We conducted in-depth interviews by digitally recorded telephone calls, and transcribed the recordings to text to accurately and reliably assess PCP KAPs. The qualitative method of an open-ended interview by telephone was chosen to ensure candid and truthful answers from participants. We composed a screening questionnaire to recruit physicians, nurse practitioners (NPs), and physician assistants (PAs) from a nationwide marketing database in the United States. We initially contacted and screened potential participants (Technical Appendix Section A) using telephones, email, and fax transmission of documents. Inclusion criteria included self-reporting of spending >50% of medical practice time in direct patient contact in a primary care setting, >30 years of age, and fluency in the English language. Interviewees were excluded from this study if they had an immediate family member who was employed in an industry that could represent a conflict of interest, including advertising or public relations, the federal government, market research, news media, or the pharmaceutical industry; if they had board certification in a subspecialty outside of primary care; or if they had practiced medicine >30 years at the time of recruitment.

Thirty-six PCPs were selected for the study. Specialties for the 27 physician participants selected included pediatrics (n = 9), family medicine (n = 9), and internal medicine (n = 9). Among PAs (n = 4) and NPs (n = 5), 6 practiced in family medicine setting and 3 practiced in pediatrics. Provider specialty, years in practice, and demographic information are described in Table 1.

Interviews were conducted during May 2013. One professional moderator who had >25 years of moderating experience conducted all interviews. Based on a discussion guide prepared by our research team (data not shown), participants were first informed of the sponsoring organization (Centers for Disease Control and Prevention, Atlanta, GA, USA), the planned use of data, and presence of listeners; they were then asked “warm-up” questions about the practice setting in which the participant worked and the patient populations whom they served. Next, each interview proceeded through an ordered list of open-ended questions on self-reported antibiotic prescribing practices, perceived prescribing practices of their peers, attitudes toward clinical practice guidelines for common bacterial infections, knowledge of narrow- versus broad-spectrum antibiotic agents, preferred resources and methods for medical education and antibiotic treatment, and attitudes toward antibiotic resistance.

Participants were provided a worksheet before the interview (Technical Appendix, Section B) which asked them to rank each of 12 factors on the basis of its perceived influence on antibiotic selection when an antibiotic is indicated (e.g., illness severity, patient demand for an antibiotic, or clinical practice guidelines). Worksheet answers were discussed and recorded during the interview.

To assess compliance with clinical practice guidelines and to evaluate clinical decision-making, each participant was given a clinical vignette about a patient (Technical Appendix, Section C) who had a diagnosis of an acute bacterial infection: acute otitis media (AOM) for pediatricians, acute bacterial rhinosinusitis for internists, acute uncomplicated cystitis for family practitioners, and group A streptococcal pharyngitis for PAs and NPs. The participant was asked to explain his or her rationale in choosing an antibiotic agent and why other PCPs might choose nonrecommended antibiotics. For the purposes of this study, antibiotics considered to be broad-spectrum include penicillins containing β-lactamase inhibitors (e.g., amoxicillin/clavulanate), second through fifth generation cephalosporins, macrolides, quinolones, and lincomycin derivatives. Narrow-spectrum agents include penicillins (e.g., amoxicillin), first generation cephalosporins, sulfonamides, and nitrofurantoin. Each participant received a cash incentive after the interview in exchange for participating.

Interviews were transcribed by project staff, and each member of the research team either listened to interview recordings or read corresponding interview transcripts. Relevant excerpts were coded into compilations of references to specific themes and were used to identify the most frequent responses to discussion topics outlined in the discussion guide. We identified common themes using both inductive and deductive methods that were reviewed by all authors. Author disagreement on theme selections were discussed until a consensus was met. If no consensus was met by most study authors, the theme was excluded from our results. We performed in-depth analyses of themes by reading and coding transcribed responses using Nvivo 9 (QSR International, Burlington, MA). This study was reviewed and approved for exemption status by the Human Research Protection Office of the Centers for Disease Control and Prevention.

We conducted 36 interviews, each lasting ≈45 minutes. Through analysis of provider in-depth interviews, several common themes regarding antibiotic prescribing and antibiotic resistance were identified (Table 2).

Our study shows that PCPs do not always adhere to guidelines because they believe broad-spectrum antibiotics may be more likely to cure an infection, a finding corroborated by the well-documented overuse of broad-spectrum antibiotics (23). However, the perception of better cure rates when using broad-spectrum agents is unfounded. For example, Streptococcus pneumoniae, a pathogen frequently implicated in bacterial respiratory infections, has a much higher prevalence of resistance to macrolides than it does to amoxicillin (24,25). Nevertheless, azithromycin is often chosen over amoxicillin for the empiric treatment for AOM, sinusitis, and other respiratory infections (25,26). Similarly, for the treatment for acute uncomplicated cystitis, widespread use of fluoroquinolones has contributed to rising antimicrobial resistance, rendering ciprofloxacin far less active against urinary tract Escherichia coli infections than the more narrow-spectrum, first-line agent nitrofurantoin (27). The perceived association between broad-spectrum antibiotic use and better cure rates may regularly contribute to inappropriate antibiotic selection and warrants further attention from appropriate antibiotic use initiatives.

There is no widely accepted definition of broad- versus narrow-spectrum antibiotics among PCPs or their professional organizations. Although a list of “antibiotics of concern” has been published by the National Committee on Quality Assurance (28) and has been used in previous research to classify antibiotics as broad-spectrum (29,30), the list was not originally intended for this purpose. Clinical practice guidelines emphasize use of narrow-spectrum antimicrobial agents instead of their broad-spectrum counterparts (4,5,31,32). However, the effect of these messages may be limited because of lack of clarity regarding what constitutes a narrow- versus broad-spectrum antibiotic. For example, few participants in our study were able to confidently categorize macrolides and penicillins, which are among the most commonly prescribed classes of antibiotics (33), as broad or narrow spectrum. Although this issue is largely one of semantics, it has critical implications for medical education, public health messaging, and community antibiotic resistance. Communication to PCPs related to antibiotic choice should not focus on dichotomous narrow- versus broad-spectrum terminology, but rather promote specific recommended first-line and targeted antibiotic therapies for individual diagnoses.

Compared with results of previous qualitative studies, PCPs participating in this study expressed greater urgency regarding antibiotic resistance. For example, in a 1998 qualitative study exploring driving factors of antibiotic misuse, a principal barrier to change in antibiotic prescribing was the attitude that antibiotic resistance was not an important problem (19). Another study published in the same year noted similar findings (21). Conversely, not a single provider in this study dismissed antibiotic resistance as being a minor issue, and several expressed grave concerns about antibiotic resistance based on their own experiences.

Modifying prescriber behavior is a complex and difficult task. Multifaceted interventions that involve a combination of interactive group meetings, outreach visits to individual physicians, physician reminders, or patient-based interventions (e.g., delayed prescribing practices) have shown the most promise in changing prescribing behaviors in ambulatory care settings (34,35). Previous studies confirm that patients desire antibiotics less frequently than providers perceive and that inappropriate prescribing is a common result of this miscommunication (21,36,37). This finding suggests that an effective target for intervention is narrowing the gap between patient expectations and clinician perception of these expectations for antibiotics. Regardless of the intervention considered for promoting appropriate antibiotic use, the concerns of PCPs highlighted in this study should be addressed. This includes reassuring providers of the efficacy of first-line and targeted therapies, clarifying the role of antibiotic prescriptions in patient satisfaction, and providing resources that streamline patient education efforts in primary care settings.

This study has limitations, however. Although in-depth interviews are an effective method to explore individual providers’ KAPs, we cannot generalize our findings to the PCP population as a whole because of the lack of external validity inherent in this type of qualitative research. Similarly, the clinicians who were screened, selected for participation, and agreed to be interviewed may hold stronger opinions about this topic than those who were excluded or declined to participate.

Our study suggests that inappropriate antibiotic selection among PCPs is not caused by lack of familiarity with guideline recommendations. Rather, this practice is the result of a complex interaction between perceived better cure rates for nonrecommended therapies and attempts to meet demands typical in primary care settings.

Future research efforts should be aimed at investigating effective incentives for appropriate antibiotic prescribing and determining alternative communication strategies to encourage use of first-line agents. Although most efforts have focused on reducing unnecessary antibiotic use, more research is needed to clarify which interventions improve antibiotic selection. Although awareness regarding antibiotic resistance appears to be improving, ongoing education efforts promoting appropriate antibiotic use among both patients and health care providers (e.g., CDC’s Get Smart: Know When Antibiotics Work Program) are critical to addressing the growing threat of antibiotic resistance.