Critical evaluation of self-rated quality of life in mild cognitive impairment and Alzheimer's disease — Further evidence for the impact of anosognosia and global cognitive impairment

Abstract

Objectives

The present study investigates the effect of anosognosia (impaired insight for an illness) and cognitive deficits on the reliability and validity of self-rated Quality of Life (QoL) in Mild Cognitive Impairment (MCI) and Alzheimer's Disease (AD).

Design

Cross-sectional study.

Setting

Cross-sectional study with a consecutive clinical sample from a memory clinic in Leipzig (Germany).

Subjects

27 patients (aged 65 years or above) with a diagnosis of either MCI (N=12) or AD (N=15), each together with a caregiver.

Measurements

The patients' QoL was measured using the Dementia Quality of Life self and proxy ratings (DEMQoL and DEMQoLproxy). The degree of anosognosia was rated by means of the Clinical Insight Rating Scale (CIR). In addition the Mini-Mental-State Examination (MMSE), and for diagnostic purposes the Bayer Activities of Daily Living Scale (B-ADL) and the Consortium to Establish a Registry of Alzheimer's Disease (CERAD) word list were applied.

Results

In accordance with the results of Ready et al. (1), patients with impaired insight were found to produce less reliable QoL ratings than those with unimpaired insight. The validity (concordance between self- and proxy QoL ratings) is influenced by cognitive deficits, anosognosia and the interaction between these factors.

Conclusions

Data which are based on dementia patients' QoL self-ratings need to be interpreted with caution when anosognosia is present.

Quality of Life (QoL) data of AD patients are increasingly requested as outcome criteria by regulatory authorities, health economists and administrative agencies outside a clinical framework in the context of clinical research. The European Medicines Agency (EMEA), for example, deems the assessment of QoL important in judging the benefits of anti-dementia drugs (2). Thus a large amount of data intended to represent QoL is being generated. These data are referred to when judging effects of medication and other therapeutic interventions and can thus form the basis for far-reaching decisions. However, there is a lack of critical evaluation of the reliability and validity of these data, which applies especially to instruments which rely on subjective ratings of QoL (3).

The original concept of QoL in dementia derives from Lawton (4). On this basis Jonker and coworkers (5) developed a more recent concept, which differentiates hierarchically between objective circumstances on a first level and their subjective evaluations that confer meaning to facts on a second level. On the third and highest hierarchical level the model conceptualises QoL as experienced at a given moment as the result of a complex, multidimensional and essentially subjective process of appraisal. For the evaluation of personal objective living circumstances (which, in the case of dementia, include memory function and everyday functional abilities) environmental factors, such as social support and material conditions, and personal value systems and coping styles are equally relevant. Objectively poor living circumstances may be perceived as less negative than may be plausible from an outsider’s perspective („satisfaction paradox”; (6)). This phenomenon is a major problem with regard to proxy-rated QoL, and thus information about an essentially subjective construct such as patient QoL cannot be obtained from external sources (such as relatives) without difficulty.

However, although this framework by Jonker and coworkers (5) is related to dementia, AD poses specific problems for the assessment of QoL according to this framework. Memory deficits and impaired reasoning and judgment which cause a significant impairment in activities of daily living (ADL) are primary symptoms of AD (7, 8). These cognitive impairments may interfere with the perception of deficits and their evaluation. Thus, forgetting instances of difficulty with regard to ADL may lead to an inaccurate perception of the level of functional impairment. In addition, an impaired capacity for judgment and evaluation may affect a patients’ evaluation of the severity of his or her perceived difficulties and lead to underestimation of their significance. This may lead to values derived from QoL instruments that do not reflect a reliable and valid judgment of the kind that is intended. In addition to cognitive deficits, the perception and evaluation of QoL in AD may also be influenced by impaired insight for deficits (anosognosia). AD patients who are unaware of their cognitive or functional impairments are likely to be unable to perform the evaluation which their QoL is supposed to be based on, since they possess no sense of a significant impairment that could be evaluated.

Therefore we expect that patients with pronounced anosognosia and significant cognitive impairment will produce less reliable and valid data when rating their QoL than patients with intact insight and less cognitive impairment. The results of a recent study by Ready et al. (1) support this hypothesis, as they demonstrate a significant effect of anosognosia and the degree of cognitive impairment on the reliability and validity of self-rated QoL.

Patients and Methods

Subjects

Twenty seven patients, aged 65 or above, diagnosed with Alzheimer’s disease (AD; n=15) or mild cognitive impairment (MCI; n=12) participated in the study. All patients were required to have a healthy caregiver who was able to give information concerning the patient’s QoL. All participants were recruited from the memory clinic at the University Clinic, Leipzig.

The following exclusion criteria were specified:

• -Another neurological illness
• -Current alcohol or drug abuse
• -Significant further psychiatric diagnosis based on ICD-10

Thirteen patients were interviewed and tested in a clinical setting while the remainder were examined at their homes. Cognitive status was assessed by means of the Mini-Mental State Examination (9) (patients were included if they scored at least 15 and no more than 27 points). AD was diagnosed according to the ICD-10 criteria (7, 8) while a diagnosis of MCI was based on the criteria by Peterson et al (10).

These require:

• -Subjective memory impairment
• -Objective memory impairment (memory performance at least one standard deviation below the mean in an objective test)
• -No impairment in activities of daily living (ADL)
• -No dementia

Memory performance was assessed using the Consortium to Establish a Registry of Alzheimer’s Disease (CERAD) word list (11) which is a standard test in memory clinics. ADL function was rated by means of the Bayer Activities of Daily Living Scale (B-ADL) (12) using both self and proxy ratings. Patients diagnosed with MCI showed no evidence for functional impairment from either self or proxy ratings.

The study design does not include a healthy control sample, as the assessment of insight into deficits presupposes the presence of deficits which by definition are not present in a healthy control sample.

Measures

Dementia Quality of Life

The Dementia Quality of Life (DEMQoL) (13) consists of a patient version (DEMQoL) with 28 items and one item for a global QoL rating, and a caregiver version (DEMQoLproxy) with 31 items and one global QoL item. This instrument is generally referred to by the acronym DEMQoL because there is another questionnaire with the same name, the Dementia Quality of Life Questionnaire (DQoL) (14). This latter instrument was used in the above-mentioned study by Ready et al. (1) and comprises five subscales (aesthetics, self-esteem, positive affect, negative affect, feelings of belonging), as opposed to the DEMQoL which covers QoL in three domains (emotions, memory and ADL).

Most items of the DEMQoL and DEMQoLproxy are similar, but each version also comprises items which are not part of the parallel version. The English original version is one of the QoL instruments which have been most thoroughly developed and validated with an emphasis on psychometric properties. There are also convincing data for the validation of the German translation which was used for the present study (15).

Patients and proxies were interviewed separately about the patients’ QoL. Similar to the procedure used by Ready et al. (1) the proxies’ ratings were substitutes for the patients’ self-ratings, i.e. proxies were instructed to rate the QoL items as they thought the patients would rate them themselves.

Clinical Insight Rating

The Clinical Insight Rating (CIR) (16) is a clinical interview for assessing the degree of anosognosia in dementia. Four aspects of insight (awareness of the reason for the visit to the clinic, awareness of memory deficits, awareness of functional deficits and awareness of the progression of the illness) are each rated by an expert on a three-point-scale (0–2) and are summed up to yield an overall score between 0 (complete insight) and 8 (no insight). In the present study the overall scores were inverted for evaluation purposes as in the study by Ready et al (1), thus higher values correspond to better insight (e.g. 8 = complete insight). The CIR has been shown to be reliable (16).

The CIR was conducted during the first screening visit at the memory clinic in order to obtain an impression of the patients’ degree of awareness before this could be influenced by the study procedure.

Data analyses

In order to obtain comparable data the analysis was based on the procedure employed by Ready et al (1). The reliability of the QoL data was assessed by means of their internal consistency (Cronbach’s alpha). In order to evaluate the effect of anosognosia on the reliability of the QoL data internal consistency reliabilities were calculated separately for patients with relatively intact insight into their illness (CIR = 7-8; n = 14) and for patients with impaired insight (CIR = 0-6; n = 13).

The sample was thus divided into two subsamples of comparable size, separating patients with intact or only slightly impaired insight from those with impaired insight. Indices for internal consistency reliabilities were assigned to categories according to the criteria of Cicchetti (17) for their clinical significance: unacceptable (below 0.70), fair (0.70-0.79), good (0.80-0.89) or excellent (0.90 and above) and were subsequently compared between subsamples.

Even though the concordance between self and proxy ratings of a patient’s QoL is no perfect measure for the validity of the self rating (see above) it was used in the current study, as the proxy rating is the best possible approximation of an objective measure, and a gold standard for the objective validation of self ratings does not and will probably never exist.

Moderated multiple regression analysis was used to determine the effects of general cognitive impairment, anosognosia and their interaction of the concordance between self and proxy ratings. All analysis were conducted on the complete sample comprising MCI and AD patients, as the primary focus was on anosognosia, rather than on effects of the diagnosis.

Impaired insight is found in both MCI and AD patients and the degree of insight can vary considerably between individuals given either diagnosis. Although according to the majority of studies on this subject the degree of anosognosia is related to dementia severity, insight can already be impaired in early dementia stages. Conversely, a significant level of insight can be preserved in later stages of the disease (for a review, see (18)).

Statistical power was also increased by running the analysis on the complete sample.

Results

Patients’and Informants Characteristics

The patients’ mean age was 76.4 years (SD = 5.2), and they had had an average of 12.4 years of education (SD = 3.3). 18 were females and 9 were males. The patients’ mean MMSE score was 22.7 (SD 3.5) out of a maximum of 30 points. The caregivers’ mean age was 67.3 years (SD = 9.4), and they had had an average of 12.3 years of education (SD = 1.8) auf. 24 were females and 3 were males. 19 caregivers were the patients’ spouses, one was the patient’s partner, 6 were patients’ children, and one was the patient’s sibling. Eight caregivers did not live together with the respective patients.

Impact of Anosognosia and Cognitive Impairment on Internal Consistency Reliabilities

Results show that there were no unacceptable values among the reliabilities for patients with relatively intact insight, while for patients with impaired insight two out of three values were unacceptable (see Table 1). The internal consistency rehabilites for patients with impaired insight have a lower clinical significance in all three DEMQoL subscales. The fact that even the largest differences, found for the memory and ADL DEMQoL subscales, are not statistically significant (DEMQoL-Memory: F = 1.99, p = 0.26; DEMQoL-ADL: F = 1.88, p = 0.22) is probably due to the small sample size and low statistical power (19).

Table 1.

Internal Consistency Reliabilities for Patients with Intact and Impaired Insight for current study and Ready et al. (1)

Current study (N= 27)
	Intact Insight	Impaired Insight
DEMQoL (Subscales)
Patient report
Emotions	0.88 good	0.79 fair
Memory	0.83 good	0.67 unacceptable
ADL	0.78 fair	0.58 unacceptable
Informant report
Emotions	0.89 good	0.82 good
Memory	0.84 good	0.57 unacceptable
ADL	0.86 good	0.83 good
Ready et al. (1) (N = 67)
	Intact Insight	Impaired Insight
DQoL (Subscales)
Patient report
Aesthetics	0.70 fair	0.55 unacceptable
Self-esteem	0.75 fair	0.65 unacceptable
Positive Affecta	0.85 good	0.61 unacceptable
Negative Affecta	0.88 good	0.76 fair
Feelings of belonging	0.58 unacceptable	0.63 unacceptable
Informant report
Aesthetics	0.74 fair	0.69 unacceptable
Self-esteem	0.76 fair	0.83 good
Positive Affect	0.84 good	0.89 good
Negative Affect	0.85 good	0.90 excellent
Feelings of belonging	0.77 fair	0.66 unacceptable

Note: DEMQoL = Dementia Quality of Life instrument. ADL = Activities of Daily Living. DQoL = Dementia Quality of Life instrument, a. Alpha significantly higher for high versus low insight

For comparison internal consistency reliabilities were also calculated for the caregivers in both groups. For caregivers of patients with impaired insight there was one unacceptable value, while there were no other differences in the clinical significance of the internal consistency reliabilities (see Table 1).

Overall the pattern found for clinical significances of internal consistency reliabilities in our data corresponds to the results of Ready et al. (1) (N = 67; see Table 1). These authors reported a higher clinical significance of reliabilities in four out of five subscales for patients with intact insight as compared to those with impaired insight, which was statistically significant for two subscales (see Table 1).

Ready et al. (1) reported another interesting pattern in their data. An explorative analysis based on a significant correlation between MMSE and CIR (r = -0.63, p < .01) indicated that insight into deficits may have a stronger effect on internal consistency reliabilities than global cognitive impairment.

As a statistically significant correlation between MMSE and CIR was also present in our sample (r = -0.57, p < .01), with a similar correlation coefficient to the one reported by Ready et al. (1), the effect of global cognitive impairment on the reliability of the QoL data was examined more closely. The sample was first divided into subsamples with higher scores (23 and above) and lower scores (22 and below). These cut-off values were based on a median split of the sample. Internal consistency reliabilities were then calculated for the three DEMQoL subscales in both subsamples. There was no evidence for lower internal consistency reliabilies among patients with lower MMSE scores. However, a small trend for lower internal consistency reliabilities was present among the caregiver of these patients (see Table 2).

Table 2.

Internal Consistency Reliabilities for Patients with Higher MMSE vs. Lower MMSE

	Higher MMSE	Lower MMSE
DEMQoL (Subscales)
Patient report
Emotions	0.85 good	0.79 fair
Memory	0.79 fair	0.82 good
ADL	0.70 fair	0.76 fair
Informant report
Emotions	0.89 good	0.76 fair
Memory	0.82 good	0.56 unacceptable
ADL	0.89 good	0.70 fair

Note: DEMQoL = Dementia Quality of Life instrument. ADL = Activities of Daily Living. MMSE = Mini-Mental State Examination

Impact of Anosognosia and Global Cognitive Impairment on the Concordance between patients’and informants’QoL ratings

In order to investigate the effects of cognitive impairment and of anosognosia on the concordance between self and proxy ratings of patients’ QoL regression analyses were conducted with proxy ratings (DEMQoLproxy) as the dependent variable and and patient ratings (DEMQoL) as the independent variable. The standardized residuals were then used as a measure of concordance. Positive residuals indicate that caregiver rated QoL higher than patients on a given subscale, while negative residuals indicate the opposite. Coefficients of determination were calculated as a standardised measure of concordance respectively the strength of the relation between self and proxy ratings for the DEMQoL subscales (DEMQol-Emotions: r² = 0.04, DEMQoL-Memory: r² = 0.02, DEMQoL-ADL: r² = 0.09, DEMQoL-Global QoL: r² = 0.02). The degree of concordance was not significant for any of the subscales.

Subsequently moderated multiple regression analyses were conducted with the standardized residuals as dependent variables. MMSE, CIR and their interaction (MMSE x CIR) were entered stepwise as predictors of the concordance between self and proxy ratings.

MMSE was entered first in order to investigate the effect of global cognitive impairment on the degree of concordance, followed by CIR and MMSE x CIR. For each added predictor the change in the amount of variance explained (for the dependent variable, i.e. the standardized residuals) was checked for statistical significance, as indicated by R² (significance of change).

Results show a trend (p<.10) for CIR as a predictor of concordance for the DEMQoL-Memory subscale (see Table 3). Higher CIR values were associated with more positive residuals (i.e. caregivers reported a higher QoL then the patients reported themselves).

Table 3.

Moderated Multiple Regression Analyses Predicting Patient-Caregiver Agreement: Impact of Cognitive Impairment and Anosognosia

	Independent Variables
	MMSE	CIR	MMSE x CIR
DEMQoL (Subscales)	R2	ΔR2	ΔR2
Emotions	0.01	0.03	0.04
Memory	0.06	0.19*	0.20
ADL	0.12*	0.17	0.35**
Global QOL	0.01	0.02	0.05

DEMQoL = Dementia Quality of Life instrument. ADL = Activities of Daily Living. MMSE = Mini-Mental-State Examination; CIR = Clinical Insight Rating Scale; QOL = Quality of Life. *p<.10. **p<.05.

There was a similar trend for the MMSE as a predictor of concordance for the DEMQoL-ADL subscale (p<.10) after the MMSE had been added to the analysis (see Table 3). Higher MMSE scores were associated with more positive residuals (i.e. caregivers reported a higher QoL than did the patients themselves). In addition the interaction between MMSE and CIR was a significant predictor of concordance for the DEMQoL-ADL subscale (p<.05) (see Table 3). Higher MMSE and CIR scores were associated with more positive residuals (i.e. caregivers reported a higher QoL than did the patients themselves).

Overall 25% of all possible values in this analyse (3 out of 12) were significant or marginally significant (see Table 3). By comparison, in the study by Ready et al. (1) 39 % of all possible values (7 out of 18) were significant or showed a trend towards significance.

Discussion

Patients with impaired insight produced less reliable QoL self-ratings than did those with intact insight into their deficits. This pattern of results supports our hypothesis and is similar to the findings of Ready et al. (1), although these authors used a different QoL instrument on a larger sample.

Despite a high correlation between anosognosia and cognitive impairment, anosognosia appears to be the decisive factor with regard to the reliability of the data, while cognitive impairment has no effect on reliability. This had been a hypothesis based on the data of Ready et al. (1) and is confirmed by our own data.

The analyses concerned with the validity of the data yielded different results, as both anosognosia and cognitive impairment as well as the interaction between both factors had effects on the concordance between QoL self and proxy ratings for the memory and ADL subscales of DEMQoL. At the same time, concordance between self and proxy ratings was low not only for these two subscales (2% for DEMQoL-Memory and 9% for DEMQoL-ADL), but also for DEMQoL-Emotion (4%) and DEMQoL-Global QoL (2%). For the latter two subscales, however, the large degree of variance or lack of concordance between self and proxy ratings was not explained by any of the predictors (anosognosia, cognitive impairment or the interaction between these two factors).

For the memory subscale intact insight was associated with higher QoL ratings by caregivers than by patients. Thus if patients are aware of their memory deficit this is a greater strain for them than caregivers assume.

For the DEMQoL-ADL subscale higher MMSE scores were associated with higher QoL ratings by caregivers than by patients. Thus if patients have less cognitive impairment caregivers assume that patients are less troubled by their functional deficits than they are.

The fact that for the DEMQoL-ADL subscale the interaction between cognitive impairment and insight was also a significant predictor for concordance adds support to this interpretation. If patients have high MMSE scores as well as a high degree of awareness of their deficits, they suffer more from their functional deficits than an observer would expect.

Overall our pattern of results regarding the impact of anosognosia and cognitive impairment on the validity of QoL data supports our hypothesis as well. Though it is slightly less pronounced than was the case in the study of Ready et al. (1), the pattern is nevertheless present.

Given these results it thus appears justified to use dementia patients’ QoL self ratings as an outcome criterion. However, it will be necessary to develop guidelines that specify conditions under which dementia patients’ self-rated QoL can safely be used or needs to be interpreted with caution. For example, it might be necessary to restrict the use of self-rated QoL data as an outcome criterion for therapeutic interventions to subgroups of dementia patients who demonstrate intact insight.

Acknowledgment: This paper was supported by the Interdisciplinary Center for Clinical Research (Interdisziplinäres Zentrum für Klinische Forschung, IZKF) at the Medical Faculty, University of Leipzig.

Financial disclosure: None of the authors had any financial interest or support for this paper.