ASCOT

ASCOT in Economic Evaluation of Adult Social Care

Economic evaluations of home care, care homes and support for unpaid carers run into a measurement problem, because a health-related quality of life measure can fail to register what a service changes when it makes a person's daily life cleaner or more under their own control. In an NIHR Health Technology Assessment project that built on earlier ASCOT work, Netten and colleagues developed the preference-weighted measure, with the stated aim of producing a measure of social care outcome that would be the equivalent of the QALY in health. The service-user instrument rates eight domains at four levels each, and preference weights anchored by a time trade-off place the overall score on a scale where 1 is the ideal state and 0 is equivalent to being dead. The interview version also asks people what their situation would be without services, so the gain attributed to care can be estimated without a control group. ASCOT here means the Adult Social Care Outcomes Toolkit developed in England, not the cardiovascular trial that shares the acronym.

Eight domains of social care-related quality of life

ASCOT measures social care-related quality of life (SCRQoL), the parts of quality of life that social care services are designed to affect. The construct is different from health-related quality of life, since a person with a stable long-term condition can gain a great deal from support that leaves health unchanged.

The final measure reported by Netten and colleagues has eight domains. They are personal cleanliness and comfort, accommodation cleanliness and comfort, food and drink, personal safety, social participation and involvement, occupation, control over daily life, and dignity. Rand and colleagues describe each domain as captured by a single item with four response levels.

The domains differ in how close they sit to basic needs, and that difference matters in practice. In a study of 366 residents of 83 English care homes for older people inspected in 2008, Netten and colleagues found that homes achieved meaningful gains in SCRQoL but did better on the basic domains than on higher-order ones such as social participation.

Response levels and the preference-weighted score

Each domain is answered at one of four outcome states. In order, these are the ideal state, in which needs and wishes are met to the level the person wants, then no needs, then some needs, and finally high-level needs, which carry immediate or longer-term implications for health. The ASCOT guidance from the University of Kent describes the response options as ordered with the ideal state at the top and high-level needs at the bottom.

The preference weights came from general population studies. Netten and colleagues first explored a discrete choice experiment and best-worst scaling with 300 members of the general population. In the main phase they carried out best-worst scaling interviews with 500 members of the general population, time trade-off interviews with a follow-up sample of 126 of those respondents, and best-worst scaling interviews with 458 people using equipment services. Best-worst scaling gives the relative value of each level in each domain, and the time trade-off rescales those values so that the overall score can be read on the same anchors as a QALY weight.

The scoring rule published in the ASCOT guidance for the English preference weights of the self-completion version (SCT4) is a linear transformation of the summed level weights.

$$S = 0.203 \times W-0.466$$

where $S$ is the SCRQoL score and $W$ is the sum of the preference weights attached to the level chosen in each of the eight domains. The guidance states that the formula rests on a time trade-off exercise with members of the public, with 0.00 anchored to being dead and 1.00 to the ideal state. Scores run between minus 0.17 and 1.00 and are rounded to two decimal places, and scores below zero describe states the general population judged worse than being dead.

The guidance gives an example in which the domain weights sum to 6.027. The arithmetic is 0.203 × 6.027 = 1.223481, and 1.223481 − 0.466 = 0.757481, which rounds to a score of 0.76.

Current, expected and the gain attributed to services

The interview version (INT4) produces three scores. Current SCRQoL is the person's quality of life with services in place. Expected SCRQoL is what the person estimates their SCRQoL would be without the care and support received, and SCRQoL gain is the difference between the two. Netten and colleagues describe the difference between current and expected SCRQoL as an indicator of service impact.

The expected score is a self-estimated counterfactual. Netten and colleagues argued that a before measure is a poor proxy for the absence of intervention, since functioning tends to decline over time, and that service users are well placed to judge what their SCRQoL would be without services. In the interview the expected question for a domain is asked only when the respondent says services affect that domain, and Sjetne and colleagues note that the counterfactual is not applied to dignity, which is scored at the no unmet needs level.

$$G = S_{c}-S_{e}$$

where $G$ is the SCRQoL gain attributed to services, $S_{c}$ is the current SCRQoL score and $S_{e}$ is the expected SCRQoL score, both on the preference-weighted scale.

The design answers a practical problem. Long-term care is rarely withdrawn at random, so a trial-style comparison with no service is seldom available, and the expected question stands in for a control group. The price is that the counterfactual is hypothetical and rests on the respondent's judgement, which is why the gain is better treated as an estimate of the effect of a service than as a measured outcome.

Versions for service users, carers and people who cannot self-report

ASCOT is a family of instruments built around the same SCRQoL construct. The choice of version changes who answers and which scores are available.

Rand and colleagues tested the SCT4 with adults with long-term physical, sensory and mental health conditions in England. They found moderate to strong correlations, all at least 0.3, between the SCT4 index and the EQ-5D-3L, the ICECAP-A or ICECAP-O, and overall quality of life, and good test-retest reliability for adults with physical or sensory disabilities using social care.

The ASCOT-Proxy was developed for people who might otherwise be left out of evaluations because of cognitive or communication difficulties. Rand and colleagues built it through cognitive interviews with unpaid carers and care workers of people with such impairments. For care homes, Towers and colleagues found the mixed-methods CH4 to be a weak unidimensional scale with acceptable internal consistency, consistent with other ASCOT measures.

The ASCOT-Carer applies the same idea to the person giving unpaid care, the subject of the carer page. Rand and colleagues tested its interview version with 387 unpaid carers of adults using publicly funded social care in England. The ASCOT-Carer has seven SCRQoL domains, each with four response levels, giving a raw unweighted total of 0 to 21, and showed satisfactory internal reliability with a Cronbach's alpha of 0.87 across the seven items. Carer scores describe the carer's own SCRQoL and sit alongside the measures covered on the caregiver quality of life page.

Preference weights are country-specific, as with any value set. Nguyen and colleagues estimated Finnish population-based weights for the service-user ASCOT, and the English weights should not be assumed to hold elsewhere.

Illustrative worked example of an ASCOT gain from a home care package

All figures in this example are illustrative. A local authority funds a home care package for an older person for one year at a net cost of GBP 6,150 compared with no package. The person completes the INT4.

1. Current SCRQoL. The preference weights for the levels the person reports with the package in place sum to 6.027. The score is 0.203 × 6.027 = 1.223481, then 1.223481 − 0.466 = 0.757481, which rounds to 0.76.

2. Expected SCRQoL. The weights for the levels the person expects without the package sum to 4.000. The score is 0.203 × 4.000 = 0.812, then 0.812 − 0.466 = 0.346, which rounds to 0.35.

3. Gain attributed to the package. The SCRQoL gain is 0.76 − 0.35 = 0.41.

4. Cost per unit of gain. If the analyst assumes the gain holds for the whole year, the package yields an illustrative 0.41 SCRQoL-adjusted years, and the cost per unit is

$$\frac{6150}{0.41} = 15000$$

where 6150 is the incremental cost in GBP and 0.41 is the SCRQoL gain over one year, giving GBP 15,000 per SCRQoL-adjusted year.

The result says the package buys a large improvement in social care-related quality of life at a modest cost per unit. The ratio does not show that the package is good value against an NHS threshold, because the denominator is not a health QALY and a cost-effectiveness threshold set for health QALYs cannot be borrowed for it. The constant-gain assumption in step 4 is the analyst's choice, and in a real evaluation the gain would be measured at several points and the area between the current and expected paths used instead.

ASCOT in economic evaluation of social care

The NICE guideline manual, Developing NICE guidelines (PMG20), sets a reference case in chapter 7 that depends on the decision problem. Its reference case table lists capability or social care-related quality of life measures, such as ICECAP or ASCOT, for interventions that result in both health and capability or social care outcomes. The same chapter says NICE will consider using 'social care QALYs' if validated, and names ASCOT in that discussion. Netten and colleagues drew on previous and parallel work developing the toolkit and the national Adult Social Care Survey in England, and tested the instrument with 30 service users from a variety of user groups and 300 older home care service users.

ASCOT can be the outcome in a cost-effectiveness ratio for a social care intervention, as in the worked example, or sit beside the EQ-5D in a cost-consequence analysis so that a decision maker sees health and social care outcomes together. Used with the ASCOT-Carer, the service-user ASCOT also lets an evaluation report outcomes for the person receiving care and for an unpaid carer.

The perspective of the analysis decides how far ASCOT adds information. An evaluation that counts only health effects will miss what ASCOT records, while one that takes in social care and carers can use ASCOT for gains that a health measure does not register.

How ASCOT differs from EQ-5D, ICECAP-O and carer measures

ASCOT, the EQ-5D and the ICECAP measures are all preference-based, but each measures a different construct. Mixing their scores without thought produces numbers that look comparable and are not.

• EQ-5D. The EQ-5D describes health-related quality of life in health dimensions. ASCOT describes SCRQoL in domains such as control over daily life and food and drink, which can change while health stays the same. The correlation between the two reported by Rand and colleagues shows overlap, not equivalence.
• ICECAP-O. The ICECAP-O measures capability wellbeing in older people. ASCOT focuses on the parts of quality of life that social care is meant to affect and offers an expected-state question, which ICECAP-O does not ask.
• Carer instruments. The ASCOT-Carer measures the carer's own SCRQoL. A proxy ASCOT completed by a carer describes the person cared for, and the two should never be pooled.
• Social care. Social care is the set of services under evaluation, while ASCOT measures what those services achieve for the people who use them. The NICE guideline manual places ASCOT beside the EQ-5D for interventions with social care outcomes, not in place of it.

Limitations and common misreadings

ASCOT has a clear purpose, but several features limit what its scores can carry in an evaluation. Several errors come from treating ASCOT as a drop-in replacement for a health utility.

• Treating SCRQoL as a health QALY weight. The 0 and 1 anchors match a QALY scale, but the descriptive system is social care-related, so ASCOT gains cannot simply be added to EQ-5D QALYs without double counting effects that both measures pick up.
• Reading the gain as observed. The expected score is a self-estimate of a state that cannot be observed, so the gain is an estimate of service impact and not a measured change.
• Ignoring dimensionality. Towers and colleagues describe ASCOT measures as weak unidimensional scales. As with any summary index, one overall score can hide a large change in a single domain, so domain-level results add information.
• Carrying English weights abroad. Countries such as Finland have their own weights, and a value set should match the population whose preferences the decision should reflect.
• Mixing versions. Self-completion, interview, proxy and Easy Read versions differ in who answers and in the scores available, and results from them are not automatically interchangeable.
• Licensing. Use of ASCOT instruments requires a licence from the University of Kent, which affects trial planning.

Sources

• Netten A, Burge P, Malley J, Potoglou D, Towers AM, Brazier J, Flynn T, Forder J, Wall B. Outcomes of social care for adults: developing a preference-weighted measure. Health Technology Assessment. 2012;16(16):1-166. doi:10.3310/hta16160
• National Institute for Health and Care Excellence. Developing NICE guidelines: the manual (PMG20), chapter 7 Incorporating economic evaluation. First published 2014. https://www.nice.org.uk/process/pmg20/chapter/incorporating-economic-evaluation
• Netten A, Trukeschitz B, Beadle-Brown J, et al. Quality of life outcomes for residents and quality ratings of care homes: is there a relationship? Age and Ageing. 2012;41(4):512-517. doi:10.1093/ageing/afs050
• Nguyen L, Jokimäki H, Linnosmaa I, Saloniki EC, Batchelder L, Malley J, Lu H, Burge P, Trukeschitz B, Forder J. Do you prefer safety to social participation? Finnish population-based preference weights for the Adult Social Care Outcomes Toolkit (ASCOT) for service users. MDM Policy and Practice. 2021;6(2). doi:10.1177/23814683211027902
• Rand SE, Malley JN, Netten AP, Forder JE. Factor structure and construct validity of the Adult Social Care Outcomes Toolkit for Carers (ASCOT-Carer). Quality of Life Research. 2015;24(11):2601-2614. doi:10.1007/s11136-015-1011-x
• Rand S, Caiels J, Collins G, Forder J. Developing a proxy version of the Adult social care outcome toolkit (ASCOT). Health and Quality of Life Outcomes. 2017. doi:10.1186/s12955-017-0682-0
• Rand S, Malley J, Towers AM, Netten A, Forder J. Validity and test-retest reliability of the self-completion adult social care outcomes toolkit (ASCOT-SCT4) with adults with long-term physical, sensory and mental health conditions in England. Health and Quality of Life Outcomes. 2017. doi:10.1186/s12955-017-0739-0
• Sjetne IS, Smith N, Towers AM, Renolen Å, et al. Norwegian translation and adaptation of the INT4 version of the Adult Social Care Outcomes Toolkit (ASCOT). Quality of Life Research. 2026. doi:10.1007/s11136-026-04209-9
• Towers AM, Rand S, Collins G, Smith N, Palmer S, Cassell J. Measuring quality of life in care homes when self-report is challenging: the construct validity, structural characteristics and internal consistency of the mixed-methods adult social care outcomes toolkit. Age and Ageing. 2023;52(9):afad168. https://academic.oup.com/ageing/article/52/9/afad168/7258776
• University of Kent. Adult Social Care Outcomes Toolkit (ASCOT) SCT4 Easy Read guidance, version 2.1. https://research.kent.ac.uk/ascot/wp-content/uploads/sites/2976/2026/07/ASCOT-ER_Guidance-v2.1.pdf