Discussion

In this systematic review, we identified key factors associated with reduced antibiotic prescription for adult uARTIs in the ED, which included provider education on antibiotics use, attendance by internal medicine–trained physicians, and comanagement with house staff. Clinical factors associated with reduced antibiotics use were a normal CRP, positive influenza test, and patient having a recent ARTI.

There is strong evidence to support interventions to reduce antibiotics use in the primary care and inpatient settings. ^{Reference Gonzales, Anderer and McCulloch31,Reference Davey, Marwick and Scott32} A proactive approach with educational outreach and workshops was more effective than a passive approach. ^{Reference Ranji, Steinman, Shojania and Gonzales33} This situation is similar to the inpatient setting, where audit and feedback, decision support tools, and educational outreach improved adherence to antibiotics prescription policies by 15%. ^{Reference Davey, Marwick and Scott32} The only study with provider education included in our systematic review highlighted a 10% reduction in antibiotics use at year 2 by adopting a similar approach. ^{Reference Metlay, Camargo and MacKenzie26} This intervention included large- and small-group education sessions on the principles of antibiotics use for the providers; feedback on antibiotic utilization; and education of patients through posters, brochures, and video kiosks. ^{Reference Metlay, Camargo and MacKenzie26} Although effect size was not provided in the study, the results suggested that evidence from other settings could likely be translated to the ED setting. Innovative methods of delivering education and the frequency of outreach could be explored to minimize the disruption to ED physicians’ work shifts.

At the provider level, older physicians and those from non–internal-medicine specialties were more likely to prescribe antibiotics. Both factors were identified by the same study. ^{Reference Aspinall, Good, Metlay, Mor and Fine20} While the reason is not well understood, time constraint could play a role. ^{Reference May, Gudger and Armstrong28} Older physicians have more responsibilities, and surgeons allocate their time between operating theatre, clinics, and wards, ^{Reference Charani, Ahmad, Rawson, Castro-Sanchèz, Tarrant and Holmes34} in addition to fear of litigation and concerns about patient access to follow-up faced by all other physicians. ^{Reference Newton17,Reference Saposnik, Redelmeier, Ruff and Tobler35} Because younger physicians and those in the internal medicine specialty are more likely to use antibiotic regularly and be updated with the antibiotic guidelines, ^{Reference Charani, Ahmad, Rawson, Castro-Sanchèz, Tarrant and Holmes34} they could be engaged in the education of other physicians on appropriate antibiotic use.

In this review, comanagement of patients by junior physicians reduced antibiotic prescribing, similar to findings from a mixed setting (primary care and emergency department). ^{Reference Pynnonen, Lynn and Kern36} In clinical decision making, 2 systems of thinking could be involved: intuitive and analytical. ^{Reference Croskerry, Petrie, Reilly and Tait37,Reference Audétat, Laurin, Dory, Charlin and Nendaz38} The intuitive thinking system is fast but error prone, while the analytical thinking system counters cognitive biases. The presence of other colleagues on a team, could activate the analytical thinking process to rationalize the use of antibiotics. Laboratory investigations also augment the decision-making process. Patients presenting with fever, purulent sputum, and abnormal respiratory examinations are often inaccurately perceived as ARTIs of bacterial etiology. ^{Reference McKay, Mah, Law, McGrail and Patrick15} Positive influenza virus test or a normal C-reactive protein level provide objective evidence that antibiotics are not required, resulting in the reduction of antibiotic prescription. ^{Reference Stearns, Gonzales, Camargo, Maselli and Metlay27,39–Reference Rosenfeld, Piccirillo and Chandrasekhar41} Given the fast working pace in EDs, tools that address cognitive biases may reduce unnecessary antibiotic prescription. ^{Reference Gonzales, Anderer and McCulloch31}

Only 1 study in this systematic review reported that the physician’s perception of the patient’s expectations of antibiotics was associated with antibiotic prescription; this study was multicentered and consistent with the findings from a systematic review at the ambulatory setting, with adjusted odds ratios ranging from 1.7 to 23.3. ^{Reference McKay, Mah, Law, McGrail and Patrick15} Potential reasons include perceived antibiotics as a convenient solution to shorten consultation time, ^{Reference Morgan, Chang, Alqatari and Pines42} as a compensation for patients’ waiting time, ^{Reference Newton17,Reference Dempsey, Businger, Whaley, Gagne and Linder43,Reference Gaarslev, Yee, Chan, Fletcher-Lartey and Khan44} or as a tool to increase patient satisfaction, which was often associated with financial incentives. ^{Reference Sharp, Shen, Kanter, Berman and Gould45} However, 1 study suggested that the reason patients desired antibiotics was for pain relief and faster recovery, ^{Reference van Driel, De Sutter and Deveugele46} both of which cannot be addressed by antibiotics. Hence, it might be worth correcting patients’ misconceptions of antibiotics by addressing their real concerns.

These findings have important public health implications. First, education of providers on appropriate antibiotics use could reduce antibiotics prescriptions for adult uARTIs. Effective strategies in the primary care and inpatient settings that have been widely accepted ^{Reference Davey, Marwick and Scott32,Reference Ranji, Steinman, Shojania and Gonzales33} can be explored in the ED setting to assess their effectiveness in this fast-paced, time-sensitive context. Second, clinical factors associated with antibiotics prescribing could potentially be addressed by laboratory investigations or other decision-support tools that differentiate viral and bacterial etiologies. Their cost-effectiveness and practicality in the real-world setting are worth exploring. Third, most of the factors identified were based on 1–2 studies. Similar studies need to be replicated in different settings because the antibiotics prescribing culture likely differ among EDs and health systems. Nurse practitioners who are empowered to prescribe antibiotics, for instance, are nonexistent in the healthcare systems of some countries. Furthermore, the medical education systems and healthcare financing models also vary worldwide.

This review has several limitations. The study designs and study populations in the various studies were heterogenous. By restricting the review to the adult population, the results are only generalizable to adult uARTI visits at EDs. For the same factor examined, different studies adopted different definitions. Furthermore, different comparator groups were used in the outcome analyses. Nonetheless, we explicitly described them in the review and considered them when interpreting the results. Lastly, we considered English-language publications only. However, our literature search only identified 52 non–English-language articles, a small proportion (2.8%) of all articles, rendering any selection bias highly unlikely.

Antibiotics need to be used judiciously in the ED to conserve the limited armamentarium against infections. We identified important clinical and nonclinical factors influencing antibiotic prescribing for uARTI in the ED at the patient, provider, and health-service levels. Further studies on interventions addressing these factors to optimize antibiotics use are warranted.