Patient Preference

Abstract

Patient preference is the relative desirability a patient assigns to the options in a health decision, and MeSH files it beneath patient satisfaction as a distinct, measurable object of study. It is the engine of shared decision making: where two or more reasonable options exist, the right choice depends on what the patient values, not on clinical facts alone. This article defines shared decision making and distinguishes it from the paternalistic and informed-choice models, separates how much information patients want from how much decision control they want, and shows how preference elicitation and decision aids turn values into value-concordant choices. A worked example and three interactive demonstrations make the mechanisms concrete: a weighted-value model in which identical options yield opposite optimal choices, the control-preference spectrum, and the measured effect of a decision aid on knowledge and choice.

Keywords: patient preference, shared decision making, decision aids

For most of medicine's history the question of what a patient preferred barely arose: the clinician determined the correct treatment and the patient received it. Patient preference names what that model left out, the relative value a patient places on the possible outcomes of a decision when more than one reasonable course exists. It is not a vague matter of taste but a structured object: a ranking of options, or a set of weights on the attributes that distinguish them, that can be elicited, measured, and acted on. MeSH classifies it beneath patient satisfaction, treating the preferences a patient holds as inputs to whether they are ultimately satisfied with the care they receive.

The concept matters most where medicine offers a genuine choice. Many decisions are preference-sensitive: two options carry similar expected survival or benefit but differ in their burdens, risks, and trade-offs, so no single choice is correct for everyone. For such decisions, ignoring preference is not neutral, it is an error, and the sections below set out what preference is, how much involvement patients actually want, and how good care elicits and supports it.

What Patient Preference Is

Patient preference is the ordering a patient would give to the available options once they understand what each entails. The crucial insight is that for a large class of decisions this ordering cannot be read off the clinical facts, because the facts underdetermine the choice. John Wennberg and Alan Gittelsohn's studies of small-area practice variation made the point unmistakable: rates of common surgeries varied several-fold between neighboring regions with similar populations, variation that could not be explained by differences in illness and that instead reflected differences in local medical opinion (Wennberg & Gittelsohn, 1973). If the correct rate were dictated by the disease, it would not swing with the clinician's zip code. The variation revealed a space of discretion that something had to fill, and the obvious candidate, the patient's own values, was the one systematically left out.

This is the territory of the preference-sensitive decision: a choice between options with comparable expected benefit but different attributes, such as a surgery that resolves a problem quickly against watchful waiting that avoids operative risk. Thomas Fried and colleagues showed how much is at stake by asking seriously ill patients how treatment burden, likely outcome, and the chance of severe functional or cognitive impairment changed what they wanted, and found that preferences shifted sharply with the probable outcome rather than holding fixed, so a patient's choice could not be predicted without asking (Fried et al., 2002). Preference, on this account, is the weighting a patient places on outcomes, and it is the missing variable that practice variation exposed.

Defining Shared Decision Making

If preference is the input, shared decision making is the process that acts on it. Cathy Charles, Amiram Gafni, and Tim Whelan gave the field its still-canonical definition, setting out four conditions: at least two participants, the clinician and the patient, both share information, both take steps to build a consensus about the preferred treatment, and both agree on the decision to implement (Charles et al., 1997). The definition did its work by contrast. On one side sits the paternalistic model, in which the clinician decides and the patient assents; on the other the informed-choice model, in which the clinician supplies facts and withdraws, leaving the patient to decide alone. Shared decision making is the collaborative middle: information flows both ways, and the decision is jointly owned.

Glyn Elwyn and colleagues turned the definition into a method clinicians could actually follow, describing a model built on three kinds of conversation, introducing the existence of a real choice, comparing the options in detail, and supporting the patient to form and express a preference (Elwyn et al., 2012). They later distilled it into the three-talk model, a staged consultation moving through team talk (establishing that a choice exists and that the patient's goals matter), option talk (laying out the alternatives and their trade-offs), and decision talk (arriving at an informed preference) (Elwyn et al., 2017). A systematic review by Hanna Bomhof-Roordink and colleagues, surveying the many models that have since appeared, found a common core across them, chiefly describing options and the eliciting of patient preferences, alongside wide variation in the margins, confirming that the field agrees on the essentials even as it proliferates frameworks (Bomhof-Roordink et al., 2019).

The worked model below makes the central claim concrete, that where options are clinically comparable, the patient's weighting of outcomes alone determines the best choice.

Both options carry equal expected survival, so the choice is preference-sensitive. Set how much the patient weights each attribute (the weights are normalized to sum to 100%). Each option’s value is the weighted sum of its attribute scores, and the higher value is the preference-concordant choice.

The two presets face identical clinical options yet reach opposite choices — B for the patient who prizes preservation, A for the one who weights avoiding radiation and recurrence worry. No single default serves both; the right answer is the patient’s own weighting.

How Much Involvement Patients Want

A natural objection to shared decision making is that many patients say they would rather the doctor decide. The objection trades on a conflation that the evidence dissolves. Raisa Deber, Nancy Kraetschmer, and Janet Irvine separated two things that are usually run together: problem-solving, the technical task of determining the facts and the options, which patients overwhelmingly want to leave to the clinician, and decision-making, the selection among options given those facts, in which patients want a far larger role (Deber et al., 1996). A patient who defers the choice to the clinician is often ceding the problem-solving, not the values-laden choice, and asking the two as one question produces a misleadingly passive picture.

Robert Say, Madeleine Murtagh, and Richard Thomson's narrative review mapped what actually varies. Desire for involvement is not uniform: it rises and falls with the patient's age, education, the severity of the illness, and the specific decision, and crucially, the desire to be informed is nearly universal even where the desire to control the final decision is not (Say et al., 2006). The practical upshot is that preference for involvement is itself a preference to be elicited, not assumed.

To make that elicitable, Lesley Degner and John Sloan built the Control Preferences Scale, which asks patients to place themselves along a single ordered spectrum of roles, from a patient-led choice through a shared one to a clinician-led one, and found that preferences spread across the whole range rather than clustering at the passive end (Degner & Sloan, 1992). The scale gives the involvement question a measurable form and underlies the control axis in the demonstration below, which separates the two axes, information and control, and shows how a patient can sit high on one and low on the other.

Set two independent desires: how fully the patient wants to be informed, and how much control they want over the final decision. The point moves in a two-dimensional space — a patient can want every fact and still want the clinician to decide.

Desire to be informed is nearly universal; desire to control the decision varies widely. Because the two axes are independent, preference for involvement is itself something to be elicited, not read off a single “do you want to decide?” question.

Eliciting and Supporting Preferences

Knowing that preference matters does not by itself reveal what a given patient prefers, and here the field moved from principle to tools. The problem Albert Mulley, Chris Trimble, and Glyn Elwyn named the silent misdiagnosis is that clinicians routinely infer patient preferences instead of eliciting them, and infer them wrongly, prescribing what they assume the patient wants (Mulley et al., 2012). The error is silent because nothing in the encounter flags it: the clinician is confident, the patient is compliant, and the mismatch surfaces only in the regret and non-adherence that follow. The remedy is to measure preference rather than guess it.

Decision aids, structured tools that present the options, their outcomes, and their probabilities and help a patient weigh them against personal values, are the best-evidenced instrument for this. Dawn Stacey and colleagues' Cochrane review, the authoritative synthesis, found that decision aids improve patients' knowledge, make their risk perceptions more accurate, lower decisional conflict, and leave patients more active in deciding, often shifting choices toward more conservative options once trade-offs are understood (Stacey et al., 2017). Anne Stiggelbout and colleagues argued that genuinely putting patients at the centre requires embedding these tools in routine care rather than treating involvement as an optional courtesy (Stiggelbout et al., 2012). The demonstration below shows the measured effect of a decision aid on the quantities the Cochrane review tracks.

Switch on a decision aid and set its quality. The four outcomes move in the directions the Cochrane review reports: knowledge, accuracy of risk perception, and active participation rise, while decisional conflict falls. The magnitudes are illustrative, not trial figures.

A decision aid does not push a particular treatment; it improves the quality of the deliberation, leaving patients who know more, perceive risk more accurately, feel less conflicted, and take a more active part in a choice that reflects their own values.

Tristan Hoffmann, Victor Montori, and Chris Del Mar closed a conceptual gap by showing that shared decision making and evidence-based medicine are not rivals but complements: evidence-based medicine is incomplete without the step that brings the patient's values to bear on the evidence, and shared decision making is unmoored without the evidence it deliberates over (Hoffmann et al., 2014). Preference elicitation is where the two meet.

Worked Example: One Diagnosis, Two Choices

Consider a woman with early-stage breast cancer choosing between two treatments of equal expected survival: mastectomy (option A) and breast-conserving surgery with radiation (option B). Because survival is equal, the choice is preference-sensitive, and the right answer depends entirely on how she weights three attributes. Each option is scored 0 to 100 for how well it delivers each attribute, and each patient assigns weights that sum to 1. The option's value is the weighted sum of its attribute scores, and the patient should choose the higher-valued option.

The fixed attribute scores are: breast preservation (A = 0, B = 100), avoiding a weeks-long radiation course (A = 100, B = 0), and peace of mind about local recurrence (A = 80, B = 60).

- Patient 1 weights breast preservation most: preservation 0.6, avoiding radiation 0.1, peace of mind 0.3. - Patient 2 weights avoiding radiation and recurrence worry: preservation 0.1, avoiding radiation 0.4, peace of mind 0.5.

Patient 1's value for A is `0.6×0 + 0.1×100 + 0.3×80 = 0 + 10 + 24 = 34`, and for B is `0.6×100 + 0.1×0 + 0.3×60 = 60 + 0 + 18 = 78`, so B wins. Patient 2's value for A is `0.1×0 + 0.4×100 + 0.5×80 = 0 + 40 + 40 = 80`, and for B is `0.1×100 + 0.4×0 + 0.5×60 = 10 + 0 + 30 = 40`, so A wins. The clinical options and their survival are identical; the best choice is opposite for the two patients, determined entirely by the weights they place on outcomes. A clinician who substituted a single default, their own or a population average, would get one of these two patients wrong, and would never know it, the silent misdiagnosis made arithmetic (Mulley et al., 2012).

Discussion

The study of patient preference has reframed a whole class of clinical decisions. Where medicine once treated variation in practice as a quality problem to be standardized away, it now recognizes that much of the variation in preference-sensitive care should exist, because it tracks legitimate differences in what patients value. The task is not to eliminate variation but to ensure it is warranted, that it follows patients' informed preferences rather than clinicians' assumptions or local habit (Wennberg & Gittelsohn, 1973). This is why preference elicitation is a measurement problem as much as an ethical one: the goal is to replace a guessed input with a measured one.

The relationship to adherence and satisfaction is direct. A patient whose care reflects their own weighted values is more likely to follow through on it and to be satisfied with the result, which is why MeSH files preference beneath satisfaction and why the broader literature on patient participation and treatment adherence connects to it so tightly. A decision a patient has genuinely authored is a decision they are more likely to keep (Charles et al., 1997).

Cognitive Implications

Patient preference is a judgment about the value of uncertain outcomes, and it inherits the properties of such judgments studied in decision making. Preferences are not simply retrieved from memory like a fixed quantity; as Paul Slovic argued, they are frequently constructed in the course of eliciting them, so the method of elicitation becomes part of what is measured (Slovic, 1995). Framing the same outcome as a survival rate or a mortality rate, presenting risks as frequencies or percentages, or changing the order of options can move the expressed preference, so a decision aid must be designed to elicit a stable, considered value rather than an artifact of presentation (Stacey et al., 2017).

This construction problem is why eliciting preference well is harder than asking a single question. The distinction Deber and colleagues drew between problem-solving and decision control has a cognitive basis: evaluating technical evidence and clarifying one's own values are different tasks that make different demands, and patients can be expert in the second while deferring entirely on the first (Deber et al., 1996). Good preference elicitation scaffolds the values task, giving the patient the structure, the comparisons, and the time to work out what they want rather than demanding an answer cold. In this sense a decision aid is a cognitive prosthesis: it externalizes the weighing that choice behavior research shows people struggle to do reliably in their heads.

Current Directions

The active frontier has moved from establishing that shared decision making works to getting it to happen reliably in routine care. France Légaré and colleagues' Cochrane review of interventions to increase clinicians' use of shared decision making found that no single strategy, whether training clinicians, giving patients decision aids, or both, reliably produces large effects, and that the evidence on what actually changes professional behavior remains uncertain (Légaré et al., 2018). Implementation, not proof of principle, is now the bottleneck, and it is a behavior-change problem in its own right.

A second current concerns measurement and standardization. The proliferation of shared-decision-making models that Bomhof-Roordink and colleagues documented is a sign of a field still negotiating its own definitions, and the drive now is toward a shared core of measurable elements so that interventions can be compared and quality assessed consistently (Bomhof-Roordink et al., 2019). Alongside this sits growing interest in integrating preference elicitation with the evidence-based-medicine workflow so that the patient's values enter at the point of care rather than as an afterthought (Hoffmann et al., 2014). The open questions are practical: how to fit deliberation into short consultations, how to reach patients with low health literacy without steering them, and how to tell genuine preference from a response manufactured by the way the choice was framed.

Common Misconceptions

Patients who defer to the doctor do not want to be involved.

Usually they are ceding the technical problem-solving, not the values-laden choice. Patients overwhelmingly want the clinician to determine the facts and options, while wanting a substantial role in selecting among them (Deber et al., 1996).

A good clinician already knows what the patient would want.

No. Clinicians routinely infer preferences and infer them wrongly, a silent misdiagnosis that surfaces only in regret and non-adherence. Preferences must be elicited, not assumed (Mulley et al., 2012).

Giving patients the facts and letting them choose is shared decision making.

That is the informed-choice model, not shared decision making. Sharing requires a two-way exchange and a jointly built consensus, not a clinician who supplies information and then withdraws (Charles et al., 1997).

Respecting preference means abandoning the evidence.

No. Shared decision making and evidence-based medicine are complements: the evidence defines the options and their probabilities, and preference decides among them. Neither is complete without the other (Hoffmann et al., 2014).

Glossary

Control Preferences Scale.

An instrument measuring how much control a patient wants over a medical decision, ranging from patient-led through shared to clinician-led.

Decision aid.

A structured tool that presents the options, outcomes, and probabilities of a decision and helps a patient weigh them against personal values.

Decisional conflict.

The uncertainty a patient feels about which course to take, which good preference elicitation and decision aids are shown to reduce.

Informed-choice model.

A model in which the clinician supplies information and then withdraws, leaving the patient to decide alone; distinct from shared decision making.

Option talk.

The stage of the three-talk model in which the clinician lays out the alternatives and their trade-offs in detail.

Paternalistic model.

A model in which the clinician determines the treatment and the patient assents, with the patient's values playing little explicit role.

Patient preference.

The relative desirability a patient assigns to the options in a health decision, expressible as weights on the outcomes that distinguish them.

Preference elicitation.

The process of measuring a patient's values and option ranking rather than inferring them, often scaffolded by a decision aid.

Preference-sensitive decision.

A choice between options with comparable expected benefit but different trade-offs, for which the right answer depends on the patient's values.

Shared decision making.

A process in which clinician and patient share information, build a consensus about the preferred option, and agree together on the decision to implement.

Silent misdiagnosis.

Mulley and colleagues' term for a clinician inferring a patient's preference wrongly, an error that goes undetected because nothing in the encounter flags it.

Three-talk model.

Elwyn and colleagues' staged account of shared decision making, moving through team talk, option talk, and decision talk.

Unwarranted variation.

Variation in care rates not explained by illness or patient preference, revealing a space of clinical discretion that preference should fill.

Value-concordant choice.

A decision that matches the patient's own weighted values, the outcome that preference elicitation and shared decision making aim to produce.

Key Researchers

Cathy Charles

(McMaster University, d. 2015). Health-policy researcher whose 1997 paper with Amiram Gafni and Tim Whelan gave shared decision making its still-canonical definition, separating it cleanly from the paternalistic and informed-choice models and establishing the conditions, shared information and a jointly built consensus, that define the process. McMaster University

Angela Coulter

(University of Oxford). Health-services researcher who made patient-centred care and the measurement of patient experience a mainstream policy concern, arguing that eliciting and respecting patient preferences is an obligation rather than a courtesy and that decision aids are the tools that make it operational. ORCID

Glyn Elwyn

(Dartmouth College). Physician-researcher who turned shared decision making from a principle into a practicable clinical method, the model of choice, option, and decision talk and the later three-talk model, and who developed the OPTION and CollaboRATE measures that let the field quantify how far a patient was actually involved. ORCID · Google Scholar · Wikipedia

Dawn Stacey

(University of Ottawa). Nurse-scientist who leads the Cochrane systematic review of patient decision aids, the most authoritative synthesis of whether supporting patients to deliberate changes what they know, how accurately they perceive risk, and whether their choices match their values. ORCID · Google Scholar

Anne M. Stiggelbout

(Leiden University Medical Center). Decision scientist whose work on preference elicitation and utility assessment clarified what it means to measure a patient's values, and whose 2012 analysis sharpened the distinction between merely informing patients and genuinely incorporating what they want. ORCID · Google Scholar

Frequently Asked Questions

What is patient preference?

It is the relative desirability a patient assigns to the options in a health decision, expressible as a ranking of options or as weights on the outcomes that distinguish them. It is a structured, measurable input to care, and MeSH classifies it beneath patient satisfaction (Charles et al., 1997).

What is a preference-sensitive decision?

It is a choice between two or more options that carry comparable expected benefit but differ in their burdens, risks, and trade-offs, so no single option is correct for everyone. For such decisions the right choice depends on the patient's values, which is why preference must be elicited (Fried et al., 2002).

What is shared decision making?

It is a process in which clinician and patient both share information, both work to build a consensus about the preferred option, and both agree on the decision to implement. It sits between the paternalistic model, in which the clinician decides, and the informed-choice model, in which the patient decides alone (Charles et al., 1997).

Don't most patients just want the doctor to decide?

Less often than it appears. Patients overwhelmingly want the clinician to handle the technical problem-solving, but many want a substantial role in the values-laden choice itself. Asking the two as one question produces a misleadingly passive picture (Deber et al., 1996).

What is the silent misdiagnosis?

It is the term Mulley and colleagues gave to a clinician inferring a patient's preference instead of eliciting it, and inferring it wrongly. The error is silent because nothing in the encounter flags it; it surfaces only later in regret and non-adherence (Mulley et al., 2012).

Do decision aids actually work?

Yes. The Cochrane review finds that decision aids improve patients' knowledge, make risk perceptions more accurate, reduce decisional conflict, and leave patients more active in deciding, often shifting choices toward more conservative options once the trade-offs are understood (Stacey et al., 2017).

Does respecting preference mean ignoring the evidence?

No. Shared decision making and evidence-based medicine are complements: the evidence defines the options and their probabilities, and the patient's preference decides among them. Each is incomplete without the other (Hoffmann et al., 2014).

Why is shared decision making still not routine?

Because making it happen reliably is a hard behavior-change problem. The Cochrane review of implementation strategies finds no single approach produces large, reliable effects, so the current bottleneck is implementation rather than proof that the approach works (Légaré et al., 2018).

References