Multidisciplinary Initiative to Increase Guideline-concordant Antibiotic Prescription at Discharge for Hospitalized Children with Uncomplicated Community-acquired Pneumonia

Abstract

Background:

Clinical practice guidelines recommend using narrow-spectrum antibiotics to treat uncomplicated pneumonia in children. This quality improvement (QI) project aimed to evaluate if QI methods could improve guideline-concordant antibiotic prescribing at hospital discharge for children with uncomplicated pneumonia.

Methods:

For this single-center QI project, we implemented QI interventions in serial plan-do-study-act cycles, focusing on the key drivers targeting general pediatric inpatient resident teams. Interventions included: (1)Small bimonthly group didactic sessions, (2)Visual job aids posted in resident work areas, and (3) A noon conference session. Balancing measures included postdischarge emergency room visits, readmission and adverse drug reactions.

Results:

To establish a baseline rate, we conducted a chart review of 112 children diagnosed with uncomplicated community-acquired pneumonia during hospitalization from July 2017 through January 2019. The average monthly percentage of children discharged with guideline-concordant antibiotics was 67%. The intervention period was from February 2019 through February 2020, with 118 children meeting the criteria after a review of 262 charts. After our interventions, the average monthly percentage of children discharged with guideline-concordant antibiotics increased to 87%, with the increase persisting for at least 12 months. There were no significant differences in balancing measures pre- and post-interventions.

Conclusions:

Our QI initiative sustained increased rates of uncomplicated community-acquired pneumonia guideline-concordant antibiotic prescribing at discharge over 12 months without an increase in balancing measures. The enduring changes in prescribing behavior suggest a lasting impact of our interventions.

INTRODUCTION

Problem Description

Before the initiation of this work, we observed that the baseline prescribing practices for discharge oral antibiotics for children with community-acquired pneumonia (CAP) at our institution were not concordant with national guidelines from the Pediatric Infectious Diseases Society (PIDS) and Infectious Diseases Society of America (IDSA). For example, the third-generation cephalosporin, cefdinir, was often prescribed at discharge. The antimicrobial spectrum of cefdinir is unnecessarily broad for most cases of CAP, and, more importantly, most oral cephalosporins have suboptimal pharmacokinetics for treating pneumonia compared with amoxicillin.¹ We were concerned about the frequent use of cefdinir at the time of discharge, the lack of guideline-concordant prescribing, and the potential effects of this practice on patient outcomes.

Available Knowledge

CAP is one of the most common diagnoses leading to pediatric hospital admission² and one of the leading causes of pediatric death worldwide.³ Pneumonia, a lower respiratory tract infection with evidence of parenchymal involvement by physical exam or radiography, may be caused by viral or bacterial processes.⁴ Difficulties determining the specific infectious etiology may factor into the large variability in antibiotic prescribing for CAP.^2,4

In 2011, the PIDS and the IDSA published evidence-based clinical practice guidelines (CPG) for managing CAP in children 3 months or older. The guidelines focused on using narrow-spectrum agents, such as amoxicillin, for empiric therapy of uncomplicated CAP.⁵ Antibiotic stewardship in the treatment of CAP was also prioritized by the American Academy of Pediatrics (AAP) as the number one consideration in the 2018 American Board of Internal Medicine Foundation’s Choosing Wisely campaign to promote the judicious use of antibiotics in respiratory illnesses and help combat antimicrobial resistance.⁶

After the release of the joint PIDS/IDSA guidelines, there is evidence of improvement in the quality of care provided for children with CAP, including an increase in appropriate, guideline-concordant antibiotic therapy and a reduction in costly, unnecessary testing.^7,8 However, significant variations in antibiotic stewardship remain, with broad-spectrum antibiotics often chosen over narrow-spectrum agents for hospitalized patients despite evidence that this does not improve clinical outcomes.^9–12

In our clinical roles, we encountered variability in prescribing and deviations from guidelines, in particular, frequent use of cefdinir as a step-down antibiotic for uncomplicated CAP. Our experiences informed this project.

Rationale

To address deviations from evidence-based care and concordance with published national treatment guidelines, we employed educational interventions and visual job aids targeted to the pediatric resident physicians in our institution. The pediatric resident physicians care for all general pediatric patients on hospital medicine services. Didactics are a recognized tool of antibiotic stewardship.^13,14 When combined with other interventions, including low-cost visual aids,^15–18 didactic teaching can result in positive and long-lasting change.^19–21

Specific aims

The primary aim of this quality improvement (QI) project was to increase the percentage of guideline-concordant antibiotic prescribing for children discharged from the Pediatric Hospital Medicine (PHM) service after inpatient treatment for uncomplicated CAP by 20% from the baseline of 67% and to sustain this increase for at least 12 months. Balancing measures included hospital readmissions and return ED visits within 30 days of hospital discharge, Infectious Diseases consultation, and documented adverse drug reactions.

METHODS

Context

This QI project focused on children hospitalized on the general PHM teams at a free-standing, tertiary care children’s hospital in the mid-Atlantic region from July 2017 through February 2020. The five PHM teams care for more than 15,000 pediatric patients yearly. These care teams can include medical students, pediatric residents, supervising PHM fellows, and attendings. This hospital’s large pediatric residency program trains about 40 resident physicians annually.

The authors were PHM and Infectious Diseases fellows at the time of project conception and conducted this study as a multidisciplinary QI project. Multiple publications describe QI projects focused on antibiotic use in pediatric CAP.^22–28 However, no published QI projects before this work focused on pediatricians’ antibiotic prescribing at hospital discharge. Our institutional electronic medical record (or “EMR”) (PowerChart, Cerner, Kansas City, Mo.) had an admission order set for patients with presumed CAP that included guideline-concordant intravenous and oral antibiotics with dosing recommendations for inpatient care but did not include discharge antibiotic guidance or prescribing capabilities.

Interventions

Upon initiating this work, our multidisciplinary team met to plan this project. The authors reviewed current data to establish a baseline and confirm the anecdotal sense of prescribing behavior. The authors identified factors contributing to the problem of nonguideline concordant prescribing (Fig. 1) and mapped interventions to the outcome of interest with a key driver diagram (Fig. 2). In developing the interventions, we addressed the feasibility of the team members conducting the planned educational interventions. With four project team members, providing two 30-minute small group teaching sessions per month was feasible. The author group would co-lead one larger didactic teaching session. The visual job aids were jointly developed and placed in designated areas described below. Our interventions are further described below as teaching sessions or visual job aids.

Fig. 1.

Fishbone diagram.

Fig. 2.

Key driver diagram for the QI project.

Teaching Sessions

The first intervention implemented (“Intervention #1” in Fig. 3) was a series of short (30 minutes), small-group teaching sessions delivered to groups of 8–10 pediatric residents assigned to the PHM teams. The teaching sessions occurred during designated blocks of time set aside by the residency program for morning teaching on the hospital medicine services. These interactive, case-based sessions focused on the AAP and IDSA diagnosis and management guidelines for uncomplicated CAP. The author group developed these sessions to cover the same learning objectives, including common etiologies, diagnosis, and treatment of CAP. The teaching sessions, led by a PHM or ID fellow from the author group, occurred approximately twice a month, starting the last Friday in January 2019 through early June 2019, resulting in 10 sessions.

Fig. 3.

Monthly guideline concordance p-chart, July 2017–February 2020. This control chart (p-chart) demonstrates special cause variation with >6 points above the baseline center line (CLB) following our QI interventions (yellow arrows #1, 2 and 3). We calculated a new center line following the demonstration of special cause variation in our interventional period, as denoted by subscript ClI. The new center line met our goal for guideline concordance, 87%, as denoted by the light green dotted line.

We taught a larger group didactic session (“Intervention #3” in Fig. 3) at a pediatric resident noon conference in March 2019. Approximately 30 pediatric residents attended this conference. During the 60-minute session, the author group focused on the key features of diagnosis, pathology, etiology, and treatment of CAP, emphasizing AAP and IDSA guidelines.

Visual job aids

We displayed visual job aids leveraging pop-culture references (“Intervention #2” in Fig. 3) in the three main pediatric resident workrooms in February 2019. The author team designed the job aids with eye-catching graphics that emphasized three predetermined high-yield teaching points:

1. Amoxicillin is the first-line option for step-down therapy at discharge.

2. Cefdinir does not have appropriate pharmacokinetics for use against pneumonia.

3. After investigation, most penicillin allergies are not true allergies.

The job aids are not published here as we did not pursue permission to publish the specific copyrighted images we used. The authors periodically confirmed that the job aids remained posted through the end of the study period (February 2020).

Study of the Interventions

This project follows the SQUIRE 2.0 reporting guidelines.²⁹ To understand the impact of our interventions on our outcome, we applied the QI methodology of plan-do-study-act. We evaluated data using standard control charts (Figs. 3 and 5).

Fig. 5.

Monthly proportion of CAP cases with ED visit within 30 days of discharge, July 2017–February 2020. This P chart shows the monthly proportion of CAP patients who had an ED visit within 30 days of discharge from their CAP hospitalization (blue dots). There was no increase in 30-day ED visits or any evidence of special cause variation following our QI interventions.

Measures

Our primary outcome measure was the percentage of children hospitalized for uncomplicated CAP discharged with a guideline-concordant antibiotic prescription, primarily defined as “high dose” amoxicillin (45 mg/kg per dose, twice daily).⁵ Other antibiotics considered appropriate were cefpodoxime, cefuroxime, cefprozil, levofloxacin, and clindamycin, with the latter two only if a patient had a documented penicillin and/or cephalosporin allergy.⁵ These antibiotics provide appropriate coverage for Streptococcus pneumoniae, which is the most prominent invasive bacterial pathogen implicated in pediatric CAP.⁵ We considered macrolides concordant if the patient was diagnosed with atypical bacterial pneumonia. While the IDSA guidelines do not specifically recommend a choice of oral antibiotics for hospital discharge, they do make recommendations for “step-down therapy.” As the transition from intravenous to oral antibiotics (“step-down”) is a marker for discharge readiness in pediatric patients, we considered the aforementioned oral antibiotics as guideline-concordant antibiotics prescribed at discharge.

We included children >3 months of age with International Classification of Diseases, Tenth Revision (ICD-10) diagnostic codes reported at hospital discharge and clinical and radiographic criteria consistent with uncomplicated CAP. If discrepancies arose during chart review, all authors reviewed the data and reached a consensus. The diagnostic criteria for CAP were similar to those used in previous studies of uncomplicated pediatric CAP.^9,30,31

We excluded patients if an alternative diagnosis was present at discharge, if they had viral pneumonia (diagnosed clinically by the primary inpatient team and managed without antibiotics) or complicated pneumonia, defined as any size pleural effusion, empyema, or need for chest tube placement.^5,32 Additional exclusion criteria included the following features that would be consistent with a more complex illness or an underlying medical history that might necessitate a deviation from standard discharge antibiotic therapy recommendations: history of chronic lung disease, history of congenital heart disease, immunocompromised state, presence of indwelling central access device, children with baseline respiratory support (home mechanical ventilation or tracheostomy or other oxygen delivery), or children identified as medically complex (5 or more home medications, cerebral palsy, chromosomal anomalies). These exclusions are similar to those found in guidance documents and previously published CAP studies.^5,9,30–32 Our study differs as we elected to include multifocal/multilobar pneumonia cases without effusion in our uncomplicated CAP cohort.

To assess the unintended negative impact on patient outcomes following the implementation of our interventions, the authors collected data on emergency department (ED) visits, readmissions and adverse medication reactions within 30 days of discharge. Additionally, we evaluated evidence of increased consultation with the Infectious Diseases service as a result of implementing our interventions.

Analysis

We collected and analyzed data after each intervention and monthly. In addition, we collected de-identified demographic data, preexisting medical conditions and allergies, clinical signs and symptoms, microbiological and other laboratory data, inpatient and discharge antibiotic regimens (agent, duration, and route), and outcomes data for all eligible patients via chart review of the EMR. We entered all data into an Excel database (Microsoft Office 365).

We used standard QI metrics (control chart/P-chart) (QI Macros for Excel) to determine special cause variation in discharge guideline-concordant prescribing rates following the interventions. In addition, Fisher’s Exact Test was used to evaluate differences in the overall rate of guideline concordance, frequency of specific antibiotics prescribed at discharge, and categorical balancing measures between the baseline and postinterventional period.

Ethical considerations

We identified no specific ethical concerns during this project’s planning, conduction, or completion. Therefore, the institutional review board did not consider this QI project human subjects research requiring review and approval.

RESULTS

During the baseline period, from July 2017 through January 2019, we reviewed the charts of 225 children hospitalized with CAP and 112 met the inclusion criteria. The baseline average monthly percentage of children with uncomplicated CAP discharged with guideline-concordant antibiotics from July 2017 through January 2019 was 67%.

The intervention data collection period was from February 2019 to February 2020. There were 262 children admitted with CAP during the interventional period, and 118 children met the inclusion criteria (the postintervention cohort). The top reasons for exclusion were patients with medical complexity and complicated or viral pneumonia diagnoses in baseline and postintervention cohorts (Fig. 4). The frequency of guideline-concordant antibiotic prescribing at the time of discharge was 89%, 83%, and 70% in the 30-day periods following interventions #1 (small group session), #2 (visual reminder posting), and #3 (noon conference), respectively. We noted special cause variation demonstrated on the statistical process control chart (Fig. 3) due to our QI interventions, with six consecutive data points above the centerline (baseline mean) after implementing our interventions. Furthermore, there was a sustained statistically significant increase in the mean monthly percentage of uncomplicated CAP patients discharged with guideline-concordant antibiotics in the 12-month postintervention period (87.2% versus 66.8%, P < 0.05).

Fig. 4.

Study patient population flow chart.

The most common “nonguideline concordant” antibiotics prescribed at discharge in the baseline and interventional period were amoxicillin-clavulanate (n = 20, baseline; n = 12, postintervention) and cefdinir (n = 12, baseline; n = 2, postintervention). The antibiotics prescribed at discharge are listed in Table 1. There was a statistically significant increase in amoxicillin prescriptions (P = 0.0116) and a decrease in cefdinir prescriptions at discharge for the postintervention group (P = 0.0048) compared with baseline.

Table 1.

Antibiotics Prescribed at Discharge, by Frequency, for Baseline and PostIntervention Cohorts

	Baseline	PostIntervention	Fisher Exact Test P value
Amoxicillin	65	88	0.0116
Amoxicillin-clavulanate	20	12	NS
Cefdinir	12	2	0.0048
Azithromycin	3	4	NS
Clindamycin	2	3	NS
Levofloxacin	1	1	NS
Cefpodoxime	1	1	NS
Amoxicillin and Azithromycin	5	5	NS
Other combination	3	2	NS
Total	112	118

When comparing balancing measures for patients hospitalized before the initiation of this QI project and after the implementation of our three interventions, there was no new special cause variation or significant increase in the percentage of subjects with 30-day posthospitalization ED visits (16 [14.6%] baseline overall versus 8 [6.6%] postintervention overall) (Fig. 5), hospitalization readmission (5 [4.4%] versus 8 [6.6%], P = 0.5), Infectious Diseases consults (4 [3.5%] versus 2 [1.6%], P = 0.43), or the number of adverse reactions to antibiotics (1 [0.9%] versus 1 [0.9%] P = 1).

DISCUSSION

Summary

Our multidisciplinary QI project demonstrated that educational interventions targeted towards pediatric residents focusing on adherence to national guidelines for treating children with CAP resulted in an increased concordance with antibiotic prescribing recommendations at hospital discharge from 67% to an average of 87%, which was sustainable. One of the greatest strengths of our QI project was that the interventions were simple, practically free, and trainee-driven, yet led to a significant change in practice among pediatric residents at our institution.

Interpretation

Similar to previous studies, we demonstrated that QI methods can improve the frequency of use of narrow-spectrum antibiotics for children with CAP on an institutional level. Since the initiation of our project, QI methods have also been used to address antibiotic prescribing for common pediatric infections on a national level through the BASiC (Better Antibiotic Selection in Children) collaborative,³³ of which our institution did not participate. These studies were conducted in various practice settings, such as ED, inpatient and outpatient, but did not specifically focus on the transition of care from inpatient to home.^22–24 Our study adds to the literature by demonstrating that hospital medicine teams can improve guideline-concordant antibiotic prescribing at discharge for children with uncomplicated CAP.

Prior studies utilized similar interventions, including institution-specific CAP CPG or pathways, EMR order sets, visual job aids, and education.^22–24 Our institution has a CPG and EMR order set for CAP that pre-dated this study. Still, prior work from our ED colleagues demonstrated that the existence and knowledge of a CPG did not necessarily promote adherence to national IDSA guidelines for treating CAP.²⁵ Other institutions have performed QI projects to improve concordance with national society guidelines in patient care.^26,34 In fact, the Joint Commission recently recommended appropriate use and concordance of care with CPGs as a new core element of measurement of the success of antimicrobial stewardship programs.³⁵ We sought to complement the preexisting CPG and order set by including education and job aid interventions.

Educational interventions, especially interactive sessions, have positively impacted the judicious prescription of antimicrobials.^{13–15,19,21} The bimonthly repetition of our sessions over 6 months allowed the team members to interact directly with interns and senior residents as they rotated through their inpatient hospital medicine rotations. The residency noon conference also allowed us to reach a larger audience, including trainees we may have missed during our smaller group sessions. One potential problem of this type of intervention is a return to prior prescribing habits after the discontinuation of the primary educational intervention.³⁶ The combination of small group interaction and the involvement of a larger number of resident trainees likely facilitated the uptake of concepts. It helped with the longevity of our impact as these trainees progress and teach their peers.

As successful QI initiatives tend to use didactic instruction in conjunction with other modalities, such as visual workroom reminders,¹⁵ peer coaching,³⁷ or case-based sessions,²¹ we deployed our visual job aids as our second intervention. There was already a trend toward guideline concordance after the initiation of our small group sessions, and the appeal and catchy nicknaming of our visual aids (colloquially referred to as the “Yoda-Batman project” by the residents due to the choice of memes used for two of our visual aids) likely aided the sustainability of our efforts. The visual aids and educational sessions reinforced concepts and acted as social influences to normalize and promote a sustained change in prescribing habits.³⁸

In planning our educational interventions, we identified a likely knowledge gap in trainees about the suboptimal pharmacokinetics of oral cefdinir compared with other medications, specifically amoxicillin, in treating pulmonary infections.⁵ Although this QI initiative aimed to improve guideline-concordant discharge prescribing practices with increased use of narrow-spectrum antibiotics, we had a specific visual job aid dedicated to discouraging cefdinir’s use to treat CAP. These efforts were effective, resulting in a substantial decrease in patients discharged with oral cefdinir after hospitalization for uncomplicated CAP and increased amoxicillin prescriptions. Another job aid prompted trainees to question a reported history of amoxicillin allergy. While we hope it helped to educate trainees about the importance of thorough allergy investigation, we did not specifically monitor for changes in allergy reporting.

Importantly, our emphasis on using narrow-spectrum antibiotics for treating CAP did not significantly increase postdischarge ED visits or hospital readmissions within 30 days of discharge. This work used return-to-care or readmission as a potential antibiotic treatment failure marker. For the patients who did return to the ED, they were younger (all were less than 2 years of age), and most had continued respiratory symptoms, leading to the diagnosis of viral lower respiratory tract infections or bronchiolitis, thus suggesting that their initial hospital diagnosis of bacterial CAP may have been incorrect and that they unnecessarily received antibiotics. Similar to published reports, our work demonstrates that narrow-spectrum enteral antibiotics, such as amoxicillin, continue to be the appropriate discharge therapy for children hospitalized for CAP, with few possible failures requiring readmission.^9,27,28,31

Limitations

This project was a QI initiative at a single institution with a modest sample size, which can limit the power of our results. In addition, the sample population only included uncomplicated bacterial CAP in otherwise healthy pediatric patients, limiting the generalizability to the pediatric population. Complicated pneumonia and the treatment of pneumonia in medically complex, critically ill, or immunocompromised children were beyond our scope. Another limitation of this work was the required timeline for project completion. We focused on the antibiotic spectrum rather than duration to set achievable targets and meet deadlines during the fellowship. The latter is a target for future interventions, particularly in light of recent publications regarding the efficacy of shorter antibiotic courses.^30,31,39 We only had 1 month between initiating each of the three interventions. While we can comment on the overall effectiveness, it is difficult to determine which of the cycles had the greatest impact. As previously mentioned, educational interventions are vulnerable to deterioration of desired behavior change, especially as a new cohort of residents enters training each year. An additional limitation was that our educational interventions were only targeted towards trainees. Although some degree of deviation due to natural variation is expected, there was a notable period of decreased guideline concordance from October to December 2019 (Fig. 3). Half of the cases in this period were nonconcordant due to prescribing an inappropriately low dose of amoxicillin (<90 mg/kg/day) on discharge, prescribed by attending physicians rather than trainees. We also acknowledge that EMR-based decision support tools to automate antibiotic prescribing work synergistically with educational interventions to effect change.¹⁵ Our EMR system cannot create order sets for discharge medications.

Conclusions

Our QI initiative effectively created a sustained change in antibiotic prescribing behaviors. With a series of didactic interventions and visual job aids, we increased guideline-concordant antibiotic prescribing at hospital discharge for pediatric patients diagnosed with CAP. The simplicity of our approach lends to adaptation to other academic centers with large numbers of trainees, making a multicenter effort appealing. Modifying this strategy for community hospitals, either without pediatric trainees or with the presence of advanced practice providers, is another potential expansion area.^40,41 The educational interventions and job aids are generalizable, with the greatest need being time for faculty and trainees to be available for teaching sessions.

Future interventions could include hardwiring the guideline recommendations into our institutional pathways and EMR so there is less reliance on posting visual job aids. Other future interventions can target attending physicians to address knowledge deficits and provide peer benchmarking for feedback about antibiotic stewardship.^42,43 With recent data regarding the noninferiority of shorter antibiotic treatment courses,^30,31,39 future interventions could also target treatment duration.

ACKNOWLEDGMENT

The authors would like to thank Dr. Nada Harik, who guided this project.