Exploring self-report and proxy-report quality-of-life measures for people living with dementia in care homes

A total of 726 resident participants were recruited (see Table 1 for demographics) from 50 care homes, of which 377 completed self-report measures. Most participants were female (536; 74%), identified as White British (702; 96%) and had an average age of 85 (range 57–102). A staff proxy was recruited for each individual alongside 197 relatives/friend proxies.

The present study compared QoL measures for people living with dementia across multiple measures with multiple raters. The inclusion of three raters allowed comparisons to be drawn between three groups of participants who may be recruited for clinical trials where QoL is an outcome. The internal consistency of measures varied from questionable (EQ-5D-5L) to good and excellent (DEMQOL), in line with previous evidence [see 28 for review]. The QOL-AD NH has been shown to have variable internal consistency [29, 30]; in the present study, however, we found the measure to have good internal consistency. This may be due to differing levels of cognitive impairment between samples, which may affect how reliably measures are completed. For example, a small sample of individuals with mild dementia was recruited for one study where the scale demonstrated good internal consistency [30], whereas a second study excluded those with advanced dementia [29]. Therefore, further research is required to establish when use of the QOL-AD NH is appropriate.

Correlations between different measures across different reporters were generally weak to moderate, in line with recent similar studies [11, 31]. This suggests that people living with dementia and those who support them do not perceive QoL in the same way, or that they may focus on different aspects of QoL, suggesting a need for several measures to be completed to ensure full coverage of perceived QoL. However, the issues may instead be due to differences in how QoL is conceptualised by people with dementia and by different types of proxy informant. This is especially important as proxy raters are thought to focus on issues such as pain and presence of neuropsychiatric symptoms, rather than QoL specifically [18]. Recent qualitative research suggests that staff members equate residents’ QoL with the quality of care delivered or the stage of their dementia, whereas relatives draw comparisons with the person’s QoL when they were younger, lived in their own home and did not have dementia [31]. It is unclear how people with dementia, particularly those who are care home residents, conceptualise their QoL compared to proxies, although those who are experiencing pain and have recently had a fall report lower QoL [18]. Future qualitative work should be undertaken to understand how QoL is conceptualised and reflected on by different types of participants when completing these measures.

Our findings broadly indicated an at best, fair agreement, between how the different raters perceived QoL for people living with dementia. The QUALID staff and relative/friend proxy ratings yielded the highest level of agreement, with fair agreement also found between staff and relative/friend proxies on the EQ-5D-5L. However, it is noted that fair agreement is not considered to be a reliable when establishing the validity of a measure, where a minimum value of .6 (substantial agreement) is recommended [32]. Notably, there was poor agreement between self-rated QoL and staff/relative proxy-rated QoL on the EQ-5D-5L. When the data were examined as Index values, agreement was not statistically above chance. This is in line with previous research, which has found discrepancies between people with dementia and their family members on this measure [33]. Further analyses, comparing the percentage of individuals who reported having problems in areas vs not having a problem in the area, revealed interesting discrepancies. Particularly, most residents reported no problems with self-care, whereas both staff and relative/friend proxies identified that most individuals had problems in this area. This may reflect additional issues with the sensitivity of this question, since people with dementia may feel uncomfortable or embarrassed stating they experience problems with self-care. Alternatively, care staff may overstate the problems individuals with dementia have, based on their own approach to provision of support for personal care, which may not be based on maximising independence, but rather on completing care tasks as efficiently as possible. Research should be conducted to explore these discrepancies in detail.

To date, a wealth of research studies have included multiple QoL and outcome measures but have not examined these systematically. For example, it has been highlighted that people with dementia are able to rate their QoL but that this differs from relative proxy ratings [34], without any exploration of why this might be. Other studies have stated that proxy ratings improve feasibility and should be used when people with dementia are unable to ‘answer by themselves’ to avoid having missing data [6], although this is presented without clear cut-offs to guide researchers. Therefore, researchers should be encouraged to examine the psychometric properties of the measures used within their studies, to help understand which are most appropriate for use with people living with dementia in care homes, with different degrees of cognitive impairment.

Previously, it has been stated that proxy-completed measures are the only option for individuals living with moderate to severe dementia [6, 9]. However, this fails to value the perspective and the insights into the QoL of individuals of people with dementia that may not be picked up by staff members, relatives or friends. Additionally, for research findings and any resultant policy changes to be meaningful, appropriate and valid QoL data must be collected [10] including recognition that people with dementia are able to provide meaningful commentary around their own QoL [7]. Researchers should, therefore, explore creative ways to work with those who struggle to communicate verbally to collect meaningful data. The burden of data collection for people with dementia needs to be considered within this, as some participants were unable to complete measures in the present study due to tiredness or boredom. Flexibility in researcher approach has been highlighted as important, providing participants with the opportunity to complete measures through several conversations or over 2 days if required [35]. Building relationships with participants with dementia can help to identify the best time of day for data collection, which could help increase the feasibility of self-completed data [35]. Furthermore, people with dementia have been shown to consistently rate their QoL higher than proxy raters. It is unclear whether this relates to an inability to accurately assess their performance or abilities against measure items or whether in fact proxies under estimate QoL based on their own, different perceptions of what is important. For example, people with dementia living in a care home may compare themselves with others living in the setting and may judge their QoL to be good comparatively or they may have reduced expectations about their own performance given their personal circumstances or may simply have a more positive outlook [34]. It is also noted that people living with dementia may benefit from overestimating their QoL, as a strategy of self-maintenance [36]. Further research is thus needed to assess why people with dementia living in care home settings make particular judgements on QoL items and what this means for assumed ‘accuracy’ and interpretation of results.

More widely, there are concerns about the quality of QoL measures in general and the feasibility of their use with people with dementia. Most of the existing dementia-related QoL measures have had limited psychometric evaluation [37, 38]. For example, to date, the relationship between the QOL-AD NH and any health-related outcomes has not been examined (criterion validity). Furthermore, the QUALID has demonstrated poor criterion validity, in both studies that have examined this [39, 40]. However, whilst these issues are concerning, in part, these may be due to methodological issues of the studies examining the psychometric properties of measures rather than highlighting underlying problems with the measures [4].

Although multiple measures were collected in the present study, only one measure was completed by residents that proxy-reporters also completed (EQ-5D-5L). Therefore, self- and proxy-rating comparisons could not be drawn from QUALID and DEMQOL. In future, where possible, the same measures should be completed by residents and proxy-reporters, in order to be able to draw more in-depth comparisons. Individuals who lived in the care home and were cared for in bed or were formally admitted to an end of life care pathway were not eligible to participate in the present study. These individuals may have been expected to have the lowest QoL and therefore the present study may not fully capture the breadth of QoL experienced by those living in care homes. Additionally, in line with previous evidence [9], those who completed self-report measures are likely to have been in the earlier stages of dementia and therefore have higher QoL than those in the later stages, which were not accounted for in the present study. Researchers need to develop alternative strategies to ensure that the perspectives of those with later-stage dementia are captured [35]. Furthermore, we do not understand why participants provided the rating that they did for each item. Collecting additional qualitative data to explore this issue would improve understanding around why differences in ratings exist.

There are demographic characteristics (for example in the qualifications and experience of staff, and relationships of relative/friends to residents) that may have affected QoL ratings [36]. For example, spouses have been found to rate QoL in people living with dementia as higher than adult children [33]. We did not stratify our analysis to explore within-group variations in QoL ratings due to poor completion of demographic details by these participants. Understanding predictors of variability in QoL ratings within relative/friend and staff proxy groups constitutes a valuable should be an on-going focus for future research.

Future research should enhance recent reviews [37, 41] and conduct a meta-analysis of different QoL measures completed by different raters over time, in order to establish which are most meaningful and suitable for use. One recent narrative review concluded that self-report and proxy-report DEMQOL and EQ-5D-5L should be used [15]. However, within the 41 studies reviewed, only four used DEMQOL; therefore, this conclusion is based on limited evidence. We found that DEMQOL had good internal consistency in the present study and scores on this measure significantly correlated with five of the additional seven measures, although these correlations were all weak, except for staff proxy-reported QUALID. In addition, QUALID correlated with all (relative/friend proxy-reporters) and all but one (staff proxy-reporters) measures. Although it demonstrated weaker internal consistency than both QOL-AD and DEMQOL, our results suggest that DEMQOL proxy may offer the most thorough and comparable measure of QoL; however, we did not collect self-report DEMQOL and cannot make a definitive judgment without this. A care home-specific version of the DEMQOL has recently been developed, in line with other measures such as the QOL-AD NH, which may provide further utility for this measure within care homes [42].

In conclusion, measuring quality of life for people with dementia is complex and often involves multiple measures completed by multiple raters. Whilst it is acknowledged that self-report data are the optimal method of data collection, the limitations of this method are also widely reported, particularly as ability to complete measures is likely to decline over time for those with dementia. The low levels of agreement between relative/friend and staff proxy raters on these measures, however, bring into question the appropriateness of proxy-rated data within this population. Given that residents may overestimate their QoL, it is difficult for researchers to establish which measure provides the most reliable or valid report of individuals’ QoL. The lack of other viable alternatives at present means that researchers should be aware of these issues and interpret their data with caution. This study highlights the need for researchers and practitioners to better understand of the impact of rater choice on QoL outcomes. It is not possible to recommend staff or relative/friend proxy ratings as more or less accurate than self-ratings, as proxy rating may be biased by factors that unduly influence perceived QoL (such as self-care ability), whilst the same factors may have little impact on QoL as experienced by the participant or resident. Therefore, researchers need to give greater consideration of the influence of raters when selecting QoL outcome measures and their completion.