Sample and design

Participants were from the CC75C study, a population-based longitudinal cohort study representative of Cambridge’s (UK) older people. Full details of the methodology has been described elsewhere [24] and can be found online at can be found on line at http://www.cc75c.group.cam.ac.uk/. The baseline survey enrolled 95% (n = 2610) of individuals aged 75 years and older, approached from six general practices in Cambridge in 1985/87. Participants were followed with surveys repeated at two (Survey 2), seven (Survey 3), ten (Survey 4), 13 (Survey 5), 17 (Survey 6) and 21 years (Survey 7). Interviews were conducted with the study participants and a proxy informant if the participant was unable to complete the interview. Proxy informants were usually relatives but could sometimes be a friend/neighbour or a warden or member of staff in a care home. Observations by the interviewer were also gathered during the course of the interviews. Core data included information on socio-demographic variables (e.g. place of residence, household structure, marital status and social contact), activities of daily living (ADLs), use of health and social services, cognitive function (e.g., Mini Mental State Examination (MMSE) [25]), health problems and medication. This data except for HRQL and cognitive assessments was retrieved from the caregivers if necessary. The interview schedule has undergone slight revisions with the addition of new sections, such as questions on affect and loneliness added in survey 3 and 4. Survey 3, conducted in the 7^th year of follow up, was the first survey that best represented all the DEMQOL domains, with questions including socio-demographic information, social relationships, mood, loneliness, anxiety, depression, ADLs and cognitive functioning. Therefore data from Survey 3 will be used in this analysis.

Dementia diagnosis

Dementia was diagnosed using a two-stage approach. At baseline, individuals who scored ≤ 23 on the MMSE, and one in three individuals with a MMSE score of 24 or 25, were assessed using the Cambridge Mental Disorders of the Elderly Examination (CAMDEX) [26], a structured schedule specifically designed to detect mild dementia. Between the baseline interview and survey 3, six assessments for dementia were conducted. Based on the CAMDEX results, different dementia types were assessed including: Alzheimer’s disease (AD), vascular dementia (VaD), mixed (AD and VaD), dementia secondary to other causes, clouded/delirious state plus AD and clouded/delirious state plus VaD. Dementia diagnoses were also sought by checking death certificates for any mention of dementia as a cause of death and checking reports from interviews with relatives for the sub-sample of participants who donated their brains to research. No additional people with dementia were identified. For the present study, participants with a last diagnosis of dementia, from any CAMDEX assessment and MMSE score > 10 (n = 110, 40 men and 70 women aged 80-99) up until one year after Survey 3 were identified for inclusion in this analysis. Among those without a MMSE score, 2 people were diagnosed with minimal, 2 with mild and 2 with moderate dementia using the CAMDEX. The staging of dementia was calculated using cut-offs on the MMSE as follows: 11–20 = moderate; 21 and above = mild [27].

Assessment of socio-demographic and functional ability

Socio-demographic variables included age (continuous), gender, marital status (currently married, widowed, separated/divorced, and never married), education (up to 14 years of age and 15 and above) and accommodation (house, warden controlled housing, residential home and long stay hospital). Functional disability included impairment in basic and instrumental ADLs (IADLs). Individuals were categorised into three groups based on their pattern of impairment including: (1) IADL disability only if they required help with cooking, housework or both; (2) ADL + IADL disability if they needed help with any of four basic ADL tasks including: bathing, dressing, getting to the toilet on time and grooming; or, (3) No ADL/IADL disability.

Questionnaire mapping and development

In this study we chose to map the DEMQOL framework as it includes the domains that have been most commonly assessed in dementia-specific HRQL measures, namely daily activities/looking after yourself, health and well-being, cognitive functioning, social relationships, and self-concept. Further, the conceptual framework underpinning the DEMQOL was developed in in the UK. Using data obtained at Survey 3, the following subdomains of the DEMQOL framework could be mapped including: health and well-being, social relationships, self-concept and items from daily activities corresponding to enjoyment of activities. Only self-reported information was used to assess HRQL. Due to the high percentage of missing responses to HRQL items in the most cognitively impaired participants, only those with a MMSE score above 10 were included in the study, as has been suggested previously [28–30].

Analysis

The psychometric properties of the final version of the questionnaire (16 items) were tested with the sample of 110 people with dementia identified as described above. Standard psychometric methods were used [31] to evaluate acceptability, reliability (internal consistency) and validity (content, convergent, discriminant and known group differences). Criteria for acceptability were: missing data for summary scores <5% and normality of the distribution of the total score (Skewness measured with Shapiro-Francia, Shapiro-Wilk and Skewness test with a p value higher than 0.05 indicating normality). Criteria for internal consistency were Cronbach’s alpha ≥0.7. Convergent and discriminant validity were measured with Spearman correlations (r). Convergent validity is the evidence that the scale is correlated with other measures of the same or similar constructs and was deemed acceptable when r was 0.20 and higher. Discriminant validity is the evidence that the scale is not correlated with measures of different constructs and was deemed acceptable when r was below 0.20. Known group differences refer to the ability of a scale to differentiate groups who are expected to differ on the construct being measured. Differences in HRQL scores by socio-demographic, functional and measures of disability and cognition were tested using the Wilcoxon Mann-Whitney tests (variables with two groups) and Kruskal-Wallis test (variables with more than two groups). Statistical significance was set at p < 0.05. A meta-analytical review was conducted in order to generate a priori hypotheses as part as the validity assessment of the measure used. This has been annexed as a separate report (Additional file 1) as a separate report. We used factor analysis to evaluate hypothesised subscales based on the conceptual framework. All analyses were performed using Stata 11 [32].

Ethics approval and consent

Each CC75C study phase was approved by Cambridge Research Ethics Committee and National Research Ethics Service Committee East of England- Cambridge Central (Reference numbers: 05_Q0108_308 and 08_H0308_3).

Written informed consent was obtained from all participants. For those deemed to lack capacity to give fully informed consent, and those unable to give consent in writing for reasons of physical disability, both the study participant’s clear assent to be interviewed and written proxy informed consent from a relative or caregiver well known to the participant were attained.

Description of sample

Results from mapping

We were able to map all the DEMQOL domains except for those that are dementia symptoms (cognition and ADLs except for enjoyment of activities). One out of seven subdomains of self-esteem, nine out of eighteen of health and well-being, three out of eight of social relationships and one out of fifteen of daily activities (one out of four of enjoyment of activities) were mapped onto the CC75C study.

Psychometric evaluation

Distribution of HRQL by socio-demographic and functional variables

Strengths

There are a number of strengths in the approach we have taken. Firstly, this is a population-based sample of older old adults with dementia. The study also includes both people with dementia living in the community and institutionalised populations. The present study did not specifically set out to measure dementia-specific HRQL. However, we have been able to maximise its value by mapping most dimensions of the DEMQOL onto this project and studying the distribution of HRQL in a representative sample of the Cambridge population over 80 years old with dementia. One of the strengths of the DEMQOL conceptual framework is that it was developed using a combination of interviews with people with mild to severe dementia and their non-professional caregivers, and a review of the literature and existing instruments. The adapted version of the DEMQOL using the CC75C study questions has been shown to have acceptable psychometric properties suggesting that this scale fulfils standard criteria for acceptability, good internal consistency, and validity, namely content validity, discriminant validity and to some extent, known group differences. These properties have been tested in a population sample, providing therefore evidence of external validity of this instrument. Something no other dementia-specific measure has proved yet [18]. Another strength of our instrument is the breadth of dimensions measured compared to dimension-specific instruments such as the Progressive deteriorations Scale (PDS) [33] or a number of other dementia-specific HRQL measures [34–38].

Weaknesses

Several limitations need to be taken into account when analysing these results. The sample used in this study was derived from several phases of a longitudinal cohort investigating HRQL in people with dementia. The use of this data may therefore result in a number of biases in the research, for example survival bias in which there might have been a selection of people with factors associated to survival or follow up such as gender, education or setting, may lead to some unrepresentative distributions of HRQL. Cognitive bias may also be affecting the results of this study given the lack of insight of people with dementia [39]. However evidence suggests that people with mild to moderate dementia are able to assess their HRQL [22, 28, 40, 41]. There are other limitations, principally arising from the methodology of exploring HRQL, previously un-researched in this sample in which HRQL had not been assessed using existing dementia-specific measures. Firstly, since the instrument has been designed a posteriori, the question stems, response options and time frame of the questions were not identical to those from the DEMQOL framework. A limitation of our instrument compared to other instruments such as the ADRQL [13], QUALIDEM [20] or DEMQOL proxy [22] is that HRQL in people with severe dementia could not be analysed given the high percentage of missing values and the lack of proxy ratings for HRQL items. Compared to the original DEMQOL [22] and Bath Assessment of Subjective Quality of Life in Dementia (BASQID) [42] our instrument did not include items on worry/satisfaction with cognition and activities of daily living. Regarding the psychometric properties, some aspects such as test-retest reliability could not be evaluated due to the study design. Causality in the associations between HRQL and covariates cannot be inferred in this study, as it is cross-sectional. Finally, the distribution by ethnicity, medication or relationship with the caregiver could not be explored. These variables have been shown to be associated in previous studies [29, 43].

Summary of distribution of HRQL

Age was not associated with HRQL, a finding consistent with previous literature [42, 44–47]. Findings such as the lower HRQL in men with dementia who are 90 years of age and older compared to younger groups may be due to differential mortality between gender [48]. The oldest age group is likely to have worst health than the other two age groups. The established relationship between married state and higher HRQL in men is found for men with dementia too. This could be related to social support mechanisms given one of the primary benefits of marriage for men is social connectedness [49]. According to the sex role hypothesis, this positive effect of marriage would not affect women because of the ungratifying nature of housework (usually conducted by women) and women’s primary responsibility for household chores [50, 51]. This lack of effect of marriage in women and the high percentage of widowed women and married men could account for the higher HRQL of men. The finding that men have higher HRQL than women could also indicate that women had higher levels of, for example depression and anxiety, as occurs in the general population [52, 53]. In fact we have found that in our sample, women have higher levels of depression than men. This finding is consistent with what is observed in clinical practice when offering support to relatives of patients with Alzheimer’s disease, and could also be related to the greater difficulties women face in terms of continuing to perform the tasks associated with their role in the family and generational context [46]. Similar results have been reported in other populations [54–56].

HRQL did not vary by levels of cognition and its association with disability is not clear. There is increasing evidence that severity of dementia may not be a major determinant of HRQL in dementia, indicating that it is possible to have good life quality at all levels of dementia severity [22, 23, 40, 57]. This finding casts doubts on the usefulness of interventions only aimed at improving or maintaining cognition and function in order to improve the quality of life of the patient with dementia. However, there is also evidence pointing at the decreased awareness of their cognitive impairments and change of behaviour by patients with dementia [58]. This also manifests as poor awareness of deficits in ADL. In general, awareness of deficits seems to decrease with an increased severity of dementia [39, 59]. This could account for the lack of association between severity and HRQL. In fact, people with moderate dementia severity who left school after age 14 had a very low HRQL score compared to those with the same level of education but higher cognition and to those who left school before that age. This may reflect that this group are more conscious of their limitations in relation to more intellectual tasks, e.g. reading for pleasure, than people with less formal education, who may be less troubled by such impairments. This could also be related with the cognitive reserve hypothesis [60]. There is evidence showing that high cognitive reserve groups have a higher cognitive decline but later [61, 62]. A systematic review, found cognitive decline to be associated to HRQL [23]. Patients with high reserve and moderate dementia might be more worried about the sudden decline of their functioning compared to the other groups.

Finally, higher HRQL in those with good self-rated health compared with people with lower self-rated health has also been previously reported [47]. Whereas it is plausible that health itself may be related to HRQL, the subjective nature of the rating of both concepts and the inclusion of a similar item as part of the concept of HRQL may partly explain this association.

Future directions

Many population-based studies on dementia exist. None of them set out to measure dementia-specific HRQL. However, a number of these studies include variables that could be used to map a dementia-specific HRQL framework into the study. Doing so would be a useful way of maximising already existing resources and allow comparison across populations and provide valuable information for policy development.