This article has a correction. Please see:
This article has a correction. Please see:
OBJECTIVE: Describe the etiology of bacteremia among a geographically diverse sample of previously well infants with fever admitted for general pediatric care and to characterize demographic and clinical characteristics of infants with bacteremia according to bacterial etiology. We hypothesized that the epidemiology of bacteremia in febrile infants from a geographically diverse cohort would show similar results to smaller or single-center cohorts previously reported.
METHODS: This was a retrospective review of positive, pathogenic blood cultures in previously healthy, febrile infants ≤90 days old admitted to a general unit. In total, there were 17 participating sites from diverse geographic regions of the United States. Cultures were included if the results were positive for bacteria, obtained from an infant 90 days old or younger with a temperature ≥38.0°C, analyzed using an automated detection system, and treated as pathogenic.
RESULTS: Escherichia coli was the most prevalent species, followed by group B Streptococcus, Streptococcus viridans, and Staphylococcus aureus. Among the most prevalent bacteria, there was no association between gender and species (Ps > .05). Age at presentation was associated only with Streptococcus pneumoniae. There were no cases of Listeria monocytogenes.
CONCLUSIONS: Our study confirms the data from smaller or single-center studies and suggests that the management of febrile well-appearing infants should change to reflect the current epidemiology of bacteremia. Further research is needed into the role of lumbar puncture, as well as the role of Listeria and Enterococcus species in infantile bacteremia.
In the past 20 years, significant advances in vaccine development, increasingly stringent food safety guidelines, and the use of prophylactic antibiotics before delivery have shifted the epidemiology of serious bacterial infection (SBI) in infancy.1–12 However, the choice of empirical antibiotic therapy has not changed over the past several decades.13–15 The accepted standard of care is to initiate broad-spectrum antibiotics, typically ampicillin and either gentamicin or a third-generation cephalosporin, before culture data availability because of the high morbidity and mortality of untreated SBI.16 However, recent data suggest that common empirical antibiotic choices may not be well matched to the current pattern of organisms that cause SBI.1,3,12,17–20 For example, Listeria monocytogenes currently makes up a small subset of cases of bacteremia in this population, with some studies failing to identify any cases of listeriosis in hospitalized infants with fever.1,2,12,20 As a sign of shifting epidemiology, group B Streptococcus (GBS), once the most common cause of bacteremia in infants, now comprises as few as 16% of cases overall.5,6 In contrast, Escherichia coli has been identified as a major pathogen, the importance of which has increased in recent years.1,2,12,20
Previous studies designed to characterize the epidemiology of bacteremia in the postvaccine era have been limited by geographic isolation and relatively small sample sizes.4,13,21,22 We previously reported epidemiologic data for febrile infants with bacteremia from a small multicenter cohort of hospitals.12 In an attempt to improve on the external validity of research in this area, we sought a larger sample from a geographically diverse group of institutions to better address this question. The objectives of this study were to describe the etiology of bacteremia among previously well infants with fever ≤90 days of age admitted for general pediatric care and to characterize demographic and clinical characteristics of infants with bacteremia according to bacterial etiology.
We hypothesized that the epidemiology of bacteremia in febrile infants would show similar results in a larger, more nationally representative sample. Data from this article represent the complete data of all participating institutions recruited to answer this question and include data from our previous smaller cohort.
This was a retrospective, multicenter, cross-sectional review of positive, pathogenic blood cultures in previously healthy, febrile infants ≤90 days old admitted to a general inpatient unit. The Pediatric Research in Inpatient Settings Network was used to facilitate site recruitment through participating hospitals. In total, there were 17 hospital systems from diverse geographic regions of the United States.
All sites used an automated blood culture detection system. The study collection period reflected available data from each site between January 2006 and January 2013. This study was independently approved by the institutional review boards at all participating institutions. Informed consent was waived.
Participating sites obtained all positive blood cultures from their microbiology laboratory in infants ≤90 days old that were drawn in their emergency department, outpatient clinics, and general inpatients units. A detailed description of culture inclusion and exclusion criteria, including full methodology of the epidemiology aspect of the cultures, has been previously published.23 Cultures were included for analysis if they were positive for bacteria, obtained from an infant 90 days of age or younger with a temperature ≥38°C recorded on presentation or reported by a caregiver, analyzed using an automated bacterial detection system, and treated as a pathogen by the medical team (defined as a full course of antibiotics intended to treat the bacteria identified in culture). Cultures were excluded if they were drawn in any method other than peripheral venipuncture, obtained from a patient in the ICU or admitted to the ICU within 5 hours after the culture, from patients with central indwelling catheters or hardware, or with a history of intraabdominal, intracranial, or intrathoracic comorbidities or surgical procedures, or drawn from a hospital or clinic outside the system of any participating site.
For cultures that met study criteria, clinical and demographic information was collected using a standardized data extraction tool. Information collected included patient age at the time of collection, gender, hospital site, antibiotics received within 24 hours preceding the culture, highest recorded temperature within 1 hour of obtaining the blood culture, and bacterial species isolated from the blood. Infants were stratified as either low risk or non–low risk based on modified Rochester Criteria. As in previous studies, infants not meeting all of the low-risk criteria or who had classifying data that was unavailable, were placed in the non–low-risk group.18,24,25 Infants were considered low risk if they had no evidence of focal infection, no previous hospitalization, treatment of hyperbilirubinemia or antibiotic administration, no history of preterm birth (<37 weeks), no chronic medical conditions, normal white blood cell count (>5000 and <15 000 cells per mm3), absolute band count of <1500 cells per mm3, and no pyuria (<10 white blood cells/high powered field). If available, urine and cerebrospinal fluid culture results were also recorded to assess for concurrent urinary tract infection (UTI) and/or meningitis.
Descriptive statistics were calculated for each positive culture included in the study. χ2 testing was used to determine variation in bacterial species related to the sites. A continuity adjusted χ2 test or Fisher’s exact test were used to examine the associations among species and gender, species and non–low risk, and species and age group among the most prevalent bacterial species. A geographic block map was created to describe the distribution of sample collection. All statistics were computed using SAS 9.3 (SAS Institute, Cary, NC). A 2-tailed P value was calculated for all tests, and P ≤ .05 was considered significant.
We previously reported detailed data regarding the numbers of cultures meeting each specific inclusion and exclusion criterion.23 In total, 392 cultures were identified from 17 sites for analysis. Figure 1 displays the geographic distribution of culture collection from our sample, with the largest number of cultures collected in Minnesota, Texas, Ohio, and Colorado, representing a fairly diverse geographic sample. Table 1 shows the demographic characteristics for each site, including the distribution of the 4 most common organisms in our sample by site. Of the 25 species recovered, E coli was the most prevalent, followed by GBS, Streptococcus viridans, and Staphylococcus aureus. The median age range in the sample was 36 days of age, with the majority (75%) collected from infants considered to be non–low risk. Figure 2 displays the variation in the most common bacterial species across the sites, with sites 13 through 17 combined into 1 group. There was a statistically significant difference in the prevalence of GBS, E coli, and the remaining bacterial species between sites (χ2 = 35.09, P < .05).
Table 2 displays the species recovered from blood culture and the corresponding demographic characteristics and rates of concurrent UTI and meningitis of each bacterial species. Table 3 displays the clinical variables associated with the most prevalent bacteria. Among the 4 most prevalent bacteria, there was no association between gender and bacterial species (P > .05). Age at presentation was associated with Streptococcus pneumoniae only, with older infants more likely than younger infants to show bacteremia due to this organism (P < .05). Infants with bacteremia due to E coli and GBS were more likely to be stratified as non–low risk (P < .05). There were no cases of L monocytogenes in the cohort during the time period sampled.
Urine cultures were obtained in our sample in 378 cases (96.4%) and were positive for pathogenic infection in 168 cases (44.4%). Of these, E coli was the most frequent, found in 142 of 157 (90.4%) cases of bacteremia. Eight infants with UTI had concurrent meningitis (4.8%), with age ranges of 3 to 85 days. Five of 8 infants(63%) were considered non–low risk on presentation. Of the 3 who were considered low risk, 1 was >30 days of age. Thus, of the 142 infants presenting with E coli bacteremia, 1 infant (0.70%) was >30 days of age and considered low risk on presentation with concurrent meningitis.
The remaining distribution of concurrent UTI is noted in Table 2.
Cerebrospinal fluid cultures were obtained in our sample in 327 cases (83.4%) and were positive in 32 cases (9.7%). Of these, GBS was the most likely organism to show concurrent infection, with 16 cases of meningitis out of 87 cases of bacteremia (18.3%).
This multiregional study assessed the epidemiology of bacteremia in previously healthy febrile infants ≤90 days of age outside of the ICU. By expanding to include 17 centers, this study expanded the results of our previously published study allowing for a more broad generalization of its findings. The total number of blood cultures in the current sample is 392, slightly more than double the previous sample of 181 cultures. This larger, more nationally diverse sample confirms the results of our initial smaller study of 6 hospital systems, identifying E coli as the predominate pathogen in bacteremic infants ≤90 days of age, followed by GBS. The emergence of E coli may represent the success of maternal intrapartum antibiotic prophylaxis for prevention of early-onset GBS sepsis.5,6 Although the overall proportion of bacteremia caused by Gram-negative organisms was lower in this sample (52%) than in other studies (63%–80%), it is consistent with our previous study (53%).1,5,12,20 As has been done previously, we considered a bacteria to be a pathogen if it was treated as such by the primary team.1,12 S viridans was the next most common organism in our sample; however, 47% (n = 15) of these cases came from a single institution. The incidence of bacteremia due to this organism in our sample, as well as its skewed site variability may reflect provider level and hospital level variation in considering the organism as a pathogen or a contaminant. S aureus, an emerging pathogen in febrile infants with skin and/or soft tissue infections was isolated from 6% of bacteremic infants.7,19,26 In our current study, bacteremia due to S aureus was more prevalent than S pneumonia, possibly reflecting an increase in skin and soft tissue infections during this time period in certain geographic locations. Interestingly, 7 of 23 infants with S aureus bacteremia were considered low risk by Rochester criteria. Although this result was not statistically significant, it does bear further investigation as to the behavior of S aureus as a pathogen in this age group. Also, similar to our previous study, S pneumoniae bacteremia was more common in infants older than 60 days (P < .05) in this larger sample.
In designating infants as low risk or non–low risk according to modified Rochester criteria, we aimed to determine if the risk status of a febrile infant could help differentiate the bacterial cause. As expected, most infants (75%) with bacteremia met non–low-risk criteria. Similar to our previous study, infants with bacteremia classified as non–low risk are more likely to have E coli or GBS as the source than infants classified as low risk. However, because of low numbers of bacteremia in each individual species, statistical comparisons between species could not be performed. Only positive blood cultures were included in this analysis; therefore, calculation of positive and negative predictive value of “risk” criteria was not possible.
Compared with our previous study, concurrent UTI was slightly less prevalent in the current sample (44% vs 49%), whereas meningitis was similar (10% vs 13%).12 The majority of infants with E coli bacteremia had a concurrent UTI (90%), but few had concurrent meningitis. Previous literature has suggested a high rate of concurrent meningitis in infantile UTI.27 More recent literature suggests that older infants (30–90 days of life) with UTI have a low rate of concurrent meningitis.28–30 It has been suggested that in well-appearing infants in this age group considered to be at low risk on presentation, lumbar puncture to assess for concurrent meningitis may not be necessary.31,32 However, because our study only examines children with bacteremia, which may indicate a higher level of illness compared with the standard child presenting with UTI, it is difficult to draw firm conclusions about the role of lumbar puncture in infants with both bacteremia and UTI.
Similar to previous studies, no cases of L monocytogenes bacteremia were identified in our population.1,12,20,33 Laboratory confirmed cases of listeriosis have decreased in recent decades, likely reflecting the success of US food standard regulations prohibiting the sale of potentially contaminated food or preventive education efforts targeting pregnant women.14 Listeriosis is a cyclical foodborne illness occurring in outbreaks, of which there were several during the time of data collection, making it unlikely that our results reflect a “lull” in Listeria activity.15,16,33 The classic teaching that bacteremia in febrile infants is caused by GBS, E coli, and L monocytogenes is not supported by our data nor by regional studies1,4,12,20,33,34 Late-onset listeriosis typically presents with meningitis without bacteremia.35–37 As our study assessed bacteremia, cases of L monocytogenes meningitis without concurrent bacteremia may have occurred in our study population. However, the vast majority of central nervous system infections occur in neonates between the second and fourth weeks of life and present with fever, toxicity, and neurologic signs.35–37 Therefore, we speculate the incidence of zero cases seen in both of our studies is a true reflection of disease burden in our population of febrile infants presenting to general pediatric services. However, given the wide variability of ICU admission criteria nationwide, it is possible that a case of L monocytogenes meningitis without bacteremia in an infant was missed in our sample. On the basis of our previous study and now this larger sample, we suggest the etiologies of bacteremia taught and written in texts should be updated to reflect the current epidemiologic data.
Current guidelines recommend empirical ampicillin and either gentamicin or a third-generation cephalosporin for the treatment of febrile infants suspected of having a SBI.16 In addition to Listeria coverage, ampicillin is included to provide coverage for Enterococcus species, an uncommon pathogen in this population. In our current sample, we identified no cases of Listeria bacteremia and found 3.6% of infants in our cohort with Enterococcus bacteremia. Given the estimated 2% risk of bacteremia in our patient population, we estimate that only 1 of every ∼1400 (95% confidence interval 924–2883) previously healthy, febrile infants who has blood cultures drawn will have enterococcal bacteremia.1,4
These epidemiologic observations related to Listeria and Enterococcus sp. as well as growing ampicillin resistance among Gram-negative rods calls into question the empirical use of ampicillin in this population. Some have suggested empirical therapy with a third-generation cephalosporin such as cefotaxime without ampicillin.12,20 Cefotaxime is well tolerated and will provide coverage for the major bacterial pathogens identified in this patient population.38,39 However, larger studies examining emerging cephalosporin resistance rates are needed before firm recommendations can be made.
Our study has several limitations. First, we conducted this study in a retrospective manner, thus limiting our ability to review data beyond what was available in a chart review. Second, our sample may have included overrepresentation of certain geographic areas and centers. More than half our data were obtained from 4 large children’s hospitals in the Midwest and Northeast with a notable absence of any representation from the Southeast and Pacific Northwest. In addition, we did find a statistically significant difference in species distribution between sites. However, with a relatively large number of sites and a small number of blood cultures at some of the sites compared with others, we suspect that this likely does not represent true statistical variability but rather Type I error due to multiple dependent variables in determining site variability. Third, our analysis was limited to hospitals that possessed automated detection systems. Fourth, we suspect that individual hospitals had varying methods for determining ICU versus non-ICU placement. In addition, the exclusion of patients admitted into ICU settings at any point during their hospitalization hinders our ability to draw conclusions regarding that subset of infants. Although it excludes a cohort of infants for analysis, it also focuses the current data set on the majority of infants who present with fever for sepsis evaluation who do not require ICU level care after admission. Fifth, in regard to S aureus bacteremia, we did not collect information on concurrent skin and soft tissue infection, limiting our ability to inform the discussion further on this organism. Finally, we defined bacteremia based on provider decision to treat a positive culture. It is possible that we either under- or overestimated the rates of pathogenic infections because of inherent subjectivity and institutional variability in using provider-based decision making to classify cases of bacteremia. The a priori approach to true bacteremia in this sample leads to the inclusion of S viridans as the third most common pathogen in our study. If the opposite approach is taken, with exclusion of all S viridans as contaminants, S aureus becomes the third most common pathogen, with S pneumoniae becoming the fourth most common. Unfortunately, because of the a priori approach to true infection versus contaminant, the true role of S viridans as a pathogen is difficult to ascertain from the current sample.
This multiregional study is the largest to date examining the epidemiology of bacteremia in infants ≤90 days of age admitted to the general care unit. The data from this large cohort of hospital systems confirm the data from our smaller study and suggest that the management of febrile well-appearing infants, and the teaching that accompanies that management for residents and medical students, should change to reflect the current epidemiology of bacteremia in this population. Further research is needed to define the role of lumbar puncture in febrile infants with UTI, as well as the true role and pathogenesis of Listeria, Staphylococcus, and Enterococcus species in infantile bacteremia.
The authors thank the executive council for the Pediatric Research in Inpatient Settings network for their help in facilitating network participation in this study.
The authors also thank the collaborators at each site who collected blood culture data for inclusion in the current study: Michael Bendel-Stenzel, MD, Children’s Hospitals and Clinics of Minnesota (Minneapolis, MN); Clifford Chen, MD, Children’s Medical Center, University of Texas Southwestern (Dallas, TX); Rianna Evans, MD, The Children’s Hospital of The King’s Daughters (Norfolk, VA); Jason French, MD, Children’s Hospital Colorado (Aurora, CO); Sara Horstmann, MD, Albany Medical Center (Albany, NY [now Carolinas Medical Center, Charlotte, NC]); Karen Jerardi MD, MEd, Cincinnati Children’s Hospital Medical Center (Cincinnati, OH); Vivian Lee, MD, Children’s Hospital Los Angeles (Los Angeles, CA); Melissa Schafer, MD, State University of New York Upstate (Syracuse, NY); Samir Shah, MD, Cincinnati Children’s Hospital Medical Center, Division of Hospital Medicine (Cincinnati, OH); Angela Statile, MD, MEd, Cincinnati Children’s Hospital Medical Center, Division of Hospital Medicine (Cincinnati, OH); Samuel C. Stubblefield, MD, Nemours/A.I. DuPont Hospital for Children (Wilmington, DE); Anne Vanden Belt, MD, St Joseph Mercy Hospital (Ann Arbor, MI).
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.
- coagulase-negative Staphylococcus
- group B Streptococcus
- lumbar puncture
- serious bacterial infection
- urinary tract infection
- Greenhow TL,
- Hung YY,
- Herz AM
- Byington CL,
- Enriquez FR,
- Hoff C,
- et al
- Verani JR,
- McGee L,
- Schrag SJ
- Bennion JR,
- Sorvillo F,
- Wise ME,
- Krishna S,
- Mascola L
- Voetsch AC,
- Angulo FJ,
- Jones TF,
- et al
- Biondi E,
- Evans R,
- Mischler M,
- et al
- Team FUS
- Fisher RG,
- Boyce TG
- 15.↵Committee on Infectious Diseases, American Academy of Pediatrics. Red Book: 2012 Report of the Committee on Infectious Diseases. 29th ed. Elk Grove Village, IL: American Academy of Pediatrics; 2012.
- Zaoutis L,
- Chiang V
- Jaskiewicz JA,
- McCarthy CA,
- Richardson AC,
- et al
- Byington CL,
- Rittichier KK,
- Bassett KE,
- et al
- Byington CL,
- Reynolds CC,
- Korgenski K,
- et al
- Schnadower D,
- Kuppermann N,
- Macias CG,
- et al
- Silk BJ,
- Date KA,
- Jackson KA,
- et al
- Fernandez M,
- Hickman ME,
- Baker CJ
- Lin F-YC,
- Azimi PH,
- Weisman LE,
- et al
- Copyright © 2015 by the American Academy of Pediatrics