A cross-sectional study was conducted in 8 nursing homes belonging to the same foundation spread over 6 cities in Spain. They involved 217 people with dementia in moderate stages who were older than 65 and had been living in the nursing home for longer than 3 months. Fourteen nursing professionals of each nursing home were asked to complete the questionnaire in the manner that the patient would have if he/she were able to answer the questions. In all the cases, five researchers collected all the data, interviewing all patients and the collaborating nurses when needed.

Inclusion criteria: over 65 years of age and institutionalized for more than 3 months, presenting a diagnosis of dementia with over six months of evolution, classified according to the GDS (inclusion levels 4 to 6 of the Global Deterioration Scale (GDS) [33]. This classification should be made by the neurologist who diagnosed or followed the evolution of the patients.

The study was approved by the Bioethics Committee of the Foundation in Madrid (COD. 2016.08) and at all directions of the centers. Informed consent was obtained from legal guardians for all participants after they were informed verbally and in writing. The principal investigator is responsible for keeping informed consent.

Sample size. The sample size was estimated with a programme for calculating sample size and the power of a contrast of hypothesis (Granmo v. 7.11, developed by Institut Municipal d'Investigació Mèdica, Barcelona, Spain), for a significance level of 0.05 and a power of 0.8, assuming equal variances). Adjusting for a dropout rate of 20%, it was determined that 208 people were needed.

Different socio-demographic variables of patients were collected, such as age, sex, marital status, and length of stay in the nursing home.

The Global Deterioration Scale (GDS) [33] assesses the degree of dementia. This scale enables the evolution of cognitive impairment from normal aging to a person with dementia. It consists of seven phases, ranging from "zero cognitive impairment" to "very severe dementia." The first two levels (GDS 1 and 2) correspond to normal states. The GDS 3 is the first phase in which there are clear deficits, although it is not considered sufficient to diagnose dementia. GDS 4, GDS 5, GDS 6, and GDS 7 reflect a cognitive impairment linked to dementia (mild, moderate, and severe dementia, respectively).

The ICECAP-O [22] [23] measures capability wellbeing using five domains or attributes (attachment, security, role, enjoyment, and control), with one question per dimension. Distinguishing four levels within each domain (levels generally range from all, a lot, a little, to not any; exact wording of levels varies per dimension). Each level distinguishes 1,024 possible “capability states.” The ICECAP-O tariffs have values between 0 (no capabilities) and 1 (full capabilities). The ICECAP-O is a five-item multiple choice questionnaire where each attribute has 4 possible response options. For this first use of the ICECAP-O in Spanish, the instruments' original author's/recommended translation guidelines were fully met. The original ICECAP-O version, as developed by Coast et al., [23] was forward-backward translated from English into Spanish by two independent translators. Subsequently it was translated back into English by two independent translators. The last version in English was sent to the original authors of the application for its validation.

In order to compute capability values, the British tariffs were applied because these do not exist for Spain. Lower scores thus represented fewer preference-based capabilities. The complete explanation as to how to calculate them was fully described on the ICECAP website: http://www.icecap.bham.ac.uk/tariffs.shtml.

Alzheimer Disease Related Quality of Life (ADRQL) [19]. We used the revised 40-item version that allows for the assessment of QoL for people with mild, intermediate, or late-stage dementia using proxy responses [34]. The dementia-specific, multi-dimensional ADRQL instrument can be completed by patients’ professional caregivers [20]. The ADRQL measures the dimensions of social interaction, awareness of self, enjoyment of activities, feelings and mood, and response to surrounding. The various dimensions range from 4 to 12 items on a dichotomous scale and each item is weighted in a range between 9.15 and 13.75, based on a judgment of importance by caregivers [20]. For each dimension, a separate subscale can be calculated and summed up in one total score ranging from 0 (lowest quality of life) to 100 (highest quality of life) [19]. The instrument exhibits good psychometric properties with adequate validity, good internal-consistency reliability, very low missing data, and good sensitivity to change [20] [34].

The EQ5D [18] Spanish version [35] measures HRQoL in terms of five dimensions (mobility, self-care, daily activities, pain and discomfort, and anxiety and depression) with three levels each (1 = no problems, 2 = moderate problems, and 3 = extreme problems) describing 243 health states. For use in people with dementia, the EQ-5D was extended with a cognitive dimension (EQ-5D+C), for which utility scores are unavailable [35] [36]. The health index component of the EQ-5D+C is therefore made up of six dimensions: mobility, self-care, usual activities, pain/discomfort, anxiety/depression, and cognition. Each dimension has three levels: no problems, some problems, and extreme problems.

The Barthel-Index [37], Spanish version [38] has 10 items of activities of daily living (ADL): feeding, grooming, bathing, dressing, bowel and bladder care, toilet use, ambulation, transfers, and stair climbing. The total score ranges from 0 to 10. The total score thus ranges between 0 (completely dependent) and 100 (completely independent) with a cutoff score of 65 indicating need for ADL assistance, with a classification in low (40–65), medium (20–40), and high (<20) care dependency groups. This is the most common test that measures residents’ abilities to perform activities of daily living by proxy- or self-report.

To establish convergent validity we expected moderate to strong and positive correlations between the ICECAP-O, the EQ-5D+C, and ADRQL (these instruments measure partial operationalizations of QoL) (H1). Furthermore, we expected a moderate and positive correlation between the ICECAP-O dimension and tariff scores and the Barthel Index (ADL) (H2). For discriminant validity, we expected to find differences in ICECAP-O tariff scores between residents suffering from severe, moderate, and mild/moderate dementia (based on the GDS), between ADL dependent (Barthel score <65) and ADL independent (Barthel score >65) residents between low, medium, and high care dependency groups (H3). A higher score on the ICECAP-O was expected for the better-off groups.

The construct validity of the ICECAP-O questionnaire was calculated through factor analysis using principal component extraction and Varimax rotation with Kaiser. Each variable was included in a single factor, according to their factor loading, by setting minimum saturation criteria of 0.50. Varimax rotation is assumed to be the most appropriate, since it is expected to discriminate between the maximum number of factors forming the scale. Estimates of sampling adequacy for Kaiser-Meyer-Olkin (KMO; range 0–1) and Barlett statistical significance (if its value is close to the unit and they are significant with p <0.05, indicating that the analysis with reduced variables is appropriate) were calculated. The reliability of the questionnaire was calculated by analyzing the internal consistency, using Cronbach Alpha, which should be interpreted as an indicator of the internal consistency of the items, because it is calculated from the covariance between them.

The demographic characteristics of residents and proxies were analyzed using descriptive statistics. Mean or median, standard deviation, were computed from the demographic data of the residents and for the different dimensions of the instruments. Correlations between the outcomes of the ICECAP-O and dimensions of the ADRQL, EQ-5D+C, and ADL were used to estimate convergent validity. Discriminant validity was analyzed using a t-test and one-way ANOVA to explore differences in means of the ICECAP-O between different demographic and dementia-related groups. For all analyses the level of significance was set at p<0.05. Data was analyzed using SPSS v.22.0.

Estimates of sampling adequacy of Kaiser-Meyer-Olkin (KMO = 0.877) and Bartlett statistical significance (p = 0.000) were calculated. Communalities analysis indicated that all items are above 0.5, and then all questions were accepted as valid to form part of the questionnaire. Using this method, the principal components’ 5 major components or the factors with eigenvalues greater than 1 were obtained, which account for 66.57% of the total variance, and an explanation percentage at appropriate acceptance levels was obtained (see Table 1).

These factors were classified as the same 5 domains or attributes as the original version [21]: attachment, security, role, enjoyment, and control. One item was included in a single factor, according to their factor loading, by setting saturation minimum criteria at 0.30. Rotated factor solutions according to VARIMAX formed a well-defined structure without overlapping. To assess the internal consistency of the questionnaire a Cronbach alpha value of 0.820 was obtained. This value was not improved by eliminating some of the items, therefore all items were maintained. The values of item-total correlation ranged between 0.52 and 0.73.

Data were collected between March and June 2016. The questionnaires were collected in 8 nursing homes by 14 nurses from a total of 217 patients with mid-stage dementia. Table 2 shows the sociodemographic variables, the characteristics of the residents, the measurement instruments, and their dimensions.

Table 3 shows that the ICECAP-O scores were strongly correlated with EQ-5DC scores, ADRQL scores, and Barthel scores. Correlations between the ICECAP-O tariff scores and the different dimensions of the EQ-5D+C were strong and significant. Correlations between the ICECAP-O and the ADRQL proved to be similarly strong and significant.

The individual ICECAP-O dimensions of role and control were strongly and significantly correlated with the EQ-5D+C and to a lesser extent with dimension mobility and self-care, usual activities, pain/discomfort, and cognition. Attachment was significantly and strongly correlated with SI, FM, and EA (ADRQL). The Barthel index was significantly correlated with all ICECAP-O dimensions with correlations between the Barthel index and the role and control dimensions being particularly strong.

In Table 4, the results of the ANOVA showed significant differences in ICECAP-O scores between patients with different dementia severity and ages (i.e., above or below 75). ANOVA results showed that the ICECAP-O tariff scores differentiated between residents classified into different care dependency levels. As expected, lower scores were observed for the more severe groups, and higher scores for the less severe groups.

This study has a number of limitations worth mentioning. First, the residents for this study were not randomly selected and therefore might have characteristics that differ from the typical population with dementia in Spain nursing homes. However, this restriction may only influence the results to a certain extent, because the focus of the study was the validation of the properties of the quality of life measurement instruments and the current sample seems adequate. Second, British tariffs were used for the ICECAP-O since Spanish tariffs are not yet available. For this reason, results may be imprecise because weights for capability dimensions might vary between countries.

The average ICECAP-O tariff score found in this study within a nursing home setting (0.61) was comparable to the scores reported in a Dutch study performed in nursing homes (0.63) [28] and in a German study also performed in nursing homes (0.63) [30], and was substantially lower than the score for community-living elderly, where the average scores ranged between 0.81–0.84 [22] [27] [31] [32] [37–41]. These findings support the reliability of our results.

The finding of the strong correlation between the ICECAP-O dimensions and the EQ-5D dimensions shows that physical health is captured to a wide extent by the ICECAP-O capability measurement, which is consistent with other findings [18] [28–32]. Overall, we found significant correlations between the ICECAP-O dimensions and the individual EQ5D+C dimensions, with the exception of attachment, which was not significantly correlated with the mobility, pain/discomfort, and cognition dimensions of the EQ5D+C. Security was only significantly correlated with the EQ5D dimension of anxiety/depression. This can be explained because people with dementia, because the disease, lose their sense of security [41] [42].

Role and control correlated significantly with all EQ5D+C dimensions with the exception of Anxiety/Depression, which could be because people with dementia, especially in moderate and advanced stages, have anxiety disorders or depression treated with non-pharmacological and pharmacological measures [43]. These findings are in line with other similar studies [28–31]. The results confirmed the expected significant correlation between the ICECAP-O and the ADRQL scores, which shows that the ICECAP-O captures both generic HRQoL and dementia specific quality of life. The correlation between ADL and ICECAP-O scores reflected that a loss of independence in ADL was associated with a decline in wellbeing. This correlation is strengthened by the same finding for the other applied quality of life measurement instruments in this study, confirming previous results that reduced ADL leads to a decrease in quality of life [44]. These significant findings point in the direction of favorable convergent validity.

The ICECAP-O dimension “security” is an exception to the other dimensions because it showed almost no significant correlation with the majority of dimensions of the other quality of life measures. The same finding was observed in the Dutch and German ICECAP-O validation studies in which no correlation with the dimension “security” was found [29, 30]. This might be because people with dementia do not generally worry about the future past a certain point, [41]. On the other hand, the moderately to highly significant correlations of the capability dimensions “enjoyment,” “role,” and “control” with the ADRQL and the EQ-5D+C dimensions support the hypothesis that the ICECAP-O captures dimensions that are relevant for the elderly. Such correlations with the EQ-5D were also found in the general British population [22] and in the German nursing homes [30] [41].

The ICECAP-O discriminated between patients based on the variables age, dementia severity, and ADL. Therefore, the group of healthier residents and the group with a higher functional status also reached the expected higher tariff scores. This suggests that the ICECAP-O is sensitive to age differences and to indicators of health, confirming that reasonable discriminant validity exists.

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. This is a particularly useful property in case populations such as institutionalized patients with dementia, distinguished by decreasing independence and multi-morbidity, potentially across different health dimensions.

Further research is needed to confirm the current favorable findings and to further explore the feasibility, validity, and usefulness of the ICECAP-O instrument, as well as in the context of economic evaluations. More research is especially encouraged in more homogeneous populations characterized by a single disease, since dementia is a syndrome with very heterogeneous diseases (eg. Alzheimer, Lewy bodies, etc).