Multicenter Study of High-Flow Nasal Cannula Initiation and Duration of Use in Bronchiolitis Free

Hospitals participating in the American Academy of Pediatrics (AAP) Value in Inpatient Pediatrics Network national QI collaborative “High Flow Interventions to Facilitate Less Overuse” (HIFLO) (AAP institutional review board: 21 CHA 02) contributed to this study. The Value in Inpatient Network is an acute care QI network of >250 hospitals across North America that aims to improve the value of health care provided to hospitalized pediatric patients. Member hospitals range widely in size, type (children’s or community), ownership model (private or nonprofit), and location (rural or urban). HIFLO was advertised via the AAP Web site and applicable listservs. Hospitals elected to participate for a variety of reasons, including an identified need to reduce overutilization of HFNC and/or fulfillment of maintenance of certification credit. All hospitals with a completed application were accepted and were required to pay a modest participation fee. This study included hospitals that submitted all baseline data at time of analysis.

Sites provided data via self-report on hospital characteristics, specific elements of local HFNC policies, and any systemwide changes to bronchiolitis and/or HFNC management that took place during the baseline period. Hospital characteristics included bed size, academic affiliation, and freestanding children’s hospital status. HFNC policy elements included whether the emergency department (ED), inpatient unit, or PICU had protocols that (1) incorporated a step to intentionally delay HFNC initiation, a “pause,” (2) restricted initial HFNC management to the PICU, or (3) promoted rapid HFNC weaning. The pause and rapid HFNC weaning definitions were broadly defined. Hospitals were asked if the ED and/or inpatient units had “any specific efforts targeting delay in HFNC initiation” and if the “inpatient bronchiolitis HFNC protocol had a rapid weaning process.” These specific policy elements were identified by the HIFLO collaborative leadership as potential QI targets on the basis of emerging evidence at the time of project development supporting substitution of low flow nasal cannula to delay and potentially avoid HFNC initiation,¹⁴  as well as HFNC rapid weaning protocols.^11,16  Hospitals also provided free-text information on systemwide HFNC changes that occurred during the included time frame.

Patients admitted for bronchiolitis in a baseline 16-week period between November 2019 and March 2020 were included (pre-coronavirus disease 2019 [COVID-19] pandemic), during which time no HIFLO collaborative-directed QI interventions occurred. Participating sites retrospectively reviewed medical records for HFNC use, initiation location, treatment duration, and LOS. Hospitals included either all bronchiolitis admissions or a random sample of 20 patients per week at larger-volume sites using a random number generator.

Eligible patients were those aged 30 days to 23 months admitted from the hospital’s ED to their inpatient unit or PICU, completed their clinical course at the participating hospital, and had a discharge diagnosis of bronchiolitis (International Classification of Diseases, 10th Revision, code J21.xx). Exclusion criteria included prematurity <32 weeks, apnea on presentation, transfer from another hospital, and comorbidities that may impact the bronchiolitis clinical course, including chronic lung disease, congenital heart disease requiring medications, or neuromuscular disorders requiring airway clearance or assistance with feeding or breathing. Additionally, patients were excluded if they required mechanical ventilation, including NIV (other than HFNC) and intubation, because several randomized controlled trials have demonstrated that early HFNC use in bronchiolitis does not impact intubation rates.^8–10  Reviewers were not given specific instructions about children with comorbid pneumonia and/or reactive airway disease. Multiple encounters from the same patient were included if the patient otherwise met criteria.

We used descriptive statistics to characterize patient demographics, hospital characteristics and markers of HFNC utilization (HFNC initiation and treatment duration), and LOS. Frequencies with proportions were used for the categorical outcome of HFNC initiation, and geometric means with confidence intervals (CIs) were used for nonnormally distributed continuous outcomes of HFNC duration and hospital LOS.

We analyzed markers of HFNC utilization by the policy elements outlined above, using generalized linear models for each policy element, and adjusting for children’s hospital status and non-PICU pediatric bed size. We controlled for children’s hospital status over university affiliation because the 2 characteristics were colinear, and patient distribution was more evenly split between freestanding and nonfreestanding children’s hospitals. Hospitals that restricted HFNC management to the PICU were excluded from inpatient analyses.

A binomial distribution with a log-link was used for HFNC initiation and a γ distribution with an identity link was used for HFNC treatment duration and hospital LOS. Sum contrasts were used in all models to set the reference to the overall study population. Odds ratios with 95% CI were derived from the effect estimates of the HFNC initiation models. Differences from the overall geometric mean with 95% CIs were derived from the effect estimates in the HFNC treatment duration and LOS models. All analyses were performed using R version 4.0.4 (Vienna, Austria).

A total of 8296 bronchiolitis hospitalizations were included from 61 hospitals across the United States and Canada. More than half (59%) were male; median age was 6 months (interquartile range, 3–12). Across all hospitals, 2734 (27%) patients were aged >12 months, with 1797 (24%) among hospitals that permit HFNC on the inpatient unit and 266 (20%) in hospitals that require HFNC management in the PICU. HFNC was initiated in 52% of hospitalizations (95% CI, 51 to 54), with a range of 11% to 93%. The most frequent HFNC initiation site was in the ED (75%). The geometric mean for HFNC treatment duration was 52.7 hours (95% CI, 51.5 to 54.0) and geometric mean LOS was 63.4 hours (95% CI, 62.2 to 64.6). Table 1 presents mean initiation rate, geometric mean HFNC duration of use (hours), and geometric mean LOS (hours) by hospital characteristic.

All hospitals included in this study had an on-site PICU and 8 required initial HFNC management to occur in the PICU. Seven hospitals reported having an ED delay protocol, and 4 of these reported both an ED and inpatient pause protocol. No hospitals that required HFNC management in the PICU reported an ED pause protocol. Twenty-nine hospitals reported modifications to local HFNC policies during the baseline time frame, with most intended to increase the use of HFNC. Free-text responses were reviewed by project leads and broadly categorized; the most common interventions included allowing increased flow rates/implementation of weight-based flow guidelines (30%), addition of weaning guidelines (30%), and allowing HFNC use on the inpatient units (17%).

Table 2 presents mean initiation rate, geometric mean HFNC duration of use (hours), and geometric mean LOS (hours) by elements of HFNC policy. Adjusted odds of HFNC initiation (Fig 1A), geometric mean HFNC duration of use (Fig 1B), and geometric mean hospital LOS (Fig 1C) differed when compared by presence of specific policy elements (Fig 1 and Supplemental Table 3). Relative to the total population, hospitals that limited initial HFNC use to the PICU had 70% lower odds of HFNC initiation (odds ratio, 0.3; 95% CI, 0.3 to 0.4) and a 4-hour decrease in geometric mean hospital LOS (−3.9 hours; 95% CI, −7.4 to −0.2), with no difference in HFNC treatment duration. Hospitals with an ED pause protocol had a 12-hour decrease in geometric mean HFNC duration (−12.3 hours; 95% CI, −15.6 to −8.8) and a 15-hour decrease in geometric mean hospital LOS (−14.9 hours; 95% CI, −18.2 to −11.6), with similar odds of HFNC initiation relative to the overall study population.

This large retrospective cohort study identified high rates of HFNC utilization in patients admitted with bronchiolitis during the 2019–2020 bronchiolitis season. Hospitals that limited management of patients on HFNC to the PICU had significantly reduced odds of initiation. Hospitals with an ED pause had reduced HFNC treatment duration and LOS, and sites with an inpatient pause had modestly reduced LOS and odds of HFNC initiation.

Previous single-center observational studies report a wide range of HFNC initiation rates (29%–67%) in similar patient populations.^11,12,14  Estimates of HFNC duration also ranged widely from 39 to 72 hours.^16–18  In this large multicenter cohort, HFNC initiation rates and duration of use were on the higher end of previously reported ranges.

Hospitals that limited HFNC management to the PICU had 70% reduced odds of initiating HFNC with similar HFNC treatment duration and shorter LOS; this is congruent with a study that reports increased HFNC usage when HFNC is expanded outside of the PICU.¹³  Similar HFNC treatment duration and shorter LOS are a surprising finding because admission to the PICU should indicate greater severity of illness, thus one might expect these patients to have longer HFNC treatment duration and LOS. There is no widely accepted way to control for severity in bronchiolitis, and criteria for use of PICU and NIV are institution-specific.^19,20  Furthermore, even if patients in the PICU-only sites are no more ill, the reduced odds of initiation with similar outcomes support limiting HFNC to the PICU as a potential QI intervention.

Participating hospitals that endorsed having an ED protocol with an HFNC initiation pause had shorter HFNC treatment duration and LOS. Although the presence of an ED initiation pause was not associated with decreased odds of initiation when compared with the group mean, hospitals without this policy element had increased odds of initiation. The number of hospitals contributing data to this subgroup was small, which likely impacted our ability to detect any difference, but still represents a reasonable QI intervention. It is unclear why hospitals with an ED pause had shorter HFNC treatment duration and LOS despite similar odds of initiating HFNC; this finding may be related to the fact that these sites were more likely to curtail HFNC utilization across multiple care areas (inpatient pause, rapid weaning protocol). In contrast to 2 recent QI studies that implemented rapid HFNC weaning protocols with a subsequent reduction in HFNC treatment duration and LOS,^11,16  hospitals that endorsed a rapid HFNC weaning protocol had similar HFNC utilization and LOS as the total population.

Our study has several limitations. First, participating hospitals were specifically interested in reducing HFNC utilization in bronchiolitis, thus the very high initiation rates may not be representative of all hospitals caring for infants with bronchiolitis. Nearly half the hospitals included in this article reported QI efforts to improve bronchiolitis care across multiple clinical settings, likely reflecting a high level of engagement that may be in part responsible for some of the significant findings.

The primary goal of this study was descriptive; comparisons of outcomes by hospital HFNC policy elements to the total population are correlative and cannot establish causality. Some findings are driven by relatively few hospitals compared with the total study population, limiting generalizability. There is no validated tool to reliably assess severity of illness in bronchiolitis, and PICU utilization and mechanical ventilation use vary widely by hospital.^19,20  It is possible that these factors influenced our findings in ways we are unable to measure. Finally, data are from a single bronchiolitis season before the COVID-19 pandemic and do not account for variations in seasonality or severity, nor predict HFNC usage in the wake of COVID-19’s effect on bronchiolitis patterns.²¹ 

In conclusion, in our study cohort of >8000 infants hospitalized with bronchiolitis, HFNC was initiated in >50% of patients. Through identification of policy elements that promote judicious HFNC utilization, our study may inform future QI endeavors. Protocolized efforts such as restricting HFNC to the PICU, pausing before HFNC initiation, and rapidly weaning HFNC may represent potential deimplementation strategies.