CHU9D

CHU9D in Paediatric Economic Evaluation

The CHU9D gives a paediatric cost-utility analysis a way to turn a child's own account of how they feel today into a utility weight, using questions whose content came from interviews with children instead of being adapted from an adult instrument. Health economists meet the instrument mainly in trials of interventions for school-age children and adolescents, and in mapping studies that predict its scores from paediatric questionnaires that have no preference weights. Its original UK weights came from 300 adults valuing health states with the standard gamble, while later Australian and Chinese algorithms used the choices of young people themselves, so the same set of answers can carry different utilities depending on the value set an analyst picks. The NICE methods manual names no preferred measure for children, which leaves the choice of instrument and value set open to scrutiny in each paediatric appraisal.

How children shaped the nine dimensions

Katherine Stevens developed the descriptive system at the University of Sheffield in stages. The first stage, reported in Qualitative Health Research, used 74 qualitative interviews with children to find out how health affects their lives, with purposive sampling balanced mainly for health within age and then for gender and ethnicity. The interviews produced eleven dimensions covering social, emotional and physical aspects of health-related quality of life. The abstract notes that these included feeling jealous and feeling tired or weak, and that parental, family and behavioural issues were left out.

The second stage tested the eleven-dimension system in 247 children drawn from general and clinical paediatric populations. Stevens reported good practicality and validity in both samples and removed two dimensions, jealous and embarrassed, so that the remaining nine could be used for health state valuation. The valuation paper that followed describes the result as a generic measure of health-related quality of life for children aged 7 to 11, the age group the development work centred on.

The way the content was chosen matters for economic evaluation. Feeling annoyed and coping with schoolwork appear because children raised them. Family functioning and behaviour are absent by design, so an intervention whose benefits run only through those routes will show little change in CHU9D utilities even when parents and teachers see a real improvement.

Questionnaire structure, recall period and proxy completion

The University of Sheffield, which holds the copyright and licenses the measure, describes the CHU9D as a paediatric generic preference-based measure made up of a descriptive system and a set of preference weights that give a utility value for each health state. Each of the nine questions has five response levels, and the recall period is today, with the sleep question referring to last night. Sheffield gives the age range as 7 to 17 years and the mode as self-completion by the child, with proxy completion available for younger children, and its licensing portal lists a free licence for non-commercial use.

Five levels on each of nine dimensions define 1,953,125 possible health states, far more than any valuation study can value directly. A value set therefore has to be a statistical model that predicts a utility for every combination from the subset of states that respondents actually valued. The answers a child gives form a profile such as 211311121, and the value set converts that profile into a single number on the scale where full health is 1 and dead is 0.

The UK standard gamble value set

The original preference weights come from Stevens's valuation study in PharmacoEconomics. Valuation interviews were held with 300 members of the UK adult general population, using both the standard gamble and a ranking exercise, and regression models were fitted to predict a value for each health state. Models estimated on the standard gamble data predicted better than the ranking model.

The preferred model was an ordinary least squares regression in which all coefficients were significant, with no inconsistencies and the best predictive performance, including a low mean absolute error. Stevens concluded that valuing the CHU9D was feasible, so that QALYs could be calculated for paediatric economic evaluation. These are adult preferences about states described by a child-centred questionnaire, and the choice of respondents is one of the points on which later value sets differ from the UK one.

Adolescent and other national value sets

Later studies asked different people, with different elicitation tasks, to value the same nine dimensions. Each produced a country-specific value set whose respondents and method an analyst needs to state, because utilities from different algorithms are not interchangeable.

Ratcliffe and colleagues asked Australian adolescents to pick the best and worst features of a series of 10 health states drawn from the CHU9D, a form of best-worst scaling. They estimated values for each level of the nine attributes and turned them into an adolescent-specific scoring algorithm on the full health equals 1 and dead equals 0 scale needed for QALYs. Because the adolescents appeared to use different decision processes for best and worst choices, the algorithm was derived from the best choices only.

Chen and colleagues built a Chinese tariff by combining best-worst scaling with the time trade-off. The best-worst survey recruited 902 students with a mean age of 13 through multi-stage random sampling of schools, and the time trade-off survey used a convenience sample of 38 undergraduates with a mean age of 18. Best and worst responses could not be pooled, so the analysis again used the best responses only, and the latent class estimates were rescaled to the QALY scale with the time trade-off values through a mapping approach.

Rowen and colleagues estimated a Dutch value set from an online discrete choice experiment with duration, completed by 1,276 adults from the Dutch general population. Conditional logit with robust standard errors gave a utility for every CHU9D state. The authors judged the approach feasible and raised normative questions about the perspective respondents should take when imagining the health of a child.

What the NICE manual says about measuring health in children

The NICE technology appraisal and highly specialised technologies manual (PMG36, published 31 January 2022 and last updated 13 October 2026) has separate paragraphs on measuring health-related quality of life in children and young people. Section 4.3.13 asks, for evaluations in which the population includes people aged under 18, that alternative measures of health-related quality of life for children be considered. Section 4.3.14 states that NICE does not recommend specific measures of health-related quality of life in children and young people, and that a generic measure shown to have good psychometric performance in the relevant age ranges should be used.

The same section acknowledges that not all paediatric health-related quality of life instruments have a UK value set and that there are methodological challenges when developing value sets for children and young people. Section 4.3.15 asks submissions to report whether measures were completed by adults with the condition, by children and young people themselves or on their behalf, for example by parents, carers or clinicians, and to report the age of the children and young people. The CHU9D is not named in these sections, so its use in a submission rests on evidence of its psychometric performance in the relevant ages and on the committee's view of its UK standard gamble value set, the kind of UK value set that section 4.3.14 notes not every paediatric instrument has.

Mapping other paediatric measures onto the CHU9D

Many paediatric trials collect a questionnaire that has no preference weights, so analysts estimate CHU9D utilities from it with a mapping function. Lambe and colleagues developed algorithms that predict CHU9D index scores from responses to the Pediatric Quality of Life Inventory (PedsQL) Generic Core Scales.

They used data from the PREDNOS randomised controlled trial of prednisolone therapy for childhood corticosteroid-sensitive nephrotic syndrome, in which health-related quality of life was collected at randomisation, week 16 and months 12, 18, 24, 36 and 48. A generalised linear model performed best. The NIHR report of the PREDNOS trial states that parents completed the proxy versions of the PedsQL and the CHU-9D, so the algorithm describes parent-reported health. A mapped utility inherits the value set used to score the CHU9D in the estimation data and the clinical population in which the algorithm was fitted, and both should be reported when the mapped values feed a cost-utility analysis.

Illustrative worked example of QALYs from a school-age trial

In a trial the CHU9D is completed at baseline and at each follow-up visit, each profile is scored with the chosen value set, and the resulting utilities are combined with time to give QALYs for each arm. The difference in mean QALYs between arms becomes the denominator of the incremental cost-effectiveness ratio. In a decision model the same utilities serve as health state utility values attached to childhood health states.

All numbers in this example are illustrative and are not drawn from any published CHU9D value set. A trial randomises children aged 8 to 11 to a new school-based intervention or to usual care, and the children complete the CHU9D at baseline, 6 months and 12 months. Mean utilities, scored with one value set, are 0.80, 0.86 and 0.88 in the intervention arm and 0.80, 0.82 and 0.83 in the usual care arm. The intervention adds GBP 650 per child, and costs and QALYs are left undiscounted over the single year.

$$Q = \sum_{k=1}^{K} \frac{u_{k-1} + u_{k}}{2} \, t_{k}$$

where $Q$ is the QALYs accrued over follow-up, $u_{0}$ is the mean utility at baseline, $u_{k}$ is the mean utility at follow-up assessment $k$, $t_{k}$ is the length in years of the interval that ends at assessment $k$ and $K$ is the number of follow-up assessments.

1. QALYs in the intervention arm. The mean of 0.80 and 0.86 is 0.83, and 0.83 × 0.5 = 0.415. The mean of 0.86 and 0.88 is 0.87, and 0.87 × 0.5 = 0.435. The total is 0.415 + 0.435 = 0.85 QALYs.

2. QALYs in the usual care arm. The mean of 0.80 and 0.82 is 0.81, and 0.81 × 0.5 = 0.405. The mean of 0.82 and 0.83 is 0.825, and 0.825 × 0.5 = 0.4125. The total is 0.405 + 0.4125 = 0.8175 QALYs.

3. Incremental QALYs and the ratio. The gain is 0.85 − 0.8175 = 0.0325 QALYs per child.

$$\text{ICER} = \frac{650}{0.0325} = 20000$$

where 650 is the incremental cost in GBP per child, 0.0325 is the incremental QALY gain per child and the result is in GBP per QALY.

4. The same answers scored with a second value set. Suppose an alternative value set gives smaller differences between the levels the children reported, so that the same profiles yield a gain of 0.025 QALYs. The ratio becomes 650 ÷ 0.025 = 26,000, so GBP 26,000 per QALY.

The children's answers did not change between steps 3 and 4, yet the ratio rose by 30 per cent because the weights did. Against an illustrative threshold of GBP 25,000 per QALY the decision would reverse, so a paediatric analysis should name its CHU9D value set and show the result under a credible alternative. If half the children at 12 months had been assessed by a parent instead of themselves, the change in respondent would sit inside the QALY gain and could not be separated from the effect of the intervention.

How the CHU9D differs from the EQ-5D family and other utility measures

The CHU9D is one example of a preference-based measure, the class of instruments whose scores carry preference weights so they can be used directly in QALYs. The value set is a separate object, the scoring model attached to the descriptive system, and one descriptive system can have several value sets, as the table above shows.

The EQ-5D is the generic measure NICE prefers for adults, with the EQ-5D-5L and a new UK value set named as the preferred approach for adults in the interim methods statement PMG51, and section 4.3.13 of PMG36 asks for alternatives to be considered when the population includes under-18s. Its youth version, the EQ-5D-Y, originally had three levels of severity in each dimension. Kreimeier and colleagues later developed a five-level descriptive system, the EQ-5D-Y-5L. The CHU9D differs in origin, since its dimensions were generated from children's own accounts, and in content, with separate items for worry, sadness, tiredness, annoyance, schoolwork and sleep. The Health Utilities Index is another generic preference-based measure used in children, and a paediatric analysis that switches between any of these instruments changes the scale on which QALYs are measured.

Limitations and common misreadings

Several errors recur when CHU9D results are read or reused. Most come from treating a CHU9D utility as a fixed property of the child, when the number also depends on who answered and whose preferences were used.

• Assuming one age range. The valuation paper describes a measure for children aged 7 to 11, while the developer now gives 7 to 17 years and offers proxy completion for younger children. A study should therefore report the ages of its participants and who completed the questionnaire, as PMG36 section 4.3.15 asks, and check that psychometric evidence exists for that age group and mode.
• Treating value sets as interchangeable. The UK set reflects adult standard gamble preferences, the Australian and Chinese sets the best choices of young people, and the Dutch set adult choices with duration. A comparison across studies that used different sets compares the weights as well as the health.
• Mixing respondents. A proxy report records a parent's view of the child's health. Pooling self-reports and proxy reports, or switching respondent during follow-up, builds the change into the QALY estimate.
• Expecting the CHU9D to capture everything. Family, parental and behavioural issues were left out by design, so effects on siblings, carers or classroom behaviour need other measures.
• Reusing a mapping algorithm outside its population. A PedsQL to CHU9D algorithm fitted to parent reports in a nephrotic syndrome trial may predict less well in another condition, age group or respondent type.

Sources

• Chen G, Xu F, Huynh E, Wang Z, Stevens K, Ratcliffe J. Scoring the Child Health Utility 9D instrument: estimation of a Chinese child and adolescent-specific tariff. Quality of Life Research. 2019;28(1):163-176. doi:10.1007/s11136-018-2032-z
• Kreimeier S, et al. EQ-5D-Y-5L: developing a revised EQ-5D-Y with increased response categories. Quality of Life Research. 2019;28(7):1951-1961. doi:10.1007/s11136-019-02115-x
• Lambe T, et al. Mapping the Paediatric Quality of Life Inventory (PedsQL) Generic Core Scales onto the Child Health Utility Index-9 Dimension (CHU-9D) score for economic evaluation in children. PharmacoEconomics. 2018;36(4):451-465. doi:10.1007/s40273-017-0600-7
• NIHR Journals Library. Economic analysis: the mapping exercise. In: Sixteen-week versus standard eight-week prednisolone therapy for childhood nephrotic syndrome: the PREDNOS RCT. NCBI Bookshelf NBK541714. https://www.ncbi.nlm.nih.gov/books/NBK541714/
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36), sections 4.3.13 to 4.3.15. Published 31 January 2022, last updated 13 October 2026. 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
• Ratcliffe J, et al. Valuing the Child Health Utility 9D: using profile case best worst scaling methods to develop a new adolescent specific scoring algorithm. Social Science and Medicine. 2016;157:48-59. doi:10.1016/j.socscimed.2016.03.042
• Rowen D, Mulhern B, Stevens K, Vermaire JH. Estimating a Dutch value set for the pediatric preference-based CHU9D using a discrete choice experiment with duration. Value in Health. 2018;21(10):1234-1242. doi:10.1016/j.jval.2018.03.016
• Stevens K. Working with children to develop dimensions for a preference-based, generic, pediatric, health-related quality-of-life measure. Qualitative Health Research. 2010;20(3):340-351. doi:10.1177/1049732309358328
• Stevens K. Assessing the performance of a new generic measure of health-related quality of life for children and refining it for use in health state valuation. Applied Health Economics and Health Policy. 2011;9(3):157-169. doi:10.2165/11587350-000000000-00000
• Stevens K. Valuation of the Child Health Utility 9D Index. PharmacoEconomics. 2012;30(8):729-747. doi:10.2165/11599120-000000000-00000
• University of Sheffield. CHU9D, measuring health and calculating QALYs for children and adolescents (licensing portal and Measuring and Valuing Health pages), accessed 17 October 2026. https://licensing.sheffield.ac.uk/product/CHU-9D and https://www.sheffield.ac.uk/scharr/sections/heds/mvh/paediatric/about-chu9d