Prevalence of Invasive Bacterial Infections in Well-Appearing, Febrile Infants

We reviewed a convenience sample of febrile, well-appearing, term infants without identifiable source of infection, bronchiolitis, or medical complexity across 75 medical centers (31 community hospitals and 44 university-affiliated hospitals) from September 2015 to December 2017. Eligible sites were recruited from hospitals that participated in the American Academy of Pediatrics Value in Inpatient Pediatrics network Reducing Excessive Variation in the Infant Sepsis Evaluation (REVISE) project.¹²  Sites ranged from small community-based hospitals to larger medical centers and freestanding children’s hospitals across the United States (Fig 1).

Full details of patient selection are described in detail elsewhere.¹²  Briefly, eligible infants were otherwise healthy, aged 7 to 60 days, had no obvious signs of focal infection, and had bacterial cultures of blood and/or cerebrospinal fluid (CSF) performed as part of their evaluation. All infants were seen in a site’s ED or inpatient setting for fever without a source (temperature ≥38.0°C) and were discharged from the site’s ED or inpatient unit. Exclusion criteria included documentation of ill appearance on presentation (defined as documentation of the patient as toxic, ill-appearing, lethargic, sick-appearing, or similar terminology¹³ ), comorbidities predisposing to severe or recurrent bacterial illness (including prematurity and genetic, congenital, chromosomal, neuromuscular, or neurodevelopmental abnormalities), a discharge diagnosis of bronchiolitis, or transfer to or from another inpatient setting. Local site investigators identified eligible subjects via medical records review, and only patients who met study criteria on the basis of manual chart review were entered into the study database. A maximum of 20 charts of eligible infants were reviewed per month at each site during the study period; sites with >20 eligible charts either selected the first 20 charts sequentially or obtained a random sampling of eligible charts for the month.

Infant age, sex, history, and laboratory obtainment data were collected as summary variables as part of the original REVISE project. Obtainment of urine, blood, and CSF bacterial testing were recorded. Microbiologic test results were reviewed by using a priori criteria and categorized as negative, growth of a contaminant organism, or growth of a pathogenic organism.^14,13  Because of limitations in the data set, bacteremia with CSF pleocytosis but negative CSF culture results could not be captured separately. Diagnosis of bacterial meningitis or bacteremia, as well as readmission to the same hospital for a missed diagnosis of bacterial meningitis or bacteremia, were also recorded. Given the lack of objective criteria for defining community hospitals,¹⁵  hospitals were classified as either a community-based and university-affiliated hospital on the basis of an individual site’s self-report.

Prevalence of bacterial meningitis, bacteremia, and IBI were estimated and compared across patient age groups and sex and between community-based and university-affiliated hospitals by using χ² analysis. All analyses were performed by using R version 3.5.1 (R Core Team, Vienna, Austria).

A total of 10 618 febrile infants met inclusion criteria, including 10 581 infants who underwent blood culture and 6747 infants who underwent CSF culture. Of these, 3530 (33%) infants were seen at community-based hospitals (Table 1). For the overall cohort, the prevalence for meningitis was 0.4% (95% confidence interval [CI]: 0.2–0.5]) and for bacteremia was 2.4% (95% CI: 2.1–2.7). Prevalence of meningitis did not differ between age groups. (Table 1) Prevalence of bacteremia was significantly higher among infants aged 7 to 30 days (3.6% vs 1.6%; P < .001).

Comparison of IBI prevalence in community-based versus university-affiliated hospitals revealed no differences for IBI (2.1% [95% CI: 1.7–2.6] vs 2.7% [95% CI: 2.3–3.1]), meningitis (0.4% [95% CI: 0.2–0.8] vs 0.3% [95% CI: 0.2–0.6]), or bacteremia (2.0% [95% CI: 1.5–2.5] vs 2.6% [95% CI: 2.2–3.0]). The most commonly detected organism among infants with bacteremia was Escherichia coli (48%) and among infants with bacterial meningitis was Streptococcus agalactiae (group B streptococci) (54%) (Table 2). No patients who were discharged and later found to have IBI were reported in this cohort. Post hoc power analysis revealed that our sample size had 80% power to detect a 1% difference in IBI between infants seen at community hospitals versus university-affiliated hospitals.

With our review of >10 000 febrile infant evaluations across 75 hospitals, we provide a contemporary estimate of IBI prevalence in community-based and university-affiliated pediatric ED and inpatient settings. We found clinically similar IBI prevalence rates among infants seen in community-based and university-affiliated hospitals. We also found that well-appearing, febrile infants aged 7 to 30 days were more likely to be diagnosed with IBI when compared with infants in the second month of life. E coli was the most common cause of bacteremia, whereas Group B streptococci was the most common cause of meningitis.

Currently, risk stratification models for febrile infants use data derived primarily from large, academic medical centers.^2,6,16  Our cohort includes 3530 infants from self-identified community-based hospitals, the location in which the bulk of children receive inpatient care in the United States.^9,16  Consequently, our prevalence estimates likely better reflect the epidemiology of bacteremia and meningitis among well-appearing, febrile infants evaluated in community-based settings than those in previous studies. Thus, the estimates from the current study can be used to inform future risk stratification models in a way that enhances their applicability to community-based acute care settings.

These estimates support previous studies from past decades in which a significant difference in bacteremia prevalence between infants in the first and second months of life was identified.¹  The prevalence of bacteremia in the first month of life in our previous systematic review and meta-analysis, derived from 11 studies and 5145 febrile infant evaluations, was 2.9%. Bacteremia prevalence was 1.6% in the second month of life on the basis of data from 5 studies and 4778 febrile infants. However, individual study quality and sample size varied widely across studies, and the resulting estimates were limited by significant heterogeneity, particularly for the febrile neonate group of studies. In the current study, we achieve comparable sample sizes to the previous meta-analysis (4031 neonates and 6587 infants in the second month of life) without concerns related to pooling data from studies of varying quality. Similar to other studies in this patient population, we detected a relatively low prevalence of meningitis in neonates and infants in the second month of life and were thus underpowered to draw further conclusions about this rare disease.

This study has limitations. First, if cultures were ordered but not successfully collected, these infants were not included in the study. Assuming these infants did not ultimately have IBI, the risk of bacteremia and meningitis may have been overestimated in our prevalence results. Additionally, identification of eligible patients may have varied across sites (including definition of well-appearing and/or presence of comorbid illnesses), leading to some heterogeneity in populations between sites. Chart abstractors received training at the beginning of the project with longitudinal reinforcement via webinars and the REVISE project e-mail Listserv to help mitigate this risk but were not assessed for interrater reliability.¹²  Some bacterial cultures could have had false-negative results, particularly when the obtained specimen was scant or suboptimal. Participating sites also tracked missed bacterial infections with return to care at the same facility, which reduces the likelihood that a delayed and/or missed diagnosis would not be included in the cohort. Not all eligible infants were sampled from participating sites every month, which could introduce a systematic bias to our cohort. This bias was mitigated by the use of standardized chart-selection techniques. A total of 15 of 75 sites had at least 1 monthly review with 20 charts (the suggested maximum); the remaining 60 sites (which contributed more than half the patients in our cohort) may have submitted only a portion of eligible cases from each month, which may also impose a systematic bias to our findings. Antibiotic pretreatment could also result in falsely negative bacterial culture results, and rates of pretreatment were not captured as part of this study. Our sample size provided adequate power to detect absolute differences in IBI prevalence of 1%; future larger-scale studies would be required to statistically confirm our observed difference in IBI prevalence of 0.6%. Finally, the definition of community-based and university-affiliated hospitals was based on self-report and not objective criteria. Given previous reports on the difficulty in defining community hospitals,^15,17  our approach is consistent with previous and existing practice.

In this contemporary study of a large cohort of well-appearing febrile infants, we provide IBI prevalence and epidemiological data that are generalizable to university-affiliated and community-based pediatric settings. The results from this study can be used to aid future clinical guidelines and risk stratification and prediction tool development.