Bronchiolitis Management and Unnecessary Antibiotic Use Across 3 Canadian PICUs Free

We conducted a retrospective cohort study including all patients ≤2 years old with a first episode of bronchiolitis admitted to 3 Canadian tertiary university-affiliated PICUs: PICU 1 (900 overall admissions per year), PICU 2 (1000 overall admissions per year), and PICU 3 (750 overall admissions per year). Study participants were admitted over 2 seasons, from November 1, 2016 to April 30, 2017 (season 1) and from November 1, 2017 to April 30, 2018 (season 2). These periods were chosen as they are the typical seasons for bronchiolitis in Canada. Trained investigators (1 at each site) screened admission diagnosis for all PICU patients admitted during the study period and reviewed medical charts to confirm the diagnosis of bronchiolitis based on the 2014 American Academy of Pediatrics (AAP) bronchiolitis guidelines.¹  The clinical criteria for admission to the PICU were similar across the 3 sites. They included the following: severe respiratory distress (defined as tachypnea, retractions, and apneas), need for HFNC and/or ventilatory support, hypercapnia, and hypoxemia.

We collected data on patient characteristics, disease severity (Pediatric Risk of Mortality Score [PRISM] III¹⁷  and Pediatric Logistic Organ Dysfunction 2 score [PELOD-2]),¹⁸  laboratory tests (white blood cells, neutrophil count, platelet count, pH, pCO₂, PaO₂, arterial PO₂, lactate, glucose, potassium, creatinine, and blood urea nitrogen) and chest radiograph results, treatments (ventilatory support initiated in the PICU [HFNC, continuous positive airway pressure (CPAP) and bilevel positive airway pressure (BiPAP)], systemic steroids, antibiotics, central venous access), and outcomes (PICU and hospital LOS and 28-day mortality). Noninvasive ventilation was defined as CPAP and/or BiPAP. PICU 1 and PICU 2 used HFNC only in the PICUs and not the pediatric wards during both seasons, whereas PICU 3 implemented the use of HFNC in the PICU only in season 2. Abnormal chest radiograph results were defined by the presence of infiltrates, consolidation, pleural effusion, hyperinflation, collapse, peri bronchial thickening, and/or atelectasis described by each hospital’s respective radiologists. Data on bronchodilators was only available for intubated patients.

We also collected bacterial and viral culture results, data on hospital-acquired infections (diagnosed by the local infection control teams), and the presence of concomitant bacterial infections. Testing for respiratory viruses was performed by multiplex polymerase chain reaction (PCR) based assays. The targets of the diagnostic panels used at each site can be found in Supplemental Table 5.

An expert committee, including 1 pediatric intensivist and 2 pediatric infectious diseases specialists, independently adjudicated the diagnosis of bacterial infections, including sepsis or septic shock, pneumonia, central nervous infections, intraabdominal infections, urinary tract infections, and otitis media, using the standardized definitions proposed by Goldstein et al (sepsis or septic shock) and the Centers for Disease Control and Prevention.^19,20  Disputes were resolved by consensus. Concomitant bacterial infections were classified as “probable” if patients presented the clinical criteria for bacterial infection without a positive bacterial culture, and as “definitive” if patients presented both the clinical criteria and ≥1 positive bacterial culture that was not adjudicated to be a bacterial contamination. The study was approved by the research ethics boards at all sites.

We summarized data using means (standard deviation [SD]) or medians (interquartile range [IQR]) for continuous data and proportions for categorical data.

We used a multivariable zero-truncated negative binomial regression model to evaluate the association between antibiotic use and hospital LOS in days. This model included the subgroup of patients adjudicated to have a viral bronchiolitis only, ie, patients for whom no probable or definitive concomitant bacterial infections were detected. Antibiotic use was defined categorically: no antibiotic prescription within 48 hours of hospital admission (reference category), antibiotics prescribed but stopped within 72 hours of hospital admission, and antibiotics prescribed and continued beyond 72 hours of hospital admission. We chose to define antibiotic prescriptions within 48 hours as this captures empirical antibiotic use and to define antibiotic discontinuation at 72 hours as bacterial and viral culture results are usually available in this time frame. Estimates from the negative binomial model were presented as LOS ratio and interpreted as the ratio of number of hospital days associated with each variable category or with a 1-unit increase in the explanatory continuous variable. After discussion with experts in the field, we adjusted the model for sex, age, presence of comorbidities, and PRISM III score. Comorbidities included cardiac congenital malformations, prematurity, chronic lung disease, tracheostomy, and use of ventilatory support at home (see Supplemental Table 4 for definitions). Analyses were conducted using R version 3.4.3.²¹ 

A total of 372 patients were included in the study, with 173 (46.5%) admitted in season 1 (Table 1). The median age was 2.1 months (IQR 1.2 – 6.6 months), and 51 (13.7%) patients were ≥12 months old (Supplemental Table 6). There were 107 (28.8%) patients with ≥1 comorbidity, most commonly prematurity (65 patients, 17.5%). The most common presenting symptoms at PICU admission were intercostal and/or subcostal retractions (332 patients; 89.2%), increased respiratory effort (294 patients; 79.0%), cough (290 patients; 78.0%), and tachypnea (271 patients; 72.8%). Leukocytosis and fever were present in 66 (17.7%) and 116 (31.1%) patients at PICU admission, respectively.

Clinical severity at PICU admission is presented in Supplemental Table 7. The median PRISM III score was 3.0 (IQR 0 – 3.0). The median PELOD-2 score on day 1 of PICU admission was 1.0 (IQR 0 – 3.0); lungs and kidneys were the organs most frequently affected.

The median values of pH and PCO₂ were 7.3 (IQR 7.3 – 7.3) and 7.0 kPa (IQR 6.1–8.0 kPa), respectively. Median respiratory rate was 43 breaths per minute (IQR 33 – 58 breaths per minute), and the median SatO₂:FiO₂ ratio was 296.7 (IQR 204.6–384.0). Overall, 304 patients (82%) had a chest radiograph performed. Of these, 269 (88%) patients had an abnormal result, with hyperinflation (109 patients; 36%) and consolidation (95 patients; 31%) being the most common findings (Supplemental Table 6).

There were 346 (93.0%) patients who received ventilatory support at PICU admission (Fig 1). The first mode of ventilatory support used at admission was HFNC for 105 (28.2%) patients, CPAP or BiPAP for 200 (53.7%) patients, and conventional invasive mechanical ventilation for 41 (11.0%) patients. Among patients initially treated with HFNC and CPAP/or BiPAP, 44 (41.9%) and 175 (87.5%) patients, respectively, did not need escalation of ventilatory support.

The total median duration of ventilatory support was 2.4 days (IQR 1.4–4.6 days). Specifically, the median duration of ventilatory support was 1.2 days (IQR 0.4–2.8 days) for HFNC, 2.0 days (IQR 0.8–3.3 days) for CPAP or BiPAP, and 3.7 days (IQR 2.4–6.8 days) for invasive mechanical ventilation (Supplemental Table 9).

Overall, 212 (57.0%) patients were prescribed antibiotics within the first 48 hours of hospital admission (Table 2), with a median duration of antibiotic treatment of 3.0 days (IQR 2.8–8.0 days). The most commonly prescribed antibiotics were ampicillin (94 patients; 44.3%) and ceftriaxone (39 patients; 18.4%) (Supplemental Table 10).

Of the 258 patients with a viral infection only, 117 (45.3%) patients were prescribed antibiotics within the first 48 hours of hospital admission (median antibiotic duration 3.0 days; IQR 2.0–6.0 days). Of these, antibiotics were stopped 72 hours from hospital admission in 79 (67.5%) patients, whereas 28 (23.9%) patients completed an antibiotic course (≥7 days).

Fifty-two (14.0%) patients received systemic steroids (Supplemental Table 9), with a median duration of 4.0 days (IQR 2.0–6.0 days). Eighteen (4.8%) patients were prescribed nebulized hypertonic saline. Sixty (16.1%) patients required a central venous line (CVL) (7.3% femoral or internal jugular line and 10.5% peripherally inserted central catheter). Of the 258 patients with viral infection only, 21 (8.1%) patients had CVLs inserted. Lastly, of the 77 intubated patients, 17 (22.1%) received bronchodilator treatment.

The median PICU LOS was 5.0 days (IQR 3.0–7.0 days) and hospital LOS was 7.0 days (IQR 5.0 – 11.0 days) (Table 2). There were no deaths during study follow-up period.

Overall, 355 (95.4%) patients had a nasopharyngeal specimen collected for viral testing, of which 298 (83.9%) were positive for at least 1 virus (Supplemental Table 11). Within the group of 249 (66.9%) patients who had mono-infection, RSV (195 patients; 52.4%) and rhinovirus (23 patients; 6.2%) were the most commonly identified pathogens. Forty-nine (13%) patients had codetections of 2 or more viruses (Supplemental Table 12).

After adjudication, 258 (69.4%) patients were found to have a viral infection only, 86 (23.1%) patients had a probable concomitant bacterial infection, and 28 (7.5%) patients had a definitive bacterial infection. The most common probable (52 patients; 14.0%) and definitive (20 patients; 5.4%) bacterial infections were pneumonia.

Fifteen (4.0%) patients developed a hospital-acquired infection, most commonly ventilator-associated pneumonia (4 patients; 26.7%) and bacteremia (3 patients; 33.3%). The median time to acquisition was 6.0 days (IQR 4.0–14.0 days) from hospital admission.

Among patients with a viral infection only (Table 3), the unadjusted association between antibiotics continued beyond 72 hours of hospital admission and hospital LOS was 1.16 (95%CI 1.00–1.37), and the adjusted association was 1.14 (95%CI 0.97–1.34).

Our study showed that CPAP or BiPAP and HFNC are commonly used in PICU patients with bronchiolitis. No escalation to invasive mechanical ventilation was needed in 87.5% and 41.9% using CPAP or BiPAP and HFNC, respectively. There were no observed deaths in our study. Furthermore, respiratory virus testing and chest radiographs were performed in most patients, and 7.5% of patients had a definitive concomitant bacterial infection. Approximately two-thirds of patients had a viral infection only, but almost half of them were prescribed antibiotics. Lastly, we observed that unnecessary antibiotic use beyond 72 hours from hospital admission in this group was not significantly associated with hospital length of stay.

The patient characteristics in our study are comparable to those described in the literature. Seventeen percent of study patients were premature, which is one of the most commonly cited comorbidities and risk factors for severe bronchiolitis.^3,12,22  In addition, the majority of patients in our study were <6 months old, which is consistent with studies that have shown the peak age for hospitalization is within the first 6 months of age.^3,23 

With no effective pharmacologic treatment available, the management of severe bronchiolitis remains largely supportive. Similarly to the survey of Canadian pediatric intensivists by Bradshaw et al, our results suggest that the use of CPAP or BiPAP and HFNC for patients admitted to a PICU with severe bronchiolitis has become the standard of care.^16,24  In 2 PICUs in Canada and France, Essouri et al showed that 77% and 16% of patients with bronchiolitis were initially treated with CPAP and HFNC, respectively, and 7% were intubated.²⁵  Mahant et al reported a significant increase in the use of noninvasive ventilation in [location redacted for review], with a rate of 4.4 in 2004–2005 to 17.2 per 1000 hospitalizations in 2017–2018.³ 

Nevertheless, current North American guidelines do not recommend the routine use of CPAP or BiPAP or HFNC for deteriorating cases of bronchiolitis, as their clinical benefit is still unclear.¹  A systematic review by Jat et al in 2015 reported that there is limited and low-quality evidence supporting CPAP use in acute or severe bronchiolitis to lower intubation rates.⁷  Furthermore, a randomized controlled trial by Franklin et al in 2018 found that patients receiving HFNC, compared with standard therapy, had significantly lower risk of escalation in respiratory support or transfer to a PICU.⁹  In contrast, in 2020 Durand et al compared HFNC to standard oxygen therapy and found no difference in the proportion of patients requiring subsequent noninvasive (16%) or invasive ventilation (14%; OR 0.66, 95%CI 0.35–1.26).¹⁰  Importantly, these trials were conducted in emergency departments and general pediatric units, and none included PICU patients. The few observational studies evaluating HFNC use in PICU patients have shown that it may be associated with lower intubation rates.^26–28  More studies are needed to assess the impact of CPAP or BiPAP and HFNC on subsequent escalation of ventilatory support and clinical outcomes to better inform evidence-based guidelines and treatment practices for severe bronchiolitis.

Despite North American bronchiolitis guidelines’ recommendations against it,^1,4  virologic testing was routinely performed in our study patients. Similarly, Kadmon et al and Carsin et al reported viral testing for 98.7% and 85% of patients <2 years old admitted to PICUs for bronchiolitis, respectively.^12,29  Tests used during our study period were conducted using reverse transcriptase polymerase chain reaction (RT-PCR) molecular technique. Previously, viral culture tests were more frequently performed. Studies have shown that in comparison with viral culture as the reference standard, RT-PCR has a sensitivity and specificity of 100% and 93%, respectively, to detect major respiratory viruses.³⁰  However, the detection of infectious virus via viral culture might be more informative in demonstrating virus infectiousness and virus shedding compared with RT-PCR.³¹  There is currently no evidence that respiratory virus testing alters management of bronchiolitis, except for cases of influenza.^4,32  Nevertheless, given the current severe acute respiratory syndrome coronavirus 2 pandemic, the rate of respiratory virus testing will likely remain high as all patients are screened on admission to hospital for infection control purposes.

In addition, the AAP bronchiolitis guidelines recommend reserving radiography for patients with severe bronchiolitis admitted to the PICU.¹  The majority of patients in our study had a CXR performed (81.7%). Similarly, Kadmon et al and Carsin et al reported the use of chest radiograph for 97.8% and 98.6%, respectively, of children <2 years old admitted to PICUs for bronchiolitis.^12,29  We speculate that the main utility for CXR in patients with bronchiolitis admitted to PICUs is to rule out other acute pathologies that may alter management, such as tension pneumothorax and pleural effusion requiring chest tube insertion, atelectasis requiring early use of positive pressure ventilation, cardiomegaly suggestive of possible myocarditis, congestive heart failure or congenital heart disease. However, in addition to increased exposure to radiation, studies have reported that the use of CXR, regardless of the results, is associated with the subsequent use of antibiotics.³³ 

Importantly, current guidelines do not recommend antibiotic use unless there is a concomitant bacterial infection, or a strong suspicion of one.^1,4  Almost 25% of patients in our study had a probable bacterial infection, with a definitive bacterial infection diagnosed in 7.5% of them after adjudication. This difference may be due to the inability of current diagnostic methods to reliably diagnose otitis media, as well as exclude a diagnosis of pneumonia.³⁴  Microbiological confirmation for the diagnosis of otitis media is rare. Moreover, while lower airway specimen samples are ideal for determining etiology in pneumonia, it was rarely feasible to obtain these samples from our patients as the majority were not intubated. Thus, it is possible that the true rate of bacterial pneumonia in our study was higher. Studies have indeed reported rates of concomitant bacterial pneumonia ranging from 10% to 38% in critically ill children with RSV bronchiolitis.^35,36 

Nevertheless, almost half of patients in our study with a viral bronchiolitis only were prescribed antibiotics. Kadmon et al similarly reported that, while bacterial pathogens were isolated in 13.4% of children <2 years old with bronchiolitis, 82.6% were treated with antibiotics. Duttweiler et al found that among 54 PICU patients with RSV bronchiolitis and no evidence of sepsis or pneumonia, 54% received antibiotics.³⁵  Such unnecessary antibiotic use likely contributes to the rise in bacterial resistance in hospital settings, and is also associated with adverse drug events, including allergic reactions, nephrotoxicity, and gastrointestinal disturbances. The development of better tests to diagnose concomitant bacterial pneumonia, as well as quality improvement interventions, may limit the unnecessary use of antibiotics in patients with viral bronchiolitis.^37,38 

Moreover, systemic steroids are still frequently used in children with a first episode of bronchiolitis, with 14% of our patients having received them. Following the AAP bronchiolitis guidelines publication, McCulloh et al reported a decrease in the use of steroids among children admitted to 2 US hospitals with acute bronchiolitis, from 26.5% to 17.5% (P < .001).³⁹  However, more recent studies show that the use of systemic steroids remains significant, reaching between 17% and 33% of hospitalized children <2 years old with bronchiolitis.^12,40–42  Hester et al studied children hospitalized with bronchiolitis and found 9.7% of patients received systemic steroids.⁴³  Evidence does not support the steroid use for severe bronchiolitis.^1,44  A meta-analysis by Hartling et al found no effect of systemic steroids compared with placebo on hospital LOS for children <2 years old hospitalized with a first episode of bronchiolitis.¹³  Similarly, van Woensel et al reported no effect of systemic steroids versus placebo on the duration of mechanical ventilation or hospital LOS for PICU patients with a lower respiratory tract infection caused by RSV.¹⁴ 

Our results on the viral epidemiology in bronchiolitis are consistent with the literature. Studies report the presence of RSV in 60% to 86% of hospitalized children with bronchiolitis.^23,29  Other common pathogens include rhinovirus (10% to 20%), and human metapneumovirus (1.5% to 8%). No pathogen is detected in up to 10% of patients, and codetection of ≥2 viruses occurs in 10% to 30% of patients.^1,23 

In contrast to recent studies, our study did not find a significant association between the unnecessary continuation of antibiotics among patients with viral infection only and hospital length of stay. However, given the estimate’s confidence intervals and our sample size, this may be due to inadequate power to detect a statistically significant result. Obolski et al studied 1003 children aged ≤2 years hospitalized for RSV bronchiolitis who lacked evidence of bacterial coinfection, and found that patients who were prescribed antibiotics had significantly longer mean hospital length of stay compared with patients who were not (4.60 days (SD 2.90) vs 3.15 days (2.47), P < .001).⁴⁵  Similarly, van Houten et al studied 284 children presenting to the emergency department or hospitalized with respiratory tract infections adjudicated to be viral only and found that patients who were prescribed antibiotics had a longer median hospitalization duration compared with patients who were not (5 days [IQR 3–9] vs 3 days [IQR 2–4], P < .001).⁴⁶ 

Our study has limitations. The results may not be generalizable to all critically ill patients with bronchiolitis in Canada as we did not explore potential variability in clinical management practices among PICUs. However, the survey by Bradshaw et al showed that management of bronchiolitis is similar across Canadian PICUs.¹⁶  Second, PICU 1 did not test for bocavirus and PICU 3 occasionally performed triplex instead of multiplex PCR, thus the frequency of bocavirus and other viruses may be underrepresented. Lastly, we could not describe the patient’s race or ethnicity or gestational age, nor the use of bronchodilators, as these variables were not available in the database used for this study.

Nevertheless, our study has important strengths. We provided a comprehensive description of the epidemiology and current management of critically ill children with bronchiolitis. Our results will help to inform the development of evidence-based care for these patients. Moreover, we conducted a rigorous adjudication process to determine the presence of probable or definitive concomitant bacterial infections using robust infection definitions.

We showed that CPAP or BiPAP and HFNC have become the standard of care for critically ill children with bronchiolitis in 3 Canadian PICUs, despite limited evidence for their use in these patients. Other nonrecommended therapies were also frequently used, including systemic steroids and antibiotics. Importantly, almost half of patients with a viral infection only were unnecessarily prescribed antibiotics. The development of evidence-based bronchiolitis guidelines that pertain specifically to critically ill children is needed to improve the care delivered to this patient population.