BACKGROUND AND OBJECTIVES

High-flow nasal cannula (HFNC) for bronchiolitis increased over the past decade without clear benefit. This quality improvement collaborative aimed to reduce HFNC initiation and treatment duration by 30% from baseline.

METHODS

Participating hospitals either reduced HFNC initiation (Pause) or treatment duration (Holiday) in patients aged <24 months admitted for bronchiolitis. Participants received either Pause or Holiday toolkits, including: intervention protocol, training/educational materials, electronic medical record queries for data acquisition, small-group coaching, webinars, and real-time access to run charts. Pause arm primary outcome was proportion of patients initiated on HFNC. Holiday arm primary outcome was geometric mean HFNC treatment duration. Length of stay (LOS) was balancing measure for both. Each arm served as contemporaneous controls for the other. Outcomes analyzed using interrupted time series (ITS) and linear mixed-effects regression.

RESULTS

Seventy-one hospitals participated, 30 in the Pause (5746 patients) and 41 in the Holiday (7903 patients). Pause arm unadjusted HFNC initiation decreased 32% without LOS change. ITS showed immediate 16% decrease in initiation (95% confidence interval [CI] −27% to −5%). Compared with contemporaneous controls, Pause hospitals reduced HFNC initiation by 23% (95% CI −35% to −10%). Holiday arm unadjusted HFNC duration decreased 28% without LOS change. ITS showed immediate 11.8 hour decrease in duration (95% CI −18.3 hours to −5.2 hours). Compared with contemporaneous controls, Holiday hospitals reduced duration by 11 hours (95% CI −20.7 hours to −1.3 hours).

CONCLUSIONS

This quality improvement collaborative reduced HFNC initiation and duration without LOS increase. Contemporaneous control analysis supports intervention effects rather than secular trends toward less use.

Viral bronchiolitis is the leading cause of hospitalization in pediatric patients <12 months of age.1  Use of high-flow nasal cannula (HFNC) in bronchiolitis increased dramatically over the last decade.2 4  Recent database studies report a 6- to sevenfold increase in noninvasive ventilation rates, ICU admissions, and costs associated with bronchiolitis care, which researchers hypothesize has been driven by HFNC.5 ,6 

HFNC was rapidly adopted before establishment of evidence to guide use in bronchiolitis, and indications for initiation remain elusive despite randomized controlled trials (RCTs). Four RCTs demonstrated no clinical benefit in early routine use of HFNC compared with low-flow nasal cannula (LFNC) in the treatment of patients with mild-to-moderate bronchiolitis.7 10  There were no differences in intubation rates or ICU utilization,7 9  and 1 study reported longer lengths of stay associated with HFNC use.10  Taken together, these data suggest that HFNC may be appropriately positioned as a rescue therapy when treatment with LFNC has failed, and lack of high-quality evidence for routine HFNC use makes it an ideal target for quality improvement (QI) efforts to reduce low-value care. Published strategies to decrease HFNC overuse have focused on guiding decision-making around HFNC utilization, including standardized rapid HFNC weaning protocols, defining initiation criteria, and reinforcing trials of LFNC before escalation to HFNC.11 15 

In 2020, the Value in Inpatient Pediatrics Network organized a national multicenter QI initiative designed to deimplement HFNC in patients admitted with bronchiolitis by providing decision support tools to influence behavior around HFNC initiation and weaning. Through this QI collaborative, entitled High-Flow Interventions to Facilitate Less Overuse (HIFLO), we sought to reduce HFNC initiation or treatment duration by 30% from baseline rates in separate intervention groups over a 16-week period.

This study was sponsored by the Value in Inpatient Pediatrics Network, a QI network within the American Academy of Pediatrics (AAP) that includes >250 hospitals across the United States and Canada, and aims to improve the value of health care provided to pediatric patients in the acute care setting. Member hospitals range widely in size, type (university-affiliated or community), and location. Retrospectively collected data from November 2019 to March 2020 were used as a project baseline. The November 2020–March 2021 season was excluded given disrupted viral transmission patterns associated with the coronavirus disease 2019 pandemic.16  Toolkit materials were released in July 2021, and the intervention data collection period was November 2021 to March 2022.

HIFLO included patients aged 30 days through 23 months admitted for mild-to-moderate bronchiolitis. Admission was defined as observation, inpatient, intermediate, or intensive care. Patients with severe bronchiolitis, defined as needing continuous positive airway pressure, bilevel positive airway pressure, or intubation at any time, were excluded. Exclusion of patients who required these therapies (typically provided in the ICU) was extrapolated from RCT data that found no change in ICU transfer rate in patients who received early HFNC therapy.8 ,9 ,17  Additional exclusion criteria included infants born <32 weeks’ gestational age, hemodynamically significant cardiac disease requiring cardiac medications, chronic lung disease on home oxygen or diuretics, significant neuromuscular disease requiring assistance with breathing or feeding, home oxygen or airway clearance therapy, and infants admitted directly from an outside facility.

In December 2020, an 11-person expert workgroup was formed, with representatives from hospital medicine, emergency medicine, and critical care. Each member collaborated with local registered nurses (RNs) and respiratory therapists (RTs) to review the available evidence and complete root cause analysis of HFNC overuse. These interprofessional teams identified barriers to HFNC deimplementation, including action bias, fear of decompensation, and variable assessments of illness severity, which were integrated into a key driver diagram to guide toolkit development (Fig 1). Through monthly virtual meetings, the workgroup used the model for improvement18  to define aims, measures, and intervention targets. The expert workgroup defined overuse as early initiation of HFNC or prolonged HFNC weaning. As a result, 2 interventions were developed, 1 targeting HFNC initiation practices (Pause) and 1 targeting HFNC weaning practices (Holiday). The study was designed such that each intervention arm targeted independent outcomes, and thus each arm also could serve as a contemporaneous control for the outcome targeted by the alternate arm.

FIGURE 1

Key driver diagrams for (A) HFNC initiation and (B) HFNC weaning.

FIGURE 1

Key driver diagrams for (A) HFNC initiation and (B) HFNC weaning.

Close modal

Recruitment began in May 2021 via e-mail announcements over AAP listservs. Sites completed an application survey that reported current HFNC practices and preferred intervention group. Hospitals were assigned to Pause or Holiday on the basis of preference, and those willing and eligible to participate in either intervention were assigned to the Pause arm. Sites were asked to form interdisciplinary leadership teams with a pediatric hospital medicine physician, emergency medicine or critical care physician as appropriate, nurse leader, and RT leader. An intervention toolkit (Pause or Holiday) was provided to hospitals in July 2021 to provide ample time to devise local implementation strategies before data collection began in November 2021. The application included a participation fee of $1000, which covered a portion of the cost of the AAP staff and data analysis.

Toolkit

Hospitals were only granted access to their assigned HIFLO toolkit, either Pause or Holiday. Each toolkit included a modular protocol designed to be integrated into their existing bronchiolitis treatment pathways to fit local needs. The Pause protocol, implemented in the emergency department (ED) and inpatient settings, outlined a series of steps recommended before the initiation of HFNC, including management of hydration, hunger, fever/discomfort, and overstimulation. The final step in the Pause was to consider a trial of LFNC before escalation to HFNC (Supplemental Fig 6A). The Holiday protocol, implemented in the inpatient setting and ICU, defined heart rate, fraction of inspired oxygen, and saturation parameters at which a patient would be eligible for a rapid wean of HFNC, with options of weaning flow by 50%, trial LFNC, or trial room air (Supplemental Fig 6B). Vital sign parameters were provided to guide teams in determining a successful Pause or Holiday trial, and hospitals could adjust parameters as locally indicated. Hospitals were encouraged to assess illness severity according to local practices. HIFLO did not recommend a clinical respiratory score because scores have poor interrater reliability and predictive validity for outcomes.19 ,20 

In addition to protocols, the HIFLO toolkit included:

  1. access to the collaborative Web site;

  2. online, interprofessional educational videos titled the “Bronchiolitis Booster Series” (www.vimeo.com/bronchiolitis), which provided an evidence review and content to foster a shared mental model of respiratory assessment;

  3. electronic medical record queries to support data acquisition;

  4. an interactive, computer-based training module for interprofessional team members simulating application of the Pause or Holiday in clinical practice;

  5. an idea list of suggested targets for local improvement efforts;

  6. access to run charts showing site- and project-level aggregate performance; and

  7. a webinar series to provide education on principles of QI and report collaborative-wide data.

Coaching

Participants were provided coaching by peer QI experts. Pods of 2 to 4 hospitals were created on the basis of region and intervention arm, and each met with their coach 5 times over the implementation season. Project leaders provided objectives for each session, including protocol implementation, garnering engagement, education efforts, data collection, and planning improvement cycles. Coaches submitted reports after each session, sharing the content discussed, challenges encountered, and successes of local teams (Supplemental Table 3). Project leaders reviewed reports in real time, using this information to inform content for subsequent meetings and to identify local teams to share their stories in the QI webinar series.

Sites collected 16 weeks of retrospective baseline data from November 2019 to March 2020 and 16 weeks of prospective intervention data from November 2021 to March 2022. Sites entered data weekly, either for all eligible patients or a random sample of 10 patients per week for hospitals with higher volumes. A randomization sequence was provided for higher-volume sites to reduce the burden of chart review and minimize the influence of large hospitals on the aggregate data set. Sites were provided with chart review manuals with step-by-step instructions for data entry. Data were deidentified and entered into the AAP’s QI Data Aggregator, a Web-based data collection instrument that allowed sites to view real-time run charts to inform local improvement efforts. The data collection instrument was the same for Pause and Holiday arms (Supplemental Fig 7).

The primary outcome measure for the Pause arm was the proportion of patients admitted for bronchiolitis who received HFNC therapy during hospitalization. The process measure was documentation of Pause completion before HFNC initiation. Hospitals devised their own method for identifying whether a Pause was performed (eg, documented in flowsheets). To capture any potential negative impact that delaying initiation of HFNC may have, geometric mean ED length of stay (LOS) and geometric mean hospital LOS were tracked as balancing measures.

The primary outcome measure for the Holiday arm was the geometric mean HFNC treatment duration. The process measure was documentation of a Holiday at any time in the HFNC weaning process. Hospitals identified their own method for determining whether a Holiday occurred. To capture any potential harmful impact of rapid discontinuation, geometric mean hospital LOS and number of clinical deterioration events after a Holiday attempt were tracked as balancing measures.

Hospital characteristics were compared between intervention groups using χ2 tests. Baseline and intervention unadjusted mean HFNC initiation rates were calculated using the individual means of each hospital site. Patient characteristics and unadjusted relative differences for each primary outcome were compared between study periods within each intervention group using χ2 tests for categorical and t tests for continuous characteristics. To account for secular trends, project outcomes were also compared between study periods within each intervention group using interrupted time series (ITS) models. ITS analysis was selected to determine if there was an immediate change in outcome at the start of the intervention and to assess differences in rate of change over time (slope) throughout the intervention season. Process measures were tracked via run charts and analyzed for special cause variation using published criteria.18  In the contemporaneous control analysis, study arm-specific outcomes (HFNC initiation for Pause and treatment duration for Holiday) were compared between intervention groups using generalized linear models controlling for week of study. All patients with an inpatient LOS <1 hour or >336 hours (<1% of observations, determined to be outliers) were excluded from all analyses. All analyses were conducted using R version 3.4.3 (R Core Team, Vienna, Austria).

This QI project was approved by the AAP institutional review board (21 CHA 02). Participating teams obtained local institutional review board approval if required at the institutional level.

A total of 103 hospitals applied to participate, 86 paid the enrollment fee, and 71 completed data entry. There were 31 hospitals in the Pause arm and 40 in the Holiday arm. There were no differences in hospital characteristics between intervention groups (Table 1).

TABLE 1

Hospital Characteristics

Hospital Characteristics
Total n (%)Pause n (%)Holiday n (%)P
Number 71 (100) 31 (44) 40 (56) n/a 
Freestanding children’s hospital .47 
 Yes 30 (42) 15 (48) 15 (38)  
 No 41 (58) 16 (52) 25 (62)  
University affiliated .59 
 Yes 52 (73) 24 (77) 28 (70)  
 No 19 (27) 7 (23) 12 (30)  
On-site PICU .38 
 Yes 66 (93) 30 (97) 36 (90)  
 No 5 (7) 1 (3) 4 (10)  
Non-ICU pediatric bedsa .19 
 ≤35 18 (25) 5 (16) 13 (32)  
 36–70 20 (28) 12 (39) 8 (20)  
 71–140 15 (21) 5 (16) 10 (25)  
 >140 18 (25) 9 (29) 9 (22)  
Hospital Characteristics
Total n (%)Pause n (%)Holiday n (%)P
Number 71 (100) 31 (44) 40 (56) n/a 
Freestanding children’s hospital .47 
 Yes 30 (42) 15 (48) 15 (38)  
 No 41 (58) 16 (52) 25 (62)  
University affiliated .59 
 Yes 52 (73) 24 (77) 28 (70)  
 No 19 (27) 7 (23) 12 (30)  
On-site PICU .38 
 Yes 66 (93) 30 (97) 36 (90)  
 No 5 (7) 1 (3) 4 (10)  
Non-ICU pediatric bedsa .19 
 ≤35 18 (25) 5 (16) 13 (32)  
 36–70 20 (28) 12 (39) 8 (20)  
 71–140 15 (21) 5 (16) 10 (25)  
 >140 18 (25) 9 (29) 9 (22)  

n/a, not applicable.

a

Defined as medical and surgical pediatrics beds and excluding ICU and well-infant nursery.

A total of 13 649 patients were included in the overall study, 5746 (42%) in the Pause arm and 7903 (58%) in the Holiday arm. There were fewer patients in the intervention period than in the baseline, although there were no clinically meaningful differences in patient demographics within each study period (Table 2).

TABLE 2

Patient Characteristics and Primary Outcomes

Patient Characteristics
TotalPause (n = 5746)Holiday (n = 7903)
BaselineInterventionPBaselineInterventionP
Patients, n (%) 13 649 (100) 3854 (67) 1892 (33) n/a 5522 (70) 2381 (30) n/a 
Age, n (%) .49  <.01 
 1–2 mo 3006 (22) 807 (21) 381 (20)  1326 (24) 492 (21)  
 3–23 mo 10 643 (78) 3047 (79) 1511 (80)  4196 (76) 1889 (79)  
Weight kg mean (SD) 8 (±3) 8 (±3) 8 (±4) <.01 8 (±3) 8 (±3) <.01 
Sex, n (%) .22  .24 
 Male 8079 (59) 2246 (58) 1136 (60)  3258 (59) 1439 (60)  
 Female 5571 (41) 1608 (42) 757 (40)  2264 (41) 942 (40)  
Patient Characteristics
TotalPause (n = 5746)Holiday (n = 7903)
BaselineInterventionPBaselineInterventionP
Patients, n (%) 13 649 (100) 3854 (67) 1892 (33) n/a 5522 (70) 2381 (30) n/a 
Age, n (%) .49  <.01 
 1–2 mo 3006 (22) 807 (21) 381 (20)  1326 (24) 492 (21)  
 3–23 mo 10 643 (78) 3047 (79) 1511 (80)  4196 (76) 1889 (79)  
Weight kg mean (SD) 8 (±3) 8 (±3) 8 (±4) <.01 8 (±3) 8 (±3) <.01 
Sex, n (%) .22  .24 
 Male 8079 (59) 2246 (58) 1136 (60)  3258 (59) 1439 (60)  
 Female 5571 (41) 1608 (42) 757 (40)  2264 (41) 942 (40)  
Primary outcomes
 HFNC initiation, n (%) 1692 (44%) 559 (30%) <.01 3170 (57%) 1391 (58%) .99 
 HFNC treatment duration geometric mean h (SD) 39.9 (37.2–42.3) 35.4 (33.0–37.8) .01 51.3 (50.0–52.6) 36.9 (35.6–38.2) <.01 
Primary outcomes
 HFNC initiation, n (%) 1692 (44%) 559 (30%) <.01 3170 (57%) 1391 (58%) .99 
 HFNC treatment duration geometric mean h (SD) 39.9 (37.2–42.3) 35.4 (33.0–37.8) .01 51.3 (50.0–52.6) 36.9 (35.6–38.2) <.01 

n/a, not applicable.

Patient characteristics for all patients included in the study (total), and for the Pause and Holiday intervention groups. Primary outcomes listed as mean HFNC initiation percentage and geometric mean (hours) of HFNC treatment duration in baseline and intervention periods for Pause and Holiday groups. P values represent within-group differences.

There was a relative reduction in the mean HFNC initiation rate of 32% compared with baseline (Table 2). The percentage change in initiation rate by hospital is shown in Supplemental Fig 8A. When accounting for changes over time in the ITS model, there was an immediate 16% absolute reduction in HFNC initiation at the start of the intervention season (95% confidence interval [CI] −27% to −5%), with no change in slope (Fig 2). In the contemporaneous control analysis, we compared the change in HFNC initiation from baseline rates (inclusive of all hospitals) to initiation rates at hospitals that implemented the Pause and initiation rates at hospitals that did not (Holiday hospitals). Compared with contemporaneous control hospitals in the Holiday arm, Pause hospitals demonstrated a 23% reduction in HFNC initiation (95% CI −35% to −10%) (Fig 3).

FIGURE 2

Pause arm ITS analysis of the proportion of patients treated with HFNC in the baseline versus intervention periods. There was an immediate 16% reduction in HFNC initiation at the start of the intervention (95% CI −27% to −5%), with no change in slope (0%; 95% CI −0.6 to 0.5%).

FIGURE 2

Pause arm ITS analysis of the proportion of patients treated with HFNC in the baseline versus intervention periods. There was an immediate 16% reduction in HFNC initiation at the start of the intervention (95% CI −27% to −5%), with no change in slope (0%; 95% CI −0.6 to 0.5%).

Close modal
FIGURE 3

Contemporaneous control analysis comparing the change in HFNC initiation from baseline rates (inclusive of all hospitals) to hospitals that implemented Pause (gray triangles) and those that did not (black circles). Compared with Holiday hospitals, hospitals that implemented the Pause demonstrated a 23% reduction in HFNC initiation at the start of the intervention (95% CI −35% to −10%), with no significant change in slope (−0.2%; 95% CI −0.7% to 0.3%).

FIGURE 3

Contemporaneous control analysis comparing the change in HFNC initiation from baseline rates (inclusive of all hospitals) to hospitals that implemented Pause (gray triangles) and those that did not (black circles). Compared with Holiday hospitals, hospitals that implemented the Pause demonstrated a 23% reduction in HFNC initiation at the start of the intervention (95% CI −35% to −10%), with no significant change in slope (−0.2%; 95% CI −0.7% to 0.3%).

Close modal

A Pause was documented before HFNC initiation in 36% of patients (Supplemental Fig 9). There was no change in hospital LOS (Supplemental Fig 10) or ED LOS (Supplemental Fig 11). When comparing unadjusted relative differences in geometric mean HFNC treatment duration between baseline and intervention seasons, there was a reduction in treatment duration of 4.5 hours (P = .01, Table 2). ITS analysis of HFNC duration in hospitals that implemented the Pause revealed a preexisting downward trend in treatment duration that continued during the intervention period. When accounting for these trends over time, there was no significant slope or intercept change (Supplemental Fig 12).

There was a relative reduction in the geometric mean HFNC treatment duration of 28% compared with baseline (Table 2). The percentage change in duration by hospital is shown in Supplemental Fig 8B. When accounting for changes over time in the ITS model, there was an immediate 11.8-hour absolute reduction in treatment duration at the start of the intervention season (95% CI −18.3 hours to −5.2 hours), with no change in slope (Fig 4). In the contemporaneous control analysis, we compared the change in HFNC treatment duration from baseline rate (inclusive of all hospitals) to treatment duration at hospitals that implemented the Holiday and duration at hospitals that did not (Pause hospitals). Compared with contemporaneous control hospitals in the Pause arm, Holiday hospitals demonstrated an 11-hour reduction HFNC duration (95% CI −20.7 hours to −1.3 hours) (Fig 5).

FIGURE 4

Holiday arm ITS analysis of the geometric mean HFNC treatment duration in the baseline versus intervention periods. There was an immediate 11.8-hour reduction in HFNC duration (95% CI −18.3 hours to −5.2 hours), with no change in slope (0.3 hours; 95% CI −0.1 hours to 0.6 hours).

FIGURE 4

Holiday arm ITS analysis of the geometric mean HFNC treatment duration in the baseline versus intervention periods. There was an immediate 11.8-hour reduction in HFNC duration (95% CI −18.3 hours to −5.2 hours), with no change in slope (0.3 hours; 95% CI −0.1 hours to 0.6 hours).

Close modal
FIGURE 5

Contemporaneous control analysis comparing the change in HFNC treatment duration from baseline rates (inclusive of all hospitals) to hospitals that implemented the Holiday (gray squares) and those that did not (black circles). Compared with Pause hospitals, hospitals that implemented the Holiday demonstrated an 11-hour reduction treatment duration at the start of the intervention (95% CI −20.7 hours to −1.3 hours), with no change in slope (0.1 hours; 95% CI −0.3 hours to 0.5 hours).

FIGURE 5

Contemporaneous control analysis comparing the change in HFNC treatment duration from baseline rates (inclusive of all hospitals) to hospitals that implemented the Holiday (gray squares) and those that did not (black circles). Compared with Pause hospitals, hospitals that implemented the Holiday demonstrated an 11-hour reduction treatment duration at the start of the intervention (95% CI −20.7 hours to −1.3 hours), with no change in slope (0.1 hours; 95% CI −0.3 hours to 0.5 hours).

Close modal

A Holiday was documented in 62% of patients during the HFNC weaning process (Supplemental Fig 13). There was no change in geometric mean hospital LOS when examining all patients in the Holiday arm (Supplemental Fig 14). For the subset of patients treated with HFNC, there was also no significant change in hospital LOS (Supplemental Fig 15). Out of 2381 Holiday attempts, there were 4 reports of clinical deterioration (0.12%). Using ITS analysis, HFNC initiation was examined in hospitals that implemented the Holiday; there was no change in HFNC initiation at the start of intervention, nor change in slope (Supplemental Fig 16).

This multisite QI collaborative demonstrated reductions in HFNC initiation and duration of use for patients with mild-to-moderate bronchiolitis without an increase in ED or hospital LOS. We achieved our specific aim of a 30% reduction from baseline HFNC initiation rates and had a 28% reduction from baseline treatment duration. Initial improvements in primary outcomes were sustained throughout the study period.

One strength of our study design was the use of each group as a contemporaneous control for the other arm, which was possible because the interventions were targeted toward unrelated aspects of HFNC use (initiation versus weaning). Hospitals that implemented the Pause demonstrated reductions in HFNC initiation, whereas the Holiday hospitals (where no interventions targeting HFNC initiation occurred) did not. Similarly, hospitals that implemented the Holiday demonstrated significant reductions in HFNC treatment duration, whereas the Pause hospitals did not. These comparisons support our conclusion that our interventions, rather than secular trends, led to the improvement in outcomes.

Interestingly, process measure documentation for the Pause was low; although this could suggest low fidelity to the Pause intervention and that changes in primary outcomes were not related to the intervention itself, this is not supported by our contemporaneous control analysis. On the basis of feedback from coaching calls (Supplemental Table 3), hospital teams were unable to reliably capture documentation and/or completion of the Pause because of variability in role assignment and documentation practices across different hospital settings. We suspect our process measure documentation was low because of this ambiguity, rather than true nonadherence to the intervention.

Our success in HFNC deimplementation is multifactorial. The expert workgroup designed each protocol to be modular so sites could easily integrate it into their local bronchiolitis treatment pathways. Educational materials were developed with the intention of engaging interprofessional care teams and building a shared mental model of respiratory assessment and evidence-based HFNC use. To further increase engagement, we offered physician maintenance of certification and continuing medical education credits, as well as RT and RN continuing education credits.

Unique to deimplementation are psychological barriers such as fear, loss, or action bias, which are concepts that were uncovered in our root cause analysis, that require specific attention when developing deimplementation projects.21  When developing our protocols, we incorporated mitigation strategies addressing these barriers, including making allowance for nonconformism and permission to change.21  We created the Pause to act as a speedbump in the path toward HFNC initiation and provided alternative actions to early HFNC initiation by substituting a series of steps for the bedside care team to complete.22  The Holiday protocol recommended discontinuation of HFNC from any flow rate once parameters were met, but also allowed for nonconformism with an alternative wean of flow by 50% or a trial of LFNC if care teams were uncomfortable with discontinuation. Attention to these barriers with targeted mitigation strategies integrated into our protocols likely contributed to our improvements.

Previous single-center QI studies that implemented rapid HFNC discontinuation trials demonstrated reduced HFNC treatment duration with associated reductions in LOS.11 13  A trial of LFNC before HFNC initiation reduced initiation rates in 1 single-center study,14  and a recent publication that used HFNC initiation criteria and preinitiation huddles reduced HFNC use and LOS.15  Our multisite study targeting HFNC overuse adds to the literature with inclusion of hospitals diverse in size, type, location, and structure. Our study demonstrates the feasibility of similar deimplementation efforts across a diverse cohort of hospitals.

It is well established in published protocols from Europe and Australia that rapid discontinuation of HFNC is the standard recommended approach and that HFNC should be reserved for patients with hypoxemia who have failed LFNC.8 10 ,23  This article contributes to a growing body of evidence from the United States that rapid HFNC weaning protocols and promotion of HFNC as a rescue therapy are safe and effective means of curtailing HFNC overuse, thereby promoting high-value care.

In contrast to single-center studies,11 15  we did not observe a statistically significant change in LOS. However, we used LOS as a balancing measure to ensure our interventions did not have a negative impact. It is likely that the smaller number of patients in the intervention period and data collection during a single viral season impacted our ability to identify a significant change in LOS. Even without reductions in LOS, reduced HFNC utilization has positive impacts on local resource use, hospital costs, and patient-centered outcomes such as comfort and oral feeding.

One limitation of the study is that all hospitals chose to participate in the collaborative and selected which intervention arm to participate in, introducing selection bias and limiting generalizability. Although the expert workgroup worked closely with local interprofessional teams, it did not include a patient representative, RN, or RT. Some of the workgroup members were also QI coaches, which could have introduced bias in those hospitals that received their more expert guidance. One RCT published between the baseline and intervention seasons could have influenced HFNC management, but its results confirm the findings of the preceding RCTs that were included in project literature.7  We did not obtain detailed information on the 4 patients with documented deterioration after a Holiday, nor did we collect additional safety data.

There was a gap of 1 season between the baseline and intervention periods because of the coronavirus disease 2019 pandemic, which may have influenced disease severity or HFNC usage. However, the use of contemporary comparisons supports that our outcomes are because of our interventions rather than secular trends. Although the 2 intervention groups served as contemporary comparisons, they were not traditional balanced control groups. Finally, because we relied on a voluntary workforce, data entry may not have been complete.

This national, multicenter QI collaborative successfully reduced HFNC initiation and treatment duration in patients hospitalized with mild-to-moderate bronchiolitis without associated increase in ED or hospital LOS. Until clear HFNC initiation criteria are defined, the deimplementation strategies presented in this study represent feasible actions hospitals may take to promote more judicious utilization of HFNC in the treatment of bronchiolitis.

Drs Byrd, Noelck, and Ralston designed and conceptualized the study, contributed to data analysis, drafted the initial draft of the manuscript, and reviewed and revised the manuscript; Dr Kerns contributed to the design of the study, acquired data and spearheaded data analysis, created all figures, and critically reviewed and revised the manuscript; Drs Bryan, Hamline, Garber, Ostrow, Riss, Shadman, Shein, and Willer contributed to the design of the study, analyzed data and results, and critically reviewed and revised the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

FUNDING: No external funding.

CONFLICT OF INTEREST DISCLOSURES: Drs Byrd and Noelck received $15 000 each for their role as project leaders. The authors have indicated they have no conflicts of interest relevant to this article to disclose.

AAP

American Academy of Pediatrics

CI

confidence interval

ED

emergency department

HFNC

high-flow nasal cannula

HIFLO

High-Flow Interventions To Facilitate Less Overuse

ITS

interrupted time series

LFNC

low-flow nasal cannula

LOS

length of stay

QI

quality improvement

RCT

randomized controlled trial

RN

registered nurse

RT

respiratory therapist

1
Hasegawa
K
,
Tsugawa
Y
,
Brown
DF
,
Mansbach
JM
,
Camargo
CA
 Jr
.
Trends in bronchiolitis hospitalizations in the United States, 2000–2009
.
Pediatrics
.
2013
;
132
(
1
):
28
36
2
Linssen
RS
,
Bem
RA
,
Kapitein
B
, et al
.
PICE Study Group
.
Burden of respiratory syncytial virus bronchiolitis on the Dutch pediatric intensive care units
.
Eur J Pediatr
.
2021
;
180
(
10
):
3141
3149
3
Linssen
RS
,
Teirlinck
AC
,
van Boven
M
, et al
.
Increasing burden of viral bronchiolitis in the pediatric intensive care unit; an observational study
.
J Crit Care
.
2022
;
68
:
165
168
4
Schlapbach
LJ
,
Straney
L
,
Gelbart
B
, et al
.
Burden of disease and change in practice in critically ill infants with bronchiolitis
.
Eur Respir J
.
2017
;
49
(
6
):
1601648
5
Pelletier
JH
,
Au
AK
,
Fuhrman
D
,
Clark
RSB
,
Horvat
C
.
Trends in Bronchiolitis ICU admissions and ventilation practices: 2010–2019
.
Pediatrics
.
2021
;
147
(
6
):
e2020039115
6
Willer
RJ
,
Coon
ER
,
Harrison
WN
,
Ralston
SL
.
Trends in hospital costs and levels of services provided for children with bronchiolitis treated in children’s hospitals
.
JAMA Netw Open
.
2021
;
4
(
10
):
e2129920
7
Durand
P
,
Guiddir
T
,
Kyheng
C
, et al
.
A randomized trial of high-flow nasal cannula in infants with moderate bronchiolitis
.
Eur Respir J
.
2020
;
56
(
1
):
1901926
8
Franklin
D
,
Babl
FE
,
Schlapbach
LJ
, et al
.
A randomized trial of high-flow oxygen therapy in infants with bronchiolitis
.
N Engl J Med
.
2018
;
378
(
12
):
1121
1131
9
Kepreotes
E
,
Whitehead
B
,
Attia
J
, et al
.
High-flow warm humidified oxygen versus standard low-flow nasal cannula oxygen for moderate bronchiolitis (HFWHO RCT): an open, phase 4, randomized controlled trial
.
Lancet
.
2017
;
389
(
10072
):
930
939
10
Franklin
D
,
Babl
FE
,
George
S
, et al
.
Effect of early high-flow nasal oxygen versus standard oxygen therapy on length of hospital stay in hospitalized children with acute hypoxemic respiratory failure: the paris-2 randomized clinical trial
.
JAMA
.
2023
;
329
(
3
):
224
234
11
Betters
KA
,
Hebbar
KB
,
McCracken
C
,
Heitz
D
,
Sparacino
S
,
Petrillo
T
.
A novel weaning protocol for high-flow nasal cannula in the PICU
.
Pediatr Crit Care Med
.
2017
;
18
(
7
):
e274
e280
12
Charvat
C
,
Jain
S
,
Orenstein
EW
,
Miller
L
,
Edmond
M
,
Sanders
R
.
Quality initiative to reduce high-flow nasal cannula duration and length of stay in bronchiolitis
.
Hosp Pediatr
.
2021
;
11
(
4
):
309
318
13
Noelck
M
,
Foster
A
,
Kelly
S
, et al
.
SCRATCH Trial: an initiative to reduce excess use of high-flow nasal cannula
.
Hosp Pediatr
.
2021
;
11
(
4
):
319
326
14
Siraj
S
,
Compton
B
,
Russell
B
,
Ralston
S
.
Reducing high-flow nasal cannula overutilization in viral bronchiolitis
.
Pediatr Qual Saf
.
2021
;
6
(
4
):
e420
15
Treasure
JD
,
Lipshaw
MJ
,
Dean
P
, et al
.
Quality improvement to reduce high-flow nasal cannula overuse in children with bronchiolitis
.
Pediatrics
.
2023
;
152
(
3
):
e2022058758
16
Hatoun
J
,
Correa
ET
,
Vernacchio
L
.
COVID-19 pandemic-related changes in pediatric seasonal respiratory infections
.
Pediatrics
.
2022
;
150
(
4
):
e2022058618
17
Lin
J
,
Zhang
Y
,
Xiong
L
,
Liu
S
,
Gong
C
,
Dai
J
.
High-flow nasal cannula therapy for children with bronchiolitis: a systematic review and meta-analysis
.
Arch Dis Child
.
2019
;
104
(
6
):
564
576
18
Provost
LP
,
Murray
SK
.
The Health Care Data Guide: Learning from Data for Improvement
.
Hoboken, NJ
:
Jossey-Bass
;
2011
19
Davies
CJ
,
Waters
D
,
Marshall
A
.
A systematic review of the psychometric properties of bronchiolitis assessment tools
.
J Adv Nurs
.
2017
;
73
(
2
):
286
301
20
Bekhof
J
,
Reimink
R
,
Bartels
IM
,
Eggink
H
,
Brand
PL
.
Large observer variation of clinical assessment of dyspnoeic wheezing children
.
Arch Dis Child
.
2015
;
100
(
7
):
649
653
21
McDaniel
CE
,
House
SA
,
Ralston
SL
.
Behavioral and psychological aspects of the physician experience with deimplementation
.
Pediatr Qual Saf
.
2022
;
7
(
1
):
e524
22
Parchman
ML
,
Palazzo
L
,
Austin
BT
, et al
.
Taking action to address medical overuse: common challenges and facilitators
.
Am J Med
.
2020
;
133
(
5
):
567
572
23
O’Brien
S
,
Craig
S
,
Babl
FE
,
Borland
ML
,
Oakley
E
,
Dalziel
SR
;
Paediatric Research in Emergency Departments International Collaborative (PREDICT) Network, Australasia
.
‘Rational use of high-flow therapy in infants with bronchiolitis. What do the latest trials tell us?’ A Paediatric Research in Emergency Departments International Collaborative perspective
.
J Paediatr Child Health
.
2019
;
55
(
7
):
746
752

Supplementary data