Eric R Coon, MD, MS

Bronchiolitis is the leading cause of hospitalization in infants aged <1 year in the United States.^1-3 Estimates suggest that 1.5% to 2.0% of US infants require hospitalization every year, with a median (interquartile range) length of stay of 2 days (1-4),³ incurring direct medical costs of $555 million annually.¹ Evidence suggests that few interventions, aside from supportive care, are effective for bronchiolitis.^4-7 Adherence to standardized clinical guidelines could improve outcomes and resource use by streamlining care and limiting ineffective interventions, thereby decreasing hospital length of stay, which is a major medical cost.^8-13 For this reason, many hospitals have adopted bronchiolitis guidelines, although institutional practices vary.^14,15

Two relatively unexplored methods to reduce the inpatient burden of bronchiolitis are the use of observation units (OU) and home oxygen therapy (HOT). Motivated by research demonstrating the safety and effectiveness of an emergency department (ED)–based HOT protocol,¹⁶ where 36 of 37 patients with mild hypoxemia discharged on HOT avoided hospital admission, our institution implemented an observation unit and home oxygen therapy (OU-HOT) protocol designed to return children with bronchiolitis home earlier from the hospital. In the first winter season of implementation (2010 to 2011), the OU-HOT protocol was associated with significant reductions in length of stay and substantial cost savings, without an increase in return visits to the ED or inpatient readmissions.¹⁷ The objectives of this study were to determine whether these encouraging initial findings persisted and to measure the long-term impact of the OU-HOT protocol.

We conducted a retrospective cohort study of children hospitalized with bronchiolitis at Primary Children’s Hospital, a freestanding children’s hospital in Salt Lake City, Utah. Discharge diagnosis and procedures codes, as well as laboratory, imaging, pharmacy, and supply costs, were obtained from the Intermountain Healthcare enterprise data warehouse. A crosswalk available from the Centers for Medicare and Medicaid Services was used to convert International Classification of Diseases (ICD)-10 discharge diagnosis and procedure codes to ICD-9 equivalents.¹⁸ This study was approved by the University of Utah institutional review board (00110419).

Patients

Children aged 3 to 24 months who were discharged with a diagnosis of bronchiolitis (466.xx) during winter seasons from 2007 to 2019 were included. A winter season was defined as November 1 to April 30. Both observation and inpatient encounters were included in the cohort. We excluded patients with discharge diagnosis or procedure codes indicating tracheostomy (519.0-519.09, V44.0, V55.0, 31.1, 31.21, 31.41, 31.74, 97.23), ventilator dependence (V46.1x), chronic lung disease (518.83, 770.7), or pulmonary hypertension (416.xx). Patients with both bronchiolitis and a concurrent diagnosis, such as otitis media or pneumonia, were included unless exclusion criteria were met.

Intervention and Process Measures

Our institution implemented the OU-HOT protocol at the start of the 2010-2011 winter season.¹⁷ The aim of the OU-HOT protocol was to discharge children with bronchiolitis home sooner by increasing use of both an OU, with frequent assessment of discharge readiness, and HOT to help children become ready for discharge. Similar to most OUs, admission to our unit was limited to patients who met hospital admission criteria, and had a short anticipated length of stay (<48 hours). As a self-contained 20-bed unit providing 24-hour dedicated pediatrician/pediatric emergency medicine physician and nursing coverage, the OU actively monitored patients’ discharge readiness, with a goal to facilitate patient throughput more akin to an ED rather than a traditional inpatient unit. Patients who could not be discharged from the OU within 48 hours were transferred to the inpatient unit. Although the OU existed at the time of protocol implementation, its use for patients with bronchiolitis was not actively encouraged until implementation.

Hospitalized patients—in either inpatient or observation units—were eligible for discharge on HOT if they met the following criteria: hypoxemia was the only indication for continued hospitalization, the child’s oxygen requirement was <0.5 L/min for at least 6 hours (0.8 L/min for children aged >1 year), the child’s caregiver(s) were willing to manage oxygen at home, and the child had reliable access to primary care provider follow up. We used two process measures across winter seasons: (1) the percentage of patients discharged from the OU, and (2) the percentage of patients discharged with HOT. The percentage of patients discharged on HOT was estimated by a manual chart review and an electronic medical record (EMR) HOT flag that came into existence with our hospital system’s adoption of a new EMR (2017-2019). Chart review randomly sampled patients from 2007-2017, totaling 457 patients. To estimate the reliability of this method, we calculated the sensitivity, specificity, positive predictive value, and negative predictive value of the EMR HOT flag using chart review as the gold standard.

Outcome Measures

The main outcome measure was mean hospital length of stay. Balancing measures were revisit rates (stratified into ED visits and readmissions) and annual per-population bronchiolitis admission rates. Visits were considered revisits if they occurred within 7 days of initial hospital discharge, and included visits to Primary Children’s Hospital as well as 22 other Intermountain Healthcare hospitals. Population estimates from the Utah Department of Health were used to calculate the annual population-based rate of bronchiolitis admissions to Primary Children’s Hospital.¹⁹ Annual admission rates were calculated per 10,000 children aged 3 to 24 months who resided in Utah each year of the study period, and were evaluated to determine if patients were admitted more frequently after OU-HOT implementation. Secondary outcome measures included the percentage of patients discharged within 24 hours and mean inflation-adjusted cost per episode of care (in 2019 dollars). Hospitalization costs were determined using Intermountain Healthcare’s internal cost accounting system, an activity-based method that aggregates costs of individual resources according to date of service.²⁰ Costs were adjusted to 2019 dollars and were defined as the total costs of a patient’s initial hospitalization as well as any 7-day revisit encounters.

Data Analysis

Demographic data were compared before and after OU-HOT protocol implementation using Pearson chi-square tests. Multivariable linear or logistic regression models were used to compare measures before and after OU-HOT protocol implementation via an interrupted time-series approach. The interrupted time-series analysis measured two types of changes after protocol implementation during the 2010-2011 winter season: (1) any immediate change in the level of an outcome (immediate effect) and (2) any change of an outcome going forward over time (change in slope).²¹ Covariates in the regression models included patient age, sex, race, ethnicity, and insurance type, as well as presence of an underlying complex chronic condition, mechanical ventilation use, and pediatric intensive care unit (PICU) admission during hospitalization. Data were analyzed in STATA 15 (StataCorp LLC).²²

A total of 7,116 patients met inclusion criteria over the study period (2,061 pre-implementation, 5,055 post-implementation). A comparison of patient characteristics before and after HOT protocol implementation is presented in Table 1. Patients were similar in terms of age, sex, and insurance type. Patients in the postimplementation period were more likely to have a complex chronic condition, require admission to the PICU, and need mechanical ventilation (P < .01). Differences between cohorts with regard to race/ethnicity distribution largely were a result of improved capture of these data elements in the postimplementation period. For example, 30% of patients were classified as “race/ethnicity unknown” in the preimplementation cohort, compared with 4% of patients in the postimplementation period.

Process Measures

Figure 1 shows trends in OU and HOT use by winter season. The percentage of patients discharged from the OU increased immediately after OU-HOT protocol implementation (absolute 26.9% immediate increase; 95% CI, 21.9-42.2). The change in the proportion of OU use per season also increased (change in slope +3.9% per season; 95% CI, 3.4%-4.4%). The percentage of patients discharged with HOT increased immediately after OU-HOT protocol implementation (26.0% immediate change; 95% CI, 18.9%-33.1%); however, the immediate increase in HOT discharges was coupled with a declining rate of HOT discharges per season in the postprotocol period compared with the preprotocol period (change in slope –4.5% per season; 95% CI, –7.5% to –1.5%). Our chart review and EMR flag included 1,354 patients, or 19.0% of our cohort. Our EMR flag for HOT in the last two seasons of the study had a positive predictive value of 100% (5 of 5 identified by EMR flag as receiving HOT were confirmed by chart review) and negative predictive value of 89% (31 of 35 identified by EMR flag as not receiving HOT were confirmed by chart review). The specificity of the EMR flag was 100% (31 of 31 of those confirmed by chart review as not receiving HOT, who were correctly identified by EMR) and the sensitivity was 55% (5 of 9 of those confirmed by chart review as receiving HOT, who were correctly identified by EMR).

Primary and Secondary Outcomes

Trends in length of stay across winter seasons are presented in Figure 2. The OU-HOT protocol was associated with an immediate reduction of 30.6 hours in mean length of stay (95% CI, –37.1 to –24.2). The rate of change in length of stay postimplementation did not differ significantly from the rate of change preimplementation (change in slope –0.6 hours per season; 95% CI, –2.3 to 1.1 hours). The percentage of patients discharged within 24 hours of admission rose immediately after protocol implementation, by 23.8 absolute percentage points (95% CI, 11.7-28.8). Slopes of the preintervention and postintervention regression lines did not differ significantly (change in slope –0.1% per season; 95% CI, –1.4% to 1.1%). Immediate decreases in length of stay were accompanied by an immediate decrease in mean cost per episode of care (–$4,181; 95% CI, –$4,829 to –$3,533). Protocol implementation also was associated with a decreased slope in cost postimplementation (change in slope –$403 per season; 95% CI, –$543 to –$264). The total cost savings, estimated by the product of the average cost savings per episode of care and the number of bronchiolitis admissions included in the study after OU-HOT implementation, amounted to $21.1 million over the 9-year period, or $2.3 million per winter season.

Balancing Measures

We observed an immediate reduction in 7-day hospital revisits (–1.1% immediate change; 95% CI, –1.8% to –0.4%), but an increasing slope in revisits after implementation (change in slope 0.4% per season; 95% CI, 0.1%-0.8%) (Figure 3). Stratifying revisits into ED visits and readmissions revealed that the revisit findings reflected changes in ED return visits, for which there was an immediate reduction at the time of implementation (–1.0% immediate change; 95% CI, –1.6% to –0.4%), but an increasing slope postimplementation (change in slope 0.5% per season; 95% CI, 0.2-0.8). Neither an immediate intervention effect (0.0% immediate change; 95% CI, –0.5% to 0.4%) nor a change in slope (change in slope 0.0% per season; 95% CI, –0.1% to 0.1%) were observed for inpatient readmissions alone. The annual rate of bronchiolitis admissions to Primary Children’s Hospital per 10,000 children who reside in Utah decreased after implementation of the OU-HOT protocol (immediate intervention effect –6.2 admissions; 95% CI, –10.8 to –1.6; change in slope –1.8 admissions per season; 95% CI, –2.8 to –0.69).

Our OU-HOT protocol was associated with immediate improvements in care delivered to children hospitalized for bronchiolitis, including decreased length of stay and cost savings. These improvements in outcomes largely have been sustained over a 9-year period. The OU-HOT protocol also appears to be safe as evidenced by a stable rate of readmissions over the study period and only a small increase in revisits to EDs across Intermountain Healthcare facilities, which see most children in the catchment area. Our OU-HOT protocol represents a combination of two interventions: (1) the creation of an OU focused on discharge within 24 to 48 hours of admission and (2) encouragement to discharge children with HOT. We found that use of the OU and a commitment to timely discharges has been sustained in recent years, while the commitment to HOT has appeared to wane.

Earlier investigations have evaluated the efficacy of HOT in the ED setting to prevent hospital admissions, finding high levels of caregiver comfort, estimating $1,300 per patient cost savings, and reporting readmission rates of approximately 5%.^16,23-25 Our study is unique in addressing HOT among a population of patients already hospitalized with bronchiolitis. The cost reductions we observed with our OU-HOT protocol were similar to those noted in the ED-based HOT protocols. However, we recorded lower readmission rates, likely because of the additional time allotted to caregivers to better gauge illness trajectory in the inpatient setting vs the ED, as well as additional time for hospitalized patients to reach the plateau or convalescent phase of illness. The small increase in ED revisits that we measured in recent years might be related to the concurrent rise in patient acuity and complexity.

Considering that length of stay has remained low despite less commitment to HOT, our results suggest that the OU might be the more impactful of the two interventions, and these data support the use of such a unit for a subset of patients with bronchiolitis. However, it is important to note that while the EMR HOT flag demonstrated high specificity, positive predictive value, and negative predictive value, the sensitivity was low (56%). As a result, it is possible that we have underestimated HOT use in the 2017-2018 and 2018-2019 seasons, the final two years of the study. Alternatively, the discrepancy between sustained outcomes and lagging use of HOT could be explained by improved identification of patients who would experience the greatest benefit with oxygen in terms of length of stay reductions, with fewer patients discharged on HOT but greater per-patient benefit. Finally, in an era that encourages reduced monitor use and less aggressive response to transient mild desaturations,^13,26,27 it is possible that fewer patients are identified with clinically actionable hypoxemia around the time they would be otherwise discharged.

Our OU-HOT model is not unprecedented. Increasingly, other formerly inpatient indications are being successfully managed in the observation, outpatient, and home setting, such as parenteral antibiotic treatment^28,29 and chemotherapy administration.³⁰ Considering the inpatient burden of bronchiolitis, similar strategies to expedite discharge are needed. Although outpatient intravenous antibiotic and chemotherapy administration have been widely adopted, we are aware of only one other pediatric health care system in the United States (Children’s Hospital Colorado) that routinely discharges inpatients with bronchiolitis on HOT.

This study has several limitations. First, although the interrupted time-series analysis is designed to account for trends that precede an intervention and covariates that differ before and after the intervention, it is possible that important unmeasured patient factors or changes in practice patterns differed between the pre- and post-intervention cohorts. There were no major changes to the OU-HOT protocol or discharge criteria after implementation, but individual practice management of bronchiolitis during the study period likely has evolved as new evidence emerges. Second, one could postulate that the increase in discharges within 24 hours and accompanying decreases in average length of stay and cost could be achieved by hospitalizing healthier patients over time, which the presence of an OU might incentivize. To the contrary, we found that population-based bronchiolitis admission rates have declined and disease severity appears to be increased since implementation of the OU-HOT protocol. The increase in medically complex children and PICU use in our postimplementation cohort aligns with recently published data suggesting these are national trends.^3,31 Third, HOT use was estimated from a sample of the cohort using a chart review and a newly available EMR flag. A low sensitivity and a small sample for the positive predictive value are limitations of the EMR flag.

Additionally, there are almost certainly unmeasured ambulatory burdens of HOT not captured by this study. ED-based protocols have estimated that patients discharged with HOT have a median of two follow-up ambulatory visits before oxygen is discontinued³²; however, the ambulatory burden associated with discharge on HOT after a hospitalization and the extent to which demographic factors affect that burden is unknown. Furthermore, one insurance company charged $94 for a month of HOT in 2019; paying even a portion of this charge represents a nontrivial financial burden for many families, even considering inpatient cost savings. Although the decision to discharge on oxygen or remain hospitalized until the child did not need oxygen was left to the parents, their posthospitalization perspectives were not assessed in this study. Although reports indicate that families largely feel positive about HOT after discharge from an ED setting, with 90% of caregivers preferring HOT use to inpatient admission and most reporting no difficulty with home management,²³ it is uncertain whether this would also apply after inpatient hospitalization.

The OU-HOT bronchiolitis protocol was associated with decreases in inpatient length of stay and cost while appearing safe to implement. The sustained use of the OU combined with declining use of HOT suggests that the OU might be the more impactful intervention. As previously inpatient indications such as parenteral antibiotics and chemotherapy increasingly have been administered in observation and outpatient settings, bronchiolitis appears ideal for a similar strategy that allows patients to spend less time in the hospital. Studies are needed to understand the outpatient burden of HOT and the generalizability of our findings.

Bronchiolitis is the leading cause of hospitalization in infants aged <1 year in the United States.^1-3 Estimates suggest that 1.5% to 2.0% of US infants require hospitalization every year, with a median (interquartile range) length of stay of 2 days (1-4),³ incurring direct medical costs of $555 million annually.¹ Evidence suggests that few interventions, aside from supportive care, are effective for bronchiolitis.^4-7 Adherence to standardized clinical guidelines could improve outcomes and resource use by streamlining care and limiting ineffective interventions, thereby decreasing hospital length of stay, which is a major medical cost.^8-13 For this reason, many hospitals have adopted bronchiolitis guidelines, although institutional practices vary.^14,15

Two relatively unexplored methods to reduce the inpatient burden of bronchiolitis are the use of observation units (OU) and home oxygen therapy (HOT). Motivated by research demonstrating the safety and effectiveness of an emergency department (ED)–based HOT protocol,¹⁶ where 36 of 37 patients with mild hypoxemia discharged on HOT avoided hospital admission, our institution implemented an observation unit and home oxygen therapy (OU-HOT) protocol designed to return children with bronchiolitis home earlier from the hospital. In the first winter season of implementation (2010 to 2011), the OU-HOT protocol was associated with significant reductions in length of stay and substantial cost savings, without an increase in return visits to the ED or inpatient readmissions.¹⁷ The objectives of this study were to determine whether these encouraging initial findings persisted and to measure the long-term impact of the OU-HOT protocol.

We conducted a retrospective cohort study of children hospitalized with bronchiolitis at Primary Children’s Hospital, a freestanding children’s hospital in Salt Lake City, Utah. Discharge diagnosis and procedures codes, as well as laboratory, imaging, pharmacy, and supply costs, were obtained from the Intermountain Healthcare enterprise data warehouse. A crosswalk available from the Centers for Medicare and Medicaid Services was used to convert International Classification of Diseases (ICD)-10 discharge diagnosis and procedure codes to ICD-9 equivalents.¹⁸ This study was approved by the University of Utah institutional review board (00110419).

Patients

Children aged 3 to 24 months who were discharged with a diagnosis of bronchiolitis (466.xx) during winter seasons from 2007 to 2019 were included. A winter season was defined as November 1 to April 30. Both observation and inpatient encounters were included in the cohort. We excluded patients with discharge diagnosis or procedure codes indicating tracheostomy (519.0-519.09, V44.0, V55.0, 31.1, 31.21, 31.41, 31.74, 97.23), ventilator dependence (V46.1x), chronic lung disease (518.83, 770.7), or pulmonary hypertension (416.xx). Patients with both bronchiolitis and a concurrent diagnosis, such as otitis media or pneumonia, were included unless exclusion criteria were met.

Intervention and Process Measures

Our institution implemented the OU-HOT protocol at the start of the 2010-2011 winter season.¹⁷ The aim of the OU-HOT protocol was to discharge children with bronchiolitis home sooner by increasing use of both an OU, with frequent assessment of discharge readiness, and HOT to help children become ready for discharge. Similar to most OUs, admission to our unit was limited to patients who met hospital admission criteria, and had a short anticipated length of stay (<48 hours). As a self-contained 20-bed unit providing 24-hour dedicated pediatrician/pediatric emergency medicine physician and nursing coverage, the OU actively monitored patients’ discharge readiness, with a goal to facilitate patient throughput more akin to an ED rather than a traditional inpatient unit. Patients who could not be discharged from the OU within 48 hours were transferred to the inpatient unit. Although the OU existed at the time of protocol implementation, its use for patients with bronchiolitis was not actively encouraged until implementation.

Hospitalized patients—in either inpatient or observation units—were eligible for discharge on HOT if they met the following criteria: hypoxemia was the only indication for continued hospitalization, the child’s oxygen requirement was <0.5 L/min for at least 6 hours (0.8 L/min for children aged >1 year), the child’s caregiver(s) were willing to manage oxygen at home, and the child had reliable access to primary care provider follow up. We used two process measures across winter seasons: (1) the percentage of patients discharged from the OU, and (2) the percentage of patients discharged with HOT. The percentage of patients discharged on HOT was estimated by a manual chart review and an electronic medical record (EMR) HOT flag that came into existence with our hospital system’s adoption of a new EMR (2017-2019). Chart review randomly sampled patients from 2007-2017, totaling 457 patients. To estimate the reliability of this method, we calculated the sensitivity, specificity, positive predictive value, and negative predictive value of the EMR HOT flag using chart review as the gold standard.

Outcome Measures

The main outcome measure was mean hospital length of stay. Balancing measures were revisit rates (stratified into ED visits and readmissions) and annual per-population bronchiolitis admission rates. Visits were considered revisits if they occurred within 7 days of initial hospital discharge, and included visits to Primary Children’s Hospital as well as 22 other Intermountain Healthcare hospitals. Population estimates from the Utah Department of Health were used to calculate the annual population-based rate of bronchiolitis admissions to Primary Children’s Hospital.¹⁹ Annual admission rates were calculated per 10,000 children aged 3 to 24 months who resided in Utah each year of the study period, and were evaluated to determine if patients were admitted more frequently after OU-HOT implementation. Secondary outcome measures included the percentage of patients discharged within 24 hours and mean inflation-adjusted cost per episode of care (in 2019 dollars). Hospitalization costs were determined using Intermountain Healthcare’s internal cost accounting system, an activity-based method that aggregates costs of individual resources according to date of service.²⁰ Costs were adjusted to 2019 dollars and were defined as the total costs of a patient’s initial hospitalization as well as any 7-day revisit encounters.

Data Analysis

Demographic data were compared before and after OU-HOT protocol implementation using Pearson chi-square tests. Multivariable linear or logistic regression models were used to compare measures before and after OU-HOT protocol implementation via an interrupted time-series approach. The interrupted time-series analysis measured two types of changes after protocol implementation during the 2010-2011 winter season: (1) any immediate change in the level of an outcome (immediate effect) and (2) any change of an outcome going forward over time (change in slope).²¹ Covariates in the regression models included patient age, sex, race, ethnicity, and insurance type, as well as presence of an underlying complex chronic condition, mechanical ventilation use, and pediatric intensive care unit (PICU) admission during hospitalization. Data were analyzed in STATA 15 (StataCorp LLC).²²

A total of 7,116 patients met inclusion criteria over the study period (2,061 pre-implementation, 5,055 post-implementation). A comparison of patient characteristics before and after HOT protocol implementation is presented in Table 1. Patients were similar in terms of age, sex, and insurance type. Patients in the postimplementation period were more likely to have a complex chronic condition, require admission to the PICU, and need mechanical ventilation (P < .01). Differences between cohorts with regard to race/ethnicity distribution largely were a result of improved capture of these data elements in the postimplementation period. For example, 30% of patients were classified as “race/ethnicity unknown” in the preimplementation cohort, compared with 4% of patients in the postimplementation period.

Process Measures

Figure 1 shows trends in OU and HOT use by winter season. The percentage of patients discharged from the OU increased immediately after OU-HOT protocol implementation (absolute 26.9% immediate increase; 95% CI, 21.9-42.2). The change in the proportion of OU use per season also increased (change in slope +3.9% per season; 95% CI, 3.4%-4.4%). The percentage of patients discharged with HOT increased immediately after OU-HOT protocol implementation (26.0% immediate change; 95% CI, 18.9%-33.1%); however, the immediate increase in HOT discharges was coupled with a declining rate of HOT discharges per season in the postprotocol period compared with the preprotocol period (change in slope –4.5% per season; 95% CI, –7.5% to –1.5%). Our chart review and EMR flag included 1,354 patients, or 19.0% of our cohort. Our EMR flag for HOT in the last two seasons of the study had a positive predictive value of 100% (5 of 5 identified by EMR flag as receiving HOT were confirmed by chart review) and negative predictive value of 89% (31 of 35 identified by EMR flag as not receiving HOT were confirmed by chart review). The specificity of the EMR flag was 100% (31 of 31 of those confirmed by chart review as not receiving HOT, who were correctly identified by EMR) and the sensitivity was 55% (5 of 9 of those confirmed by chart review as receiving HOT, who were correctly identified by EMR).

Primary and Secondary Outcomes

Trends in length of stay across winter seasons are presented in Figure 2. The OU-HOT protocol was associated with an immediate reduction of 30.6 hours in mean length of stay (95% CI, –37.1 to –24.2). The rate of change in length of stay postimplementation did not differ significantly from the rate of change preimplementation (change in slope –0.6 hours per season; 95% CI, –2.3 to 1.1 hours). The percentage of patients discharged within 24 hours of admission rose immediately after protocol implementation, by 23.8 absolute percentage points (95% CI, 11.7-28.8). Slopes of the preintervention and postintervention regression lines did not differ significantly (change in slope –0.1% per season; 95% CI, –1.4% to 1.1%). Immediate decreases in length of stay were accompanied by an immediate decrease in mean cost per episode of care (–$4,181; 95% CI, –$4,829 to –$3,533). Protocol implementation also was associated with a decreased slope in cost postimplementation (change in slope –$403 per season; 95% CI, –$543 to –$264). The total cost savings, estimated by the product of the average cost savings per episode of care and the number of bronchiolitis admissions included in the study after OU-HOT implementation, amounted to $21.1 million over the 9-year period, or $2.3 million per winter season.

Balancing Measures

We observed an immediate reduction in 7-day hospital revisits (–1.1% immediate change; 95% CI, –1.8% to –0.4%), but an increasing slope in revisits after implementation (change in slope 0.4% per season; 95% CI, 0.1%-0.8%) (Figure 3). Stratifying revisits into ED visits and readmissions revealed that the revisit findings reflected changes in ED return visits, for which there was an immediate reduction at the time of implementation (–1.0% immediate change; 95% CI, –1.6% to –0.4%), but an increasing slope postimplementation (change in slope 0.5% per season; 95% CI, 0.2-0.8). Neither an immediate intervention effect (0.0% immediate change; 95% CI, –0.5% to 0.4%) nor a change in slope (change in slope 0.0% per season; 95% CI, –0.1% to 0.1%) were observed for inpatient readmissions alone. The annual rate of bronchiolitis admissions to Primary Children’s Hospital per 10,000 children who reside in Utah decreased after implementation of the OU-HOT protocol (immediate intervention effect –6.2 admissions; 95% CI, –10.8 to –1.6; change in slope –1.8 admissions per season; 95% CI, –2.8 to –0.69).

Our OU-HOT protocol was associated with immediate improvements in care delivered to children hospitalized for bronchiolitis, including decreased length of stay and cost savings. These improvements in outcomes largely have been sustained over a 9-year period. The OU-HOT protocol also appears to be safe as evidenced by a stable rate of readmissions over the study period and only a small increase in revisits to EDs across Intermountain Healthcare facilities, which see most children in the catchment area. Our OU-HOT protocol represents a combination of two interventions: (1) the creation of an OU focused on discharge within 24 to 48 hours of admission and (2) encouragement to discharge children with HOT. We found that use of the OU and a commitment to timely discharges has been sustained in recent years, while the commitment to HOT has appeared to wane.

Earlier investigations have evaluated the efficacy of HOT in the ED setting to prevent hospital admissions, finding high levels of caregiver comfort, estimating $1,300 per patient cost savings, and reporting readmission rates of approximately 5%.^16,23-25 Our study is unique in addressing HOT among a population of patients already hospitalized with bronchiolitis. The cost reductions we observed with our OU-HOT protocol were similar to those noted in the ED-based HOT protocols. However, we recorded lower readmission rates, likely because of the additional time allotted to caregivers to better gauge illness trajectory in the inpatient setting vs the ED, as well as additional time for hospitalized patients to reach the plateau or convalescent phase of illness. The small increase in ED revisits that we measured in recent years might be related to the concurrent rise in patient acuity and complexity.

Considering that length of stay has remained low despite less commitment to HOT, our results suggest that the OU might be the more impactful of the two interventions, and these data support the use of such a unit for a subset of patients with bronchiolitis. However, it is important to note that while the EMR HOT flag demonstrated high specificity, positive predictive value, and negative predictive value, the sensitivity was low (56%). As a result, it is possible that we have underestimated HOT use in the 2017-2018 and 2018-2019 seasons, the final two years of the study. Alternatively, the discrepancy between sustained outcomes and lagging use of HOT could be explained by improved identification of patients who would experience the greatest benefit with oxygen in terms of length of stay reductions, with fewer patients discharged on HOT but greater per-patient benefit. Finally, in an era that encourages reduced monitor use and less aggressive response to transient mild desaturations,^13,26,27 it is possible that fewer patients are identified with clinically actionable hypoxemia around the time they would be otherwise discharged.

Our OU-HOT model is not unprecedented. Increasingly, other formerly inpatient indications are being successfully managed in the observation, outpatient, and home setting, such as parenteral antibiotic treatment^28,29 and chemotherapy administration.³⁰ Considering the inpatient burden of bronchiolitis, similar strategies to expedite discharge are needed. Although outpatient intravenous antibiotic and chemotherapy administration have been widely adopted, we are aware of only one other pediatric health care system in the United States (Children’s Hospital Colorado) that routinely discharges inpatients with bronchiolitis on HOT.

This study has several limitations. First, although the interrupted time-series analysis is designed to account for trends that precede an intervention and covariates that differ before and after the intervention, it is possible that important unmeasured patient factors or changes in practice patterns differed between the pre- and post-intervention cohorts. There were no major changes to the OU-HOT protocol or discharge criteria after implementation, but individual practice management of bronchiolitis during the study period likely has evolved as new evidence emerges. Second, one could postulate that the increase in discharges within 24 hours and accompanying decreases in average length of stay and cost could be achieved by hospitalizing healthier patients over time, which the presence of an OU might incentivize. To the contrary, we found that population-based bronchiolitis admission rates have declined and disease severity appears to be increased since implementation of the OU-HOT protocol. The increase in medically complex children and PICU use in our postimplementation cohort aligns with recently published data suggesting these are national trends.^3,31 Third, HOT use was estimated from a sample of the cohort using a chart review and a newly available EMR flag. A low sensitivity and a small sample for the positive predictive value are limitations of the EMR flag.

Additionally, there are almost certainly unmeasured ambulatory burdens of HOT not captured by this study. ED-based protocols have estimated that patients discharged with HOT have a median of two follow-up ambulatory visits before oxygen is discontinued³²; however, the ambulatory burden associated with discharge on HOT after a hospitalization and the extent to which demographic factors affect that burden is unknown. Furthermore, one insurance company charged $94 for a month of HOT in 2019; paying even a portion of this charge represents a nontrivial financial burden for many families, even considering inpatient cost savings. Although the decision to discharge on oxygen or remain hospitalized until the child did not need oxygen was left to the parents, their posthospitalization perspectives were not assessed in this study. Although reports indicate that families largely feel positive about HOT after discharge from an ED setting, with 90% of caregivers preferring HOT use to inpatient admission and most reporting no difficulty with home management,²³ it is uncertain whether this would also apply after inpatient hospitalization.

The OU-HOT bronchiolitis protocol was associated with decreases in inpatient length of stay and cost while appearing safe to implement. The sustained use of the OU combined with declining use of HOT suggests that the OU might be the more impactful intervention. As previously inpatient indications such as parenteral antibiotics and chemotherapy increasingly have been administered in observation and outpatient settings, bronchiolitis appears ideal for a similar strategy that allows patients to spend less time in the hospital. Studies are needed to understand the outpatient burden of HOT and the generalizability of our findings.

Bronchiolitis is the leading cause of hospitalization in infants aged <1 year in the United States.^1-3 Estimates suggest that 1.5% to 2.0% of US infants require hospitalization every year, with a median (interquartile range) length of stay of 2 days (1-4),³ incurring direct medical costs of $555 million annually.¹ Evidence suggests that few interventions, aside from supportive care, are effective for bronchiolitis.^4-7 Adherence to standardized clinical guidelines could improve outcomes and resource use by streamlining care and limiting ineffective interventions, thereby decreasing hospital length of stay, which is a major medical cost.^8-13 For this reason, many hospitals have adopted bronchiolitis guidelines, although institutional practices vary.^14,15

Two relatively unexplored methods to reduce the inpatient burden of bronchiolitis are the use of observation units (OU) and home oxygen therapy (HOT). Motivated by research demonstrating the safety and effectiveness of an emergency department (ED)–based HOT protocol,¹⁶ where 36 of 37 patients with mild hypoxemia discharged on HOT avoided hospital admission, our institution implemented an observation unit and home oxygen therapy (OU-HOT) protocol designed to return children with bronchiolitis home earlier from the hospital. In the first winter season of implementation (2010 to 2011), the OU-HOT protocol was associated with significant reductions in length of stay and substantial cost savings, without an increase in return visits to the ED or inpatient readmissions.¹⁷ The objectives of this study were to determine whether these encouraging initial findings persisted and to measure the long-term impact of the OU-HOT protocol.

We conducted a retrospective cohort study of children hospitalized with bronchiolitis at Primary Children’s Hospital, a freestanding children’s hospital in Salt Lake City, Utah. Discharge diagnosis and procedures codes, as well as laboratory, imaging, pharmacy, and supply costs, were obtained from the Intermountain Healthcare enterprise data warehouse. A crosswalk available from the Centers for Medicare and Medicaid Services was used to convert International Classification of Diseases (ICD)-10 discharge diagnosis and procedure codes to ICD-9 equivalents.¹⁸ This study was approved by the University of Utah institutional review board (00110419).

Patients

Children aged 3 to 24 months who were discharged with a diagnosis of bronchiolitis (466.xx) during winter seasons from 2007 to 2019 were included. A winter season was defined as November 1 to April 30. Both observation and inpatient encounters were included in the cohort. We excluded patients with discharge diagnosis or procedure codes indicating tracheostomy (519.0-519.09, V44.0, V55.0, 31.1, 31.21, 31.41, 31.74, 97.23), ventilator dependence (V46.1x), chronic lung disease (518.83, 770.7), or pulmonary hypertension (416.xx). Patients with both bronchiolitis and a concurrent diagnosis, such as otitis media or pneumonia, were included unless exclusion criteria were met.

Intervention and Process Measures

Our institution implemented the OU-HOT protocol at the start of the 2010-2011 winter season.¹⁷ The aim of the OU-HOT protocol was to discharge children with bronchiolitis home sooner by increasing use of both an OU, with frequent assessment of discharge readiness, and HOT to help children become ready for discharge. Similar to most OUs, admission to our unit was limited to patients who met hospital admission criteria, and had a short anticipated length of stay (<48 hours). As a self-contained 20-bed unit providing 24-hour dedicated pediatrician/pediatric emergency medicine physician and nursing coverage, the OU actively monitored patients’ discharge readiness, with a goal to facilitate patient throughput more akin to an ED rather than a traditional inpatient unit. Patients who could not be discharged from the OU within 48 hours were transferred to the inpatient unit. Although the OU existed at the time of protocol implementation, its use for patients with bronchiolitis was not actively encouraged until implementation.

Hospitalized patients—in either inpatient or observation units—were eligible for discharge on HOT if they met the following criteria: hypoxemia was the only indication for continued hospitalization, the child’s oxygen requirement was <0.5 L/min for at least 6 hours (0.8 L/min for children aged >1 year), the child’s caregiver(s) were willing to manage oxygen at home, and the child had reliable access to primary care provider follow up. We used two process measures across winter seasons: (1) the percentage of patients discharged from the OU, and (2) the percentage of patients discharged with HOT. The percentage of patients discharged on HOT was estimated by a manual chart review and an electronic medical record (EMR) HOT flag that came into existence with our hospital system’s adoption of a new EMR (2017-2019). Chart review randomly sampled patients from 2007-2017, totaling 457 patients. To estimate the reliability of this method, we calculated the sensitivity, specificity, positive predictive value, and negative predictive value of the EMR HOT flag using chart review as the gold standard.

Outcome Measures

The main outcome measure was mean hospital length of stay. Balancing measures were revisit rates (stratified into ED visits and readmissions) and annual per-population bronchiolitis admission rates. Visits were considered revisits if they occurred within 7 days of initial hospital discharge, and included visits to Primary Children’s Hospital as well as 22 other Intermountain Healthcare hospitals. Population estimates from the Utah Department of Health were used to calculate the annual population-based rate of bronchiolitis admissions to Primary Children’s Hospital.¹⁹ Annual admission rates were calculated per 10,000 children aged 3 to 24 months who resided in Utah each year of the study period, and were evaluated to determine if patients were admitted more frequently after OU-HOT implementation. Secondary outcome measures included the percentage of patients discharged within 24 hours and mean inflation-adjusted cost per episode of care (in 2019 dollars). Hospitalization costs were determined using Intermountain Healthcare’s internal cost accounting system, an activity-based method that aggregates costs of individual resources according to date of service.²⁰ Costs were adjusted to 2019 dollars and were defined as the total costs of a patient’s initial hospitalization as well as any 7-day revisit encounters.

Data Analysis

Demographic data were compared before and after OU-HOT protocol implementation using Pearson chi-square tests. Multivariable linear or logistic regression models were used to compare measures before and after OU-HOT protocol implementation via an interrupted time-series approach. The interrupted time-series analysis measured two types of changes after protocol implementation during the 2010-2011 winter season: (1) any immediate change in the level of an outcome (immediate effect) and (2) any change of an outcome going forward over time (change in slope).²¹ Covariates in the regression models included patient age, sex, race, ethnicity, and insurance type, as well as presence of an underlying complex chronic condition, mechanical ventilation use, and pediatric intensive care unit (PICU) admission during hospitalization. Data were analyzed in STATA 15 (StataCorp LLC).²²

A total of 7,116 patients met inclusion criteria over the study period (2,061 pre-implementation, 5,055 post-implementation). A comparison of patient characteristics before and after HOT protocol implementation is presented in Table 1. Patients were similar in terms of age, sex, and insurance type. Patients in the postimplementation period were more likely to have a complex chronic condition, require admission to the PICU, and need mechanical ventilation (P < .01). Differences between cohorts with regard to race/ethnicity distribution largely were a result of improved capture of these data elements in the postimplementation period. For example, 30% of patients were classified as “race/ethnicity unknown” in the preimplementation cohort, compared with 4% of patients in the postimplementation period.

Process Measures

Figure 1 shows trends in OU and HOT use by winter season. The percentage of patients discharged from the OU increased immediately after OU-HOT protocol implementation (absolute 26.9% immediate increase; 95% CI, 21.9-42.2). The change in the proportion of OU use per season also increased (change in slope +3.9% per season; 95% CI, 3.4%-4.4%). The percentage of patients discharged with HOT increased immediately after OU-HOT protocol implementation (26.0% immediate change; 95% CI, 18.9%-33.1%); however, the immediate increase in HOT discharges was coupled with a declining rate of HOT discharges per season in the postprotocol period compared with the preprotocol period (change in slope –4.5% per season; 95% CI, –7.5% to –1.5%). Our chart review and EMR flag included 1,354 patients, or 19.0% of our cohort. Our EMR flag for HOT in the last two seasons of the study had a positive predictive value of 100% (5 of 5 identified by EMR flag as receiving HOT were confirmed by chart review) and negative predictive value of 89% (31 of 35 identified by EMR flag as not receiving HOT were confirmed by chart review). The specificity of the EMR flag was 100% (31 of 31 of those confirmed by chart review as not receiving HOT, who were correctly identified by EMR) and the sensitivity was 55% (5 of 9 of those confirmed by chart review as receiving HOT, who were correctly identified by EMR).

Primary and Secondary Outcomes

Trends in length of stay across winter seasons are presented in Figure 2. The OU-HOT protocol was associated with an immediate reduction of 30.6 hours in mean length of stay (95% CI, –37.1 to –24.2). The rate of change in length of stay postimplementation did not differ significantly from the rate of change preimplementation (change in slope –0.6 hours per season; 95% CI, –2.3 to 1.1 hours). The percentage of patients discharged within 24 hours of admission rose immediately after protocol implementation, by 23.8 absolute percentage points (95% CI, 11.7-28.8). Slopes of the preintervention and postintervention regression lines did not differ significantly (change in slope –0.1% per season; 95% CI, –1.4% to 1.1%). Immediate decreases in length of stay were accompanied by an immediate decrease in mean cost per episode of care (–$4,181; 95% CI, –$4,829 to –$3,533). Protocol implementation also was associated with a decreased slope in cost postimplementation (change in slope –$403 per season; 95% CI, –$543 to –$264). The total cost savings, estimated by the product of the average cost savings per episode of care and the number of bronchiolitis admissions included in the study after OU-HOT implementation, amounted to $21.1 million over the 9-year period, or $2.3 million per winter season.

Balancing Measures

We observed an immediate reduction in 7-day hospital revisits (–1.1% immediate change; 95% CI, –1.8% to –0.4%), but an increasing slope in revisits after implementation (change in slope 0.4% per season; 95% CI, 0.1%-0.8%) (Figure 3). Stratifying revisits into ED visits and readmissions revealed that the revisit findings reflected changes in ED return visits, for which there was an immediate reduction at the time of implementation (–1.0% immediate change; 95% CI, –1.6% to –0.4%), but an increasing slope postimplementation (change in slope 0.5% per season; 95% CI, 0.2-0.8). Neither an immediate intervention effect (0.0% immediate change; 95% CI, –0.5% to 0.4%) nor a change in slope (change in slope 0.0% per season; 95% CI, –0.1% to 0.1%) were observed for inpatient readmissions alone. The annual rate of bronchiolitis admissions to Primary Children’s Hospital per 10,000 children who reside in Utah decreased after implementation of the OU-HOT protocol (immediate intervention effect –6.2 admissions; 95% CI, –10.8 to –1.6; change in slope –1.8 admissions per season; 95% CI, –2.8 to –0.69).

Our OU-HOT protocol was associated with immediate improvements in care delivered to children hospitalized for bronchiolitis, including decreased length of stay and cost savings. These improvements in outcomes largely have been sustained over a 9-year period. The OU-HOT protocol also appears to be safe as evidenced by a stable rate of readmissions over the study period and only a small increase in revisits to EDs across Intermountain Healthcare facilities, which see most children in the catchment area. Our OU-HOT protocol represents a combination of two interventions: (1) the creation of an OU focused on discharge within 24 to 48 hours of admission and (2) encouragement to discharge children with HOT. We found that use of the OU and a commitment to timely discharges has been sustained in recent years, while the commitment to HOT has appeared to wane.

Earlier investigations have evaluated the efficacy of HOT in the ED setting to prevent hospital admissions, finding high levels of caregiver comfort, estimating $1,300 per patient cost savings, and reporting readmission rates of approximately 5%.^16,23-25 Our study is unique in addressing HOT among a population of patients already hospitalized with bronchiolitis. The cost reductions we observed with our OU-HOT protocol were similar to those noted in the ED-based HOT protocols. However, we recorded lower readmission rates, likely because of the additional time allotted to caregivers to better gauge illness trajectory in the inpatient setting vs the ED, as well as additional time for hospitalized patients to reach the plateau or convalescent phase of illness. The small increase in ED revisits that we measured in recent years might be related to the concurrent rise in patient acuity and complexity.

Considering that length of stay has remained low despite less commitment to HOT, our results suggest that the OU might be the more impactful of the two interventions, and these data support the use of such a unit for a subset of patients with bronchiolitis. However, it is important to note that while the EMR HOT flag demonstrated high specificity, positive predictive value, and negative predictive value, the sensitivity was low (56%). As a result, it is possible that we have underestimated HOT use in the 2017-2018 and 2018-2019 seasons, the final two years of the study. Alternatively, the discrepancy between sustained outcomes and lagging use of HOT could be explained by improved identification of patients who would experience the greatest benefit with oxygen in terms of length of stay reductions, with fewer patients discharged on HOT but greater per-patient benefit. Finally, in an era that encourages reduced monitor use and less aggressive response to transient mild desaturations,^13,26,27 it is possible that fewer patients are identified with clinically actionable hypoxemia around the time they would be otherwise discharged.

Our OU-HOT model is not unprecedented. Increasingly, other formerly inpatient indications are being successfully managed in the observation, outpatient, and home setting, such as parenteral antibiotic treatment^28,29 and chemotherapy administration.³⁰ Considering the inpatient burden of bronchiolitis, similar strategies to expedite discharge are needed. Although outpatient intravenous antibiotic and chemotherapy administration have been widely adopted, we are aware of only one other pediatric health care system in the United States (Children’s Hospital Colorado) that routinely discharges inpatients with bronchiolitis on HOT.

This study has several limitations. First, although the interrupted time-series analysis is designed to account for trends that precede an intervention and covariates that differ before and after the intervention, it is possible that important unmeasured patient factors or changes in practice patterns differed between the pre- and post-intervention cohorts. There were no major changes to the OU-HOT protocol or discharge criteria after implementation, but individual practice management of bronchiolitis during the study period likely has evolved as new evidence emerges. Second, one could postulate that the increase in discharges within 24 hours and accompanying decreases in average length of stay and cost could be achieved by hospitalizing healthier patients over time, which the presence of an OU might incentivize. To the contrary, we found that population-based bronchiolitis admission rates have declined and disease severity appears to be increased since implementation of the OU-HOT protocol. The increase in medically complex children and PICU use in our postimplementation cohort aligns with recently published data suggesting these are national trends.^3,31 Third, HOT use was estimated from a sample of the cohort using a chart review and a newly available EMR flag. A low sensitivity and a small sample for the positive predictive value are limitations of the EMR flag.

Additionally, there are almost certainly unmeasured ambulatory burdens of HOT not captured by this study. ED-based protocols have estimated that patients discharged with HOT have a median of two follow-up ambulatory visits before oxygen is discontinued³²; however, the ambulatory burden associated with discharge on HOT after a hospitalization and the extent to which demographic factors affect that burden is unknown. Furthermore, one insurance company charged $94 for a month of HOT in 2019; paying even a portion of this charge represents a nontrivial financial burden for many families, even considering inpatient cost savings. Although the decision to discharge on oxygen or remain hospitalized until the child did not need oxygen was left to the parents, their posthospitalization perspectives were not assessed in this study. Although reports indicate that families largely feel positive about HOT after discharge from an ED setting, with 90% of caregivers preferring HOT use to inpatient admission and most reporting no difficulty with home management,²³ it is uncertain whether this would also apply after inpatient hospitalization.

The OU-HOT bronchiolitis protocol was associated with decreases in inpatient length of stay and cost while appearing safe to implement. The sustained use of the OU combined with declining use of HOT suggests that the OU might be the more impactful intervention. As previously inpatient indications such as parenteral antibiotics and chemotherapy increasingly have been administered in observation and outpatient settings, bronchiolitis appears ideal for a similar strategy that allows patients to spend less time in the hospital. Studies are needed to understand the outpatient burden of HOT and the generalizability of our findings.

In recent years, mounting evidence has emerged questioning the practice of using prolonged intravenous antibiotic therapy to treat certain serious bacterial infections in children, including complicated appendicitis, osteomyelitis, and complicated pneumonia. Historically, treatment of these conditions was often completed intravenously after hospital discharge using peripherally inserted central catheters (PICCs). Line infections, clots, mechanical problems, and general discomfort complicate PICCs, which led to their removal in more than 20% of children in one study.¹ Oral antibiotics avoid these complications and are less burdensome to families.² Recently, a series of multicenter studies showed no difference in outcomes between oral and postdischarge intravenous antibiotic therapy (PD-IV) for complicated appendicitis, osteomyelitis, and complicated pneumonia.^3-5

Despite a growing body of evidence suggesting that oral therapy ought to be the default treatment strategy rather than PD-IV, the extent to which practices have changed is unknown. In this study, we measured national trends in PD-IV use and variation by hospital for complicated appendicitis, osteomyelitis, and complicated pneumonia.

We performed a retrospective cohort study of children discharged from hospitals that contributed data to the Pediatric Health Information System (PHIS) database from January 2000 through December 2018. PHIS is an administrative database of children’s hospitals managed by the Children’s Hospital Association (Lenexa, Kansas) and contains deidentified patient-­level demographic data, discharge diagnosis and procedure codes, and detailed billing information, including medical supply charges.

The cohorts were defined using International Classification of Diseases, 9th and 10th Revisions (ICD-9 and ICD-10) discharge diagnosis and procedure codes. Patients admitted through September 2015 were identified using ICD-9 codes and patients admitted from October 2015 through December 2018 were identified using ICD-10 codes. The Centers for Medicaid & Medicare Services crosswalk was used to align ICD-9 and ICD-10 codes.⁶ Inclusion and exclusion criteria identifying cohorts of children hospitalized for complicated appendicitis, osteomyelitis, or complicated pneumonia were based on prior studies using the PHIS database.^3-5 These studies augmented the PHIS administrative dataset with local chart review to identify patients from 2009-2012 with the following inclusion and exclusion criteria: Patients with complicated appendicitis were defined by a diagnosis code for acute appendicitis and a procedure code for appendectomy, with postoperative length of stay lasting between 3 and 7 days. Patients with osteomyelitis had a diagnosis code of acute or unspecified osteomyelitis with a hospital length of stay between 2 and 14 days. Patients with complicated pneumonia were defined by a diagnosis code for both pneumonia and pleural effusion with one of these as the primary diagnosis. Patients were excluded if they were older than 18 years or if they were younger than 2 months for osteomyelitis and complicated pneumonia or younger than 3 years for appendicitis. For all three conditions, children with a complex chronic condition⁷ were excluded. Only the index encounter meeting inclusion and exclusion criteria for each patient was included. PD-IV therapy was defined using procedure codes and hospital charges during the index hospitalization. This definition for PD-IV therapy has been validated among children with complicated pneumonia, demonstrating positive and negative predictive values for PICC exposure of 85% and 99%, respectively.⁸

Trends in the percentage of patients receiving PD-IV were adjusted for age, race, insurance type, intensive care unit days, and hospital-level case mix index with use of Poisson regression. Calculated risk ratios represent the change in PD-IV across the entire 19-year study period for each condition (as opposed to an annual rate of change). An inflection point for each condition was identified using piecewise linear regression in which the line slope has one value up to a point in time and a second value after that point. The transition point is determined by maximizing model fit.

Some hospitals were added to the database throughout the time period and therefore did not have data for all years of the study. To account for the possibility of a group of high– or low–PD-IV use hospitals entering the cohort and biasing the overall trend, we performed a sensitivity analysis restricted to hospitals continuously contributing data to PHIS every year between 2004 (when a majority of hospitals joined PHIS) and 2018. Significance testing for individual hospital trends was conducted among continuously contributing hospitals, with each hospital tested in the above Poisson model independently.

For the most recent year of 2018, we reported the distribution of adjusted percentages of PD-IV at the individual hospital level. Only hospitals with at least five patients for a given condition are included in the percent PD-IV calculations for 2018. To examine the extent to which an individual hospital might be a low– or high–PD-IV user across conditions, we divided hospitals into quartiles based on PD-IV use for each condition in 2017-2018 and calculated the percent of hospitals in the lowest- and highest-use quartiles for all three conditions. All statistics were performed using Stata 15 (StataCorp).

Among 52 hospitals over a 19-year study period, there were 60,575 hospitalizations for complicated appendicitis, 24,753 hospitalizations for osteomyelitis, and 13,700 hospitalizations for complicated pneumonia. From 2000 to 2018, PD-IV decreased from 13% to 2% (RR, 0.15; 95% CI, 0.14-0.16) for complicated appendicitis, from 61% to 22% (RR, 0.41; 95% CI, 0.39-0.43) for osteomyelitis, and from 29% to 19% (RR, 0.63; 95% CI, 0.58-0.69) for complicated pneumonia (Figure 1). The inflection points occurred in 2009 for complicated appendicitis, 2009 for complicated pneumonia, and 2010 for osteomyelitis. The sensitivity analysis included 31 hospitals that contributed data to PHIS for every year between 2004-2018 and revealed similar findings for all three conditions: Complicated appendicitis had an RR of 0.15 (95% CI, 0.14-0.17), osteomyelitis had an RR of 0.34 (95% CI, 0.32-0.36), and complicated pneumonia had an RR of 0.55 (95% CI, 0.49-0.61). Most individual hospitals decreased PD-IV use (complicated appendicitis: 21 decreased, 8 no change, 2 increased; osteomyelitis: 25 decreased, 6 no change; complicated pneumonia: 14 decreased, 16 no change, 1 increased). While overall decreases in PD-IV were observed for all three conditions, considerable variation remained in 2018 for use of PD-IV (Figure 2), particularly for osteomyelitis (median, 18%; interquartile range [IQR] 9%-40%) and complicated pneumonia (median, 13%; IQR, 3%-30%). In 2017-2018, 1 out of 52 hospitals was in the lowest PD-IV–use quartile for all three conditions, and three hospitals were in the highest-use quartile for all three conditions.

Over a 19-year period, we observed a national decline in use of PD-IV for three serious and common bacterial infections. The decline in PD-IV is notable given that it has occurred largely in the absence of nationally coordinated guidelines or improvement efforts. Despite the overall declines, substantial variation in the use of PD-IV for these conditions persists across children’s hospitals.

The observed decrease in PD-IV use is a natural example of deimplementation, the abandonment of medical practices found to be harmful or ineffective.⁹ What is most compelling about the deimplementation of PD-IV for these infectious conditions is the seemingly organic motivation that propelled it. Studies of physician practice patterns for interventions that have undergone evidence reversals demonstrate that physicians might readily implement new interventions with an early evidence base but be less willing to deimplement them when more definitive evidence later questions their efficacy.¹⁰ Therefore, concerted improvement efforts backed by national guidelines are often needed to reduce the use of a widely accepted medical practice. For example, as evidence questioning the efficacy of steroid use in bronchiolitis mounted,¹¹ bronchiolitis guidelines recommended against steroid use¹² and a national quality improvement effort led to reductions in exposure to steroids among patients hospitalized with bronchiolitis.¹³ Complicated intra-abdominal infection guidelines acknowledge oral antibiotic therapy as an option,¹⁴ but no such national guidelines or improvement projects exist for osteomyelitis or complicated pneumonia PD-IV.

What is it about PD-IV for complicated appendicitis, osteomyelitis, and complicated pneumonia that fostered the observed organic deimplementation? Our findings that few hospitals were in the top or bottom quartile of PD-IV across all three conditions suggest that the impetus to decrease PD-IV was not likely the product of a broad hospital-wide practice shift. Most deimplementation frameworks suggest that successful deimplementation must be supported by high-quality evidence that the intervention is not only ineffective, but also harmful.¹⁵ In this case, the inflection point for osteomyelitis occurred in 2009, the same year that the first large multicenter study suggesting efficacy and decreased complications of early oral therapy for osteomyelitis was published.¹⁶ A direct link between a publication and inflection points for complicated pneumonia and appendicitis is less clear. It is possible that growth of the field of pediatric hospital medicine,¹⁷ with a stated emphasis on healthcare value,¹⁸ played a role. Greater understanding of the drivers and barriers to deimplementation in this and similar contexts will be important.

Our study has some important limitations. While inclusion and exclusion criteria were consistent over the study period, practice patterns (ie, length of stay in uncomplicated patients) change and could alter the case-mix of patients over time. Additionally, the PHIS database largely comprises children’s hospitals, and the trends we observed in PD-IV may not generalize to community settings.

The degree of deimplementation of PD-IV observed across children’s hospitals is impressive, but opportunity for further improvement likely remains. We found that marked hospital-­level variation in use of PD-IV still exists, with some hospitals almost never using PD-IV and others using it for most patients. While the ideal amount of PD-IV is probably not zero, a portion of the observed variation likely represents overuse of PD-IV. To reduce costs and complications associated with antibiotic therapy, national guidelines and a targeted national improvement collaborative may be necessary to achieve further reductions in PD-IV.

In recent years, mounting evidence has emerged questioning the practice of using prolonged intravenous antibiotic therapy to treat certain serious bacterial infections in children, including complicated appendicitis, osteomyelitis, and complicated pneumonia. Historically, treatment of these conditions was often completed intravenously after hospital discharge using peripherally inserted central catheters (PICCs). Line infections, clots, mechanical problems, and general discomfort complicate PICCs, which led to their removal in more than 20% of children in one study.¹ Oral antibiotics avoid these complications and are less burdensome to families.² Recently, a series of multicenter studies showed no difference in outcomes between oral and postdischarge intravenous antibiotic therapy (PD-IV) for complicated appendicitis, osteomyelitis, and complicated pneumonia.^3-5

Despite a growing body of evidence suggesting that oral therapy ought to be the default treatment strategy rather than PD-IV, the extent to which practices have changed is unknown. In this study, we measured national trends in PD-IV use and variation by hospital for complicated appendicitis, osteomyelitis, and complicated pneumonia.

We performed a retrospective cohort study of children discharged from hospitals that contributed data to the Pediatric Health Information System (PHIS) database from January 2000 through December 2018. PHIS is an administrative database of children’s hospitals managed by the Children’s Hospital Association (Lenexa, Kansas) and contains deidentified patient-­level demographic data, discharge diagnosis and procedure codes, and detailed billing information, including medical supply charges.

The cohorts were defined using International Classification of Diseases, 9th and 10th Revisions (ICD-9 and ICD-10) discharge diagnosis and procedure codes. Patients admitted through September 2015 were identified using ICD-9 codes and patients admitted from October 2015 through December 2018 were identified using ICD-10 codes. The Centers for Medicaid & Medicare Services crosswalk was used to align ICD-9 and ICD-10 codes.⁶ Inclusion and exclusion criteria identifying cohorts of children hospitalized for complicated appendicitis, osteomyelitis, or complicated pneumonia were based on prior studies using the PHIS database.^3-5 These studies augmented the PHIS administrative dataset with local chart review to identify patients from 2009-2012 with the following inclusion and exclusion criteria: Patients with complicated appendicitis were defined by a diagnosis code for acute appendicitis and a procedure code for appendectomy, with postoperative length of stay lasting between 3 and 7 days. Patients with osteomyelitis had a diagnosis code of acute or unspecified osteomyelitis with a hospital length of stay between 2 and 14 days. Patients with complicated pneumonia were defined by a diagnosis code for both pneumonia and pleural effusion with one of these as the primary diagnosis. Patients were excluded if they were older than 18 years or if they were younger than 2 months for osteomyelitis and complicated pneumonia or younger than 3 years for appendicitis. For all three conditions, children with a complex chronic condition⁷ were excluded. Only the index encounter meeting inclusion and exclusion criteria for each patient was included. PD-IV therapy was defined using procedure codes and hospital charges during the index hospitalization. This definition for PD-IV therapy has been validated among children with complicated pneumonia, demonstrating positive and negative predictive values for PICC exposure of 85% and 99%, respectively.⁸

Trends in the percentage of patients receiving PD-IV were adjusted for age, race, insurance type, intensive care unit days, and hospital-level case mix index with use of Poisson regression. Calculated risk ratios represent the change in PD-IV across the entire 19-year study period for each condition (as opposed to an annual rate of change). An inflection point for each condition was identified using piecewise linear regression in which the line slope has one value up to a point in time and a second value after that point. The transition point is determined by maximizing model fit.

Some hospitals were added to the database throughout the time period and therefore did not have data for all years of the study. To account for the possibility of a group of high– or low–PD-IV use hospitals entering the cohort and biasing the overall trend, we performed a sensitivity analysis restricted to hospitals continuously contributing data to PHIS every year between 2004 (when a majority of hospitals joined PHIS) and 2018. Significance testing for individual hospital trends was conducted among continuously contributing hospitals, with each hospital tested in the above Poisson model independently.

For the most recent year of 2018, we reported the distribution of adjusted percentages of PD-IV at the individual hospital level. Only hospitals with at least five patients for a given condition are included in the percent PD-IV calculations for 2018. To examine the extent to which an individual hospital might be a low– or high–PD-IV user across conditions, we divided hospitals into quartiles based on PD-IV use for each condition in 2017-2018 and calculated the percent of hospitals in the lowest- and highest-use quartiles for all three conditions. All statistics were performed using Stata 15 (StataCorp).

Among 52 hospitals over a 19-year study period, there were 60,575 hospitalizations for complicated appendicitis, 24,753 hospitalizations for osteomyelitis, and 13,700 hospitalizations for complicated pneumonia. From 2000 to 2018, PD-IV decreased from 13% to 2% (RR, 0.15; 95% CI, 0.14-0.16) for complicated appendicitis, from 61% to 22% (RR, 0.41; 95% CI, 0.39-0.43) for osteomyelitis, and from 29% to 19% (RR, 0.63; 95% CI, 0.58-0.69) for complicated pneumonia (Figure 1). The inflection points occurred in 2009 for complicated appendicitis, 2009 for complicated pneumonia, and 2010 for osteomyelitis. The sensitivity analysis included 31 hospitals that contributed data to PHIS for every year between 2004-2018 and revealed similar findings for all three conditions: Complicated appendicitis had an RR of 0.15 (95% CI, 0.14-0.17), osteomyelitis had an RR of 0.34 (95% CI, 0.32-0.36), and complicated pneumonia had an RR of 0.55 (95% CI, 0.49-0.61). Most individual hospitals decreased PD-IV use (complicated appendicitis: 21 decreased, 8 no change, 2 increased; osteomyelitis: 25 decreased, 6 no change; complicated pneumonia: 14 decreased, 16 no change, 1 increased). While overall decreases in PD-IV were observed for all three conditions, considerable variation remained in 2018 for use of PD-IV (Figure 2), particularly for osteomyelitis (median, 18%; interquartile range [IQR] 9%-40%) and complicated pneumonia (median, 13%; IQR, 3%-30%). In 2017-2018, 1 out of 52 hospitals was in the lowest PD-IV–use quartile for all three conditions, and three hospitals were in the highest-use quartile for all three conditions.

Over a 19-year period, we observed a national decline in use of PD-IV for three serious and common bacterial infections. The decline in PD-IV is notable given that it has occurred largely in the absence of nationally coordinated guidelines or improvement efforts. Despite the overall declines, substantial variation in the use of PD-IV for these conditions persists across children’s hospitals.

The observed decrease in PD-IV use is a natural example of deimplementation, the abandonment of medical practices found to be harmful or ineffective.⁹ What is most compelling about the deimplementation of PD-IV for these infectious conditions is the seemingly organic motivation that propelled it. Studies of physician practice patterns for interventions that have undergone evidence reversals demonstrate that physicians might readily implement new interventions with an early evidence base but be less willing to deimplement them when more definitive evidence later questions their efficacy.¹⁰ Therefore, concerted improvement efforts backed by national guidelines are often needed to reduce the use of a widely accepted medical practice. For example, as evidence questioning the efficacy of steroid use in bronchiolitis mounted,¹¹ bronchiolitis guidelines recommended against steroid use¹² and a national quality improvement effort led to reductions in exposure to steroids among patients hospitalized with bronchiolitis.¹³ Complicated intra-abdominal infection guidelines acknowledge oral antibiotic therapy as an option,¹⁴ but no such national guidelines or improvement projects exist for osteomyelitis or complicated pneumonia PD-IV.

What is it about PD-IV for complicated appendicitis, osteomyelitis, and complicated pneumonia that fostered the observed organic deimplementation? Our findings that few hospitals were in the top or bottom quartile of PD-IV across all three conditions suggest that the impetus to decrease PD-IV was not likely the product of a broad hospital-wide practice shift. Most deimplementation frameworks suggest that successful deimplementation must be supported by high-quality evidence that the intervention is not only ineffective, but also harmful.¹⁵ In this case, the inflection point for osteomyelitis occurred in 2009, the same year that the first large multicenter study suggesting efficacy and decreased complications of early oral therapy for osteomyelitis was published.¹⁶ A direct link between a publication and inflection points for complicated pneumonia and appendicitis is less clear. It is possible that growth of the field of pediatric hospital medicine,¹⁷ with a stated emphasis on healthcare value,¹⁸ played a role. Greater understanding of the drivers and barriers to deimplementation in this and similar contexts will be important.

Our study has some important limitations. While inclusion and exclusion criteria were consistent over the study period, practice patterns (ie, length of stay in uncomplicated patients) change and could alter the case-mix of patients over time. Additionally, the PHIS database largely comprises children’s hospitals, and the trends we observed in PD-IV may not generalize to community settings.

The degree of deimplementation of PD-IV observed across children’s hospitals is impressive, but opportunity for further improvement likely remains. We found that marked hospital-­level variation in use of PD-IV still exists, with some hospitals almost never using PD-IV and others using it for most patients. While the ideal amount of PD-IV is probably not zero, a portion of the observed variation likely represents overuse of PD-IV. To reduce costs and complications associated with antibiotic therapy, national guidelines and a targeted national improvement collaborative may be necessary to achieve further reductions in PD-IV.

In recent years, mounting evidence has emerged questioning the practice of using prolonged intravenous antibiotic therapy to treat certain serious bacterial infections in children, including complicated appendicitis, osteomyelitis, and complicated pneumonia. Historically, treatment of these conditions was often completed intravenously after hospital discharge using peripherally inserted central catheters (PICCs). Line infections, clots, mechanical problems, and general discomfort complicate PICCs, which led to their removal in more than 20% of children in one study.¹ Oral antibiotics avoid these complications and are less burdensome to families.² Recently, a series of multicenter studies showed no difference in outcomes between oral and postdischarge intravenous antibiotic therapy (PD-IV) for complicated appendicitis, osteomyelitis, and complicated pneumonia.^3-5

Despite a growing body of evidence suggesting that oral therapy ought to be the default treatment strategy rather than PD-IV, the extent to which practices have changed is unknown. In this study, we measured national trends in PD-IV use and variation by hospital for complicated appendicitis, osteomyelitis, and complicated pneumonia.

We performed a retrospective cohort study of children discharged from hospitals that contributed data to the Pediatric Health Information System (PHIS) database from January 2000 through December 2018. PHIS is an administrative database of children’s hospitals managed by the Children’s Hospital Association (Lenexa, Kansas) and contains deidentified patient-­level demographic data, discharge diagnosis and procedure codes, and detailed billing information, including medical supply charges.

The cohorts were defined using International Classification of Diseases, 9th and 10th Revisions (ICD-9 and ICD-10) discharge diagnosis and procedure codes. Patients admitted through September 2015 were identified using ICD-9 codes and patients admitted from October 2015 through December 2018 were identified using ICD-10 codes. The Centers for Medicaid & Medicare Services crosswalk was used to align ICD-9 and ICD-10 codes.⁶ Inclusion and exclusion criteria identifying cohorts of children hospitalized for complicated appendicitis, osteomyelitis, or complicated pneumonia were based on prior studies using the PHIS database.^3-5 These studies augmented the PHIS administrative dataset with local chart review to identify patients from 2009-2012 with the following inclusion and exclusion criteria: Patients with complicated appendicitis were defined by a diagnosis code for acute appendicitis and a procedure code for appendectomy, with postoperative length of stay lasting between 3 and 7 days. Patients with osteomyelitis had a diagnosis code of acute or unspecified osteomyelitis with a hospital length of stay between 2 and 14 days. Patients with complicated pneumonia were defined by a diagnosis code for both pneumonia and pleural effusion with one of these as the primary diagnosis. Patients were excluded if they were older than 18 years or if they were younger than 2 months for osteomyelitis and complicated pneumonia or younger than 3 years for appendicitis. For all three conditions, children with a complex chronic condition⁷ were excluded. Only the index encounter meeting inclusion and exclusion criteria for each patient was included. PD-IV therapy was defined using procedure codes and hospital charges during the index hospitalization. This definition for PD-IV therapy has been validated among children with complicated pneumonia, demonstrating positive and negative predictive values for PICC exposure of 85% and 99%, respectively.⁸

Trends in the percentage of patients receiving PD-IV were adjusted for age, race, insurance type, intensive care unit days, and hospital-level case mix index with use of Poisson regression. Calculated risk ratios represent the change in PD-IV across the entire 19-year study period for each condition (as opposed to an annual rate of change). An inflection point for each condition was identified using piecewise linear regression in which the line slope has one value up to a point in time and a second value after that point. The transition point is determined by maximizing model fit.

Some hospitals were added to the database throughout the time period and therefore did not have data for all years of the study. To account for the possibility of a group of high– or low–PD-IV use hospitals entering the cohort and biasing the overall trend, we performed a sensitivity analysis restricted to hospitals continuously contributing data to PHIS every year between 2004 (when a majority of hospitals joined PHIS) and 2018. Significance testing for individual hospital trends was conducted among continuously contributing hospitals, with each hospital tested in the above Poisson model independently.

For the most recent year of 2018, we reported the distribution of adjusted percentages of PD-IV at the individual hospital level. Only hospitals with at least five patients for a given condition are included in the percent PD-IV calculations for 2018. To examine the extent to which an individual hospital might be a low– or high–PD-IV user across conditions, we divided hospitals into quartiles based on PD-IV use for each condition in 2017-2018 and calculated the percent of hospitals in the lowest- and highest-use quartiles for all three conditions. All statistics were performed using Stata 15 (StataCorp).

Among 52 hospitals over a 19-year study period, there were 60,575 hospitalizations for complicated appendicitis, 24,753 hospitalizations for osteomyelitis, and 13,700 hospitalizations for complicated pneumonia. From 2000 to 2018, PD-IV decreased from 13% to 2% (RR, 0.15; 95% CI, 0.14-0.16) for complicated appendicitis, from 61% to 22% (RR, 0.41; 95% CI, 0.39-0.43) for osteomyelitis, and from 29% to 19% (RR, 0.63; 95% CI, 0.58-0.69) for complicated pneumonia (Figure 1). The inflection points occurred in 2009 for complicated appendicitis, 2009 for complicated pneumonia, and 2010 for osteomyelitis. The sensitivity analysis included 31 hospitals that contributed data to PHIS for every year between 2004-2018 and revealed similar findings for all three conditions: Complicated appendicitis had an RR of 0.15 (95% CI, 0.14-0.17), osteomyelitis had an RR of 0.34 (95% CI, 0.32-0.36), and complicated pneumonia had an RR of 0.55 (95% CI, 0.49-0.61). Most individual hospitals decreased PD-IV use (complicated appendicitis: 21 decreased, 8 no change, 2 increased; osteomyelitis: 25 decreased, 6 no change; complicated pneumonia: 14 decreased, 16 no change, 1 increased). While overall decreases in PD-IV were observed for all three conditions, considerable variation remained in 2018 for use of PD-IV (Figure 2), particularly for osteomyelitis (median, 18%; interquartile range [IQR] 9%-40%) and complicated pneumonia (median, 13%; IQR, 3%-30%). In 2017-2018, 1 out of 52 hospitals was in the lowest PD-IV–use quartile for all three conditions, and three hospitals were in the highest-use quartile for all three conditions.

Over a 19-year period, we observed a national decline in use of PD-IV for three serious and common bacterial infections. The decline in PD-IV is notable given that it has occurred largely in the absence of nationally coordinated guidelines or improvement efforts. Despite the overall declines, substantial variation in the use of PD-IV for these conditions persists across children’s hospitals.

The observed decrease in PD-IV use is a natural example of deimplementation, the abandonment of medical practices found to be harmful or ineffective.⁹ What is most compelling about the deimplementation of PD-IV for these infectious conditions is the seemingly organic motivation that propelled it. Studies of physician practice patterns for interventions that have undergone evidence reversals demonstrate that physicians might readily implement new interventions with an early evidence base but be less willing to deimplement them when more definitive evidence later questions their efficacy.¹⁰ Therefore, concerted improvement efforts backed by national guidelines are often needed to reduce the use of a widely accepted medical practice. For example, as evidence questioning the efficacy of steroid use in bronchiolitis mounted,¹¹ bronchiolitis guidelines recommended against steroid use¹² and a national quality improvement effort led to reductions in exposure to steroids among patients hospitalized with bronchiolitis.¹³ Complicated intra-abdominal infection guidelines acknowledge oral antibiotic therapy as an option,¹⁴ but no such national guidelines or improvement projects exist for osteomyelitis or complicated pneumonia PD-IV.

What is it about PD-IV for complicated appendicitis, osteomyelitis, and complicated pneumonia that fostered the observed organic deimplementation? Our findings that few hospitals were in the top or bottom quartile of PD-IV across all three conditions suggest that the impetus to decrease PD-IV was not likely the product of a broad hospital-wide practice shift. Most deimplementation frameworks suggest that successful deimplementation must be supported by high-quality evidence that the intervention is not only ineffective, but also harmful.¹⁵ In this case, the inflection point for osteomyelitis occurred in 2009, the same year that the first large multicenter study suggesting efficacy and decreased complications of early oral therapy for osteomyelitis was published.¹⁶ A direct link between a publication and inflection points for complicated pneumonia and appendicitis is less clear. It is possible that growth of the field of pediatric hospital medicine,¹⁷ with a stated emphasis on healthcare value,¹⁸ played a role. Greater understanding of the drivers and barriers to deimplementation in this and similar contexts will be important.

Our study has some important limitations. While inclusion and exclusion criteria were consistent over the study period, practice patterns (ie, length of stay in uncomplicated patients) change and could alter the case-mix of patients over time. Additionally, the PHIS database largely comprises children’s hospitals, and the trends we observed in PD-IV may not generalize to community settings.

The degree of deimplementation of PD-IV observed across children’s hospitals is impressive, but opportunity for further improvement likely remains. We found that marked hospital-­level variation in use of PD-IV still exists, with some hospitals almost never using PD-IV and others using it for most patients. While the ideal amount of PD-IV is probably not zero, a portion of the observed variation likely represents overuse of PD-IV. To reduce costs and complications associated with antibiotic therapy, national guidelines and a targeted national improvement collaborative may be necessary to achieve further reductions in PD-IV.

Children hospitalized for bronchiolitis frequently require admission to the intensive care unit (ICU), with estimates as high as 18%^1,2 and 35%³ in two prospective, multicenter studies. The indication for ICU admission is nearly always a need for advanced respiratory support, which historically consisted of continuous or bilevel positive airway pressure (CPAP and BiPAP, respectively) or mechanical ventilation. High-flow nasal cannula (HFNC) is a recent addition to the respiratory support armamentarium, delivering heated and humidified oxygen at rates of up to 60 L/min and allowing for clinicians to titrate both flow rate and fraction of inspired oxygen (FiO₂).⁴

Several studies have demonstrated that HFNC is capable of decreasing a child’s work of breathing,^5-8 and it has the potential advantage of being better tolerated than other forms of advanced respiratory support.^9,10 These case-series physiologic studies informed early ward-based HFNC protocols for bronchiolitis, which were adopted to decrease ICU utilization. Since then, single center observational studies examining the association between ward-based HFNC protocols and subsequent ICU utilization have come to discordant conclusions.^11-14 Studying the effect of employing HFNC outside of the ICU is challenging in the context of a randomized, controlled trial (RCT) because it is difficult to blind healthcare providers to the intervention and because crossover from the control group to HFNC is frequent. Two unblinded RCTs published in 2017 and 2018 found that children randomized to conventional nasal cannula were frequently escalated to HFNC (flow rates of 1-2 L/kg per minute), but neither trial found a difference in ICU admission.^15,16 Sample sizes substantially larger than those present in currently published or registered RCTs would be required to evaluate the impact of ward-based HFNC protocols on the outcome that inspired the protocols in the first place, namely ICU utilization.¹⁷

Children’s hospitals have adopted ward-based HFNC protocols at different time points over the last decade, which allows for a natural experiment—a promising alternative study design that avoids the challenges of blinding, crossover, and modest sample sizes. In order to have sufficient postadoption data for analyses, the present study is limited to ward-based HFNC protocols adopted prior to 2016, which we have termed “early” ward-based HFNC protocols. Among children with bronchiolitis, our objective was to measure the association between hospital-level adoption of a ward-based HFNC protocol and subsequent ICU utilization, using a multicenter network of children’s hospitals.

We conducted a multicenter retrospective cohort study using the Pediatric Health Information System (PHIS) database. The PHIS database is operated by the Children’s Hospital Association (Lenexa, Kansas) and provides deidentified patient-level information for children who receive hospital care at 55 US children’s hospitals. Available data elements include patient demographic data, discharge diagnosis and procedure codes, and detailed billing information, such as laboratory, imaging, pharmacy, and supply charges. At the patient level, the use of HFNC vs standard oxygen therapy circuits cannot be discriminated.

Exposure

The study exposure was a hospital’s first ward-based HFNC protocol, with adoption measured at the hospital level at each PHIS site via direct communication with leaders in hospital medicine. In most cases, first contact was made with the pediatric hospital medicine division chief or fellowship program director, who then, if necessary, connected study investigators to local HFNC champions aware of site-specific historical HFNC protocol details. Contact with a hospital was made only if the hospital had contributed at least 6 consecutive years of data to PHIS. Hospitals were classified as “adopting” hospitals if their HFNC protocol met all of the following criteria: (a) allows initiation of HFNC outside of the ICU (on the floor or in the ED), (b) allows continued care outside of the ICU (on the floor), (c) not limited to a small unit like an intermediate care unit, and (d) adopted during a specific, known respiratory season. Hospitals for which ward-based HFNC protocols were adopted but did not meet these criteria were excluded from further analysis. Our intent was to identify large scale, programmatic protocol launches and exclude hospitals with exceptions that might preclude a sizable portion of our cohort from being eligible for the protocol. Hospitals for which inpatient use of HFNC remains limited to the ICU were defined as “nonadopting” hospitals. Respondents at adopting hospitals were asked to share details about their protocol, including patient eligibility criteria and maximum HFNC rates of flow permitted outside of the ICU.

Patient Characteristics

Patients aged 3 to 24 months who were hospitalized at adopting and nonadopting hospitals were included if an International Classification of Diseases, Ninth Revision (ICD-9) discharge diagnosis code for bronchiolitis (466.XX) was present in any position (not limited to a primary diagnosis). The lower age limit of 3 months was chosen to match the most restrictive age eligibility criteria of provided HFNC protocols (Appendix 1). A crosswalk available from the Centers for Medicare & Medicaid Services¹⁸ was used to convert ICD-10 diagnosis and procedure codes from recent years to ICD-9 diagnosis and procedure codes. Patients were excluded if their encounter contained a diagnosis or procedure code signifying a complex chronic condition,¹⁹ if their hospitalization involved care in the neonatal ICU, or if their admission date occurred outside of the respiratory season. Respiratory season was defined as November 1 through April 30.

Outcomes

Outcomes were measured during three respiratory seasons leading up to adoption and during three respiratory seasons after adoption. The primary outcome was ICU utilization, including the proportion of patients admitted to the ICU and ICU length of stay, expressed as ICU days per 100 patients. Secondary outcomes included mean total length of stay and the proportion of patients who received mechanical ventilation. Lengths of stay were measured in days, the most granular unit of time provided in PHIS over the entire study period. As such, partial days of care are rounded up to 1 full day. A previously published strict definition for mechanical ventilation that limits false positives was used, requiring that patients have a procedure or supply code for mechanical ventilation and a pharmacy charge for a neuromuscular blocking agent.²⁰

Primary Analysis

The primary analysis was restricted to adopting hospitals. An interrupted time series approach was used to measure two possible types of change associated with HFNC protocol adoption: an immediate intervention effect and a change in the slope of an outcome.²¹ The immediate intervention effect represents the change in the level of the outcome that occurs in the period immediately following the introduction of the protocol. The change in slope is the extent to which the outcome changes on a per season basis, attributable to the protocol. Interrupted time series estimates were adjusted for patient age, gender, race, ethnicity, and insurance type; linear regression was used for continuous outcomes and logistic regression for dichotomous outcomes. An ordinary least squares time series model was used to adjust for autocorrelation and Newey-West standard errors were employed.²² Analyses were performed using STATA version 14 (Stata-Corp, College Station, Texas).

Supplementary Analyses

Two preplanned supplementary analyses were conducted. Supplementary analysis 1 was identical to the primary analysis, with the exception that the first season after adoption was censored. The rationale for censoring the first adoption season was to account for a potential learning effect and/or delayed start to full protocol implementation. Supplementary analysis 2 used the nonadopting hospitals as a control group and subtracted the effects measured from an interrupted time series analysis among nonadopting hospitals from the effects measured among adopting hospitals. The rationale for this approach was to control for unmeasured secular (eg, availability of ICU beds) and temporal (eg, severity of a given bronchiolitis season) factors that may have coincidentally occurred with HFNC adoption seasons. The only modification to the interrupted time series approach for supplementary analysis 2 was to provide the nonadopting hospitals with an artificial interruption point because nonadopting hospitals, by definition, did not have an adoption season that could be used in an interrupted time series approach. The interruption point for nonadopting hospitals was set at the median adoption season for adopting hospitals.

Exposure

Leaders at 44 hospitals were contacted regarding their hospital’s use of HFNC outside of the ICU (Figure 1). Responses were obtained for 41 hospitals (93% response rate), 18 of which were classified as nonadopting hospitals. Of the 23 hospitals where the presence of ward-based HFNC protocols were reported, 12 met inclusion criteria and were classified as adopting hospitals. HFNC protocols were adopted at these hospitals in a staggered fashion between the 2010-2011 and 2015-2016 respiratory seasons (Figure 2). The median adoption season was the 2013-14 respiratory season.

Nine adopting hospitals were able to provide details about their first HFNC protocols (Appendix 1). No two protocols were identical, but they shared many similarities. Minimum age requirements ranged from birth to a few months of age. Exclusion criteria were particularly variable, with a history of chronic lung disease or apnea being the most common criteria. Maximum allowed rates of flow ranged from 4 to 10 liters per minute. Criteria for transfer to the ICU were consistently based on an elevated FiO₂ and duration of HFNC exposure.

Patient Characteristics

A total of 32,809 bronchiolitis encounters occurred at adopting hospitals during qualifying respiratory seasons, of which 6,556 (20%) involved patients with a complex chronic condition and were excluded. Of the 26,253 included bronchiolitis encounters, 12,495 encounters occurred prior to ward-based HFNC protocol adoption and 13,758 encounters occurred after adoption. The median age of patients was 8 months (interquartile range, 5-14 months). Most patients were on government insurance (64%), male (58%), of white (56%) or black (18%) race, and of non-Hispanic ethnicity (72%). Pre- and postadoption patient demographics were similar (Appendix 2).

Primary Analysis

Shifts in the level of ICU use and ICU length of stay were observed at the time of adoption of a ward-based HFNC protocol (Figure 3). Specifically, ward-based HFNC protocol adoption was associated with an immediate 3.1% absolute increase (95% CI, 2.8%-3.4%) in the proportion of patients admitted to the ICU and a 9.1 days per 100 patients increase (95% CI, 5.1-13.2) in ICU length of stay (Table). The slope of ICU admissions per season was increasing after HFNC protocol adoption (1.0% increase per season; 95% CI, 0.8%-1.1%). When examined at the individual-hospital level (Appendix 3), seven hospitals were found to have significant increases in ICU admissions (immediate intervention effect or change in slope) after adoption, and one hospital was found to have a significant decrease in ICU admissions (change in slope only). Neither immediate intervention effects nor changes in the slopes of total length of stay and mechanical ventilation were observed, with mean total length of stay approximately 3 days and just over 1% of patients receiving mechanical ventilation (Figure 3).

Supplementary Analyses

Supplementary analyses were largely consistent with the primary analysis. Associations with increased ICU utilization were again observed, although the immediate change in ICU length of stay for supplementary analysis 1 was not significant and the slope for ICU length of stay in supplementary analysis 2 was down trending (Table). Changes in total length of stay and mechanical ventilation were not observed in either supplementary analysis, with the lone exception being an increase in the proportion of patients receiving mechanical ventilation per season (increase in slope) in supplementary analysis 1.

This is the largest multicenter study to date evaluating ICU utilization after adoption of a ward-based HFNC protocol for patients with bronchiolitis. While a principal goal of allowing HFNC use outside of the ICU is to reduce the time that patients with bronchiolitis spend in the ICU, we found that early protocols were, paradoxically, associated with increased ICU utilization. Ward-based HFNC protocols were not associated with changes in hospital length of stay or need for mechanical ventilation. Our findings are particularly relevant given that the majority of children’s hospitals in our sample have adopted ward-based HFNC protocols to care for patients with bronchiolitis.

The increase in ICU utilization measured in our study is a novel finding, seemingly in contradiction to existing literature. Early pilot studies inspired hope that employing HFNC on the general ward might prevent a portion of children from needing ICU care.^11,12 Subsequent larger observational studies did not demonstrate decreases in ICU utilization after adoption of ward-based HFNC protocols.^13,14 The two RCTs comparing low-flow and high-flow nasal cannula use outside of the ICU did not measure a statistically significant effect on ICU utilization, an exploratory outcome in both trials.^15,16 However, the reported point estimates for absolute differences in ICU admission were 2% to 3% higher among patients randomized to HFNC, which is consistent with the 2% to 4% increase in ICU admission measured in the present study.

What might explain this surprising finding? While our observational study cannot speak to mechanism, the protocol details examined in the present study suggest that initial adoption of a ward-based HFNC protocol is often coupled with specific ICU transfer criteria that were unlikely in place prior to protocol initiation. For example, most protocols recommended consideration of ICU transfer for elevated FiO₂ or prolonged duration of HFNC. Transfer to the ICU for prolonged HFNC duration is only possible in the setting of a ward-based HFNC protocol and transfer for elevated FiO₂ was probably unnecessary prior to protocol adoption given that low-flow nasal cannula generally delivers 100% FiO₂. It is also possible that with HFNC comes a perception of increased acuity. For example, medical providers may see patients on HFNC as sicker than patients with the same amount of work of breathing but off HFNC, which makes providers more likely to seek ICU admission for patients on HFNC. The combination of unchanged total length of stay and increased ICU utilization suggests that early ward-based HFNC protocols were an ineffective instrument to improve hospital bed availability during the peak census times that often occur in bronchiolitis season.

The large sample size afforded our study by its multicenter, retrospective design also allowed for a meaningful assessment of the association between a ward-based HFNC protocol and the need for mechanical ventilation. Early indications suggested a lack of substantial association between HFNC use outside of the ICU and rates of mechanical ventilation, but prior studies were limited by small numbers of patients receiving mechanical ventilation (<30 patients in each study).^13,14,16 The present study, in which 783 patients received mechanical ventilation, supports the lack of association between early ward-based HFNC protocols and the need for mechanical ventilation. It should be noted that other studies have measured decreases in mechanical ventilation in association with ICU-based HFNC use.^23-26 In addition to examining HFNC use in a different clinical context, decreases in mechanical ventilation measured after HFNC implementation in the ICU could be explained by preexisting practice trends to limit invasive ventilation and/or selection bias resulting from an increase in less severely ill patients being admitted to the ICU over time. The interrupted time series approach and the staggered adoption of HFNC protocols make the present study less susceptible to biases from preexisting trends and the inclusion of patients cared for both on the ward and within the ICU reduces selection bias.

Our study has several important limitations. First, all hospitals included in the analysis were US children’s hospitals and these findings may not generalize to other practice environments, including community hospitals and other countries. Second, our cohort and outcomes were defined using administrative billing data, which have been incompletely validated, making some degree of misclassification likely. Third, we measured HFNC exposure at the hospital level, but could not examine the extent to which individual patients were exposed to HFNC because such data are not present in PHIS. Even if we had access to patient-level HFNC exposure data, we would have still compared outcomes among all patients with bronchiolitis (not just those who received HFNC), to avoid selection bias. However, knowing HFNC exposure status at the patient level would have allowed for weighting of the effects measured at each hospital according to the extent of HFNC exposure. Fourth, there are likely other, unmeasured secular and temporal factors that could affect study outcomes. To some degree, the interrupted time series approach, observed staggered adoption of protocols, and nonadopting hospital supplementary analysis mitigate this risk of bias. Fifth, while the pre- and postadoption populations appeared demographically similar, it is possible that the populations might have differed by other unmeasured factors. Finally, early ward-based HFNC protocols have likely undergone iterative changes since adoption. We compared pre- and postadoption outcome slopes and censored the first adoption season in a supplementary analysis to attempt to account for this potential limitation.

In conclusion, our findings suggest that initial implementation of ward-based HFNC protocols were not successful at reducing ICU utilization for children with bronchiolitis. Future research should examine whether more evolved HFNC protocols that use higher flow rates, more generous ICU transfer criteria, and more rapid weaning criteria can reduce ICU utilization.

We thank Dr Vineeta Mittal (University of Texas Southwestern Medical Center) for providing feedback regarding the manuscript.

Children hospitalized for bronchiolitis frequently require admission to the intensive care unit (ICU), with estimates as high as 18%^1,2 and 35%³ in two prospective, multicenter studies. The indication for ICU admission is nearly always a need for advanced respiratory support, which historically consisted of continuous or bilevel positive airway pressure (CPAP and BiPAP, respectively) or mechanical ventilation. High-flow nasal cannula (HFNC) is a recent addition to the respiratory support armamentarium, delivering heated and humidified oxygen at rates of up to 60 L/min and allowing for clinicians to titrate both flow rate and fraction of inspired oxygen (FiO₂).⁴

Several studies have demonstrated that HFNC is capable of decreasing a child’s work of breathing,^5-8 and it has the potential advantage of being better tolerated than other forms of advanced respiratory support.^9,10 These case-series physiologic studies informed early ward-based HFNC protocols for bronchiolitis, which were adopted to decrease ICU utilization. Since then, single center observational studies examining the association between ward-based HFNC protocols and subsequent ICU utilization have come to discordant conclusions.^11-14 Studying the effect of employing HFNC outside of the ICU is challenging in the context of a randomized, controlled trial (RCT) because it is difficult to blind healthcare providers to the intervention and because crossover from the control group to HFNC is frequent. Two unblinded RCTs published in 2017 and 2018 found that children randomized to conventional nasal cannula were frequently escalated to HFNC (flow rates of 1-2 L/kg per minute), but neither trial found a difference in ICU admission.^15,16 Sample sizes substantially larger than those present in currently published or registered RCTs would be required to evaluate the impact of ward-based HFNC protocols on the outcome that inspired the protocols in the first place, namely ICU utilization.¹⁷

Children’s hospitals have adopted ward-based HFNC protocols at different time points over the last decade, which allows for a natural experiment—a promising alternative study design that avoids the challenges of blinding, crossover, and modest sample sizes. In order to have sufficient postadoption data for analyses, the present study is limited to ward-based HFNC protocols adopted prior to 2016, which we have termed “early” ward-based HFNC protocols. Among children with bronchiolitis, our objective was to measure the association between hospital-level adoption of a ward-based HFNC protocol and subsequent ICU utilization, using a multicenter network of children’s hospitals.

We conducted a multicenter retrospective cohort study using the Pediatric Health Information System (PHIS) database. The PHIS database is operated by the Children’s Hospital Association (Lenexa, Kansas) and provides deidentified patient-level information for children who receive hospital care at 55 US children’s hospitals. Available data elements include patient demographic data, discharge diagnosis and procedure codes, and detailed billing information, such as laboratory, imaging, pharmacy, and supply charges. At the patient level, the use of HFNC vs standard oxygen therapy circuits cannot be discriminated.

Exposure

The study exposure was a hospital’s first ward-based HFNC protocol, with adoption measured at the hospital level at each PHIS site via direct communication with leaders in hospital medicine. In most cases, first contact was made with the pediatric hospital medicine division chief or fellowship program director, who then, if necessary, connected study investigators to local HFNC champions aware of site-specific historical HFNC protocol details. Contact with a hospital was made only if the hospital had contributed at least 6 consecutive years of data to PHIS. Hospitals were classified as “adopting” hospitals if their HFNC protocol met all of the following criteria: (a) allows initiation of HFNC outside of the ICU (on the floor or in the ED), (b) allows continued care outside of the ICU (on the floor), (c) not limited to a small unit like an intermediate care unit, and (d) adopted during a specific, known respiratory season. Hospitals for which ward-based HFNC protocols were adopted but did not meet these criteria were excluded from further analysis. Our intent was to identify large scale, programmatic protocol launches and exclude hospitals with exceptions that might preclude a sizable portion of our cohort from being eligible for the protocol. Hospitals for which inpatient use of HFNC remains limited to the ICU were defined as “nonadopting” hospitals. Respondents at adopting hospitals were asked to share details about their protocol, including patient eligibility criteria and maximum HFNC rates of flow permitted outside of the ICU.

Patient Characteristics

Patients aged 3 to 24 months who were hospitalized at adopting and nonadopting hospitals were included if an International Classification of Diseases, Ninth Revision (ICD-9) discharge diagnosis code for bronchiolitis (466.XX) was present in any position (not limited to a primary diagnosis). The lower age limit of 3 months was chosen to match the most restrictive age eligibility criteria of provided HFNC protocols (Appendix 1). A crosswalk available from the Centers for Medicare & Medicaid Services¹⁸ was used to convert ICD-10 diagnosis and procedure codes from recent years to ICD-9 diagnosis and procedure codes. Patients were excluded if their encounter contained a diagnosis or procedure code signifying a complex chronic condition,¹⁹ if their hospitalization involved care in the neonatal ICU, or if their admission date occurred outside of the respiratory season. Respiratory season was defined as November 1 through April 30.

Outcomes

Outcomes were measured during three respiratory seasons leading up to adoption and during three respiratory seasons after adoption. The primary outcome was ICU utilization, including the proportion of patients admitted to the ICU and ICU length of stay, expressed as ICU days per 100 patients. Secondary outcomes included mean total length of stay and the proportion of patients who received mechanical ventilation. Lengths of stay were measured in days, the most granular unit of time provided in PHIS over the entire study period. As such, partial days of care are rounded up to 1 full day. A previously published strict definition for mechanical ventilation that limits false positives was used, requiring that patients have a procedure or supply code for mechanical ventilation and a pharmacy charge for a neuromuscular blocking agent.²⁰

Primary Analysis

The primary analysis was restricted to adopting hospitals. An interrupted time series approach was used to measure two possible types of change associated with HFNC protocol adoption: an immediate intervention effect and a change in the slope of an outcome.²¹ The immediate intervention effect represents the change in the level of the outcome that occurs in the period immediately following the introduction of the protocol. The change in slope is the extent to which the outcome changes on a per season basis, attributable to the protocol. Interrupted time series estimates were adjusted for patient age, gender, race, ethnicity, and insurance type; linear regression was used for continuous outcomes and logistic regression for dichotomous outcomes. An ordinary least squares time series model was used to adjust for autocorrelation and Newey-West standard errors were employed.²² Analyses were performed using STATA version 14 (Stata-Corp, College Station, Texas).

Supplementary Analyses

Two preplanned supplementary analyses were conducted. Supplementary analysis 1 was identical to the primary analysis, with the exception that the first season after adoption was censored. The rationale for censoring the first adoption season was to account for a potential learning effect and/or delayed start to full protocol implementation. Supplementary analysis 2 used the nonadopting hospitals as a control group and subtracted the effects measured from an interrupted time series analysis among nonadopting hospitals from the effects measured among adopting hospitals. The rationale for this approach was to control for unmeasured secular (eg, availability of ICU beds) and temporal (eg, severity of a given bronchiolitis season) factors that may have coincidentally occurred with HFNC adoption seasons. The only modification to the interrupted time series approach for supplementary analysis 2 was to provide the nonadopting hospitals with an artificial interruption point because nonadopting hospitals, by definition, did not have an adoption season that could be used in an interrupted time series approach. The interruption point for nonadopting hospitals was set at the median adoption season for adopting hospitals.

Exposure

Leaders at 44 hospitals were contacted regarding their hospital’s use of HFNC outside of the ICU (Figure 1). Responses were obtained for 41 hospitals (93% response rate), 18 of which were classified as nonadopting hospitals. Of the 23 hospitals where the presence of ward-based HFNC protocols were reported, 12 met inclusion criteria and were classified as adopting hospitals. HFNC protocols were adopted at these hospitals in a staggered fashion between the 2010-2011 and 2015-2016 respiratory seasons (Figure 2). The median adoption season was the 2013-14 respiratory season.

Nine adopting hospitals were able to provide details about their first HFNC protocols (Appendix 1). No two protocols were identical, but they shared many similarities. Minimum age requirements ranged from birth to a few months of age. Exclusion criteria were particularly variable, with a history of chronic lung disease or apnea being the most common criteria. Maximum allowed rates of flow ranged from 4 to 10 liters per minute. Criteria for transfer to the ICU were consistently based on an elevated FiO₂ and duration of HFNC exposure.

Patient Characteristics

A total of 32,809 bronchiolitis encounters occurred at adopting hospitals during qualifying respiratory seasons, of which 6,556 (20%) involved patients with a complex chronic condition and were excluded. Of the 26,253 included bronchiolitis encounters, 12,495 encounters occurred prior to ward-based HFNC protocol adoption and 13,758 encounters occurred after adoption. The median age of patients was 8 months (interquartile range, 5-14 months). Most patients were on government insurance (64%), male (58%), of white (56%) or black (18%) race, and of non-Hispanic ethnicity (72%). Pre- and postadoption patient demographics were similar (Appendix 2).

Primary Analysis

Shifts in the level of ICU use and ICU length of stay were observed at the time of adoption of a ward-based HFNC protocol (Figure 3). Specifically, ward-based HFNC protocol adoption was associated with an immediate 3.1% absolute increase (95% CI, 2.8%-3.4%) in the proportion of patients admitted to the ICU and a 9.1 days per 100 patients increase (95% CI, 5.1-13.2) in ICU length of stay (Table). The slope of ICU admissions per season was increasing after HFNC protocol adoption (1.0% increase per season; 95% CI, 0.8%-1.1%). When examined at the individual-hospital level (Appendix 3), seven hospitals were found to have significant increases in ICU admissions (immediate intervention effect or change in slope) after adoption, and one hospital was found to have a significant decrease in ICU admissions (change in slope only). Neither immediate intervention effects nor changes in the slopes of total length of stay and mechanical ventilation were observed, with mean total length of stay approximately 3 days and just over 1% of patients receiving mechanical ventilation (Figure 3).

Supplementary Analyses

Supplementary analyses were largely consistent with the primary analysis. Associations with increased ICU utilization were again observed, although the immediate change in ICU length of stay for supplementary analysis 1 was not significant and the slope for ICU length of stay in supplementary analysis 2 was down trending (Table). Changes in total length of stay and mechanical ventilation were not observed in either supplementary analysis, with the lone exception being an increase in the proportion of patients receiving mechanical ventilation per season (increase in slope) in supplementary analysis 1.

This is the largest multicenter study to date evaluating ICU utilization after adoption of a ward-based HFNC protocol for patients with bronchiolitis. While a principal goal of allowing HFNC use outside of the ICU is to reduce the time that patients with bronchiolitis spend in the ICU, we found that early protocols were, paradoxically, associated with increased ICU utilization. Ward-based HFNC protocols were not associated with changes in hospital length of stay or need for mechanical ventilation. Our findings are particularly relevant given that the majority of children’s hospitals in our sample have adopted ward-based HFNC protocols to care for patients with bronchiolitis.

The increase in ICU utilization measured in our study is a novel finding, seemingly in contradiction to existing literature. Early pilot studies inspired hope that employing HFNC on the general ward might prevent a portion of children from needing ICU care.^11,12 Subsequent larger observational studies did not demonstrate decreases in ICU utilization after adoption of ward-based HFNC protocols.^13,14 The two RCTs comparing low-flow and high-flow nasal cannula use outside of the ICU did not measure a statistically significant effect on ICU utilization, an exploratory outcome in both trials.^15,16 However, the reported point estimates for absolute differences in ICU admission were 2% to 3% higher among patients randomized to HFNC, which is consistent with the 2% to 4% increase in ICU admission measured in the present study.

What might explain this surprising finding? While our observational study cannot speak to mechanism, the protocol details examined in the present study suggest that initial adoption of a ward-based HFNC protocol is often coupled with specific ICU transfer criteria that were unlikely in place prior to protocol initiation. For example, most protocols recommended consideration of ICU transfer for elevated FiO₂ or prolonged duration of HFNC. Transfer to the ICU for prolonged HFNC duration is only possible in the setting of a ward-based HFNC protocol and transfer for elevated FiO₂ was probably unnecessary prior to protocol adoption given that low-flow nasal cannula generally delivers 100% FiO₂. It is also possible that with HFNC comes a perception of increased acuity. For example, medical providers may see patients on HFNC as sicker than patients with the same amount of work of breathing but off HFNC, which makes providers more likely to seek ICU admission for patients on HFNC. The combination of unchanged total length of stay and increased ICU utilization suggests that early ward-based HFNC protocols were an ineffective instrument to improve hospital bed availability during the peak census times that often occur in bronchiolitis season.

The large sample size afforded our study by its multicenter, retrospective design also allowed for a meaningful assessment of the association between a ward-based HFNC protocol and the need for mechanical ventilation. Early indications suggested a lack of substantial association between HFNC use outside of the ICU and rates of mechanical ventilation, but prior studies were limited by small numbers of patients receiving mechanical ventilation (<30 patients in each study).^13,14,16 The present study, in which 783 patients received mechanical ventilation, supports the lack of association between early ward-based HFNC protocols and the need for mechanical ventilation. It should be noted that other studies have measured decreases in mechanical ventilation in association with ICU-based HFNC use.^23-26 In addition to examining HFNC use in a different clinical context, decreases in mechanical ventilation measured after HFNC implementation in the ICU could be explained by preexisting practice trends to limit invasive ventilation and/or selection bias resulting from an increase in less severely ill patients being admitted to the ICU over time. The interrupted time series approach and the staggered adoption of HFNC protocols make the present study less susceptible to biases from preexisting trends and the inclusion of patients cared for both on the ward and within the ICU reduces selection bias.

Our study has several important limitations. First, all hospitals included in the analysis were US children’s hospitals and these findings may not generalize to other practice environments, including community hospitals and other countries. Second, our cohort and outcomes were defined using administrative billing data, which have been incompletely validated, making some degree of misclassification likely. Third, we measured HFNC exposure at the hospital level, but could not examine the extent to which individual patients were exposed to HFNC because such data are not present in PHIS. Even if we had access to patient-level HFNC exposure data, we would have still compared outcomes among all patients with bronchiolitis (not just those who received HFNC), to avoid selection bias. However, knowing HFNC exposure status at the patient level would have allowed for weighting of the effects measured at each hospital according to the extent of HFNC exposure. Fourth, there are likely other, unmeasured secular and temporal factors that could affect study outcomes. To some degree, the interrupted time series approach, observed staggered adoption of protocols, and nonadopting hospital supplementary analysis mitigate this risk of bias. Fifth, while the pre- and postadoption populations appeared demographically similar, it is possible that the populations might have differed by other unmeasured factors. Finally, early ward-based HFNC protocols have likely undergone iterative changes since adoption. We compared pre- and postadoption outcome slopes and censored the first adoption season in a supplementary analysis to attempt to account for this potential limitation.

In conclusion, our findings suggest that initial implementation of ward-based HFNC protocols were not successful at reducing ICU utilization for children with bronchiolitis. Future research should examine whether more evolved HFNC protocols that use higher flow rates, more generous ICU transfer criteria, and more rapid weaning criteria can reduce ICU utilization.

We thank Dr Vineeta Mittal (University of Texas Southwestern Medical Center) for providing feedback regarding the manuscript.

Children hospitalized for bronchiolitis frequently require admission to the intensive care unit (ICU), with estimates as high as 18%^1,2 and 35%³ in two prospective, multicenter studies. The indication for ICU admission is nearly always a need for advanced respiratory support, which historically consisted of continuous or bilevel positive airway pressure (CPAP and BiPAP, respectively) or mechanical ventilation. High-flow nasal cannula (HFNC) is a recent addition to the respiratory support armamentarium, delivering heated and humidified oxygen at rates of up to 60 L/min and allowing for clinicians to titrate both flow rate and fraction of inspired oxygen (FiO₂).⁴

Several studies have demonstrated that HFNC is capable of decreasing a child’s work of breathing,^5-8 and it has the potential advantage of being better tolerated than other forms of advanced respiratory support.^9,10 These case-series physiologic studies informed early ward-based HFNC protocols for bronchiolitis, which were adopted to decrease ICU utilization. Since then, single center observational studies examining the association between ward-based HFNC protocols and subsequent ICU utilization have come to discordant conclusions.^11-14 Studying the effect of employing HFNC outside of the ICU is challenging in the context of a randomized, controlled trial (RCT) because it is difficult to blind healthcare providers to the intervention and because crossover from the control group to HFNC is frequent. Two unblinded RCTs published in 2017 and 2018 found that children randomized to conventional nasal cannula were frequently escalated to HFNC (flow rates of 1-2 L/kg per minute), but neither trial found a difference in ICU admission.^15,16 Sample sizes substantially larger than those present in currently published or registered RCTs would be required to evaluate the impact of ward-based HFNC protocols on the outcome that inspired the protocols in the first place, namely ICU utilization.¹⁷

Children’s hospitals have adopted ward-based HFNC protocols at different time points over the last decade, which allows for a natural experiment—a promising alternative study design that avoids the challenges of blinding, crossover, and modest sample sizes. In order to have sufficient postadoption data for analyses, the present study is limited to ward-based HFNC protocols adopted prior to 2016, which we have termed “early” ward-based HFNC protocols. Among children with bronchiolitis, our objective was to measure the association between hospital-level adoption of a ward-based HFNC protocol and subsequent ICU utilization, using a multicenter network of children’s hospitals.

We conducted a multicenter retrospective cohort study using the Pediatric Health Information System (PHIS) database. The PHIS database is operated by the Children’s Hospital Association (Lenexa, Kansas) and provides deidentified patient-level information for children who receive hospital care at 55 US children’s hospitals. Available data elements include patient demographic data, discharge diagnosis and procedure codes, and detailed billing information, such as laboratory, imaging, pharmacy, and supply charges. At the patient level, the use of HFNC vs standard oxygen therapy circuits cannot be discriminated.

Exposure

The study exposure was a hospital’s first ward-based HFNC protocol, with adoption measured at the hospital level at each PHIS site via direct communication with leaders in hospital medicine. In most cases, first contact was made with the pediatric hospital medicine division chief or fellowship program director, who then, if necessary, connected study investigators to local HFNC champions aware of site-specific historical HFNC protocol details. Contact with a hospital was made only if the hospital had contributed at least 6 consecutive years of data to PHIS. Hospitals were classified as “adopting” hospitals if their HFNC protocol met all of the following criteria: (a) allows initiation of HFNC outside of the ICU (on the floor or in the ED), (b) allows continued care outside of the ICU (on the floor), (c) not limited to a small unit like an intermediate care unit, and (d) adopted during a specific, known respiratory season. Hospitals for which ward-based HFNC protocols were adopted but did not meet these criteria were excluded from further analysis. Our intent was to identify large scale, programmatic protocol launches and exclude hospitals with exceptions that might preclude a sizable portion of our cohort from being eligible for the protocol. Hospitals for which inpatient use of HFNC remains limited to the ICU were defined as “nonadopting” hospitals. Respondents at adopting hospitals were asked to share details about their protocol, including patient eligibility criteria and maximum HFNC rates of flow permitted outside of the ICU.

Patient Characteristics

Patients aged 3 to 24 months who were hospitalized at adopting and nonadopting hospitals were included if an International Classification of Diseases, Ninth Revision (ICD-9) discharge diagnosis code for bronchiolitis (466.XX) was present in any position (not limited to a primary diagnosis). The lower age limit of 3 months was chosen to match the most restrictive age eligibility criteria of provided HFNC protocols (Appendix 1). A crosswalk available from the Centers for Medicare & Medicaid Services¹⁸ was used to convert ICD-10 diagnosis and procedure codes from recent years to ICD-9 diagnosis and procedure codes. Patients were excluded if their encounter contained a diagnosis or procedure code signifying a complex chronic condition,¹⁹ if their hospitalization involved care in the neonatal ICU, or if their admission date occurred outside of the respiratory season. Respiratory season was defined as November 1 through April 30.

Outcomes

Outcomes were measured during three respiratory seasons leading up to adoption and during three respiratory seasons after adoption. The primary outcome was ICU utilization, including the proportion of patients admitted to the ICU and ICU length of stay, expressed as ICU days per 100 patients. Secondary outcomes included mean total length of stay and the proportion of patients who received mechanical ventilation. Lengths of stay were measured in days, the most granular unit of time provided in PHIS over the entire study period. As such, partial days of care are rounded up to 1 full day. A previously published strict definition for mechanical ventilation that limits false positives was used, requiring that patients have a procedure or supply code for mechanical ventilation and a pharmacy charge for a neuromuscular blocking agent.²⁰

Primary Analysis

The primary analysis was restricted to adopting hospitals. An interrupted time series approach was used to measure two possible types of change associated with HFNC protocol adoption: an immediate intervention effect and a change in the slope of an outcome.²¹ The immediate intervention effect represents the change in the level of the outcome that occurs in the period immediately following the introduction of the protocol. The change in slope is the extent to which the outcome changes on a per season basis, attributable to the protocol. Interrupted time series estimates were adjusted for patient age, gender, race, ethnicity, and insurance type; linear regression was used for continuous outcomes and logistic regression for dichotomous outcomes. An ordinary least squares time series model was used to adjust for autocorrelation and Newey-West standard errors were employed.²² Analyses were performed using STATA version 14 (Stata-Corp, College Station, Texas).

Supplementary Analyses

Two preplanned supplementary analyses were conducted. Supplementary analysis 1 was identical to the primary analysis, with the exception that the first season after adoption was censored. The rationale for censoring the first adoption season was to account for a potential learning effect and/or delayed start to full protocol implementation. Supplementary analysis 2 used the nonadopting hospitals as a control group and subtracted the effects measured from an interrupted time series analysis among nonadopting hospitals from the effects measured among adopting hospitals. The rationale for this approach was to control for unmeasured secular (eg, availability of ICU beds) and temporal (eg, severity of a given bronchiolitis season) factors that may have coincidentally occurred with HFNC adoption seasons. The only modification to the interrupted time series approach for supplementary analysis 2 was to provide the nonadopting hospitals with an artificial interruption point because nonadopting hospitals, by definition, did not have an adoption season that could be used in an interrupted time series approach. The interruption point for nonadopting hospitals was set at the median adoption season for adopting hospitals.

Exposure

Leaders at 44 hospitals were contacted regarding their hospital’s use of HFNC outside of the ICU (Figure 1). Responses were obtained for 41 hospitals (93% response rate), 18 of which were classified as nonadopting hospitals. Of the 23 hospitals where the presence of ward-based HFNC protocols were reported, 12 met inclusion criteria and were classified as adopting hospitals. HFNC protocols were adopted at these hospitals in a staggered fashion between the 2010-2011 and 2015-2016 respiratory seasons (Figure 2). The median adoption season was the 2013-14 respiratory season.

Nine adopting hospitals were able to provide details about their first HFNC protocols (Appendix 1). No two protocols were identical, but they shared many similarities. Minimum age requirements ranged from birth to a few months of age. Exclusion criteria were particularly variable, with a history of chronic lung disease or apnea being the most common criteria. Maximum allowed rates of flow ranged from 4 to 10 liters per minute. Criteria for transfer to the ICU were consistently based on an elevated FiO₂ and duration of HFNC exposure.

Patient Characteristics

A total of 32,809 bronchiolitis encounters occurred at adopting hospitals during qualifying respiratory seasons, of which 6,556 (20%) involved patients with a complex chronic condition and were excluded. Of the 26,253 included bronchiolitis encounters, 12,495 encounters occurred prior to ward-based HFNC protocol adoption and 13,758 encounters occurred after adoption. The median age of patients was 8 months (interquartile range, 5-14 months). Most patients were on government insurance (64%), male (58%), of white (56%) or black (18%) race, and of non-Hispanic ethnicity (72%). Pre- and postadoption patient demographics were similar (Appendix 2).

Primary Analysis

Shifts in the level of ICU use and ICU length of stay were observed at the time of adoption of a ward-based HFNC protocol (Figure 3). Specifically, ward-based HFNC protocol adoption was associated with an immediate 3.1% absolute increase (95% CI, 2.8%-3.4%) in the proportion of patients admitted to the ICU and a 9.1 days per 100 patients increase (95% CI, 5.1-13.2) in ICU length of stay (Table). The slope of ICU admissions per season was increasing after HFNC protocol adoption (1.0% increase per season; 95% CI, 0.8%-1.1%). When examined at the individual-hospital level (Appendix 3), seven hospitals were found to have significant increases in ICU admissions (immediate intervention effect or change in slope) after adoption, and one hospital was found to have a significant decrease in ICU admissions (change in slope only). Neither immediate intervention effects nor changes in the slopes of total length of stay and mechanical ventilation were observed, with mean total length of stay approximately 3 days and just over 1% of patients receiving mechanical ventilation (Figure 3).

Supplementary Analyses

Supplementary analyses were largely consistent with the primary analysis. Associations with increased ICU utilization were again observed, although the immediate change in ICU length of stay for supplementary analysis 1 was not significant and the slope for ICU length of stay in supplementary analysis 2 was down trending (Table). Changes in total length of stay and mechanical ventilation were not observed in either supplementary analysis, with the lone exception being an increase in the proportion of patients receiving mechanical ventilation per season (increase in slope) in supplementary analysis 1.

This is the largest multicenter study to date evaluating ICU utilization after adoption of a ward-based HFNC protocol for patients with bronchiolitis. While a principal goal of allowing HFNC use outside of the ICU is to reduce the time that patients with bronchiolitis spend in the ICU, we found that early protocols were, paradoxically, associated with increased ICU utilization. Ward-based HFNC protocols were not associated with changes in hospital length of stay or need for mechanical ventilation. Our findings are particularly relevant given that the majority of children’s hospitals in our sample have adopted ward-based HFNC protocols to care for patients with bronchiolitis.

The increase in ICU utilization measured in our study is a novel finding, seemingly in contradiction to existing literature. Early pilot studies inspired hope that employing HFNC on the general ward might prevent a portion of children from needing ICU care.^11,12 Subsequent larger observational studies did not demonstrate decreases in ICU utilization after adoption of ward-based HFNC protocols.^13,14 The two RCTs comparing low-flow and high-flow nasal cannula use outside of the ICU did not measure a statistically significant effect on ICU utilization, an exploratory outcome in both trials.^15,16 However, the reported point estimates for absolute differences in ICU admission were 2% to 3% higher among patients randomized to HFNC, which is consistent with the 2% to 4% increase in ICU admission measured in the present study.

What might explain this surprising finding? While our observational study cannot speak to mechanism, the protocol details examined in the present study suggest that initial adoption of a ward-based HFNC protocol is often coupled with specific ICU transfer criteria that were unlikely in place prior to protocol initiation. For example, most protocols recommended consideration of ICU transfer for elevated FiO₂ or prolonged duration of HFNC. Transfer to the ICU for prolonged HFNC duration is only possible in the setting of a ward-based HFNC protocol and transfer for elevated FiO₂ was probably unnecessary prior to protocol adoption given that low-flow nasal cannula generally delivers 100% FiO₂. It is also possible that with HFNC comes a perception of increased acuity. For example, medical providers may see patients on HFNC as sicker than patients with the same amount of work of breathing but off HFNC, which makes providers more likely to seek ICU admission for patients on HFNC. The combination of unchanged total length of stay and increased ICU utilization suggests that early ward-based HFNC protocols were an ineffective instrument to improve hospital bed availability during the peak census times that often occur in bronchiolitis season.

The large sample size afforded our study by its multicenter, retrospective design also allowed for a meaningful assessment of the association between a ward-based HFNC protocol and the need for mechanical ventilation. Early indications suggested a lack of substantial association between HFNC use outside of the ICU and rates of mechanical ventilation, but prior studies were limited by small numbers of patients receiving mechanical ventilation (<30 patients in each study).^13,14,16 The present study, in which 783 patients received mechanical ventilation, supports the lack of association between early ward-based HFNC protocols and the need for mechanical ventilation. It should be noted that other studies have measured decreases in mechanical ventilation in association with ICU-based HFNC use.^23-26 In addition to examining HFNC use in a different clinical context, decreases in mechanical ventilation measured after HFNC implementation in the ICU could be explained by preexisting practice trends to limit invasive ventilation and/or selection bias resulting from an increase in less severely ill patients being admitted to the ICU over time. The interrupted time series approach and the staggered adoption of HFNC protocols make the present study less susceptible to biases from preexisting trends and the inclusion of patients cared for both on the ward and within the ICU reduces selection bias.

Our study has several important limitations. First, all hospitals included in the analysis were US children’s hospitals and these findings may not generalize to other practice environments, including community hospitals and other countries. Second, our cohort and outcomes were defined using administrative billing data, which have been incompletely validated, making some degree of misclassification likely. Third, we measured HFNC exposure at the hospital level, but could not examine the extent to which individual patients were exposed to HFNC because such data are not present in PHIS. Even if we had access to patient-level HFNC exposure data, we would have still compared outcomes among all patients with bronchiolitis (not just those who received HFNC), to avoid selection bias. However, knowing HFNC exposure status at the patient level would have allowed for weighting of the effects measured at each hospital according to the extent of HFNC exposure. Fourth, there are likely other, unmeasured secular and temporal factors that could affect study outcomes. To some degree, the interrupted time series approach, observed staggered adoption of protocols, and nonadopting hospital supplementary analysis mitigate this risk of bias. Fifth, while the pre- and postadoption populations appeared demographically similar, it is possible that the populations might have differed by other unmeasured factors. Finally, early ward-based HFNC protocols have likely undergone iterative changes since adoption. We compared pre- and postadoption outcome slopes and censored the first adoption season in a supplementary analysis to attempt to account for this potential limitation.

In conclusion, our findings suggest that initial implementation of ward-based HFNC protocols were not successful at reducing ICU utilization for children with bronchiolitis. Future research should examine whether more evolved HFNC protocols that use higher flow rates, more generous ICU transfer criteria, and more rapid weaning criteria can reduce ICU utilization.

We thank Dr Vineeta Mittal (University of Texas Southwestern Medical Center) for providing feedback regarding the manuscript.