| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
Correspondence: Address correspondence to Dr. Richard Trigg, Division of Psychology, Nottingham Trent University, Nottingham NG1 4BU, England. E-mail: Richard.trigg{at}ntu.ac.uk
 |
Abstract |
|---|
 TOP Abstract MethodResultsDiscussionReferences |
|---|
Key Words: Psychometrics • Self-report • Alzheimer's disease • Outcome assessment • Well-being
Measuring QoL in Dementia
In dementia research and practice there has been a long-unchallenged assumption that people with dementia are unable to give a reliable account of their own QoL (Cotrell & Schulz, 1993). Measurement has therefore focused on observable aspects of QoL, such as symptom severity and function, with only fleeting reference to the person's subjective perceptions (Stewart, Sherbourne, & Brod, 1996). Consequently, many QoL measures risk being little more than health status assessments that replicate much of the information that can be obtained from many other disability and disease burden measures (Gill & Feinstein, 1994; Leplege & Hunt, 1997). There is also the danger that by conceptualizing QoL for those with dementia in terms of negative constructs such as loss and disability, QoL research focuses on minimizing negative outcomes to the exclusion of maximizing the potential for positive life experiences.
The medical model of QoL places most emphasis on functional capacity, stressing the ability to perform everyday tasks and fulfill premorbid social and occupational roles (Leplege & Hunt, 1997). Implicit within the medical model is the notion that there is an optimum level of functioning to which all people should aspire, whereby those who are impaired or disabled have, by definition, a poorer QoL. This leads to the questionable assumption that one cannot achieve positive QoL in the presence of physical deficits. Similarly, although health is undoubtedly an important component of QoL, the two terms should remain distinct (Hunt, 1997), as evidence has suggested that it is possible for people in poor health to report a good QoL (Albrecht & Devleiger, 1999). In developing QoL measures, experts must take care to differentiate between causal variables (e.g., symptoms and reduced function) that affect QoL and outcome variables that reflect QoL (Fayers & Machin, 2000). By weighting QoL assessments heavily toward the presence or absence of symptoms like memory loss and disability, researchers exclude the possibility that some people may be able to adapt or adjust to their health problems, so maintaining or even improving their QoL (Ettema, Droes, de Lange, Mellenbergh, & Ribbe, 2005). Indeed, evidence has indicated that there is little or no association between QoL and severity of cognitive impairment for people with mild- to moderate-stage dementia (Hoe, Katona, Roch, & Livingston, 2005; Logsdon, Gibbons, McCurry, & Teri, 1999; Ready, Ott, Grace, & Fernandez, 2002; Thorgrimsen et al., 2003). Just because an individual's cognition worsens, experts cannot assume that this inevitably leads to a worsening of QoL.
Some have argued that QoL can only be adequately measured by determining the opinions of patients, rather than relying on the views of "experts" (e.g., clinicians or carers; Gill & Feinstein, 1994), which may be reliable for observable aspects of behavior and ability but are unsuitable for inferring subjective experience and feelings (Berkowitz, Du, Kazis, & Lewis, 1995). Recent research has suggested that with careful attention to the wording, structure, and format of questions, obtaining self-reported QoL is possible using standardized measures (Brod, Stewart, Sands, & Walton, 1999; Logsdon, Gibbons, McCurry, & Teri, 2002). Moreover, it provides consistent and reliable information regarding QoL issues (Feinburg & Whitlatch, 2001; Mozley et al., 1999).
Reviews of QoL measures that could be used in dementia have highlighted a variety of suitable tools (Ettema et al., 2005; Ready & Ott, 2003), but few tools have been developed specifically for dementia, and even fewer are designed to gather information directly from the person with the condition. The Dementia Quality of Life Instrument (DQoL; Brod et al., 1999) is a measure designed solely for patient administration; it assesses a person's sense of well-being and aesthetics. It excludes common indicators of QoL such as social interaction, mobility, and activity performance with the aim of minimizing respondent burden and because these domains are seen as more suited to objective assessment. The DQoL therefore focuses primarily on feeling states and mood. However, Jennings (1999) suggested that QoL judgements based on feelings such as happiness are too simplistic; it is not plausible to say that a person's QoL is good just because pleasurable sensations outnumber the unpleasant ones. An evaluation of life areas such as satisfaction is central to the concept and assessment of QoL (Lawton, 1997). The Quality of Life–Alzheimer's Disease scale (QoL-AD; Logsdon et al., 1999) is a brief measure designed for both patient and proxy completion, with respondents asked to rate general life areas such as physical health, energy, mood, living situation, memory, family, marriage, friends, chores, fun, money, self, and life as a whole on a scale from poor to excellent. However, one might argue that this approach differs from an evaluation of satisfaction in that it does not allow for the possibility that the person may rate his or her performance in a particular life domain as poor but report higher levels of satisfaction due to a low perceived importance of that domain or to the successful use of coping and adaptation.
There is therefore a need to develop a new measure of QoL for use by people with mild- to moderate-stage dementia that subjectively assesses a range of QoL domains particularly relevant to this population (including positive and negative feeling states) with judgement or evaluation of various life areas. The measure should avoid inclusion of causal variables such as symptom severity that do not necessarily reflect QoL status per se and should promote the assessment of positive constructs, such as feelings of self-worth and life satisfaction.
Development of the Bath Assessment of Subjective Quality of Life in Dementia (BASQID)
We developed the BASQID from the perspective of the person with dementia, and this philosophy remained central to each stage of development. These stages had previously included the qualitative development of a conceptual framework, generated from in-depth interviews, with 30 people with mild- to moderate-stage dementia, exploring in their own language and ideas those issues that were relevant and important to their QoL and the ways in which dementia impacted these areas. Furthermore, structured interviews with people with dementia examined understanding, and the relative importance, of the emergent QoL domains and provided an opportunity to explore the potential format of questions and response choices. Results indicated that presenting questions and response scales both visually and orally may be beneficial in helping the respondent attend to questions and respond to choices. An initial pool of 44 items was produced based on the domain and facet structure from the conceptual framework.
The conceptual framework and details of item reduction are presented elsewhere (Trigg, Jones, & Skevington, 2007). This article reports on the psychometric properties of the final BASQID measure, including the relationship of BASQID scores to other clinical indicators such as cognition.
|
Method |
|---|
|
TOPAbstractMethodResultsDiscussionReferences |
|---|
A second field test (Stage 2) followed, when a shortened version of the BASQID (21 items) was administered with the Mini-Mental State Examination (MMSE; Folstein, Folstein, & McHugh, 1975). We used this information to produce the final BASQID instrument and explore the contended relationship between QoL and cognition.
Nested within the Stage 2 sample was a modest longitudinal study of 36 participants who were beginning treatment on one of the three available acetylcholinesterase inhibitors for mild–moderate Alzheimer's disease (AD; Sample 2a). They were tested twice, at 3-month intervals, to provide information on validity and responsiveness of the BASQID scores to changes in clinical condition.
Participants
The Stage 1 field test included 60 people with dementia (Sample 1), as 60 was calculated as being sufficient to detect significant correlations (>0.3) between questionnaire items and between measures for the purpose of construct validity testing. We reassessed 30 of these 2 weeks later and used the scores to examine test–retest reliability, as 30 would allow for the detection of significant correlations (>0.7) between the two administrations of the measure, albeit with large standard errors. For Sample 2, we recruited 150 patients, as this was sufficient to allow for detailed analysis of the structure and scaling properties of the BASQID using factor analytic techniques. MacCallum, Widman, Zhang, and Hong (1999) indicated that between 100 and 200 cases is adequate when there are few factors and when variables have communalities (common variance) >0.5. Nested within Sample 2 was a subgroup of 36 participants who were beginning treatment on one of the three available acetylcholinesterase inhibitors for mild–moderate AD. They were assessed on the first occasion that these drugs were prescribed, and then again 3 months later.
Participants were consecutive admissions to a memory clinic during a 15-month period who satisfied the inclusion criteria of (a) a positive diagnosis of dementia according to the Diagnostic and Statistical Manual of Mental Disorders (4th ed.; American Psychiatric Association, 1994) and (b) an MMSE score of 12 or above, signifying a mild to moderate stage (National Institute for Clinical Excellence, 2001). Previous research has shown this to be the level at which patients are able to directly report on QoL issues (Logsdon et al., 1999; Mozley et al., 1999). Exceptionally, some patients with scores below 12 were recruited if the clinical staff judged that they might be able to complete the assessment and if their low MMSE score was due to poor sight or hearing. Participants were excluded if they had already participated in the BASQID development or did not have English as their first language. In Stage 1 we requested the participation of a spouse or caregiver of the person (where available) to complete proxy assessments on behalf of the patient for validation purposes.
Materials
BASQID
The number of items contained within the measure was different for the two field tests: 44 items for Stage 1, in which the 21 items tested at Stage 2 were embedded. The items were administered via interview, with the interviewer presenting each question visually and orally to the person with dementia. Each item is printed on an individual card in large sans serif font. Response scales are printed on individual cards (same font size) and are set out horizontally, with vertical lines separating the five scalar points. The BASQID uses two response scales: not at all satisfied, a little satisfied, satisfied, very satisfied, extremely satisfied; and not at all, a little, a moderate amount, quite a lot, a great deal. Only words define each point on the scale, not the scores associated with each response. All items are scored 0 to 4, so that low scores indicate poor QoL.
GSS-15
The 15-item GDS (Sheikh & Yesavage, 1986) has demonstrated acceptable reliability (
=.81) as a screening instrument for depression. It has advantages over shorter versions in that researchers can use the total score as a measure of the severity of the depressive episode (Almeida & Almeida, 1999).
WHOQOL-BREF
The WHOQOL-BREF is a 26-item short form version of the WHOQOL-100 (WHOQOL Group, 1998b). It is a generic assessment of QoL across four domains: physical health, psychological, social relationships, and environment (Skevington, Lofty, & O'Connell, 2004). The WHOQOL-BREF demonstrated internal consistency reliability, with Cronbach's alpha values between.66 and.88 for the four domains, in an international sample of well and sick people. The WHOQOL-BREF was comparable to the WHOQOL-100 in terms of discriminating between the QoL of ill and well respondents.
MMSE
The MMSE (Folstein et al., 1975) is a widely used measure of cognitive function. The MMSE has a maximum score of 30 points. It assesses aspects of orientation to time and place, registration, attention, calculation, recall, language, and visual construction. The MMSE scores validly discriminate between people with dementia, depression, or cognitive impairment with depression (Folstein et al., 1975). Test–retest reliability was measured at 0.89, with interrater reliability at 0.83. Researchers have reported validation of the measure (Tombaugh & McIntyre, 1992), however we must also note that some have criticized the MMSE for its dependence on education and age (Tombaugh & McIntyre, 1992) and on race, ethnicity, and language (Ramirez, Teresi, Holmes, Gurland, & Lantigua, 2006).
Clinician Global Change Rating (CGCR) Forms
A Clinician Global Change Rating (CGCR) of QoL was completed for each participant in Sample 2a as part of the assessment of responsiveness. Responsiveness testing requires the comparison of measure changes to an accepted indication of change as the external standard (Husted, Cook, Farewell, & Gladman, 2000). However, for QoL measurement no such gold-standard measure exists. Terwee, Dekker, Wiersinga, Prummel, and Bossuyt (2003) suggested that in such cases, global ratings of change by clinicians might be appropriate. Although it is admittedly an imperfect measure, because a physician saw all participants, we saw this as the most appropriate form of proxy QoL rating in the current study. The CGCR consisted of a transitional question in which the physician used a 5-point scale to rate whether the participant's QoL was a lot worse, a little worse, the same, a little better or a lot better compared to 3 months previously.
Procedure
In Stage 1, interviews took place in the participants' homes or at the memory clinic, depending on the preference of patient and carer. Both parties provided informed consent. The carer completed the WHOQOL-BREF on behalf of the patient. The BASQID and then the GDS-15 were administered via interview to the patient. Interviewers revisited 30 respondents 2 weeks after the initial administration, when patients completed the BASQID alone to provide data for calculation of test–retest reliability.
All assessments in Stage 2 took place in the memory clinic as part of the patients' scheduled visit. Routine cognitive assessments and evaluations completed by memory clinic staff included administration of the MMSE. The researcher was then introduced to the patient and explained the purpose of the QoL assessment. Informed consent was obtained from the participant, and the BASQID administered.
For participants in Sample 2a, the procedure was completed twice. The first assessment was prior to the participant being prescribed an acetylcholinesterase inhibitor for the treatment of AD. The second assessment was at their 3-month follow-up appointment, where the physician supervising the prescription of the intervention evaluated the effectiveness of the drug using the CGCR form.
Statistical Analysis
We revisited data from Stage 1 once the final BASQID scales had been produced as part of construct validity and test–retest reliability analysis. We examined the relationship between BASQID scores, the GDS, and the four subscales of the WHOQOL-BREF by using Pearson correlation coefficients. We used intraclass correlation coefficients to explore the temporal stability of the BASQID.
We used data from Stage 2 to develop the final BASQID instrument, retaining only the items and subscales that demonstrated sound psychometric properties. We explored the dimensionality of BASQID items by using principal components analysis (PCA) with a Varimax, orthogonal rotation. In an iterative process, we retained factors if eigenvalues exceeded 1.0. Items that failed to load at least 0.4 on any factor or that cross-loaded on more than one factor (at above 0.4) were considered for possible rejection from the measure. Once items had been eliminated, we reran the best fitting model. We note that these analyses are exploratory and that future analyses should examine other models (e.g., a bifactor [Gibbons et al., 2007] or a random intercept model [Maydeu-Olivares & Coffman, 2006]). The latter model is appropriate when item wording may affect item response, and the bifactor model is informative with respect to the presence of a general factor in addition to two subfactors that capture residual covariation.
We examined internal consistency of the resultant scales by using Cronbach's coefficient alpha. We conducted secondary analysis to investigate the effect of participant cognitive status on internal consistency. We divided Sample 2 into tertiles according to MMSE score (<16, 16–20, >20) and recalculated Cronbach's alpha for the BASQID scales within each tertile (Logsdon et al., 2002).
We assessed responsiveness of the scales through calculation of effect size statistics (change score for the instrument divided by the standard deviation of the baseline scores; Kazis, Anderson, & Meenan, 1989). We compared these effect sizes to the CGCR to assess whether the BASQID was able to detect clinically important observed changes in patients' QoL. Spearman's correlations examined agreement between the effect sizes recorded on the BASQID, and changes in QoL as rated on the CGCR form. Secondary analysis involved (a) splitting the sample into three groups according to clinician judgements about improved QoL, worse QoL, or the same QoL compared to 3 months ago; and (b) comparing effect sizes across the groups.
We explored the construct validity of the BASQID thorough examination of Pearson's correlations between BASQID scores and participant cognitive status, age, GDS-15, and the carers' proxy judgement using the WHOQOL-BREF. An independent samples t test explored differences in BASQID scores for men and women. We hypothesized that BASQID scores would show low to moderate significant correlations with proxy ratings of QoL on the WHOQOL-BREF. As there is a strong relationship between self-reported QoL and mood (Logsdon et al., 2002; Thorgrimsen et al., 2003), McDowell and Newell (1996) suggested that the correlation between QoL and depression may be as high as 0.6, so we expected a moderate correlation between BASQID scales and GDS-15. In line with results from other self-report QoL instruments, we did not expect significant correlations between BASQID scales and participant MMSE scores (Logsdon et al., 1999; Ready et al., 2002; Thorgrimsen et al., 2003) or between BASQID scores and participant age or gender. We used one-way analysis of variance to explore differences in BASQID score across age and MMSE tertiles.
|
Results |
|---|
|
TOPAbstractMethodResultsDiscussionReferences |
|---|
|
The item reduction procedure is published elsewhere (Trigg et al., 2007). The analysis in this article was conducted on the 14 extracted items that were found to be the "best" items out of 44 in psychometric terms and that form the final measure.
BASQID Scales
The BASQID measure contains 14 items. In Stage 1, the level of missing data for these items was minimal, with any one question having no more than three missing values (5%). For Stage 2 the level of missing data within the 14 items was no more than 4.66% (7/150). We employed listwise deletion of cases where there were missing data in the variables used for analyses.
One derives a total score for the 14-item scale (BASQID), from 0 to 100, by multiplying the sum score by 100/[m x (k – 1)], where m represents the number of items in the scale and k represents the number of response choices. The distribution of scores on the BASQID had a mean of 61.47 (SD = 14.25, range = 23.21–89.29).
We used PCA to explore the dimensionality of the 14 items using data (n = 143) from Stage 2. The Kaiser-Meyer-Olkin statistic had a value of.90, indicating an excellent level of sampling adequacy. Bartlett's test of sphericity was significant, indicating that PCA was appropriate. Both the Kaiser criterion of retaining factors with an eigenvalue of greater than one and the scree test (see Figure 1) suggested that the items within the BASQID may be organized according to two orthogonal dimensions. Although the test of scree suggested an essentially unidimensional result (a ratio of 4:1 of the first to the second eigenvalue, with the second only slightly over 1), we can not rule out the presence of two components without further confirmatory analyses with a cross-validation sample and other models. The content of the items within each dimension suggested that Component 1 (eigenvalue = 5.90) represented a Life Satisfaction (LS) subscale and Component 2 (eigenvalue = 1.36) represented a Feelings of Positive QoL (FPQ) subscale. Table 2 displays the content of these items and loadings on the two components.
|
|
Reliability
Internal Consistency
We calculated the internal consistency of the BASQID by using data from Stage 2. Cronbach's alpha for the BASQID was.89, and alphas for the LS and FPQ subscales were.84 and.83, respectively (n = 143), indicating acceptably high internal consistency. When we divided Sample 2 into tertiles on the basis of MMSE score, internal consistency was acceptable for each of the three groups, with the lowest at.78 for the LS scale in the MMSE <16 group (n = 47).
Reproducibility
Using Sample 1 data, test–retest reliability for the BASQID (n = 29) demonstrated a nonsignificant difference between the two administrations (Time 1 – Time 2) of 0.65 (SD = 3.89, t = –0.85, p >.05), with an intraclass correlation coefficient of 0.85 (95% confidence interval [CI] = 0.70–0.93). For the LS subscale there was a nonsignificant difference in mean scores between the two administrations of –2.58 (SD = 8.43, t = 1.65, p >.05), with an intraclass correlation coefficient of 0.79 (95% CI = 0.59–0.89). Difference in mean scores for the FPQ subscale was –0.30 (SD = 8.48, t = 0.18, p >.05), with an intraclass correlation coefficient of 0.85 (95% CI = 0.70–0.93).
Validity
Content Validity
The content of the BASQID is based on a framework derived from qualitative work that conceptualized subjective QoL as the person's evaluation of multiple QoL domains of health, function, leisure, sleep, energy, mobility, environment, mood, and social interaction, as well as feelings of need fulfilment, identity, and affect; we included all components of this framework as items in the initial pool (Trigg et al., 2007). The 14 items in the BASQID relate to the QoL domains of health, social interaction, function, mobility, being occupied, energy, and psychological well-being.
Construct Validity
As expected, scores on the BASQID showed moderate correlation with the GDS-15 and low to moderate association with the proxy-completed WHOQOL-BREF (see Table 3). The social subscale of the WHOQOL-BREF and the GDS-15 displayed the highest level of association with BASQID scales.
|
Responsiveness
Clinicians judged 10 of the 36 people in Sample 2a to have an improved QoL after 3 months; 23 participants remained the same, and 3 deteriorated. We calculated effect size statistics for the BASQID and the two subscales (see Table 4) and found significant correlations between clinician global ratings of QoL change effect size statistics: for BASQID (r =.50, p <.01), for LS (r =.49, p <.01), for FPQ (r =.33, p <.05).
|
|
Discussion |
|---|
|
TOPAbstractMethodResultsDiscussionReferences |
|---|
The internal consistency of the 14 items indicates that practitioners may use the BASQID as a single scale of QoL. However, preliminary results from the PCA suggest that the 14 BASQID items may contain two independent subscales: LS and FPQ. We must acknowledge that the separation of items into a satisfaction component alongside a component relating to feelings and mood may be an artefact induced by the item stems, as wording can induce spurious factors (e.g., positively and negatively worded and ordering of items; see Marsh, 1996). However, existing conceptualizations of QoL support this separation of constructs (Brod et al., 1999; Ferrans & Powers, 1992; Jennings, 1999; Lawton, 1997), which may allow a more detailed investigation of QoL in people with dementia. Further evaluation of the dimensionality of the BASQID using confirmatory analyses is warranted, with larger samples and other models, to confirm the independence of these subscales.
Overall, the BASQID displays strong psychometric properties and is brief and acceptable to patients. The BASQID displays high internal consistency and acceptable test–retest reliability in line with other measures such as the QoL-AD and DQoL. As with the DQoL and QoL-AD, the BASQID does not show any association with MMSE scores, but it does display a strong association with depression. Construct validity was supported through significant correlations with proxy ratings of QoL. Although significant, these correlations were low to moderate in size, which concurs with the literature (Novella et al., 2001). As with any new measure, ongoing validation with large samples is needed in order to determine the usefulness of the measure in different settings with different groups of individuals. The samples used in the development of the BASQID were recruited from a memory clinic and therefore included people more likely to be receiving support and intervention from health and social services and accustomed to undergoing formal assessment. The participants were English-speaking, mainly Caucasian individuals, and therefore there is the possibility that the measure does not represent QoL issues important to other ethnic groups.
Although sensitive to the effects of dementia, the BASQID does not simply reflect the effects of worsening symptoms or changes in disease. Although changes in cognition and function may influence QoL, they should not be taken as a direct measure of QoL. There is a low level of association between the BASQID and the physical domain on the WHOQOL-BREF, but perhaps more significantly there is a lack of association between the BASQID and MMSE scores. This concurs with current research evidence (Hoe et al., 2005; Logsdon et al., 1999; Ready et al., 2002; Thorgrimsen et al., 2003) and adds weight to the argument that cognitive ability is not necessarily a predictor of subjective QoL. Indeed, the work of Kitwood (1995) has demonstrated that it is possible that the person with dementia can experience positive long-term changes.
We also explored the responsiveness of BASQID scores to changes in QoL over a 3-month period experienced by people receiving an acetylcholinesterase inhibitor for the treatment of AD. Effect sizes for the BASQID showed significant associations with the QoL change ratings of clinicians. Data from this study are strongly indicative of the fact that the BASQID is responsive to changes in QoL, but they are somewhat limited due to the relatively small sample size employed in this study and the use of physician ratings of QoL as the external standard of QoL change. Further work on responsiveness is needed, with longer follow-up periods to chart changes in QoL brought about through changes in other variables, such as lifestyle, living arrangements, care provision, and nonpharmacological therapeutic interventions. Studies such as this should combine quantitative and qualitative methodologies in order to assess how changes in BASQID scores relate to accounts of significant life events reported by patients and carers.
As a self-report measure, the BASQID should complement existing objective measures of health, disability, and QoL, giving a more complete appraisal of the QoL of people with dementia. We developed the BASQID to provide a means of better understanding the experiences, perceptions, and beliefs of people with dementia. This measure does this through acknowledgement of the many influences on QoL over and above health status. By adopting a biopsychosocial framework, the BASQID allows people to report QoL levels that might differ from those suggested by objective indices of health and disability. This subjective viewpoint on issues of life satisfaction and feelings of positive QoL may provide caregivers and researchers an important insight into treatments and therapies that have personal benefits for the person with dementia. With the availability of measures such as the BASQID, there is no longer a reason to ignore the perspective of the person with dementia when evaluating approaches to dementia care.
|
Footnotes |
|---|
We are grateful to all of the patients, carers, and staff who participated in this study at the Research Institute for the Care of the Elderly, St. Martin's Hospital, Bath. Copies of the BASQID measure and manual are available from http://www.rice.org.uk/research_BASQID.php
1 Department of Psychology, World Health Organization Centre for the Study of Quality of Life, University of Bath, England.
2 Research Institute for the Care of the Elderly, Bath, England.
Decision Editor: William J. McAuley, PhD
Received for publication January 4, 2007. Accepted for publication April 19, 2007.
|
References |
|---|
|
TOPAbstractMethodResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
