QALY Shortfall

QALY Shortfall in NICE Severity Assessment

The NICE severity modifier rests on one comparison. Expected remaining QALYs for people who have a condition and receive current NHS care are set against expected remaining QALYs for a general population of the same age and sex, and the gap between the two numbers is the QALY shortfall. NICE reads that gap in two ways, as a count of QALYs (absolute shortfall) and as a share of the QALYs the matched population would expect (proportional shortfall), and either reading can raise the weight on a technology's QALY gains to 1.2 or 1.7. Both readings share the same two inputs, so any choice that moves either input, such as the life tables, the general population utilities or the model of current care, moves both measures at once. In August 2026 NICE's interim methods statement PMG51 changed one of those inputs by moving general population utilities to the UK EQ-5D-5L value set, for topics that start evaluation after its publication, while evaluations already under way keep the EQ-5D-3L values.

Expected QALYs with and without the condition

A QALY shortfall compares two forecasts of the same future. The first is the health a group of people would be expected to have over the rest of their lives if they did not have the condition. The second is the health the same group is expected to have living with the condition, treated as the NHS currently treats it. NICE defines severity in the technology appraisal and highly specialised technologies manual (PMG36) as the future health lost by people living with the condition with standard care in the NHS, and measures that loss in quality and length of life together.

Each forecast is an expected quantity of quality-adjusted life, and both are built the same way, by weighting the chance of being alive in each future year by the health-related quality of life expected in that year and discounting the result.

$$Q = \sum_{t=0}^{T} \frac{S(t)\, u(t)}{(1 + r)^{t}}, \qquad \text{QALY shortfall} = Q_{h} − Q_{c}$$

where $Q$ is expected remaining discounted QALYs, $S(t)$ is the probability of being alive in year $t$, $u(t)$ is the mean health state utility value of those alive in year $t$, $r$ is the annual discount rate, $T$ is the last year of remaining life, $Q_{h}$ is $Q$ for the matched general population and $Q_{c}$ is $Q$ for people with the condition under current care.

The general population forecast uses survival from national life tables and utilities from population surveys, combined by age and sex. McNamara and colleagues built quality-adjusted life expectancy norms for England in this way, combining EQ-5D-5L responses from 14,412 Health Survey for England participants (2017 and 2018 waves), mapped to the EQ-5D-3L value set with NICE's preferred mapping, with Office for National Statistics life tables for 2017 to 2019 through the Sullivan method, and reported the norms at discount rates of 0, 1.5 and 3.5 per cent. The condition forecast comes from the same evidence as the economic model, since current care is the comparator against which the new technology is judged.

Both forecasts look forward only. Health already lost before the decision point does not enter either number, so two groups with very different histories can have the same QALY shortfall.

Two ways NICE reads the same shortfall

QALY shortfall is the umbrella term for the gap itself. Absolute shortfall is that gap counted in QALYs, and proportional shortfall is the same gap divided by the matched population's expected QALYs. PMG36 calculates absolute QALY shortfall by subtracting the total QALYs that people living with the condition would be expected to have with current treatment over their remaining lifetime from the total QALYs that a general population with the same age and sex distribution would be expected to have. Proportional QALY shortfall divides that absolute figure by the remaining QALYs of the same matched general population.

$$\text{AS} = Q_{h} − Q_{c}, \qquad \text{PS} = \frac{Q_{h} − Q_{c}}{Q_{h}}$$

where AS is the absolute shortfall in QALYs, PS is the proportional shortfall as a fraction between 0 and 1, and $Q_{h}$ and $Q_{c}$ are as defined above.

Table 6.1 of PMG36 turns the two readings into a QALY weight. The committee applies the weight implied by whichever measure shows the greater severity, and a value that falls exactly on a cut-off takes the higher severity level.

Because both measures share the numerator $Q_{h} − Q_{c}$, they differ only in what the loss is scaled against. The absolute measure keeps the scale of the matched population's remaining health, which shrinks with age, while the proportional measure removes it. A loss of 7 QALYs is a large share of the future for a group that expects 8, and a small share for a group that expects 25.

Data choices that move both measures

The QALY shortfall is an estimate, and several inputs can push a condition across a cut-off. The NICE Decision Support Unit (DSU) guide to calculating severity shortfall, Technical Support Document 23 (TSD 23), sets out the data NICE expects, and PMG36 states that the estimates should focus on the specific population for which the technology will be used and be based on established clinical practice in the NHS.

For the general population side, TSD 23 says life expectancy by age and sex should come from Office for National Statistics life tables for England, weighted by the proportions of men and women in the patient population, and the January 2024 version of the guide recommended period life tables for 2017 to 2019 because of uncertainty about the long-term effect of Covid-19 on the data for 2020. Utilities are the second component. PMG51, published 27 August 2026, says the EQ-5D-5L value set should also be used to calculate estimates of health-related quality of life in the general population, including the utility values used to calculate absolute and proportional QALY shortfall for the severity modifier. NICE applies this to topics that start evaluation after PMG51 was published, which for appraisals means topics with an invitation to participate issued after that date, and evaluations already under way continue with 3L values. Before PMG51, NICE methods required 5L responses to be mapped to utilities based on the 3L value set, so the same population can have a different $Q_{h}$ depending on which set an evaluation follows.

On the condition side, $Q_{c}$ inherits every survival extrapolation and utility assumption in the model of current care. TSD 23 states that carers' quality of life, where a model includes it, is excluded from the QALYs for current care and from the incremental QALYs that attract a weight other than 1. It also states that the reference-case discount rate of 3.5 per cent should be used for severity calculations, so $r$ in the equation above takes that value for both forecasts. The DSU also publishes a severity shortfall calculator as a spreadsheet alongside TSD 23.

Worked example with two illustrative populations

All figures in this example are illustrative, discounted at 3.5 per cent, and not drawn from any appraisal. Population A has a mean age in mid-life and population B is much older. The steps show that the same shortfall idea can trigger the weight through different measures.

1. Population A. The matched general population expects 17.5 discounted QALYs and people with the condition under current care expect 4.9.

$$\text{AS} = 17.5 − 4.9 = 12.6, \qquad \text{PS} = \frac{12.6}{17.5} = 0.72$$

where the absolute shortfall of 12.6 QALYs lies in the 12 to 18 band and the proportional shortfall of 0.72 lies below 0.85, so the absolute measure sets the weight at 1.2.

2. Population B. The matched general population expects 8.0 discounted QALYs and people with the condition expect 0.36.

$$\text{AS} = 8.0 − 0.36 = 7.64, \qquad \text{PS} = \frac{7.64}{8.0} = 0.955$$

where the absolute shortfall of 7.64 QALYs is below 12 and the proportional shortfall of 0.955 is at least 0.95, so the proportional measure sets the weight at 1.7. With only 8.0 QALYs left in the matched population, population B could never reach an absolute shortfall of 12, even if the condition caused immediate death.

3. A different general population input. Suppose a different set of population utilities gave population A a matched value of 16.8 QALYs, with $Q_{c}$ unchanged.

$$\text{AS} = 16.8 − 4.9 = 11.9, \qquad \text{PS} = \frac{11.9}{16.8} = 0.708$$

where both measures now fall below their lowest cut-offs, so the weight drops to 1. A change of 0.7 QALYs on the general population side alone, with no change in the evidence about the condition, removes the modifier.

4. Effect on the cost per QALY. In population A a new technology costs GBP 21,000 more than current care and gains 0.6 QALYs, an unweighted ICER of 21,000 / 0.6 = 35,000, or GBP 35,000 per QALY. Weighted at 1.2, the gain counts as 0.6 × 1.2 = 0.72 QALYs and the ratio becomes 21,000 / 0.72 = 29,167, or GBP 29,167 per QALY. In population B a technology costs GBP 17,000 more and gains 0.4 QALYs, an unweighted ratio of 17,000 / 0.4 = 42,500. Weighted at 1.7, the gain counts as 0.4 × 1.7 = 0.68 QALYs and the ratio becomes 17,000 / 0.68 = 25,000.

The two populations reach a weight through different measures, so a rule that reads the shortfall both ways gives extra weight to conditions that one of the two measures alone would miss. Step 3 shows that the general population inputs matter as much as the clinical evidence when a shortfall sits near a cut-off, and step 4 shows that the weight multiplies only the incremental QALYs, never $Q_{h}$ or $Q_{c}$ themselves.

How the shortfall enters a funding decision

A QALY shortfall changes nothing on its own. It decides which weight applies, and the weight then multiplies the QALYs the new technology gains over current care, which in effect raises the cost-effectiveness threshold for that technology in proportion to the weight. PMG36 treats the extent of unmet need as part of the severity definition, so a condition with poor current care tends to show a larger shortfall.

Because the shortfall is calculated against current care based on established NHS practice, a technology that becomes established practice and improves outcomes can raise $Q_{c}$ for later evaluations in the same condition, which would lower the shortfall for that condition. PMG36 also limits where the weighting is used. The severity modifier is not initially applied to technology appraisals of HealthTech, and highly specialised technologies do not receive the additional weight.

QALY shortfall and neighbouring measures

Several measures that also count health lost sit close to QALY shortfall. A QALY gain is a difference between two treatments for the same people, while a QALY shortfall is a difference between people with and without the condition when both receive what they would receive now. Years of life lost and the DALY count health lost across a whole population, usually against a standard life expectancy, and feed estimates of the burden of illness. A QALY shortfall is a per-person expected value for the group a technology would treat, includes quality of life as well as length of life, and is matched to that group's own age and sex mix.

Life expectancy and quality-adjusted life expectancy are inputs to the shortfall, not measures of severity in themselves. The severity modifier is the weight, and the shortfall is the evidence that sets it. Absolute and proportional shortfall are the two readings of the shortfall that NICE uses. Stolk and colleagues illustrated an equity adjustment procedure using proportional shortfall in 2004.

Limitations and common misreadings

The shortfall framework embeds value judgements as well as data choices. Giving more weight to QALYs for people with a larger expected loss means fewer QALYs for other patients whose care is displaced. The calculation is also open to several errors of method and reading.

• Treating the shortfall as fixed. $Q_{h}$ depends on the life tables and value set used, and $Q_{c}$ depends on the comparator model, so a reported shortfall is only as stable as those inputs. Near a cut-off, small changes on either side can switch the weight, as step 3 of the example shows.
• Weighting the wrong quantity. The weight applies to incremental QALYs of the new technology. Multiplying total QALYs, or the shortfall itself, overstates the effect of the modifier.
• Mismatched comparison groups. A general population that does not match the patients' age and sex distribution gives the wrong $Q_{h}$, which shifts both measures at once.
• Discounting inconsistently. TSD 23 states that 3.5 per cent should be used for severity calculations, so a shortfall calculated undiscounted, or at another rate, does not match the cut-offs.
• Reading severity as total lifetime disadvantage. The calculation looks forward only and matches on age and sex alone, so it ignores health lost before the decision point and any difference in expected health between social groups.
• Counting carers. Carer QALYs included in a model for other reasons are taken out of the shortfall calculation and out of the weighted gain under TSD 23.

Sources

• 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. Section 6.2 and Table 6.1. https://www.nice.org.uk/process/pmg36/chapter/committee-recommendations-2
• National Institute for Health and Care Excellence. Interim methods statement: implementing the EQ-5D-5L value set (PMG51). Published 27 August 2026. Statement on using the EQ-5D-5L value set. https://www.nice.org.uk/process/pmg51/chapter/statement-on-using-the-eq-5d-5l-value-set
• NICE Decision Support Unit. Technical Support Document 23: a guide to calculating severity shortfall for NICE evaluations. University of Sheffield. January 2024; listed by the DSU as updated September 2026. https://sheffield.ac.uk/nice-dsu/tsds/severity-shortfall-tsd
• McNamara S, Schneider PP, Love-Koh J, Doran T, Gutacker N. Quality-adjusted life expectancy norms for the English population. Value in Health. 2023;26(2):163-169. doi:10.1016/j.jval.2022.07.005
• Stolk EA, van Donselaar G, Brouwer WB, Busschbach JJ. Reconciliation of economic concerns and health policy: illustration of an equity adjustment procedure using proportional shortfall. PharmacoEconomics. 2004;22(17):1097-1107. doi:10.2165/00019053-200422170-00001