The influence of social relationships on clinician interpretation of performance data in an outpatient antibiotic stewardship intervention

Abstract

Background

Audit and feedback (A&F) of performance is a common strategy to improve antibiotic prescribing, yet it is variably successful. Little is known about the impact of social relationships on the implementation of A&F interventions in antibiotic stewardship. Our objective was to identify how social context influences primary care clinicians' perceptions of A&F received as part of a stewardship intervention.

Methods

We conducted semi-structured interviews with primary care clinicians who participated in a stewardship intervention in the Eastern United States. Interviews were conducted in the year following the intervention and respondents were purposively sampled by their performance: (1) sustained high performance, (2) improved performance, and (3) no improvement. Snowball sampling was used to deepen our understanding of practice context. Interviews were analyzed using the framework method.

Results

Interviews were conducted with 32 clinicians from 15 practices. Two clinicians were recruited via snowball. Thirty respondents were exposed to the intervention (of 183 clinicians in the intervention). Of these 30 respondents, 13 sustained high performance, 7 improved, and 10 did not improve. The content of responses by group varied in descriptions of the influence of social relationships on clinician understanding of feedback. Three relationships were identified as important: 1) with the health system, 2) with the practice, and 3) with patients. Clinicians with consistently high or improved performance described being understanding of and responsive to the demands of the health system and having a unified approach to judicious prescribing in their practice. They also expressed confidence in navigating patient demand for antibiotics. Clinicians whose performance did not improve were either annoyed by, anxious about, or dismissive of the health system's expectations, had low levels of interactions with colleagues in their practice surrounding antibiotics and prescribing norms at the practice that catered to patient expectations for antibiotics. They also reported difficulty in negotiating patient demand for antibiotics.

Conclusions

This study examines clinician reactions to the implementation of a feedback intervention to improve antibiotic prescribing in primary care. Our study demonstrates that clinicians use social relationships to make sense of performance feedback, which may influence the formation and execution of behavioral intentions in response to feedback.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13756-025-01686-4.

Introduction

Audit and feedback (A&F) of performance data is a commonly used implementation strategy to encourage the uptake of evidence-based recommendations in clinical practice [1]. It is commonly used to improve antibiotic prescribing [2–4], especially in the outpatient setting where the vast majority of human antibiotic consumption occurs [5–7]. A 2025 systematic review and meta-analysis of 56 randomized controlled trials of A&F interventions in primary care found the strategy was associated with an 11% relative reduction in antibiotic prescribing volume along with reductions in unnecessary antibiotic initiation, prolonged duration, and broad-spectrum antibiotic selection, although variation in impact across studies was substantial [8]. This finding is echoed in the broader A&F literature across clinical domains, which demonstrates improvement with marked variation in impact that is only partly explained by included effect modifiers [1, 9, 10]. To advance science in this area, research is needed that focuses on discerning ways to optimize the effectiveness of A&F as a strategy [11].

One way to optimize A&F is to integrate social science theories that specify how feedback works to change professional practice into the design and evaluation of interventions [12, 13]. For A&F to improve clinical practice clinicians must see the feedback, understand it, accept it, form an intention to change behavior based on the information contained in it, and then translate that intention into reliable behavior change [14]. The successful achievement of each step in the cycle is influenced by characteristics of the feedback, the recipient of the feedback, and the context surrounding them. Recent work suggests that the social context in which clinicians are embedded may shape the success of feedback interventions [2, 15]. Social context refers to the setting in which interactions between people take place. It includes the cultural, historical, economic, and political factors that shape people’s experiences. The contextual characteristics of healthcare organizations such as leadership, teamwork, policies, and community served, have been demonstrated to influence the impact of patient safety interventions [16].

Despite suggestions that social context may matter for the success of A&F interventions, it has not been well-studied in the literature evaluating the implementation dynamics surrounding outpatient antibiotic stewardship feedback-based interventions. The few studies that have examined clinician responses to A&F interventions have focused on the way clinicians think about the validity and reliability of performance data and the clarity of feedback report design [17–19]. Less is known about the way the social context of outpatient settings may shape clinician reactions to prescribing feedback. Given the well-documented contribution of patient expectation for unnecessary antibiotics to antibiotic overuse in outpatient settings, social context may be particularly relevant [20].

To address this gap in knowledge, we conducted a qualitative study using semi-structured interviews with primary care clinicians who had participated in a feedback-based antibiotic stewardship intervention to improve prescribing for respiratory tract diagnoses in the previous year. The intervention reduced overall antibiotic prescribing, but considerable variation in impact was observed between practices and clinicians [21]. Our objective was to examine how clinicians from different prescribing performance groups perceived the intervention, with a specific focus on the role of social context.

Methods

Study design, setting and participants

This qualitative study was conducted in the post-implementation period of a stewardship intervention that combined education with A&F with peer comparison to improve antibiotic use for respiratory tract diagnoses in primary care [21]. The intervention was conducted in 31 primary care practices in an academic health system located throughout the greater Philadelphia region. The practices varied in terms of teaching status (trainee physicians present or not), family medicine specialty (versus internal medicine), and geographic location (urban practices located in Philadelphia versus suburban community practices located outside the city). This grant-funded project was a research study led by academic infectious diseases clinicians. The intervention included two components: (1) a one-time in-person educational session delivered by infectious diseases physicians and (2) monthly individual feedback reports sent via email. The feedback reports, described in detail previously, included individual prescriber performance compared to their health system peers on two respiratory tract infection stewardship metrics [21, 22]. Clinicians were also given a laminated pocket reference card that summarized evidence-based recommendations. The last feedback report was sent in October 2018.

Interviews were conducted in the year following the intervention, from July 1, 2019 to February 1, 2020, between 8 and 16 months post-intervention. To better understand the intervention outcomes we used stratified purposeful sampling to identify respondents based on their performance [23]. Although 183 unique clinicians were exposed to the intervention, to increase the efficiency of recruitment, we randomly selected 108 of them to approach for participation. These 108 clinicians were stratified into three groups based on their prescribing performance before and during the intervention. The first group included clinicians in the highest-performing quartile of appropriate prescribing both before and during the intervention (Consistent High Performers). The second group included those in the lowest performing quartile prior to the intervention who stayed in the lowest-performing quartile during the intervention (Did Not Improve Performers). The third group included those in the lowest-performing quartile prior to the intervention who improved by a minimum of one quartile during the intervention (Improved Performers).

Physician investigators who led the intervention (KWH, LD) identified eligible clinicians in each group (labeled Groups 1, Group 2, Group 3) and created a spreadsheet of names and contact information. The investigative team responsible for qualitative data collection and analysis (LWG, BMM, JES) remained blind to the performance group of each respondent. Clinicians were recruited via email to participate in an in-person interview in which they would be asked for their opinions of the recently completed stewardship intervention. Respondents provided informed consent prior to starting their interview. To bolster our sample size at each practice we utilized snowball sampling. At the conclusion of the interview respondents were asked to refer practice colleagues to participate in an interview. Although we were referred in some cases to clinicians who had not been exposed to the intervention because they had joined the practice recently, we included them in our final sample because they provided important contextual information on practice culture.

Data collection

Interviews were conducted by a Postdoctoral Fellow (LWG), a female sociologist with extensive expertise in qualitative research on health. The interviews were conducted using a semi-structured guide (Supplemental Material). The guide was created by the research team, which included experts in antibiotic stewardship, infectious diseases, primary care, sociology, and implementation science. The design of the guide was informed by our prior work investigating the implementation of antibiotic stewardship interventions in primary care and in research on the social determinants of antibiotic use [17, 24, 25]. It included questions in the following domains: clinician decision-making surrounding antibiotics, perceptions of stewardship in primary care, interactions with patients about antibiotics, and opinions about the antibiotic stewardship intervention (educational session and prescribing feedback reports). All but two of the interviews were conducted in person at the clinician’s practice. Interviews were recorded with respondent permission and professionally transcribed. Respondents received a $50 gift card at the conclusion of their interview. The Postdoctoral Fellow wrote a data collection memo following the interview, summarizing initial impressions and emergent themes across each domain of the guide. This document, along with discussions amongst the research team, was used to determine when we had reached thematic saturation and could stop recruitment. The study protocol was approved by the University of Pennsylvania Institutional Review Board (Protocol #832834).

Data analysis

Interviews were professionally transcribed and uploaded into NVivo 12 qualitative data analysis software (QSR International) for management and coding. Interview transcripts were analyzed in a multistage process using the framework method [26]. Three investigators participated in the data analysis process, the Postdoctoral Fellow, a Research Analyst, and the Principal Investigator of the study (LWG, BMM, JES), all of whom are social science trained (sociology, anthropology) and have extensive experience conducting qualitative research on antibiotic prescribing and the implementation of stewardship interventions. First, an index codebook was created that contained the major domains of the interview guide augmented with codes derived from the literature. The codes were defined clearly and discussed amongst the analytic team. Then, two coders (LWG, BMM) applied the codebook to a subset of five interview transcripts. Intercoder reliability was assessed. The codebook was modified through discussion with the analytic team to include additional inductive codes, and to clarify code definitions to ensure consistent application across transcripts. Then, one coder (LWG) applied the codebook to all interview transcripts.

To examine potential qualitative differences in the responses from clinicians across different performance groups, we isolated a subset of codes for in-depth review. These codes included respondent perception of their prescribing performance, opinions about the intervention, and social context codes (including interactions with patients and interactions with colleagues). The coded excerpts were independently reviewed by two members of the investigative team (LWG, JES) to identify potential differences in content and tone across performance groups. Analysts remained blinded to respondent performance as they independently reviewed each coded excerpt. They wrote analytic memos during their review to identify supporting and falsifying data across performance categories.

The analysts met weekly over two months to discuss and synthesize findings across performance categories using framework matrices [27]. A matrix was created with the respondents as columns clustered by performance group. Rows included key analytic domains including discussions of prescribing performance, opinions about the intervention (with a separate column for discussions of feedback reports), interactions with colleagues and interactions with patients. Blinding was maintained until after the first framework matrix was created. Then, the physician investigators shared with the analysts which performance categories were reflected in the three groups. After unblinding, the analysts reviewed the framework matrix to synthesize observed variation in tone and content across performance groups. The analysts began by identifying notable differences by performance group on their own based on the first phase of analysis and review of the analytic memos. Then, they met together to discuss their observations and to identify areas of agreement and disagreement. The analysts met with the entire investigative team to review observations, share exemplar excerpts of interview data to support key points and finalize the results. This research was performed in accordance with the Consolidated Criteria for Reporting Qualitative Research (Supplemental Material) [28].

Results

Characteristics of respondents

Interviews were conducted with 32 clinicians from 15 practices. Two respondents had not been exposed to the intervention because they were hired during the implementation phase. Of the 30 respondents who participated in the intervention, 43.3% (13) were consistent high performers (from 6 distinct practices), 23.3% (7) improved performance (from 5 distinct practices) and 33.3% (10) did not improve (from 6 distinct practices) following the intervention (Table 1). Respondents had been in practice for an average of 17 years, with a range of 1 to 40 years in practice (median 13 years). Interviews averaged 33 min in length with a range of 10–66 min (median 32 min).

Table 1.

Characteristics of interview respondents by performance group

	Consistent high performers (n = 13)	Did not improve (n = 10)	Improved performers (n = 7)	Not exposed to intervention (n = 2)
Degree
MD	12	3	5	1
DO	0	1	0	0
NP	0	5	2	1
PA	1	1	0	0
Health system practice location*
City	13	1	2	0
Community	0	9	5	2
Gender
Male	8	4	2	0
Female	5	6	5	2

*Health system practices are in the greater Philadelphia, Pennsylvania region. This includes practices located in the city of Philadelphia, where the health system headquarters is located (anchored by the health system’s large teaching hospital). There are also practices located in suburban and rural communities outside the city. We refer to practices located in Philadelphia as “City” practices and those practices located outside of the city as “Community” practices

The influence of social context on clinician interpretation of feedback reports

We identified variation in the way clinicians conceptualized the influence of social context on their understanding of the feedback and their antibiotic decision-making. Respondents from different performance groups talked differently about their relationships with patients and with colleagues in their practice, which we asked about in the interviews. A third social context domain emerged inductively: the clinician’s relationship with and perception of the health system implementing the intervention. Below we describe how clinicians from different performance groups talked about these three different social relationships in relation to antibiotic decision-making and their perceptions of the antibiotic stewardship intervention.

Relationship 1: health system

We observed differences in the way respondents talked about the health system in relation to the intervention and the goal of improving antibiotic use. Although we did not ask specifically about the role of the health system, it was spontaneously mentioned by 27 respondents. The tone of the responses varied by performance group. Though respondents from all groups were aware of the health system as a surveilling presence, consistent high performers and those who improved expressed a sense of gratitude and understanding about the importance of performance monitoring. Those who did not improve expressed a anxiety, irritation, or dismissiveness.

Three respondents referred to the health system as “big brother,” invoking the Orwellian character symbolic of totalitarian mass surveillance. Two were in the group that did not improve. One described their response to the prescribing feedback reports:

I eventually was like, “I'm not looking at the reports. I'm gonna throw them right into the shredder.” 'Cause when you get constant bad feedback and you're stressed already, and you're trying to do what's best for your patient and your patients aren't complaining, but yet you're still getting bad feedback from an outside source, that was hard… I liked the pocket card. I mean, this is like my bible. I love the fact that I can use it and then hand it to the patient who wants antibiotics and say, “this is why I have to do this because [Health System] is watching us.” I can blame it on Big Brother. “I'm being watched, and if I don't do what's right for you, I'm going to get in trouble.” [R#30, Nurse Practitioner, Community Practice, Did Not Improve]

This respondent felt strongly about the feedback reports. Feeling discouraged by poor performance evaluations from “an outside source,” they eventually ignored the reports. The way they discuss using the educational pocket card as a defense against patient demand for unnecessary antibiotics demonstrates their perception of the health system as a punitive entity.

Another respondent who did not improve expressed keen awareness of being watched by the health system. The antibiotic prescribing metric was one in a system of surveillance in which many behaviors were being monitored:

I do think they made me try to be a little more thoughtful about prescribing antibiotics, just knowing that it was being kept track of. I will tell you that probably after the first one, I stopped looking at them and I would just delete them 'cause I couldn't deal with it. But it did make me think about it... I don't know. We're always trying to make [Health System] happy, too, as well as patients happy. So, the same way I'm trying to close my charts on time and do stuff like that, I’m like, okay. This is one more thing I have to think about….of all the things that get counted from [Health System] that I’m doing. Like Big Brother watching stuff. Do you know what I mean? Just monitoring all the reports we get about, I don't know, like Press Ganey, likely to recommend, and then percent of charts that are closed, how many patients you see, RVUs. I feel like... But that's part of working in a secure place like [Health System] and a huge place like [Health System] is that of course, everything's counted, and analyzed. But I guess it was just one more thing that I'm like, “oh, they're watching.” Which is good; they should be watching. It's okay. [R#27, Physician Assistant, Community Practice, Did Not Improve]

Although this respondent affirms that at the end of this statement that it is “okay” for the health system to be watching, they found the feedback overwhelming and discouraging especially in the context of other performance metrics. They mention patient satisfaction scores (Press Ganey), an administrative metric capturing workflow efficiency in the electronic health record (percent of charts closed), workload volume, and a financial reimbursement measure (Relative Value Units [RVUs]). There is a sense of fatigue expressed—“this is one more thing.” Contrast this to the tone of the response from the third person who referred to the health system as Big Brother, a physician whose performance improved:

So, there’s the big picture of medicine and then Big Brother [Health System] and then our practice. We have to listen to lots of different vested interests. But at the end of the day, you just wanna make sure that the person gets the right treatment. But when [Health System] writes us a letter, it feels a little more like, “Oh, God, I've been identified.” So, when I get—you know, well, because they have my e-mail address, they know I work for them. If it comes from a national board, I pay no attention, it doesn’t matter how I'm doing compared to all the physicians in the United States. But when they compare you to your colleagues, that feels very personal. And that feels like, at some point, it could be used to help me or against me. So, I do pay a little more attention to that…So, I'm glad that there’s some accountability to this, because we all need to help in various ways, and antibiotic use is one of them. [#9, Physician, Community Practice, Improved]

This person describes the role of the health system in surveillance as an important and helpful force, even if the reports stimulate an anxious response at first—“Oh, God, I’ve been identified” and “it feels like it could be used to help me or against me.”

Other respondents in the group that did not improve were frankly dismissive of the role of the health system and the impact of the intervention:

Some of my partners did not feel good about it [the report]. And they said “well, I just don’t understand this.” And I’m like “don’t worry about it, nobody’s grading you.” But everybody wants an A, so that is what we strive for. Until people came around to realizing it is not meaningful. It didn’t mean anything, it didn’t do anything, and it didn’t really matter, so people started ignoring it. It really didn’t impact my prescribing. I wasn’t going to change anything. I liked the talk, that girl who came out from [Health System] and gave it to us, that cute little card, which I’ve lost. I did like it but I think mine went in the laundry or something. [R#22, Physician, Community Practice, Did Not Improve]

This respondent is clear about not changing their prescribing because of the intervention, saying that the reports did not mean anything and had no consequences. Their description of the rest of the components of the intervention has a condescending tone—the “cute” pocket card that was lost in the laundry and referring to the “girl” who came to give the educational session.

Consistent high performers, many of whom were based in academic city practices, tended to describe themselves as being identified with and representative of the health system:

I think mostly the culture—I’ve been embedded in [Health System] for the last 25 years or so. And I do residency training. So, if we were not being responsible stewards of antibiotics, we would be subject to all kinds of public shaming. It does really kill me, though, that—like one of my patients did a video visit with another clinician within [Health System] and got an antibiotic that was totally inappropriate. And I sent a note and said, “What's up with this?” The person I sent it out to is, the ID department had a major antibiotic stewardship thing where they actually send people around. They have a road show. And they came and did a couple presentations to us. So, when I saw that this patient had called with like a runny nose—called, and then got a prescription for a Z-Pak. I sent that to him, and I said, “Have you guys taken your road show over to these guys?” [Laughter] He wrote back to me and said, “No.” [Laughter] I think they had already folded up that part of the operation. [R#8, Physician, City Practice, Consistent High Performer]

This respondent identifies strongly with the University health system, having worked in the system for over two decades. The presence of trainees functions as an additional layer of surveillance—where senior clinicians demonstrate adherence to evidence-based recommendations lest they be “subject to all kinds of public shaming.” This respondent also expresses taking on a surveillance role—speaking up about the prescribing behavior of another clinician in the system.

Relationship 2: practice colleagues

All respondents were asked about the social context of their practice in the interviews. Variation was observed in the way that respondents described the interactions they had with their colleagues about the feedback reports. Consistent high performers and those who improved described talking to their colleagues about the feedback reports. As one respondent who improved explained:

Our discussions about the reports were not so much about oh, “how did you do?” We discussed strategies for dealing with patients. You know, there are some people you can persuade, some people you can't persuade. We have had that discussion about lots of things, whether it's antibiotics or it's opiates. Some patients are just more intense than others. But the way we talked about the reports was not so much with regard to “how did you do,” but more “how are we going to approach it?” Or “how do you approach it?” [R#16, Physician, Community Practice, Improved]

In this scenario, the reports stimulated reflection and discussion amongst colleagues about how to address a common barrier to judicious prescribing—patient demand. A high performing physician explained:

When they presented the idea to us and we got the report cards people were very encouraged by it and were appreciative of it. My personal colleagues generally buy into the idea, and they liked the way the reports were formatted and talked about the reports. So, I do report cards for other things and people really liked that, like highest and lowest scores and kind of where you were in comparison with others. We actually reformatted one of our other feedback reports, on high-risk meds in the elderly, we made it look like the antibiotic stewardship report. [R #1, Physician, City Practice, Consistent High Performer]

This response demonstrates a high degree of acceptability and engagement with the feedback reports from the practice as a collective. It is notable that the antibiotic stewardship report was so well-received that the practice used it as a model for other clinical metrics.

In contrast, respondents who did not improve said that they did not discuss their reports with their colleagues:

Maybe I wasn't reading it right, I don't even know. And you know what? I'm pretty busy. So, that kind of thing eventually, I've got to focus on getting work done. I didn’t talk with anyone about the report, maybe others did. I don’t know. I don’t think so. [R#17, Nurse Practitioner, Community Practice, Did Not Improve]

Others who did not improve believed their colleagues ignored the reports:

So, I mean, most of my colleagues with that whole study, they just ignored it, never even looked at it. I mean, I'm one of the few people that try to take it seriously, because it is a problem… most people here, at this practice, are gonna say, “Why would I change, I'm just gonna practice how I'm practicing, make sure people are happy.” Like, yeah, it’s overall better for them to—overall better for the patients, too, for their health to not be taking antibiotics, but it’s not worth the pushback, so. [R#10, Physician, Community Practice, Did Not Improve]

This respondent perceives themselves as one of the “few people” in their practice to take the issue of judicious antibiotic prescribing seriously. This response reflects another pattern of variation we observed across performance groups: respondents from the high performing and improved groups tended to describe more unified prescribing practices than those who did not improve. As one consistent high performing clinician said:

In this practice, we talk about it, sure. We joke how patients want antibiotics all the time because they want to feel better, and we just say no. And pretty much, most of my colleagues are pretty good at that. I think we are fairly similar, slight variations, but there’s nobody in my practice that gives antibiotics without a fight, so to speak. There is a pretty unified culture around antibiotic prescribing. I think we all have a very similar practice style for that, especially for respiratory infections, which are, of course, the things we get the most of it. [R#6, Physician, City Practice, Consistent High Performer]

The description that this respondent gives about joking within the practice about the way they handle patient demand for unnecessary antibiotics indicates a shared norm around prescribing. Humor serves several functions for groups in clinical practice, including facilitating camaraderie, promoting conformity, and managing emotions [29].

A consistent high performer describes the differences between academic city practices and community practices around judicious prescribing and professional identity:

There is definitely a culture, absolutely a culture of prescribing, I think sometimes within practices, sometimes within divisions. I think within my division, which is general internal medicine there's a real culture about not overprescribing antibiotics and I think that the culture might be very different out in the community and in community practices. And I think that most of my people I work with in the school seem to really pride themselves about being careful about not overprescribing and we've totally bought the message of Dr. X [ID Physician] and others. We believe that and so that is almost a question of pride—“I don't overprescribe.” And as you know we get these ratings on how often we use antibiotics and I think people are very proud that they weren't overprescribing antibiotics, people see that as a marker of being a good doctor in our division, not overprescribing antibiotics, and it partly means for us that we're not just trying to please the patient, like we're really trying to do what we think is the right thing. I think like if you're in solo practice or you're sort of in a practice where people don’t care or give in all the time—you don't have the reinforcement of your colleagues to back you up. [R#14, Physician, City Practice, Consistent High Performer]

We did observe that respondents in community practices who did not improve described prescribing practices oriented towards “giving in” to patients or lack of a unified approach to antibiotic prescribing where people have strong individualistic attitudes about practice:

I think it varies prescriber to prescriber, definitely. Yeah. I think for our practice we tend to cater to our patients a lot, I think, and part of that catering is maybe saying yes when we shouldn't say yes. We order a lot of labs here. But I think we are a Mainline-y, make you happy kind of a practice. [R#27, Nurse Practitioner, Community Practice, Did Not Improve]

This respondent practices at a clinic in a wealthy suburb outside of Philadelphia. Their reference to being a “make you happy kind of practice” indicates a more lenient approach to prescribing.

One physician respondent from a community practice who was not exposed to the intervention because she joined the practice partway through the study (so had no prescribing data available to create a performance report) but was colleagues with two of our respondents who did not improve described their early experiences in the practice like this:

So, in the beginning of my time here I had a patient walk out in the middle of my conversation because I wasn't giving her antibiotics. So, it doesn't go that well all the time. I would have patients call in the next day saying—when I was newer, saying, "Dr. X didn't give me antibiotics, and I'm sick." And they would call in the next day and send a message to their PCP. Because we see each other's patients for sick visits. But they'd call in and be like, "Dr. X didn't give me antibiotics. I'm sick…” In the beginning when I actually first started working here, I did. I was giving out—I actually remember texting one of my friends. I was like, "Guys, I've given out antibiotics more in one month than I did in three years at my last office." Because I started in September, so it was cold season. It's difficult because when other providers in the office don't go along with the same way that you're taught, trained, it can be difficult especially when your patients overlap, and stuff like that, it can be difficult. It's difficult when patients are like, "Well, so-and-so gave me a Z-Pak. Why aren’t you giving me one?” [R#21, Physician, Community Practice, No Performance Data]

This respondent provides a valuable perspective for understanding how antibiotic prescribing norms in a practice can exert an influence on individual decision-making through patient expectations and interactions, as patients are seen by multiple clinicians.

Relationship 3: patients

Respondents across all performance groups described patient demand for unnecessary antibiotics as a barrier to judicious prescribing, but the content and tone of their responses varied in two ways. First, in discussions about the reasons why patients demand antibiotics unnecessarily, we found that respondents who were high performers and those who improved acknowledged that clinicians were partly responsible for this phenomenon:

A problem is that there is an expectation, I think, bred by us, the medical community over the years that if you're sick, you probably need antibiotics. It’s a problem of our own making. I can remember twenty five years ago when I first interviewed for the position in this practice, the guy who was then senior partner saying, "Yeah, when these people have a little bit of purulent spew, I just find you put them on amoxicillin and they get better faster. And what's more, they like being on something." I remember, even back then, thinking I'm not so sure it works like that. But today, you get people coming in. "Well, this scratchy throat showed up just two nights ago. I want to nip it in the bud. So, if you can get me on something, I'm sure I'll be doing better." That's the discussion, but it’s on us. [Respondent #16, Physician, Community Practice, Improved]

In contrast, the answers from respondents who did not improve universally suggested that intense patient consumerism was the primary reason why antibiotics were overused, or that it was the fault of other medical specialties. They tended to portray clinicians as service providers who had little autonomy to stand up to patient demand, especially in the context of pressure to ensure high patient satisfaction. As one respondent said:

This is the main reason why I wanted to do this interview, honestly, because I really don’t think people—like, people, researchers, or policy people—understand how much pressure there is to meet patient demands. And just, like—yeah, if patients feel like they need an antibiotic, it’s very, very hard to talk them down from that. So, antibiotics are definitely over prescribed. And, you know, it’s easier to say yes rather than taking time and, like, so much time to say no. [R#10, Physician, Community Practice, Did Not Improve]

Respondents who did not improve more frequently pointed to the role of other clinicians, especially those practicing in urgent care or emergency departments as the main driver of patient demand leading to antibiotic overuse:

But it's just frustrating because even when you try to educate them and explain this is the reason, they still either listen to you, like I said, and then turn around and ask somebody else for it, or they go to urgent care, and they get it from there, or whatever. So, you know they will get it somewhere else if you don’t give it to them. [R#20, Nurse Practitioner, Community Practice, Did Not Improve]

When mentioning urgent care centers, respondents who did not improve tended to justify giving unnecessary antibiotics because they felt the patients who really wanted them would get them somewhere else. Respondents who were consistent high performers or who improved did mention this same dynamic, but did not give it as a reason why they, individually, would prescribe:

Patients will choose. They're not dumb. It's like having a child that knows to go to which parent for which thing. I'll get this from Mom; I'll get this from Dad. Which one you gonna choose? If kids are smart enough to figure that out, you got to believe our patients are smart enough to figure that out. So, if they really want antibiotics, they can find them. That's a whole 'nother challenge. But we should still do the right thing here and not prescribe even if you realize that. [R#18, Physician, City Practice, Improved]

Second, respondents who did not improve described greater difficulties than those who improved or were consistent high performers in managing patient demand for antibiotics. They more often reporting “giving in” to patient demand:

Sometimes patients don't push back on my saying they don’t need antibiotics. But most of the time they do. Most of the time they put up some form of a fight. They say, "Look. I know. I get it. I understand. I've got this thing coming up. I really need them. Please give them to me." In some form or another. So that's usually what happens. And that's usually how they respond. It's hard. It's hard. You may say, "Well, look. Why don't you just wait it out for a few days, and call me back if things aren't getting better, and we'll talk?" And they say, "Oh, I've done that before. It never works. Give it to me now." Or something along those lines. And I don't do well with conflict. I'm not here to really debate you, necessarily, about this. And I don't want you walking out the door angry at me, or upset, which sometimes happens. But it ends up being a bit of a stressful situation for me. I have had patients storm out, get up, slam the door, curse and say, “why am I here?” in so many words. [R#19, Physician, Community Practice, Did Not Improve]

Respondents who did not improve said they gave in to patient demands because they were worried about negative ramifications from disappointing a patient, ranging from the hassle of getting complaints to disciplinary actions or financial penalties:

If I do gently push back, it generally doesn’t go well. I will often find myself compromising and giving the prescription and saying “give it five more days before you take it.” Then, “here's a Z-Pak” or “here's whatever” and “if you feel like it's not improving, then take it.” I know they go out and fill that Z-Pak the next day and immediately start taking it. Yeah. I feel like that's a compromise, but I know it's really me giving in. Not giving in exactly but doing probably what they don't need…I also think, for me at least, prescribing will happen because I know that if I don't give that antibiotic, I will have a message in my inbox the next day saying, I don't feel better. Almost very much like clockwork, I can just tell. I'll see this kind of patient and be like, “and there it is.” I didn't give them the Z-Pak they wanted and here's my phone call that I have to deal with now. That makes me be like, “I know you're gonna call me tomorrow…” [R#27, Physician Assistant, Community Practice, Did Not Improve]

Respondents who were consistent high performers or who improved reported more confidence and less fear about patient complaints about not receiving antibiotics:

When you don’t give an antibiotic that a patient wants, they are unsatisfied, and they want to give feedback, and of course by chance they will be the ones picked out for feedback or a survey. That's just what it is. I mean when things get—you get these reports, you get this feedback from patients that say—and these are very common—“this doctor is hard to prescribe an antibiotic.” I see this as, okay, that's good. Or “hard to prescribe a pain medicine,” and I would say, "Okay, that's good." [R# 11, Physician, Community Practice, Improved]

This respondent does not fear getting patient complaints or feedback about not receiving unnecessary medication and instead sees that as an affirming sign of making the right decisions for patients.

We found that the tone of responses about patient pushback on the “no antibiotic” recommendation from consistent high performers was confident and involved specific descriptions of strategies they use in this scenario:

I tell 'em that, "This is a viral infection. There's no role here for antibiotics," and a good portion of them say, "Okay, great. I have no desire to take medication unnecessarily," and many are relieved. Then there are these stubborn people who say, "Well, my other doctor always gave me antibiotics, or my friend got antibiotics, or my hairdresser said I must," so then you have a little bit of a challenge on your hands. And I tend to be pretty firm—"I'm not gonna do it, but here's what I'm gonna do for you: You get worse, you call me. I'm not turning my back on you. I'm not throwing you out the door. If this takes a turn for the worse—"occasionally I'll mention some specific parameters to look for—"please call the practice." [R#25, Physician, City Practice, Consistent High Performer]

This respondent describes an approach to the demanding patient that acknowledges the social challenges embedded in this potentially tense interaction—“I’m not turning my back on you” and providing a contingency plan if things get worse, an evidence based strategy to improve antibiotic prescribing [30, 31].

Discussion

Our qualitative interview study examined the way social context influences primary care clinicians’ perceptions of performance data received as part of an intervention to improve antibiotic prescribing for respiratory tract infections. We found that clinicians drew on their social relationships to make sense of feedback reports at multiple levels including their relationship with the health system, their practice, and patients. By examining the responses of clinicians from different prescribing performance groups we were able to explore variation in attitudes and beliefs around the role of social relationships in making sense of A&F data. The novel contribution of our findings is that there is a relational dimension to the way that clinicians interpret antibiotic prescribing performance feedback. Although the provision of feedback appears as an individual-level behavioral intervention, clinicians draw on their relationships with people around them to make sense of feedback. This sensemaking may influence the success of the feedback cycle [14]. These findings have implications for the implementation of stewardship A&F interventions.

We found that clinicians who remained high performers and those who improved because of the feedback intervention described being responsive to the demands of the health system and having a unified approach to judicious prescribing at their practice. They positively interacted with practice colleagues about the feedback reports. In contrast, clinicians who did not improve during the intervention were either anxious about or dismissive of the health system’s expectations, had low levels of interactions with their colleagues about the stewardship intervention and described a permissive approach to prescribing at their practice, that frequently catered to patient demand for unnecessary antibiotics. Respondents from different performance groups varied in the way they talked about their interactions with patients about antibiotics, with high performers and those who improved expressing more confidence in countering patient demand for antibiotics than those who did not improve.

Research on feedback-based antibiotic stewardship interventions in outpatient settings have identified numerous barriers to clinician engagement with and acceptance of prescribing feedback reports. Challenges include skepticism about data validity, belief that data are not relevant to their clinical practice, feedback reports that are difficult to understand, and unmotivating peer comparators [17, 18, 32, 33]. Most of the research on clinician reactions to A&F interventions in antibiotic stewardship focuses on characteristics of the feedback itself and less on the context in which the feedback is delivered. Our study is different from most existing qualitative work in this area by examining the role of social relationships in the way clinicians think about the credibility and actionability of prescribing feedback. Our study contributes to this literature by identifying potential mechanisms by which the social context shapes how clinicians perceive feedback [34].

There are two key stages in the feedback cycle that determine whether an A&F effort will change behavior [14]. First is the information to intention stage, where the recipient sees information on their performance and determines that a change in their behavior is necessary to improve. Second is the intention to behavior stage, where the recipient renders into action the behavioral intentions they formulated in response to the performance information. Our findings suggest that respondents draw on their social relationships to make sense of antibiotic prescribing feedback in a way that may influence the formation of behavioral intentions and the ability to consistently enact those intentions in everyday practice.

First, our finding that the health system was so often invoked by respondents to make sense of their feedback reports suggests this is an important driver of perceived report credibility, which is a precursor to forming a behavioral intention [34]. Our team did not think to ask specifically about the health system in our interviews because the intervention was part of a research study led by academic infectious diseases physicians with an extramural grant. While the health system had given permission for the study and we did state in the introduction to feedback reports that the system was supportive of the initiative to optimize prescribing for respiratory infections, this was not an operational health-system project.

That our respondents so commonly invoked the health system in relation to the reports is, in retrospect, not wholly surprising. However, it points to the importance of how recipients of feedback perceive the “feedback architect” in their sensemaking about performance data [35, 36]. Respondents who improved or stayed high performers were more accepting of the role of the health system as a performance monitor than those who did not improve, who often felt overwhelmed or unsupported by system level leadership. This finding underscores the importance of understanding how a feedback report’s design and implementation might be perceived by feedback recipients.

In addition to increasing the impact of A&F on clinical performance, attention to leader behavior in relation to the delivery of feedback behavior may also mitigate potential unintended negative consequences of performance feedback on clinician well-being. A recent field experiment of a feedback with peer comparison intervention to primary care clinicians demonstrated that the physicians who received feedback did not improve performance and experienced a significant decrease in job satisfaction and increase in burnout [37]. The study found that the mechanism underlying the negative effects was that the intervention inadvertently signaled a lack of support from leadership. Engagement of feedback recipients in the design of performance reports, paying close attention to the messaging about the feedback initiative especially as it relates to other system-wide initiatives, and enlisting a trusted local champion may improve the impact of A&F interventions [1, 34, 38].

Second, our finding about the role of practice level relationships indicates that although decision-making about antibiotics in the outpatient setting seems like a highly individualized experience (one clinician making a decision about one patient, relatively free of the direct influence of their colleagues in that moment, compared to inpatient care delivery), there are collective social dynamics that may shape how well clinicians are able to execute behavioral intentions formed in response to a performance feedback report [24]. A potential social mechanism is the degree to which individuals feel identified with practice-level social norms. Our findings suggest that if clinicians feel they are part of a collective that is unified and identified with the goal of judicious antibiotic prescribing that may generate feelings of self-efficacy in making the right decision even if it may lead to a difficult interaction. If they feel they are part of a practice that frequently uses antibiotics when they are not needed to please a demanding patient population, this may generate feelings of futility around withholding antibiotics, especially if others are going to do it too.

Literature about best practices for the design of A&F interventions emphasize the value of positive social interaction around performance feedback to aid in the formation of behavioral intentions and the development of strategies to enact those intentions [2, 14, 15]. We noted that prescribers who were consistent high performers and those that improved reported talking to their colleagues about the feedback reports, even though they were delivered to individuals. Respondents who did not improve described not discussing their reports with others in their practice or perceived they were alone in caring about the goal of improving antibiotic prescribing. This individual versus collective orientation to the feedback reports at the practice level may play a role in implementation success.

In our prior work on readiness to change and the success of a pediatric system-wide A&F antibiotic stewardship intervention [39], we found that high readiness practices trended towards greater improvements in prescribing than low readiness practices. Qualitative data gathered in that study indicated that clinicians in high readiness practices reported more open communication, active group change processes, and supportive underlying group cohesion. Clinicians in low readiness practices functioned more independently, with less interaction around antibiotic prescribing practices with their colleagues. Our findings in the present study indicate the potential importance of practice-level social dynamics on clinician sensemaking about performance feedback reports. Providing opportunities for social interaction at the practice level around antibiotic prescribing reports may increase clinician engagement with feedback reports and may encourage group brainstorming about strategies to improve antibiotic use [2].

It is also worth nothing that our sample exhibited striking difference in performance by geographic location of the practice—with more high performing respondents working in city practices while more of those who did not improve performance working in community practices. This is consistent with the finding in the intervention, which demonstrated that practices located outside the city had overall worse prescribing performance [21]. Possible contextual features of the health system’s practice network may have contributed to this observation, including the concentration of clinical trainees at city practices and different patient demographics in community practices. Previous research has demonstrated, for example, that judicious antibiotic prescribing is higher in clinical settings where trainees are present [40]. This could be because the training environment encourages more up to date evidence-based practice. Many of the community practices in the study were in more affluent areas, which may have a more demanding patient population, including patients who desire antibiotics that are not necessary [20].

Third, our finding about the way that clinicians from different performance groups talked about the navigation of patient demand for unnecessary antibiotics suggests that this social relationship may influence a clinician’s ability to consistently execute behavioral intentions formed in response to performance feedback. Patient demand for unnecessary antibiotics is consistently identified as a barrier to judicious prescribing across studies [20, 41, 42]. While most outpatient clinicians will report experiencing patient demand for unnecessary antibiotics, our findings suggest they may have different levels of comfort in responding to these demands, and differential sensitivity to potential negative consequences from withholding expected antibiotics (e.g. patient complaints, negative patient satisfaction scores, increased workload). They also may hold differential beliefs about the inevitability of a demanding patient getting access to an antibiotic from somewhere else. More research is needed to understand cognitive schema that may underlie these observations, and whether communication training tailored to these schema would help clinicians better navigate patient demand for antibiotics [43].

Our study has several strengths. While there have a been a number of qualitative studies investigating clinician responses to antibiotic stewardship A&F interventions in the outpatient setting [17, 18, 32], these studies gather data from clinicians whom have been exposed to a stewardship intervention but do not purposively sample based on performance. Variation in the impact of A&F at the individual prescriber level has been observed but research investigating reasons for this has been sparse. Our study represents an innovation in qualitative research on clinician response to a feedback-based stewardship intervention in that it examines variation in attitudes by performance group.

This study has some limitations to note. As a qualitative study of 32 clinicians from one health system the results should be interpreted as exploratory, rather than establishing a causal relationship. Despite limited generalizability, many of the relationships we identify as being important are likely transferable to other contexts, especially the impact of relationships with practice colleagues and patients. Although we recruited widely, it is possible that those respondents who agreed to participate have different perspectives than nonresponders. Finally, we interviewed busy primary care clinicians between 8 and 16 months after they were exposed to the intervention. It is possible that respondents forgot aspects of the intervention or their interpretation of the feedback reports changed over time in the post-intervention period. The interview method also has inherent limitations in that a respondent works to portray themselves as credible and, in talking about past events, may be sharing post-hoc rationalizations of performance. Therefore, we are limited in our ability to understand actual behavior as it occurred during the invention [44].

Practical implications

Our findings have numerous practical implications for the design and implementation of A&F interventions aimed at improving antibiotic prescribing in primary care settings. First, it is important for those designing interventions that are to be implemented across a system-wide network to develop a communication plan for how the purpose of the intervention will be conveyed [45]. Aligning the intervention with more general system-wide goals, providing supportive strategies to help clinicians improve performance (e.g. user-friendly clinical decision support or patient-facing materials to affirm the system’s commitment to judicious prescribing), and ensuring that senior leaders convey the importance of the intervention are potential approaches to address how clinicians perceive the purpose of A&F. Second, given our finding about the importance of practice-level social dynamics, there is potentially value in pre-implementation evaluation of practice dynamics, including norms about antibiotic use and the degree to which improving antibiotic use is a priority for the practice. Understanding practice-level dynamics prior to implementation could be used to develop more targeted strategies to promote shared understanding of and engagement with the A&F intervention. Such strategies could include identification of a local opinion leader to champion the intervention, group-based discussions about ways to improve performance, or communication training about ways to manage patient demand for unnecessary antibiotics [43].

Finally, our findings can help implementers re-interpret resistance that they encounter from clinicians or practices to an A&F intervention. Specifically, adopting a relational lens to an inquiry about why someone might object to an intervention can reveal unanticipated concerns that can be addressed while an intervention is ongoing. Adaptive challenges can be difficult to predict in advance as they often concern deeper issues such as competing priorities, legacy ways of working, people being pushed beyond their capacity, and deeply held loyalties [46]. A relational approach helps an implementer think about their intervention in context, which can identify social dynamics that might shape the impact of A&F.

Conclusions

This study provides important insights on the role of social context in the implementation of an A&F antibiotic stewardship intervention in the primary care setting. Although providing clinicians with performance feedback may seem like a highly individualized behavioral intervention, it is filtered through the social context of a clinician’s work organization, relationships with colleagues, and interactions with patients. Behavioral interventions are social interventions [35], and attending to the social context when implementing them is a critically important step to promote engagement and impact. We demonstrate that social relationships shape the way clinician make sense of antibiotic prescribing feedback, which may influence whether they accept the feedback and form a behavioral intention in response to the feedback. Future research on ways to harness the social context in the implementation of antibiotic stewardship performance feedback interventions is needed to advance the impact of these efforts on care delivery [2].

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

A&F

Audit and feedback

CP-FIT

Clinical performance feedback intervention theory

RVU

Relative value units

Author contributions

JES led the conceptualization and design of the study, obtained funding, lead the qualitative data analysis, wrote and substantially revised the initial draft of the manuscript. LWG recruited all interview respondents, conducted all interviews, participated actively in all phases of data analysis, and contributed to writing the first draft of the manuscript. BMM assisted in obtaining ethics approval for the study, assisted in the creation of the interview guide, participated in data analysis, and critically reviewed and revised the manuscript. EL obtained funding, participated in critically refining the conceptualization and design of the study, assisted in gaining access to study sites, and critically reviewed and revised the manuscript. KWH participated in critically refining the conceptualization and design of the study, co-led the quantitative analysis to identify eligible interview respondents by performance category, assisted in gaining access to study sites, and critically reviewed and revised the manuscript. LD participated in critically refining the conceptualization and design of the study, co-led the quantitative analysis to identify eligible interview respondents by performance category, assisted in gaining access to study sites, and critically reviewed and revised the manuscript. All authors have read and approved the manuscript.

Funding

This work was also supported by the US Centers for Disease Control and Prevention Cooperative Agreement FOA#CK-20-004-Epicenters for the Prevention of Healthcare Associated Infections. Dr. Szymczak also received support from the Agency for Healthcare Research and Quality (R01HS029328). The content is solely the responsibility of the authors and does not necessarily represent views of the US Centers for Disease Control and Prevention or Agency for Healthcare Research and Quality.

Availability of data and materials

The data generated and analyzed during this study are not publicly available due to the sensitive nature of the data and ethics restrictions on data sharing. Respondents did not consent to have their data publicly shared. A de-identified dataset may be available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval for this study was obtained from the University of Pennsylvania Institutional Review Board (IRB Protocol #832834). Potential interview respondents were sent an information sheet explaining the study via email at the time they scheduled their interview. Before each interview, an informed consent document was reviewed with respondents and they had ample time to ask questions and make a decision about participation. Verbal consent was obtained prior to the interview. All interviews were audio-recorded with the permission of the respondent.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.