Antibiotics are unique therapeutic agents often referred to as “societal” medications due to their ability to simultaneously affect the patient being treated and the community at large. Reference Schiff, Wisniewski, Bult, Parada, Aggarwal and Schwartz1,Reference Owens2 Inappropriate use of antibiotics in healthcare settings is most often characterized as resulting from a failure to adhere to best-practice guidelines and/or diagnostic error. This gap in care quality has been identified as a primary, modifiable contributor to the global increase in antibiotic-resistant bacterial infections. Reference Ventola3 Thus, there have been multiple “calls to action” related to antibiotic stewardship, including those targeting the emergency department (ED). Reference May, Cosgrove and L’Archeveque4
Skin and soft-tissue infections (SSTIs) account for ∼3% of all ED encounters (>3 million annual visits), and inappropriate antibiotic prescribing for this condition occurs frequently in this setting. Reference Rui, Kang and Ashman5–Reference Pallin, Camargo and Schuur8 There is a clear need to identify interventions that can optimize antibiotic use in the management of SSTIs in the ED. To successfully improve prescribing, interventions must be informed by key drivers of behavior, which can vary by provider type and setting. Reference Charani, Castro-Sánchez and Holmes9 Although much is known about drivers of guideline-discordant antibiotic use for other conditions (eg, respiratory tract infections) and settings, the literature for SSTIs and the ED setting is comparatively limited. Reference Haran, Wu, Bucci, Fischer, Boyer and Hibberd10,Reference May, Gudger and Armstrong11
The International Federation for Emergency Medicine published a report characterizing the ED as a unique clinical environment regarding quality and safety interventions. Reference Hansen, Boyle and Holroyd12 The report emphasizes the need for human factors and systems engineering informed approaches to successfully overcome these barriers. Therefore, we sought to characterize barriers and facilitators to optimal antibiotic use in the management of SSTIs in the ED using a systems engineering framework and to match them with targeted stewardship interventions.
Methods
Sampling
We conducted semistructured interviews at a national emergency medicine (EM) conference. To achieve conceptual saturation, we conducted additional interviews with EM physicians working at university and community EDs in the Midwest. Reference Bowen13 To be eligible, physicians needed to be actively practicing clinical EM in the United States and have completed or be in the final year of an EM residency program. All participants received $100 in financial incentive following the interview.
We recruited participants through conference e-mails and brochures included with conference materials. We also recruited participants from EDs in Wisconsin by direct e-mails. We selected potential participants by purposeful criterion sampling to ensure that perspectives from a range of settings (ie, urban, suburban, rural), geographic locations, years of experience, sex, and size of the ED were represented. Reference Palinkas, Horwitz, Green, Wisdom, Duan and Hoagwood14 Interviews and analyses were conducted over a 2-year period spanning 2017–2019. Our institutional review board approved all study activities.
Design and procedure
Using semistructured interviews, we explored broad themes around the diagnostic and antibiotic decision-making process for SSTIs that would be applicable across practice settings. Interview questions were primarily open ended so the participant could respond with what came to mind first. Probing follow-up questions were based on elements of the Systems Engineering Initiative for Patient Safety (SEIPS) framework and were utilized to identify themes within each element of the framework (see Supplement 1: Interview Guide). SEIPS was developed to comprehensively assess elements of healthcare work systems that affect care processes and outcomes. SEIPS has been successfully applied to characterize various quality of care and patient safety challenges (eg, antibiotic stewardship and diagnostic errors). Reference Tischendorf, Brunner and Knobloch15–Reference Carayon, Hundt and Karsh17 The element at the center of the model is the person (provider or patient) with surrounding elements (ie, physical environment, tasks, organization, and tools and technology) that operate within an external environment. The elements of the work system interact when preforming healthcare processes, which produce outcomes that feed back into the work system.
A nonclinical, study-team member with 5 years of experience in qualitative methods (R.J.S.) conducted one-on-one interviews in a private room. The principle investigator, a practicing EM physician with advanced training in systems engineering and qualitative methods (M.S.P.), attended 2 of the initial interviews to observe, ask additional clarifying questions, and facilitate minor modifications of the interview guide. We pilot tested the semistructured interview guide with 2 EM physicians at our institution. As interviews progressed, we refined questions and incorporated more pointed follow-up questions to encourage physicians to elaborate on the emerging themes.
We audio recorded all interviews, and a private company professionally transcribed audio files verbatim, which the study team reviewed for accuracy. Prior to starting the interview, we asked physicians demographic (eg, sex and years of experience) and practice-setting questions (eg, type of ED, teaching versus nonteaching, the annual ED volume per year, and the geographic region of the country where they worked). We proceeded with sampling, data collection and data analysis concurrently. We stopped collecting data when sufficient heterogeneity in participant answers was achieved as indicated by the responses becoming redundant and targeted probes failing to uncover new themes (ie, conceptual saturation). Reference Bowen13
Content analysis
We used deductive directed content analysis guided by the SEIPS model. Reference Hsieh and Shannon18 Researcher R.J.S. wrote an initial memo after each interview to capture emerging concepts and general observations; we used these memos as we generated the code book. Reference Birks, Chapman and Francis19 The study team developed a preliminary code book based on the domains of the interview questions and the elements in the SEIPS model. Reference Saldana20 Two study team members (M.S.P. and R.J.S.) used the preliminary code book and coded 6 interviews independently. Next, the coders met to review codes, add new codes, and refine code definitions. The study team continued to use memos during the coding process to track how code definitions evolved and to track divergent cases. We conducted this process for 6 interviews. For the remaining interviews, R.J.S. completed primary coding and M.S.P. conducted a secondary review, adding codes as needed. Any discrepancies in coding were resolved by discussion and consensus. Reference Barry, Britten, Barber, Bradley and Stevenson21 The finalized code book is included in Supplement 2. We used Dedoose, qualitative data software, to facilitate the coding process. 22
Intervention development
Once coding was complete, we generated a list of codes representing potentially modifiable barriers and facilitators or strategies. Following established intervention development methods, each identified modifiable barrier and/or facilitator or strategy was matched with a proposed intervention. Reference Wight, Wimbush, Jepson and Doi23,Reference Craig, Dieppe, Macintyre, Michie, Nazareth and Petticrew24 The interventions were then presented to a diverse group of 12 stakeholders from the author’s affiliated healthcare system in a series of either small group (n = 3) or individual meetings (n = 5) to elicit feedback. Stakeholders were identified and verbally invited to participate. The group was selected based on a goal of having diverse and multidisciplinary perspectives considered. As such, the group included 3 emergency physicians, 1 emergency medicine resident, 1 mid-level emergency provider, 2 ED nurses, 1 member of the ED clinical operations team, 2 infectious diseases physicians, and 2 infectious diseases pharmacists on the hospital antimicrobial stewardship committee. The meetings ranged from 20 to 40 minutes in length, and suggested revisions on the structure of the proposed interventions were captured by detailed notetaking. Revisions of the interventions and associated descriptions continued until there was group consensus that no further edits were necessary.
Results
In total, 39 physicians expressed interest in participating, and we conducted 20 interviews. No one refused to participate. The average interview lasted 48 minutes. The demographic and practice-setting characteristics are summarized in Table 1. Purposeful sampling yielded representation from a range of settings (urban, suburban, rural), geographic locations, years of experience, sex, and size of ED. The results that follow are organized according to the primary work system code of the SEIPS model with the last section of the results describing cross-cutting themes and targeted interventions proposed.
Description of Physician and Practice Setting Characteristics (n=20)
Barriers
We identified barriers to optimal antibiotic prescribing within the person (provider and patient), task, organization, tool and technology and external-environment work-system elements of the SEIPS framework (Table 2).
Barriers to Optimal Antibiotic Prescribing, Corresponding Work System Element, Infection Type and Representative Quote
a Both includes cellulitis and abscess.
Person-level barriers
Physicians described how patient expectation of treatment for both cellulitis and abscess was a barrier to optimal antibiotic prescribing because it led physicians to give patients antibiotics even if they did not always think they were necessary (Q1, Q2). Physicians described an increased willingness to prescribe antibiotics and to prescribe multiple antibiotics for SSTIs for patients who have an increased risk profile (eg, diabetes, recurrent infections), even if there was considerable diagnostic uncertainty (Q3, Q4). Providers also described how provider fear of treatment failure, including the development of a more serious infection with delayed treatment (eg, sepsis) and relapses were barriers to optimal antibiotic prescribing (Q5, Q6). Concerns over the chance of treatment failure were prioritized over the potential harms related to unnecessary antibiotics (Q7, Q8).
Task-level barriers
Diagnostic uncertainty was one of the primary barriers to optimally utilizing antibiotics for cellulitis. Physicians described utilizing antibiotics for a suspected cellulitis even if they had low levels of diagnostic certainty (Q9–Q11). Physicians described how optimal antibiotic usage for cellulitis was challenging because there is no objective diagnostic test (Q12). For abscess, diagnostic certainty was not a barrier except as it related to not knowing whether the causative organism was methicillin-resistant Staphylococcus aureus (MRSA). This uncertainty often led to the prescription of multiple antibiotics to achieve expanded spectrum of coverage (Q13, Q14).
Organization-level barriers
Physicians described an organizational culture where it is unacceptable to miss a bacterial infection (Q15). This culture encouraged physicians to ‘err on the side of caution’ and prescribe in cases of diagnostic uncertainty. Specifically, physicians cited pressure from hospital or department administration to ‘do something’ for patients as being a barrier to optimal antibiotic utilization (Q16). This pressure was particularly apparent when providers’ institutions emphasized patient satisfaction scores and if the provider believed the patient expected antibiotics (Q17). Physicians described poor access to ED follow-up care as a barrier to optimal antibiotic utilization because in many cases they could not count on a patient being seen for reevaluation in a day or 2 and were thus more likely to treat these patients with antibiotics in the ED (Q18). Finally, physicians described how time pressures in a busy ED was a barrier to optimal antibiotic prescribing because they simply did not have time to talk with patients about appropriate antibiotic use, including risks and benefits (Q19).
Tool and technology-level barriers
Physicians also described the need for diagnostic tools and how the absence of these tools made it hard to optimally diagnose SSTIs. For abscess, physicians were interested in having a rapid diagnostic test that could detect the presence of MRSA (Q20). Likewise, for cellulitis, physicians described the need for new tools to help them accurately diagnose cellulitis (Q21).
Environment-level barriers
We identified several external environment barriers to optimal antibiotic prescribing. Many physicians sensed that the current standard of practice in EM is to utilize antibiotics to treat cellulitis if there is any degree of clinical suspicion and indicated that it is challenging to go against historical standard of practice (Q22). Likewise, with abscess, many physicians described equipoise in the literature regarding the optimal utilization of antibiotics, which can make it hard for physicians to know how to optimally utilize antibiotics (Q23).
Facilitators
The physicians also described person and task-level facilitators (Table 3).
Facilitators to Optimal Antibiotic Prescribing, Corresponding Work System Element, Infection Type and Representative Quote
Person-level facilitators
Many physicians described having a shared decision-making conversation with patients. They felt that if they had enough time, they could often obtain buy-in to plans that did not involve prescribing an antibiotic (Q24). This finding contrasts with the patient-expectation barrier described previously, in which many physicians felt like patients always expected antibiotics no matter how much they discussed the idea of not prescribing with patients. A second person-level facilitator was physicians who self-identified as ‘antibiotic stewards.’ These physicians expressed the importance of antibiotic stewardship and in cases of uncertainty were more likely to consider the risk to benefit ratio related to antibiotics (Q25, Q26).
Task-level facilitators
Physicians described many task-level facilitators that helped them optimally prescribe antibiotics for skin infections. For abscess, physicians described how they could routinely convince patients that they did not need an antibiotic after completing an incision and drainage because they had done an intervention, drained the infection (Q27). For cellulitis, ruling out mimics was a facilitator that physicians used to help them optimally use antibiotics (Q28). Additionally, physicians described using the facilitator, watchful waiting, where they would not give a patient an antibiotic but instead put in place a plan for a recheck if the infection worsened (Q29). Finally, some physicians described providing a wait-and-fill prescription in which they would prescribe an antibiotic only to take under certain circumstances (eg, expanding erythema) (Q30).
Intervention mapping
We selected modifiable barriers and operationalizable facilitators and strategies identified by the physicians, and developed proposed interventions that could mitigate the barrier or enhance the facilitator (Table 4). Each intervention has a detailed description that underwent multiple rounds of refinement using input from a multidisciplinary group of stakeholders. These interventions cut across many of the identified SEIPS work-system elements, and they address several concerns: lack of access to ED follow-up care, patient expectations, diagnostic uncertainty (eg, MRSA and pseudocellulitis); fear of adverse outcomes, perceived clinical equipoise, and provider knowledge gaps. They range from systems-level programs (eg, community paramedicine follow-up programs) to novel diagnostics and clinical decision support tools.
Mapped Skin and Soft-Tissue Infection Stewardship Interventions for the Emergency Department
Discussion
In this analysis, we present the first qualitative assessment of perceived barriers and facilitators to optimal antibiotic prescribing for SSTIs from the perspective of emergency physicians. Utilizing the SEIPS systems engineering framework enabled us to identify barriers beyond the patient and provider themselves. This process directly addresses calls to develop quality improvement interventions (eg, antibiotic stewardship) that are grounded in systems engineering and behavior change theory and that are informed by data collected from frontline providers. Reference Charani, Castro-Sánchez and Holmes9,Reference Charani, Cooke and Holmes25–Reference Baker, Camosso-Stefinovic and Gillies27 Key identified barriers to optimal antibiotic prescribing for SSTIs included poor access to follow-up care (organization), need for more definitive diagnostic tools (tools and technology), and fear over adverse outcomes related to missed infections (person).
One unexpected finding of our analysis was the identification of knowledge gaps and skepticism of the literature. For instance, many providers held the view that antibiotics should now be given to all patients with uncomplicated abscesses based on recent trial data. Reference Daum, Miller and Immergluck28,Reference Talan, Mower and Krishnadasan29 There was a lack of awareness about the high number needed to treat in these trials and recent calls for a more nuanced approach to antibiotic prescribing for uncomplicated abscesses. Reference Pulia and Fox30,Reference DeBlieux31 Additionally, most providers doubted the validity of literature citing a 30% misdiagnosis rate of cellulitis in the ED. Reference Weng, Raff and Cohen7 One potential technological solution to these knowledge gaps would be clinical decision support embedded in the electronic health record as best-practice alerts.
SSTIs pose a particular diagnostic challenge given the absence of a gold-standard test. Providers expressed that the treatment decision must be made despite significant diagnostic uncertainty. Most providers opted to ‘err on the side of caution,’ which involved prescribing an antibiotic(s) even if the perceived likelihood of bacterial infection and/or their diagnostic certainty was low. This was especially true when other barriers were present such as poor access to follow-up care or a high-risk patient profile, which essentially lowered the bar to prescribe an antibiotic. The perceived patient safety and professional risk related to failing to provide antibiotics for an actual SSTI typically outweighed the acknowledged risk of adverse drug reactions and detrimental impact on public health related to unnecessary antibiotic use. Providers felt that evidence-based diagnostic tools that would make the SSTI evaluation process more objective would enable them to avoid prescribing in cases of low clinical suspicion.
Interestingly, several evidence-based interventions would fit this need that have not been extensively studied or adopted. For instance, rapid MRSA assays for purulent infections that strongly correlate with traditional cultures and improve tailored prescribing have been available for years. Reference May, Rothman and Miller32 Although not as well established, risk stratification scores (ALT-70) and surface thermal imaging have demonstrated potential to accurately differentiate cellulitis from pseudocellulitis. Reference Ko, Raff and Garza-Mayers33,Reference Raff, Weng and Cohen34
The perception among providers that patients generally expect antibiotics has been documented across a variety of healthcare settings, including the ED. Reference May, Gudger and Armstrong11,Reference Zanichelli, Tebano and Gyssens35 However, research examining the expectations of patients with respiratory tract infections (RTIs) in the ED did not find that patients routinely expect antibiotics. Reference May, Gudger and Armstrong11,Reference Ong, Nakase and Moran36 Although RTIs are a distinct clinical syndrome when it comes to antibiotic decision making, our findings suggest that a perceived expectation of antibiotics also plays a role in SSTIs. In addition to encouraging providers not to assume patients expect an antibiotic, a potential intervention would be for healthcare organizations to exclude encounters involving demands for nonindicated antibiotics from patient-satisfaction metrics. Alternatively, a more patient-centered approach towards education and shared decision making could potentially avoid this issue altogether. The development of a shared decision-making tool to facilitate patient–provider communication, such as has been demonstrated effective in reducing low value workups for low-risk chest pain in the ED, could enable clarification of the patient’s actual expectations (if any) while educating them about their individual risk and the providers level of diagnostic certainty (or lack thereof). Reference Bakhit, Del Mar, Gibson and Hoffmann37,Reference Gafni-Pappas, DeMeester and Boyd38
Barriers related to the external environment need to be addressed at a healthcare-system level. For instance, providers often ‘lower the bar’ to treat patients who have known difficulties with access to follow-up care. Ensuring that the patient could have a repeat assessment in a timely fashion to ensure any progression of the condition is identified as soon as possible could increase provider comfort in withholding antibiotics. With the rapid expansion of telehealth services due to the COVID-19 pandemic, it is more feasible than ever to incorporate either synchronous or asynchronous follow-up visits into ED SSTI care protocols.
This study had several limitations. Our recruitment strategy was an opt-in system, and it is possible that physicians who were already informed and interested in managing infections in the ED were the participants in the study. Because the primary aim of this analysis was to identify unifying themes, it is important to note that our findings do not represent an exhaustive set of emergency physician perspectives on this phenomenon. The proposed interventions were based on a mapping process guided by the identified themes, but we did not attempt to ascertain the magnitude of their potential impact or feasibility in different practice settings.
Using a systems engineering informed qualitative approach, we were able to characterize a number of barriers and facilitators to optimal antibiotic use for SSTIs specific to the ED work system. The developed mapped interventions span multiple components of the ED work system and should inform future efforts to improve antibiotic stewardship for SSTIs in this setting.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ash.2022.316
Acknowledgments
We thank all participants for sharing their thoughts and opinions with us.
Financial support
This work was supported by the Central Society for Clinical & Translational Research (CSCTR) Early Career Development Award, by the Agency for Healthcare Research and Quality (AHRQ award no. K08HS024342), and by a Clinical and Translational Science Award (CTSA) program, through the National Institutes for Health (NIH) National Center for Advancing Translational Sciences (NCATS grant no. UL1TR002373). The content is solely the responsibility of the authors and does not necessarily represent the official views of the CSCTR, the AHRQ, or the NIH.
Conflicts of interest
All authors report no conflicts of interest relevant to this article.
Open access
