Dementia is highly prevalent in today’s society. It was estimated that worldwide 50 million people lived with dementia in 2018, and with the ageing population this number is expected to increase to 82 million by 2030 and further to 152 million by 2050 [1
]. With these increasing numbers comes an increase in the magnitude of care required. Family members, who are most often elderly spouses, siblings or friends, are frequently the ones to provide (part of) this care [2
]. These informal carers may be unprepared for the physical and emotional demands that caring entails and many carers experience considerable strain and well-being losses due to their caregiving tasks [3
At the same time, governments struggle with the rising demand for care of people with dementia, those with other (chronic) diseases, and their carers, and with limited health care resources. Regarding decisions for optimal spending, economic evaluation is a useful decision-making aid. It is increasingly applied in many Western healthcare systems. Usually in economic evaluations, health-related quality of life (HrQoL) is used to measure intervention benefits. However, in areas such as mental health and care of older people, improving health outcomes is not necessarily the main focus of care interventions [5
]. Broader outcome measures, which go ‘beyond health’ [7
], capturing the effects of interventions in terms of well-being, for both patients and carers, may be more appropriate to capture relevant benefits of such interventions.
Several well-being measures have recently been developed, broadening the evaluative scope beyond health in different ways. Some of these have focused on context-specific elements of well-being beyond health, such as care-related quality of life in carers [8
] or social care-use-related well-being [10
]. Analogous with measuring disease-specific quality of life rather than generic health-related quality of life, context-specific measures may be more sensitive to specific changes, but this comes at the expense of comparability across situations and populations. Generic measures of well-being in principle allow comparisons, across interventions, diseases and situations. The ICECAP-O, a capability well-being measure for older people, is such a generic well-being measure [11
], which is increasingly used. Capability in this context refers to the extent to which a person is able to function in a particular way, whether or not he or she chooses to do so [11
]. The ICECAP-O consists of five important general capability well-being dimensions: attachment, security, role, enjoyment and control, and has four answering levels per domain. Values for the states described with the instrument were derived from a sample of older people in England, using best–worst scaling [12
]. ICECAP-O questionnaires and further information can be found on the website: www.birmingham.ac.uk/research/activity/mds/projects/HaPS/HE/ICECAP/ICECAP-O
As a generic well-being measure, the ICECAP-O is well suited for economic evaluation of care interventions in elderly populations, especially those who are suffering from chronic diseases [13
]. Given the generic nature of the instrument, it may be suitable to measure well-being not only in patients but also in informal carers. Using the ICECAP-O to measure well-being in both patients and their informal carers would facilitate comparisons and aggregation of outcomes within the same economic evaluation. This is also relevant for a scenario in which two interventions are being compared, and one of these interventions impacts informal carers. Without a comparable outcome between patients and informal carers, for example if the ICECAP-O and the CarerQoL were used, respectively, it would be necessary to perform multiple-criteria decision analysis, which may be more costly and time consuming.
Before the ICECAP-O can be used as a well-being measure in economic evaluations of care interventions, it needs to be validated in relevant populations. Validation is performed to assess the extent to which a measure evaluates what it sets out to represent. The most frequently used validity tests for health status measures in the literature are construct validity tests. These examine the extent to which the measure indeed captures the concept it intends to measure [14
]. Construct validity consists of both convergent and discriminant validity. Convergent validity refers to the extent to which a measure correlates with related concepts [14
], while discriminant validity refers to the extent to which relevant differences in (sub-) groups are adequately reflected by the measure [9
So far, the ICECAP-O has been validated for various populations such as older people in England [12
], psycho-geriatric older people in nursing homes in the Netherlands [19
], post-hospitalized older people in the Netherlands [20
] and older people with dementia in Germany [5
], mostly with favourable results. In these studies, sample sizes typically have been relatively small and taken from only one country. To date, to our knowledge, no study has validated the ICECAP-O as a well-being measure in informal carers.
In this study, we therefore add to the literature in a number of ways. This is the first study to validate the ICECAP-O in a sample of informal carers, using a rich dataset. We use data from a relatively large sample of carers obtained in eight European countries in the context of the Actifcare project [21
], which aims to analyse the pathways to care for people with dementia and their families. This paper therefore considers the construct validity of the ICECAP-O in a sizeable international population of informal carers for people with dementia. Furthermore, validating the ICECAP-O in this kind of population allows those performing economic evaluations to consider the ICECAP-O when measuring carer well-being.
Data were collected in eight European countries: Germany, Ireland, Italy, the Netherlands, Norway, Portugal, Sweden and the UK. Care receivers adhering to the specified criteria ("Appendix 1
") and their informal carers were invited to complete the questionnaires, available in seven different languages. For all measures, including the ICECAP-O, nationally validated versions were used, or, if not available, the measure was translated, back translated and pilot tested following a translation protocol [21
]. The data collection consisted of different parts. People with dementia and carers were interviewed by trained interviewers about their socio-demographic characteristics and the comorbidities and health care resource use of the former. Carers completed questionnaires covering a variety of outcome measures and were interviewed about the caregiving situation, their resource use, and the person with dementia’s health. Finally, the interviewer completed questionnaires about the health, quality of life and care needs of the person with dementia [21
Demographic characteristics included age, gender, nationality, ethnicity, marital status, level of education and state of employment. Before listing the health and well-being measures used in our analyses, it is worth describing the ICECAP-O in further detail. As mentioned above, the ICECAP-O is a general capability well-being measure, consisting of five dimensions: attachment, security, role, enjoyment and control. These dimensions are sometimes described in a little more detail as ‘love and friendship’, ‘thinking about the future’, ‘doing things that make you feel valued’, ‘enjoyment and pleasure’ and ‘independence’. The ICECAP-O and has four answering levels per domain which are no capability, a little capability, a lot of capability and full capability. The ICECAP-O provides us with separate scores for each domain, meaning there are 1024 different possible ‘capability states’. By attaching the designated utilities to each attribute, we are provided with a final ICECAP-O tariff score with a range between zero (no capability) and one (full capability) [15
There are health and well-being measures used in our analysis to test the validity of the ICECAP-O. The first measures are those answered by the carer, about the carer and their environment. These are CarerQol [9
], EQ-5D-5L [24
], Positive Affect Index (PAI) [25
], Perseverance Time (PT) [26
] and the Lubben Social Network Scale (LSNS) [27
]. The next measures are those answered by the carer and/or an interviewer as a proxy about the person with dementia. These are Clinical Dementia Rating (CDR) [28
], DemQoL-U [29
] (proxy-rated), Quality of Life in Alzheimer’s disease (QoL-AD) [31
], Resource Utilization in Dementia (RUD) [32
], and finally a subset of questions regarding unmet need from the Camberwell Assessment of Need for the Elderly (CANE) [33
]. These measures are discussed and referenced in Table 1
. Here it is important to note that when referring to ‘CANE Unmet Need’ we are referring to a measure taken from the CANE instrument, which in this case was collected by an interviewer talking with the carer and person with dementia. In the measure, we sum the number of times ‘unmet need’ is chosen out of the 24 questions asked. Summary statistics of all continuous variables used are shown in "Appendix 2"
To measure health-related quality of life [24
Mobility, self-care, usual activities, pain/discomfort, anxiety/depression. [24
] Each dimension has five response categories ranging from ‘no problems’ to ‘severe problems’ [43
Can be converted to a single utility score with range − 0.281 to 1, with 0 equal to being dead and 1 equal to perfect health. Can also be converted to a single health problems index by summing the scores on the five levels, with range of 0 to 20, with 0 being no health problems [44
To measure health-related quality of life.
One visual analogue scale which patients use to place their ‘self-rated health’ [45
Range of 0 to 100 where endpoints are labelled ‘worst imaginable health state’ (0) and ‘best imaginable health state’ (100) [45
To measure and value the impact of providing informal care on carers [9
Relational problems, mental health problems, problems combining daily activities with care, financial problems, physical health problems, fulfillment from caregiving, support with lending care [9
Weighted sum score with range 0 to 100, where 100 is the best caregiving situation [46
To measure the self-reported happiness of carers [9
One visual analogue scale which respondents use to rate their ‘self-rated happiness’ [9
Range of 0 to 10 where 0 is ‘completely unhappy’ and 10 is ‘completely happy’ [9
To measure relationship quality [25
Five items that reflect the respondent’s feelings towards their relative/friend: communication, understanding, trust, fairness, respect and affection [25
A range of 6 to 30, with 30 being the best score [25
To stage the progression of dementia [27
Uses information provided by patient and carer [27
Can be one of five stages: (0) no dementia, (0.5) very mild, (1) mild, (2), severe, (3) severe [47
To evaluate health-related quality of life in mild to moderate dementia [28
Information taken from questionnaire responses [28
A range of 0 to 1, with 1 being full health [28
To describe the relationship of QoL to demographic characteristics, cognitive and functional status, depression and pleasant level activity [30
13-item self- and proxy-report [30
A range of 13 to 52, with 52 being the best outcome [30
CANE Unmet Need
To identify needs for old age people with a mental illness or cognitive problems [33
24 domains (bio-psycho-social needs), and 2 items referring to carer needs [33
]. Original with responses being (0) no need, (1) met need, (2) unmet need
We sum the number of times ‘unmet need’ is chosen out of the 24 questions asked, thereby giving a range of 0 to 24
To gauge social isolation in older adults [26
Measures perceived social support received by family and friends [26
A range of 0 to 30, where a score of 12 or less suggests being at risk of social isolation [26
To assess formal and informal resource use [31
Number of carers involved in a patient’s care, carer time, carer work status, and patient and carer health resource utilization [31
To measure the period of time at which the informal carer indicates to be able to maintain current care if the situation remains the same [32
Six levels: (1) 1 week or less, (2) between 1 week and 1 month, (3) between 1 and 6 months, (4) between 6 months and 1 year, (5) between 1 and 2 years, and (6) more than 2 years [32
A range of ‘1 week or less’ to ‘more than 2 years’
To test whether the ICECAP-O is a valid measure of capability well-being, two main sections of analysis were performed: convergent validity and discriminant validity. A priori expected correlations and relationships between the ICECAP-O and other variables from the questionnaires, discussed below, were drawn from previous literature, if available. In the analyses, correlation strength levels were taken from Cohen’s Set Correlation and Contingency Tables [34
]. Correlations are considered strong if the coefficient is above 0.5, moderate if the coefficient is between 0.3 and 0.5, and weak if the coefficient is below 0.3. A p
value of 0.05 was taken to signify statistical significance.
To test convergent validity, Spearman correlation coefficients of the tariff scores and dimensions of the ICECAP-O were compared against the EQ-5D-5L results (utility tariff, health problems index, and VAS) [35
], CarerQol-7D tariff scores and CarerQol-VAS scores, respectively. It was anticipated that there would be a moderate positive correlation between the ICECAP-O scores and the EQ-5D-5L utility tariff scores and VAS scores of carers, a moderate negative correlation between the ICECAP-O scores and the EQ-5D-5L health problems index of carers, and a strong positive correlation between the ICECAP-O and the CarerQol scores.
For discriminant validity, sub-groups were defined based on characteristics that previously were shown to be related to informal carer outcomes. For measures that have no pre-defined cut-off points for high or low, in this case the EQ-5D-5L tariff and VAS scores, the cut-off points between sub-groups were primarily based on a face valid classification in relatively similar group sizes. Education was split unto three sub-groups based on primary school only (low), up to high-school education (medium), and higher education (high).
Student’s t tests (for two sub-groups) or ANOVA (for more than two sub-groups) were performed to identify significant differences in ICECAP-O scores. Then, a multivariate regression model was estimated for the ICECAP-O tariff scores using all variables in which the ICECAP-O could discriminate at a P value of 0.1 or less, to gain insight into the magnitude and significance of the variables that were associated with the ICECAP-O scores. There are exceptions to this exclusion rule: the variables age, gender, education, relationship between the carer and person with dementia, and carer daily hours. We include age, gender, education, and the type of relationship because these are basic demographic factors. It was pre-defined by the authors that carer daily hours would be included in the multivariate regression as it is a key variable in the care giving context. A second model was estimated including country dummies, to account for country-level effects. In this regression, Germany was used as the reference country as it had the lowest mean ICECAP-O score among carers.
Several hypotheses were generated regarding carer, care receiver and caregiving context variables and their relationship with the ICECAP-O. It is important to note that this literature did not necessarily refer to informal caregivers, or carers of people with dementia. Regarding carer variables, employed carers were expected to have a significantly higher ICECAP-O score than those unemployed [36
], carers with higher health status (i.e. a higher EQ-5D-5L score) were expected to have significantly higher ICECAP-O scores than those with lower health status [37
], and carers with a higher PAI score were expected to have a significantly higher ICECAP-O score than those with a lower PAI score [38
]. Furthermore, there was insufficient evidence to form a hypothesis on the effect of carer age on the ICECAP-O [20
]. There was no expectation for the ICECAP-O to score differently for different levels of carer education [20
]. Regarding care receiver variables, carers for a person with dementia with a higher health status (i.e. a higher EQ-5D-5L, DemQoL-U and QoL-AD score, or a lower CDR score) were expected to have a significantly higher ICECAP-O score than carers for persons with lower health status. Finally, regarding caregiving context variables, carers with a low care burden (i.e. fewer daily care hours, lower CANE unmet needs in the person with dementia, higher PT and/or higher RUD scores) were expected to have a significantly higher ICECAP-O score than those with a higher care burden [40
], and carers with a higher LSNS score were expected to have a significantly higher ICECAP-O score than those with a lower LSNS score [41
All tariff scores (for ICECAP-O, EQ-5D-5L and CarerQol-7D) were calculated using UK value sets because of both their availability and the need for consistency. The proxy ratings of the informal carers were used for the EQ-5D-5L, QoL-AD and DemQoL-U of people with dementia. All analyses were performed in STATA 14.
The aim of this paper was to determine the validity of the ICECAP-O in a relatively large, eight-country population sample of informal carers for people living with dementia. Validation was performed using convergent and discriminant validity tests, followed by multivariate analysis. As hypothesized, there were significant moderate-to-strong correlations in the expected directions between the ICECAP-O scores and carers’ EQ-5D-5L utility tariff score and health problems index, EQ-VAS scores, as well as the CarerQol-7D and the CarerQol-VAS scores. The multivariate regressions showed that the age of the person with dementia, the EQ-5D-5L health index of the person with dementia, carer–patient relationship, care recipient CDR, carer LSNS Score, the carer PAI score, and Perseverance Time all had a significant relation with the carer ICECAP-O score. The fact that age of the person with dementia had a non-linear relationship with the carer’s ICECAP-O score may be explained by older people with dementia having more health and behavioural problems that were not captured in the multivariate regression. The reason for age of the person with the dementia being correlated with an increase in the ICECAP-O until age 79 is still a somewhat surprising result, perhaps explained by younger people wanting to take part in more activities or work than their older counterparts. Somewhat surprisingly, the ICECAP-O did not have a significant relationship with the number of daily care hours, even though it was assumed these would have an impact on carer well-being. This may be due to the selection of the sample as only people with mild to moderate dementia were included. The results also showed that living in certain countries may be of importance for the carer ICECAP-O scores.
Even though we presented the first validation of the ICECAP-O instrument in a sample of informal carers, our results were quite comparable to results from previous ICECAP-O validation studies [5
]. While many addressed the specific dimensions of the ICECAP-O rather than the ICECAP-O tariff scores, several of the results found in our study were similar to those of previous studies. Almost all studies found that the ICECAP-O could discriminate effectively between groups of different ages [5
]. Additionally, all studies found moderate-to-strong convergent validity between the ICECAP-O and health (quite frequently using the EQ-5D as measure) although not necessarily for every dimension of both measures. Makai et al. [20
] also found that the ICECAP-O could discriminate between older people who had more or fewer opportunities for social interaction, which is in line with the significance of the LSNS score in our analysis. Most previous validation studies conclude that the ICECAP-O may be a promising patient outcome measure in economic evaluations, although it may not completely cover physical health [18
]. Based on the results of the above analysis, this paper comes to the same conclusion for the validity of the ICECAP-O in carers (of persons with dementia). An interesting finding is that in Sweden the ICECAP-O was not correlated with the EQ-5D-5L utility tariff score and health problems index. One reason for this may be that in our sample, the lowest EQ-5D-5L utility tariff score in Sweden is approximately 0.37, which is far higher than the lowest score from the full sample (− 0.1).
The main strength of this study is that it is the first to validate the ICECAP-O in carers: in a sample both large in size and country variety. While several validation studies of the ICECAP-O have been executed, they all used relatively small sample sizes and only focused on one country. The eight-country nature of this sample allowed a more comprehensive overview of the ICECAP-O’s validity in Europe (in carers). Another strength is that extensive data were provided on both carers and care receivers. None of the previous studies looked at convergent validity between the ICECAP-O and the CarerQol, CANE, RUD, PAI, QoL-AD and Perseverance Time. Moreover, this study was the first to validate the ICECAP-O in carers.
Some limitations of our study need to be mentioned as well. First is the lack of variation within the sample. Large percentages of both carers and care receivers were relatively healthy and most carers seemed to experience a relatively low care burden, which may be the result from selection bias as carers who experience a high care burden may be less likely to participate in the study. Therefore, a detailed analysis of validity of the ICECAP-O in those carers who are less healthy or feel higher care burden is not possible here. Second, carer proxy scores were used for some of the outcome measures for people with dementia (i.e. EQ-5D-5L, QoL-AD and DemQoL-U). The correlation of ICECAP-O scores with the harder to observe variables for persons with dementia (such as the DemQoL-U items) may be less reliable than those with more easily observed variables (such as the EQ-5D-5L items). In addition, due to the neurodegenerative nature of dementia and the stress experienced by carers, proxies may give more negative answers regarding care receivers’ health and well-being [42
]. While this most likely does not affect our regression results, as from the measures of health of persons with dementia only CDR was used, it is worth bearing in mind for future studies. Another limitation is the use of UK value sets for both the ICECAP-O and EQ-5D-5L-related measures. This was done for consistency, as value sets were not available for all countries in the sample; however, it may be partially responsible for Sweden-specific EQ-5D-5L results being uncorrelated with the ICECAP-O. Finally, we used the ICECAP-O in the complete sample of carers. However, as can be derived from Table 2
, nearly half of carers were younger than 65. The ICECAP-O (Older) was designed to capture the capability well-being of people age 65 and over. For people aged under 65, the ICECAP-A (Adults) [6
]—which covers the five capability well-being dimensions attachment, stability, achievement, enjoyment and autonomy—would in principle be more suitable. This was not feasible in the current study. It is unclear how accurately the ICECAP-O measures the well-being of people aged under 65.
It has been shown in previous validation studies that the ICECAP-O is a worthy contender as a patient outcome measure in economic evaluations regarding care of the older people, due to its broad, well-being-focused nature. In this study, the ICECAP-O has shown good convergent and discriminant validity as a well-being outcome measure in carers of people with dementia. These findings suggest that the ICECAP-O potentially is a relevant and useful measure for economic evaluation in samples of elderly informal carers, especially when considering interventions that have impacts ‘beyond health’. If it is used in both carers and care receivers, this allows comparisons of outcomes across interventions and aggregation of outcomes within interventions. Before being able to recommend this, a number of important issues need to be resolved. First, the ICECAP-O needs to be further validated as an outcome measure among people with dementia and their carers. This would include linguistic validation of translations of the ICECAP-O, currently being analysed in Germany and Portugal as part of the Actifcare project [21
]. It would be beneficial to conduct linguistic validations in other countries where psychometric validations have been conducted [48
]. Second, future studies need to confirm our results and expand on them, to increase the evidence of the validity of the ICECAP-O. Third, a choice needs to be made whether the use of the generic ICECAP-O (which is aimed at older people) is to be preferred over the use of more carer-specific well-being measures, such as the CarerQol. While the results of the latter may be less easily aggregated with, for example, ICECAP outcomes in patients, they may provide more precise estimates of care-related quality of life and more detailed information. Finally, while the CarerQoL is aimed at carers (regardless of age), the ICECAP measures would need to be tailored to age groups of carers, which raises questions of aggregation and comparison of ICECAP-A and ICECAP-O scores.
Further research of the ICECAP-O in samples of informal carers for people with different chronic illnesses would also be useful. It would allow investigation into whether the ICECAP-O is also a valid measure and shows similar relationships to other outcomes in the context of diverse chronic illnesses. If the ICECAP-O is to be used as a well-being measure in economic evaluations, it would also be of interest to conduct further research into its sensitivity to change and Minimal Clinically Important Difference.
The ICECAP-O is a capability well-being measure that has been proven to be of use for economic evaluations of care of older people. This study adds that the ICECAP-O may be useful in economic evaluations of interventions considering elderly informal carers, where a broader measure of well-being is more relevant than a narrower health-related quality of life measure such as the EQ-5D-5L.
The Actifcare Consortium partners are as follows: Coordinator: Maastricht University (NL): Frans Verhey, professor (scientific coordinator, WP1 leader). Consortium members: Maastricht University (NL): Marjolein de Vugt, Claire Wolfs, Ron Handels, Liselot Kerpershoek. Martin Luther University Halle-Wittenberg (DE): Gabriele Meyer (WP2 leader), Astrid Stephan, Anja Bieber. Bangor University (UK): Bob Woods (WP3 leader), Hannah Jelley. Nottingham University (UK): Martin Orrell. Karolinska Institutet (SE): Anders Wimo (WP4 leader), Anders Sköldunger, Britt-Marie Sjölund. Oslo University Hospital (NW): Knut Engedal, Geir Selbaek (WP5 leader), Mona Michelet, Janne Røsvik, Siren Eriksen. Dublin City University (IE): Kate Irving (WP6 leader), Louise Hopper, Rachael Joyce. CEDOC, Nova Medical School, Faculdade de Ciências Médicas. Universidade Nova de Lisboa (PT): Manuel Gonçalves-Pereira, Maria J. Marques, Conceição Balsinha, Ana Machado, on behalf of the Portuguese Actifcare team. Alzheimer’s Research Unit-Memory Clinic, IRCCS Centro S.Giovanni di Dio (IT): Orazio Zanetti, Daniel Portolani.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.