Risk of Serious Bacterial Infection in Children With Fever Without a Source in the Post–Haemophilus influenzae Era When Antibiotics Are Reserved for Culture-Proven Bacteremia FREE

FEVER IS A COMMON pediatric complaint and accounts for approximately 20% of all pediatric emergency department (ED) visits.¹ Children aged 2 to 36 months with fever without a source (FWS) are at risk of developing occult bacteremia (OB) and may subsequently develop serious bacterial infection (SBI).^{2 - 6} The evaluation and treatment of these children remains a challenge to physicians. Previous recommendations suggest empiric treatment to prevent SBI in children at risk for OB.^{7 - 8} Bacterial meningitis, sepsis, and death are the outcomes that frequently drive empiric antibiotic treatment in children with suspected OB. However, based on the results of earlier studies,^{9 - 12} these outcomes are rare events, with an overall combined rate of 0.1% for bacterial meningitis and sepsis in children with FWS. Since the introduction in 1993 of practice guidelines by Baraff et al,^{7 - 8} widespread Haemophilus influenzae type b immunizations and continuous blood culture monitoring systems have been implemented.¹³ Although recent meta-analyses^{14 - 15} have demonstrated a trend toward reduced risk of SBI with empiric antibiotic treatment, statistical significance was not reached despite the large sample sizes. Therefore, the utility of empiric antibiotic treatment may be less than previously believed.

The objective of our study is to determine the rate of SBI in children aged 2 to 36 months with FWS in the post–H influenzae era, when antibiotic therapy is reserved for culture-proven bacteremia.

A retrospective review was performed of the medical records of febrile patients aged 2 to 36 months who were evaluated in the ED or urgent care center (UCC) of Children's Hospital of Wisconsin, an urban tertiary care children's hospital. The study was approved by the institutional human rights review board of Children's Hospital of Wisconsin.

Patients aged 2 to 36 months, evaluated between January 1, 1995, and July 31, 2000, in the ED or UCC with fever (International Classification of Diseases, Ninth Revision, code 780.6) as one of the discharge diagnoses were identified from the hospital medical record database. Our usual coding practice is to assign a fever code along with other diagnoses where applicable if the patient presented with a complaint of fever. Thus, we could identify most patients eligible to have blood samples taken because of FWS in the ED and UCC from the hospital database alone. The eligibility of patients' enrollment in the study was determined by reviewing the entire ED and UCC visit records identified from the hospital medical record database by the investigators. To ensure the enrollment of all patients from whom a blood sample was drawn in the ED or UCC, we also reviewed the hospital microbiology laboratory database. Immunocompetent children with a documented ED or UCC rectal temperature of 39.0°C or greater were eligible for enrollment. Exclusion criteria included (1) ED or UCC temperature less than 39.0°C; (2) identifiable source of infection (otitis media, sinusitis, pneumonia, urinary tract infection, meningitis, osteomyelitis, septic arthritis, cellulitis, bronchiolitis, croup, gingivostomatitis, varicella, roseola, and fifth disease); (3) current or recent use of antibiotics within 1 week preceding the ED or UCC visit, antibiotic therapy in the ED or UCC, or prescription for use at home; and (4) hospital admission. Patients with viral upper respiratory tract infections and gastroenteritis were not excluded from our study.

During the study, an ED clinical practice guideline for patients aged 2 to 24 months with FWS was in place that recommended obtaining a catheterized urine specimen for urinalysis and culture in boys up to 6 to 12 months of age and girls up to 12 to 24 months of age with a temperature of 39.0°C or greater. Chest radiography was recommended only for patients with respiratory tract symptoms, and complete blood cell count was not a routine test. Recommendation for blood culture included all patients with a temperature of 41.0°C or greater and, at the physician's discretion, for patients with a temperature of 39.0°C to 40.9°C. Use of antibiotics was recommended only for patients with an apparent bacterial focus of infection, suspected SBI requiring hospital admission, or culture-proven bacteremia. Antibiotics were not recommended for patients with FWS alone. Patients who were discharged from the hospital received a printed comprehensive home care teaching sheet for febrile children and were advised to return to the ED if they continued to have a temperature of 39.0°C or greater, developed other symptoms, or appeared ill. A designated callback nurse provided telephone follow-up within 24 to 48 hours of an ED visit. Routine follow-up advice included continuation of antipyretic therapy if the patient was still febrile or return to the ED if he or she appeared irritable, lethargic, or had decreased intake/output or difficulty breathing.

Enrolled patients were classified according to Figure 1. Patients with an ED temperature of 39.0°C or greater were assigned to either group 1 (blood culture obtained) or group 2 (no blood culture). Blood cultures were performed using sterile techniques and were inoculated into pediatric blood culture bottles (Bactec Peds Plus; Becton, Dickinson and Co, Franklin Lakes, NJ). A single bottle containing an enriched soybean-casein digest broth was inoculated with 0.5 to 1.0 mL of blood. Blood culture bottles were immediately taken to the laboratory and loaded into a continuous blood culture monitoring system, Bactec 9240 (Becton, Dickinson and Co), which monitored the production of carbon dioxide every 10 minutes through a fluorescent sensor. Bottles identified as positive were immediately removed from the instrument for gram staining and were subcultured for bacterial identification. The ED was notified immediately of all positive culture findings and of the gram stain results by the laboratory technicians who are on duty 24 hours a day. A designated callback nurse on each shift was responsible for informing the ED attending physician of and contacting the families or the primary physician with positive blood culture results. The standard follow-up advice for a patient with positive blood culture results was a return ED visit. However, some primary care physicians elected a follow-up visit in their offices instead.

Demographic characteristics, associated symptoms, duration and height of fever at home, ED or UCC temperature, blood culture results, disposition, ED or UCC revisits, final diagnosis, and the development of SBI were abstracted from the medical records. Additional data for patients with positive blood culture results included time to positive culture findings, date and time of follow-up visit, discharge diagnosis, disposition, and results of repeated blood culture. Serious bacterial infection was defined as meningitis, septic shock, or death due to an infectious cause within 2 weeks of the initial ED visit. Children's Hospital of Wisconsin accounts for 93% of all pediatric hospital admissions, including all critical care admissions, in the region (Children's Hospital of Wisconsin, unpublished institutional data, 2000). One of the faculty is a member of the county death review board, so it is unlikely that a case of SBI as defined would be missed.

Bacteria that were considered pathogenic included Streptococcus pneumoniae, Staphylococcus aureus, group A streptococci, Enterococcus species, Neisseria meningitides, Salmonella species, Moraxella catarrhalis, Pseudomonas species, H influenzae, Campylobacter organisms, and Escherichia coli. Bacteria that were considered contaminants were coagulase-negative Staphylococcus species, α-hemolytic streptococci, nonpathogenic streptococcus, micrococcus, Clostridium species, Corynebacterium species, other gram-positive rods, and nonpathogenic Neisseria species. Time to positive culture results was measured in hours.

Continuous variables were compared using the unpaired, 2-tailed t test, and categorical data were compared using the χ² test. The Fisher exact test was used to evaluate small sample size data. The 95% confidence interval (CI) was calculated when appropriate. The κ statistic was calculated on a random sample of 100 medical records to determine the agreement rate on inclusion and exclusion criteria between the investigators. All analyses were performed using statistical software (SPSS v10.05; SPSS Inc, Chicago, Ill).

During the study, 9241 febrile patients were identified: 2641 (29%) met the enrollment criteria and 6600 (71%) were excluded (Figure 1). Reasons for exclusion were identifiable focus of infection (58%), ED or UCC temperature of less than 39.0°C (30%), antibiotic administration in the ED or UCC or prescription for use at home or within 1 week preceding the ED or UCC visit (9%), and hospital admission at the time of the initial ED or UCC visit (3%). The demographic characteristics of enrolled patients are summarized in Table 1. Blood cultures (group 1) were obtained more frequently in younger children with higher mean temperatures and a longer mean duration of fever. Also, a larger proportion of patients in group 1 underwent urinalysis/culture and chest radiography (Table 1).

Thirty-seven patients (3%) in group 1 (n = 1202) had culture-proven OB (95% CI, 2.2%-4.2%). A higher rate of OB was observed in patients aged 13 to 24 months (4.0%) compared with those aged 2 to 6 months (1.8%). Most pathogens (84%) were S pneumoniae (Table 2). Of the 37 patients with OB, 30 (81%) returned to the ED after notification from our callback nurse, and 28 of these (93%) had another blood culture performed. Seven (25%) of these 28 patients had persistent S pneumoniae bacteremia on the repeated blood culture. Sixteen patients with OB (53%) were febrile at the time of the follow-up visit and did not have an identifiable focus of bacterial infection. Seven patients with OB had follow-up by their primary physician. No further microbiology laboratory data were available for these patients, and no adverse outcome or hospital admissions were documented on medical record review.

Thirty-nine patients (3.2%; 95% CI, 2.3%-4.4%) had a contaminated blood culture; 32 (82%) were coagulase-negative staphylococcus, 5 (13%) were α-hemolytic streptococcus, and 2 (5%) were micrococcus or other gram-positive rods. Twenty-two patients (56%) returned to the ED on notification; 17 of them (77%) had another blood culture performed and all were negative.

The hospital admission rate for patients with OB was 32% (n = 12) and for those with contaminated blood culture it was 26% (n = 10). Reasons for hospital admission included toxic or ill appearance, dehydration with persistent vomiting or diarrhea, or respiratory tract symptoms requiring continuous monitoring.

The mean ± SD time to positive culture results for pathogenic bacteria was 17.5 ± 8.4 hours and for contaminated blood culture was 40.4 ± 37.4 hours (mean difference, 23.0 hours; 95% CI, 8.0-37.2 hours). The median time from report of positive blood culture result by the microbiology laboratory to ED follow-up was 3.1 hours (range, 1.0-8.0 hours).

Two patients (0.08%; 95% CI, 0.009%-0.27%) developed SBI in our study population (n = 2641). A 23-month-old boy with gram-negative diplococci in blood samples had follow-up with his primary physician and was given oral antibiotic treatment. He received a dose of intramuscular ceftriaxone sodium on day 2 at the same physician's office for persistent fever and returned to the ED on day 3 in septic shock. His initial blood culture in the ED was positive for N meningitides, and his subsequent blood and cerebrospinal fluid culture results were negative. The second patient was a 2-month-old girl who had a blood culture positive for S pneumoniae. On her return visit to the ED she was febrile and appeared ill. A complete sepsis evaluation was performed, she had cerebrospinal fluid pleocytosis, and both her cerebrospinal fluid and repeated blood cultures were positive for S pneumoniae. Both patients recovered completely.

Seventy-seven patients (5%) from group 2 (n = 1439) who did not have a blood culture performed during the initial visit returned to the ED or UCC within 1 to 2 days for persistent fever. There was no identifiable focus of bacterial infection noted in any of these patients. Twenty-six (34%) of these patients had blood cultures performed at the second visit and all had negative results. The κ statistic for agreement on inclusion and exclusion criteria was 0.9.

In 1993, Baraff et al^{7 - 8} published practice guidelines recommending empiric antibiotic treatment for children at risk for OB. However, there have been no prospective studies to validate these recommendations, and the treatment of these children remains controversial.¹⁶ In 4 previously published prospective randomized clinical trials^{9 - 12} undertaken before the widespread use of H influenzae type b vaccines (Table 3), the combined rate of meningitis and sepsis in children treated with empiric antibiotics was 0.08% (95% CI, 0.03%-0.17%). The overall rate of SBI in those 4 studies is calculated according to our study definition of SBI being bacterial meningitis, sepsis, or death due to an infectious cause. It is computed among patients who received either parenteral or oral antibiotics during their initial visit and pending blood culture results. Therefore, the reported rate of SBI was 0.08% (95% CI, 0.03%-0.17%; 6/7485).^{9 - 12} Although the total number of patients who developed SBI was 8, 2 were from the no treatment group.¹²

In a meta-analysis, Bulloch and colleagues¹⁴ reported that the use of either oral antibiotics or intramuscular ceftriaxone did trend toward a reduced risk of serious infection; however, neither reached statistical significance (odds ratio [OR], 0.60; 95% CI, 0.10-3.49; and OR, 0.38; 95% CI, 0.12-1.17, respectively). The authors concluded that 414 patients should receive empiric antibiotic therapy to prevent 1 SBI. In a similar meta-analysis, the rate of SBI was compared in individuals with S pneumoniae bacteremia who received parenteral vs oral empiric antibiotic treatment. There was no difference in the rate of SBI between the 2 groups (OR, 1.48; 95% CI, 0.5-4.3).¹⁵ In our study, the SBI rate, based on 2 of 2641 patients, is 0.08% (95% CI, 0.009%-0.27%) and is comparable with previously published data.^{9 - 12} The Centers for Disease Control and Prevention and American Academy of Pediatrics recommendations^{17 - 18} for judicious use of antibiotics were followed in our study, with a comparable outcome to that reported in the previously mentioned trials. Even assuming that empiric treatment would be 100% effective in preventing SBI compared with our expectant approach, our findings suggest the need to treat 1250 patients to prevent an additional case of SBI.

In contrast to previous studies on OB, we excluded children with otitis media, which may have affected the rate of OB in our study.^{3 ,9 - 13 ,19} However, the risk of OB in children with otitis media does not necessarily increase their risk of developing SBI as defined in our study.²⁰ Furthermore, oral antibiotics prescribed to treat otitis media may be effective in treating bacteremia as well.⁴ Because our objective was to determine the rate of SBI in children with FWS when antibiotic therapy is reserved until blood culture results are positive, we excluded this group of patients from our analysis.

In estimating the rate of SBI, 2 of 2641 in our population (0.08%; 95% CI, 0.009%-0.27%), we strictly defined SBI as bacterial meningitis, sepsis, or death due to an infectious cause within 2 weeks of illness. We included children who did not have a blood culture performed in the ED (group 2) in the denominator when we calculated the rate of SBI. This could lead to underestimation of the true rate of SBI. However, both of the children with SBI had blood cultures performed in the ED (group 1), and there were no cases of SBI in group 2. Therefore, our estimate represents the true SBI rate in our entire population. As noted previously, Children's Hospital of Wisconsin is the only freestanding children's hospital in the state and accounts for most of the pediatric admissions in the region. In addition, all children with fever are routinely contacted by telephone within 48 hours by the ED follow-up nurse. Although the chance of losing patients to follow-up existed, it would be highly unlikely for an SBI to occur at any outside facility without involving our system.

Part of the success of our approach may be attributable to use of a continuously monitoring blood culture system²¹ with around-the-clock notification, which allows most of the pathogenic OB to be detected within 18 hours. This is consistent with the results of another recently published study.¹³ Furthermore, we have an effective callback system in the ED with a dedicated nurse in place before implementing the clinical practice protocol for febrile children aged 2 to 36 months. As a result, follow-up was achieved within 8 hours of the time of report of OB in our study population (median, 3.1 hours; quartile range, 2.4-3.4 hours).

The 3% rate of OB in our study (95% CI, 2.2%-4.2%), and the distribution of causative organisms, is consistent with recently reported data.^{13 ,19} As in another study¹³ performed since the introduction of the H influenzae type b vaccine in the United States, no cases of invasive H influenzae type b infection were reported in our population.

Because blood cultures were not performed universally in patients with a temperature between 39.0°C and 40.9°C, our estimate of the rate of OB may be biased. When we compared our practice pattern with the 1993 practice guidelines, we found that only 46% of our eligible patients had a blood culture performed in the ED. However, because our study population may be considered, according to published guidelines, at high risk for OB based on temperature alone, the anticipated effect of any such bias would be to inflate the apparent rate of OB. An additional limitation is our reliance on hospital discharge diagnosis to identify patients with fever. All patients who had blood culture performed, as listed in the laboratory database, were successfully identified by discharge diagnosis, suggesting that our diagnosis coding is consistent. It is possible that some otherwise eligible patients who did not have blood culture performed may have been missed. This would, however, lead to an underestimate of the true population at risk (denominator) and again would have the effect of overestimating the rate of OB.

The rate of contamination in our study (3.2%) is also consistent with previously published rates.^{22 - 23} This rate gains significance as we found that the rates of performing a second blood culture and rates of hospitalization were similar in patients with a pathogenic OB and with contaminated growth in blood, resulting in higher health care costs.^{24 - 26} The effect of these false-positive results must be considered as part of the decision-making process in managing FWS.^{27 - 29}

Our study provides up-to-date data on the effectiveness of a selective, expectant approach to patients with FWS. The low risk of SBI is similar to that reported with empiric antibiotic treatment. We believe that ensuring adequate follow-up and withholding antibiotic administration until culture results are known is a safe practice for children with FWS at risk for OB. Such an approach could decrease unnecessary antibiotic use, believed to be a contributory factor to the emergence of resistant S pneumoniae in certain communities across the United States.³⁰

It is believed that the widespread use of conjugate pneumococcal vaccine in the future will likely lead to a rapid and sustained drop in the rate of OB in children.³¹ The introduction of the conjugate pneumococcal vaccine may be the first step toward a new era, when the issue of empiric antibiotics and OB will finally be resolved.³²

Accepted for publication January 10, 2002.

We thank Marc H. Gorelick, MD, MSCE, and Christine M. Walsh-Kelly, MD, for their contribution to manuscript preparation and Sue C. Kehl, PhD, for her assistance with the microbiology laboratory database.

Children aged 2 to 36 months with suspected OB are at risk of developing SBIs. Bacterial meningitis and sepsis, although rare events, are the outcomes that frequently drive empiric antibiotic treatment in children with suspected OB. Previous recommendations, based on results of earlier studies, suggest empiric treatment to prevent SBI in children at risk for OB. The results of this study suggest that an expectant approach rather than unnecessary empiric antibiotic treatment is effective without any increase in adverse outcome. When blood samples taken from children with FWS and suspected OB are analyzed in a modern, continuously monitored blood culturing system, and those children can be contacted to return promptly for follow-up, then the rates of bacterial meningitis, sepsis, and death are not different from those in previously published literature that used antibiotics empirically.

Corresponding author and reprints: Subhankar Bandyopadhyay, MD, Department of Pediatrics, Medical College of Wisconsin, 9000 W Wisconsin Ave, Mailstop 677, Milwaukee, WI 53226 (e-mail: sbandy@mcw.edu).