The results of this study extend the development of the EOLD scales by demonstrating the “internal consistency” reliability of the measures in a second cohort, and showing for the first time, convergent validity with other instruments measuring similar constructs. These findings further establish the psychometric integrity of the EOLD scales and support their usefulness as outcome measures in advanced dementia research.
There are important similarities and differences between subjects analyzed in the CASCADE study and the original development cohort. The baseline sample sizes were comparable (CASCADE, N = 189 vs. original, N = 154). In both studies, most of the patients with advanced dementia were white; however, the CASCADE patient cohort was older (mean age, 84.8 vs. 81.2 y), and had a greater proportion of females (84% vs. 45%). All CASCADE patients resided in NHs, whereas subjects in the original study lived in either the community or an institution. Nonetheless, the functional disability of the 2 study cohorts was almost identical as measured by the mean BANS-S score (20.7 vs. 20.4). With respect to the dyad relationship, the HCPs in CASCADE were mostly offspring (69%), whereas family caregivers in the original study were predominately spouses (64%), and only 34% were offspring. Family caregivers in the original study were also more often female (75%) compared to the HCPs in CASCADE (61%).
There are also some notable differences between the data collection methods used in the 2 studies. In the original study, all measures were ascertained by a mail-in survey administered to caregivers (ie, family members) up to 1 year after the patient with dementia had died. In contrast, in the CASCADE study, the SWC-EOLD was administered to the HCPs in a telephone interview although the resident was still alive to ascertain satisfaction with care during the prior 90 days. Similarly, the SM-EOLD was obtained from a face-to-face nursing interview while the resident was alive to measure symptoms in the prior 90 days. In both studies the CAD-EOLD was obtained after death. However, in the CASCADE study it was collected from a nursing interview rather than the caregiver.
Despite the differences in subjects and data collection methods, the 3 EOLD scales demonstrated very good reliability in the CASCADE study sample, and these reliability estimates were similar to the estimates reported in the original study. This finding is important as it further establishes the generalizabilty of the psychometric properties of these scales in advanced dementia, particularly in the NH setting where approximately 70% of patients with advanced dementia die.22
As expected, the α coefficients were slightly lower in this validation cohort compared to the original developmental cohort for the SWC-EOLD (0.83 vs. 0.90), SM-EOLD (0.68 vs. 0.78), and the CAD-EOLD (0.82 vs. 0.85). Satisfaction with care, as measured by the average SWC-EOLD scores, was very similar in 2 studies (30.9 vs. 30.5). However, the mean SM-EOLD score in the CASCADE sample was much greater than in the original cohort (36.4 vs. 21.3), indicating more comfort. The mean CAD-EOLD score (33.6 vs. 31.4) was also slightly higher in CASCADE. Moreover, the distributions of the SM-EOLD and CAD-EOLD scales were skewed toward higher scores in CASCADE, whereas these scales approached a normal distribution in the original study. The reason for these differences is not clear. It is possible that persons dying with advanced dementia in the CASCADE study experienced greater comfort compared to those in the original study. Alternatively, it is possible that differences in how these measures were collected in the 2 studies influenced the reporting of symptoms and signs in this population.
This study has several limitations worthy of discussion. The CASCADE study consists of residents who are predominately white, as was true of the original study. It is not clear whether the reliability and validity of the scales generalize to people of other racial or ethnic backgrounds. Second, the analyses presented in this study are cross-sectional, as was true of the original study. The next step in establishing the psychometric properties of these scales, particularly the SWC-EOLD and SM-EOLD, would be to examine the response patterns over time. The repeated, longitudinal measures currently being collected in the on-going CASCADE study will provide an opportunity to conduct these analyses.
Measuring the quality of health care, particularly at the end-of-life, is a difficult, but critical step toward improving care. The findings of this study demonstrated that SWC-EOLD, SM-EOLD, and the CAD-EOLD scales are valid and reliable tools that measure key palliative care outcomes in advanced dementia. In particular, this study extends the usefulness of these scales to research conducted in the NH setting, and while the patients are actually receiving end-of-life care, rather than after they have died.