Multisite Emergency Department Inpatient Collaborative to Reduce Unnecessary Bronchiolitis Care
BACKGROUND AND OBJECTIVES: There is high variation in the care of acute viral bronchiolitis. We sought to promote collaboration between emergency department (ED) and inpatient (IP) units with the goal of reducing unnecessary testing and treatment.
METHODS: Multisite collaborative with improvement teams co-led by ED and IP physicians and a 1-year period of active participation. The intervention consisted of a multicomponent change package, regular webinars, and optional coaching. Data were collected by chart review for December 2014 through March 2015 (baseline) and December 2015 to March 2016 (improvement period). Patients <24 months of age with a primary diagnosis of bronchiolitis and without ICU admission, prematurity, or chronic lung or heart disease were eligible for inclusion. Control charts were used to detect improvement. Achievable benchmarks of care were calculated for each measure.
RESULTS: Thirty-five hospitals with 5078 ED patients and 4389 IPs participated. Use of bronchodilators demonstrated special cause for the ED (mean centerline shift: 37.1%–24.5%, benchmark 5.8%) and IP (28.4%–17.7%, benchmark 9.1%). Project mean ED viral testing decreased from 42.6% to 25.4% after revealing special cause with a 3.9% benchmark, whereas chest radiography (30.9%), antibiotic use (6.2%), and steroid use (7.6%) in the ED units did not change. IP steroid use decreased from 7.2% to 4.0% after special cause with 0.0% as the benchmark. Within-site ED and IP performance was modestly correlated.
CONCLUSIONS: Collaboration between ED and IP units was associated with a decreased use of unnecessary tests and therapies in bronchiolitis; top performers used few unnecessary tests or treatments.
- AAP —
- American Academy of Pediatrics
- ABC —
- achievable benchmark of care
- B-QIP —
- Quality Collaborative for Improving Hospital Compliance with the American Academy of Pediatrics Bronchiolitis Guideline
- ED —
- emergency department
- IP —
- IQR —
- interquartile range
- LOS —
- length of stay
- QI —
- quality improvement
- SPC —
- statistical process control
- VIP —
- Value in Inpatient Pediatrics
Acute viral bronchiolitis is among the most common illnesses seen in the emergency department (ED) and is 1 of the top 3 reasons for pediatric admissions to the hospital.1–3 The available research has not revealed any therapy as providing a significant impact on outcomes in the disease.4–8 The authors of numerous studies have demonstrated that there is high variation in use of tests and therapies for bronchiolitis across health care settings,9,10 and the most recent American Academy of Pediatrics (AAP) guideline includes recommendations against the routine use of most testing and strong recommendations against bronchodilators and steroids.11
The members of the Value in Inpatient Pediatrics (VIP) Network use a “virtual collaborative” model to do quality improvement (QI) involving volunteer provider-led, hospital-based teams interacting via telephone, e-mail, and the Internet to promote evidence-based practice. Previous work using this model has revealed significant reductions in unnecessary care for inpatient (IP) bronchiolitis across multiple hospitals12; however, the care continuum for patients with bronchiolitis crosses multiple clinical settings, and choices about treatment and therapy in one setting can influence physician, staff, and parental expectations in other settings in ways that make guideline adherence challenging.13
Recognizing that collaboration and coordination between the ED and IP settings may strongly influence the success of improvement efforts for acute bronchiolitis, this collaborative, entitled Stewardship in Bronchiolitis, was established with the overall goal of reducing unnecessary testing and treatment across the care continuum, in accordance with published guidelines.11
The specific aims of the collaborative were, for all eligible patients with acute viral bronchiolitis in the ED or IP setting, to (1) decrease the percentage of patients receiving bronchodilators to 20% (ED) and 10% (IP), (2) decrease the number of doses after the initial dose by 50% for patients treated with bronchodilators, (3) decrease the usage of corticosteroids to 0%, (4) decrease the percentage of patients receiving chest radiographs to 30% (ED), (5) decrease the percentage of patients receiving viral testing to 25% (ED), and (6) decrease the percentage of patients receiving antibiotics to 20% (ED) in 1 year. Additional aims were to establish benchmark values for usage metrics in the ED and IP settings and to describe any relationship between ED and IP performance at the site level.
The design of the project was modeled on a previous VIP Network bronchiolitis collaborative12 centered around webinars, online data sharing, a project listserv, and individual site coaching by e-mail and telephone. The participation period was from August 2015 to July 2016. Target size for the collaborative was 35 hospitals. Minimum eligibility criteria for participation were (1) a minimum of 50 bronchiolitis ED visits or IP visits per year, (2) colocated ED and IP units, and (3) an identified project leader from both the ED and IP practice areas. ED and IP coleaders were expected to lead improvement in their respective practice areas and were encouraged to coordinate improvement activities. Sites were selected to prioritize settings where strategies for improving collaboration between units could be articulated, settings where context factors perceived by the site leaders to contribute to success (such as leadership buy-in) were present, and settings with an overall need for standardization.
In accordance with the Model for Understanding Success in Quality framework,14 the project was conceived to leverage several context factors as a means to promoting overall improvement: (1) maintenance of certification requirements as an external motivator for physicians, (2) AAP project sponsorship, and (3) provision of Web-based data infrastructure. Collaboration between the ED and IP units was postulated to be important for success on the basis of qualitative interviews with the previous collaborative.12,13
Project planning began with an expert group meeting including members with expertise in bronchiolitis and QI, specifically including both IP and ED physicians. The goals of the meeting were to define measures for the project on the basis of anticipated updates to the AAP clinical practice guideline for bronchiolitis,11 to establish site recruitment criteria, and to identify components of the change package intervention.
At each site, core improvement teams were formed that comprised at least 1 member of each of the following: a hospital medicine physician, an ED physician, and a respiratory therapist or nurse. Site teams were encouraged to include other relevant stakeholders as appropriate.
The change package consisted of examples of evidence-based pathways and order sets, a respiratory scoring tool, and sample communication verbiage for parents and referring physicians; this open-access package also included recorded webinars.15 Teams planned their individual interventions after review of their baseline data to address local gaps and opportunities. At the collaborative level, implementation was supported by learning and information-sharing webinars, a dedicated project listserv, and optional coaching. Group aggregate performance was reviewed during the webinars, and individual site teams locally performed plan-do-study-act cycles by using their data and group comparison data.
Study of the Intervention(s)
ED and hospital encounters for patients <2 years of age with a primary diagnosis of acute viral bronchiolitis (International Classification of Diseases, Ninth Revision codes 466.11 and 466.19) were eligible for inclusion after chart review to confirm the diagnosis. Encounters were excluded for prematurity (<35 weeks’ gestational age); congenital heart disease; chronic lung disease; genetic, congenital, or neuromuscular abnormalities; and pediatric intensive care admission.
Measures and Data Collection
Usage measures were chosen on the basis of the AAP 2014 clinical practice guideline. Goal rates for the collaborative were set by using achievable benchmarks of care (ABCs)16,17 from the literature, when available, or were extrapolated by consensus from guideline recommendations.11 In addition, site teams were encouraged to use a 50% reduction in usage as an alternative goal if this met local needs better than project benchmarks (as in the case of high usage).
ED measures included the percentage of sampled patients receiving a chest radiograph, viral testing (except influenza), antibiotics, steroids, or a bronchodilator; and measures included the number of bronchodilator doses per patient when this medication was received. Patients tested for influenza via a combination panel were recorded as having had a viral test. IP measures included the percentage of sampled patients receiving a bronchodilator or steroid and the number of bronchodilator doses received during admission when used. Lengths of stay (LOSs) in the ED and IP units were chosen as balancing measures because providers may choose to observe patients for a greater duration when fewer interventions are used. In the previous QI literature on bronchiolitis, readmissions were not impacted and were generally rare12,18; thus, we chose not to track them as a balancing measure in this project.
Hospital characteristics were collected by using a preproject survey completed by leaders at each site. Patient data collection was accomplished by manual chart review and included only the information necessary to compute usage rates on performance metrics and patient ages in months. Teams input data into the AAP Quality Improvement Data Aggregator, a Web-based data repository. The intended sampling strategy was to include the first 20 charts per month that met inclusion criteria from each area, that is, 40 charts per hospital (20 ED charts and 20 IP charts). If fewer than 20 patients met inclusion criteria per area, then all patients were included. Charts were reviewed by using a structured chart review tool, and the data were entered into the Web aggregator. Training for using the chart review tool consisted of a live webinar that was also recorded and available for repeat viewing. Data were collected from December 2014 through March 2015 to establish a baseline and from December 2015 through March 2016 for the improvement period.
For each measure, data were plotted on 2 run charts: (1) a run chart for the individual site and (2) a combined run chart for the entire collaborative. For the combined run chart, weekly data points were used, whereas the site-level run chart used monthly time points, given the smaller sample size. Site-level data (monthly medians) were available in real time for each site team to use in improvement efforts; selected collaborative-level run charts with weekly medians were shared during a webinar session at the midpoint of the improvement period and via the listserv as the data became available. For the final analysis of usage measures at the collaborative level, run charts were converted to statistical process control (SPC) charts and then centerline-adjusted according to the rules for identifying special cause.19,20
ABC values were computed for each metric in a manner consistent with described methodology.16,17 At the conclusion of data entry, sites were ranked in descending order on the basis of postintervention performance. Top sites were then selected such that the total number of patients from those sites totaled at least 10% of the population. The ABC was then calculated by averaging the performance of that group.
For the bronchodilator and steroids measures, we tested the association between IP and ED site performance in relation to project goals by using a Fisher’s exact test. Sites were classified as IP and ED goals met, IP goals met, ED goals met, or no goals met on the basis of the stated collaborative specific aim for the measure or site goal (50% decrease in usage).
The average LOS for the entire collaborative was analyzed for differences between the baseline and improvement time periods by using the Wilcoxon rank test. Statistical comparison between the 2 periods was computed by using R version 3.2.4 (R Foundation for Statistical Computing, Vienna, Austria).21
This project was approved by the AAP Institutional Review Board. Written informed consent was obtained from each site’s improvement team leadership. Local site teams handled individual institutional review board applications as deemed necessary by the leaders of the participating institution. No protected health information or patient identifiers were collected for the project, and sites were deidentified in public presentations of data.
Applications from 66 hospitals were received, and 35 hospitals were selected for participation. Site characteristics are described in Table 1. One site team provided only IP data and was included in the SPC analysis for IP performance but excluded from further analysis of ED and IP performance correlations. A total of 9467 charts were reviewed, 4739 during baseline (2537 ED, 2202 IP) and 4728 in the improvement period (2541 ED, 2187 IP). The average age of ED patients was 7.6 months (median: 6, interquartile range [IQR]: 3–11) during the baseline period and 8.0 months (median: 6, IQR: 4–12) during the improvement period. The average age of IP patients was 6.5 months (median: 5, IQR: 2–10) during the baseline period and 6.7 months (median: 5, IQR: 2–10) during the improvement period.
Control charts depicting results for collaborative-level performance are presented in Fig 1 (ED and IP bronchodilator and steroid usage) and Fig 2 (ED chest radiography, viral testing, and antibiotic use). The mean percentage of patients receiving a bronchodilator decreased from 37.1% to 24.5% in the ED units and from 28.4% to 17.7% for IP units. The number of bronchodilator doses per patient remained at 1.4 doses per patient in the ED units and decreased from 5.8 to 4.2 doses per patient for IP units. The percentage of patients receiving steroids remained at 7.6% in the ED units and decreased from 7.2% to 4.0% for IP units. In the ED units, the mean percentage of viral testing decreased from 42.6% to near the goal at 25.4%, the chest radiography use remained near the goal at 30.9%, and antibiotic prescribing met the goal but remained at 6.2%. The ED LOS decreased by 6.6 minutes (P < .001). The IP LOS decreased by 6.9 hours (P < .001).
Distribution of the change package was temporally associated with special cause for improvement for the percentage of patients receiving bronchodilators in the ED and IP units, for IP doses of bronchodilators per patient and IP percentage receiving steroids, and for ED percentage receiving viral testing. Special cause for improvement was evident again at 28 weeks (January 31, 2016) for the percentage receiving bronchodilators in the ED.
Site performance during the baseline and improvement periods is described in Table 2, with comparison of project goals and ABCs for each measure. More than a third of site teams met a project goal for all measures.
In Fig 3, we plotted ED versus IP performance on project goals for any use of a bronchodilator and steroid measures. We found an association in performance on goals for the bronchodilator measure (P = .03), with 10 of 34 (29.4%) site teams meeting goals in both ED and IP settings and 14 site teams (41.2%) meeting neither criteria. We found no significant association in performance on steroid use goals (P = .16), with 8 (23.5%) site teams meeting goals in both locations and 14 (41.2%) site teams meeting goals in neither location.
In this multisite QI collaborative with more than 9000 patients at 35 hospital sites, we demonstrated improved performance in care of acute viral bronchiolitis for 2 of 6 ED measures and 3 of 3 IP measures. We also established ABC values for ED and IP usage measures, which indicate that evidence-based care was achieved in multiple settings. A novel aspect of our study was the requirement that ED and IP physicians colead improvement at each site for the purpose of promoting collaboration. We demonstrated an association in performance on project goals (both ED and IP meeting goals or not meeting goals) within sites for any use of a bronchodilator but not for any use of steroids.
It is notable that the final performance on ED measures in our collaborative is better than previously published ED usage rates. By using 2005 data from a nationally representative cross section of EDs, Knapp et al22 found that 53% of bronchiolitis patients treated in the ED received antibiotics (compared with 6.2% in this report) and 72% of patients received chest radiographs (compared with 30.9%). By using data from the same survey for the years 2001–2009, Johnson et al23 found that 53.8% of patients received bronchodilators (compared with 24.5% in our study) and that 20.4% received corticosteroids (compared with 7.6%).
Four of 6 ED measures (the rates of chest radiography, steroid, and antibiotic usage, as well as the number of bronchodilator doses given per patient) did not improve over the course of the project. Reasons for this lack of improvement may include the fact that the usage rates of some of these tests and therapies was already low. Nevertheless, a performance gap continues to exist between mean use and the ABC benchmarks for the collaborative. For example, although the 7.6% rate of steroid use in the ED in our study is low compared with that shown in published literature, the ABC was 0.0%, indicating that improvement from that level was not unrealistic. More systematic collection of failure modes and context data are likely necessary to achieve higher performance.
The mean IP performance in this collaborative also compares favorably with that shown in the literature. On the bronchodilator measure, baseline (28.4%) and postproject (17.7%) performance were both superior to the rate of 32.7% achieved in the previous VIP Network–sponsored Quality Collaborative for Improving Hospital Compliance with the American Academy of Pediatrics Bronchiolitis Guideline (B-QIP),12 although the magnitude of improvement and rates of systemic corticosteroid usage were similar between the 2 projects (4% and 2%, respectively). Comparable single-center QI work reported by Mittal et al24 revealed a final performance of 14% on the bronchodilator measure and 11% on the steroid measure.
Possible explanations for the lower overall rate of mean bronchodilator use in this collaborative include national secular trends or selection bias in our site recruitment, although the lower rate in our collaborative is consistent with our initial hypothesis that better alignment between ED and IP providers would result in more evidence-based care. Approximately two-thirds of site teams in this collaborative self-described as university-affiliated, compared with approximately half in the B-QIP. Although it has been hypothesized and sometimes demonstrated25 that sites with an academic affiliation may outperform national averages, performance benchmarks in the B-QIP in fact compared favorably with data that included only academic children’s hospitals26,27; and in our collaborative, site teams without an academic affiliation made up a percentage of the top quartile of performers that was similar to the percentage found in the overall collaborative.
The possibility of a national secular trend toward decreasing unnecessary therapies for bronchiolitis as a contributing factor is interesting and perhaps encouraging, but data on the existence of such a trend is conflicting. Parikh et al28 analyzed data from 2004–2012 before and after publication of the 2006 AAP guideline29 and found statistical trends toward decreasing unnecessary diagnostic and therapeutics for hospitalized children at children’s hospitals, but Florin et al30 found no decrease in 3 of 5 resources not recommended for routine use in bronchiolitis and only a small decrease in chest radiography use and steroid use for the years after the publication of the same guideline. For patients treated in the ED, Johnson et al23 found no statistical trend in bronchodilator, steroid, or antibiotic use for the years 2001–2009 but did find a decrease in chest radiography use.
Our study has several limitations. First, it is possible that implementation of the data collection strategy varied among sites. Although the search strategy was specified a priori and chart reviewers received training, some variation in coding strategies may be expected among sites, and some selection bias may have occurred. In addition, sample size variation occurred among sites, with some site teams failing to reach the recommended 20 charts per month because of low census and others choosing to enter more than the recommended 20 charts per month to gain improved understanding of local performance. This may have caused some site-level confounding on the aggregated data. Third, although the use of SPC represents an improved method of analysis compared with pre-post design, intervention timing and content varied at a local level, and critical interventions were difficult to identify. This is partially mitigated by the simplicity of the change package; in other words, context factors likely account for much of the improvement, and there are not a large number of specific interventions to study. Finally, our study was primarily concerned with promoting practice improvement and can only really be hypothesis-generating around the possibility that collaboration between the ED and IP units improved outcomes. As interest in participation in similar collaborative grows, future work should seek to take advantage of large collaborative size to study site-level variation in context factors more rigorously as a predictor of performance improvement.
In a multisite bronchiolitis improvement collaborative emphasizing collaboration between ED and IP units, the usage of several unnecessary therapies decreased in both settings. Although usage rates were lower than previously reported, performance gaps between average and top sites reveal that continued improvement is still possible. The correlation between meeting ED and IP goals suggests that within-site collaboration may be an important context factor for successful improvement.
- Accepted September 13, 2017.
- Address correspondence to Grant M. Mussman, MD, MHSA, Cincinnati Children’s Hospital Medical Center, 3333 Burnet Ave, MLC 3024, Cincinnati, OH 45229. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Funded by the American Academy of Pediatrics Friends of Children Fund.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Hasegawa K,
- Tsugawa Y,
- Brown DF,
- Mansbach JM,
- Camargo CA Jr
- Bourgeois FT,
- Monuteaux MC,
- Stack AM,
- Neuman MI
- Knapp JF,
- Simon SD,
- Sharma V
- Conway PH,
- Edwards S,
- Stucky ER,
- Chiang VW,
- Ottolini MC,
- Landrigan CP
- Ralston SL,
- Garber MD,
- Rice-Conboy E, et al; Value in Inpatient Pediatrics Network Quality Collaborative for Improving Hospital Compliance with AAP Bronchiolitis Guideline (BQIP)
- Kaplan HC,
- Provost LP,
- Froehle CM,
- Margolis PA
- Stewardship in Bronchiolitis Expert Group
- Ralston S,
- Comick A,
- Nichols E,
- Parker D,
- Lanter P
- Benneyan JC,
- Lloyd RC,
- Plsek PE
- Langley GJ,
- Moen RD,
- Nolan KM,
- Nolan TW,
- Norman CL,
- Provost LP
- R Foundation for Statistical Computing
- Knapp JF,
- Simon SD,
- Sharma V
- Johnson LW,
- Robles J,
- Hudgins A,
- Osburn S,
- Martin D,
- Thompson A
- Parikh K,
- Hall M,
- Mittal V, et al
- American Academy of Pediatrics Subcommittee on Diagnosis and Management of Bronchiolitis
- Copyright © 2018 by the American Academy of Pediatrics