The shared decision-making model

Medical decision-making is an established field with its own textbooks, a professional society and several journals; shared decision-making is its central paradigm. While the medical and philosophical literature vary considerably in their understanding of the type of doctor–patient relationship in which medical decisions should be made,5 there is consensus about the major common elements of shared decision-making.3 The practice is built on the doctrine of patient autonomy, while acknowledging the valuable contribution that clinicians can make based on their professional expertise. One representative definition describes shared decision-making as an approach in which “the patient receives information about the available…treatment options, including the associated benefits and risks. The clinicians and patient then consider these options in light of the patient's circumstances, goals, and preferences. Working together, they then select the best course of action.”6

The process of shared decision-making has been studied empirically,4 analysed philosophically7 and dissected psychologically.8 The result has been the discovery of a lengthy list of barriers to its implementation and a litany of complaints about its feasibility.9 While strategies to surmount each barrier have been proposed10 and approaches to enhance practicality advocated,3 the cumulative number of fixes needed is enormous, resulting in a process that is unwieldy at best.

Barriers to shared decision-making

The barriers to shared decision-making include patient factors, physician factors and system factors. In the patient arena, perhaps the most profound obstacles arise from cognitive biases. First identified by Tversky and Kahneman as applying to decision-making in general8 and subsequently applied to medical decision-making, these include anchoring, framing and availability biases. Patients are more influenced by the first piece of evidence offered than by subsequent data; they are influenced by the way data are presented (a treatment that confers an 80% chance of living is more likely to be chosen than a treatment described as conferring a 20% chance of dying, even though they are mathematically equivalent) and patients are affected by their recent experiences, even if these are not representative. The solution to patients’ use of shortcuts in making decisions about complex issues is to present the information in multiple distinctive ways, with the expectation that the combination will achieve balance and neutrality.10

Shortly after cognitive biases were identified, researchers became aware of innumeracy. Patients did not merely resort to heuristics for psychological reasons; they made flawed judgements because of lack of familiarity with graphs, probability and statistical reasoning.11 Innumeracy proved to be widespread, though it, too, could be surmounted with specially designed decision aids that rely on simple pictorial illustrations. If innumeracy could be overcome, however, the problem of health literacy would need to be tackled. Independent of their educational background, and quite apart from their sophistication in understanding probability, patients were often unfamiliar with the basic vocabulary of medicine.12 The solution to this barrier required adapting the language used by physicians, a strategy easier to implement in written materials than in face-to-face interactions.

In addition to cognitive biases and innumeracy, cultural factors proved to affect the practicability of shared decision-making: some patients, by virtue of their ethnic background, rejected the very notion of becoming informed about their diagnosis and prognosis.13 A possible solution to this obstacle was to ask patients if they wished to participate in decision-making and to give them the opportunity to select an alternative decision-maker.

Beyond patient-related barriers, there are physician-related obstacles to achieving shared decision-making. Physicians have expressed concerns about shared decision-making from its inception, claiming that it is their prerogative, based on their unique expertise, to determine the best possible medical treatment for a given patient. In the interest of beneficence, they argue, they should make medical decisions on behalf of patients. Moreover, physicians have expressed a paternalistic wish to protect patients from the pain of knowing what they see as emotionally burdensome facts about diagnosis and prognosis. Some physicians have indicated they have neither the know-how, the predilection nor the time to engage in the kind of dialogue that shared decision-making demands.9 The usual solution to overcoming professional barriers to shared decision-making is physician education. The assumption is that if doctors only learnt how to involve patients, they would be happy to do so. However, while a variety of educational programmes have been designed to steep physicians in joint decision-making, many of the results have been disappointing. A recent study examining the results of simulation-based communication training on patient-assessed quality of communication, for example, demonstrated no improvement.14

Even if patients and physicians could in principle become skilled in communication, the system-related forces conspiring against full implementation of the shared decision-making paradigm are formidable. Economic incentives stimulate short patient/doctor encounters, not the lengthy and repeated interactions that may be required for shared decision-making. Other disciplines vie for time in the medical school curriculum and during residency training, with those specialists who command a disproportionate share of research funding, such as cardiologists and oncologists, liable to drive out the less technologically intensive physicians who teach subjects such as ethics and patient/doctor communication.

Finally, there is an imbalance of both power and expertise between physicians and patients that threatens to derail the ideal of egalitarian decision-making.15 Many patients want more guidance in deciding about their own care than the fully collaborative model allows: hence the question, universally posed by patients to their physicians, “what would you do if it were your mother?”16 Not “what would you recommend for me?” but rather, what would you advise for someone you care about. Even strong proponents of patient autonomy report the prevailing approach personally unsatisfactory when they are faced with difficult medical decisions: in a book describing the experiences of seven medical ethicists who developed cancer or who cared for a spouse with cancer, Rebecca Dresser writes: “We found that our ideas about patient autonomy were too abstract, too cerebral. Making treatment decisions was more difficult than we had expected, and we were surprised at how much we depended on doctors to help us make treatment choices.17” Decision aids are perhaps the best hope for rescuing shared decision-making from the fate of being a great idea that failed. Decision aids incorporate the remedies for a number of major barriers, seeking in one fell swoop to surmount cognitive biases, innumeracy, health illiteracy and ill-equipped physicians. But decision aids are seldom adopted in practice, even when well-designed, validated instruments are available.18 The inescapable conclusion is that the approach delimited by the shared decision-making paradigm is not working in practice.

Rescuing shared decision-making through re-engineering

The acknowledgement that the practice of shared decision-making is problematic does not undercut the goal of engaging patients in their own healthcare, beginning with decisions about which tests and treatments to undergo. It does not reject the underlying desire to promote individual autonomy by granting patients the ultimate power to choose their medical care. It upholds the assertion that only patient preference can adjudicate among various treatment options when several viable alternatives exist. The problem is the process, which as currently construed asks patients to act like doctors—evaluating highly technical data—and asks doctors to act like friends—serving as a sounding board for their patients. By re-engineering the process, perhaps we can rescue shared decision-making, a concept whose validity is well established in both ethics and the law.

The era of shared decision-making began with the recognition that in contemporary medicine, there are both effective medical therapies and, in many instances, alternative approaches to a given clinical condition. Every approach comes with its own set of benefits and burdens, each of which in turn has its own probability of occurring. Because different patients place different values on each of these benefits and burdens and demonstrate varying levels of risk-aversion, the choice of therapy depends as much on the patient's values and preferences as it does on a correct diagnosis and on knowledge of the available treatments. Therefore, patients need to be involved in those medical decisions for which there is more than one medically reasonable approach to treatment, that is, decisions that are ‘preference-sensitive’.6 In the contemporary version of shared decision-making, this translates into patients’ picking and choosing among particular options: surgical versus medical treatment, chemotherapy versus radiation and hospital care versus hospice care. But what if, instead of patients being asked to decide on the specific treatment, with the clinician's input, patients were asked to elucidate their goals of care and physicians were expected to help translate those goals into actual medical care?

For patients with serious illness, this strategy entails asking patients to prioritise their goals of care and relies on physicians to explain the implications of those goals for treatment. Instead of starting by asking patients if they would favour attempted cardiopulmonary resuscitation, dialysis or major surgery in the event of acute illness, as often occurs today, the conversation would begin with the patient's goals. Patients might, for example, wish to focus on life-prolongation at any cost, or they might want treatment that promotes their daily functioning—walking, hearing, seeing and thinking—or they might want an exclusive emphasis on comfort. Since most people want all three—life-prolongation, maximisation of comfort and maintenance of function—but choosing treatment typically requires making trade-offs, patients need to prioritise their goals. Once patients have rank ordered their goals, shepherded through this process by their physician, the goals can then be translated into an actual treatment plan by the physician, with the patient retaining the ultimate authority to accept or reject the proposed treatment.

The idea of eliciting the goals of care for patients with advanced illness is not new: palliative care specialists are often consulted for the specific purpose of having a ‘goals of care discussion’ and family meetings are often held to encourage patients, together with their families, to formulate their goals. As the end of life approaches, physicians often advise ‘redirecting care’ or moving from a focus on cure to a focus on palliation.19 But while articulating goals is an important first step, it does not necessarily enable patients to choose among alternative treatments. Figuring out what treatment is most consistent with their goals is not easy and the evidence until now suggests that patients have difficulty translating how just two goals, life-prolongation and comfort, should affect the choice of specific interventions such as cardiopulmonary resuscitation, major surgery and antibiotics.20 The translation process requires a deep understanding of how life is likely to unfold with each of the possible treatments. And that, arguably, is just where the physician's expertise comes into play. The patient's job is to think about what is most important personally, given some understanding of their medical condition and how that condition is likely to evolve over time, a task that relies on values. The physician's job is to translate the value-based goals into practice.

This approach to shared decision-making changes the focus of the endeavour from determining what treatments patients want to ascertaining what they hope the treatment will accomplish. For it to work, physicians can no more begin the conversation with the question ‘what are your goals of care?’ than they can (as is so often the reality today) by starting out with “these are your options, which one do you want?” Goal-centric shared decision-making must be embedded in a process that begins with the physician clarifying the patient's underlying health status. What this means is assuring that patients understand the diagnosis, the prognosis and the likely trajectory of disease in the context of their other medical problems. While physicians today are expected to share the patient's diagnosis and usually do, they are far less likely to discuss prognosis: patients with life-limiting diseases, for example, are often unaware of the natural history of their condition, whether they have advanced cancer21 or advanced heart failure.22 Discussion of the experience of illness, with and without various proposed treatments—whether the patient will remain independent or will require assistance with basic daily activities, what it is like for a patient with progressive dementia to undergo treatment in an intensive care unit—is likely even rarer.

Once physicians have discussed and patients appreciate their health status, physicians can move on to a discussion of the goals of care. Goals are intimately linked to the position in the life cycle: whether undergoing burdensome treatment in exchange for a small possibility of life-extension makes sense or not typically hinges on age (whether or not the patient is approaching the end of life, no matter what treatment choice is made) and on the presence or absence of comorbidities (whether the patient is otherwise in good health or has many other illnesses that affect physical function, cognitive function, quality of life and life-expectancy). While some patients, based on religious beliefs or cultural views, may favour life-prolonging treatment regardless of their clinical situation and others, based on their personal philosophy, may espouse a comfort-oriented approach whatever their stage of life, for many if not most people, goals are situational and evolve over the life course. Several approaches have been suggested for initiating a conversation about goals, including asking trigger questions such as “what gives your life meaning” and “what are your biggest fears”.23 Physicians must then translate the responses into the three major medical goals of care: life-prolongation, maintenance of function and maximising comfort. This can be accomplished by taking the patient’s words, for example, “I don't want to have to depend on anyone,” and saying “It sounds as though living on your own is important to you, which means maintaining critical abilities like vision and mobility.” Patients can accept or reject the revised formulation; if it is rejected, physicians must modify their understanding and reformulate the goal of care accordingly.

After goals have been elicited and prioritised, physicians must translate the prioritisation into a specific treatment based on their knowledge of the consequences of the various treatment alternatives. For example, patients with advanced cancer for whom third-line chemotherapy or hospice are two competing options would be advised which option is most consistent with the goal they have indicated is of prime importance. Patients who favour maximising comfort would be told that hospice care is the alternative most likely to succeed in achieving relief of symptoms. Patients whose preference is for life-prolongation would be told that third-line chemotherapy is the treatment that offers the best chance, however small, of producing a remission of the cancer.

A view of shared decision-making that asks patients to define their goals and physicians to figure out the practical implications of those goals alters the division of labour between patients and physicians. It assigns the values component to the patient and the technical component to the physician without fundamentally altering the concept of patient engagement. It seeks to enable patients to participate meaningfully in decision-making about their medical care without placing unfair burdens on them and without diminishing the importance or the sophistication of the physician. It promotes patient autonomy by recognising that self-determination hinges on a patient's freedom to choose how his or her life will unfold. The ability to make choices is thus a marker rather than the source of autonomy: what the individual needs to be able to do in order to shape his life is to take actions that are likely to result in the life he wants. Assisting patients in determining those actions is not coercive paternalism; it is expert guidance that facilitates autonomy.

Evidence supporting re-engineered shared decision-making

It is one thing to argue that the current model isn't working; it is quite another to claim that this re-engineered version would be any more likely to succeed. The proposed variant still requires that patients understand something about their diagnosis and their prognosis—after all, the goals of care are not immutable throughout life, but may well change as the patient's health status changes. It still requires doctors to talk to patients—in the case of serious illness, they will need to elicit the goals of care and help patients prioritise them, a delicate undertaking that necessitates far more than merely laying out three possible goals and asking patients to pick one. In the case of more mundane decisions, they will need to ascertain the patient's general style of addressing life's problems. The chief difference between this model and the approach recommended today is that patients are not expected to select a treatment. Doctors recommend the treatment based on their knowledge of which treatment corresponds best to the patient's stated goals. As a final check on the integrity of the process, patients can veto the decision. So what evidence is there that this approach is any more likely to succeed than the version it is meant to supplant?

For this approach to succeed, patients need to be able to select a goal of care, they need to be able to prioritise their goals and to endorse a pathway to care that embodies their prioritisation, and physicians need to be able to translate the goals of care into specific treatments in the event of an acute illness. Patients have little difficulty selecting a principal goal: in a study of the goal of care chosen by middle-aged outpatients for themselves in the event they developed advanced dementia, subjects readily chose between life-prolonging care, comfort care and basic care (an approach between the two extremes) after watching a video depicting the features of advanced dementia.24 Even extremely elderly and impaired patients living in a nursing home, or their surrogates, are able to prioritise their goals of care. Moreover, once they have done so, they are comfortable endorsing a ‘pathway to care’, a summary of the kinds of treatments that are consistent with their goals.25 Finally, there is preliminary evidence indicating that when given guidelines, physicians are able to translate the goals patients have selected into specific treatments in a variety of clinical scenarios: after physicians at a long-term care facility were given a single lecture about the translation process, followed by a pen-and-paper test asking which of four possible treatment options they would recommend in each of five common clinical scenarios, given a predetermined prioritisation of goals, there was excellent concordance between their responses and a reference answer if the primary goal was comfort or life-prolongation, and a narrowing of the options from four to two if the primary goal was maintaining function.26