ReQoL-UI

ReQoL-UI in Mental Health Cost-Utility Analysis

The ReQoL-UI gives mental health researchers a utility score built from questions written for people using mental health services, so that a trial or a service in mental health economics can count QALYs without depending only on the EQ-5D. Keetharuth and colleagues developed the Recovering Quality of Life questionnaires partly because generic measures such as the EQ-5D and SF-6D had been criticised as unsuitable for people with mental health conditions, and the utility index was the step that let those questionnaires feed a cost-utility analysis. The UK scoring algorithm rests on time trade-off interviews in which 305 members of the UK public valued 64 health states, and the utilities it produces lie between −0.195 for the worst state and 1 for the best. Because the index asks about hope and self-perception as well as physical health, its QALYs can differ from EQ-5D QALYs collected from the same patients, and that difference can change a mental health cost-effectiveness result.

The ReQoL questionnaires behind the index

The ReQoL-UI sits inside two longer questionnaires, the ReQoL-10 and the ReQoL-20, and it cannot be understood apart from them. Both are self-reported outcome measures designed around recovery, a view of mental health care that values a meaningful life as well as fewer symptoms.

Keetharuth and colleagues published the two versions in 2018 in the British Journal of Psychiatry. They built the item set through qualitative item development followed by quantitative item reduction, factor analysing data from more than 6,500 service users and using item response theory models to inform which items to keep. The items are worded both positively and negatively and cover seven themes, namely activity, hope, belonging and relationships, self-perception, well-being, autonomy and physical health. The authors reported that both versions achieved acceptable internal consistency, test and retest reliability above 0.85, known-group differences, convergence with related measures and responsiveness over time.

The ReQoL-10 and ReQoL-20 produce profile scores on a recovery scale. Those scores are not anchored at full health and dead, so on their own they cannot weight life years in a QALY calculation. The utility index closes that gap.

How the ReQoL-UI classification and the UK value set were built

The index has its own descriptive system, smaller than either questionnaire. The developers' site states that the ReQoL-UI uses six mental health items and one physical health item that are contained in both the ReQoL-10 and the ReQoL-20, which is why preference weights can be calculated from either version. Each item has five response levels. Thai and colleagues describe the six mental health items as sharing the same five levels running up to most or all of the time, while the physical health item runs from no problems to very severe problems.

Keetharuth, Rowen, Bjorner and Brazier described the valuation in four stages. These were the derivation of the classification system, the choice of health states for valuation using item response theory, a time trade-off survey and the modelling of preference weights. The survey sample was representative of the UK public by age, sex and region, and interviewers administered the time trade-off tasks face to face with props. The 305 participants valued 64 health states between them.

Several regression models were fitted to the time trade-off data. The authors chose a random effects model because its coefficients were significant and consistent and it gave the best fit, and the resulting value set assigns utilities between −0.195 and 1. The paper appeared online in November 2020 and in the 2021 volume of Value in Health. The developers distribute the scoring algorithms in Excel, SPSS and Stata formats through Oxford University Innovation.

Turning item responses into a utility

The UK scoring model does not attach one weight to each item. Keetharuth and colleagues chose one item from each of the six mental health themes and added the physical health item, and they treated the six mental health items as a single unidimensional component. Responses to those six items are summarised as an item response theory score for mental health, and the time trade-off values were regressed on that score together with dummy variables for the levels of the physical health item.

A respondent's answers therefore define one health state, and the algorithm gives that state a utility on the scale where 1 is the best possible state and values below 0 are judged worse than dead. The UK coefficients come with the developers' scoring algorithm, and an analysis that rebuilds them from summary tables risks small errors in every utility. The structure can be written as follows.

$$U = f(\theta) + \sum_{k=2}^{5} \beta_k D_k$$

where $U$ is the utility of the health state, $\theta$ is the item response theory mental health score from the six mental health items, $f(\theta)$ is the estimated relationship between that score and utility, $D_k$ equals 1 when the physical health item is answered at level $k$ and 0 otherwise, and $\beta_k$ is the change in utility for that level against the best level.

Utilities measured at trial visits become QALYs through the area under the utility curve. With linear change between two visits the calculation is as follows.

$$Q = \frac{u_0 + u_1}{2} \times T$$

where $Q$ is the QALYs accrued in the interval, $u_0$ and $u_1$ are the utilities at the start and end of the interval and $T$ is the length of the interval in years.

Illustrative worked example comparing ReQoL-UI and EQ-5D QALYs

All numbers here are illustrative and do not reproduce the published UK coefficients. They show the steps from item responses to a cost per QALY and why the choice of utility measure can change the answer.

1. Scoring one health state. A participant's six mental health answers give an illustrative mental health score that the algorithm places at 0.80 on the utility scale, and the answer to the physical health item falls at a level carrying an illustrative coefficient of −0.12. The utility is 0.80 − 0.12 = 0.68.

2. QALYs over 26 weeks with the ReQoL-UI. Both arms of a trial of a psychological therapy start with a mean ReQoL-UI utility of 0.70. At 26 weeks, or 0.5 years, the therapy arm reaches 0.76 and the usual care arm 0.74. The therapy arm accrues (0.70 + 0.76) / 2 = 0.73, and 0.73 × 0.5 = 0.365 QALYs. The usual care arm accrues (0.70 + 0.74) / 2 = 0.72, and 0.72 × 0.5 = 0.36 QALYs. The difference is 0.365 − 0.36 = 0.005 QALYs.

3. The same trial scored with the EQ-5D-5L. Suppose the EQ-5D-5L utilities start at 0.60 in both arms and reach 0.72 with therapy and 0.66 with usual care. The therapy arm accrues (0.60 + 0.72) / 2 = 0.66, and 0.66 × 0.5 = 0.33 QALYs. The usual care arm accrues (0.60 + 0.66) / 2 = 0.63, and 0.63 × 0.5 = 0.315 QALYs. The difference is 0.33 − 0.315 = 0.015 QALYs.

4. Cost per QALY. The therapy adds GBP 450 per person. With ReQoL-UI QALYs the ICER is 450 / 0.005 = 90,000, or GBP 90,000 per QALY. With EQ-5D-5L QALYs the ratio is 450 / 0.015 = 30,000, or GBP 30,000 per QALY.

The same patients and costs produce ratios three times apart. In this illustration the EQ-5D-5L registers more of the change, which is the direction Franklin and colleagues found in a published trial of a mental health intervention, where they attributed it to the EQ-5D-5L's better responsiveness and its anxiety or depression and usual activities items. The example does not show which measure is right, and another trial could go the other way. A submission that reports both measures as scenarios when both were collected is therefore more informative, and a ratio cannot be read without knowing which utility measure lies under it.

Evidence on how the ReQoL-UI performs against the EQ-5D-5L

Published head-to-head comparisons test different properties. Two of them give a sense of what changes when the ReQoL-UI is used in place of the EQ-5D-5L.

Franklin and colleagues reported in 2022 on a trial in people with anxiety or depression in which 361 participants were randomised in a 2 to 1 ratio, with QALYs calculated over 8 weeks. The EQ-5D-5L crosswalk produced higher incremental QALYs than either the value set for England or the ReQoL-UI. The value set for England and the ReQoL-UI gave similar unadjusted QALYs but contrasting baseline-adjusted QALYs. The probability that the intervention was cost-effective below GBP 30,000 per QALY ranged between 6 per cent, with complete-case ReQoL-UI baseline-adjusted QALYs, and 64.3 per cent, with multiple-imputation EQ-5D-5L crosswalk unadjusted QALYs.

Xu and colleagues surveyed 500 members of the Hong Kong general population by telephone in July 2020 and compared the EQ-5D-5L, the ReQoL-UI and the ICECAP-A. Mean scores were 0.92 for both the ReQoL-UI and the EQ-5D-5L and 0.85 for the ICECAP-A. The authors judged all three measures psychometrically robust, with the EQ-5D-5L more suitable for physical health-related quality of life and the ICECAP-A and ReQoL-UI more appropriate for interventions aimed at well-being and mental health.

Value sets beyond the UK

The original ReQoL-UI scores rest on UK preferences, and other countries have since produced their own sets. A country-specific value set can change both the range of utilities and the form of the scoring model.

Thai and colleagues estimated an Australian value set in 2025 using a discrete choice experiment with a duration attribute, completed online by 1,019 adults aged 18 and over. The worst state was valued at −0.585, well below the UK floor of −0.195. Unlike the UK model, the Australian set is a discrete choice model with a coefficient for each dimension, and physical health problems were the most important dimension and self-perception the least important. An analysis that applies the UK algorithm to an Australian trial, or the reverse, is therefore using another population's preferences, and the gap between the two worst-state values shows how far the elicitation method and the population can move the lower end of the scale.

Mapping symptom scores to the ReQoL-UI when the questionnaire was not collected

Many mental health datasets hold symptom scores but no utility measure. A mapping function can predict ReQoL-UI utilities from those scores, at the cost of extra uncertainty.

Franklin and Hernández Alava published mapping models in 2023 that predict ReQoL-UI or EQ-5D-5L utilities from the PHQ-9 and the GAD-7, using trial data from Improving Access to Psychological Therapies services in England, now NHS Talking Therapies. They estimated adjusted limited dependent variable mixture models and beta mixture models with the PHQ-9, GAD-7, age and sex as covariates. The best-fitting adjusted limited dependent variable mixture models had four components, and the authors concluded that the functions can predict utilities from variables routinely collected in mental health services or trials. Predicted utilities carry the error of the mapping model, so a probabilistic analysis should draw from that error as well as from sampling uncertainty in the trial.

NICE, NHS England and routine data

The ReQoL-UI matters to HTA bodies mainly as an alternative to the measure they already prefer. Section 4.3.6 of the NICE technology appraisal and highly specialised technologies manual (PMG36) makes the EQ-5D the preferred measure of health-related quality of life in adults, for consistency across evaluations. When the EQ-5D is not available or not appropriate, figure 4.1 of the manual lists another generic preference-based measure as the first alternative. The interim methods statement PMG51, published on 27 August 2026, names the EQ-5D-5L descriptive system and the UK EQ-5D-5L value set for adult reference-case analyses. The PMG36 overview says PMG51 takes precedence until the manual is updated, and sections 4.3.6 and 4.3.12, with figure 4.1, are flagged for update after PMG51. NICE has not adopted the ReQoL-UI as a reference-case measure, so in a NICE submission ReQoL-UI QALYs would normally sit beside the EQ-5D results as supporting evidence.

NHS England's commissioner guidance for adult community mental health rehabilitation services (published June 2025) names DIALOG, the ReQoL-10 and goal-based outcomes as the recommended suite for community mental health services. Because the ReQoL-UI items are contained in the ReQoL-10, routinely collected ReQoL-10 data can in principle be scored as utilities, which opens the way to QALY estimates from service data as well as from trials.

Limitations and common misreadings

Several errors recur when the ReQoL-UI is used or reported. Most come from treating the index as interchangeable with either the ReQoL total score or the EQ-5D.

• Total score read as a utility. A ReQoL-10 or ReQoL-20 total is a profile score on a recovery scale. Only the scoring algorithm turns the seven ReQoL-UI items into a health state utility value, and simple rescaling of the total does not reproduce it.
• All items assumed to count. The index uses only six of the mental health items in the ReQoL-10 or ReQoL-20, so the remaining mental health items do not enter the utility at all, while the physical health item does.
• Utilities pooled across measures. ReQoL-UI and EQ-5D-5L utilities come from different descriptive systems and value sets. The trial reported by Franklin and colleagues shows that the choice between them, together with the handling of baseline and missing data, can move the probability of cost-effectiveness across a wide range.
• One set treated as universal. The UK algorithm reflects UK general population preferences elicited by time trade-off, and the Australian set reflects Australian preferences elicited by a discrete choice experiment, with a very different worst-state value.
• Symptom change equated with utility change. The index values recovery-related quality of life, so a treatment that reduces symptoms without changing activity, hope or relationships may show little utility gain, which is a property of the measure and not evidence of no benefit.

Sources

• Franklin M, Hernández Alava M. Enabling QALY estimation in mental health trials and care settings: mapping from the PHQ-9 and GAD-7 to the ReQoL-UI or EQ-5D-5L using mixture models. Quality of Life Research. 2023;32(10):2763-2778. doi:10.1007/s11136-023-03443-9. https://pubmed.ncbi.nlm.nih.gov/37314661/
• Franklin M, Hunter RM, Enrique A, Palacios J, Richards D. Estimating cost-effectiveness using alternative preference-based scores and within-trial methods: exploring the dynamics of the quality-adjusted life-year using the EQ-5D 5-level version and Recovering Quality of Life Utility Index. Value in Health. 2022;25(6):1018-1029. doi:10.1016/j.jval.2021.11.1358
• Keetharuth AD, Brazier J, Connell J, et al. Recovering Quality of Life (ReQoL): a new generic self-reported outcome measure for use with people experiencing mental health difficulties. British Journal of Psychiatry. 2018;212(1):42-49. https://pmc.ncbi.nlm.nih.gov/articles/PMC6457165/
• Keetharuth AD, Rowen D, Bjorner JB, Brazier J. Estimating a preference-based index for mental health from the Recovering Quality of Life measure: valuation of Recovering Quality of Life Utility Index. Value in Health. 2021;24(2):281-290. doi:10.1016/j.jval.2020.10.012
• 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 4.3.6 and figure 4.1. https://www.nice.org.uk/process/pmg36/chapter/economic-evaluation-2
• National Institute for Health and Care Excellence. Interim methods statement: implementing the EQ-5D-5L value set (PMG51). Published 27 August 2026. https://www.nice.org.uk/process/pmg51/chapter/statement-on-using-the-eq-5d-5l-value-set
• NHS England. Commissioner guidance for adult community mental health rehabilitation services. Published 9 June 2025, updated 18 June 2025. https://www.england.nhs.uk/long-read/commissioner-guidance-adult-community-mental-health-rehabilitation-services/
• ReQoL (University of Sheffield). ReQoL-UI preference weights. Accessed 17 October 2026. https://www.reqol.org.uk/p/reqol-ui-preference-weights.html
• Thai T, Engel L, Ride J, Mulhern B, et al. Developing an Australian value set for the Recovering Quality of Life-Utility Index instrument using discrete choice experiment with duration. Value in Health. 2025;28(3):460-469. https://pubmed.ncbi.nlm.nih.gov/39827911
• Xu RH, et al. Measuring health-related quality of life and well-being: a head-to-head psychometric comparison of the EQ-5D-5L, ReQoL-UI and ICECAP-A. European Journal of Health Economics. 2022;23:165-176. doi:10.1007/s10198-021-01359-0