DISCUSSION

This quality improvement study revealed that guideline-discordant antibiotic prescribing in the outpatient setting was common. The overall rate of guideline discordance was 60%. Largely, guideline-discordant therapy was a result of prescribing antibiotics when an antibiotic was not indicated.

Our results are comparable to other studies evaluating appropriate antibiotic prescribing for URIs.Reference Schroeck, Ruh, Sellick, Ott, Mattappallil and Mergenhagen ^{3 ,} Reference Fleming-Dutra, Hersh and Shapiro ⁴ Schroek et alReference Schroeck, Ruh, Sellick, Ott, Mattappallil and Mergenhagen ³ evaluated antibiotic use for respiratory tract infections within a single VA healthcare system and found that 64.2% of patients received guideline-discordant therapy. In a nationwide CDC-sponsored study, Fleming-Dutra et al estimated half of antibiotic prescriptions for URIs in the outpatient setting from 2010 to 2011 in the United States may have been unnecessary. Sinusitis and pharyngitis are among the 3 most common diagnoses associated with antibiotic prescriptions.Reference Fleming-Dutra, Hersh and Shapiro ⁴ In contrast to our study, this survey did not validate the accuracy of the physician’s diagnosis and may have underestimated the frequency of overprescribing.

Diagnostic inaccuracy likely contributed to our rates of guideline-discordant therapy. In our study, all cases not meeting diagnostic criteria were reviewed by the infectious disease physician and were reclassified as appropriate. For example, all of the patients in the ASB cohort were identified through an ICD-9 code for cystitis, and all of the patients with nonspecific URIs were identified through an ICD-9 code for pharyngitis or sinusitis.Reference Schroeck, Ruh, Sellick, Ott, Mattappallil and Mergenhagen ³ Distinguishing between bacterial and viral etiologies can be difficult; however, diagnostic guidelines are available to help providers make this distinction.

We observed an association between physicians in training and guideline-concordant antibiotic prescribing. In our study, residents and fellows were approximately twice as likely to prescribe guideline-concordant therapy as staff physicians. A few prior studies have also evaluated the influence of physician-level experience on antibiotic prescribing patterns. Roumie et alReference Roumie, Halasa, Edwards, Zhu, Dittus and Griffin ⁵ also found that residents in hospital clinics prescribed antibiotics for respiratory diagnoses less often than staff physicians, NPs, and PAs. Likewise, Gonzales et alReference Gonzales, Camargo and MacKenzie ⁶ found that ED visits associated with physicians in training were less likely to be associated with antibiotic prescriptions. Cadieux et alReference Cadieux, Tamblyn, Dauphinee and Libman ⁷ concluded that inappropriate antibiotic prescribing increased with time in practice and practice volume. Decreased guideline concordance over time may be associated with a provider’s increasing submission to patient expectations and reduced focus on patient education to ease work load.Reference Butler, Rollnick and Pill ⁸ Over time, providers may also become less aware of guidelines, which tend to be updated frequently.

Our study revealed that patients presenting to the medical center’s clinic were more likely to receive guideline-concordant therapy compared to patients presenting to the ED, an outlying clinic, or a CBOC. Historically, ED providers have misused antibiotics more frequently than primary-care providers.Reference Gonzales, Camargo and MacKenzie ⁶ Ong et alReference Ong, Nakase, Moran, Karras, Kuehnert and Talan ⁹ postulated that ED prescribers may be more likely to overprescribe antibiotics due to the more severe clinical presentations of ED patients, questionable follow-up, and greater patient expectations. Nonetheless, this same study reported that patient satisfaction was related to better knowledge regarding their diagnosis, not receipt of an antibiotic.Reference Ong, Nakase, Moran, Karras, Kuehnert and Talan ⁹

Several explanations for the CBOCs having more frequent guideline-discordant prescribing are likely. In contrast to the VAMC clinics, the CBOCs did not benefit from close affiliation with the local medical school, the support of physicians-in-training, or the ready availability of diagnostic testing. In addition, providers at CBOCs may have erred on the side of overtreatment because patients who worsened in the absence of antibiotics could have a greater distance to travel for seeking care.

Our results indicate an opportunity to expand our antimicrobial stewardship in the outpatient setting. The greatest area for improvement across the IC-VAHCS is in not prescribing antibiotics when antibiotics are not warranted, and our program is currently exploring strategies for addressing this problem. Audit and feedback, communication training, point-of-care testing, decision support, delayed prescribing, and education for both providers and patients have all been studied as interventions to improve outpatient antibiotic use and have demonstrated varying degrees of success.Reference Drekonja, Filice and Greer ¹⁰

The strengths of this study include assessing the accuracy of the ICD-9–coded diagnosis and the guideline concordance of the overall prescription, including antibiotic, dose, and duration. Also, we collected provider and setting data that is not well defined in the literature. Specifically, separating physicians in training from staff physicians and separating locations (ie, VAMC, ED, and CBOC) were all unique features of this study.

This study has several limitations. First, it was completed at a single center with a population of mostly elderly white males. Our results may not be generalizable to other populations. Second, we did not assess additional factors that could have contributed to prescribing decisions, including distance from the nearest hospital and time of day. Third, because this study included a retrospective chart review, providers may have failed to sufficiently document symptoms that would have altered the appropriateness of diagnosis and prescribed antibiotic. Fourth, all departures from guidelines were treated equally. Finally, the rapid antigen detection test for the diagnoses of group A streptococcal pharyngitis was not available at all VA clinics. In the absence of this test, Centor criteria were utilized, which could have led to the overestimation of a bacterial etiology of pharyngitis.

In conclusion, despite national practice guidelines for the treatment of cystitis and URIs, most patients received guideline-discordant therapy. Our identification of provider and setting determinants of guideline-discordant prescribing will inform future outpatient stewardship interventions at our center.