Decision Making

Decision Making in Health Care Resource Allocation

Every funding recommendation, formulary listing and treatment choice in a health system is the end point of a process that combines evidence, analysis and judgement, whether the choice falls to a national committee, an insurer or a clinician with a patient. The process matters because the same cost-effectiveness evidence can lead to different choices, depending on who decides, which criteria they apply, how they treat uncertainty and how the evidence is presented to them. The term is used here in that resource allocation sense rather than as a general account of the psychology of choice. Normative, descriptive and procedural accounts of it ask different questions, and in NICE technology appraisal economic analysis stops at a point where committee judgement begins, as an illustrative example in which the analysis brackets the decision shows.

Who makes health care resource allocation decisions

Health care decisions are made at several levels, and the decision maker shapes the criteria that apply. A national health technology assessment body or ministry decides on access for a whole population, a commissioner, insurer or hospital formulary committee decides for its own members or budget, and a clinician and patient decide on treatment for one person within the options those higher-level decisions leave open.

Daniels notes that in mixed systems such as that of the United States, decisions on funding new technologies are made by public agencies and by private insurers and managed care organisations, which make coverage decisions, while in most universal coverage systems they are made by public agencies or authorities. In England, according to the NICE manual (PMG36), an appraisal committee makes technology appraisal recommendations to NICE, which is responsible for publishing the final guidance.

Normative, descriptive and procedural accounts

Writing on decision making in health care answers three different questions, and confusing them is a common source of error. A normative account says how a decision maker with a stated objective and budget ought to choose. A descriptive account reports how decision makers actually choose, including the shortcuts and inconsistencies that appear in practice. A procedural account asks what makes a decision process fair and legitimate when people reasonably disagree about the principles that should govern it.

A normative model sets a benchmark, descriptive evidence shows where practice departs from it, and procedural rules govern how the judgements that a model cannot make are reached and explained.

The normative benchmark of expected net benefit under a budget

One normative approach used in HTA starts from a health system that wants the most health from a fixed budget. Spending on one option means health forgone somewhere else. Each option is therefore valued by the health it produces minus the health that its cost would have produced elsewhere, measured in quality-adjusted life years (QALYs) as net health benefit, defined below.

$$NHB_j = Q_j - \frac{C_j}{\lambda}$$

where $NHB_j$ is the net health benefit of option $j$ in QALYs, $Q_j$ is its expected health effect in QALYs, $C_j$ is its expected cost and $\lambda$ is the cost at which the health system can produce one QALY elsewhere, the cost-effectiveness threshold. The rule chooses the option with the highest expected net health benefit; with two options, this is the same as adopting the more costly and more effective option only when its ICER is below $\lambda$.

The NICE manual (PMG36) ties this benchmark to opportunity cost. Given the fixed budget of the NHS, the appropriate maximum acceptable ICER is that of the opportunity cost of programmes displaced by new, more costly technologies. The decision page explains why the choice itself is made on expected values and how uncertainty bears on whether to seek more evidence, using the expected value of perfect information.

Where analysis stops and committee judgement begins

Economic evaluation informs a decision without making it. The NICE manual (PMG36) states that the committee does not use a precise maximum acceptable ICER above which a technology would automatically be judged not cost effective, and that consideration of cost effectiveness is necessary but not the only basis for decision making. It uses a range instead, so that other factors carry more influence when the ICER is closer to the top of the range.

The committee also chooses which analysis to believe. The manual lists the factors that influence its decisions on cost effectiveness, including the plausibility of the model's inputs and assumptions, the committee's preferred modelling approach, the range and plausibility of the ICERs, and the likelihood of decision error and its consequences. The "most plausible ICER" that NICE compares with its range is therefore a committee judgement, not a figure taken directly from any one submission. The manual adds that the committee will want to be increasingly certain of cost effectiveness as the impact of adopting the technology on NHS resources increases.

Value judgements enter the same way. The manual states that the committee considers advice from NICE on making scientific and social value decisions, and that substantial effects on health inequalities can justify flexibility in the range normally considered cost effective. The full list of factors, including the severity modifier, is set out on the decision criteria page; the formal procedure for comparing an estimate with a threshold is covered on the decision rule page.

Deliberation, transparency and legitimacy

Because the analysis leaves room for judgement, the way that judgement is reached and explained becomes part of the decision. A deliberative committee has to show how the evidence and testimony before it led to its conclusion.

In the NICE manual (PMG36), the committee bases its recommendations on the evidence presented, including stakeholder statements and expert views, and may release provisional recommendations for consultation before amending them. The manual states that the credibility of NICE guidance depends on the transparency of the committee's decision-making process, and that the reasoning is explained in the committee discussion section of the guidance.

Daniels explains why such procedures matter. Where reasonable people disagree about principles, such as how much priority to give the worst off, a fair process allows agreement on what is legitimate. Key elements include transparency about the grounds for decisions, rationales that fair-minded people accept as relevant to meeting health needs fairly, and procedures for revising decisions in the light of challenges, which together make up "accountability for reasonableness". He cites the creation of NICE as a sign of awareness in the United Kingdom of the need for clear process, and places the approach between explicit and implicit rationing, in which principles are not fixed in advance, but the reasoning must be public.

How decisions are made in practice

Descriptive research shows that real decision makers depart from the normative benchmark in predictable ways. Kahneman distinguishes a fast, automatic intuitive mode of judgement from a slower, deliberate controlled mode, and reviews evidence that intuitive choices are shaped by how a problem is described.

The effect appears in clinical choices. Kahneman reports that describing outcomes of surgery and radiation therapy as survival rates rather than mortality rates produced a substantially higher preference for surgery, and that the effect was no less pronounced among experienced physicians than among patients. That evidence comes from patients and physicians; the same risk plausibly applies to how model results are framed in committee papers. The cognitive bias page covers these mechanisms.

A second descriptive approach treats past decisions as data. Dakin and colleagues modelled NICE decisions published by December 2011 as binary choices, a revealed preference approach to the weights implicit in deliberation. Cost-effectiveness alone correctly predicted 82% of decisions, and their most accurate model suggested a 50% chance of rejection at £40,000 per QALY, higher than the £20,000 to £30,000 range then stated, which the authors note may reflect other factors that are hard to quantify. Such findings describe past behaviour; they do not show what the threshold ought to be.

Worked example in which the analysis brackets the decision

The figures below are illustrative and are not taken from any appraisal. A new technology B is compared with current care A for an annual cohort of 1,500 eligible patients. The company's base case estimates an incremental cost of £14,000 and an incremental gain of 0.48 QALYs per patient, both discounted over the model's time horizon.

1. Calculate the ICER. The incremental cost-effectiveness ratio is calculated as follows.

$$ICER = \frac{14000}{0.48} \approx 29167$$

where $ICER$ is the incremental cost per QALY gained in pounds, rounded to the nearest pound, 14,000 is the incremental cost in pounds and 0.48 is the incremental QALY gain per patient. The estimate lies inside NICE's range of £25,000 to £35,000 per QALY.

2. Calculate net health benefit at both ends of the range. The NICE manual (PMG36) asks for expected net health benefits to be presented alongside ICERs, using values of £25,000 and £35,000 per QALY. At £25,000 the result is as follows.

$$NHB = 0.48 - \frac{14000}{25000} = 0.48 - 0.56 = -0.08$$

where $NHB$ is the incremental net health benefit per patient in QALYs and 0.56 QALYs is the health displaced by spending £14,000. At £35,000 the result is as follows.

$$NHB = 0.48 - \frac{14000}{35000} = 0.48 - 0.40 = 0.08$$

where 0.40 QALYs is the health displaced at that value. For the annual cohort of 1,500 patients, the total effect ranges from a loss of 120 QALYs to a gain of 120 QALYs, and the total incremental cost is 1,500 × £14,000 = £21,000,000, both discounted over the model's time horizon.

3. Apply the committee's preferred assumption. Suppose the committee judges that the company's assumption about how long the treatment effect lasts is optimistic and prefers a scenario with a gain of 0.40 QALYs, with the incremental cost assumed unchanged. The ICER becomes £14,000 ÷ 0.40 = £35,000 per QALY, at the top of the range, and the net health benefit is zero at £35,000 and −0.16 QALYs per patient at £25,000, a loss of 240 QALYs for the cohort.

4. Interpret. The analysis does not settle the choice. In the base case the sign of the net health benefit depends on where within the range a QALY is valued, and under the committee's preferred assumption the technology sits at the top of the range, above which the manual requires an increasingly stronger case. Within the range the committee must make explicit reference to the certainty around the ICER, uncaptured benefits and non-health factors, and health inequalities, and a resource impact of £21,000,000 per cohort may lead it to look for more certainty. The recommendation, which might be restricted to a subgroup or, if the technology is a medicine, made with managed access, rests on that deliberation and on the reasons published for it.

Decision making compared with related concepts

Several wiki pages deal with parts of the same process. The table separates what each one refers to.

The general cognitive sense of decision making, the mental process of choosing among alternatives, overlaps with the descriptive account above but leaves out the normative and procedural parts that matter most in resource allocation.

Common misunderstandings

Several errors recur when decision making in health care is discussed. Most come from treating one of the three accounts as if it were the whole process.

• The model makes the decision. An ICER or net benefit estimate is an input. NICE states that cost effectiveness is necessary but not the only basis for its decisions, and the committee chooses which analysis it finds most plausible.
• Revealed thresholds are the right thresholds. Estimates of how a committee has behaved describe past decisions; whether the implied value per QALY reflects opportunity cost is a separate question.
• Deliberation means unexplained discretion. The NICE manual requires the committee's reasoning to be published, and accountability for reasonableness also requires a route for challenge and revision.
• Behavioural findings apply only to patients. Framing effects have been shown among experienced physicians, and the same risk plausibly extends to how evidence is presented to committees.

Sources

• Dakin H, Devlin N, Feng Y, Rice N, O'Neill P, Parkin D. The influence of cost-effectiveness and other factors on NICE decisions. Research Paper 13/06. London: Office of Health Economics; 2013. Later published in Health Economics. 2015;24(10):1256-1271. doi:10.1002/hec.3086
• Daniels N. Accountability for reasonableness. BMJ. 2000;321(7272):1300-1301. doi:10.1136/bmj.321.7272.1300
• Kahneman D. Maps of bounded rationality: a perspective on intuitive judgment and choice. Nobel Prize lecture, 8 December 2002. In: Les Prix Nobel 2002. Stockholm: Nobel Foundation; 2003. pp. 449-489.
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). London: NICE; 2022, last updated 13 October 2026. Chapters 4 and 6.