Decision Problem

Decision Problem: Framing the Choice an Economic Evaluation Informs

Every economic evaluation and every health economic model is built to answer one particular choice, and the decision problem is the written statement of that choice. The framing fixes which options are compared and which evidence counts as relevant, so a badly framed decision problem can produce a precise answer to the wrong question. This page sets out what the statement contains, how NICE uses the term formally through its scope and the company's decision problem table, how it shapes a model, and where usage differs between sources. It ends with a published appraisal in which the company's decision problem departed from the scope, and with the framing errors that modelling guidelines warn against.

What a decision problem specifies

A decision problem turns a policy or clinical question into a form that analysis can answer. The modelling guideline by Philips and colleagues asks for a clear statement of the decision problem prompting the analysis, covering the disease or condition, the patient group and the diagnostic or treatment pathways. It also asks for the objective of the evaluation (the maximand) to be defined, for the primary decision maker to be named because the decision maker affects which data are relevant, and for the options under evaluation to be clearly defined within the statement.

The table combines the clinical elements of a PICO question with the economic elements that an evaluation adds. In NICE appraisals the perspective comes from the reference case, which takes costs from the perspective of the NHS and personal social services, while the time horizon is one of the items the scope lists.

How NICE uses the term

NICE's formal use of "decision problem" in a technology appraisal centres on two documents that are meant to line up. The first is the scope, which NICE writes; the second is the company's own statement in its evidence submission.

The NICE manual (PMG36) states that scoping aims to define what question the evaluation will answer and what will and will not be included. Its listed contents include the population, the technology, the care pathway, the comparators, the principal outcome measures, the costs, the time horizon and subgroups. In the reference case summary of the same manual, the element "Defining the decision problem" is met by "The scope developed by NICE".

The economic evaluation itself then has to begin from the decision problem. The manual states that it "should start with a clear statement of the decision problem that defines the technologies being compared and the relevant patient groups", consistent with the scope, and that any differences must be justified. For medicines, once the submission arrives, the NICE technical lead and the external assessment group check whether the decision problem is specified appropriately with reference to the final scope, and a clarification letter follows if it is not.

The company states its decision problem in a table at section 1.1 of the evidence submission template, with reference to the final scope, as the NICE user guide for the template (PMG24) instructs. In published submissions, such as the one for romosozumab discussed below, the table has three columns: the final scope issued by NICE, the decision problem addressed in the company submission, and the rationale where the two differ. When the committee later weighs the economic model, the manual says it considers carefully whether the model reflects the decision problem at hand.

From decision problem to model structure

A model applies evidence to a specified decision problem, so its design follows from the statement rather than the reverse. The ISPOR-SMDM task force on model conceptualisation divides the work into two steps: conceptualising the problem, which turns knowledge of the care process or decision into a representation of the problem, and then conceptualising the model, which matches a modelling method to that problem.

The Philips guideline states that the appropriate model type, such as a decision tree or a Markov model, will be dictated by the stated decision problem and the causal relationships chosen within the model. The NICE manual asks for the type of model and the model structure to be justified for each new decision problem, and states that prior use of a structure in published work or earlier NICE submissions is not enough on its own.

The decision problem also sets the relevance test for evidence and parameters. The NICE manual links the critical appraisal of each study to its relevance to the decision problem, and asks for the external validity of published economic evaluations to be assessed against it. Disagreement about how the problem should be represented, for example which disease states or treatment sequences to include, is one source of structural uncertainty.

Two decisions inside one problem

In the decision-analytic approach set out by Claxton and colleagues, one decision problem contains two related choices. Claxton argued that decisions should rest on expected net benefit whether or not differences are statistically significant, and that the spread of net benefit matters for a separate question: whether more information should be acquired. Claxton, Sculpher and Drummond later set out a framework for NICE that keeps apart whether a technology should be adopted on current evidence and whether more research should be demanded to support future decisions.

The analysis of a decision problem therefore addresses both the adoption choice and whether further evidence is worth acquiring. Value of information analysis works on the second of these.

How usage differs between sources

The term is used more narrowly or more broadly depending on the source. In NICE appraisals the decision problem is a named document element with a fixed relationship to the scope, so a departure from the scope is visible and has to be justified in the table. In the modelling literature, as in the Philips guideline, the decision problem also carries the objective of the evaluation and the identity of the decision maker, which a clinical question usually leaves out.

The EU regulation on health technology assessment (Regulation (EU) 2021/2282) uses "assessment scope" for a related idea in joint clinical assessments. It defines the assessment scope as the set of parameters in terms of patient population, intervention, comparators and health outcomes requested jointly by Member States. A joint clinical assessment is a comparative analysis of the available clinical evidence, its report contains no value judgement or conclusions on overall clinical added value, and Member States remain responsible for conclusions on the value of a technology. The EU assessment scope therefore corresponds to the clinical part of a decision problem, while costs, perspective and the economic question stay with national bodies.

A decision problem is also wider than a research question framed with the PICO framework. PICO fixes the clinical question; the decision problem adds the perspective, time horizon, costs and decision maker that an economic evaluation needs. In a NICE appraisal the HTA scope is the body's formal statement of the decision problem, which the company's own statement must follow or depart from with justification. A model scope, the limits or boundaries of a model, follows from the decision problem: the Philips guideline notes that some of those boundaries, such as the patient group, are defined within the statement of the decision problem. The decision problem is in this sense the frame that holds together the population, intervention, evidence, parameters and resources of one analysis.

A published case: romosozumab for severe osteoporosis

The committee papers for NICE's appraisal of romosozumab for severe osteoporosis (TA791) show how a company's table records departures from the scope. The scope population was postmenopausal women with severe osteoporosis at high risk of fracture, and the company narrowed it to those who had had a recent major osteoporotic fracture within 24 months, describing them as the part of the licensed population with the greatest unmet need. The intervention became romosozumab for 12 months followed by alendronate, in line with the licensed 12-month course. The scope listed bisphosphonates, other active drugs and no active treatment as comparators, but the company's base case compared romosozumab with alendronate only, with scenario analyses against the other comparators in the scope by network meta-analysis, except ibandronic acid. Each departure changes the decision the analysis can support, and the table makes it visible before the model is reviewed.

Where decision problems go wrong

Errors in a decision problem can surface later as errors in the model or as disputes at committee. The Philips guideline asks for all feasible and practical options relating to the stated decision problem to be evaluated, and for justification whenever a feasible option is excluded. It adds that the options should not be constrained by the decision maker's prior views on outcomes, data availability or the constraints of current practice. A related point in the same guideline is that decisions have to be made regardless of data availability, so the evidence that happens to exist should not by itself define the question, even if missing data later limit the scope of the model.

Sources

• Claxton K. The irrelevance of inference: a decision-making approach to the stochastic evaluation of health care technologies. Journal of Health Economics. 1999;18(3):341-364.
• Claxton K, Sculpher M, Drummond M. A rational framework for decision making by the National Institute For Clinical Excellence (NICE). The Lancet. 2002;360(9334):711-715.
• European Parliament and Council of the European Union. Regulation (EU) 2021/2282 of 15 December 2021 on health technology assessment and amending Directive 2011/24/EU. Official Journal of the European Union, OJ L 458, 22.12.2021, p. 1. Recital 26 and Articles 2(6), 2(9), 8(6) and 9(1).
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). London: NICE; 2022, updated March 2026. Sections 2.1, 3.3, 3.4, 4.2, 4.6, 5.6 and 6.2.
• National Institute for Health and Care Excellence. Single technology appraisal and highly specialised technologies evaluation: user guide for company evidence submission template (PMG24). London: NICE; 2015, updated March 2026. Section 1.1.
• National Institute for Health and Care Excellence. Romosozumab for treating severe osteoporosis (TA791): committee papers. London: NICE; 2021. Document A, company evidence submission summary, Table 3.
• Philips Z, Ginnelly L, Sculpher M, Claxton K, Golder S, Riemsma R, et al. Review of guidelines for good practice in decision-analytic modelling in health technology assessment. Health Technology Assessment. 2004;8(36). Chapter 3.
• Roberts M, Russell LB, Paltiel AD, Chambers M, McEwan P, Krahn M. Conceptualizing a model: a report of the ISPOR-SMDM Modeling Good Research Practices Task Force-2. Medical Decision Making. 2012;32(5):678-689.