Discussion

While 75% of encounters in which an antibiotic was prescribed were associated with a first-line antibiotic and 70% with an appropriate duration, only 53% of encounters were associated with a first-line antibiotic prescription for an appropriate duration. This indicates substantial opportunity to improve respiratory UC antibiotic prescribing beyond the decision to prescribe an antibiotic. Our analysis also revealed differences in first-line prescribing across age, gender, and race and differences in appropriate duration between gender, race, and ethnicity.

Unsurprisingly, a documented penicillin allergy impacted first-line antibiotic selection. In our study, penicillin allergy conferred the highest risk of receiving a non–first-line antibiotic. Collectively, patients with a documented penicillin allergy only accounted for 11.8% of encounters in which an antibiotic was prescribed, but more than 40% of non–first-line prescriptions. This suggests that delabeling penicillin allergies, many of which are inaccurate for a variety of reasons, ^{Reference Trubiano, Adkinson and Phillips23–Reference Inglis, Caughey, Smith and Shakib25} could have a large effect on improving outpatient prescribing as previously reported. ^{Reference Hersh, Shapiro, Zhang and Madaras-Kelly26} Interestingly, in our study 545 (8.5%) encounters with penicillin allergy received first-line therapy suggesting that some clinicians may have questioned the veracity of the documented allergy. Penicillin allergy also conferred a higher likelihood of an extended duration prescription despite guidelines suggesting similar durations of therapy for most alternative second-line agents recommended for penicillin-allergic patients. Further study is needed to understand this unexpected finding. One potential explanation could be that certain clinical characteristics differ between patients with and without allergy which in turn lead clinicians to treat penicillin-allergic patients with longer antibiotic courses. The differences we observed do not appear to be entirely related to allergy prevalence: penicillin allergy in non-White patients receiving an antibiotic was 8.3% compared to 12.0% of White patients. Prior studies have demonstrated higher rates of penicillin and beta-lactam allergy in adult and pediatric White patients. ^{Reference Taylor, Joerger and Li27,Reference Moskow, Cook, Champion-Lippmann, Amofah and Garcia28} The rate of penicillin allergy in our overall population was 11.6%, similar to that of the general population. ^{Reference Shenoy, Macy, Rowe and Blumenthal29,Reference Castells, Khan and Phillips30}

Older patients were less likely to receive a first-line agent or a first-line agent for an appropriate duration. While patients ≥65 years old represented approximately 5% of our study population, only 39.1% of encounters in this age category received a first-line antibiotic for an appropriate duration. ^{Reference Poole, Shapiro, Fleming-Dutra, Hicks, Hersh and Kronman31} Despite elderly patients being a small proportion of our population, these findings highlight the opportunity to further improve antibiotic prescribing in these patients. Our findings for pediatric patients, supported by age and clinic-type data, are consistent with existing literature. ^{Reference Poole, Shapiro, Fleming-Dutra, Hicks, Hersh and Kronman31} Gender and race were predictors for first-line antibiotic selection along with appropriate duration prescriptions, and ethnicity was also associated with appropriate prescription duration. There may be differences in patient expectations or demands not captured by our study contributing to these findings.

Some studies suggest higher prescribing rates by APCs ^{Reference Schmidt, Spencer and Davidson32} while others demonstrate similar prescribing rates for MD/DO and APC providers. ^{Reference Knobloch, Musuuza, Baubie, Saban, Suda and Safdar33,Reference Sanchez, Hersh, Shapiro, Cawley and Hicks34} We identified similar practices for both clinician types. While it was not a primary or secondary objective of our study, we noted similar interprovider variability for measured outcomes among both provider types (Supplementary Material, Figure 1). This finding could reflect the IH UC system involving all clinicians in education, tracking, and an organizational commitment to stewardship efforts. Both MD/DO and APC providers participate equally in education and tracking efforts and receive the same academic detailing communications. Recent data suggest shorter courses of therapy are sufficient for sinusitis and AOM in certain situations, ^{Reference King, Sanchez, Bartoces, Hicks and Fleming-Dutra10,Reference Kozyrskyj, Klassen, Moffatt and Harvey35} and our findings emphasize the opportunity to improve the duration of antibiotic prescriptions among all clinicians.

Overall, the rate of first-line prescribing remained less than 80%. 80% has been proposed as the target rate for first-line antibiotic prescribing, providing some margin for when non–first-line therapy may be appropriate due to first-line failure and penicillin allergy. ^{36–Reference Capra, Lieu, Black, Shinefield, Martin and Klein38} Despite institutional guidelines suggesting durations of 5–7 days for sinusitis and AOM, encounters for penicillin-allergic patients were associated with a longer-duration prescription. These findings provide an opportunity to emphasize and reinforce first-line selection while also providing education regarding the appropriate duration of antibiotic therapy.

While some of our study encompassed the COVID-19 pandemic, we did not observe an impact on antibiotic selection or prescription duration. Despite an increase in azithromycin use nationwide, we noted a decrease in azithromycin use from 4.5% in July 2019 to 1.2% in June 2020 and during March 2020–June 2020, azithromycin prescribing rates for respiratory UC encounters in our system remained less than 3%. ^{Reference Stenehjem, Wallin and Willis21} Our study has limitations. First, we relied solely on ICD10 codes and did not assess for concordance between coding and the clinical diagnosis despite using a validated methodology. Second, we defined the appropriate duration for AOM in patients older than 2 years of age as 7 days or less while our institutional guideline recommends 10 days of therapy for severe AOM. However, our data extraction methods were unable to account for severity which could have led to an underestimation of appropriate duration prescriptions. Similarly, for our analysis, we defined the appropriate duration for sinusitis as 7 days when our pediatric guideline recommended 10 days, though efforts are underway to update this guideline to recommend a shorter duration. The rates of sinusitis in the pediatric population were low (7.7%), but this also may have underestimated the rate of appropriate duration prescriptions. Patient-preferred language, geography, and socioeconomic variables not assessed could have created a differential impact on access to care or expression of expectations and thus affected antibiotic selection and prescription duration as well. We did not examine what percentage of non–first-line therapy was due to failure or adverse events on first-line antibiotics. However, 1,353 (2.5%) of all encounters with an antibiotic occurred within 30 days of a previous encounter with an antibiotic prescription. Of these, 528 (39.1%) received a first-line antibiotic. For these second antibiotic-associated encounters, we were unable to capture if the prescription associated with the initial encounter was filled, or how much of the initial prescription was consumed. While we did not exclude encounters occurring within 30 days of an initial encounter, the first-line prescribing rate for encounters where there was not a subsequent antibiotic-associated encounter within 30 days was 75.4%, similar to our findings when examining all antibiotic-associated encounters. Lastly, we did not examine the appropriateness of antibiotic initiation and it is possible a proportion of prescriptions were not indicated.

Even in large vertically integrated healthcare systems with robust stewardship programs and amidst sustainable interventions that reduce respiratory antibiotic prescribing, ^{Reference Stenehjem, Wallin and Willis21} there exist significant opportunities to improve prescribing practices beyond the Antibiotic Use for Respiratory Conditions Healthcare Effectiveness Data and Information Set (HEDIS) metric. ³⁹ Our findings emphasize the importance of delabeling antibiotic allergies when feasible and appropriate. Healthcare systems should also consider strategies such as prospective audit and feedback and peer comparison dashboards to improve the duration of antibiotic prescriptions given the success of these approaches in other areas of outpatient stewardship. ^{Reference Doernberg, Abbo and Burdette40} First-line antibiotic prescriptions for an appropriate duration could also be a useful measure to evaluate antibiotic use for conditions where antibiotics may or may not be indicated while also being actionable and relevant to clinicians. Optimizing prescriptions for first-line antibiotics should not be considered separate from optimizing the duration of therapy: both are key components of antibiotic stewardship and a dual focus on antibiotic selection and duration should be considered as antibiotic stewardship interventions and metrics are further developed.