Equity-Weighted Analysis

Equity-Weighted Analysis of QALYs in Economic Evaluation

Equity-weighted analysis multiplies the QALYs gained by chosen groups by a weight, usually above one for priority groups, before costs and benefits are compared. A prominent current use is the NICE severity modifier, which lets an appraisal committee multiply QALYs by 1.2 or 1.7 for treatments of severe conditions. Weights change both the ICER and net health benefit. The NICE manual (PMG36) sets rules for their use, public preference studies give evidence on their size, and the method is distinct from distributional cost-effectiveness analysis.

From "a QALY is a QALY" to weighted QALYs

Standard cost-effectiveness analysis adds up health gains without asking who receives them, a rule Cookson and colleagues summarise as "a QALY is a QALY". The NICE reference case (PMG36, Table 4.1) states that an additional QALY has the same weight whatever the other characteristics of the person receiving it, except in specific circumstances.

Equity-weighted analysis relaxes that rule. Cookson and colleagues describe two ways of doing it, either equity weights applied to the health benefits of people with given characteristics, or a single equity parameter that sets how much concern for reducing health inequity is traded against total health. They note that many proposed weighting systems rest on health characteristics, especially current severity of illness or lifetime health, rather than on social characteristics such as income. An equity parameter, such as the inequality aversion of a social welfare function, implies weights that change with the existing distribution of health.

How the weights enter the calculation

In the NICE approach, the weights act on the incremental QALY gain of the technology, not on costs or on total QALYs. Wailoo and colleagues describe NICE severity weights in the same way, as applied to the incremental QALYs estimated in the cost-effectiveness assessment. For a technology whose effects differ between groups, the weighted gain, ICER and net health benefit are

$$\Delta E_w = \sum_{g} w_g \, \Delta E_g, \qquad \text{ICER}_w = \frac{\Delta C}{\Delta E_w}, \qquad \text{NHB}_w = \Delta E_w - \frac{\Delta C}{\lambda}$$

where $\Delta E_w$ is the weighted QALY gain, $w_g$ is the equity weight for group $g$, $\Delta E_g$ is the QALY gain in group $g$, $\Delta C$ is the incremental cost, $\text{ICER}_w$ is the weighted incremental cost-effectiveness ratio, $\text{NHB}_w$ is the weighted net health benefit and $\lambda$ is the value placed on a QALY.

For a single group, dividing the ICER by a weight is the same as multiplying the cost-effectiveness threshold by it. Against the range of £25,000 to £35,000 per QALY in PMG36 section 6.3.7, a weight of 1.7 is equivalent to a range of £42,500 to £59,500. Cookson and colleagues report that a Norwegian committee took this route in 2014, proposing different threshold ranges for three categories of disease severity.

Wailoo, Tsuchiya and McCabe warn that multiplying unweighted QALYs after the analysis, the approach NICE's modifier and the example below take, will in many situations not reflect the preferences on which the weights are based, and that such results need further work before they are a fit basis for resource allocation.

The NICE severity modifier as equity weights in practice

The NICE manual treats severity as a decision-making modifier. PMG36 section 6.2.11 defines modifiers as factors that go beyond the QALY, and value judgements, and allows them to be applied qualitatively through committee discussion or quantitatively through QALY weighting. Severity is the future health lost by people living with the condition on current NHS care (section 6.2.12), measured by absolute shortfall and proportional shortfall in QALYs against the general population of the same age and sex.

Whichever measure implies the greater severity applies, and a value on a cut-off takes the higher level (section 6.2.18). The unmet need page works through a shortfall calculation.

The modifier replaced the end-of-life modifier in January 2022. Wailoo and colleagues, writing with NICE staff, describe the end-of-life rule as a weight of around 1.7 for treatments that extend life by at least three months in patients with a life expectancy under 24 months. The severity cut-offs and weights were set from analysis of past appraisals so that the change would be broadly opportunity cost neutral, matching the average QALY weight per decision. Applied retrospectively to 464 decisions made before 2022, the new bands give a mean severity weight of 1.116, against 1.125 across 91 decisions under the new rules; the authors found no significant difference and concluded that the change had been approximately cost neutral.

In NICE technology appraisal, social characteristics do not earn a QALY weight. PMG36 section 4.12.7 states that a distributional analysis should not weigh costs or benefits differently by the social characteristics of the people affected. Where health inequality impacts are substantial, section 6.2.38 lets the committee apply flexibility to the range of acceptable ICERs instead.

What public preference studies show about the weights

Empirical weights come mainly from stated preference surveys of the general public. Gu and colleagues reviewed 64 studies using eight methods; the discrete choice experiment and the person trade-off were the most popular, but only 34% of studies calculated distributional weights, mostly with the person trade-off. Despite heterogeneity, the young tended to be favoured over the old and the more severely ill over the less severely ill, while people with self-induced illness or high socioeconomic status tended to receive lower priority. Most studies of end-of-life care found weak or no support for a premium.

The size of a weight depends on the method used to elicit it. In the Social Value of a QALY project, Baker and colleagues interviewed 587 members of the public in England. Discrete choice questions showed no strong effect of age and severity beyond the QALY gain, whereas matching questions gave weights three to four times higher for some groups than others, depending on how responses were aggregated. In an online discrete choice survey of 3,669 members of the UK general population, Rowen and colleagues found a small and inconsistent effect of burden of illness but a consistently positive one for end of life.

The NICE weights were calibrated on past appraisals, not taken from such surveys. Hausman argues that the preference literature uses a variety of notions of severity, so it is unclear how far it supports weights based on absolute or proportional shortfall. Elicited weights for socioeconomic inequality, expressed as inequality aversion, are covered on the equity-efficiency trade-off and Atkinson index pages.

Worked example with one medicine and two subgroups

All figures in this example are illustrative. A new medicine is appraised in two subgroups. In subgroup A, 300 patients have a proportional shortfall of 0.96, which earns a weight of 1.7; each gains 0.6 QALYs at an extra cost of £30,000. In subgroup B, 700 patients have a proportional shortfall of 0.80 and an absolute shortfall below 12, so the weight is 1; each gains 0.8 QALYs at an extra cost of £24,000.

1. Unweighted ICERs. For subgroup A the ICER is 30,000 / 0.6 = £50,000 per QALY, and for subgroup B it is 24,000 / 0.8 = £30,000 per QALY.

2. Weighted ICERs. Subgroup A has weighted QALYs of 1.7 × 0.6 = 1.02, so the weighted ICER is 30,000 / 1.02 = £29,412 per QALY. Subgroup B is unchanged at £30,000.

3. Net health benefit per patient at £35,000 per QALY, one of the two values PMG36 section 6.3.2 asks for. The displaced health is 30,000 / 35,000 = 0.857 QALYs in A and 24,000 / 35,000 = 0.686 in B.

At £25,000 per QALY the displaced health is 1.2 QALYs per patient in A and 0.96 in B, so net health benefit is negative in both, at −0.18 weighted and −0.6 unweighted in A, and −0.16 in B.

4. Find the switching weight. Subgroup A breaks even when $w \times 0.6 = 30000 / \lambda$, so $w^{*} = 30000 / (0.6 \times 35000) = 1.43$ at £35,000 and $w^{*} = 30000 / (0.6 \times 25000) = 2.0$ at £25,000.

5. Interpretation. The weight moves subgroup A from £50,000 to below £30,000 per QALY, but the real cost and the displaced care do not change. Recommending the medicine for subgroup A at £35,000 per QALY gives a weighted net health benefit of 48.9 QALYs while total population health falls by 77.1 QALYs. That loss is the health given up to act on the severity judgement. A weight of 1.43 would be enough at the top of the range, but no NICE severity weight reaches the 2.0 needed at £25,000.

Arguments for and against equity weighting

The main case for weighting is that it puts a number on judgements that would otherwise stay implicit. Cookson and colleagues recommend equity-weighting analysis as an aid to deliberation, run as a sensitivity analysis over different weights. Transparency is also the reason PMG36 section 6.3.2 asks for net health benefit with and without the QALY weighting.

The case against has several strands.

1. Opportunity cost. The weight raises the value of the gains, but the health displaced elsewhere stays unweighted. PMG36 section 6.3.2 accepts that a technology with negative unweighted net health benefit, which reduces overall health, may still be recommended on an ethical and moral rationale. Cookson and colleagues add that the distribution of opportunity cost matters as well, since cuts to other public services may fall more heavily on poorer people.
2. Double counting. A weight should apply only to value outside the QALY, as section 6.2.11 implies. PMG36 applies no severity weight to highly specialised technologies because severity is already captured in topic selection (section 6.2.20), and Sandman and Hofmann argue that weighting both unmet need and severity counts the same thing twice.
3. Weak and unstable evidence. Weights vary with the elicitation method, and Gu and colleagues call for further work before robust distributional weights exist.
4. The basis of the weight. Hausman argues that NICE's severity weighting looks only at future health and only at health, not well-being in general, so it is not well supported by egalitarian or prioritarian philosophy. Wailoo and colleagues note that shortfall-based severity has an intrinsic link with age, since life expectancy enters the calculation.

Equity weighting and distributional cost-effectiveness analysis

Equity weighting and distributional cost-effectiveness analysis (DCEA) both serve health equity but do different jobs. Weighting changes the value of a QALY for chosen groups and still reports one weighted total. DCEA estimates how costs, health gains and health opportunity costs fall across population groups, as PMG36 section 4.12.1 and Cookson and colleagues describe it, and the resulting equity impact can then be set against total health, for example through an equity-weighted index of social welfare.

The two meet when a DCEA uses an inequality aversion parameter, because that parameter implies weights for each group. NICE keeps them apart in technology appraisal. PMG36 section 4.12.3 limits DCEA to supporting evidence, while the committee can apply severity weights to the QALYs in its decision. A severity weight therefore says nothing about deprivation, because it looks at health lost to a condition, not at the social position of the people who have it.

Sources

• Baker R, Bateman I, Donaldson C, Jones-Lee M, Lancsar E, Loomes G, et al. Weighting and valuing quality-adjusted life-years using stated preference methods: preliminary results from the Social Value of a QALY Project. Health Technology Assessment. 2010;14(27):1-162. https://doi.org/10.3310/hta14270
• Cookson R, Mirelman AJ, Griffin S, Asaria M, Dawkins B, Norheim OF, Verguet S, Culyer AJ. Using cost-effectiveness analysis to address health equity concerns. Value in Health. 2017;20(2):206-212. https://pmc.ncbi.nlm.nih.gov/articles/PMC5340318/
• Gu Y, Lancsar E, Ghijben P, Butler JR, Donaldson C. Attributes and weights in health care priority setting: a systematic review of what counts and to what extent. Social Science and Medicine. 2015;146:41-52. https://doi.org/10.1016/j.socscimed.2015.10.005
• Hausman DM. Problems with NICE's severity weights. Social Science and Medicine. 2024;348:116833. https://doi.org/10.1016/j.socscimed.2024.116833
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). London: NICE; published 31 January 2022, last updated 31 March 2026. Table 4.1 and sections 4.12.1, 4.12.3, 4.12.7, 6.2.10 to 6.2.20, 6.2.38, 6.3.2 and 6.3.7. https://www.nice.org.uk/process/pmg36
• Rowen D, Brazier J, Mukuria C, Keetharuth A, Risa Hole A, Tsuchiya A, et al. Eliciting societal preferences for weighting QALYs for burden of illness and end of life. Medical Decision Making. 2016;36(2):210-222. https://doi.org/10.1177/0272989X15619389
• Sandman L, Hofmann B. Why we don't need "unmet needs"! On the concepts of unmet need and severity in health-care priority setting. Health Care Analysis. 2019;27(1):26-44. https://doi.org/10.1007/s10728-018-0361-2
• Wailoo A, Bell Gorrod H, Dunning L, Kenny J, Leckenby E, Shah K. The NICE experience of designing and utilising severity weights. Health Policy OPEN. 2026;10:100156. https://doi.org/10.1016/j.hpopen.2025.100156
• Wailoo A, Tsuchiya A, McCabe C. Weighting must wait: incorporating equity concerns into cost-effectiveness analysis may take longer than expected. PharmacoEconomics. 2009;27(12):983-989. https://doi.org/10.2165/11314100-000000000-00000