Shared Decision Making

Abstract

Shared decision making is a type of decision making: a process in which a clinician and a patient reach a health decision together, combining the clinician's knowledge of the options and their outcomes with the patient's informed preferences. It applies most directly to preference-sensitive choices, where two or more reasonable options differ in benefits and harms that patients weigh differently, so no single decision is right for everyone. Cognitive psychology bears on the process at three points: how a consultation elicits and combines the two parties' contributions, how evidence about options is communicated so it is understood rather than merely stated, and how decision aids change what patients know and how confidently they choose. This article sets out the leading process models, the psychology of risk communication and numeracy, the evidence on decision aids, and the barriers to routine practice.

Keywords: shared decision making, patient-centered care, decision aids, risk communication, numeracy

Shared decision making names both an ethical standard and a describable cognitive process: the standard that patients should participate in decisions about their own care, and the process by which a clinician and patient actually pool what each knows to arrive at a choice (Barry & Edgman-Levitan, 2012). The two are not the same, and cognitive psychology is concerned with the second, because a decision can honor the ethical standard in name while failing as a process if the patient never understood the numbers or never had their values elicited (Stiggelbout et al., 2015).

What Shared Decision Making Is

Shared decision making is defined by the sharing of both information and the decision itself. In the founding analysis, Charles, Gafni, and Whelan set four conditions: at least two participants, the clinician and the patient, are involved; both take steps to participate in the process; information is shared in both directions; and a treatment decision is reached that both parties agree to (Charles et al., 1997). It is positioned against two alternatives it is not. In the paternalistic model the clinician decides and informs the patient; in the pure informed-choice model the clinician supplies information and the patient decides alone. Shared decision making is the middle course in which deliberation is genuinely joint (Charles et al., 1999).

The approach is reserved above all for preference-sensitive decisions. When the options differ in outcomes that patients value differently, for example a treatment that extends life at the cost of its quality against one that does the reverse, the right choice depends on the patient's priorities, and no amount of clinical expertise can supply those priorities from outside (Stiggelbout et al., 2012). This genuine balance among reasonable options, a state of equipoise, is what marks a decision as preference-sensitive and so suited to being shared. This is why shared decision making has been called the pinnacle of patient-centered care: it is the practice in which caring for the patient as a person, rather than treating the disease in the abstract, becomes operational (Barry & Edgman-Levitan, 2012).

Figure 1

Patients differ in how much of the decision they want to control, and that preferred role is itself information the clinician needs. Early work measuring control preferences found that patients vary widely, from wanting to decide alone to wanting the clinician to decide, with many preferring a shared role, and that clinicians routinely misjudge which a given patient wants (Degner & Sloan, 1992). Shared decision making does not force a fixed division of labor; it asks the clinician to calibrate to the role the patient actually wants, which may itself change over the course of a serious illness.

Models of Shared Decision Making

Several process models decompose the consultation into definable components. Makoul and Clayman, reviewing the many published definitions, distilled an integrative model whose essential elements recur across them: defining the problem, presenting options, discussing benefits and harms, eliciting the patient's values and preferences, checking understanding, and reaching and implementing a decision (Makoul & Clayman, 2006). A later systematic review of shared-decision-making models confirmed that a small set of components, describing options, discussing preferences, and making the decision, appear in nearly all of them, while elements such as explicitly defining the clinician's role are far less consistently included (Bomhof-Roordink et al., 2019).

Elwyn and colleagues translated these components into a model designed to be used at the point of care. Their first formulation set out three steps, choice talk, option talk, and decision talk, with a decision-support role for the clinician throughout (Elwyn et al., 2012). The refined three-talk model relabels the stages as team talk, option talk, and decision talk: the clinician first works with the patient to establish that there is a choice and that they will approach it together, then compares the options and their consequences, and finally elicits and integrates the patient's informed preferences into a decision (Elwyn et al., 2017). The model is explicitly iterative, allowing the conversation to return to an earlier stage as the patient's understanding grows.

Several persistent misunderstandings attach to these models. Légaré and Thompson-Leduc catalogued twelve recurring myths, among them that shared decision making is appropriate only for major decisions, that it necessarily lengthens the consultation, that patients do not want it, and that simply providing information amounts to sharing the decision (Légaré & Thompson-Leduc, 2014). The last is the most consequential for cognitive psychology: handing a patient a pamphlet is not shared decision making if the patient cannot interpret its numbers or was never asked what matters to them, which is why the presentation of evidence and the elicitation of preferences are treated here as core, not peripheral.

Risk Communication and Numeracy

A decision can only be shared if the patient grasps the quantitative stakes, and whether they do depends heavily on how the numbers are presented. Gigerenzer and Edwards showed that clinicians and patients alike reason far more accurately about risk when it is expressed as natural frequencies, for example a count such as 10 in 1,000, than as conditional probabilities or percentages, because frequencies preserve the base-rate information that probability formats strip away (Gigerenzer & Edwards, 2003). The same evidence can mislead when framed as a relative rather than an absolute risk: a reduction described as cutting risk by 20% sounds dramatic even when it means moving from 5 to 4 cases in 100, so the format is not cosmetic but part of what the patient decides on.

Fagerlin and colleagues assembled these findings into practical guidance, recommending absolute risks over relative ones, consistent denominators, visual formats such as icon arrays, and the removal of nonessential information that competes for attention (Fagerlin et al., 2011). The need for such care is sharpened by the fact that numeracy, the ability to understand and use numbers, is limited even among educated adults: Lipkus, Samsa, and Rimer found that highly educated samples made frequent errors on basic numeracy items such as converting between percentages and proportions (Lipkus et al., 2001).

Numeracy does not merely gate comprehension; it shapes the decision. Peters and colleagues showed that less numerate people draw less meaning from numbers and are correspondingly more influenced by how information is framed, so that two presentations of the same risk lead them to different choices (Peters et al., 2006). Reyna's fuzzy-trace theory explains part of why: people encode both a precise verbatim representation and a simpler gist, the bottom-line meaning, and they tend to reason and decide from the gist, so effective communication must convey the right gist and not only the exact figures (Reyna, 2008). Underlying all of this is the finding that preferences are often constructed in the moment rather than retrieved intact, so the way options and risks are framed can shape the very values a patient expresses (Slovic, 1995). The same constructive tendency appears in everyday reasoning, where people lean on fast heuristics that can be well adapted or systematically biased depending on the environment (Gigerenzer & Gaissmaier, 2011).

Decision Aids and the Evidence

Patient decision aids are tools, from booklets to interactive programs, that present the options and outcomes for a specific decision and help patients clarify what they value. They are the most heavily evaluated instrument in the field. The Cochrane systematic review led by Stacey, pooling well over a hundred randomized trials, found that decision aids improve patients' knowledge of the options, give them more accurate risk perceptions, reduce decisional conflict arising from feeling uninformed or unclear about values, and leave patients more active in decision making, with no evidence of worse health outcomes (Stacey et al., 2017). This is the strongest quantitative warrant the approach has.

For a decision aid to promote shared decision making rather than simply inform, Agoritsas and colleagues argue it must be usable within the clinical encounter and genuinely support deliberation between clinician and patient, not substitute a tool for a conversation (Agoritsas et al., 2015). The distinction matters because the mechanism of benefit runs through engagement: Hibbard and Greene's synthesis of patient activation, the degree to which patients have the knowledge, skills, and confidence to manage their care, found that more activated patients have better outcomes and experiences, locating part of the value of these tools in the engagement they build rather than the information alone (Hibbard & Greene, 2013). Momentum has also come from policy and professional endorsement, with commentators arguing that converging ethical, evidentiary, and system pressures had made this the moment for shared decision making to enter routine care (Spatz et al., 2017).

Worked Example

Consider a preventive medication whose benefit is described, as it often is in marketing, by its relative risk reduction, and follow the same quantities the risk-communication demonstration above manipulates. Suppose that over ten years, 50 of every 1,000 untreated patients like this one will have the event the drug prevents, a baseline risk of 5%. The drug carries a relative risk reduction of 20%.

Applying the relative reduction to the baseline gives the treated risk: 20% of 50 is 10, so the drug prevents 10 events per 1,000 and the treated group has 50 minus 10, or 40 events per 1,000, a risk of 4%. The absolute risk reduction is therefore 50 minus 40, which is 10 in 1,000, or 1 percentage point. The number needed to treat is the reciprocal of the absolute risk reduction: 1,000 divided by 10, which is 100, meaning that on average 100 patients must take the drug for ten years for one of them to avoid the event.

The psychology is in the gap between two true statements about the identical result. The claim that the drug lowers risk by 20% and the statement that if 100 similar patients take it for ten years, one will be spared the event while the other 99 would have had the same outcome either way describe the same 10-in-1,000 effect, yet patients and clinicians alike systematically perceive the first as a larger benefit than the second (Gigerenzer & Edwards, 2003). A shared decision requires the absolute, natural-frequency form, because only it lets the patient weigh the real size of the benefit against the burden and harms of taking the drug. Presenting the relative figure alone is not neutral information; it tilts the decision before the patient's values are ever consulted.

Discussion

If the evidence for decision aids is strong, the evidence on uptake is sobering: shared decision making remains far from routine, and the obstacles are well mapped. Légaré and colleagues, reviewing health professionals' own reports, found the dominant barriers to be time pressure, a perceived lack of applicability to a given patient or clinical situation, and the belief that one already practices it; the chief facilitators were provider motivation and the conviction that the process would improve the outcome or the care process (Légaré et al., 2008).

The barriers are not only the clinician's. Joseph-Williams and colleagues, synthesizing patients' reported experiences, found that giving patients knowledge is not enough: a power asymmetry in the relationship leaves many reluctant to question the clinician, state a preference, or disagree, so that even well-informed patients may not participate unless they are explicitly invited and given permission to deliberate (Joseph-Williams et al., 2014). This reframes the task as changing the relationship and not only the patient's information, and it explains why providing a decision aid without a cultural and relational shift often fails to produce shared decisions.

The deeper claim about why the effort is worth making is that shared decision making completes evidence-based medicine rather than competing with it. Evidence establishes what the options do on average; only the patient can supply the weighting of outcomes that turns an average effect into a decision for this person, so eliciting preferences is a necessary step in applying evidence, not a departure from it (Stiggelbout et al., 2015).

Current Directions

The research front has moved from demonstrating that shared decision making works to understanding why it so rarely becomes routine. A systematic review of implementation in hospitals by Waddell and colleagues found that barriers and facilitators recur across organizational levels, from individual clinician attitudes to team culture to institutional policy, and concluded that interventions aimed at a single level, typically the clinician, are insufficient without accompanying change in workflow and organizational support (Waddell et al., 2021). This has shifted attention toward multilevel implementation strategies and the measurement of whether shared decision making actually occurs, rather than whether tools are merely available.

A second, more conceptual direction reexamines what the process is for. Gulbrandsen and colleagues argue that reducing shared decision making to a procedure of information exchange and preference elicitation misses its deeper aim, which is to restore the autonomous capacity of a person destabilized by illness, framing the encounter as an existential as much as an informational one (Gulbrandsen et al., 2016). The two directions are complementary: one asks how to embed the process in real systems, the other what the process should be understood to achieve, and both push beyond the original demonstration that giving patients good information and a say improves the decisions they reach.

Key Researchers

Angela Coulter

(contemporary). A leading voice for patient-centered care and patient engagement who argued that withholding the information patients need to participate is itself a failure of care quality, and that shared decision making requires system-level change rather than willing clinicians alone. ORCID

Glyn Elwyn

(contemporary). Translated the conceptual model of shared decision making into practice, defining the three-talk model of team, option, and decision talk and co-developing the OPTION scale and Option Grid decision aids that operationalize how clinician and patient deliberate. ORCID

Gerd Gigerenzer

(contemporary). Showed that expressing risk as natural frequencies rather than conditional probabilities sharply improves both clinicians' and patients' statistical reasoning, giving shared decision making a concrete method for communicating benefits and harms. ORCID

France Legare

(contemporary). Built the implementation science of shared decision making, cataloguing the barriers and facilitators clinicians report, running the systematic reviews of interventions to embed it in routine care, and developing interprofessional models that extend it beyond the doctor-patient dyad. ORCID

Victor M. Montori

(contemporary). Originated minimally disruptive medicine and the careful-and-kind care model, designing encounter-based decision aids intended for use during the consultation itself and arguing that evidence-based medicine is completed, not contradicted, by eliciting what matters to the patient. ORCID

Ellen Peters

(contemporary). Demonstrated that numeracy shapes how people weigh medical risks and benefits, that the less numerate are more swayed by framing, and that well-designed numeric presentation can make the key quantities both usable and affectively meaningful. ORCID

Valerie F. Reyna

(contemporary). Developed fuzzy-trace theory, which holds that people reason from the simplest meaningful gist of information rather than its verbatim detail, explaining why patients often decide on the bottom-line meaning of a risk. ORCID

Paul Slovic

(contemporary). Founded the modern psychological study of risk perception and showed that preferences are often constructed rather than retrieved, which explains why the framing of options and risks can shape the very values a patient expresses. ORCID

Dawn Stacey

(contemporary). Leads the Cochrane systematic review of patient decision aids, assembling the cumulative randomized-trial evidence that decision aids improve knowledge, calibrate risk perceptions, and reduce decisional conflict, and maintains the Ottawa Decision Support Framework. ORCID

Anne M. Stiggelbout

(contemporary). Advanced the measurement and practice of preference-sensitive care, clarifying what shared decision making requires of clinicians, how patient preferences can be elicited rigorously, and why preference-sensitive decisions should turn on the informed values of the patient. ORCID

Glossary

Absolute Risk Reduction.

The arithmetic difference between the event rate without a treatment and the event rate with it, the true size of the benefit, as opposed to the relative reduction.

Decision Aid.

A tool, from a booklet to an interactive program, that presents the options and outcomes for a specific decision and helps a patient clarify what they value before choosing.

Decisional Conflict.

The state of uncertainty about a course of action that arises from feeling uninformed, unclear about personal values, or unsupported, and which decision aids measurably reduce.

Equipoise.

A state of genuine uncertainty or balance among the reasonable options, which signals that a decision is preference-sensitive and so suited to being shared.

Fuzzy-Trace Theory.

The theory that people encode both a precise verbatim representation and a simpler gist of information, and tend to reason and decide from the gist, the bottom-line meaning.

Gist.

In fuzzy-trace theory, the simplified bottom-line meaning a person extracts from information, which typically drives their decision more than the exact figures.

Icon Array.

A visual display of risk showing a grid of figures in which a highlighted subset represents those affected, making a frequency immediately legible without calculation.

Natural Frequency.

A risk expressed as a count out of a reference population, such as 10 in 1,000, which preserves base-rate information and is understood far more accurately than a conditional probability.

Number Needed to Treat.

The reciprocal of the absolute risk reduction, the average number of patients who must receive a treatment for one of them to benefit.

Numeracy.

The ability to understand and work with numbers and probabilities, which is limited even in educated adults and shapes how risk information is interpreted and used.

Patient Activation.

The degree to which a patient has the knowledge, skills, and confidence to manage their own health and care, associated with better outcomes and experiences.

Preference-Sensitive Care.

A decision in which two or more reasonable options differ in outcomes that patients value differently, so the right choice depends on the individual's informed preferences.

Relative Risk Reduction.

The proportional reduction in an event rate produced by a treatment, which tends to be perceived as larger than the same effect expressed as an absolute reduction.

Three-Talk Model.

Elwyn and colleagues' model of the shared-decision-making consultation as three kinds of conversation: team talk, option talk, and decision talk.

Values Clarification.

The process, often supported by a decision aid, of helping a patient recognize and weigh what matters most to them among the outcomes an option affects.

Frequently Asked Questions

What is shared decision making?

It is a process in which a clinician and patient reach a health decision together, combining the clinician's knowledge of the options and their outcomes with the patient's informed preferences about those outcomes (Charles et al., 1997).

When is shared decision making most appropriate?

It is most appropriate for preference-sensitive decisions, where two or more reasonable options differ in benefits and harms that patients value differently, so there is no single choice that is right for everyone (Stiggelbout et al., 2012).

What is the three-talk model?

It is Elwyn and colleagues' practical model of the consultation as three conversations: team talk to agree to collaborate, option talk to compare the choices, and decision talk to elicit preferences and decide (Elwyn et al., 2017).

Why does the way risk is presented matter?

Because the same evidence is understood differently depending on format: natural frequencies and absolute risks are grasped far more accurately than conditional probabilities or relative-risk figures (Gigerenzer & Edwards, 2003).

How does numeracy affect medical decisions?

Less numerate patients draw less meaning from numbers and are more influenced by how information is framed, so the same risk presented two ways can lead them to different choices (Peters et al., 2006).

Do patient decision aids actually work?

A large Cochrane review found that decision aids improve patients' knowledge, give them more accurate risk perceptions, reduce decisional conflict, and leave them more active in decisions, with no evidence of worse outcomes (Stacey et al., 2017).

Why is shared decision making still not routine?

Reported barriers include clinician time pressure and the belief that one already practices it, low patient numeracy, and a power asymmetry that leaves even informed patients reluctant to participate without explicit invitation (Joseph-Williams et al., 2014).

Does shared decision making conflict with evidence-based medicine?

No; evidence establishes what options do on average, but only the patient can supply the weighting of outcomes that turns an average effect into a decision for them, so eliciting preferences completes rather than competes with evidence (Stiggelbout et al., 2015).

References