OBJECTIVES. The goals were to describe the (1) frequency of sepsis evaluation and empiric antibiotic treatment, (2) clinical predictors of management, and (3) serious bacterial illness frequency for febrile infants with clinically diagnosed bronchiolitis seen in office settings.
METHODS. The Pediatric Research in Office Settings network conducted a prospective cohort study of 3066 febrile infants (<3 months of age with temperatures ≥38°C) in 219 practices in 44 states. We compared the frequency of sepsis evaluation, parenteral antibiotic treatment, and serious bacterial illness in infants with and without clinically diagnosed bronchiolitis. We identified predictors of sepsis evaluation and parenteral antibiotic treatment in infants with bronchiolitis by using logistic regression models.
RESULTS. Practitioners were less likely to perform a complete sepsis evaluation, urine testing, and cerebrospinal fluid culture and to administer parenteral antibiotic treatment for infants with bronchiolitis, compared with those without bronchiolitis. Significant predictors of sepsis evaluation in infants with bronchiolitis included younger age, higher maximal temperature, and respiratory syncytial virus testing. Predictors of parenteral antibiotic use included initial ill appearance, age of <30 days, higher maximal temperature, and general signs of infant distress. Among infants with bronchiolitis (N = 218), none had serious bacterial illness and those with respiratory distress signs were less likely to receive parenteral antibiotic treatment. Diagnoses among 2848 febrile infants without bronchiolitis included bacterial meningitis (n = 14), bacteremia (n = 49), and urinary tract infection (n = 167).
CONCLUSIONS. In office settings, serious bacterial illness in young febrile infants with clinically diagnosed bronchiolitis is uncommon. Limited testing for bacterial infections seems to be an appropriate management strategy.
Practice guidelines have recommended laboratory testing and empiric antibiotic treatment for selected febrile infants ≤3 months of age with no identifiable focus of infection.1 Guidelines for laboratory testing and antibiotic treatment for infants with bronchiolitis are less clear, stating that “antibacterial medications should be used only in children with bronchiolitis who have specific indications of the coexistence of a bacterial infection.”2 Most studies have shown relatively low rates of serious bacterial illness (SBI) in children evaluated in emergency departments or hospitalized with fever and bronchiolitis or respiratory syncytial virus (RSV) infection, with SBI being identified in 0% to 10% of cases.3–11 Infants ≤2 months of age have the greatest risk of associated bacterial infections, particularly urinary tract infections (UTIs).3,6,11
Sepsis evaluation prolongs stays and significantly increases costs for infants hospitalized with bronchiolitis.4 Results from published studies of patients with bronchiolitis drawn from hospitals and emergency departments may not be generalizable to pediatric office settings, because they may represent sicker subsets of infants or groups for whom follow-up care presents challenges that are less prevalent in primary care practice. Research has shown that office-based physicians use less laboratory testing and less-aggressive therapies when treating febrile infants than recommended in guidelines developed to address infants treated in emergency departments.12,13 No studies of the office-based care of febrile infants with clinically diagnosed bronchiolitis have been published, leaving primary care practitioners without generalizable guidance for management.
This report from the Pediatric Research in Office Settings (PROS) Febrile Infant Study addresses the following questions. (1) What are the frequency and clinical predictors of sepsis evaluation of febrile infants with clinically diagnosed bronchiolitis seen in the office setting? (2) What are the frequency and clinical predictors of empiric antibiotic treatment? (3) What is the frequency of associated SBI among patients with bronchiolitis who were tested and treated?
PROS is a practice-based research network of the American Academy of Pediatrics (AAP). Previously published articles described the methods of the PROS Febrile Infant Study in detail.12,14–16 The PROS Febrile Infant Study involved 573 practitioners from 219 practices in 44 states, the District of Columbia, and Puerto Rico. Practitioners in the study were representative of the AAP general pediatric membership at the time (as determined by the 1995 Periodic Survey of Fellows17), except that fewer practitioners practiced in inner-city areas (7.3% vs 12%; P < .001). In addition, study practitioners were slightly older, more likely to be white, and less likely to be female and Asian American.14
The Febrile Infant Study was a prospective cohort study of febrile infants ≤3 months of age conducted from February 28, 1995, to April 25, 1998. Eligible infants had temperatures of ≥38°C in the office or in the preceding 24 hours at home and were previously healthy, without major comorbidity. Clinicians treated infants according to their customary practice; therefore, informed consent from subjects was not required. The Committee on Human Research of the University of California, San Francisco, approved the study protocol. Practitioners and office staff members recorded clinical and demographic data on standardized study forms. The study protocol required practitioners to document initial physical examination findings, diagnostic impressions, and assessment of severity of illness before obtaining radiograph or laboratory test results. Practitioners and office staff members also collected outpatient and inpatient follow-up information on study forms. PROS staff members at the AAP and at the University of California, San Francisco, assessed data quality and edited missing or erroneous data after contacting PROS practice coordinators regarding problematic information, as described previously.14,15 The final group included 3066 infants who met the eligibility criteria, of 3131 consecutively enrolled infants.
Practitioners were required to keep a log of patients who were enrolled and patients who were eligible but not enrolled. On the basis of this document, selection bias regarding infant age, gender, payment source, race, and ethnicity was assessed. Fifty-four percent of practices had perfect enrollment (ie, they enrolled all eligible patients in a specified time period). Of the 12 practices that missed patients, 58% (n = 7) missed only 1 or 2 eligible patients. In addition, enrolled and nonenrolled infants did not differ with respect to standard demographic or clinical features.
Participating practitioners made clinical diagnoses before obtaining laboratory results, and they recorded the diagnoses on the study form as initial clinical impressions. The Febrile Infant Study instruction manual, which was provided as a guide to clinician study participants, defined bronchiolitis as follows: “an infection of the bronchioles characterized by wheezing, tachypnea, fever, and cough, and is usually associated with respiratory viruses, in particular RSV. Conclusive diagnosis includes isolation of RSV from nasopharyngeal washings or positive RSV antigen. Parainfluenza A and B are also common causes.” The study instruction manual also provided definitions of other diseases, including bacterial and viral pneumonia, meningitis, bacteremia, and UTI, as a guide to clinicians. The study guide prompted clinicians to make a specific initial diagnosis, and fever was coded as “fever without recognizable illness,” to reinforce the idea of choosing a specific diagnosis when possible.
Independent Predictor Variables
Predictor variables, which were selected on the basis of association with severity of febrile illness in previous studies, were recorded on paper forms by PROS practitioners or their staff members.12,18,19 Demographic variables included age (in days), gender, ethnicity, and type of insurance. History variables included presence of ill family members and perinatal risk factors. Physical examination variables included patient's initial appearance (coded as mildly, moderately, or severely ill), maximal temperature, findings of cyanosis, dehydration, respiratory distress, inattentiveness, abnormal cry, inconsolability, or no smile, and presence of multiple signs of infant distress. Laboratory variables included abnormal white blood cell counts (<5000 cells per μL or ≥15 000 cells per μL), abnormal urinalysis results (dipstick test positive for leukocyte esterase or nitrite or >5 white blood cells per high-powered field), abnormal chest radiograph findings, and positive RSV test results.
Dependent Outcome Variables
We examined 3 main outcome variables for the subset of febrile patients with bronchiolitis, compared with febrile infants without clinically diagnosed bronchiolitis, including the frequency of complete sepsis evaluation (urine, blood, and cerebrospinal fluid cultures) and use of parenteral antibiotic therapy (both markers for most aggressive management) and the incidence of SBI, including the dichotomous variables UTI, bacteremia, meningitis, and combined bacteremia/meningitis. We then developed statistical models to identify clinical predictors for complete sepsis evaluation and for use of parenteral antibiotic therapy.
We used Stata 8.2 (Stata, College Station, TX) for all statistical analyses. We performed bivariate analyses of patient characteristics, clinical presentation, and laboratory testing for patients with and without bronchiolitis. We used logistic regression analyses to estimate the effect of a clinical diagnosis of bronchiolitis on management decisions, with adjustment for age, temperature, initial ill appearance, signs of infant distress, history of ill family members, and diagnosis of otitis media. We then entered potential predictor variables into backward, stepwise, multivariate, logistic regression analyses to identify significant predictors of complete sepsis evaluations and use of parenteral antibiotic therapy among infants with clinically diagnosed bronchiolitis. We performed all logistic regression analyses by using the “cluster” option in the statistical software, to adjust for the effect of infants enrolled by the same practitioner. We determined the goodness of fit of the logistic model by using the Hosmer-Lemeshow method.20 We assessed the discrimination of the models by using the c statistic, equal to the area under the receiver operating characteristic curve.20
Patient Characteristics and Clinical Presentation
Practitioners made an initial clinical diagnosis of bronchiolitis (before obtaining any test results) for 218 (7%) of 3066 infants. A total of 182 (83.5%) of 218 patients in the clinically diagnosed bronchiolitis group had respiratory distress, chest-related findings (including wheezing, retractions, rhonci, rales, decreased breath sounds, and prolonged expiration) or a cough, compared with 337 (12%) of 2848 patients in the group without bronchiolitis, which suggests that clinicians appropriately followed the study manual guidelines in making the clinical diagnosis of bronchiolitis. The study protocol asked practitioners to write in clinically important physical findings but did not require documentation of wheezing specifically; therefore, we could not use this specific finding to check the accuracy of the clinical bronchiolitis diagnosis.
Table 1 shows the patient characteristics and clinical presentation of febrile infants with and without bronchiolitis. Patients with bronchiolitis were significantly older (mean age: 8.1 vs 6.9 weeks; P < .001) and were more likely to have Medicaid insurance. Clinical history more often included vomiting, ill household members, and fever duration of ≥24 hours, compared with findings for febrile infants without bronchiolitis. Physical examination findings associated with bronchiolitis included fewer patients with high fever, more patients who appeared moderately or very ill, and more patients with findings suggesting serious illness, as well as a trend toward increased signs of infant distress. As expected, respiratory distress, chest-related findings, and otitis media were all associated with bronchiolitis.
Table 2 shows the frequency and results of laboratory testing for patients with and without clinically diagnosed bronchiolitis. Patients with bronchiolitis were significantly less likely to have urine testing (risk ratio: 0.6), abnormal urinalysis results (risk ratio: 0.2), cerebrospinal fluid cultures (risk ratio: 0.5), and complete sepsis evaluations (risk ratio: 0.5). They were significantly more likely to have chest radiographs (risk ratio: 2.8), abnormal chest radiograph findings (risk ratio: 3.1), RSV testing (risk ratio: 7.7), positive RSV test results (risk ratio: 2.9), oxygen saturation testing (risk ratio: 6.4), and abnormal oxygen saturation results. There were no differences in rates of complete blood counts, complete blood count results, or rates of blood cultures.
Eighty-eight percent of patients (192 of 218 patients) with clinically diagnosed bronchiolitis had abnormal chest radiograph results, respiratory distress, significant chest-related findings on examination, or a cough. RSV testing was performed for 102 patients with bronchiolitis, and results were positive for 70 patients. In the group without clinically diagnosed bronchiolitis, 41 patients were subsequently found to have positive RSV test results. Practitioners clinically diagnosed patients in this group as having upper respiratory tract infection in the majority of cases. Other initial impressions in the RSV-positive, nonbronchiolitis group included otitis media, pneumonia, pharyngitis, and fever without source. These clinical syndromes are frequently seen in infants with RSV infections, and the findings suggest that clinicians appropriately sorted RSV-infected patients on the basis of symptoms.
Table 3 describes the care of febrile infants with and without bronchiolitis. Patients with bronchiolitis were significantly less likely to undergo complete sepsis evaluations (14% vs 28%; P < .001). This difference in management persisted after adjustment for age, temperature, initial ill appearance, signs of infant distress, history of ill family members, and diagnosis of otitis media. As a group, however, infants with bronchiolitis were significantly more likely to be hospitalized (50% vs 34%; P < .001) and received more follow-up visits (mean: 1.9 vs 1.7 visits; P < .001). The overall use of antibiotics did not differ between the groups; however, patients with bronchiolitis were more likely to receive orally administered antibiotics (26% vs 11%; P < .001) and less likely to receive parenterally administered antibiotics. The increased use of orally administered antibiotics observed for patients with bronchiolitis persisted after adjustment for concurrent otitis media, age, temperature, ill appearance, signs of infant distress, and history of ill family members. After the same logistic regression adjustment, patients with bronchiolitis were significantly less likely to receive parenterally administered antibiotics (33% vs 45%; P < .001).
Table 4 shows multivariate predictors of complete sepsis evaluation in clinically diagnosed bronchiolitis. Significant predictors included younger age, higher maximal temperature, and RSV testing (regardless of the result). The model for sepsis evaluation had good discrimination (c = 0.80), with an excellent fit (Hosmer-Lemeshow χ82 = 7.05; P = .53).
Table 5 demonstrates multivariate predictors of parenteral antibiotic therapy for clinically diagnosed bronchiolitis. Significant predictors included initial ill appearance, age of <30 days, higher maximal temperature, and general signs of infant distress. Patients with bronchiolitis with specific signs of respiratory distress were less likely to receive parenterally administered antibiotics. The model for parenteral antibiotic treatment also had good discrimination (c = 0.78) and fit (Hosmer-Lemeshow χ82 = 3.0; P = .93).
Table 6 outlines the prevalence of SBI in febrile infants with and without bronchiolitis. There were no cases of UTI, bacteremia, or meningitis in any of the infants with cultures in the clinically diagnosed bronchiolitis group (0 of 125 patients; 95% confidence interval [CI]: 0%–2.4%). The risk differences were statistically significant for UTI (P = .001), bacteremia and bacterial meningitis combined (P = .031), and any SBI (P < .001).
Final Primary Diagnoses
Table 7 shows the final primary diagnoses for patients with clinically diagnosed bronchiolitis. The initial clinical impression of bronchiolitis was consistent with the final diagnosis for 78%. Because hospitalized infants with bronchiolitis commonly have infiltrates noted on chest radiographs,21 it is not surprising that pneumonia was the final diagnosis for 11% of our patients. Similarly, upper respiratory tract symptoms and otitis media are frequently associated with bronchiolitis22 and thus are not unexpected as final diagnoses for patients with clinically diagnosed bronchiolitis.
Practitioners performed a full sepsis evaluation, including urine, blood, and cerebrospinal fluid cultures, one half as often for patients with bronchiolitis, compared with other febrile infants. They were most likely to test patients <1 month of age and those with high fever. RSV testing (regardless of the result) also correlated with sepsis evaluation. We postulate that practitioners performed complete laboratory testing (including RSV testing) for infants they deemed to be high risk. In addition, they might have included RSV testing for infection-control purposes among infants they planned to admit to the hospital. Patients with bronchiolitis were almost twice as likely to be hospitalized but were less likely to receive parenteral antibiotic treatment. Predictors for parenteral antibiotic therapy included ill appearance, younger age, higher fever, and signs of infant distress. No cases of UTI, bacteremia, or meningitis were identified in our study of febrile infants with bronchiolitis diagnosed clinically at presentation in an office setting.
Previous studies examining the risk of SBI in bronchiolitis were all conducted in either emergency departments or hospitals. This is the first large-scale study to examine primary care treatment and associated bacterial infections among febrile infants with clinically diagnosed bronchiolitis. Most published studies reported a low risk of bacterial infection, but rates varied from 0% to 10%, with the highest risk being in febrile infants <2 months of age.3–6,8–11 UTIs have been the most commonly reported infections in infants with bronchiolitis and/or RSV infection, with rates ranging from 0% to 6.1%.3–6,8–11 Although bacteremia was rare (<1%) in most studies, a report by Levine et al6 documented bacteremia in 3.7% of a subset of RSV-infected infants who were <28 days of age and were all evaluated in pediatric emergency departments. Reports of meningitis complicating bronchiolitis have been extremely rare, with only 2 cases described in all published series; both cases involved hospitalized patients.4,8 Practice guidelines for the care of febrile infants advocate extensive laboratory testing for those ≤3 months of age.1 Despite the reported low risk of bacterial infections, no consensus exists regarding the care of febrile infants with bronchiolitis who are <3 months of age. The 2006 AAP guidelines for management of bronchiolitis acknowledge low rates of bacterial infection and recommend limiting antibiotic treatment to documented cases of bacterial coinfection; however, the guidelines make no clear recommendations regarding indications for laboratory testing.2 Our large, primary care-based study demonstrates that the rate of bacterial infection is low for practitioner-diagnosed bronchiolitis seen in office settings. The office practitioner approach, as shown in this study, of limited testing for bacterial infections in infants with bronchiolitis seems appropriate. PROS practitioners were most likely to test infants <28 days of age with fever of >38.5°C, which is the group at highest risk for infection.
This study has several limitations. Variability might have occurred in the criteria practitioners used to diagnose bronchiolitis clinically, despite the definition in the study guide, especially because PROS clinicians were not required to document the presence or absence of wheezing as a physical finding for the study. If this is the case, however, then it likely reflects variability in how bronchiolitis is diagnosed in community settings, thus enhancing the generalizability of the results. Our study data support this assertion. Eighty-eight percent of cases (192 of 218 cases) of clinically diagnosed bronchiolitis involved abnormal chest radiograph findings, respiratory distress, significant chest-related findings on examination, or cough. In addition, the majority of cases (78%) initially defined as clinically diagnosed bronchiolitis had this diagnosis confirmed as a final primary diagnosis after inclusion of all laboratory results and follow-up information.
PROS practitioners did not enroll all eligible infants in the Febrile Infant Study. However, as noted previously, eligible infants who were not enrolled did not differ significantly from the infants in the study, other than being an average of 4 days older.12 Therefore, recruitment bias is unlikely. We think that our findings are representative of the management and outcomes of clinically diagnosed bronchiolitis in most pediatric primary care settings, except that our results included limited data from urban inner-city practice sites. Our results may not be generalizable to emergency department and hospital settings, where the patient population may be sicker and thus may have different rates of bacterial complications.
This study was conducted almost a decade ago and thus might reflect less accurately the current office care of febrile infants with bronchiolitis. We think this is unlikely, however, because, other than the increasing availability of rapid RSV tests, no significant bronchiolitis management strategies have been introduced in the past 10 years. Furthermore, regular use of the conjugate pneumococcal vaccine since June 2000 would not significantly change the incidence of SBI in our study population, because of the already low rate of pneumococcal infections seen in this study.12
Our study might have missed cases of SBI in patients with clinically diagnosed bronchiolitis, because the majority of patients did not undergo a full sepsis evaluation. However, because the subjects who were tested were likely to have been at higher risk than those who were not tested, this issue becomes one of completeness of ascertainment and sample size, rather than sampling bias. We consider it unlikely that cases of meningitis or bacteremia were missed, because the study design required follow-up monitoring for all patients and no untoward events occurred in patients not tested or treated, but we cannot exclude the possibility of spontaneously resolving bacteremia. A previous report from this study suggested that several additional UTIs would have been diagnosed in the patients with bronchiolitis if all had had cultures. That report found that there were only 2 infants with UTIs diagnosed after the initial visit among 807 infants who were not initially tested or treated with antibiotics (0.24%; 95% CI: 0.03%–0.89%), whereas ∼61 UTIs (7.6%) would have been predicted. This low rate of late UTIs was presumably attributable to spontaneous resolution of most undiagnosed UTIs,15 which also might have occurred among some of the infants with bronchiolitis who did not receive urine cultures. Finally, our study did not have sufficient sample size to be definitive regarding some infrequent but serious infections, although our results suggest that SBI complications are rare for patients with clinically diagnosed bronchiolitis.
This study is consistent with previously published data reporting low rates of SBI in patients with clinically diagnosed bronchiolitis. Our study adds to current knowledge in that patients with bronchiolitis treated in primary care settings seemed to have even lower rates of SBI, compared with infants treated in emergency departments and hospitals. The fact that PROS practitioners performed a full sepsis evaluation one half as often for patients with bronchiolitis, compared with other febrile infants, seems justified. Management guidelines for young febrile infants with bronchiolitis who are seen in office settings should be updated to reflect the low rates of bacterial infection seen in these infants.
- Accepted February 22, 2008.
- Address correspondence to Lynn M. Luginbuhl, MD, MS, Middlebury Pediatric and Adolescent Medicine, 1330 Exchange St, Suite 201, Middlebury, VT 05753. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
What's Known on This Subject
Febrile infants with bronchiolitis who are hospitalized or evaluated in emergency departments have low reported rates of serious bacterial illness. Infants who are ≤2 months of age have the greatest risk of associated bacterial infections, particularly urinary tract infections.
What This Study Adds
Patients with bronchiolitis treated in primary care settings received a full sepsis evaluation one half as often as other febrile infants and seemed to have even lower rates of serious bacterial illness than those treated in emergency departments and hospitals.
- ↵Baraff LJ, Schriger DL, Bass JW, et al. Practice guideline for the management of infants and children 0 to 36 months of age with fever without source. Pediatrics.1993;92 (1):1– 12
- ↵Lieberthal AS, Bauchner H, Hall CB, et al. Diagnosis and management of bronchiolitis. Pediatrics.2006;118 (4):1774– 1793
- ↵Levine DA, Platt SL, Dayan PS, et al. Risk of serious bacterial infection in young febrile infants with respiratory syncytial virus infections. Pediatrics.2004;113 (6):1728– 1734
- ↵Titus MO, Wright SW. Prevalence of serious bacterial infections in febrile infants with respiratory syncytial virus infection. Pediatrics.2003;112 (2):282– 284
- ↵Young PC. The management of febrile infants by primary-care pediatricians in Utah: comparison with published practice guidelines. Pediatrics.1995;95 (5):623– 627
- ↵Bergman DA, Mayer ML, Pantell RH, Finch SA, Wasserman RC. Does clinical presentation explain practice variability in the treatment of febrile infants? Pediatrics.2006;117 (3):787– 795
- ↵Newman TB, Bernzweig J, Takayama JI, Finch SA, Wasserman RC, Pantell RH. Urine testing and urinary tract infections in febrile infants seen in office settings: the Pediatric Research in Office Settings' Febrile Infant Study. Arch Pediatr Adolesc Med.2002;156 (1):44– 54
- ↵American Academy of Pediatrics, Division of Heath Services Research. Periodic Survey of Fellows 32. Elk Grove Village, IL: American Academy of Pediatrics; 1995
- ↵McCarthy P, Jekel J, Stashwick C. Further definition of history and observation variables in assessing febrile children. Pediatrics.1981;67 (5):687– 693
- ↵Stata. Stata Statistical Software Manual: Release 5.0. College Station, TX: Stata; 1997
- ↵Rudolph CD, Rudolph AM, Hostetter MK, Lister G, Siegel NJ, eds. Rudolph's Pediatrics. 21st ed. New York, NY: McGraw-Hill; 2003
- Copyright © 2008 by the American Academy of Pediatrics