Communication practices that encourage and constrain shared decision making in health‐care encounters: Systematic review of conversation analytic research

Abstract Background Shared decision making (SDM) is generally treated as good practice in health‐care interactions. Conversation analytic research has yielded detailed findings about decision making in health‐care encounters. Objective To map decision making communication practices relevant to health‐care outcomes in face‐to‐face interactions yielded by prior conversation analyses, and to examine their function in relation to SDM. Search strategy We searched nine electronic databases (last search November 2016) and our own and other academics' collections. Inclusion criteria Published conversation analyses (no restriction on publication dates) using recordings of health‐care encounters in English where the patient (and/or companion) was present and where the data and analysis focused on health/illness‐related decision making. Data extraction and synthesis We extracted study characteristics, aims, findings relating to communication practices, how these functioned in relation to SDM, and internal/external validity issues. We synthesised findings aggregatively. Results Twenty‐eight publications met the inclusion criteria. We sorted findings into 13 types of communication practices and organized these in relation to four elements of decision‐making sequences: (i) broaching decision making; (ii) putting forward a course of action; (iii) committing or not (to the action put forward); and (iv) HCPs' responses to patients' resistance or withholding of commitment. Patients have limited opportunities to influence decision making. HCPs' practices may constrain or encourage this participation. Conclusions Patients, companions and HCPs together treat and undertake decision making as shared, though to varying degrees. Even for non‐negotiable treatment trajectories, the spirit of SDM can be invoked through practices that encourage participation (eg by bringing the patient towards shared understanding of the decision's rationale).


| BACKGROUND
Shared decision making (SDM) '…is a process in which clinicians and patients work together…with the aim of reaching mutual agreement on the best course of action' (p.2). 1 SDM is advocated as an ideal model of health-care decision making 2,3 and is associated with better health-care efficiency, quality and outcomes and highly valued by patients. [4][5][6] However, implementation is not universal despite HCPs' claims to be doing SDM. 7,8 SDM involves engaging in decision making or plan-making collaboratively wherein both patient (and/or companion) and HCP contribute. We drew on the conceptual framework proposed by Entwistle and Watt 9 which extends beyond a focus on the "selection from a menu of health-care options" (p.276) and, therefore, is more broadly applicable to all decisions (ie spanning those with multiple reasonable courses of action and where there is only one course). It includes, but is not restricted to, recognition of patients' perspectives and contributions, being committed to a goal/activity, communicating significant issues and being informed.
Arguably, the concept of SDM has received more attention than its actual implementation in real-life health-care episodes. To help cast some light, we synthesised one body of evidence-that from conversation analytic studies of health-care encounters. Conversation analysis (CA) is a systematic and methodologically distinctive approach to studying interaction. It elucidates both the structural forms and the functional consequences of communication practices by studying recordings of actual interactions. 10,11 The recording process affects the interaction to some extent, 12 but considerable evidence suggests this does not preclude valid, useful findings. 13 CA does not try to understand communication by imputing psychological states; rather, it builds understandings of what people accomplish (together) through communication.
It is reasonable to understand all communications during healthcare encounters as integral to decision making. However, in this review, we purposely narrow the focus to commitment points: where it becomes relevant for patients to commit-or not-to a course of action (eg immediately after a HCP's proposal or suggestion). This is because decisions are internal matters that can only be gotten at through verbal claims and observable behaviours (ie commitment). We examine communication practices that happen during, shortly before and shortly after commitment points. CA research on communication in relation to health-care decision making is not comprehensive-some settings and decision types have been extensively studied, others minimally or not at all.
Communications included a variety of health-care matters: prescribing/altering pharmaceuticals, surgery, vaccination, psychotherapeutic or radiological intervention(s) or equipment; ordering/offering clinical/screening tests; setting therapeutic goal(s); and lifestyle adjustments. Our key objectives were as follows:

To identify communication practices entailed in decision making
in health-care interactions.

2.
To highlight patients'/companions' actions which contribute to their participation in decision making. Participation includes patients/companions having opportunity to discuss and/or influence decision making, having their points of view taken into consideration and/or opportunities for consultation and/or negotiation.

3.
To examine how HCPs' practices encourage and constrain participation.

| METHODS
We used an approach developed previously for systematically reviewing conversation analytic and discourse analytic research. 14 The rationale and process of this reviewing approach are described in a dedicated paper. 15 We used an aggregate approach to map findings across the structure that emerged (rather than undertaking a re-analysis).

| Study selection
One author (VL) undertook searching and initial screening of titles and abstracts and excluded publications clearly not meeting these criteria: 1. Audio/audio-visual recording of naturalistic health-care interactions with co-present patients/companions.

In English.
3. Both data and analysis examined broaching, considering, planning and/or deciding health/illness-related actions.

4.
CA as a primary analytic approach.

5.
Published in books or peer-reviewed journals (no date restrictions).
Remaining records (see Figure 1) were independently assessed by two reviewers (VL and RP); disagreements were resolved through discussion. PsycINFO; Scopus; and Sociological Abstracts CSA (Table 1 details search terms). Following contemporary guidance, 16,17 we searched additional sources: our own and other academics' reference collections; specialist bibliographies; and online discussion groups.

| Data collection, appraisal and synthesis
We developed, piloted and then used a customized data extraction form 15 to extract study characteristics, aims, findings relating to communication practices, how these functioned in relation to SDM and internal/external validity issues for study appraisal. We synthesized findings aggregately through discussion within the research team and via consultations with clinicians and researchers (both individually with academics/colleagues and also at seminars for sharing our work-in-progress).

| FINDINGS
Twenty-eight records  were identified (see Table 2 for publication characteristics). We organized findings in chronological order: actions prior to commitment point(s) being reached (termed "broaching"); putting forward a course of action (commitment becomes relevant); how patients convey commitment (or not); and HCPs' responses to patients' resistance or withholding of commitment. Table 3 summarizes the practices, their functions, and the settings and publications in which they were documented.

| Broaching decision making: actions occurring prior to any commitment point being reached
We term activities relevant to decision making but before a commitment point is reached broaching activities. Four ways of broaching were documented: flagging up that a commitment point is approaching; eliciting patient perspectives about decisions; encouraging patient agreement with proposals; and patient lobbying for a specific treatment/test.

| Flagging up
In four publications, 22,36,44,45 HCPs make an announcement to indicate an approaching commitment point. This "flagging up" does not stipulate any course of action; it encourages patients to move into the activity of deciding but does not push for one specific outcome.
Nevertheless, announcements can indicate other aspects, including whether there are multiple options or whose decision it is ( Figure 2).

| Encouraging patient agreement
There are several practices used in isolation or combination in service of achieving patient agreement with a not-yet-specified course of action while indicating the nature of that action. Six publications 20,27,28,33,37,38 document these practices (although three from one programme of research) from various settings: orthopaedics; oncology; neonatal ICU; and neurological physiotherapy. These practices include long turns, "brightside" formulations, logical inferences, general case descriptions and accounting.

Long turns
Orthopaedic surgeons projected longer turns through prefacing or requiring an element of the initial talk to be unpacked or by inserting parenthetical talk, thereby minimizing a patient's opportunity to disagree with the parenthetical information. 20 These long turns, found in non-surgical recommendations, allow surgeons to "concurrently manage multiple (competing) contingencies and actively work to anticipate and pre-empt possible problems with the action underway" (p.397).
"Brightside" formulations Across the 28 publications, the settings included were as follows: eight primary care, three neurorehabilitation, three orthopaedic clinics, two oncology clinics, two outpatient clinics, two antenatal clinics, two mental health, two outpatient epilepsy, two ICUs (one neonatal), one ENT oncology and primary care diabetes, and one dietician clinic. A HCP may specifically rule out a particular option. This is generally less straightforward-both its design and reception-than affirmatively putting forward a course of action. By ruling out, the HCP produces the treatment that is ruled out as known to the patient, expectable and also possibly the preferred treatment option.
physiotherapy setting, therapists sometimes use accounts when the upcoming recommendation is counter to patient's expectation or report of current activity. 33 In sum, these practices forecast and strengthen the recommendation/suggestion prior to its production. Used particularly when the recommendation is liable to resistance or counter to patients' expectations, they function to achieve patient agreement in a potentially challenging environment. Indeed, resistance was more likely when forecasting activities were absent. 20

| Patient lobbying for specific treatment prior to commitment point
Prior to HCPs referring to specific courses of action or making commitment relevant, patients may reference a particular course of action to seek pre-emptively to influence the treatment trajectory. Documented in five publications, 18,25,26,41,42 this is the only patient/companioninitiated broaching activity described in the included publications.
If a patient knows a diagnosis projects a treatment, challenging the diagnosis may be a way of lobbying for a desired treatment. In consultations with children with upper respiratory illnesses, parental resistance to a viral diagnosis may be a resource for resisting the projected non-antibiotic treatment. 42 This pre-emptive subtle influence was also identified in two single-case analyses. In a hospital outpatient clinic, by inquiring about the availability of a test and describing a previous positive experience in a similar situation with the doctor's predecessor, the "patient exerts subtle but persistent pressure for a diagnostic test" (p.451) before the doctor's recommendation. 25 By lobbying, patients position themselves as having a role in determining the decision. Nevertheless, generally this pressure is applied subtly (attentive to being heard as possibly treading into HCPs' territory 41 ) and designed not to oblige the HCP to offer or decline the lobbied for treatment/test.

| Putting forward the course of action (the commitment point)
The next phase-although decision making may begin here if there are no broaching activities-is putting forward or ruling out possible paths of action. This is a commitment point as it obliges the patient to make or (implicitly or explicitly) avoid commitment. This activity is solely within the HCP's domain in the studies reviewed.

| Single option
The most common way HCPs reach a commitment point is by putting forward a single course of action (27 of 28 publications); these practices are imbued with varying levels of assumption that the patient should/will follow that course of action. A HCP may make an explicit recommendation (Extract 6) or even build in presumption of agreement (Extracts 7-8).
[taken from Excerpt 6 in, 23  A HCP may produce a suggestion (eg "my suggestion would be…") 45 which conveys their stance but reduces their authority to require the patient take the particular action. They may be structured to indicate a shared decision (Extract 9) or entirely the patient's choice (Extract 10).
[taken from, 31  Putting forward a single path-however openly phrased-is likely to be heard as HCP-endorsed (see 34 ). Sometimes a HCP may be able to offer a single option only (likely to vary significantly depending on setting) and doing so does not preclude recognition of patient Non-lexical occurrences autonomy. These turns bring the interaction to a commitment point.
Patient agreement here is sufficient to reach a decision. Although interactionally more difficult, patients may reject the course of action.
This difficulty may be compounded by the format used (eg it is more difficult to reject announcements than suggestions).

| Ruling out a single option (primary treatment)
HCPs may specifically rule out a particular option (eight publications). 18,20,[27][28][29]39,41,42 These appear less common than affirmative recommendations: in one study, initial recommendations against a treatment were found in 29 of 309 consultations (compared to 252 initial affirmative recommendations). 39 Compared with affirmative recommendations, a narrower spectrum of formats are used to rule out including "I don't think we need to…"; "I don't recommend…"; and "you certainly do not need…." An example from paediatric acute care shows a whole class of treatments being ruled out with "I don't think we need to put her on any medication." 39  Seven of the eight publications show ruling out occurring in environments in which there is an orientation to a primary treatment (eg surgery, antibiotics). The ruled-out treatment is treated as known to the patient, expectable and possibly preferred. Sometimes HCPs also offer an affirmative alternative (treated as less preferable). They usually occur after the rule out and as a result of the patient's response to it rather than designed that way from the outset. 39 Often HCPs engage in activities for seeking agreement prior to ruling out the primary and/or offer an alternative to create an auspicious environment for patient agreement. The rule out might be produced as a temporary decision (Extract 11) which preserves the primary as a possible future option. Indeed, in an orthopaedic surgery consultation, the rule out is achieved by referencing surgery as on offer in the future ("delay your surgery"). 20 Ruling out a course of action is less straightforwardboth its design and reception-than affirmative recommendations.

| Multiple options
Less frequently-primary finding in five publications, 22,36,38,44,45 subsidiary in one 30 -HCPs put forward multiple options from the outset (rather than offering options in response to withholding commitment to a single option). This practice (ostensibly) provides clear opportunity for patient participation. Usually, HCPs announce that multiple options are about to be listed, perhaps because otherwise the recipient might be primed to hear the first option as a single option.
Multiple options tend to be presented with multiturn units detailing benefits, risks, effects or rationale of each of the options and with opportunities for patient responses. After the list, the HCP may elicit the patient's view (eg "what do you think"). 45 The options may be fairly neutral or may display-strongly or weakly-a stance. In a neonatal ICU, some options were more persuasively presented than others. 38 In mental health consultations, a psychiatrist flagged up three choices, discounted the first two before producing the third: the structure of The upcoming decision may be flagged to indicate the patient's participation is expected and that there are multiple options available.
Examples: 'we've got a couple of choices' (diabetes consultation) [22] 'we've got three choices' (psychiatric consultation) [36] The upcoming decision may be flagged to indicate multiple options without indicating patient involvement (although this is not precluded either) Example: 'there's choice' (neurology consultation) [43] Does not indicate multiple options There were no examples which included the patient but that did not suggest multiple options were available.
The upcoming decision may be flagged without indicating that patient involvement is anticipated and without suggesting there are multiple options Examples: 'it means changing your treatment' (diabetes clinic) [22] 'you need to have something done' (oncology clinic) [22] offering multiple options can thus be a vehicle to recommend a single path. 36 Another study showed that in six of 15 ICU cases, the presentation of options was "shaded": not all options were present or physicians' preferences were strongly indicated. 30

| Committing
The nature of the initiating action shapes what constitutes a relevant next action. A single option makes relevant commitment (or avoiding commitment), and lists make relevant a selection. 44 When a single option is put forward (as is most common), patients/companions and HCPs jointly treat patient commitment as the necessary next action.
At the commitment point, all parties treat patient involvement as crucial, although this can involve very short utterances. Commitment involves accepting rather than merely acknowledging: "treatment recommendations are routinely accepted with objects such as period intoned 'Okay.' or 'Alright.'; 'Let's do that.'; 'That's fine.'; and assessments such as 'Good'" (p. [46][47]. 40 Patients may produce themselves as involved even if they say very little. In addition, patients may implicitly commit by continuing to the next activity. In oncology consultations, a patient's implicit agreement is shown by moving to a question about treatment location, however, 'very rarely…do these unfold with so little input from the patient' (p.88). 37 The severity of the condition and complexity of decision may have a strong bearing on this. In an ICU setting, consensus regarding decisions pertaining to removal of life-sustaining equipment was a topic in its own right. 30 Acceptance is only sufficient when a single path has been put forward. In cases of multiple options, the relevant next action is selection from the list. However, patients may challenge the option-listing format by seeking a recommendation instead. 44 In some cases, commitment to a course of action is not required in that interaction. Three oncology consultations were examined in which decisions regarding adjuvant ther- apy were left open. 37 In these cases "visits are treated as opened-ended sessions in which there is no expectation for an on-the-spot decision" (p.102).  Nine 19,23,29,31,37,39,40,42,43 of the 28 publications explicitly discuss how patients/companions withhold commitment, often referring to this as "passive resistance, " with a further 12 publications 18,20,27,28,30,32,33,36,38,41,44,45 making implicit reference. Withholding commitment obliges HCPs to stay within the decision making phase however, if commitment is still not achieved, patients may move to "active resistance."

| Active resistance
After a commitment point has been reached, the patient/companion may question or challenge the proposed course of action. In the publications reviewed, this "active resistance" occurs as an escalation from initially withholding commitment. Two main practices for actively resisting were identified.

Questions/concerns
First, patients may raise questions or concerns about the medical problem or the proposed treatment/plan (six publications). 19,29,31,37,38,40 Extract 12 shows a patient questioning the diagnosis (lines 19-21, 27-29 and 31-32) and raising a previously unarticulated concern (lines 45-46). 29 These instances of active resistance often indicate the nature of the barrier to commitment.
[taken from

| Pursue agreement without changing course
Eight publications 18,20,29,31,32,36,37,40 document pursuit of commitment after resistance. In five of these, pursuit involves treating patients' problems as obstacles to overcome before agreement is achieved.
Patient participation is evident despite the unchanged treatment trajectory as patients "postpone acceptance until their treatment preferences and concerns are satisfied" (p.1110). 29 The remaining three publications show HCPs pursuing commit- In another of these three publications, a doctor pressurizes a patient to commit to a medication change in a mental health consultation. 36 The persuasion is strong, the patient orients to it as pressure, and their eventual agreement is grudging. Yet the analysis shows that the patient engages in activities (eg retrospectively orienting to doctor's recommendation as advice, reluctant agreement) which convey that a shared decision is taking place.

| Modify the potential course of action (pursuing agreement by changing course)
HCPs may modify the course of action (five publications). 23,29,36,40,41 Taken from an oncology consultation, Extract 14 demonstrates a dosage recommendation modified to a recommendation to "work up to" that dosage and furthermore only "if you can": [taken from Excerpt 6 in, 23

| Leave the decision open
Documented in four publications, HCPs may attend to patients'/ companions' resistance by leaving "open" the decision, either by deferring it until another time 23,30 or by offering to review/revise at a later date. 31,40 In an oncology consultation, resistance to gall bladder removal resulted in a recommendation to "think about" surgery in the future, 23 and in an ICU, lack of consensus about withdrawing a relative's ventilator resulted in deferral until agreement can be reached. 30 Two publications show doctors offering the option of revisiting the decision if it turns out to be unsuitable. 31,40 The offer of revisiting the decision is in response to resistance, that is, these are distinct from interactions that are designed from the outset as not requiring commitment during the encounter or framed from the outset as temporary.

| Limitations
The categories identified are broad because the 28 publications span a wide variety of conditions: acute to chronic; minor to life-threatening; those with multiple treatment paths to those for which protocols or urgency dictate one path. The evidence-while rich in detail-is concentrated in specific areas (eg acute care, particularly antibiotic prescription; orthopaedic surgery; oncology). Practices may operate differently depending on the setting (eg ruling out) which can be problematic for transposing findings to other settings. Therefore, we offer the practices presented here as a mapping rather than a definitive structure. However, despite differences, common activities exist.
We attempted to weave appraisal findings into our syntheses but it is difficult to assign relative weight to CA studies' findings due to their qualitative nature and associated small data sets.
For practical reasons, we included English language publications using English language data only. This is disappointing as several excellent studies using non-English language data would contribute to the findings, for example Norwegian work exploring differences across different health-care settings 48 or analysis of Swiss physiotherapy data which contributes to understanding goal elicitation. 49 Similarly, we did not include grey literature.

| Applications for practitioners
By lobbying for specific treatments/plans, patients produce themselves as having a role in determining the decision. The subtlety of this lobbying orients to the delicacy of potentially stepping into the professional's domain. Patients' requests therefore may not look like typical requests, and practitioners can be responsive to this without having to grant the request while also providing reassurance. 26 Patients' resistance often provides opportunities for HCPs to address specific problems, thereby treating them as involved participants.
Patients are skilled at doing this in ways that avoid confrontation, and it would be beneficial to HCPs to be able to recognize these. 19 Eliciting patient perspectives and ensuring that information is genuinely taken into consideration generally result in patients expe- Recognition that patient resistance is a resource for participation means that using the interactional slot after resistance to invite patients to collaboratively construct an acceptable decision is "a candidate best practice" (p.1111). 29 Exploring patients' reasons for resistance-even when protocol means there is no alternative-validates patients' participation. Even where the patient eventually agrees to the original recommendation, where reasons are explored, they will have still participated in the decision making process. Pursuing agreement without engaging with patients' reasoning for withholding is less encouraging of patients' participation and may be treated as coercive.
Where the option for modifying recommendations is possible, this allows for greater patient participation in terms of influencing the final decision. However, as patients/companions become increasingly proactive in their health-care, HCPs balance the encouragement of participation with the importance of need to not being pressured to give inappropriate treatment. 41

| Future research
Finally, we discuss three opportunities for future research. First, HCPs give treatment and decision relevant information at various points (eg prior to recommending, during offering single/multiple options, after patients withhold commitment) but this has received limited attention as a phenomenon in its own right. This is particularly important as information sharing is central to patient participation. Second, existing studies (particularly those with extensive data sets) have been concentrated in a few specific areas, for example primary care. Given that setting and condition can shape the operation of these practices, it would be valuable to explore a range of secondary care settings and also settings in which successful outcome is arguably more subjectivesuch as maternity care, palliative medicine or plastic surgery. Third, the actions that we have outlined here may be achieved by a range of practices that has not yet been fully documented.

| CONCLUSION
Decision making encompasses more than the turn in which a course of action is put forward and patient's immediate response to it.