The narrow conception that we construct and critique is not intended to represent any particular published definition or model of shared decision making in its entirety. Rather, it is a set of propositions that have been advanced or implied, separately or together, in various writings about, and efforts to promote, shared decision making (see refs 3,4 for reviews of definitions):
In summary, our critique is that proposition 1 is unduly simplistic, and in ways that undermine the rationales for shared decision making represented in proposition 2, and that lead to inappropriate limitations on professional–patient engagement as represented in proposition 3. We contend that in some instances it can be respectful, serve good ends and be otherwise ethically legitimate for health professionals to question, challenge and influence the preferences of patients (including both ‘evidence informed’ preferences for particular interventions and expressed or apparent preferences for not discussing their options and preferences). Phrased in these carefully qualified terms, we do not imagine this contention will seem unreasonable. It may even seem non-contentious to some people. It does, however, involve taking issue with the propositions listed earlier and hence with some current understandings of shared decision making.
We consider each of the above propositions in turn. To avoid overcomplicating our discussion, we talk as if about single ‘discrete’ decisions and simple health professional–patient dyads. We recognize that, in practice, decisions occur in more complex clusters and decision making is more widely socially distributed.5 The points we make still apply when this complexity is acknowledged.
Eliciting patient preferences
Proposition 1 has two elements: that patients have preferences and that, in the context of clinical decision making, it is the job of health professionals to elicit them and take them seriously. (We are also using ‘preferences’ as shorthand for ‘beliefs and ‘values’ etc.) It would be foolish to assert that patients never have preferences or to deny that some of their preferences are relevant to clinical decision making. People often have preferences relating to particular health-care interventions and/or their possible outcomes (including their implications for biography and well-being). These are widely recognized as relevant to decision making, as are preferences relating to risk aversion or risk taking. So-called ‘meta-preferences’6 relating to processes and styles of professional–patient interaction and deliberation and decision making are also likely to be salient, and we, of course, agree that, where such preferences exist, it is important that they are reflected in health-care provision. This will usually require that health professionals have the skill and inclination to discover them and take them into account.
However, it is only sometimes the case that patients have clear and stable decision-relevant preferences or meta-preferences that are consistent both with each other and with the way the world is. Both illness and health-care trajectories present patients with novel and evolving situations in which it might be properly said that they do not know what they prefer or that they have multiple, and possibly competing or even conflicting, preferences. In addition, the preferences patients develop may be based on a lack of understanding, or misunderstandings, about their health states or possible interventions.7–9
In these circumstances, it is misleading to think that health professionals who seek to take preferences into account need simply to listen to patients’ expressions of preference or skilfully ‘elicit’ their pre-existing preferences as if they were hidden (unmentioned) nuggets, and then to feed these unproblematically into a decision process.
In other words, proposition 1a needs broadening. It makes sense to think of patients either having or developing or ‘constructing’ preferences, that may be more or less well-founded and that may not form a single coherent set. If the goal is to make choices on the basis of reasons (including preferences) that will be sustained on reflection,10 people may need to find ways of ‘checking’ their preferences, identifying which are likely to be more longstanding, or more ultimate (or which represent their ‘deeply held values’)11 and find ways of making trade-offs or otherwise reconciling different preferences (including different kinds and ‘levels’ of preferences).
This suggests the need to make a knock-on broadening modification to proposition 1b. It may be (part of) the job of health professionals to help patients construct, check and prioritize their preferences in domains important for decisions for which health professionals have some responsibility7– although both the need for and appropriateness of particular approaches to such ‘co-construction’, preference elicitation and/or values clarification is likely to be context dependent.8 We stress that we do not take the language of ‘co-construction’ to mean that professionals and patients should have an ‘equal say’ in what counts as patient preferences; nor do we want to lend credence to any idea that professionals have a licence to ‘re-engineer’ patients’ preferences. What we envisage is essentially a facilitative role. We are acknowledging that there may be circumstances in which it is appropriate for professionals to explore carefully, check, encourage reflection on and, in that sense, ‘co-create’ preferences.
Why patient preferences matter
Proposition 2 asserts that patient preferences ought to be taken seriously for both intrinsic and instrumental reasons – because doing so respects the autonomy of patients and because it provides data that are useful, or even necessary, for good decision making. It follows from what we said about proposition 1, however, that this second proposition needs careful consideration. Not all preferences seem to be of equal importance, and if we have any reason to suppose that preferences may be unstable, given inappropriate weight in decisions, or susceptible to being ‘trumped’ by other preferences – perhaps ones that are currently less conspicuous – then both the intrinsic and the instrumental rationale are problematic.
The ‘respect for autonomy’ rationale needs further consideration for other reasons too. As Rebecca Walker12 has argued, what really matters in health care is respect for autonomous persons. In normal cases, this will entail respecting the preferences and choices of persons with capacity for autonomy, but this is not because health professionals are engaged in, or are necessarily capable of, discerning which preferences and choices are themselves ‘autonomous’. It may well be that some of the preferences and choices of patients reflect some form of ‘compulsion’ or ‘weakness of will’ or otherwise fail to be ‘autonomous’. If we insist on the importance of respecting such preferences and choices, it is because we decide that the demands of respect for persons as autonomous override our concerns about any shortcomings of these specific preferences or choices. We are not, that is, ethically bound to respect preferences as in themselves expressions of autonomy,13 and there may be occasions in which the best means of respecting the person (and supporting their capacity for autonomy)14 is by questioning or even challenging their expressed or apparent preferences – as long as we stop well short of coercing patients or forcing particular choices upon them.
The point of questioning or challenging patients’ expressed or apparent preferences would be to help them develop and check their own preferences as discussed under proposition 1. For example, a clinician might help a patient view their clear and stable dislike of injections as something arguably comparatively trivial with regard to their overall life project and autonomy in a particular situation and help them develop preferences about more fundamental treatment directions with potentially life-altering implications in the longer term. More generally, professionals can, through experience of similar cases, be aware of factors that may come to seem salient to patients at future times and have insights into common misconceptions about possible future events and states.15 By sharing their awareness and insights, they may help avoid the obvious dangers of just focussing on any immediate preferences that are proactively voiced by or easily elicited from patients.
Thus, proposition 2a can only be sustained if we add a broader notion of what might count as both respecting and supporting autonomy to a more problematized notion of what counts as identifying preferences.
Broadly, the same considerations apply to the instrumental justification in proposition 2b as to the intrinsic one in 2a: we can only decide on the value of preferences by discriminating between them. Patients’ preferences often provide useful data for improved decision making because they can indicate what counts as a good treatment or outcome from particular patients’ points of view. However, while it is rational for health professionals who want to make good decisions to investigate and pay attention to what their patients seem to value or want, not all preferences are ‘fit for purpose’. To be instrumentally valuable for decision making, preferences must have a reasonable degree of predictive validity, that is, provide a good indication of how patients will come to evaluate the interventions and possible consequences that the preferences are about. In practice, it would, no doubt, be seen as high handed and disrespectful for doctors to discount preferences that they have reason to believe would fail this test but, purely from the instrumental perspective of proposition 2b, if it were possible to make these judgements correctly and confidently, this discounting could be warranted.
When and what to negotiate
Proposition 3 relates to when health professionals should negotiate with patients and what they should negotiate about. The idea of professionals working with patients is clearly a crucial dimension of shared decision making. However, the processes of collaboration and/or negotiation that feature in some narrow conceptions of shared decision making are heavily circumscribed, and this, we think, can be problematic.
Proposition 3 suggests that health professionals and patients should discuss health-care options and deliberate about them together (a) only if the patient wants to and (b) by seeking to make a decision that reflects both research evidence about the effects of different options and the patient’s preferences (as informed or modified by patients in the light of the evidence base).
The (quite strong) emphasis that is sometimes placed on condition (a) reflects the importance that is often attached to respect for preferences as expressions of patient autonomy. Some accounts suggest that patients’ meta-preferences relating to approaches to decision making are, or should be, considered first and/or given simple lexical priority over their substantive preferences for particular interventions or the outcomes and probabilities associated with these.6,16 For the reasons we rehearsed when considering proposition 1, this is problematic. Even leaving aside the considerable practical problems of eliciting (and ordering) meta-preferences, there seems no obvious reason to treat them as any more well-founded, stable and non-negotiable than substantive preferences. And there are considerable dangers in doing so if certain expressed meta-preferences are responded to in ways that preclude the possibility of attending to substantive preferences about intervention options and outcomes.
As with expressed preferences about intervention options and outcomes, however, there are other good reasons to accept patients’ expressed meta-preferences. It is generally likely to be more courteous and more effective for a health professional to go with the grain of what the patient wants in terms of communication and engagement – although as explained in relation to proposition 2a, this acceptance is not always required by respect for autonomy. Expressed meta-preferences might reflect a lack of awareness that there are several reasonable courses of action and that the question of which is ‘best’ depends on what an individual patient thinks about a particular issue. They might also reflect a lack of experience of having been given information and support that could enable meaningful participation in discussions. Both possibilities suggest further scope for autonomy-supportive checks or ‘challenges’ from health professionals to initial and apparently undeliberated meta-preferences – especially those that tend to work ‘against’ patient participation. However, in flagging up the potential defensibility of such checks and challenges, we do not wish to imply that enhanced participation is always to be preferred or that it does not have possible negative consequences.17
We have argued elsewhere that the nature of the negotiation or joint deliberation advocated in shared decision making is relatively unspecified,18 but some specifications do emerge if we attend to the rationale behind narrow conceptions of shared decision making, as, for example, represented by proposition 3b. Such accounts rest on a particular division of labour between professionals and patients. They envisage each party bringing something specific to the consultation/negotiation, playing a defined role in it and leaving it with their independence intact. The usual assumptions are that the health professional has relevant specialist knowledge (including of research evidence about effectiveness), the patient has his or her preferences (including hopes and aspirations etc.), and three overlapping kinds of transaction occur. First, both parties share or pool their respective insights. Second, the patient may revise their preferences in the light of the specialist knowledge shared with them, and on hearing more about the patient’s preferences, the professional may need to identify and discuss more relevant options and associated evidence. Third, the two parties agree on a course of action that properly reflects both sets of insights. (On some accounts, failure to reach such an agreement can still be compatible with shared decision making but may necessitate the patient seeking another professional to work with.)2
According to proposition 3b, it is not part of a health professional’s role to question, challenge or influence the patient’s preferences except insofar as the provision of disinterested impersonally constructed evidence might act as a cognitive stimulus or ‘challenge’ to patients’ initial preference sets. The considerations mentioned above, however, suggest that a broader role may be appropriate.
Broader conceptions of shared decision making: thinking about relationships
In discussing broader conceptions of shared decision making, we do not intend to specify a particular set of propositions or to build a specific model. Rather, we are exploring what might ‘lie beyond’ some of the possible limitations of narrow conceptions. Nor are we asserting that broader conceptions are inherently better, in some unqualified sense, than narrow conceptions. Rather, we wish to defend the relevance and importance of broader conceptions in principle, while acknowledging significant problems with their potential operationalization.
It is striking how different, and how much more restricted, the conception of dialogue built into narrow conceptions of shared decision making is from other familiar examples of dialogue. In areas where roles are less clearly defined – most obviously in friendships and also, for example, in parent–child and some teacher–learner relationships – it is often seen as desirable for both preferences and meta-preferences to be questioned and challenged. Indeed, in many instances, the relationship can be properly viewed as a context for exploring and questioning preferences such that we might say that one fails to be a good friend, parent or teacher unless one succeeds in doing this.
Of course, we recognize that doctor–patient relationships are different from these other relationships. Our consideration of these other kinds of relationships does not rest upon the notion that what is possible in them could be directly ‘translated’ into health-care arenas. Rather, it stems from the desire to identify some comparator cases that illuminate what ‘dialogue’ and ‘shared decision making’ within relationships can mean. As we have indicated earlier, a good case can be made for exploring broader ideas about shared decision making in health care encounters too.
There is support for this suggestion in the wider literature on clinical relationships and ethics. In their classic analysis of ‘four models of the physician–patient relationship’, Emmanuel and Emmanuel,19 who also make comparisons with friendship and teaching, advocate ‘the deliberative model’ as ‘Descriptively and prescriptively’…’the ideal physician-patient relationship’ (p2225). This model involves doctors challenging and being prepared to co-construct patient preferences and could be thus considered to provide a notable example of a broad conception of shared decision making.
The examples of friendship and teaching illustrate that dialogue within relationships has fundamentally important purposes that extend well beyond the kinds of transactions envisaged within narrow conceptions of shared decision making. In our conversations with our friends, children or students we do not confine ourselves to helping them to articulate their preferences or to questioning the factual assumptions lying behind or shaping their preferences. Rather – to simplify – we try to combine an emotionally supportive acknowledgement of ‘where they are coming from’ with a critical engagement with the things they affirm – including their expressed preferences. In doing so, we constantly have to walk a careful line, but if our critical engagement is conscientious, we do not believe that it is a threat to their autonomy. Rather the opposite: it is because we care about them, and take their present and future autonomy seriously, that we are ready to take issue with what they say they want. In these cases, we are unquestionably concerned that the conclusions and choices that our friends, children and students end up with are very much their own but this does not rule out a willingness to help them construct, reconstruct and co-construct their preferences through dialogue. It is in the nature of these exchanges that we risk ‘invading the territory’ of a friend, child or student, but because we are concerned to ‘meet’ them as people and not just as ‘contractors’, we aspire to be responsive to them with the full range of humanity and not just with practical or technical expertise.
Taking this thought further, we can see that the contrasts and tensions between narrower and broader conceptions of shared decision making arise in large part from what we might label the ‘open-endedness’ of relationships. Friendships are paradigmatic of ‘open-endedness’– a friend (or at least a good friend) is involved in many areas of our lives, over a long period, and is typically trusted to get involved in, and participate to varying degrees in, key events and choices in our lives. Although we can draw boundaries around what we want to involve our friends in, a key characteristic of friendship is that these boundaries are not prescribed by the role, are not socially defined or clear-cut and are often quickly dissolved when friends need help and support. (Parent–child relationships have many similarities but often lack the key axis of ‘broad equality’ characteristic of friendships. In that respect, parent–child relationships also have parallels with the power dynamic in some professional–patient relationships and hence the notorious threat of paternalism associated with the latter).
Clinical relationships are much less open-ended than friendships, but there are different degrees of open-endedness within them, depending upon various dimensions including the time period over which the relationship could and does take place (a known professional supporting the long-term management of a chronic condition vs. a one-off appointment with an ‘unknown’ professional); the flexibility/fixedness of role boundaries in the relationship (a GP taking blood and chatting, vs. a dedicated hospital phlebotomist taking blood and chatting); the depth or penetration of probes and discussions (for example, exploring more fundamental value orientations rather than focusing on preferences at quite superficial levels); and the scale or length of the ‘horizon’ which frames what is at stake (e.g. short-term vs. whole ‘future life’ considerations).
Generally speaking as we move from less to more open-ended relationships, two things change: first, the individuals involved become less ‘substitutable’; second, the relevant notions of expertise broaden out. These two factors overlap but can be distinguished. The non-substitutability of our friends can be partly understood as an expression of their commitment to us (and vice versa), and partly by their richly informed knowledge of us, as people.
In narrower conceptions of shared decision making, which rely on a relatively fixed and constrained view of the appropriate scope and function of dialogue, and which limit health professionals’ roles to the sharing of impersonal technically decision-relevant knowledge, professionals are plausibly substitutable one with another. When ‘non-interference’ is emphasized (or required), substitutability can be seen as an indicator of success. However, when the kinds of support associated with more ‘committed’ and open-ended relationships are valued, substitutability can be taken to indicate a deficit.
In less open-ended relationships, it is possible to conceive of the knowledge being practised as essentially technical knowledge and of the practitioner as essentially following a protocol which anyone with the same kind of technical expertise could follow. As relationships become more open-ended, more factors come into play, including an indefinitely large number of considerations connected to the personhood and life circumstances of both parties. Here, it is quite common to refer to the need for intellectual and moral ‘virtues’– the qualities of character that enable us to respond to, and wisely balance together, multiple and often competing considerations.20
Considerations of principles in practice
If we simply consider idealized and ‘perfectly virtuous’ professionals, then it is relatively easy to argue for the value of broader conceptions of shared decision making. Wholly capable, scrupulous and wise professionals could help patients to formulate, reflect upon and adapt their preferences (and underlying values) as part of genuinely shared decision making and do so without being inappropriately swayed by professional or managerial norms and without undermining (and indeed rather by supporting) patient autonomy. But, of course, not all actual people (including health professionals) are ‘perfectly virtuous’– and just about all actual people (with the possible exception of saints or other great leaders) are subject to pressures from institutional norms. Furthermore, substantial power imbalances exist in professional–patient relationships, and these, along with the absolute or relative vulnerability of patients, are, rightly, often cited as threats to the autonomy of patients.
It is in this context, and with the desire to protect patients from either straightforward abuses of power or more well-intentioned forms of paternalistic pressure, that narrow conceptions of shared decision making make sense. However, the limitations of narrow conceptions of shared decision making also mean that such protection comes at a price. While the divisions of labour within narrow conceptions aim to protect patients from power by avoiding the risk of unwarranted interference, the narrow conceptions of shared decision making – at least as expressed in propositions 1–3 above – shut off from patients some of the most important health-care resources available, namely those that arise from the personhood of professionals and the broad potential of supportive human relationships.
We are thus left facing the longstanding and recalcitrant conundrum that gives rise to the various competing ideals of health-care relationships. How can the full resources of the (typically relatively powerful) professional be made available to serve the interests of the (typically relatively vulnerable) patient without those interests being compromised as a result? How can patient’s clinical decision making be supported by the professionals involved in their care and protected from those professionals at the same time? ‘Agency’ models of professional–patient relationships acknowledge the conflicts of interest that arise for health professionals when their employment and broader social responsibilities are taken into account, but these are not readily compatible with the kinds of partnership aspired to in shared decision making.21
We have suggested that broader conceptions of shared decision making provide a useful ‘corrective’ in principle to the problems of narrower conceptions by allowing for more open-ended and fully dialogical ways of relating. In these broader conceptions, the ethical compromise between unduly paternalistic and unduly consumerist approaches depends essentially on the skills and virtues of individual health professionals and on the creation of climates that can support the development and exercise of both professional virtues and patient capabilities. In practice, broader conceptions of shared decision making would clearly make great demands on health professionals as they work within particular clinical encounters. The tensions between broader and narrower conceptions of shared decision making can therefore partly be seen as tensions between ‘principle’ and ‘practice’.
Even in Emmanuel and Emmanuel’s skilful advocacy of the deliberative model, there are clues to the substantial problems of operationalizing broader conceptions of shared decision making. First, while they are happy to propose that doctors should be able to evaluate patient preferences and to attempt to persuade patients about the ‘health-related values’ that ought to apply from case to case, Emmanuel and Emmanuel19 stress that doctors should (i) stay on side of reasonable persuasion and not stray into coercion and (ii) confine themselves to persuasion about ‘health-related values’ and not broader values. Second, they argue that this model requires us to ‘implement change in medical care and education’ because ‘it may well be that many physicians currently lack the training and capacity to articulate the values underlying their recommendations and persuade patients that these values are worthy.’(p2226) We all know, for example, from personal experience of ‘counselling’ friends faced with health-related decisions, that it is not easy to find the balance between reasonable advice and unreasonable pressure. But things are much harder for health professionals following a broad conception of shared decision making. They are not only legitimate parties in clinical decisions, but also socialized into responsibility for such decisions and are legally accountable for their role in decision making. In principle, we might argue that health professionals should be able to respect the distinctions between health-related and other values (assuming this is a meaningful and a clear enough distinction) and between persuasion and coercion (or as we would say between autonomy-supportive and autonomy undermining forms of communication).14 But in practice – as Emanuel and Emanuel suggest – we cannot reasonably expect all health professionals to have (and consistently apply) the knowledge, skills and inclinations needed for these extraordinarily difficult balancing acts. The challenge is therefore to know what to say about the possibility of broader conceptions of shared decision making in policy terms: are these ideals largely irrelevant to the ‘real world’ or do they capture something about the direction in which we ought to be heading?