Implementation of an Antibiotic Stewardship Initiative in a Large Urgent Care Network

Key Points Question What is the association between a multifaceted antibiotic stewardship initiative with antibiotic prescribing for respiratory conditions in a large urgent care (UC) network? Findings This quality improvement study of 493 724 total UC encounters found a decrease in antibiotic prescribing for respiratory conditions from 48% (baseline) to 33% (intervention), with reductions of 22% at the start of the intervention and 5% per month throughout the 1-year intervention period. Meaning This study’s findings suggest that a multifaceted antibiotic stewardship initiative was associated with reduced antibiotic prescribing for UC respiratory conditions, and that such initiatives in large UC networks may decrease inappropriate antibiotic prescribing.


Introduction
The majority of antibiotic prescriptions in the United States result from outpatient encounters; up to 30% may be unnecessary. 1,2Despite awareness of problems associated with antibiotic overuse (eg, antibiotic resistance and antibiotic-associated adverse effects), 3 there has been only modest improvement in the rate of unnecessary prescribing over time. 2 Urgent care (UC) is one of the fastest growing sites of outpatient care delivery in the United States, with the encounter volumes increasing by 50% or more in recent years. 4,5These UC encounters result in more antibiotic prescriptions overall and more unnecessary antibiotic prescriptions compared with other outpatient settings. 6Infectious conditions, especially respiratory tract infections, which often lead to inappropriate antibiotic prescribing, are the most common types of diagnoses managed in UC. 7 Recognition that unnecessary antibiotic prescribing is common in outpatient settings led the Centers for Disease Control and Prevention (CDC) to develop the Core Elements of Outpatient Antibiotic Stewardship. 8This guidance provides a framework for outpatient stewardship implementation.Most outpatient stewardship interventions have been designed and evaluated in primary care (PC) settings or emergency departments (EDs), with few interventions being applied specifically to UC. [9][10][11] Urgent care incorporates features common to PC settings (eg, uncomplicated low-acuity respiratory conditions) and EDs (eg, expanded hours, walk-in visits, rotating clinicians, and expectations for fast turnaround).Therefore, stewardship programs that best meet the needs of patients and clinicians in UC settings should understand and incorporate these characteristics.Substantial variability in rates of antibiotic prescribing for respiratory conditions exists between clinics and clinicians across the Intermountain Healthcare (IH) UC system. 7To improve prescribing, we developed an antibiotic stewardship program specifically designed for UC settings based on CDC Core Elements.The objective of the present study was to evaluate the association of this program with antibiotic use for respiratory conditions in UC encounters.

Setting and Study Design
Intermountain Health is a nonprofit, integrated, health care delivery system that includes 24 hospitals and more than 185 outpatient clinics throughout the Mountain West.During the quality improvement study, IH operated 38 UC clinics, including 32 for patients of all ages (InstaCare) and 6 providing care exclusively to children younger than 18 years (KidsCare).InstaCare and KidsCare clinics are predominantly staffed by physicians.In addition, IH operates a direct-to-consumer telemedicine UC clinic (Connect Care) primarily staffed by advanced practice clinicians (APCs).This study followed the Standards for Quality Improvement Reporting Excellence (SQUIRE) reporting guideline.The IH institutional review board approved this study and waived the need for obtaining informed consent for this minimal risk, evidence based, quality improvement study.The study is registered at ClinicalTrials.gov. 12pre-post quality improvement study design was used to evaluate the association of a systemwide antibiotic stewardship initiative with antibiotic prescribing for respiratory conditions in the IH UC network.The prespecified primary and secondary analyses compared a 12-month baseline period (July 1, 2018, through June 30, 2019) with a 12-month intervention period (July 1, 2019, through June 30, 2020).In addition, we included a 12-month sustainability period (July 1, 2020, through June 30, 2021) to evaluate subsequent prescribing trends.

Intervention
To inform the intervention, the study team (eAppendix in Supplement 1) performed site visits to assess the UC environment and to replicate the patient experience.Patient and clinician interviews were performed to identify barriers and possible solutions to improve antibiotic prescribing. 13

Financial Incentive
The IH UC service line has annual quality measures tied to compensation.UC clinicians given the goal to individually prescribe antibiotics in fewer than 50% of their respiratory encounters.This target was identified by a preintervention analysis that showed 50% to be the median respiratory antibiotic prescribing rate for the UC service line. 7breviations: EHR, electronic health record; IH, Intermountain Health; UC, urgent care.
a Not a formal component of the intervention.
introduced the antibiotic prescribing measure into their bundle of quality measures, 14 with an individual target of lower than 50% prescribing (Box).

Study Outcomes
Data were electronically extracted from the EHR for each encounter and included patient-reported demographic characteristics, clinician specialty (MD, DO, or APC), discharge International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes, and antibiotic prescriptions ordered.Each encounter was categorized into 1 of 5 clinical categories (respiratory, skin or skin structure, gastroenterology, genitourinary, and other) based on ICD-10-CM codes. 7The respiratory category was further subcategorized and assigned tiered groups based on whether antibiotics were indicated: tier 1, antibiotics indicated (eg, pneumonia); tier 2, antibiotics sometimes indicated (eg, sinusitis); and tier 3, antibiotics not indicated (eg, bronchitis). 1 If diagnosis codes from multiple tiers were present, the category defaulted to the lowest tier (eg, when tier 2 and tier 3 codes were present, the category was defined as tier 2).Encounters were excluded if no ICD-10-CM code was billed, if codes from multiple clinical categories were present (eg, respiratory and genitourinary), or if the clinician totaled fewer than 25 visits during the study period.
Antibiotic prescriptions generated or administered (eg, intravenous antibiotics) during UC encounters were captured as electronic orders, and topical or inhaled formulations were excluded.
Delayed antibiotic prescriptions were created in the EHR for antibiotics traditionally used for sinusitis and acute otitis media (AOM), allowing for direct measurement of their use.Delayed antibiotic prescriptions counted as an antibiotic prescription in all measures.
The primary study outcome was the change in antibiotic prescribing for respiratory conditions between the baseline and intervention periods.Secondary outcomes included the change in antibiotic prescribing for tier 3 respiratory conditions and the use of first-line antibiotics for AOM, sinusitis, and pharyngitis.First-line antibiotics were defined as penicillin or amoxicillin for pharyngitis and amoxicillin or amoxicillin-clavulanate for acute sinusitis and AOM.Additional outcomes included the percentage of respiratory conditions prescribed azithromycin and the use of delayed prescriptions for AOM and sinusitis.
Balancing measures related to the potential unintended consequences of the intervention included patient satisfaction and ED visits or hospitalizations within 14 days of a UC respiratory condition.Patient satisfaction was assessed using data from an IH patient satisfaction survey administered by Press Ganey. 15Data were extracted for the "Overall rating of care received during your visit" on a scale of 1 (very poor) to 5 (very good).Mean patient satisfaction scores for respiratory conditions were evaluated from January 1, 2019, through June 30, 2020, based on data availability.

Statistical Analysis
An interrupted time series model was used to assess the association of the intervention with the primary and secondary outcomes. 16Binomial generalized estimating equations were used to account for clustering among the clinicians and in the clinic.For the prespecified primary and secondary outcomes comparing the intervention with baseline period, regression models included time-based terms to estimate the slope per month for antibiotic prescribing in the baseline period, the change in prescribing at the start of the intervention, and the change from the period before the intervention to the intervention period.The same approach was used to compare the sustainability period with the intervention period.For each model, an odds ratio (OR) was estimated for antibiotic prescribing per 1 month change in the baseline period (the baseline period slope), the change in odds of antibiotic prescribing from the baseline period to intervention (the immediate change associated with the intervention initiation during July 2019), and the OR for antibiotic prescribing per 1 month change during the intervention (the intervention period slope).This OR was calculated using the linear combination of the terms for baseline slope and the change in slope over time from baseline to the intervention using estimate statements in PROC GENMOD.Because the intervention period overlapped with the onset of the COVID-19 pandemic (March 2020), we performed a sensitivity

JAMA Network Open | Infectious Diseases
Implementing an Antibiotic Stewardship Initiative in a Large Urgent Care Network analysis to exclude any association with study outcomes by limiting the intervention period to July 1, 2019, through February 28, 2020.All modeling was performed in SAS, version 9.4 (SAS Institute Inc).
Statistical significance was defined as a 2-sided P < .05 or 95% CIs excluding 1.
The intervention period showed an increase in the proportion of Connect Care visits (3.8% at baseline to 8.4% during the intervention) with a corresponding decrease in InstaCare and KidsCare visits.There was no change in the proportion of visits by MD or DO clinicians compared with APCs.
Clinical and demographic characteristics for the baseline and intervention periods are compared in Table 1.

Primary Outcome
Among respiratory conditions, antibiotic prescribing decreased from 47.8% during the baseline period to 33.3% in the intervention period (Table 1).Table 2 provides a summary of the interrupted time series analyses, and fitted models are shown in the Figure.Antibiotic prescribing did not change during the baseline period (OR, 0.99; 95% CI, 0.99-1.00;P = .11).During the initial month in which the intervention was implemented (July 2019), a 22% reduction in antibiotic prescribing (OR, 0.78; 95% CI, 0.71-0.86;P < .001) was observed.Antibiotic prescriptions continued to decrease by 5% each month during the course of the intervention (OR, 0.95; 95% CI, 0.94-0.96;P < .001).Similar decreases in antibiotic prescribing were observed across all clinic types (eg, InstaCare, KidsCare, and Connect Care).All clinics showed a decrease in antibiotic prescribing for respiratory conditions (range, −4.8% to −37.5%).Of clinicians who were present in both the baseline and intervention periods, 95% had a decrease in their individual antibiotic prescribing for respiratory conditions.
During the baseline period, 38.5% of clinicians had a prescribing rate higher than 50%, compared with 10.2% during the intervention period (ie, only approximately 10% of clinicians did not meet the quality measure for the financial incentive).In a sensitivity analysis excluding the onset of the COVID-19 pandemic from the intervention period, the monthly decrease in antibiotic prescribing was slightly higher (OR, 0.93; 95% CI, 0.93-0.94vs OR, 0.95; 95% CI, 0.94-0.96).

Secondary Outcomes
Among tier 3 respiratory encounters, antibiotic prescribing decreased from 18.7% during the baseline period to 7.5% in the intervention period (Table 1).Tier 3 antibiotic prescribing was already slightly decreasing during the baseline period (OR, 0.99; 95% CI, 0.97-1.00;P = .04);however, when the intervention was activated, a 47% reduction in antibiotic prescribing was observed for tier 3 encounters (OR, 0.53; 95% CI, 0.44-0.63;P < .001) in its initial month.Antibiotic prescriptions for tier 3 encounters continued to decrease by 4% each month during the course of the intervention (OR, 0.96; 95% CI, 0.94-0.98;P < .001)(Table 2 First-line antibiotic prescriptions for AOM, sinusitis, and pharyngitis increased from 70.7% during the baseline period to 74.5% during the intervention period (Table 1).First-line antibiotic prescriptions were stable during the baseline period with an OR of 1.01 (95% CI, 1.00-1.02;P = .10).
When the intervention was implemented, an 18% increase in first-line antibiotic prescribing was observed (OR, 1.18; 95% CI, 1.09-1.29;P < .001),but no further changes were observed over time (OR, 1.00; 95% CI, 0.99-1.01;P = .48)(Table 2;  the intervention period.Among sinusitis and AOM encounters, when an antibiotic was prescribed, a delayed prescription was used 24.8% of the time during the intervention period.

Balancing Measures
Among respiratory conditions, patient satisfaction changed minimally.During the baseline period, the mean response rating was 4.4 (of 2533 surveys) compared with 4.3 (of 2189 surveys) during the intervention period.For patients who received an antibiotic, the baseline period mean response rating was 4.5 compared with 4.4 during the intervention.For patients who did not receive an antibiotic, the baseline period mean response rating was 4.2 compared with 4.2 during the intervention.
Hospitalizations within 14 days of a UC encounter occurred in 0.4% of encounters during the baseline period compared with 0.5% of encounters during the intervention period.This small a First-line antibiotic prescriptions among sinusitis, otitis media, and pharyngitis encounters when an antibiotic was prescribed.First-line antibiotics were defined as penicillin or amoxicillin for pharyngitis and amoxicillin or amoxicillin-clavulanate for acute sinusitis and acute otitis media.Antibiotic allergies were not taken into consideration.
b Delayed prescription use among sinusitis and acute otitis media encounters when an antibiotic was prescribed.Delayed prescriptions were unable to be measured during the baseline period.
increase was observed among tier 1 encounters (0.3% for baseline vs 0.4% for intervention), in which antibiotic prescribing did not change.Encounters in EDs within 14 days of a UC encounter (with no subsequent hospitalization) occurred in 8.1% of encounters during the baseline period and 8.6% of encounters during the intervention period.This small increase was observed after tier 1 UC encounters (6.5% for baseline vs 7.2% for intervention) and tier 2 encounters (7.2% for baseline vs 7.7% for intervention).There was no change among tier 3 encounters (9.6% for baseline vs 9.6% for intervention).

Sustainability Period
The 1-year sustainability period included 391 608 UC encounters, of which 95 221 (24.3%) were respiratory conditions.Patient demographic characteristics were similar to the baseline and intervention periods.A greater proportion of patients visited Connect Care during the sustainability period (11.5%) compared with the intervention period (8.4%).Antibiotic prescribing for respiratory conditions was 25.5% in the sustainability period (vs 33.3% during intervention) (Table 1).
Interrupted time series modeling showed a continued decrease in antibiotic prescribing at the start of the sustainability period, with minimal change thereafter (Table 2; Figure).Minimal change was observed with tier 3 prescribing (6.2% for sustainability vs 7.5% for intervention); first-line antibiotic therapy for AOM, sinusitis, and pharyngitis (72.5% for sustainability vs 74.5% for intervention); delayed prescriptions for AOM or sinusitis (24.6% for sustainability vs 24.8% for intervention), and azithromycin prescriptions (0.8% for sustainability vs 1.2% for intervention) (Table 1 and Table 2;

Discussion
We developed an antibiotic stewardship program for UC in a large integrated health system.By the end of its initial 1-year implementation period, we observed an absolute reduction in antibiotic prescribing for respiratory conditions of 15%.Reductions in antibiotic prescribing were observed across all clinic sites and among nearly all clinicians.Improvements were also realized in antibiotic selection and prescribing for tier 3 respiratory conditions.Despite the substantial reduction in antibiotic prescribing, balance measures of ED visits and hospitalizations within 14 days of the UC encounter and patient satisfaction scores remained unchanged following the introduction of the intervention.
[17][18][19] Features of successful programs include the use of audit and feedback, clinician and patient education, EHR tools, and peer comparison or benchmarking.The interventions used in our program were integrated into the UC service line in a a First-line antibiotic prescriptions among sinusitis, otitis media, and pharyngitis encounters where an antibiotic was prescribed.First-line antibiotics were defined as penicillin or amoxicillin for pharyngitis and amoxicillin or amoxicillin-clavulanate for acute sinusitis and acute otitis media.Antibiotic allergies were not taken into consideration.
large health system. 20Although the antibiotic stewardship interventions themselves were novel for this health system, within this context, clinicians were familiar with several of the core components of the program, including regular updates to clinical guidelines that adhere to evidence-based practice, clinical decision support tools embedded in the EHR, and regular performance evaluations using standardized practice metrics.The intervention was implemented July 1, 2019.Plots show the change in antibiotic prescribing rates for all respiratory encounters (A), all tier 3 respiratory encounters (B), and first-line antibiotics for sinusitis, acute otitis media, and pharyngitis encounters when an antibiotic was prescribed (C).The circles indicate the observed percentage of encounters receiving an antibiotic per month; dark blue lines indicate the fitted interrupted time series model.
We found that antibiotic prescribing rates did not revert to baseline rates after the intervention period.Unlike inpatient stewardship programs, which generally require longitudinal budgeted financial support to sustain implementation, an ongoing challenge for outpatient stewardship is program durability after implementation. 18,21,22When outpatient stewardship interventions are incorporated into an integrated system already deeply engaged in quality initiatives, sustainability may be enhanced.
Although not an explicit feature of our intervention, UC leadership added achieving an antibiotic stewardship quality measure as part of clinician compensation, along with the other elements of our intervention.Financial incentives have been shown to have a modest positive effect on reducing antibiotic prescribing 23,24 and may have contributed to our findings.Notably, financial incentives have not been featured in multiple previous studies of outpatient stewardship interventions, nor are they included in the CDC Core Elements.
The last 3 months of the intervention period overlapped with the onset of the COVID-19 pandemic.This overlap was associated with substantial health care disruptions and a reduction in outpatient antibiotic prescriptions nationwide. 256][27] Changes in the case mix of patients seeking care in UC centers could potentially have influenced prescribing rates.However, a sensitivity analysis conducted in the present study excluding the period of the pandemic showed no difference in the primary outcome.
Our study evaluated a novel HEDIS (Healthcare Effectiveness Data and Information Set) metric (namely, antibiotic use for respiratory conditions) for antimicrobial stewardship in the outpatient setting that avoids confounding due to changes in coding practices, and it is now endorsed by the National Committee for Quality Assurance. 14In addition to reductions in the overall rate of antibiotic prescribing for respiratory conditions, several other areas of antibiotic prescribing improved, including the use of azithromycin and delayed prescriptions.Because these areas were explicitly targeted by components of our intervention, they likely represent changes in clinical practice and were sustained beyond the intervention period.
We found no differences in several balancing measures of patient safety and patient satisfaction when comparing the baseline and intervention periods.This is important because of the relatively large reduction in antibiotic use for respiratory conditions that occurred with the intervention.9][30][31] Assessing hospitalization rates after the implementation of an ambulatory antibiotic stewardship initiative is critical because small increases in hospitalizations could change the acceptability of these initiatives.

Limitations
Our study has limitations.It was not a randomized design, which limits interpretation of causality.Our study was performed in 1 large integrated health system, a single geographic region, and required substantial resources.These factors limit its generalizability to other UC settings.The multifaceted nature of the intervention did not allow for determination of each component's independent contribution to the measured outcomes.The findings regarding patient satisfaction were limited by the small sample size.Although our results suggest that there was no increase in severe disease as measured by ED visits and hospitalizations, because we did not prospectively enroll patients and assess symptoms over time, we cannot exclude the possibility that some illnesses not treated with antibiotics could have been prolonged.The intervention period was only 1 year; thus, its long-term sustainability and penetrance when confronted with changes in health care delivery demands are uncertain.

Conclusion
The findings of this quality improvement study of a multifaceted UC antibiotic stewardship initiative in a large integrated health care system indicated a significant reduction in antibiotic prescribing for respiratory conditions by the end of the study period that was sustained with no detectable changes in unintended consequences.This study provides a model for UC stewardship in a large integrated health care system.
Our interventions were developed using 4 categories based on the CDC Core Elements: education focused on UC clinicians and patients (eAppendix in Supplement 1); electronic health record (EHR; Cerner) tools to assist clinicians in ordering antibiotic prescriptions correctly and more efficiently documenting encounters; a transparent clinician benchmarking dashboard; and media targeting patients and clinicians.The Box provides a full description of the interventions.
Figure).The percentage of respiratory conditions where azithromycin was prescribed decreased from 5.1% during the baseline period to 1.2% during JAMA Network Open | Infectious Diseases Implementing an Antibiotic Stewardship Initiative in a Large Urgent Care Network JAMA Network Open.2023;6(5):e2313011.doi:10.1001/jamanetworkopen.2023.13011(

Figure . Fitted
Figure.Fitted Interrupted Time Series Models for Baseline, Intervention, and Sustainability Periods

Box. Components of Antibiotic Stewardship Intervention Deployed in the IH UC Network and Concomitant Independent Antibiotic Utilization Financial Incentive Intervention Domain: Education
If quality measures are met by clinicians, they are eligible to receive additional compensation.Occurring in parallel and independent of the development of the antibiotic stewardship interventions, IH UC leadership

Table 2 .
Summary of ITS Analysis Evaluating the Association of the Intervention With Antibiotic Prescribing in the Intermountain Urgent Care Network