Normalization of a conversation tool to promote shared decision making about anticoagulation in patients with atrial fibrillation within a practical randomized trial of its effectiveness: a cross-sectional study

The institutional review boards of the Mayo Clinic (approval 16-005409) and two other participating sites (Park Nicolette-HealthPartners and Hennepin Healthcare System) approved all study procedures. We recruited clinicians for a randomized trial (Shared Decision Making for Atrial Fibrillation [SDM4Afib]) evaluating the effectiveness of the Anticoagulation Choice Decision Aid tool versus usual care in patients with nonvalvular atrial fibrillation at risk for thromboembolic strokes and considering anticoagulation (ClinicalTrials.gov registration number NCT02905032) [7].

The tool, which can be found at anticoagulationd​ecisionaid.​mayoclinic.​org, was designed to support collaborative conversations between patients and clinicians to determine whether to use anticoagulation, given the patient’s risk of stroke, and how to anticoagulate (using either warfarin or direct oral anticoagulants) based on the relative merits (bleeding risk, dosing, need for monitoring, reversibility, cost, and interactions with foods and medications) of these options.

The SDM4Afib trial protocol [7] and the design of the Anticoagulation Choice Decision Aid tool have been described previously [6]. Briefly, we recruited clinicians from three hospital sites in Minnesota: an academic medical center, a suburban community group practice, and a safety net health system. Clinicians were eligible if they were involved in conversations about anticoagulation with patients with atrial fibrillation. Eligible clinicians were invited to participate in the trial via e-mail and during established clinic meetings at primary care, emergency, cardiology, and dedicated anticoagulation practices at participating sites. After consenting to participate in the trial, the clinicians took part in a brief training session, either individually or in groups. In each session, study staff demonstrated how to use the Anticoagulation Choice Decision Aid tool with a patient through role playing with the clinician and, when feasible, by reviewing a brief video-recorded demonstration. Some of these training sessions took place “just in time” prior to appointments with eligible patients (which sometimes prohibited a more extensive training session with video recording), while others were scheduled outside the time of clinical duties. The clinicians were then asked to complete a questionnaire. During and after training sessions, participating clinicians were invited to explore issues of normalization of the SDM tool into their practice with prompts such as, “How does this tool differ from your everyday work?” and “What resources do you have for integrating this tool into practice?” Spontaneous and prompted comments made by clinicians during training sessions were recorded with their consent. Feedback and questions from training sessions and interviews were discussed with the study principal investigators. Recommendations from those conversations were shared with study staff via a live electronic database and monthly in-person/phone study staff meetings (Fig. 1, gray panel).

The clinician questionnaire was used to collect information about demographic and professional characteristics, burnout symptoms, and clinicians’ perceptions of the Anticoagulation Choice Decision Aid’s normalization potential (Additional file 1).

Burnout was assessed using two items of the Maslach Burnout Inventory (MBI), “I feel burned out from my work” and “I have become more callous toward people since I took this job,” that have strong associations with the full MBI and key outcomes such as suicidality and desire to drop out of training [8]. The MBI items were scored on a 7-point scale where “never” contributed 0 points and “every day” contributed 6 points. Similar to previously reported methods, burnout was classified as “high,” “moderate,” or “low” [8‐10]. High burnout was classified as feeling burned out from work or feeling more callous toward others once or more per week (4 points or greater). Medium burnout was having feelings of burnout or callousness once or more per month but less often than once per week (2 or 3 points). Low burnout was classified as feeling burned out and more callous toward others a few times per year or less (1 point or less).

Clinicians answered questions regarding their familiarity with the tool and the Normalization Measure Development (NoMAD) instrument after training [11]. Familiarity was rated on 10-point Likert scale anchored by responses such as “still feels very new” and “feels completely familiar” at the extremes. The NoMAD instrument elicits stakeholder views about how interventions impact their work and whether the new intervention could become normalized in their setting by probing the four domains of NPT (Table 1) [12, 13]. In our study, we omitted questions about “reflexive monitoring” because participants had not yet had a chance to integrate the SDM tool into their practice. The 15 NoMAD items, comprising 3 domains (5 items in each), were scored on a 5-point Likert scale, where “strongly agree” was assigned a score of 5 and “strongly disagree” was assigned a score of 1. Psychometric testing of this instrument supports taking the average score for all items within each domain [11].

Training across all sites captured feedback from 112 clinicians, 40 (36%) of whom received individual training and were interviewed one-to-one. We used a computer-generated simple random sample of 30 training recordings for a theory-led framework analysis. Proceedings were transcribed and classified using inductive coding into NPT domains (Table 1). The four coders (AT, GSB, OJP, and PO) calibrated their judgments by analyzing two recordings independently and successively, guided by a preliminary codebook. Disagreements were discussed, and the codebook was iterated accordingly with the addition of helpful examples. The remaining 28 recordings were analyzed independently in pairs, with disagreements discussed and resolved between reviewer pairs or by a third coder when necessary.

In describing the possibility to integrate the SDM tool into their practices, 168 clinicians responded to the whole 0–10 scale with a mean of 6.6 (SD 2.5); cardiologists gave, on average, ratings that were 2 points higher. The mean NPT domain scores were 4.1 of 5 (SD 0.5) for coherence, 4.3 (0.5) for cognitive participation, and 3.9 (0.5) for collective action (Table 3). Using ANCOVA adjusted for clinician characteristics, there was no significant association between level of burnout and NPT domain scores or sites (results not shown).

Embedding an intervention in clinical routines is necessary to evaluate its effectiveness in practice. This process takes individual and collective work subject to barriers and facilitators. Identifying and addressing these factors is critical in the design and implementation of practical clinical trials. This is particularly important when barriers and facilitators may differ across sites of a multisite trial. Here, we explored the embedding of an SDM intervention in routine clinical consultations as we recruited clinicians for a multicenter trial evaluating it. Clinicians rated the normalization potential of the intervention highly; for cardiologists, using the SDM tool in care made more sense than for other types of clinicians. Clinicians endorsed all domains of normalization to the same extent, regardless of site, clinician characteristics, or burnout or callousness ratings. Ninety-two percent of clinicians saw the value of using the Anticoagulation Choice Decision Aid tool, and 96% were open to working with colleagues to improve the use of it in practice. Fewer clinicians saw how the Anticoagulation Choice Decision Aid tool differed from their normal work (72%), agreed that there were drivers leading the implementation of the conversation aid (75%), or responded in agreement with items related to collective action (70–77%). Qualitatively, clinicians paid most attention to making sense of the tool. Tool buy-in seemed to depend heavily on clinicians’ ability to see the tool as accurate and “evidence-based” and on their having time to use it during consultations. They mentioned that the Anticoagulation Choice Decision Aid tool might lead to standardization of clinical encounters. Collective action issues focused on who has time and information to use the tool (e.g., enrolling pharmacists and the information to which they have access when including patient out-of-pocket costs in SDM conversations). Few clinicians commented on the concept of “shared decision making” per se (less than 1% of clinician utterances).

While clinicians found coherence between their job and using the SDM conversation tool, they were less likely to agree that using the Anticoagulation Choice Decision Aid tool differed from their normal work. This is consistent with a metasynthesis of qualitative studies which found that for conversations about anticoagulation for atrial fibrillation, clinicians reported engaging in SDM routinely, even when patients were more likely to report a less participatory approach [14]. This was also supported by an evaluation of real-world distribution of decision aids; qualitative data suggested that clinicians did not have a shared understanding of the purpose of the decision aid [1].

Similar to the findings of the systematic review by Legaré et al., who documented that clinicians identify time constraints and lack of pertinence of the SDM intervention to patients and clinical situations as barriers to SDM implementation [5]. The design of the Anticoagulation Choice Decision Aid tool as an SDM conversation aid enables clinicians to tailor the tool to the patient’s situation [6], dealing with the latter barriers, and it may, in fact, efficiently guide a complex conversation and save time. Although SDM conversation tools demand dedicated and structured work that may change the way the clinician feels about time during the encounter, a recent systematic review of these tools demonstrated that, on average, these tools do not significantly prolong the clinical encounter [15].

Previous studies have also found burnout to be negatively associated with adaptive reserve [16], “practice features that enhance resilience, such as relationships” [17]. Burnout has also been associated with less empathic concern and decreased “effort to adopt the point of view of another person” [18]. While SDM has been proposed as an intervention to increase meaningful encounters and decrease burnout [19], our findings do not directly support an association between burnout and enthusiasm regarding SDM. However, given the highly positive response by clinicians to openness to collaboration with colleagues, their multiple inquiries regarding other clinicians using the Anticoagulation Choice Decision Aid tool, encouraging collaboration and innovation between team members, and involving local leadership may improve implementation of complex interventions such as SDM tools.

Process evaluations during trials may provide insights about the factors that promote or inhibit the implementation of complex interventions that are different from those gained from evaluations performed outside of research endeavors [2]. Research studies are temporary, require participant consent, and allow participant withdrawal without penalties (Fig. 1). Temporary add-on trial procedures, for instance, may overwhelm clinicians, who then opt out of the research, or, conversely, they may elicit weaker responses than permanent or mandatory practice changes. Similarly, clinicians choosing to participate in the trial may have been more enthusiastic about the intervention than those who declined. This study was conducted at the beginning of the trial, before participating clinicians had an opportunity to use the tool with patients. Therefore, we captured mostly clinicians’ expectations about using the SDM tool, informed by very brief training and demonstration when clinicians may have been more enthusiastic about the process. We did not assess “reflexive monitoring,” the fourth domain of NPT, which requires responding to questions after having regularly used the intervention, such as, “Given the changes observed as a result of the intervention, should we keep doing it?” For example, the organic monitoring, appraisal, and endorsement by clinical staff (represented by white arrows and white panel in Fig. 1) were not measured in our study. These limitations affect inferences about the normalization of SDM tools in practice to a greater extent than about the procedures for ongoing clinician recruitment into our trial, the purpose with which we designed this process evaluation. Finally, while participants were informed that the results would be anonymized, it is possible that clinicians may have felt uncomfortable reporting high levels of burnout for fear of stigma or professional repercussions.

It has been suggested that SDM may decrease clinician burnout by creating more meaningful patient–clinician encounters [19]. A clinician communication skills training intervention was associated with a significant and small decrease in burnout [20]. Yet, trials testing SDM tools rarely collect the clinician perspective, with no trials collecting measures of burnout or clinician well-being [15, 19, 21]. Our study explored the interplay between SDM implementation and contextual factors such as burnout and practice resources such as time in consultation. A key insight that participants offered and deserves further exploration is that the normalization of SDM tools in practice may benefit from supporting collective action. We found that people were enthusiastic about working with colleagues, a process hindered by their lack of awareness of who else was participating in it and who was driving the implementation of the tool in their clinics. Perhaps future research should ascertain not just clinician burnout but also isolation.

This process evaluation contributed to the rollout of the trial. Clinicians were invited to provide feedback and ask questions about trial procedures and the tool. When study staff (research assistants and study coordinators) were unable to answer questions in real time (e.g., about bleeding risk estimates), questions were referred to the principal investigators. The principal investigators, topic experts at their own sites, contacted participating clinicians to respond to these queries, communicating their buy-in in the use of the SDM tool. Internally, study staff maintained a live database of questions and topics that came up during clinician consent or NPT interviews. The study staff recorded and regularly updated crowdsourced answers (Fig. 1) to maintain coherence regarding the tool and trial within the study team and between study sites. In these interactions, we had to maintain a balance between facilitating the use of the SDM tool and maintaining a sense of uncertainty about its relative efficacy (rather than advocating for its value), which justified the clinical trial.

Clinicians made several requests to modify the content of the tool. For example, the tool was designed to support a conversation between clinicians and their patients; patient and clinician expertise and experience should enhance the tool and tailor its effect to each patient and situation, contributing to a pertinent and useful SDM conversation. In this way, we resisted adding more information to the tool, a common request, because this may have reduced clinician participation in the conversation and promoted instead an interaction between the patient and the SDM tool. Other changes could be accommodated. For example, we added the ability to toggle between 5-year and 1-year risk estimates.

Finally, an apparent disconnect emerged between the communication goals of the research and of clinicians, the former more focused on cocreation and the latter on conveying information. A previous study showed that even within clinic teams, there are dissimilar interpretations of SDM [22]. In our random sample of 30 recordings (112 clinicians), only 2 comments mentioned SDM, and these arose from making sense of using SDM tools rather than a philosophical discussion on the purpose of “shared decision making.” In this way, far from having arrived at a broad consensus of definition, purpose, and value of SDM, our study suggests that clinicians continue to negotiate the purpose and value of SDM in their practices. Future studies should continue to explore the clinician perspective, particularly when SDM implementation must take place within individual and collective workflows. Clinicians can both give meaning to SDM tools and preclude their presence in clinical encounters.