Preference

Preference in Health Economics

In health economics a preference is a comparative evaluation held by a person, which places one option, such as a health state or a treatment, above, below or level with another. Health utilities and willingness-to-pay values are both numerical summaries of preferences, so preferences are the raw material of economic evaluation. Welfarist economics counts a change as a benefit when people move to options they rank higher, which means the preferences an evaluation measures settle what it counts as benefit. The NICE manual (PMG36) takes reference-case QALY weights from the preferences of a representative sample of the UK public, while patients describe their own health, so whose preferences count is a live question in every appraisal.

Strict preference, indifference and weak preference

Hansson and Grüne-Yanoff, writing in the Stanford Encyclopedia of Philosophy, treat preferences as subjective comparative evaluations of the form "agent A prefers X to Y". The agent may be an individual or a collective. Economists often take the items to be bundles of goods, and health economists take them to be health states or treatments.

Preference logic usually takes weak preference, read as at least as good as, to be the primitive relation and derives strict preference and indifference from it.

$$A \succ B \iff (A \succcurlyeq B) \wedge \neg(B \succcurlyeq A) \qquad A \sim B \iff (A \succcurlyeq B) \wedge (B \succcurlyeq A)$$

where $A$ and $B$ are options, $\succcurlyeq$ is weak preference, $\succ$ is strict preference and $\sim$ is indifference.

Economics adds completeness, so that any two options can be compared, and transitivity, so that if A is at least as good as B and B at least as good as C, then A is at least as good as C. Choice theory builds its model of the rational chooser on these conditions. Hansson and Grüne-Yanoff note that some incompleteness cannot be resolved, because the options differ in ways that cannot be put on the same footing, and a patient asked to compare long survival with a disabling side effect against shorter survival without it may have no ranking ready. The usual defence of transitivity is the money pump, in which a person with cyclic preferences who pays a small sum for each preferred swap can be traded round the cycle indefinitely and loses money at every step.

Turning a preference ordering into utility numbers

A utility number records no more than the preference behind it unless more is assumed. Under an ordinal representation a utility function gives the preferred option the higher number.

$$A \succ B \iff u(A) > u(B)$$

where $u$ is a utility function and $u(A)$ and $u(B)$ are the numbers it assigns to options $A$ and $B$. Any positive monotone transformation of $u$ represents the same preferences, so only the order of the numbers carries meaning.

The encyclopedia entry adds that a preference relation has an ordinal representation only if it is complete and transitive, and that measuring how much one option is preferred to another needs a measurement procedure and a cardinal scale. Most such work has used interval scales built from lotteries, the approach of Ramsey and of von Neumann and Morgenstern. The standard gamble applies it to health states, giving an expected utility scale on which differences between health state utility values can be compared, as QALY arithmetic requires. Ordinal and cardinal utility differ in whether the numbers record only the order or also the strength of preference.

Choices, survey answers and the evidence for preferences

Revealed preference theory infers preferences from observed choices. The encyclopedia entry notes that Samuelson proposed in 1938 to define preference by choice alone, a step economists have largely not followed, and that people hold preferences over options they cannot choose between, which can be found by asking. Many health outcomes are of that kind, since a patient cannot buy a different health state, so health economics leans on stated methods such as the time trade-off, discrete choice experiments and contingent valuation, compared under preference elicitation.

The HM Treasury Green Book (2026) ranks valuation methods by reliability, with market prices first and stated preference alongside wellbeing valuation at the bottom.

Preference satisfaction as the welfarist measure of benefit

Crisp, in the encyclopedia entry on well-being, links the dominance of desire theories of well-being to welfare economics, whose practitioners came to see well-being as the satisfaction of preferences revealed by choices. Brouwer and colleagues list individual sovereignty among the four tenets of the welfarist framework, the view that individuals are the best judges of what contributes to their utility, and add that welfarists do not usually ask how preferences came about or judge them on grounds of decency or ethics. That is the logic of cost-benefit analysis, where a programme is valued by the willingness to pay of the people affected and changes are judged by the Pareto efficiency or compensation tests set out under welfare economics.

Hansson and Grüne-Yanoff note that the wish-based view of welfare struggles with misinformed and self-defeating wishes, and that the usual answer is to launder or refine preferences first, at the risk of paternalism if others decide which preferences survive. They also discuss second-order preferences, such as an addict who prefers not to prefer smoking. A welfarist appraisal of a tobacco tax that counts smokers' lost consumer surplus at face value treats the first-order preference as welfare, and giving weight to the second-order preference instead is a value judgement outside narrow welfarism.

Whose preferences count in extra-welfarist evaluation

Brouwer and colleagues trace one root of extra-welfarism to the decision-making approach of Sugden and Williams, in which decision makers are the source of values and weights. Preferences may still matter in extra-welfarist calculation, they write, though not necessarily those of the direct beneficiaries, and QALY values may be grounded in the preferences of representative members of the population served without being utilities in the welfarist sense.

The NICE reference case follows that pattern. Changes in health-related quality of life should be reported directly by patients, while their valuation should rest on public preferences from a representative sample of the UK population using a choice-based method (sections 4.3.1, 4.3.3 and 4.3.4). The manual lists subgroups based solely on individual utilities for health states and patient preference among those not considered relevant (section 4.9.12). The QALY in a cost-utility analysis therefore carries a social judgement about health states, applied through a value set.

The patient versus public question remains contested. Dolan argued that framing it as a choice between patients and the general public is a false dichotomy, and that a more important issue is whether adaptation to poor health should be taken into account when values are given to those states. Ubel, Loewenstein and Jepson list reasons the two groups' values differ, among them adaptation to illness that the public does not predict and a focusing illusion in which people forget obvious aspects of unfamiliar states.

Worked example of patient and public values for one health state

All numbers in this example are illustrative. A treatment restores full health for 5 years to people who would otherwise spend those years in a chronic state S, with no effect on survival, at an incremental cost of GBP 36,000 per patient. Discounting is left out for clarity. A public valuation exercise gives S a value of 0.55, while patients living in S give it 0.80.

1. Compare the orderings. Both groups rank full health above S and S above death, so the orderings are identical and only the size of the gap between S and full health differs.

2. Work out the QALY gain. The gain per patient is

$$\Delta Q = T \times (1 - u_S)$$

where $\Delta Q$ is the QALY gain, $T$ is the 5 years spent in the improved state and $u_S$ is the value given to S. With public values $\Delta Q = 5 \times 0.45 = 2.25$ QALYs, and with patient values $\Delta Q = 5 \times 0.20 = 1.00$ QALY.

3. Work out the ICER. The ICER with public values is $\text{ICER} = 36000 / 2.25 = 16000$, or GBP 16,000 per QALY, and with patient values it is $\text{ICER} = 36000 / 1.00 = 36000$, or GBP 36,000 per QALY.

4. Work out the net health benefit. At GBP 25,000 per QALY the opportunity cost of the spending is $\text{health forgone} = 36000 / 25000 = 1.44$ QALYs, so the net health benefit is $\text{NHB} = 2.25 - 1.44 = 0.81$ QALYs with public values and minus 0.44 QALYs with patient values.

5. Find the switching values. The ICER equals GBP 25,000 when the gain is 1.44 QALYs, at $u_S = 1 - 1.44 / 5 = 0.712$. It equals GBP 35,000 when the gain is $\Delta Q = 36000 / 35000 = 1.029$ QALYs, at $u_S = 1 - 1.029 / 5 = 0.794$.

6. Interpret the result. Under PMG36, below a most plausible ICER of GBP 25,000 per QALY the decision is normally based on the cost-effectiveness estimate and the acceptability of the technology (section 6.3.4), while above GBP 35,000 the committee will need an increasingly stronger case (section 6.3.8). The manual reads negative net health benefit as overall health loss (section 6.3.2). The clinical evidence is identical, so the result turns on whose preferences value S. Any value between about 0.71 and 0.79 would put the ICER inside the GBP 25,000 to GBP 35,000 range covered by section 6.3.7.

Framing, adaptation and unstable preferences

Stated preferences in health respond to how a question is put. McNeil, Pauker, Sox and Tversky gave 238 ambulatory patients, 491 graduate students and 424 physicians summary data on surgery and radiation therapy for lung cancer. In all three groups surgery looked substantially more attractive when the treatments were named, when outcomes were given as life expectancy instead of cumulative probabilities, and when the problem was framed as the probability of living instead of dying. The same clinical facts produced different choices, which makes any single elicitation a weak guide to a settled preference.

Hansson and Grüne-Yanoff also cite evidence that preferences change with habit formation and with physiological change. If people adapt to a chronic condition, a value elicited soon after diagnosis and one elicited years later may both be sincere, and an evaluation has to choose between them. Preferences over the timing of costs and outcomes are covered under time preference.

Preference beside utility, value judgements and social choice

Several nearby terms are often run together with preference. Utility is a numerical representation of a preference ordering, and utility maximisation is the assumption that people choose the affordable option they rank highest. A value judgement about whose preferences should count, or about whether health should replace preference satisfaction as the outcome, is a normative choice that no quantity of preference data can settle. Combining many individual orderings into a social ordering is the subject of social choice theory and of the social welfare function.

Common misreadings of preference in health evaluation

Most errors come from treating a measured preference as more settled, or more closely tied to welfare, than the theory allows. Four recur in health evaluation.

1. Preference and choice are the same thing. Choices also reflect beliefs and the options on offer.
2. A ranking is enough for a QALY. QALY arithmetic needs the size of the gaps between states, which an ordinal ranking does not supply.
3. NICE ignores patients. PMG36 asks patients to describe their health-related quality of life and uses public preferences to value it.
4. Welfarism judges preferences. It takes preferences as given, and laundering them or replacing them with health is a step outside narrow welfarism.

Sources

• Brouwer WBF, Culyer AJ, van Exel NJA, Rutten FFH. Welfarism vs. extra-welfarism. Journal of Health Economics. 2008;27(2):325-338. https://doi.org/10.1016/j.jhealeco.2007.07.003
• Crisp R. Well-being. In: The Stanford Encyclopedia of Philosophy. First published 6 November 2001, substantive revision 13 January 2026. https://plato.stanford.edu/entries/well-being/
• Dolan P. Whose preferences count? Medical Decision Making. 1999;19(4):482-486. https://doi.org/10.1177/0272989X9901900416
• Hansson SO, Grüne-Yanoff T. Preferences. In: The Stanford Encyclopedia of Philosophy. First published 4 October 2006, substantive revision 14 March 2022. https://plato.stanford.edu/entries/preferences/
• HM Treasury. The Green Book: UK government guidance on appraisal. London: HM Treasury; 2026. Chapter 8, paragraph 8.6 and Figure 8. https://www.gov.uk/government/publications/the-green-book--2
• McNeil BJ, Pauker SG, Sox HC, Tversky A. On the elicitation of preferences for alternative therapies. New England Journal of Medicine. 1982;306(21):1259-1262. https://doi.org/10.1056/NEJM198205273062103
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). Published 31 January 2022, last updated 13 October 2026. Sections 4.3.1, 4.3.3, 4.3.4, 4.9.12, 6.3.2, 6.3.4, 6.3.7 and 6.3.8. https://www.nice.org.uk/process/pmg36
• Ubel PA, Loewenstein G, Jepson C. Whose quality of life? A commentary exploring discrepancies between health state evaluations of patients and the general public. Quality of Life Research. 2003;12(6):599-607. https://doi.org/10.1023/A:1025119931010