Convergent validity of EQ-5D with core outcomes in dementia: a systematic review

Table 3 lists the measures that were used to measure the core dementia outcomes in the studies. Details of the measures are provided in Additional file 5. In total there were 12 distinct measures: Two cognition measures (MMSE and ADAS-Cog), six measures of function (Katz ADL, ADCS-ADL, Barthel index, Lawton scale, DAD and BADLS) and four behaviour/mood measures (NPI, CSDD, GDS and CMAI). Where it was reported, the measures were completed either by proxy, researcher observation, or a combination of information such as self or proxy information and recent care records (administration details are provided in Additional file 5). The most predominant measure was the MMSE (n = 25), followed by the NPI (n = 12).

Table 3 Core dementia outcome measures

The MMSE and ADAS-Cog measures commonly capture cognitive impairment, however the latter is administered via direct observation, resulting in a longer administration duration. There two types of daily activities – basic activities of daily living (BADLs) and instrumental-ADLs. Instrumental ADLs are not necessary for fundamental functioning, they are generally more complex activities that allow a person to live independently, e.g., managing one’s own finances. Basic ADLs are fundamental skills, typically related to basic physical needs, e.g., toileting and eating [49]. Of the six function measures, two captured basic ADLs (BADL) alone (Katz ADL and Barthel index), one captured instrumental ADLs (IADL) alone (Lawton scale) and the remaining three included both BADL and IADL items (ADCS-ADL, DAD and BADLS). Of the behavioural measures, two measured depression (CSDD and GDS), while the others captured agitation (CMAI) and general neuropsychiatric symptoms (NPI).

EQ-5D convergent validity with cognition

It was hypothesised that cognition would have a positive correlation with EQ-5D whereby greater cognitive impairment would be associated with lower EQ-5D index scores (lower MMSE scores indicate greater cognitive impairment). Additional file 6 provides complete details of the empirical relationship between cognition and EQ-5D. In total, eighteen studies assessed the convergent validity between EQ-5D index scores and the cognitive measures (MMSE, n = 17; ADAS-Cog, n = 2; one study collected both measures [41]). Three studies reported a different relationship between cognition and EQ-5D by rater type [28, 35, 41].

Within only seven distinct studies a statistically significant relationship (p < 0.05) between cognition and EQ-5D was reported, all of which were positive correlations [28, 30, 35, 40, 41, 43, 46], and three of these seven studies had a sample size of greater than 300 [28, 35, 46]. Of the studies that reported statistically significant findings, the rater type was predominantly an informal caregiver (5/7 studies), and were studies that had included participants spanning the entire dementia severity range (mild-to-severe). The one study that reported a statistically significant association between self-reported EQ-5D and cognition was within a mild-stage study sample [41]. Figure 2 shows the proportion of studies that demonstrated a relationship between cognition and EQ-5D in both directions (see Additional file 6 for more details).

Fig. 2figure 2

Direction of convergent validity between EQ-5D and cognition. *Total number of studies is n = 19 as 2 studies report convergent validity twice

EQ-5D convergent validity with function

For the convergent validity between EQ-5D and the measures of function, a positive correlation was hypothesised whereby greater functional independence would be associated with higher EQ-5D index scores (see Additional file 5 for details of function instrument scoring). Twenty distinct studies provided empirical evidence of the convergent validity between EQ-5D and the measure of function within the study (ADCS-ADL, n = 2; BADLS, n = 2; Barthel index, n = 7; DAD, n = 6; Lawton scale, n = 4; one study collected both Barthel index and Lawton scale [41]).

Two studies reported a difference in relationship between function and EQ-5D by rater type, whereby the Lawton index showed a positive and significant correlation with proxy EQ-5D and a negative and non-significant correlation with self-rated EQ-5D [38, 41]. These two studies were the only reports of a negative association; both of these studies had a mainly mild-stage sample size of < 200 participants. The remaining studies all reported a positive correlation between function and EQ-5D, of which the majority (15/16 studies) were statistically significant. Two studies explored function as a dependent variable within regression analyses, both of which found it to be a significant (p < 0.01) determinant of proxy reported EQ-5D [40, 43], but not self-reported EQ-5D [40]. One study had reported a positive correlation between ADCS-ADL and EQ-5D for both rater types, but was only statistically significant for proxy-report [20], and was again within a mild-stage study sample [20]. Figure 3 shows the characteristics of the studies that reported convergent validity evidence between function and EQ-5D (Additional file 7 provides complete details).

Fig. 3figure 3

Characteristics of studies reporting convergent validity between EQ-5D and function. *Indicates non-significant correlation (p > 0.05). Y axis = sample size

EQ-5D convergent validity with behaviour/mood

For the behaviour/mood measures, higher scores indicate greater severity (see Additional file 5). Therefore, it was hypothesised that the measures would have negative correlations with EQ-5D, whereby more behavioural disturbance is associated with lower EQ-5D index scores. Seventeen distinct studies reported empirical evidence of the convergent validity between EQ-5D and the measure of behaviour/mood, four studies collected multiple measures [20, 28, 46, 47] (CSDD, n = 2; CMAI, n = 1; GDS, n = 8; NPI, n = 10).

Only one study captured agitation (via CMAI), reporting a negative correlation with EQ-5D which was only statistically significant for proxy-report [36].

Ten studies measured depression (via CSDD and GDS). All ten studies reported a negative correlation between the measure of depression and EQ-5D, whereby statistically significant results were found with self-rated EQ-5D only n = 4 [31, 41, 44, 47]; proxy-EQ-5D only, n = 2 [37, 46] and both rater types, n = 2 [20, 28]. One study did not report statistical significance, but rather strength of correlation coefficients – indicating moderate convergent validity between EQ-5D index scores and GDS [33].

The NPI captures 12 broad neuropsychiatric symptoms and was administered in ten of the reviewed studies. Two of these studies reported a difference in relationship between NPI and EQ-5D by rater type; one study found a negative correlation with self-rated EQ-5D, but a positive correlation with proxy-EQ-5D [42], while the other study found the inverse [39]. However, neither of these findings were statistically significant. The remaining studies all reported negative correlations, whereby statistical significance was found only with proxy-EQ-5D [20, 21, 25, 27, 28, 35, 46]. Figure 4 shows the characteristics of the studies that reported convergent validity evidence between the behaviour/mood measure and EQ-5D (Additional file 8 provides complete details).

Fig. 4figure 4

Characteristics of studies reporting convergent validity between EQ-5D and behaviour. *Indicates non-significant correlation (p > 0.05). Y axis = sample size

Convergent validity evidence by EQ-5D dimension

A total of seven distinct studies reported empirical evidence of convergent validity of the pre-defined core dementia outcome measures with EQ-5D dimensions – summarised in Table 4. Cognition (via MMSE) was associated with self-rated anxiety/depression [19], and people with more cognitive impairment self-reported fewer problems across all EQ-5D dimensions [42].

Table 4 Empirical evidence of relationships between outcome measures and EQ-5D dimensions

Function via the Katz index was associated with self-rated mobility, self-care, usual activities, and pain/discomfort; no relation was found with anxiety/depression [19]. Function via BADLS was correlated with proxy rated mobility, self-care and usual activities. Stronger correlations were observed for informal carer reports of self-care and usual activities, while the clinician rated mobility correlation was stronger [34]. Function via Barthel index was significantly correlated with self and proxy mobility, self-care and usual activities [42], and was associated with reporting problems in all EQ-5D dimensions minus anxiety/depression [23].

Depression (via CSDD) was associated with reporting problems in anxiety/depression [23]; and depression (via GDS) showed evidence of moderate convergent validity with mobility, self-care, usual activities and anxiety/depression [33]. NPI summary scores were associated with proxy rated anxiety/depression [34, 42] and mobility [42].

Inter-rater agreement

To further understand the potential impacts of rater-type upon EQ-5D assessment, information related to the inter-rater agreement was extracted and is summarised in Table 5.

Table 5 Evidence of inter-rater agreement

Nine studies, representing samples across the entire dementia-severity range, commented on the inter-rater agreement between self and proxy rated EQ-5D index scores [19, 28, 29, 32, 35,36,37,38, 41]. Proxy-EQ-5D index scores were found to be significantly lower than self-report [28, 29, 37, 38, 41] and had stronger correlations with clinical variables [20, 35, 36].

Of the EQ-5D dimensions, it was reported in two distinct studies that the mobility dimension had the strongest inter-rater agreement, produced at an acceptable level (kappa > 0.4) (versus the other dimensions) [19, 37]. One of the studies reported that agreement between formal and informal proxies was also highest for mobility (kappa = 0.61), and all other dimensions remained below the usually accepted level [19].

Agreement between reports of the usual activities dimension was the lowest, with self-report reflecting more optimistic reports [29, 32, 37, 41]. Agreement in the pain/discomfort dimension was low, whereby proxies rated more problems than PwD themselves [32, 37, 41]. Evidence of agreement in the anxiety/depression dimension was mixed; one study found that PwD self-rated this dimension more optimistically than proxies [41], while another study reported that this was the only dimension that PwD had self-rated more problems (than proxies) [32].

Quality appraisal

Of the 30 papers included within the review, 18 were of high quality, and the remaining 12 were considered to be of medium quality (see Additional file 4 for full quality appraisal).

留言 (0)

沒有登入
gif