KIDSCREEN-10

KIDSCREEN-10 in Child and Adolescent Outcome Measurement

A single number summarising how a child felt over the last week is attractive to trialists and awkward for health economists, and the KIDSCREEN-10 shows why. Its score places a child relative to other European children, with a mean of about 50 and a standard deviation of about 10, and is not a utility anchored at full health and dead. Health economists therefore meet it in school, public health and population studies that collected a short wellbeing outcome, and in mapping studies that predict utilities from it. Chen and colleagues built such an algorithm for the CHU9D from the answers of 590 Australian adolescents aged 11 to 17, so a KIDSCREEN-10 trial can reach a cost-utility analysis only through a prediction step with its own error. KIDSCREEN is a family of questionnaires with 52, 27 and 10 items, and the topic here is the ten-item index.

Origins in the European KIDSCREEN project

The European KIDSCREEN project developed self-report and proxy versions with 52, 27 and 10 items simultaneously in 13 countries. A 2014 overview by Ravens-Sieberer and colleagues reports that the KIDSCREEN had by then been found reliable, valid and appropriate as a quality of life measure in 38 countries or languages. The 2010 validation paper describes the ten-item version as a short measure of wellbeing and health-related quality of life in children and adolescents.

Ravens-Sieberer and colleagues tested the child self-report and parent report versions of the KIDSCREEN-10 in 22,830 European children and adolescents aged 8 to 18 and 16,237 parents. They concluded in 2010 that the index provides a valid measure of a general health-related quality of life factor, but that it does not represent well most of the single dimensions of the original KIDSCREEN-52. An analyst who needs separate information on, for example, physical wellbeing or relations with parents therefore needs the longer versions.

The KIDSCREEN Group holds the copyright to the KIDSCREEN questionnaires and their translations, including the KIDSCREEN-10 Index in child, adolescent and parent or proxy forms. Its terms of use state that the instruments are now fully open access, with no collaboration form required, while distribution or transfer of the questionnaires to third parties, such as by sale, rental or licensing, is not permitted.

Ten questions about the last week

Each of the ten items asks the child to think about the last week and answer on a five-point response scale. In the list given by Ravens-Sieberer and colleagues, the first four items ask whether the child has felt fit and well, full of energy, sad and lonely. The parent version asks the same questions about the child.

The answers are not reported item by item. Only the ten-item version yields a total index score formed from all its items, and the KIDSCREEN Group's scoring guidance allows at most one missing answer when the T-value is computed. Answers to negatively worded items, such as feeling sad, are recoded before summing, and a higher index means better health-related quality of life.

Rasch scoring and the European norm data

The scoring rests on a Rasch model, a measurement model in which the raw sum of answers carries all the information about the respondent's position on a single latent scale. The KIDSCREEN Group describes the steps as converting each possible sum score to a Rasch person parameter and then transforming the person parameters into values with a mean of about 50 and a standard deviation of about 10. The European norm data behind that scale come from the international KIDSCREEN sample, and national norm tables stratified by age and gender sit in the manual appendices.

$$T = 50 + 10 \, \frac{\theta-\bar{\theta}}{s_{\theta}}$$

where $T$ is the KIDSCREEN-10 T-value, $\theta$ is the Rasch person parameter that corresponds to the child's sum score, $\bar{\theta}$ is the mean person parameter in the European norm sample and $s_{\theta}$ is its standard deviation.

The transformation fixes the meaning of a score. Scored against the European norms, a T-value of 40 places a child one norm-sample standard deviation below the European mean, whatever the country in which the questionnaire was completed. It says nothing about how much that child's health is worth relative to full health, which is the information a QALY needs.

Reliability, validity and screening cut points

The 2010 validation reported a Cronbach alpha of 0.82 for the self-report version and 0.78 for the parent version, with test-retest reliability over two weeks giving an intraclass correlation of 0.70 for self-report and 0.67 for proxy report. The correlation between child and parent scores was 0.54. Correlations between the KIDSCREEN-10 and the scales of the KIDSCREEN-52 ranged from 0.24 to 0.72 for the self-report version and 0.27 to 0.72 for the parent version.

Later work has examined how well the Rasch assumptions hold. Müller and Haenni Hoti fitted Rasch models to 2017 data from a study of primary school children and their teachers, and found that the standard Rasch model did not fit while a graphical loglinear Rasch model did. Four pairs of items were locally dependent, and items behaved differently by gender and by citizenship. The Cronbach alpha of 0.79 in their data overestimated the true internal consistency because of the local dependence, and the items on feeling fit and well and on feeling sad did not fit the model.

Hirschfeld, von Brachel and Thiele derived screening cut points from the German KiGGS study, covering 2,566 children, 2,136 younger adolescents and 2,740 older adolescents. Classification of mental health problems was acceptable in every version, with areas under the curve of 0.77 to 0.79, and the optimal cut points were 41.13 for the parent version, 42.52 for younger adolescents and 40.29 for older adolescents. Only the parent version worked for general health, and no version worked for chronic disease.

Predicting CHU9D utilities from KIDSCREEN-10 answers

Because the index carries no preference weights, a cost-utility analysis with only KIDSCREEN-10 data needs a mapping function to estimate utilities. Chen, Stevens, Rowen and Ratcliffe developed an algorithm of this kind, predicting CHU9D utilities from KIDSCREEN-10 item scores.

Their 590 Australian adolescents completed both questionnaires. The authors compared ordinary least squares, censored least absolute deviations, a robust MM-estimator and a generalised linear model, judging accuracy by mean absolute error and root mean squared error. The MM-estimator with stepwise-selected KIDSCREEN-10 items predicted best on mean absolute error, the equivalent ordinary least squares model predicted best on root mean squared error, and the authors chose the MM-estimator as accurate enough for cost-utility analyses in which only KIDSCREEN-10 data exist.

Chen and colleagues scored the CHU9D with the Australian adolescent-specific algorithm, not the UK adult value set, so the mapped utilities carry adolescent preferences as well as the age range of the sample and the fact that the adolescents reported for themselves. The NICE technology appraisal and highly specialised technologies manual (PMG36, last updated 13 October 2026) states in section 4.3.14 that NICE does not recommend specific measures of health-related quality of life in children and young people and that a generic measure with good psychometric performance in the relevant age ranges should be used. For the KIDSCREEN-10 that makes the published psychometric evidence by age, and whether the child or the parent version was used, part of the case for using the index.

Illustrative worked example of a school programme evaluation

All numbers in this example are illustrative and do not come from any published study or algorithm. A cluster-randomised trial of a school wellbeing programme for pupils aged 12 to 14 records the self-report KIDSCREEN-10 at 12 months. The mean T-value is 52.4 in the programme schools and 49.4 in the control schools, the groups were balanced at baseline, and the programme costs GBP 240 per pupil. Costs and outcomes are left undiscounted.

1. Express the difference in norm-sample units. The difference is 52.4 − 49.4 = 3.0 points, and 3.0 ÷ 10 = 0.3, so the programme moves the average pupil by 0.3 of a European norm-sample standard deviation.

2. Predict a utility difference. Suppose a mapping algorithm implies that one T-value point corresponds to 0.004 on the CHU9D utility scale. Then 3.0 × 0.004 = 0.012, a utility gain of 0.012.

3. Convert the gain into QALYs. If the gain lasts the whole year, the QALY gain is 0.012 × 1 = 0.012. The ratio is 240 ÷ 0.012 = 20,000, so the ICER is GBP 20,000 per QALY.

4. Change the assumption about duration. If the gain lasts only six months, 0.012 × 0.5 = 0.006 QALYs and 240 ÷ 0.006 = 40,000, so GBP 40,000 per QALY.

A difference of 0.3 standard deviations looks respectable on the T-value scale, yet whether the programme appears cost-effective turns on the mapping slope and on how long the effect lasts, neither of which the KIDSCREEN-10 itself supplies. A sensitivity analysis over both, and a comparison with a directly collected preference-based measure where one exists, keeps the result from resting on one assumed conversion.

How the KIDSCREEN-10 differs from neighbouring paediatric measures

The KIDSCREEN-10 is a generic patient-reported outcome index without preference weights. That separates it from preference-based measures such as the CHU9D, whose value sets give utilities for QALYs, and its single index separates it from profile measures that report several scale scores.

The PedsQL 4.0 Generic Core Scales are the nearest comparison. The PedsQL reports four scale scores and three summary scores on a 0 to 100 metric, with a past-month recall period, separate forms for ages 2 to 18 and parent proxy-report for the youngest children. The KIDSCREEN-10 gives one Rasch-scaled T-value for ages 8 to 18 about the last week. Within its own family, the KIDSCREEN-27 and KIDSCREEN-52 report separate dimensions, and the 2010 validation found that the ten-item index does not reproduce most of them.

Limitations and common misreadings

Most problems arise when a KIDSCREEN-10 score is asked to do work it was not designed for. The index summarises general wellbeing well and carries little information about specific domains or about value.

• Reading a T-value as a utility. A T-value of 50 is the European average, not a utility of 0.50, and the scale has no anchor for full health or dead.
• Expecting sensitivity to physical illness. The KiGGS cut-point study found no version that classified chronic disease adequately, so the index should not be read as a screen for chronic physical illness.
• Assuming a perfect Rasch fit. Local dependence and differential item functioning by gender and citizenship mean that reliability can be overstated and that some group comparisons need checking at item level.
• Mixing respondents. Child and parent scores correlated at 0.54 in the validation sample, so a trial should keep self-report and parent report apart and not switch between them during follow-up.
• Reusing a mapping algorithm out of context. The CHU9D algorithm came from Australian adolescents aged 11 to 17 and predicts utilities on the Australian adolescent scoring algorithm, so prediction error is likely to grow for younger children, parent reports or other countries, and the utilities are not on the UK value set.

Sources

• Chen G, Stevens K, Rowen D, Ratcliffe J. From KIDSCREEN-10 to CHU9D: creating a unique mapping algorithm for application in economic evaluation. Health and Quality of Life Outcomes. 2014;12:134. doi:10.1186/s12955-014-0134-z
• Hirschfeld G, von Brachel R, Thiele C. Screening for health-related quality of life in children and adolescents: optimal cut points for the KIDSCREEN-10 for epidemiological studies. Quality of Life Research. 2020;29:529-536. doi:10.1007/s11136-019-02324-4
• KIDSCREEN Group. Evaluation by hand, and Terms of use and copyright, kidscreen.org, accessed 17 October 2026. https://www.kidscreen.org/english/analysis/evaluation-by-hand/ and https://www.kidscreen.org/english/terms-of-use/
• Müller M, Haenni Hoti A. Item analysis of the KIDSCREEN-10 using Rasch modelling. Health and Quality of Life Outcomes. 2020;18. doi:10.1186/s12955-020-01596-6
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36), section 4.3.14. Published 31 January 2022, last updated 13 October 2026. https://www.nice.org.uk/process/pmg36/chapter/economic-evaluation-2
• Ravens-Sieberer U, Erhart M, Rajmil L, et al. Reliability, construct and criterion validity of the KIDSCREEN-10 score: a short measure for children and adolescents' well-being and health-related quality of life. Quality of Life Research. 2010;19(10):1487-1500. doi:10.1007/s11136-010-9706-5
• Ravens-Sieberer U, Herdman M, Devine J, et al. The European KIDSCREEN approach to measure quality of life and well-being in children: development, current application, and future advances. Quality of Life Research. 2014;23(3):791-803. doi:10.1007/s11136-013-0428-3