BACKGROUND: Fever in infants is a common clinical dilemma. The objective of this study was to present data from hospital systems across the northeast, southeast, mid-west, and western United States to identify the pathogens causing bacteremia in febrile infants admitted to general care units.
METHODS: This was a retrospective review of positive blood culture results in febrile infants aged ≤90 days admitted to a general care unit across 6 hospital systems. Data were collected from January 1, 2006 through December 31, 2012 from emergency departments and general inpatient units. Cultures from ICUs, central lines, or infants who had complex comorbidities were excluded, as were repeat cultures positive for the same bacteria. Common contaminants were considered pathogens if they were treated as such.
RESULTS: We identified 181 cases of bacteremia in 177 infants. The most common pathogen was Escherichia coli (42%), followed by group B Streptococcus (23%). Streptococcus pneumoniae was more likely in older infants (P = .01). Non-low-risk bacteremic infants were more likely to have E coli or group B Streptococcus than low-risk bacteremic infants. We identified no cases of Listeria monocytogenes. Variation between sites was minimal.
CONCLUSIONS: This is the largest and most geographically diverse study to date examining the epidemiology of bacteremia in infants. We suggest E coli is the most common cause of bacteremia in previously healthy febrile infants admitted to a general inpatient unit. We identified no cases of L monocytogenes and question whether empirical therapy remains necessary for this pathogen.
- AMC —
- Albany Medical Center
- CHCM —
- Children’s Hospitals and Clinics of Minnesota
- CHKD —
- The Children’s Hospital of The King’s Daughters
- CHLA —
- Children’s Hospital Los Angeles
- CHOI —
- Children’s Hospital of Illinois
- CoNS —
- coagulase-negative staphylococcus
- CSF —
- cerebrospinal fluid
- GBS —
- group B Streptococcus
- GCHS —
- Golisano Children’s Hospital at Strong
- SBI —
- serious bacterial infection
- SSTI —
- skin or soft-tissue infection
- UTI —
- urinary tract infection
What’s Known on This Subject:
Bacteremia occurs in 2.2% of febrile infants who have a blood culture drawn. Regional data suggest that Escherichia coli, group B Streptococcus, and Staphylococcus aureus are leading causes; however, the geographic boundaries of these data limit universal applicability.
What This Study Adds:
This is the first national study examining epidemiology of bacteremia in febrile infants admitted to a general inpatient unit. The most common pathogens were Escherichia coli (42%), group B Streptococcus (23%), and Streptococcus pneumoniae (6%). No Listeria monocytogenes was identified.
Fever in infants is a common problem faced by pediatricians and is often the only sign of serious bacterial infection (SBI) in this age group.1 It is estimated that 2% of infants presenting with fever without a source will have bacteremia, but it is often difficult to clinically predict which infants have bacteremia before obtaining blood culture results.2–6 For this reason, empirical broad spectrum antibiotic coverage is classically recommended until specific pathogens can be isolated or excluded; however, the recommended antibiotics have changed very little in the last few decades.7,8
Recent regional data suggest that the epidemiology of bacteremia in term infants is shifting, and that Escherichia coli is now the most common cause of bacteremia in febrile infants ≤90 days old.2 Once considered the most common cause of bacteremia in infants, estimates suggest that group B Streptococcus (GBS) may now account for as few as 16% of cases and was only the third most common cause of bacteremia in 1 study.6,9 Early-onset GBS has declined dramatically since the early 1990s from 1.7 cases per 1000 live births to 0.34 to 0.37 in recent years.10 Additionally, although evidence-based guidelines regarding empirical antibiotics continue to include coverage for Listeria monocytogenes,8 there is evidence that the incidence of listeriosis has been decreasing.11
The subpopulation of febrile infants who are well-appearing enough to be admitted to a general inpatient unit is important. These infants are exposed to empirical antibiotic coverage pending culture results, colloquially termed the “rule out sepsis” evaluation, but rarely demonstrate true bacteremia. A review of the literature identified no multicenter studies examining this issue, and regional studies were limited to <100 cases of bacteremia.5,6,10 The low prevalence of bacteremia in this group, and potential geographic variability, make it difficult for regional or single-site studies to provide generalizable epidemiological data, and thus warrant a national examination. Furthermore, although patients stratified as low or non-low risk by the Rochester criteria12 can differ in risk for SBI by greater than eightfold,5 a literature review did not identify a study examining bacterial epidemiology by risk.
The purpose of this study was to provide geographically diverse data on the epidemiology of bacteremia in previously healthy, febrile infants admitted to a general inpatient unit; to determine the incidence of L monocytogenes in this population; and to determine whether age or infant risk can predict the epidemiology of bacteremia.
This was a multicenter, retrospective review of positive, pathogenic blood cultures in previously healthy, febrile infants aged ≤90 days admitted to a general inpatient unit between January 1, 2006 and December 31, 2012. This study was independently approved by the Institutional Review Boards at all 6 participating institutions. Informed consent was waived.
The central site for this study was Golisano Children’s Hospital at Strong (GCHS) at the University of Rochester Medical Center, a tertiary care children’s hospital located in Rochester, New York. Participating sites included The Children’s Hospital of Illinois (CHOI), The Children’s Hospital of The King’s Daughters (CHKD), Children’s Hospitals and Clinics of Minnesota (CHCM), The Children’s Hospital at Albany Medical Center (AMC), and Children’s Hospital Los Angeles (CHLA). Table 1 describes the demographic attributes of the participating sites. All sites use a BACTEC automated blood culture detection system.
Each participating site obtained from their microbiology laboratory an initial database of all positive blood cultures drawn from infants ≤90 days old within their hospital system. Inter-site study dates varied secondary to availability of data and feasibility of collection.
A preliminary review of culture data allowed exclusion of cultures drawn in an ICU and a full chart review was then performed on the remaining cultures to determine whether study criteria were met. To qualify for inclusion, cultures must have been positive for bacteria, drawn from an infant ≤90 days of age, treated as a pathogen, and drawn from a patient who had either a fever per history or a recorded temperature of ≥38.0°C. Cultures drawn in any manner other than peripheral venipuncture, from children requiring an ICU level of care (defined as drawn in an ICU or from a patient admitted directly to an ICU), from patients with central lines, intra-abdominal, intra-cranial, or intra-thoracic surgical histories, or repeat cultures growing the same bacteria from the same infant were excluded. Cultures that grew common contaminants such as diphtheroids, Propionibacterium sp., or coagulase-negative Staphylococcus (CoNS) were generally excluded unless the culture was treated by the attending physician as a pathogen (eg, a full course of antibiotic therapy).
For cultures meeting study criteria, clinical and demographic information was collected using a standardized extraction tool. Abstracted information included age at the time of culture, gender of the infant, bacterial species isolated from the blood, and infant risk based on modified Rochester Criteria similar to that used by others.12,13 Each infant was stratified as either low risk or non-low risk. Low risk infants were defined as those without evidence of focal infection, previous antibiotic administration, treatment of hyperbilirubinemia, previous hospitalization, history of preterm birth (<37 weeks), chronic medical conditions, abnormal white cell count (<5000 or >15 000 white blood cells per mm3), absolute band cell count of >1500 per mm3, or urinalysis results (>10 white blood cells/high power field).7,8 As in previous studies, patients who did not meet all of the low risk criteria, or who had classifying data that were unavailable, were placed in the non-low risk group.12–14 When available, cerebrospinal fluid (CSF) and urine culture results were recorded if drawn within 24 hours of the positive blood culture to assess for concurrent urinary tract infection (UTI) and/or meningitis. In cases of Staphylococcus aureus bacteremia, a further chart review was done to determine whether there was concern for skin or soft-tissue infection (SSTI) at the time of blood culture. The principal investigator at each site was responsible for ensuring that cultures from their site met study criteria and data were reported to GCHS for analysis.
Prevalence of bacterial species among infants was calculated and reported along with a basic statistical summary. χ2 was used to determine variance in bacterial prevalence among the 6 sites (the 2 smallest sites were combined) for prevalence of E coli compared with all other bacterial species and GBS compared with all other bacterial species. The 95% confidence interval for L monocytogenes was determined by using a method proposed for numerators of zero.15,16 χ2 or Fisher’s exact tests were used to test dichotomous variables, with statistical significance set at P < .05. The statistical program used was SPSS version 21 (IBM SPSS Statistics, IBM Corporation).
We identified 2901 positive blood cultures from infants ≤90 days of age within the study period (Fig 1). Eighteen hundred were excluded because they were drawn from patients who were in a neonatal or pediatric ICU, admitted directly to the ICU, or who had central lines. Of the remaining cultures, 811 were either identified as a contaminant or not treated as a pathogen and therefore did not meet inclusion criteria, the majority of which were CoNS (59%). Because of complex comorbidities, an additional 96 cultures were excluded. Of the remaining 194 cultures, 9 were not obtained from an infant who had a fever history and 4 were duplicate cultures growing the same bacteria from the same patient, leaving 181 cultures for analysis obtained from 177 infants. Table 2 displays demographic information for cultures by site and in total. E coli was the most prevalent pathogen in all but 1 hospital system. There was no statistically significant difference in the prevalence of E coli, GBS, or all other bacteria species between sites (χ2 = 9.9, P = .28).
Nineteen different bacterial species were identified (Table 3). The most common pathogen recovered from all blood cultures was E coli (76/181, 42%), followed by GBS (41/181, 23%), Streptococcus pneumoniae (10/181, 6%), and S aureus (9/181, 5%). Of the infants we identified with S aureus bacteremia, 56% (5/9) had evidence of SSTI on presentation. There were no cases of L monocytogenes (95% confidence interval, 0%–2% of total cases).
Urine cultures were obtained in 172 (95%) cases and 85 (49%) of these were positive for bacteria. Of the 76 blood cultures that grew E coli, 75 (99%) had an associated urine culture collected and 68 (91%) of these were found to have concurrent E coli UTI. Of the 105 blood cultures positive for other bacterial species, 97 (92%) had associated urine cultures and 16 (15%) had concurrent UTI.
CSF cultures were obtained in 151 (83%) cases, 20 (13%) of which were positive for bacteria. GBS was the bacterial species most likely to cause concurrent meningitis (10/37, 27%). There was also 1 case of Pantoea sp. bacteremia with concurrent GBS meningitis. Of the 76 cases of E coli bacteremia, 63 had CSF cultures obtained and 5 (7%) had concurrent meningitis. Two infants who had E coli bacteremia and E coli UTI also had concurrent E coli meningitis (2/68, 3%). Of the 10 cases of S pneumoniae bacteremia, 4 had CSF cultures obtained and 1 of these had concurrent meningitis.
In the majority of cases, infants were non-low risk (144/181, 80%). Four (3%) of these were placed into this category owing to missing data. There was no difference in gender (P = .88) or age (P = .97) between risk groups. Table 4 highlights the characteristics of infants who had bacteremia from the 5 most common pathogens. Non-low risk bacteremic infants were significantly more likely than low risk bacteremic infants to have E coli (P = .001) or GBS (P = .01). In the non-low risk cohort, 4% (8/181) of cultures grew Klebsiella sp. as compared with none (0/37) in the low risk infant cohort. Infants who had S pneumoniae bacteremia were significantly more likely to be older than infants who had other causes of bacteremia (P = .01).
This multiregional study examined the epidemiology of bacteremia in previously healthy febrile infants outside of the ICU. A national approach is particularly salient to the question of bacterial epidemiology in our patient population for 2 reasons. First, bacteremia in this group is a rare event and combining data from 6 centers allows us a more rigorous examination than would otherwise be possible. Second, patterns of bacterial prevalence may vary geographically, potentially limiting generalizability of regional data.
Our study demonstrates an increasingly prominent role for E coli as a cause of bacteremia in febrile infants in the United States. These results are largely consistent with single-region observations, although we do demonstrate a slightly lower percentage of bacteremia caused by Gram-negative organisms (53%) than previous studies (63%–80%).2,10,17
The majority of infants who had E coli bacteremia had concurrent UTI. It has been suggested that infants who have E coli UTI may not require a lumbar puncture.18 However, in infants who had E coli UTI and bacteremia, we identified 2 cases (3%) of concurrent meningitis. GBS remains a common cause of bacteremia and was the most common bacteria to cause concurrent meningitis (27%).
Consistent with previous regional data, we found no cases of L monocytogenes bacteremia.2,10,17 There has been an overall decrease in laboratory-confirmed listeriosis in recent decades, likely owing to prohibitions regarding the sale of potentially contaminated food and public education campaigns targeted at high-risk populations, including pregnant women.19 Although L monocytogenes is generally considered a cyclical pathogen, making it possible that our data were gathered during a “lull” in incidence, we collected geographically diverse data over several years during which time there were several nationally reported outbreaks of listeriosis.20 Historically, physicians have been taught that bacteremia in young infants is caused by GBS, E coli, and L monocytogenes.21,22 This classic teaching stems from studies performed decades ago when L monocytogenes was a more common cause of food-borne infection.23,24 Based on our data and regional studies performed previously, we suggest that the etiologies of SBI taught to physicians in training and documented in written text should be revised to reflect current epidemiology.9,18
Classifying each culture as coming from either a low or non-low risk infant was done in an attempt to determine whether children who had certain bacteria were more likely to be “sicker.” As would be expected, most infants who had bacteremia were classified as non-low risk by the Rochester criteria. The finding of bacteremia in low risk infants is not a critique of the performance of published guidelines for the management of low risk febrile infants,12 as we examined only children who had bacteremia. Therefore, we are not attempting to provide data regarding risk for bacteremia in either risk cohort. What our data suggest is that, in infants who have bacteremia, those classified as non-low risk are more likely to have bacteremia attributable to E coli or GBS than low risk infants. It is notable that, although sample size limits statistical significance, we did not find a single case of Klebsiella sp. in our low risk infant population.
Current recommendations for empirical antibiotic coverage for febrile infants typically include the combination of either ampicillin and gentamicin or ampicillin and a third-generation cephalosporin.21,22 The importance of ampicillin in the empirical antibiotic regimen for febrile infants has been maintained for coverage of both L monocytogenes and Enterococcus sp.25 Although a large prospective examination is necessary before firm recommendations regarding a change in empirical therapies can be made, we identified no cases of L monocytogenes bacteremia and only 4% of infants had bacteremia caused by Enterococcus sp. Given that the overall risk for bacteremia in our population is estimated at 2%,2,5 we suggest that incidence of Enterococcus bacteremia may be <0.1%. Future prospective studies examining resistance patterns and incidence may confirm that empirical monotherapy with a third-generation cephalosporin, such as cefotaxime, may be adequate. Cefotaxime is well tolerated in young infants and should provide coverage for the vast majority of pathogens we identified.26,27 This study also highlights the emergence of S aureus as a leading pathogen in bacteremia in young infants, similar to that seen previously,2,10 and it should be considered particularly for febrile infants in whom SSTI is suspected.
Our decision to use treatment with a full course of antibiotics as an aspect of our study criteria warrants some explanation, although this is not the first time that treatment has been used in some way to determine contamination.2 Although there are bacteria that should always represent a true infection,28 and some that almost always represent contamination in an otherwise healthy child,28,29 using bacterial species alone to determine contamination is unlikely to be sufficient, because many common contaminants can also represent true bacteremia.28–30 For this reason we feel that, although we did potentially exclude up to 3 cases of true bacteremia (1 culture grew several species including S aureus, and 2 others grew E coli; per chart review, none were treated and all did well), our method of determining contaminants allowed us to avoid including many cases that likely did not represent true bacteremia. One example was a febrile infant sent home from the emergency department without antibiotics after having blood cultures drawn that eventually grew Enterococcus sp. Further review revealed that the repeat cultures from the outpatient office showed no growth, and that the child was afebrile and doing well at 1 week follow-up. Our study criteria allowed us to exclude this patient, who would otherwise have been mistakenly included.
Our study had several limitations. First, the retrospective nature of the review inhibited our ability to identify negative blood cultures, so we are unable to provide information regarding the overall risk for bacteremia. This limits our ability to provide firm recommendations regarding empirical antibiotic therapy and a prospective, national study is necessary. Second, criteria for ICU admission vary from institution to institution and this may have introduced some unidentified heterogeneity in our patient population. Third, when analyzing the difference in bacterial species between risk cohorts, statistical significance of some individual species was limited by small sample sizes (eg, Klebsiella sp.). Additionally, we assigned risk retrospectively after the infant was known to have bacteremia; risk classification is typically employed as a way to predict which infants will have an SBI before 1 being identified. Finally, the start date of data collection varied between sites by up to 14 months owing to limitations in the availability of data within each institution’s microbiology database. However, our analysis found no statistically significant variability in the prevalence of E coli, GBS, or all other species (P = .28), which suggests uniformity across the dates of study.
We present the largest study to date examining the question of bacterial epidemiology in infants admitted to the general care unit with bacteremia. These data support the trend seen in previous regional publications that E coli has become the most common cause of bacteremia in febrile infants admitted to the general care unit, with GBS and S pneumoniae the second and third most common causes, respectively. Together, these 3 species account for 71% of positive blood cultures in our patient population. Not a single case of L monocytogenes was identified. In infants who have bacteremia, E coli and GBS are more likely to be identified as the cause in non-low risk than in low risk infants and S pneumoniae is more likely to be found in older infants. We suggest a prospective evaluation to determine whether a change in typical empirical antibiotic regimens is necessary.
- Accepted September 12, 2013.
- Address correspondence to Eric Biondi, MD, Department of Pediatrics, University of Rochester Medical Center, 601 Elmwood Ave, Box 667, Rochester, NY 14620. E-mail:
Dr Biondi conceptualized and designed the study, collected data, served as the central site principal investigator, coordinated data collection at all 6 study sites, drafted much of the initial manuscript, and participated in statistical analysis; Dr Evans aided in study conceptualization and design, collected data at her institution, served as participating site principal investigator, and drafted parts of the initial manuscript; Dr Mischler aided in study conceptualization and design, collected data at his institution, served as participating site principal investigator, drafted parts of the initial manuscript, and aided in statistical analysis; Dr Bendel-Stenzel aided in data collection, served as principal investigator at his institution, and reviewed drafts of this manuscript; Dr Horstmann aided in data collection, served as principal investigator at her institution, drafted parts of the initial manuscript, and reviewed drafts of the manuscript; Dr Lee aided in data collection, served as principal investigator at her institution, and reviewed drafts of the manuscript; Dr Aldag provided statistical analysis of the compiled data and critically reviewed and drafted aspects of the methodology and results sections; Dr Gigliotti aided in conceptualization and study design, provided supervision and mentorship at the central site, and critically reviewed the manuscript; and all authors approved the final manuscript as submitted.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Bachur RG,
- Harper MB
- Byington CL,
- Enriquez FR,
- Hoff C,
- et al
- Kourtis AP,
- Sullivan DT,
- Sathian U
- ↵Team FUS. Cincinnati Children’s Hospital Medical Center: Evidence-based clinical care guideline for fever of uncertain source in infants 60 days of age or less. Guideline 02, 1. Available at: www.cincinnatichildrens.org/svc/alpha/h/health-policy/ev-based/default.htm. Accessed October 14, 2010
- Byington CL,
- Rittichier KK,
- Bassett KE,
- et al
- Verani JR,
- McGee L,
- Schrag SJ,
- Division of Bacterial Diseases, National Center for Immunization and Respiratory Diseases, Centers for Disease Control and Prevention (CDC)
- Voetsch AC,
- Angulo FJ,
- Jones TF,
- et al.,
- Centers for Disease Control and Prevention Emerging Infections Program Foodborne Diseases Active Surveillance Network Working Group
- Jaskiewicz JA,
- McCarthy CA,
- Richardson AC,
- et al.,
- Febrile Infant Collaborative Study Group
- Bennion JR,
- Sorvillo F,
- Wise ME,
- Krishna S,
- Mascola L
- ↵Cartwright EJ, Jackson KA, Johnson SD, et al. Listeriosis outbreaks and associated food vehicles, United States, 1998–2008. Emerg Infect Dis. Available at: http://dx.doi.org/10.3201/eid1901.12039. Accessed May 28, 2013
- ↵Fisher RG, Boyce TG. Moffet’s Pediatric Infectious Diseases, A Problem-Oriented Approach. 4th ed. Philadelphia, PA: Lippincott Williams and Wilkins; 2005:355–366, 650–651
- Zaoutis LB,
- Chiang VW
- ↵Committee on Infectious Diseases, American Academy of Pediatrics. Pickering LK, ed. Red Book: 2012 Report of the Committee on Infectious Diseases. 29th Ed. Elk Grove Village, IL: American Academy of Pediatrics; 2012
- Lin FY,
- Azimi PH,
- Weisman LE,
- et al
- Fernandez M,
- Hickman ME,
- Baker CJ
- Weinstein MP,
- Towns ML,
- Quartey SM,
- et al
- Weinstein MP
- Hall KK,
- Lyman JA
- Copyright © 2013 by the American Academy of Pediatrics