Bidirectional learning opportunities: How GP‐supervisors and trainees exchange knowledge

Abstract Introduction Workplace‐based learning conversations can be a good opportunity for supervisors and trainees to learn from each other. When both professionals discuss their specific knowledge openly with each other, learning conversations may be a useful educational tool, for instance for learning how to apply evidence‐based medicine (EBM) in the workplace. We do, however, need a better understanding of how the exchange of knowledge provides opportunities for such bidirectional learning. The aim of this study was therefore to analyse how trainees and supervisors currently handle bidirectional learning opportunities by describing in detail how supervisors respond to knowledge expressed by trainees during a learning conversation. Method We video‐recorded learning conversations between supervisors and trainees in general practice (GP). Within these learning conversations, EBM discussions on medical topics were selected and transcribed. We then identified, analysed using Conversation Analysis (CA) and categorised each expression of knowledge by the trainee and the supervisor's subsequent response. Results We found that when a trainee expresses knowledge during the learning conversation, supervisors either (a) refute the expressed knowledge, (b) immediately suggest an alternative or (c) pose (additional) questions. These responses have consequences for the learning opportunities of both trainee and supervisor: it is only when supervisors pose further questions that trainees are encouraged to elaborate on their knowledge, leading to a bidirectional learning opportunity. Discussion Improving EBM learning opportunities for both supervisors and trainees requires more than simply instructing trainees to express knowledge‐based—for instance—on recent evidence more often. Inflexible institutional roles related to historical claims of supervisors’ epistemic authority hamper bidirectional learning. Posing open questions during learning conversations enhances the flexibility of institutional roles while also creating bidirectional learning opportunities.


| INTRODUC TI ON
Medical education relies a great deal on workplace-based learning, in which trainees learn through, for instance 'supported participation' and dialogue. 1 Dialogue between trainee and supervisor can be seen as one of the most important aspects of workplace-based learning, being deliberative yet informal. [2][3][4][5] One specific workplace dialogue that could be a good learning opportunity for both supervisor and trainee is the 'learning conversation', a standard practice in general practitioner (GP) specialty training. Learning conversations are regularly scheduled meetings in which clinical supervisor and trainee discuss clinical questions, medical topics or personal development. 6 They combine debriefing, supervision and feedback, in this way assigning the 'learner' role largely to the trainee while the supervisor advises, comments or instructs. 7 Knowledge on how these dialogues precisely occur and how both trainee and supervisor can learn from exchanging knowledge during such dialogues is lacking.
Although workplace learning conversations are mainly seen as an opportunity for trainees to learn from supervisors, supervisors can also learn from trainees. Previous research on bidirectional learning opportunities showed that supervisors acknowledge that they could learn from their trainee during learning conversations. 8,9 Also research on collaborative learning has shown that learning between peers or professionals can lead to valuable learning outcomes due to the exchange of knowledge and perspectives. 10,11 However, collaborative learning is mostly researched in a setting of two peers. The concept of bidirectional learning, defined as reciprocal learning between supervisor and trainee, has gained surprisingly little attention within medical education research. So it remains unclear whether and how bidirectional learning opportunities during learning conversations are seized.
To analyse how bidirectional learning opportunities are approached, it is a good starting point to look at how trainees and supervisors learn to apply evidence-based medicine (EBM). While traditionally, learning of EBM has been mainly focused at how to search and appraise clinical evidence, recent papers advocate for a more holistic approach in which the contextual application of EBM at the workplace is essential. 12,13 This involves learning how to take decisions about individual patients by combining (a) the best available evidence, (b) the patient's preferences and (c) the clinician's clinical expertise. 12 logues. In order to get a better view on the potential of bidirectional learning, it is essential to know how knowledge exchanges between supervisor and trainee currently occur.
To study bidirectional learning opportunities, we have applied the Conversation Analysis (CA) method. By analysing how people respond to each other's utterances, CA allows us to identify how people deal with asymmetries in knowledge, adding valuable new insights into learning and learning opportunities in medical education. [16][17][18] By describing in detail how supervisors and trainees exchange EBM knowledge together during their conversations, this study aimed to analyse how trainees and supervisors currently approach bidirectional learning opportunities.

| Setting
In the Netherlands, postgraduate GP specialty training takes 3 years.
In the first and last year, trainees work alongside an experienced GP, that is, their supervisor. One hour of the daily routine is set aside for workplace-based learning conversations covering a range of topics, from medical cases in daily practice and training institute assignments to personal development. Trainees are expected to set the agenda for the conversation, but the supervisor can also add topics.

| Data collection
Between September 2016 and April 2017, we used convenience sampling to select a heterogeneous group of nine established pairs of Dutch GP supervisors and trainees affiliated with the GP training institute in Utrecht, the Netherlands. 19 The group differed in terms of the trainee's stage of training, the supervisor's age and experience, the length of collaboration between supervisor and trainee and the type of practice-solo, duo or health centre-in which they worked. Each pair was asked to video-record two learning conversations, both of which had to include at least one discussion of a medical topic, since that is how we defined an EBM dialogue. Since Conversation Analysis focuses on naturally occurring talk and naturalistic data the pairs received as little guidance as possible on which conversations to record and on the purpose of the study. 20 They were aware that the purpose of the study was to gain more insight on EBM learning at the workplace.
To ensure a diverse range of medical topics, we selected one discussion of a medical topic per learning conversation, producing a dataset of 18 medical discussions in total, which we transcribed verbatim. These discussions lasted between 5 and 20 minutes and showed how a trainee and a supervisor discussed a medical question.
The discussions selected, for example involved discussions regarding the appropriate medication for benign prostatic hyperplasia (BPH) or protocols on administering vitamin D supplements to the elderly.
The discussion started with the trainee asking a question or introducing the topic, followed by a dialogue on that question or topic, and concluded with a wrap-up or transition to another topic. We examined our dataset of 18 medical discussions for opportunities for bidirectional learning. Since learning conversations are traditionally seen as an opportunity for trainees to learn from supervisors, we defined bidirectional learning opportunities as moments presenting learning opportunities not only for trainees but also for supervisors, with trainees expressing their knowledge to their supervisor.

| Analytical procedure
Since CA is an inductive method, we started by taking an open approach to our dataset, focusing on the utterances in which trainees expressed knowledge. 21 We organised a data session in which ten CA researchers from various backgrounds commented on and analysed specific fragments, using the verbatim transcripts and anonymised video-and audio material. 22 We were then able to describe our phenomenon of interest in greater detail. CA focuses on how a conversation unfolds, turn by turn, emphasising that language is co-constructed and happens according to fixed patterns. 17,20,21,23 CA sees knowledge as a socially constructed process that occurs in interaction with others through mutual social actions. 20,[24][25][26] Since CA looks at previous and subsequent turns in the interaction (why that now 27 ), we began our analysis by considering the trainee's expression of knowledge. We subsequently focused on how the supervisor responds to the trainee's expression of knowledge, as this provides interactional learning opportunities. [28][29][30][31] We also included the trainee's utterance following the supervisor's response. Figure 1 depicts the interactional phenomenon, including the three steps that we analysed.
We then selected one fragment from each supervisor-trainee pair (nine in total) in which this interactional phenomenon was present. We transcribed these fragments in detail using Jeffersonian transcription conventions and analysed the fragments according to CA standards. 32,33 Individual case analysis was undertaken, with a focus on the design of individual turns at talk and the relationship between turns. 34,35 Grouping fragments that showed a similar response by the supervisor to the trainee's utterance of knowledge allowed us to form three sub-collections. 20,21 We were thus able to identify patterns in how supervisors respond to an expression of knowledge by a trainee and whether this response promoted bidirectional learning opportunities.
To verify whether the identified patterns also applied to other sequences, we returned to the 18 medical discussions as a whole.
Within these 18 discussions, two researchers (LW and LdC) looked for all sequences in which the trainee expressed knowledge. A total of 25 moments could be identified. All 25 moments fit within the three defined categories, leading us to conclude that the results and fragments presented in the Results section are illustrative of the 25 sequences in the complete dataset. The researchers discussed all analyses and conclusions at length and in detail within the research group. 36

| Ethical considerations
This study was part of a larger research project on EBM learning in the GP workplace. Approval for the research project as a whole, in which also Belgian GP supervisors and trainees took part, was

| RE SULTS
Three kinds of responses by the supervisor could be identified.
Supervisors (a) refute the expressed knowledge, (b) immediately suggest an alternative or (c) pose (additional) questions. These three kinds of responses will be described below, using fragments that are illustrative of the interactional phenomena.

| Refuting
The first response can be described as the supervisor refuting the expressed EBM knowledge (Box 1). In this fragment, the trainee and F I G U R E 1 Illustration of the interactional phenomenon that was analysed, including relevant sequences supervisor discuss the amount of levothyroxine that the trainee prescribed to a female patient. An explanation of the transcription symbols can be found in Table 1 below.
In lines 1-6, the trainee expresses her knowledge by reading aloud the guideline to account for her deciding on a dosage of 150 (line 6). While the supervisor confirms the explicit reference to the guideline in line 5, the trainees' conclusion in line 6 is refuted by the supervisor in line 8. This refutation is not directed to the specific, theoretical knowledge presented by the trainee in lines 1-4, but refers to how to interpret and use this information in practice.
The supervisor formulates his refutation in line 8 by starting with an agreement token (yes) and continuing with his own opinion, using the first-person perspective (I) and the verb 'think' ('I think that's a lot'), but without immediately explaining the source of his opinion. The addition of the word 'well' (line 8) illustrates a correction of the previous turns or information. 37 Following this refutation, the trainee responds with an accepting 'okay' (line 11) and proceeds to draw a conclusion about what to do in any future encounter (line 13), in this way accepting the supervisor's refutation.
The trainee gives a nervous laughter during this turn, demonstrating a tension or consciousness that the two ideas (supervisor's and trainee's) are not (as yet) aligned. 38 Even though the trainee has already accepted his suggestion, the supervisor proceeds to substantiate his opinion by first referring to the guideline, using 'they' ('what do they say again', line 20) and adding a description of his own approach (line 28: 'I often titrate up a bit') and mentioning this patient's specific situation (lines [29][30]. The supervisor concludes by suggesting in lines 33-34 and lines 36-38 how to proceed. He does with the phrase 'I can imagine' (line 33), implying that his approach might be an option instead of an obligation, but, on the other hand, also guiding the trainee towards a preferred response in the next turn ('yes') (line 35). 39 The trainee acknowledges these substantiations and suggestions multiple times with confirmatory responses such as 'no', 'yes' and 'okay' (eg lines 32, 35 and 39).
When the supervisor immediately refutes the trainee's expressed knowledge, a discussion ensues in which the trainee no longer accounts for or expands on her knowledge and the supervisor does not ask for additional knowledge, missing an opportunity for bidirectional learning. The trainee accepts the supervisor's utterances without asking additional questions to improve her understanding, while the supervisor holds on to his 'teacher' role.

| Immediately suggesting an alternative
Another way in which supervisors deal with trainees' expressed knowledge is to simply ignore it and immediately suggest an alternative. A detailed example is provided in Box 2, in which supervisor and trainee discuss a case handled by the latter in which a female patient with Parkinson's disease has sleeping problems.   41 The trainee responds to his suggestions by referring to the external source again: she repeats that she has checked the specific guidelines (lines [43][44][45][46][47] and it says 'just ordinary benzos'. By bringing up the guideline, the trainee appears to be appealing to an external authority and thus minimising the disagreement. 42 In the end, supervisor and trainee do not align; the trainee ultimately states that it's unnecessary to consult these specialists because they will reach the same conclusion. In this example, we see that ignoring the trainee's expressed knowledge by immediately suggesting alternatives prevents a knowledge exchange in which trainee and supervisor come to a shared decision or a general consensus on which knowledge is decisive, the medical guidelines or practical experience. They end up in a learning conversation in which they miss an opportunity for bidirectional learning. The excerpt presented in Box 4 shows a supervisor asking a question to seek clarification of a specific case rather than to get the trainee to elaborate on (theoretical) knowledge. Supervisor and trainee discuss the risk of inducing hypotension when starting ACE inhibitors and diuretics at the same time to treat hypertension.

| Posing questions
The trainee describes what he did during a recent case in lines 1-6. In line 7, the supervisor utters a declarative question about starting the two kinds of medication 'all at once'. In line 8, the trainee confirms that he was allowed to start these medicines 'all at once'.
Who allowed him to do so is not made clear. The supervisor rephrases the question in line 9, which the trainee once again confirms.

| S TRENG TH S AND LIMITATI ON S
This study uses Conversation Analysis to offer a fine-grained examination of the learning opportunities that arise during conversations between trainees and supervisors in the workplace. As a data-driven method with a basis in actual rather than idealised workplace conversations, the usage of this methodology can be seen as a strength. As Peräkylä (2003)

| CON CLUS I ON S AND RECOMMENDATIONS
Conversation Analysis of learning conversations between supervisors and trainees in general practice shows that bidirectional EBM learning opportunities are not always handled successfully.
Improving EBM learning opportunities for both supervisors and trainees requires more than simply instructing trainees to express knowledge based-for instance -on recent evidence more often.

ACK N OWLED G EM ENTS
The authors wish to thank all GPs and trainees that participated in the study.

CO N FLI C T S O F I NTE R E S T
The authors declare that they have no conflicts of interests. Esther de Groot https://orcid.org/0000-0003-0388-385X