time to positivity. Pediatrics.2000;106 :251– 255
- Kumar Y, Qunibi M, Neal TJ, Yoxall CW. Time to positivity of neonatal blood cultures. Arch Dis Child Fetal Neonatal Ed.2001;85 :F182– F186
- ↵Kaplan RL, Harper MB, Baskin MN, Macone AB, Mandl KD. Time to detection of positive cultures in 28- to 90-day-old febrile infants. Pediatrics.2000;106(6) . Available at: www.pediatrics.org/cgi/content/full/106/6/e74
- Copyright © 2004 by the American Academy of Pediatrics