Concept Architecture
Concept
Theoretically, a Choice Experiment is a stated preference method used to quantify individuals' preferences by asking them to choose between hypothetical alternatives defined by multiple attributes. The method is founded on random utility theory, which assumes that individuals select the alternative that provides the greatest utility. In health economics, choice experiments are widely used to estimate preferences for healthcare interventions, service delivery, treatment attributes and health policies when revealed preference data are unavailable.
Mathematically, choice experiments model the utility associated with each alternative as the sum of a systematic component and a random error term. Utility is expressed as a function of the attributes describing each alternative, and model parameters are estimated using discrete choice models such as the conditional logit, mixed logit or multinomial logit model. Estimated coefficients quantify the relative importance of individual attributes and facilitate estimation of trade-offs and willingness to pay where a cost attribute is included.
In practice, respondents complete a series of experimentally designed choice tasks, each requiring selection of one preferred alternative. Responses are analysed using maximum likelihood estimation to estimate preference weights. Choice experiments are extensively applied in health technology assessment, health service design, patient preference studies and valuation of healthcare programmes.
Purpose
Used to quantify preferences for healthcare interventions and their attributes by analysing choices between hypothetical alternatives, thereby supporting economic evaluation, health policy and resource allocation.
Mathematical Formulae
Primary Formula
Utility function:
U?? = V?? + �??
where:
- U?? = utility of alternative j for individual i
- V?? = systematic utility
- �?? = random error component
Supporting Formulae
Systematic utility:
V?? = ?? + ??X? + ??X? + ... + ??X?
Conditional logit probability:
P?? = exp(V??) / ?exp(V??)
Marginal willingness to pay:
WTP = ??????????? / ?c???
Related Mathematical Methods
- Random Utility Theory
- Conditional Logit Model
- Mixed Logit Model
- Multinomial Logit Model
- Maximum Likelihood Estimation
- Discrete Choice Modelling
Example
A choice experiment evaluates patient preferences for diabetes treatments using four attributes: administration route, reduction in HbA1c, risk of hypoglycaemia and monthly cost. Analysis using a conditional logit model estimates a cost coefficient of ?0.020 and a coefficient for avoiding injections of 0.600.
Willingness to Pay = ?0.600 / ?0.020 = �30 per month
Patients are therefore estimated to be willing to pay approximately �30 per month to avoid injections, all else being equal.
Excel Implementation
| Function | Example Formula | Health Economics Application |
|---|---|---|
| EXP | =EXP(B2) | Calculates the exponential component of the utility function. |
| SUM | =SUM(C2:C5) | Computes the denominator of the choice probability. |
| LET | =LET(U,EXP(B2),Den,SUM(C2:C5),U/Den) | Calculates predicted choice probabilities. |
| SUMPRODUCT | =SUMPRODUCT(AttributeRange,CoefficientRange) | Calculates systematic utility for an alternative. |
| LN | =LN(P2) | Computes log-likelihood components for model estimation. |
VBA (Optional)
VBA can automate calculation of utilities, predicted choice probabilities and summary outputs for discrete choice experiment datasets exported from statistical software.
Sources
- Louviere JJ, Hensher DA, Swait JD. Stated Choice Methods: Analysis and Applications. Cambridge University Press.
- Lancsar E, Louviere J. Conducting discrete choice experiments to inform healthcare decision making. Pharmacoeconomics.
- Bridges JFP, Hauber AB, Marshall D, et al. Conjoint Analysis Applications in Health?a Checklist. Value in Health.
- ISPOR Conjoint Analysis Good Research Practices Task Force Reports.
- Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the Economic Evaluation of Health Care Programmes. Oxford University Press.
Related Concepts (3)
Library
Publications
1
Conjoint Analysis Applications in Health — A Checklist: A Report of the ISPOR Good Research Practices for Conjoint Analysis Task Force — Bridges, Hauber, Marshall, Lloyd, Prosser, Regier, Johnson & Mauskopf, Vol. 14, No. 4 ed., 2011 (Value in Health)
The ISPOR good-practice checklist for conjoint analysis and discrete-choice experiments in health — the stated-preference methods used to elicit patient and public preferences over treatment attributes for value assessment and priority-setting.
Journal ArticleView source →
Frequently Asked Questions (6)
What is a choice experiment?
A stated preference method in which respondents repeatedly pick their preferred option from sets of hypothetical alternatives that vary across defined attributes.
Source: Louviere, Hensher & Swait 2000
How is a choice experiment designed?
Attributes and their levels are defined, and combinations are assembled into alternatives which are grouped into choice sets. The combinations shown are selected by a formal experimental design rather than at random, so that the effect of each attribute can be separated from the others and the maximum information is extracted from a limited number of questions. Designs are constructed to be efficient given assumptions about the likely parameter values, and including a no-treatment or opt-out alternative allows the value of the service as a whole to be estimated.
Source: Louviere, Hensher & Swait 2000
How many respondents and tasks does a choice experiment need?
The requirement depends on the number of parameters to be estimated, the design efficiency and the precision sought, rather than on any fixed rule, and sample size calculations for choice experiments are less standardised than for trials. Each respondent typically completes multiple choice sets, which multiplies the observations without multiplying recruitment, at the cost of fatigue after a point that varies with task complexity. Published health studies commonly use samples in the low hundreds, which supports overall estimates and not detailed subgroup analysis.
Source: de Bekker-Grob, Ryan & Gerard 2012
How is a choice experiment analysed?
The analysis assumes each respondent attaches a value to each attribute level, chooses the alternative with the highest total, and that an unobserved random component accounts for departures from that rule. Estimating the model yields a weight for each level. Where cost is included as an attribute, the ratio of another attribute's weight to the cost weight expresses it in money. More elaborate specifications allow the weights to vary between respondents, which is usually a better description of the data than assuming everyone values things identically.
Source: Lancsar & Louviere 2008
What is a choice experiment used for in health?
Its distinctive use is predicting response to a service that does not yet exist, since the estimated weights can be applied to any combination of attribute levels rather than only to those shown. That supports questions such as how uptake of a screening programme would change if it moved to a different setting, or how much waiting time a population would accept in exchange for a closer facility. It is also used to establish the relative weight patients place on effectiveness against side effects, to inform workforce policy through the job characteristics that influence where staff choose to work, and to derive values for use where a monetary framework is required.
Source: de Bekker-Grob, Ryan & Gerard 2012
What threatens the validity of a choice experiment?
The most consequential threat is that the design bounds the answer, since any characteristic the analyst did not include is valued at zero by construction and the levels chosen determine how important each attribute can appear. Beyond that, hypothetical choices carry no consequence, so stated willingness to pay exceeds what people actually pay, and tasks exceeding what respondents can process are simplified in ways the model does not represent. Whether the resulting estimates predict real behaviour is tested rarely, and the studies that have tested it report mixed results, which is why the method is better used for relative weights than for absolute values.
Source: Bridges et al. 2011
Trust Record
Verified by Dr Darrin Baines
British health economist
Professional identity: darrinbaines.org
Verification date: 31 Jul 2025
Content version: 1.0.0
Canonical Identity
- Persistent URI
- https://healtheconomics.wiki/concept/choice-experiment
- Term code
- HE-EE-CBA-007
Stable URI · Machine-readable · Resolvable · CC BY 4.0