Prevalence of and Risks for Bacterial Infections in Hospitalized Children With Bronchiolitis

We performed a single-center retrospective cohort study of children aged 1 day to 24 months hospitalized at our institution with a diagnosis of bronchiolitis from January 1, 2012 to December 31, 2019. Our institution is a free-standing children’s hospital that serves as the community children’s hospital for a large metropolitan area and as a quaternary referral center in the Western United States. Before November 30, 2014, all children receiving high-flow nasal cannula required PICU admission; after November 30, 2014 children receiving ≤2 L per minute per kilogram of high-flow nasal cannula did not require PICU admission. We identified children with bronchiolitis by querying our institutional Electronic Data Warehouse (EDW) for children with the previously validated International Classification of Disease (ICD) 9 codes (466*) and ICD-10 codes (J20*, J21*) for bronchiolitis.^12,13  Children with existing tracheostomy tubes and cystic fibrosis were excluded from analysis, as were children who were admitted to the cardiac ICU or the NICU.

Approval for this study was obtained and a waiver of informed consent granted from the institutional review board at our institution.

We collected demographic, laboratory, imaging, microbiologic, antibiotic, clinical outcome, and diagnostic code data from our EDW and direct chart review for selected patient populations. Laboratory data included laboratories collected from 24 hours before admission through hospital day 5.

A single author (N.C.) reviewed chest x-ray (CXR) interpretations provided by pediatric radiologists at our institution to categorize CXR results as definite, possible, or unlikely to have bacterial pneumonia based on standard verbiage used in the CXR read (Supplemental Table 5).¹⁴ 

cSBI – children diagnosed with bacterial pneumonia (hereafter referred to as pneumonia), bacteremia, UTI and/or meningitis.

Pneumonia – (1) positive respiratory culture from an endotracheal tube (defined as 3+ or 4+ growth or greater than 10⁴ CFU/mL of a pathogenic organism) (Supplemental Table 6) or a positive blood culture obtained during the first 5 days of the child’s stay and a CXR categorized as definite or possible pneumonia, (2) a positive pleural fluid culture, or (3) a CXR categorized as definite pneumonia.

Bacteremia – a positive blood culture, excluding likely contaminants (Supplemental Table 7), obtained during the first 5 days of the child’s stay. For blood cultures with suspected contaminants, charts were reviewed to confirm a repeat blood culture was negative and/or the child did not receive treatment of the organism identified.

UTI – (1) positive urine culture (defined as growth greater than 10⁵ CFU/mL) obtained during the first 5 days of the child’s stay and urinalysis with evidence of pyuria (≥5 white blood cells or +leukocyte esterase) or +nitrites per American Academy of Pediatrics guidance.¹⁵  As recent literature suggests lowering the culture threshold for UTI to 10⁴ CFU/mL, we performed a sensitivity analysis to identify additional cases of UTI with the 10⁴ CFU/mL threshold.¹⁶ 

Meningitis – (1) a positive cerebrospinal fluid (CSF) culture, (2) positive BioFire FilmArray Meningitis Encephalitis Panel result for a bacterial pathogen, or (3) ICD 9 or 10 code for meningitis with direct chart review identifying culture or polymerase chain reaction (PCR) results obtained from hospitals outside our institution. Children admitted with a primary diagnosis of meningitis who were diagnosed with a viral infection, without symptoms of bronchiolitis, were excluded.

For children with an ICD 9 or 10 code for a cSBI but lacking diagnostic study data in our EDW, we performed chart review to identify diagnostic studies performed outside our institution to confirm cSBI status. Children with multiple diagnosed cSBIs were categorized by their likely source of primary infection in the order of meningitis, pneumonia, UTI, and bacteremia.

Analysis occurred at the level of hospital admission; thus, a child would be included in this study for each admission attributable to bronchiolitis over the study period. We summarized continuous variables with medians and interquartile ranges (IQRs) and categorical variables with counts and percentages. We defined abnormal white blood cell count (WBC) as <5000 mL or ≥15 000 mL, absolute neutrophil count (ANC) ≥5000 mL, and C-reactive protein (CRP) ≥4 mg/dL based on previously published data.^17–20  We used Wilcoxon rank sum test to compare continuous variables and χ-squared or Fisher’s exact tests to compare categorical variables between children with and without cSBI. We fitted univariable and multivariable logistic regression models to assess the association between presence of any cSBI and explanatory variables of interest. Variables with a P value ≤ .10 in univariable analysis were included in a multivariable model. We report odds ratios (ORs) and 95% confidence intervals (CIs). We used the generalized variance inflation factor (GVIF) to assess multicollinearity among covariates in our multivariable model. Multicollinearity was considered tolerable if the GVIF was <2.24, which is equivalent to a conventional VIF <5. We used a 2-sided significance level of 0.05. Statistical analyses were performed using R v. 4.0.3 (R Core Team, 2020).

We identified 7871 hospital admissions for bronchiolitis occurring in 6334 children over the study period. Median age was 8.6 months (IQR 6.9–13.7), 57% were male and 22% had 1 or more chronic medical conditions; 40% were admitted to the PICU. The overall prevalence of cSBI by our definition was 4.2% (333 of 7871). Children with and without cSBI did not differ with regards to age, sex, or race and ethnicity. PICU admission (69% vs 38%, P < .001) and invasive mechanical ventilation (44% vs 11%, P < .001) occurred more frequently in those with a cSBI. Children with cSBI were also more likely to have 1 or more complex chronic conditions (CCC), as defined by Feutdner et al.¹⁷  Children with cSBI had significantly longer hospital and PICU lengths of stay (LOS) (Table 1).

By our cSBI definitions, the occurrence of individual cSBIs included 277 (3.5%) cases of pneumonia, 31 (0.4%) cases of bacteremia, and 23 (0.3%) cases of UTI. Meningitis was rare, occurring in 2 children; 1 was infected with Salmonella and another Haemophilus influenzae. In both cases, the child developed clinical symptoms of meningitis after admission for symptoms consistent with bronchiolitis. Table 2 summarizes demographic characteristics by cSBI. In terms of workup across the entire cohort, 4092 (52%) received a CXR, 2013 (26%) had blood cultures, 1862 (24%) had urine studies, and 699 (9%) had CSF studies. Tables 3 and Supplemental Table 8 further summarize diagnostic testing. Pneumonia was diagnosed by culture in 62 children (22% of pneumonia cases) and by CXR findings alone in 215 (78% of pneumonia cases). In children with pneumonia diagnosed by culture, 9 children had a positive blood culture, 10 children had a positive pleural fluid culture, 1 child had both a positive blood and pleural fluid culture, 43 children had a positive respiratory culture, and 2 children had both a positive blood and respiratory culture. Supplemental Table 6 lists the organisms recovered from cultures. There were 635 children (8% of all children) with a “possible” CXR but no culture criteria to meet our pneumonia definition. In our sensitivity analysis utilizing a urine culture threshold of 10⁴ CFU/mL, we identified an additional 4 cases of UTI, bringing the total count to 27 (0.3%). Fifteen children met diagnostic criteria for multiple cSBIs: 9 with pneumonia and bacteremia, 3 with UTI and bacteremia, and 3 with pneumonia and UTI.

Children with cSBI were more likely to have a respiratory virus identified than children without cSBI (70% vs 60.1%, P < .001, Table 3). Additionally, human metapneumovirus (hMPV), influenza, adenovirus, and parainfluenza infection occurred more frequently in children with a cSBI. Compared with children without a cSBI, children with a cSBI had CRP values with median 5.9 mg/dL (IQR 1.6–13.2) vs 2.4 mg/dL (IQR 0.8–6.2), P value < .001 (Table 3). The maximum WBC count was similar between the 2 groups. Procalcitonin was not routinely collected throughout the study period and therefore was not included in this analysis.

Table 4 summarizes the results of univariate and multivariate logistic regression models. In the univariate models, PICU admission, need for noninvasive positive pressure ventilation or invasive mechanical ventilation, WBC, ANC, CRP, number of CCCs, and virus type were all associated with cSBI. In multivariable models, invasive mechanical ventilation (odds ratio [OR] 2.53, 95% confidence interval [CI] 1.78–3.63), an elevated CRP ≥4 mg/dL (OR 2.2, 95% CI 1.47–3.32), presence of CCCs (OR 1.67, 95% CI 1.22–2.25), and admission to the PICU (OR 1.46, 95% CI 1.02–2.07) remained associated with cSBI after adjusting for other variables. GVIF of all variables included in multivariable models was less than 2.24, indicating no evidence of multicollinearity.

We performed a retrospective cohort study to determine the prevalence of cSBI in children hospitalized with bronchiolitis and to identify potential risk factors associated with cSBI. We found that cSBI occurred in 4.2% of children admitted with bronchiolitis. Pneumonia was the most common cSBI with 3.7% of children meeting our definition of pneumonia. Invasive mechanical ventilation, elevated CRP, presence of CCCs, and admission to the PICU were all associated with cSBI.

Our rates of cSBI are similar to those found in other studies and subsets of children with bronchiolitis. Byington et al studied cSBI in febrile infants <90 days presenting to our institution between 1996 and 2002 and found a cSBI prevalence of 4.2% with low rates of bacteremia and meningitis, similar to our cohort of children admitted between 2012 and 2019.²¹  Kupperman et al found a cSBI rate of 1.9%, all from UTI, in children with bronchiolitis presenting to emergency departments.²²  Although the rate of UTI was higher than our cohort, a recent systematic review examining concurrent UTI utilizing a combined culture and urinalysis criteria similar to ours noted a UTI rate of 0.8%, more similar to our findings.²³  In one of the largest studies of inpatient bronchiolitis, Wilson et al found an overall cSBI rate of 40.9%, with 7.6% meeting their definition of pneumonia and 5.6% diagnosed with bacteremia or sepsis. This rate is significantly higher than our study, however, Wilson et al determined cSBI rate based on ICD 9 codes rather than through analysis of culture data, imaging findings, and chart review and included acute otitis media as a cSBI, likely contributing to these differences.⁵ 

There is minimal literature investigating factors associated with cSBI in bronchiolitis. Kneyber et al studied children with bronchiolitis who required PICU admission and found no difference in WBC and CRP levels between those with negative and positive blood or respiratory cultures.²⁴  However, we found that CRP ≥4 mg/dL was associated with cSBI. The smaller sample size in the study by Kneyber et al likely explains why they did not identify significant differences. Our finding of the association with CRP and bacterial infection fits with febrile infant literature noting CRP as one variable that can differentiate bacterial from nonbacterial infection.^20,21  We also found that hMPV, influenza, and parafinfluenza viruses, but not RSV, were associated with cSBI. This is similar to results from studies examining children cared for in emergency rooms or PICUs that found no association between RSV infection and cSBI.^25,26  The large sample size of this study likely allowed us to detect the association between several other viruses and cSBI. We also noted an association between CCC and cSBI. Prior work notes an increased prevalence of UTI, sepsis, and pneumonia in pediatric patients with CCC in general.^27–29  The association we found could be because of causative interactions between a CCC and a viral infection, increased baseline risk of a bacterial infection independent of a viral infection, or both the bacterial and viral infection being coincidentally diagnosed as a type of indication bias if clinicians have a lower threshold to obtain diagnostic tests in children with CCC. However, our careful review and use of strict UTI, bacteremia, and pneumonia diagnostic criteria reduced this bias by avoiding incidental findings of potential contaminant organisms or equivocal results (eg, asymptomatic bacteriuria), leading to a false cSBI classification.

Identifying cSBI in patients with bronchiolitis can be difficult and lead to over testing and treatment.^7,30  Use of a targeted set of diagnostic tests with quick turnaround times could limit unnecessary testing and antibiotics. Our results suggest that when a patient’s clinical status (illness course, level of support, underlying medical conditions) raises clinical suspicion for a cSBI, a reasonable diagnostic approach would include a CXR and a CRP given the higher prevalence of bacterial pneumonia compared with other cSBIs and lower diagnostic yield for bacteremia, UTI, and meningitis testing. Although not assessed in our study because of low rate of use during the study period at our institution, procalcitonin is another potential first line test shown to identify children at risk for pneumonia.^31–33  Additional testing, such a urinalysis, with reflex to urine culture and blood culture could be considered based on the results of the CXR and CRP, clinical context, and individual risk factors for a UTI or blood stream infection.^14,34  Future study to validate and refine the cSBI risk factors we found can further inform which patients need additional cSBI diagnostics and which tests are highest yield.

There are some limitations to our study. First, the single center design may limit generalizability and could reflect local disease prevalence and epidemiology, although, when compared with other studies, our cSBI prevalence appears similar.^2–5  Further, in addition to being the community pediatric hospital for an urban area, our hospital is also a quaternary referral center that may preferentially admit children who are sicker with higher rates of cSBI and need for PICU care. Second, given the inherent limitations of a retrospective chart review, we may have missed some children who were diagnosed with cSBI outside our institution. We mitigated this by reviewing ICD 9 and 10 codes for cSBI and direct chart review for children without culture data in the EDW. Reliance on chart review also limits our ability to account for confounders such as a child’s clinical appearance or pace of illness that may affect the decision to obtain diagnostic tests, such as viral studies, and may lead to indication bias in our associations between test results and cSBI. There is also potential that some of the children categorized with a cSBI developed the condition after admission. We intentionally limited our study period to the first 5 days of the child’s admission in an attempt to eliminate hospital acquired conditions. Additionally, it is possible undiagnosed congenital immunodeficiencies may contribute to a higher risk for cSBI, which we were unable to evaluate in our study because of limitations of chart review and/or lack of immunologic evaluation in our cohort. However, immunodeficiency diagnoses are quite rare and would be unlikely to impact the outcomes in a study this large. Finally, our cohort did not include data from during or after the coronavirus disease 2019 pandemic. It is possible that cSBI prevalence changed during this period as recent work noted changes in overall bronchiolitis hospitalizations.³⁵  Future multicenter studies are needed to understand how the coronavirus disease 2019 may change cSBI prevalence.

A major limitation to studying bacterial coinfection in cohorts of children with bronchiolitis is accurately identifying bacterial infections. A strength of our study was the utilization of gold standard diagnostic tests and definitions and chart review to accurately identify blood stream infections (blood culture) and urinary tract infections (urinalysis and urine culture). Assigning bacterial pneumonia status is more difficult as there is no single “clinical definition” and more definitive diagnostic tests, such as pleural fluid culture or bronchial lavage fluid, culture are infrequently obtained. To minimize false-positive classification of pneumonia, we used a clinically relevant and feasible definition based on graded radiographic evidence and available microbiologic data to assign bacterial pneumonia status.¹⁴  It is possible this strategy underestimates the true prevalence of bacterial pneumonia, as some children in our cohort had “possible” pneumonia in CXR interpretations but lacked culture findings that met our definition of pneumonia. However, this combined radiographic and microbiologic approach, as opposed to reliance on diagnostic codes, is based on the approach used in the EPIC study, a large, multisite prospective cohort study of children hospitalized with pneumonia.³⁶  The EPIC study identified a viral-bacterial coinfection in 7% of all children. This rate compares similarly to our pneumonia rate of approximately 4% and suggests our pneumonia definition does not lead to an excess of false positive bacterial pneumonia classifications.³⁶ 

In our study, we found that cSBI occurs in 4.2% of children hospitalized with bronchiolitis. Pneumonia was the most common cSBI. The need for invasive mechanical ventilation, elevated CRP, presence of CCCs, and admission to the PICU are associated with cSBI in children hospitalized with bronchiolitis.