Background
Economic evaluation of healthcare services aims to inform policy makers by comparing the costs and benefits of alternative health care interventions. In such an evaluation, it is crucial that besides all costs, all benefits of healthcare services are captured. Capturing such benefits can be challenging, since healthcare services such as elderly care, long-term mental health, and public health may impact individuals health and health related quality of life, as well as their wellbeing more generally [
1‐
4].
Health can be defined as a multidimensional construct of physical, psychological and social dimensions [
5]. These health dimensions can be inter-related, for example decreased mobility may lead to a decrease in social contacts and depression [
6,
7], subsequently impacting social and psychological dimensions of health [
7]. Health related quality of life (HrQol) tries to capture how health impacts individuals’ Quality of Life (Qol) [
8]. In economic evaluations, benefits are frequently assessed by changes in health-related quality of life combined with the duration an individual spends in various health states. Duration and HrQol are then subsequently combined in Quality-Adjusted Life-Years (QALYs), and thus arguably capture the effect of healthcare services on physical, psychological and social dimensions of health. Aspects of broader wellbeing, such as maintaining independence, dignity, and comfort [
1], however, arguably are not captured by the concept of HrQol in its entirety. This can cause problems in capturing the full benefits of interventions, in particular in the evaluation of social care interventions, as well as integrated health and social care services [
9]. For example, specific social care interventions like day care and meals on wheels may improve wellbeing, but not health, or at least not only health [
9]. As a consequence, such services cannot be evaluated in the same manner as other healthcare services such as medicines [
9] where using HrQol seems more appropriate in many cases. Otherwise, the benefits of these provisions may be undervalued [
10].
Therefore, broadening the evaluative space of economic evaluations by a wider measurement of benefits has been suggested in evaluation of elderly care [
1,
11], using dimensions of wellbeing such as independence, attachment, or the ability to pursue valued activities [
10] in addition to health dimensions. In that context, a proposed alternative to measuring HrQol is to measure capabilities. Capabilities may be seen as a conceptualization of wellbeing [
1], defined as the capacity to perform certain actions and achieve certain states (irrespective of actually doing so). Capability wellbeing assesses what individuals can do instead of focusing on functioning, i.e. what individuals actually do [
1]. Capability-wellbeing captures a variety of health and non-health dimensions, which may be difficult to separate [
12].
In order to measure capability wellbeing, two instruments have been developed to date, the ICECAP-O [
10,
13] (ICEpop (Investigating Choice Experiments for the Preferences of Older People)) CAPability measure for Older people above 65 and the ICECAP-A [
1] for the general population. Both instruments are intended as outcome measures for economic evaluations of both health and social services, where beyond health, wellbeing aspects have to be considered as well [
1,
9,
10]. In order to be useful for economic evaluations, instruments should be sufficiently validated in terms of their convergent and discriminant validity. While the ICECAP-A has been validated in the UK only [
14], the ICECAP-O has been validated in a number of settings: in the British general elderly population [
10], in an Australian population of post-hospitalized elderly receiving residential care [
15], in a Canadian population of elderly visiting a fall-prevention clinic [
16] and a proxy version has been validated in Dutch nursing home settings [
17].
However, to date, the ICECAP-O has not been validated in a population of post-hospitalized older-people in the Netherlands. Post-hospitalized elderly are increasingly recognized as a population in which health improvements can be achieved [
18] through geriatric interventions. In the Netherlands, in the context of the National Care for the Elderly Program significant efforts are made to improve health and quality of life outcomes in frail elderly, for instance through the Prevention and Reactivation Care Programme among older patients who are admitted to a hospital [
19]. For elderly populations, hospitalization increases the risk of functional decline, defined customarily as a decrease in (instrumental) activities of daily living ((I)ADL) [
20]. Although elderly may be hospitalized due to function decline resulting from illness, such functional decline is also frequent after admission: 35% of 70 year olds and 65% of 90 year olds experience such a decline. Functional decline is therefore influenced by hospital care as well [
20], through increased complications [
21] or through less aggressive treatment regimens than customary in younger populations [
18]. In a group of post-hospitalized older people, a wide range of differences in health, capabilities and well-being problems may be expected due to (differences in) age, physical function, and other characteristics of the elderly such as multi-morbidity and support from their direct environment. As a result, this population is likely to receive various forms of publicly funded healthcare, as well as being the recipients of other social services. Furthermore, there is little research on how the ICECAP-O is related to other conceptualizations of wellbeing and the relationships between the ICECAP-O and measures of health (physical, psychological and social) remain underexplored. Exploring such issues is preferably done in a group in which a variety of health and well-being problems may be expected such as post-hospitalized elderly. Therefore, the aim of this study is to validate the convergent and discriminant validity of the ICECAP-O in a Dutch community-dwelling population discharged from a hospital in the prior three months. We further study the discriminant validity of the ICECAP-O by performing sub-group analyses, highlighting the differences in ICECAP-O scores between groups of elderly.
Discussion
Summary of main results
As hypothesized, the capability wellbeing instrument ICECAP-O tariffs were significantly correlated with other measures of wellbeing (Cantril’s ladder, the SPF-IL) as well as with all health measures (EQ5D dimensions and utilities, IADL, GDS, SF-20 Social Activity limitation). Contrary to expectations based on the type of instrument, the strength of the correlation between the ICECAP-O and the wellbeing measures was fairly similar as that with health measures. The individual ICECAP-O dimensions were also correlated with the overall scores of the different health and wellbeing measures. Overall, we found significant correlations between the ICECAP-O dimensions and the individual EQ5D dimensions, with the exception of Attachment, which was not significantly correlated with the Pain/Discomfort and Anxiety/Depression dimensions of the EQ5D and Security, which was not significantly correlated with the EQ5D dimensions Mobility and Self-care. As hypothesized, the ICECAP-O discriminated between the following measures in the bivariate and multivariate analyses: depressed and non-depressed elderly, IADL dependent and non IADL dependent elderly and between those with social activity limitations and without social activity limitations. In the exploratory analysis the ICECAP-O discriminated between multi-morbid and other elderly and between elderly with high and low EQ5D scores. Regarding measures of wellbeing, the ICECAP-O is significantly related to both Cantril’s ladder and the SPF-IL, even when correcting for health variables.
Limitations
This study has a number of limitations worth mentioning. First, our sample of elderly was not representative, but consisted of post-hospitalized elderly, who were previously admitted to a single hospital, living in one region of the Netherlands. Elderly in our sample are frailer than the general community-dwelling elderly population, reporting lower levels of mobility on the EQ5D [
36‐
38] than customary for the age group. Such reduced mobility suggests that our population is characterized by functional decline, consistent with frailty. In addition, patients in our sample were characterized by a broad range of diseases and multiple chronic conditions, with heart failure and osteoporosis being the most common diagnoses. Such a relative high number of elderly with multi-morbidity is also consistent with frailty. Associations between capabilities, health and well-being may be weaker in a general sample of frail elderly due to less variation in measurements. However, we have no indication that the selection of respondents drives the results regarding validity. Future research in other community-dwelling elderly populations also in other countries than the UK is necessary to further test this and validate the instrument. Second, we used a stepwise regression to identify explanatory variables of the ICECAP-O scores, which has limitations. In order to avoid rejecting possible significant variables, we used a relatively high p-value (0.2) for excluding variables. Additionally, we performed a regression analysis with all possible independent variables, which confirmed the results from the stepwise regression. It is worth noting moreover that, given the modest sample size, some subgroups were relatively small. This may lead to lack of power in establishing significant relationships.
Comparability with other findings
Compared to previous studies [
10,
15‐
17,
34], the values for the individual dimensions and overall scores of the ICECAP-O in this current study are similar to those obtained in the general elderly population and substantially higher than those obtained in a Dutch nursing home [
17]. The current scores are comparable to the British and Australian reference values [
10,
15,
34], with the exception of the attachment dimension, where the British and Australian studies [
10,
15,
34] report a higher percentage of older people at full capability (57% British and Australian studies vs. 36% current study) and the security dimension, where this current study has a far higher percentage of older people at full capability (53% current study vs. 18% British study vs. 37% Australian study). The differences in the attachment dimension cannot be explained by differences in the fraction of married elderly, which is quite similar across the studies. However, the elderly in the current study are a worse-off group (i.e. in terms of mobility) than the general elderly population in the UK, which may partly explain the lower scores on the attachment dimension. Differences on the security dimension may be explained by cultural differences in answering this question. Indeed, this is the second study in the Netherlands in which relatively high scores were found for the security dimension [
17]. Hence, Dutch elderly either have fewer concerns about the future than UK elderly or are less likely to share their concerns about the future. It also seems important to further investigate whether the translation of the description of the security dimension may lead to the observed differences. The average overall scores found here i.e. 0.84 were comparable to those obtained in the British and Canadian population (0.82), the Australian population (0.81) and substantially higher than for older people in a nursing home (0.63). Comparison of the overall scores suggest that on average the ICECAP-O scores of the Dutch community-living elderly are comparable to the general population in Australia and the UK, and are substantially better than elderly living in nursing homes in the Netherlands.
Furthermore, the correlations between the ICECAP-O, Cantril’s Ladder and EQ5D show broadly similar results as reported in previous studies, with a number of exceptions. Unlike the British validation study [
10] but in line with the Australian study [
15], we found a statistically significant though moderate correlation between ICECAP-O attachment dimension and the EQ5D dimensions mobility, self-care and usual activities. In addition, unlike the British study we found a significant correlation between the ICECAP-O’s security dimension and the EQ5D dimensions usual activities and pain. It must be noted that these are quite weak correlations, and significance may or may not be reached due to minor differences in sampling variation. Such minor differences in sample variation may be related to differences in the respective samples; here we approached previously hospitalized elderly, while the British study was performed in a sample from the general elderly population. Our correlation results were also comparable to a Dutch study using proxy respondents in nursing homes [
17]. There, however, the correlation between the ICECAP-tariffs and the EQ5D was somewhat stronger then found here, which may be due to differences between self-report and proxy responses. In this study the ICECAP-O is unrelated to Sex and Education level, which is consistent with previous findings.
Relationship between health and wellbeing and the ICECAP-O
Comparing the performance of the ICECAP-O to that of other health and wellbeing instruments, some aspects deserve mentioning. Given the strong correlations between the ICECAP-O measure of capability wellbeing and the other two wellbeing measures, as well as between the ICECAP-O measure and the EQ5D HrQol measure, ICECAP-O scores are related to both health and other wellbeing scores. The ICECAP-O scores are moreover related to individual health dimensions in terms of physical functioning, psychological functioning and social functioning. The tests of discriminant validity confirm this relationship between health measures and the ICECAP-O scores. Even though the ICECAP-O does not have an explicit physical dimension [
39], it seems that it is capable of capturing the effect of decreased physical function on capability wellbeing to a large degree, primarily through the control and role dimensions. With respect to the wellbeing instruments, the strong correlation between the ICECAP-O and Cantril’s ladder as well as the SPF-IL suggests that the ICECAP-O is related to these wellbeing measures as well, which is also confirmed in multivariate analyses. Table
4 does suggest that GDS has an influence on SPF-IL and Cantril’s ladder beyond what is captured by the ICECAP-O. This may be related to the concept of capability wellbeing or to the ICECAP-O instrument’s insensitivity for depression.
Implications for policy and future research
The ICECAP-O is a measure of wellbeing, and therefore has the potential to broaden the evaluative space of economic evaluations in health care by focusing on more than health alone. As such, it can potentially compare the benefits across a large number of sectors which (primarily) aim to improve wellbeing, such as (parts of) social care [
2], institutionalized elderly care [
40], public health [
3], and mental health [
4]. This is a particularly useful property in case of populations such as frail elderly characterized by decreasing independence and multi-morbidity, potentially across different health dimensions. The ICECAP-O measures (one conceptualization of) wellbeing. In doing so, its outcomes are, expectedly, related to health outcomes. The ICECAP-O moreover discriminates between various better off and worse-off groups. In this current study, in a post-hospitalized group significant insights were gained in terms of the relationship between capability wellbeing, life satisfaction, SPF_IL and various health measures. On the basis of our findings, we advocate the further use of the ICECAP-O measure in the context of economic evaluations, especially in those circumstances where broader well-being effects are expected and in combination with other measures. It can also be used in large scale surveys aimed at identifying depraved populations in order to identify groups which may benefit from interventions, as has been done previously [
34]. Nonetheless, a number of issues need to be explored further.
Further research is required to confirm the current favorable findings and to further explore the feasibility, validity and usefulness of the ICECAP-O instrument, also in the context of economic evaluations. In that context, larger studies would be helpful, allowing more subgroup analyses, as well as studies in different contexts (e.g. specific disease areas, living environments or cultural settings). Further research is especially encouraged in more homogeneous population characterized by a single disease. Furthermore, since the performance of the ICECAP-O has not been widely explored in longitudinal studies, the sensitivity to changes of the ICECAP-O is currently unclear. Whether the ICECAP-O comprehensively captures health and wellbeing changes, including depression, also deserves further attention. Additionally, further research is necessary to establish a causal relationship between health and wellbeing as measured by the ICECAP-O, and to explore ways in which the capabilities of older people can be improved.
Competing interests
The authors declare that they have no competing interests.
Authors’ contribution
All authors made substantial contribution to the design, analysis, interpretation of data, have been involved in drafting and revising the manuscript and have approved the final version.