Discussion

Prior studies have identified high rates of antibiotic prescribing for nonbacterial upper respiratory tract infections with rates reaching up to 50% in the general population and up to around 70% among US veterans. ^{Reference Silverman8} Therefore, we recognized antibiotic prescribing for URIs as a crucial outpatient stewardship target for our center.

Our active intervention was multi-faceted and targeted multiple key elements that we identified as essential to decreasing prescribing rates. We provided baseline and distributed peer comparison reports to department leaders that helped each department identify prescribing patterns that could be improved, for which as-needed education was provided. Utilization of the EZ ID ambulatory order panel increased from an average of 1.5 uses per month in 2021 to 13.3 uses per month in 2022. These syndrome-specific order panels in the electronic medical record list appropriate non-antibiotic interventions for different respiratory diagnoses which may have helped reduce prescribing rates. Similarly, the viral prescription pad, which contains appropriate non-antibacterial therapies that patients can use for symptomatic relief may have helped reduce patient demand for antibiotics. ^{Reference Martinez9} The individual contribution of each of the measures cannot be determined, but when used in combination, these interventions significantly reduced the inappropriate prescribing of antibiotics for viral upper respiratory tract infections. Similar findings were observed at the Mayo Clinic Enterprise level. ^{Reference Ilges6}

Our study had limitations. First, we were unable to account for delayed (ie, watch and wait) prescribing or prescribing outside of a provider visit. Second, antibiotics prescribed in the outpatient setting within our system did not require an associated diagnosis, which limited the information we had about such prescriptions. Third, our Tier 3 diagnoses included a diagnosis of URI not otherwise specified; it is possible that antibiotics may have been appropriate for some of these encounters. Fourth, we are unable to completely account for the possibility of providers avoiding Tier 3 diagnoses to avoid detection, although there did not appear to be appreciable shift in diagnosis tiers (figure 3). Fifth, we excluded all COVID-19-related diagnoses as antimicrobial prescribing rate for ambulatory COVID-19 patients can be as low as 3%, and inclusion of these encounters could artificially reduce the URI prescribing rate. ^{Reference Stevens, Jensen, O’Horo and Shah10} Finally, the SlicerDicer model included only respiratory infection codes and antibiotics typically prescribed for respiratory infections, but other indications for antibiotics such as a urinary tract infection were not excluded. It is thus possible that some patients with a Tier 3 primary diagnosis who received an antibiotic prescription for a secondary diagnosis of another bacterial infection (eg, cellulitis, urinary tract infection) were incorrectly identified as inappropriate; however, a prior validation of the model demonstrated that this phenomenon occurred rarely. Moreover, the exclusion of other indications for antibiotics would have decreased the inappropriate antibiotic prescribing rate for Tier 3 respiratory infections further.

In conclusion, a multi-faceted intervention, which included baseline education, promotion of syndrome-specific order panels, dissemination of resources for symptomatic management, and distribution of quarterly peer comparison reports, resulted in significant reduction of inappropriate antibiotic prescribing for URIs.