Emergency department visits for anaphylaxis have increased considerably over the past few decades, especially among children. Despite this, anaphylaxis management remains highly variable and contributes to significant health care spending. On the basis of emerging evidence, in this quality improvement project we aimed to safely decrease hospitalization rates, increase the use of cetirizine, and decrease use of corticosteroids for children with anaphylaxis by December 31, 2019.
A multipronged intervention strategy including a revised evidence-based guideline was implemented at a tertiary children’s teaching hospital by using the Model for Improvement. Statistical process control was used to evaluate for changes in key measures. Length of stay and unplanned return visits within 72 hours were monitored as process and balancing measures, respectively. As a national comparison, hospitalization rates were compared with other hospitals’ data from the Pediatric Health Information System.
Hospitalizations decreased significantly from 28.5% to 11.2% from preimplementation to implementation, and the balancing measure of 72-hour revisits was stable. The proportion of patients receiving cetirizine increased significantly from 4.2% to 59.7% and use of corticosteroids decreased significantly from 72.6% to 32.4% in patients without asthma. The proportion of patients meeting length of stay criteria increased from 53.3% to 59.9%. Hospitalization rates decreased nationally over time.
We reduced hospitalizations for anaphylaxis by 17.3% without concomitant increases in revisits, demonstrating that unnecessary hospitalizations can be safely avoided. The use of a local evidence-based guideline paired with close outcome monitoring and sustained messaging and feedback to clinicians can effectively improve anaphylaxis management.
Anaphylaxis is a serious allergic reaction.1 Emergency department (ED) visits for anaphylaxis have doubled in 10 years and hospitalizations have increased sevenfold, with the greatest increases occurring in children.2–5 Management varies considerably across institutions; hospitalization rates in some locations are as high as 98%, with a median of 40% to 50%.4,6,7 However, hospitalization rates are <10% in Canada and Australia, and emerging evidence suggests the need for late intervention in anaphylaxis is rare.8,9 Children are frequently hospitalized to observe for biphasic anaphylaxis, but these reactions likely occur in as few as 5% of patients.10,11 Moreover, most biphasic reactions occur within 4 hours of the initial reaction, and late biphasic reactions are most often mild.10,12,13 Given this evidence, an opportunity to decrease hospitalizations in the Unites States exists.
There is also substantial variability in pharmacologic management of anaphylaxis.14 Epinephrine is the only medication that is consistently recommended by guidelines for anaphylaxis treatment,15,16 and there remains a dearth of evidence supporting the routine use of adjunctive therapies. Despite that, corticosteroids may be used as or more frequently than epinephrine, in some studies by as much as 20% to 30%.17–20 Second-generation H1 antihistamines such as oral cetirizine appear to be as effective as first-generation H1 antihistamines like diphenhydramine in symptom management but are less likely to cause sedation.21 Their longer duration of action requires less frequent dosing, which is also desirable for ease of administration. However, there is variability in their adoption as a first-line oral antihistamine as adjunctive therapy for anaphylaxis.8,22
Overall, as anaphylaxis becomes increasingly common in children, strategies are needed to reduce hospitalizations and improve uniformity of approach. Literature has shown the effectiveness of local evidence-based guidelines (EBGs) in safely reducing hospitalization rates and decreasing variability in management for patients with anaphylaxis.23–25 At our institution, we first embarked on improving anaphylaxis care in 2011 with a quality improvement (QI) intervention aimed at reducing hospitalization rates and improving discharge education and epinephrine prescription rates.24 This initiative included the creation of our original EBG, which was based on recommendations from the National Institute of Allergy and Infectious Diseases anaphylaxis guidelines26 and developed through an iterative process involving key stakeholders. The adoption of this EBG contributed to a decrease in hospitalization rate from 54% to 36%.24
Since the original QI intervention, the 2015 World Allergy Organization anaphylaxis guidelines and new evidence on the management of anaphylaxis have emerged.14,27–32 Recognizing the need to match our local management to updated guidelines, our QI group established the following SMART aims: through the development of a revised EBG, public promotion, and ongoing provider contact, by December 31, 2019, ED providers would (1) further reduce hospitalization rates safely for children with anaphylaxis to 10%, (2) increase the use of cetirizine as the first-line adjunctive antihistamine for children with anaphylaxis to 90%, and (3) decrease corticosteroid use in children with anaphylaxis by 50%.
Methods
Context
Boston Children’s Hospital (BCH) is an urban, tertiary, 415-bed children’s teaching hospital with an annual ED volume of ∼60 000 visits. On average, there are >225 visits to the BCH ED each year for anaphylaxis. The EBG program at BCH began in 2008. There has been strong leadership support of EBGs, and providers currently have access online to >40 ED-specific EBGs as well as order sets and other support features within the electronic medical record (EMR).
Intervention
To address changes and updates in anaphylaxis care, we assembled a multidisciplinary group in 2018, beginning with an iterative review process to overhaul and comprehensively update the EBG. This group included ED physicians, a pediatric allergist, ED nurses, QI specialists, and ED pharmacists. The EBG underwent 5 rounds of revisions by all stakeholders over a 3-month review period. These revisions used a variety of QI tools, including run charts of medication administration, histograms for ED length of stay (LOS), and control charts to detect changes in hospitalization rates over time. We concluded that the hospitalization criteria and medication recommendations would need to change and that maintaining these improvements would require ongoing monitoring and provider education from the QI team.
Ultimately, the revised EBG (Supplemental Fig 5) recommended that children meeting criteria for anaphylaxis with ongoing symptoms receive intramuscular epinephrine and oral cetirizine, and that those with wheezing also receive albuterol and dexamethasone. In contrast, the original EBG included diphenhydramine and corticosteroids as standard treatment. Whereas the original EBG recommended hospitalization for patients requiring 2 or more doses of epinephrine and for those with hypotension or wheezing at any time, the revised EBG recommended hospitalization for patients with hypotension at any time, persistent wheezing or other system involvement without improvement, and/or 3 or more doses of intramuscular epinephrine. This updated version was posted in the online EBG library on May 23, 2018, accompanied by revision to the anaphylaxis order set in the EMR and a presentation of the new EBG at a faculty meeting. The adoption of this EBG was complemented by ongoing efforts to improve its visibility and usage, including practice reminders, clinician education, individual provider performance audit with feedback, and annual minor guideline revisions. The anaphylaxis QI team met regularly to assess statistical process control (SPC) charts and histograms for outcomes of interest. They assessed for special cause variation, determined if measures should be retired, and reviewed individual cases. These meetings led to real-time EBG modifications and were accompanied by individualized audits and discussion with providers, including review of nonadherent cases to understand overarching causes of variation and allow for coaching and encouragement. Overall progress was communicated to the broader group by e-mail, as were additional reminders, advice, and teaching videos to further promote use and adherence.
Study of the Intervention
We included all patient encounters with children between ages 6 months and 18 years presenting to the BCH ED between January 1, 2014, and December 31, 2019 with International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) or International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) diagnosis codes for anaphylaxis (ICD-9-CM 995.0, 995.60-995.69; ICD-10-CM T78.00-T78.09, T78.2, T88.6). This start date was chosen to overlap with the baseline established following our original intervention. We excluded any visit with a diagnosis code of mast cell activation syndrome, mastocytosis, hereditary angioedema, or a transfusion/contrast reaction (ICD-9-CM 279.8, 202.6, 238.5, 277.6, 999.80, 999.89, 995.29; ICD-10-CM D89.4, Q82.2, C96.2, D47.0, D84.1, T80.92, T50.8X5A, E947.8). Patients with unknown outcomes were also excluded.
We queried our institution’s electronic health record database (Cerner Corporation, Kansas City, MO) to identify eligible encounters and collect all variables. For a comparative analysis between BCH and other children’s hospitals, data for this study were obtained from the Pediatric Health Information System (PHIS) database (Children’s Hospital Association, Overland Park, KS). The PHIS administrative database contains encounter-level data from >50 not-for-profit, tertiary care pediatric hospitals affiliated with the Children’s Hospital Association (Lenexa, KS). Participating hospitals provide discharge and encounter data including demographics, diagnoses, and procedures. Data are deidentified at the time of data submission and are subjected to reliability and validity checks before being included in the database. For this study, data from 48 hospitals were included, comprising all recorded visits for anaphylaxis during the study period.
Measures
The primary outcome measure was hospitalization rate. Secondary outcome measures were cetirizine use, overall corticosteroid use, and corticosteroid use among patients with and without asthma, the last of which was added 12 months after the revised EBG was implemented. The original revision had the unintended consequence of decreasing corticosteroid use among patients with asthma; the new measure was added when this impact was detected in a QI team meeting. The process measure was ED LOS, which was chosen for inclusion as a key safety measure that had undergone a change from the original EBG. Both EBGs stipulated a minimum 4-hour observation period from epinephrine administration, but the revised version included a minimum observation from time of symptom resolution if no treatment was provided in the ED. The balancing measure was ED revisits for anaphylaxis, defined as all unplanned return visits to the hospital within 72 hours after discharge from the ED for patients with anaphylaxis. Patient characteristics collected included age, sex, type of insurance, codiagnoses, and medications administered. Patients were determined to have a complex chronic condition (CCC) by using codes developed by Feudtner et al.33
Analysis
In this study, we compared the prerevision phase (January 1, 2014, to May 22, 2018) with the implementation phase (May 23, 2018, to December 31, 2019). Patient characteristics were stratified by phase and compared using a Fisher’s exact test. We compared each of our outcomes between the two phases using the Fisher’s exact test and a risk difference between phases. For analyses not employing SPC methodology, all statistical tests were two-tailed and P < .05 was considered to be statistically significant.
Hospitalization rate, return ED visits, antihistamine usage, and corticosteroid usage were all monitored over time by using SPC methodology.34 Initial centerlines were set at the mean of the prerevision phase. Eight consecutive points on one side of the existing mean line signified special cause variation and resulted in a center line shift, and the mean of those first 8 points was established as the new centerline.35 Median ED LOS was monitored by using a run chart. SPC charts and run charts were created by using SQCPack version 7.0 (PQ Systems, Dayton, OH).
To evaluate trends over time in hospitalization rates between BCH and other children’s hospitals, we performed an interrupted time series analysis using PHIS data. We estimated multivariate logistic regression models using hospitalization rate as the dependent variable and time, site (BCH or non-BCH hospitals), and a time-site interaction term as the independent variables. To standardize for local differences in observation stays, this analysis was limited to inpatient hospitalizations only. Two different models were created, one for each phase, and each model was stratified by BCH and non-BCH hospitals. Given that patients with CCCs may be at different risk of hospitalization than those without, a subanalysis was also conducted excluding such patients.
Data analysis was performed in R statistical software (R Foundation for Statistical Computing, Vienna, Austria). This research was considered exempt by the institutional review board. We adhered to SQUIRE 2.0 guidelines for QI reports.36
Results
Study Population
We included 1449 patient encounters, of which 2 were excluded because of underlying conditions (both with medication and transfusion reactions). We analyzed 1447 encounters (Table 1). Sex, age, insurance, proportion of patients with asthma, proportion of patients with CCC, and number of doses of epinephrine received did not discernibly differ by phase.
Demographic Characteristics of the Study Cohort
| Phase | P | |||
|---|---|---|---|---|
| Overall | Prerevision | Implementation | ||
| Total | 1447 | 1055 | 392 | — |
| Number male (%) | 792 (54.7) | 592 (56.1) | 200 (51.0) | .095 |
| Age, median (IQR) | 6.99 (2.59–13.3) | 7.11 (2.68–13.4) | 6.60 (2.35–13.0) | .20 |
| Insurance (%) | .59 | |||
| International | 21 (1.5) | 16 (1.5) | 5 (1.3) | — |
| Private | 977 (67.5) | 702 (66.5) | 275 (70.2) | — |
| Public | 443 (30.6) | 333 (31.6) | 110 (28.1) | — |
| Unknown | 6 (0.4) | 4 (0.4) | 2 (0.5) | — |
| Previous asthma diagnosis (%) | 239 (16.5) | 174 (16.5) | 65 (16.6) | 1.00 |
| CCC (%) | 39 (2.7) | 25 (2.4) | 14 (3.6) | .28 |
| Doses of epinephrine received (%) | .081 | |||
| 0 | 862 (59.6) | 613 (58.1) | 249 (63.5) | — |
| 1 | 549 (37.9) | 418 (39.6) | 131 (33.4) | — |
| 2 or more | 36 (2.5) | 24 (2.3) | 12 (3.1) | — |
| Phase | P | |||
|---|---|---|---|---|
| Overall | Prerevision | Implementation | ||
| Total | 1447 | 1055 | 392 | — |
| Number male (%) | 792 (54.7) | 592 (56.1) | 200 (51.0) | .095 |
| Age, median (IQR) | 6.99 (2.59–13.3) | 7.11 (2.68–13.4) | 6.60 (2.35–13.0) | .20 |
| Insurance (%) | .59 | |||
| International | 21 (1.5) | 16 (1.5) | 5 (1.3) | — |
| Private | 977 (67.5) | 702 (66.5) | 275 (70.2) | — |
| Public | 443 (30.6) | 333 (31.6) | 110 (28.1) | — |
| Unknown | 6 (0.4) | 4 (0.4) | 2 (0.5) | — |
| Previous asthma diagnosis (%) | 239 (16.5) | 174 (16.5) | 65 (16.6) | 1.00 |
| CCC (%) | 39 (2.7) | 25 (2.4) | 14 (3.6) | .28 |
| Doses of epinephrine received (%) | .081 | |||
| 0 | 862 (59.6) | 613 (58.1) | 249 (63.5) | — |
| 1 | 549 (37.9) | 418 (39.6) | 131 (33.4) | — |
| 2 or more | 36 (2.5) | 24 (2.3) | 12 (3.1) | — |
IQR, interquartile range; —, not applicable.
Primary Outcome Measure: Hospitalization Rate
In the prerevision phase, 28.5% of children were hospitalized compared with 11.2% in the implementation phase (risk difference: −17.3%, 95% confidence interval [CI]: −21.4% to −13.2%, P < .001). Special cause variation in hospitalization rate was detected immediately following the implementation of the revised EBG with a new center line established at 14.2% and sustained through the implementation phase (Fig 1).
P-chart of monthly anaphylaxis hospitalization rates, with key interventions shown. The date of EBG revision implementation was May 23, 2018. UCL, upper control limit.
P-chart of monthly anaphylaxis hospitalization rates, with key interventions shown. The date of EBG revision implementation was May 23, 2018. UCL, upper control limit.
Secondary Outcome Measures: Medication Use
In the implementation phase, the proportion of patients receiving cetirizine increased from 4.2% to 59.7% (P < .001, Fig 2A). Use of corticosteroids in patients without asthma decreased significantly, from 72.6% to 32.4% (P < .001, Fig 2B). There was a significant decrease in the proportion of patients with asthma receiving corticosteroids, from 74.1% to 53.8% (P = .005, Supplemental Fig 6).
P-charts of monthly medication usage rate for children presenting to the ED with anaphylaxis, with key interventions shown. The left panel (A) shows cetirizine use. The right panel (B) shows corticosteroid use among patients without diagnosis of asthma. LCL, lower control limit; UCL, upper control limit.
P-charts of monthly medication usage rate for children presenting to the ED with anaphylaxis, with key interventions shown. The left panel (A) shows cetirizine use. The right panel (B) shows corticosteroid use among patients without diagnosis of asthma. LCL, lower control limit; UCL, upper control limit.
Process Measure: LOS
Following implementation of the revised EBG, the proportion of encounters with LOS ≥240 minutes increased from 53.3% to 59.9% (P = .024, Supplemental Fig 8).
Balancing Measure: Unplanned Return to Hospital
We did not detect special cause variation in the balancing measure of revisits within 72 hours during the implementation phase (Fig 3). Overall 72-hour, unplanned revisit rates for patients discharged from the ED with a diagnosis of anaphylaxis were 4.2% before and 2.6% after revision (P = .23). Nevertheless, the g-chart illustrated that the frequency of unplanned revisits was decreasing, with only 9 total revisits during the implementation phase and an increase in cases between revisits.
G-chart of number of cases between ED revisits for patients with anaphylaxis who were discharged from the hospital from the ED. CL, center line; UCL, upper control limit.
G-chart of number of cases between ED revisits for patients with anaphylaxis who were discharged from the hospital from the ED. CL, center line; UCL, upper control limit.
Comparison with National Trends
The interrupted time series analysis demonstrated that nationally the hospitalization rate decreased over the whole study period (Fig 4). During the implementation phase, hospitalizations were stable. There were no significant differences in hospitalization rates before and after EBG implementation between BCH and other children’s hospitals nationally (Table 2). Similarly, there were no significant differences in trends in hospitalizations over time (Fig 4).
Hospitalization rates and 95% CIs over time at BCH (blue) and other PHIS hospitals (orange) for children with anaphylaxis. Hospitalization rate estimates were derived by using logistic regression separately for each study phase.
Hospitalization rates and 95% CIs over time at BCH (blue) and other PHIS hospitals (orange) for children with anaphylaxis. Hospitalization rate estimates were derived by using logistic regression separately for each study phase.
Estimated Hospitalization Rates Immediately Before and After Implementation of an EBG
| Patient Sample | Preimplementation, % (95% CI) | Postimplementation, % (95% CI) |
|---|---|---|
| All patients | ||
| National sample | 4.9 (4.5–5.4) | 4.9 (4.1–5.7) |
| BCH | 3.2 (1.8–5.4) | 2.1 (0.5–6.3) |
| Patients without a CCC | ||
| National sample | 4.3 (3.8–4.8) | 4.2 (3.5–5.1) |
| BCH | 2.3 (1.1–4.3) | 0.7 (0.0–3.8) |
| Patient Sample | Preimplementation, % (95% CI) | Postimplementation, % (95% CI) |
|---|---|---|
| All patients | ||
| National sample | 4.9 (4.5–5.4) | 4.9 (4.1–5.7) |
| BCH | 3.2 (1.8–5.4) | 2.1 (0.5–6.3) |
| Patients without a CCC | ||
| National sample | 4.3 (3.8–4.8) | 4.2 (3.5–5.1) |
| BCH | 2.3 (1.1–4.3) | 0.7 (0.0–3.8) |
BCH is compared with a national sample of other children’s hospitals from the PHIS.
Discussion
With the use of a major local EBG revision and sustained provider messaging and feedback, we safely reduced hospitalizations for anaphylaxis by 17.3 percentage points over 19 months to 11.2% without a concomitant increase in revisits within 72 hours. This decrease was close to, but ultimately short of, our aim of 10% hospitalization rate. On the basis of the immediate, downward, sustained shift in hospitalization rates, we believe that the EBG revision and subsequent efforts to maintain the practice contributed to this observed decrease. Over this same period, we increased the proportion of patients receiving cetirizine to 60%, which was short of our goal of 90%. The relative risk of receiving corticosteroids decreased by 55.4%, greater than our objective of 50%. The EBG mandated a 4-hour observation period, and the modestly increased proportion of patients with LOS ≥4 hours may indicate increased EBG adherence.
Health care spending on pediatric anaphylaxis has skyrocketed over the past decade.28 By safely increasing the number of patients discharged from the ED, we prevent unnecessary hospitalizations and help curtail these rising costs while also reducing the significant emotional and economic stress families experience with hospitalizations. Importantly, we did not observe an increase in 72-hour revisits to the ED, suggesting there was not an associated increase in clinically significant cases of biphasic reaction despite the decrease in corticosteroid use. On the contrary, the observed decrease in 72-hour revisits suggests improved control that may be enhanced by the accompanying education, follow-up, and medication provision recommended by the EBG. For instance, because of the longer half-life of cetirizine compared with diphenhydramine, patients may have been less symptomatic and thus returned to the ED less frequently. This finding is in line with what others have previously reported10,13 and further indicates that many current hospitalizations could be safely avoided. Comparison of these results to the PHIS data suggests that national hospitalization rates are trending downward and that BCH in general had lower hospitalization rates both before and after the QI initiative. The 11.2% hospitalization rate from this study is similar to levels reported in Canadian literature8 and markedly lower than previously reported national hospitalization rates.4,6,7
There are several reasons that this project was successful. First, there is widespread acceptance and use of EBGs at our institution. This culture has been cultivated over more than a decade and likely facilitates the success of this EBG as well as others. The combination of support from ED leadership and interdisciplinary participation in the creation and improvement of guidelines has created an environment that further fosters evidence-based medicine. The EBG itself is also accompanied by an EMR order set that facilitates and promotes increased adherence. Close outcome monitoring and announcements showing the initial success of the initiative likely generated additional buy-in for this specific revision. In addition, this measure was tracked on an individual basis for faculty with peer comparisons, perhaps inspiring improved adherence and outcomes. This combination of culture, strong communication, and careful outcome tracking creates an environment wherein EBGs can be quickly adopted and realize impacts rapidly.
Improving anaphylaxis care in the future will depend on research that will clarify the natural history of anaphylaxis and determine appropriate management. Epinephrine remains the mainstay of therapy, and new guidelines emphasize that corticosteroids have no role in acute management of anaphylaxis whereas H1 antihistamines continue to be used as second line agents in anaphylaxis for symptomatic treatment of cutaneous symptoms.31 Through our intervention, we decreased corticosteroid use and increased oral cetirizine use. This led to an unintended decrease in corticosteroid use among patients with asthma, resulting in a further EBG revision for this population. Tracking corticosteroids use among these patients going forward will be an important balancing measure. We fell short of our cetirizine use goal, which may have been due to some patients with milder presentations that providers did not think warranted treatment. Additionally, we were unable to measure whether patients received antihistamines before arrival that obviated the need for additional intervention. Studies demonstrating clearer parameters governing administration of these adjunct interventions could help further tailor management to individual patients.
We saw an initial and sustained decline in hospitalization rate yet fell short of our goal that had been informed by low hospitalization rates internationally.8 Part of this difference may be due to our specific patient population, but part may be due to imperfect use of the EBG. Qualitative review of provider feedback revealed that most deviations occur when clinicians deem hospitalization necessary despite more permissive EBG recommendations. In all such cases, no further treatments were needed after hospitalization. This likely represents an area for continued provider education. Future studies could benefit from tracking of additional process measures such as time to epinephrine administration to promote further guideline adherence and improved patient outcomes. Finally, the fact that most patients are able to be discharged from the ED suggests that some patients may be good candidates to have their anaphylaxis managed in the outpatient setting, which may be especially important in light of the ongoing coronavirus disease 2019 pandemic.37
There are some limitations to this study. Given the robust engagement of faculty in evidence-based care and the local QI context, this may not be generalizable to other settings. Some patients likely received medications before presentation, and we are unable to account for those medications. National data in this study were from PHIS, an administrative database that does not provide clinical detail. Because of this, we are unable to determine the severity of presentations accurately and cannot account for prehospital epinephrine doses, among other factors that may confound hospitalization rates. The PHIS analysis was further limited by sample size. Finally, observation stays are difficult to classify as either hospitalizations or not, and this may have biased our national analysis.
Conclusions
Through implementing a QI initiative centered on use of an updated local anaphylaxis EBG, we safely reduced hospitalization rates and corticosteroid usage without a concomitant increase in ED revisits. The use of local guidelines along with sustained messaging to clinicians can effectively reduce hospitalizations for anaphylaxis in children.
Drs Gaffney and Michelson conceptualized and designed the study, performed the data analysis, and drafted the initial manuscript; Mr Porter collected data and assisted with data analysis; Ms Gerling conducted data analysis; Drs Schneider and Shah helped design the intervention; Dr Stack helped design the intervention and assisted with data interpretation; and all authors revised the manuscript critically for important intellectual content, approved the final manuscript as submitted, and agree to be accountable for all aspects of the work.
FUNDING: Dr Michelson received funding through award 1K08HS026503 from the Agency for Healthcare Research and Quality. The other authors received no external funding. The funder provided research time for Dr Michelson to work on this project. Funded by the National Institutes of Health (NIH).
- BCH
Boston Children’s Hospital
- CCC
complex chronic condition
- CI
confidence interval
- EBG
evidence-based guideline
- ED
emergency department
- EMR
electronic medical record
- ICD-10
International Classification of Diseases, 10th Revision, Clinical Modification
- ICD-9
International Classification of Diseases, Ninth Revision, Clinical Modification
- LOS
length of stay
- PHIS
Pediatric Health Information System
- QI
quality improvement
- SPC
statistical process control
Comments