Behavioural Science Section / Subjective Memory Complaints: Their Nature, Correlates and Consequences Received: September 1, 2014 Accepted: December 3, 2014 Published online: March 19, 2015
Gerontology 2015;61:251–257 DOI: 10.1159/000371347
How Valid Are Subjective Ratings of Prospective Memory in Mild Cognitive Impairment and Early Dementia? Claire L. Thompson a Julie D. Henry b Peter G. Rendell c Adrienne Withall d Henry Brodaty e
a
James Cook University (Australia), Singapore Campus, Singapore; b The University of Queensland, St. Lucia, Qld., Australian Catholic University, Melbourne Campus, Melbourne, Vic., and d School of Public Health and Community Medicine, Faculty of Medicine, and e Centre for Healthy Brain Ageing and Dementia Collaborative Research Centre, The University of New South Wales, Sydney, N.S.W., Australia c
Abstract Background: Prospective memory refers to memory for future intentions and is a critical predictor of functional capacity in late adulthood. For many other cognitive abilities, selfand informant-rated methods of assessment are routinely used to guide clinical decision-making. However, little is known about the validity (and consequently the clinical utility) of subjective reports of prospective memory difficulties. Objective: The aim of this study was to compare clinical [mild cognitive impairment (MCI), dementia] and nonclinical older adults (healthy controls) on self- and informant-rated versions of prospective and retrospective memory function, as well as objective measures of prospective memory. Critical here was not only the assessment of between-group differences, but also whether these different methods of assessing memory function would show appropriate convergent and discriminant validity. Methods: A total of 138 participants aged between 64 and 92 years, diagnosed with
© 2015 S. Karger AG, Basel 0304–324X/15/0613–0251$39.50/0 E-Mail [email protected] www.karger.com/ger
dementia (n = 37), MCI (n = 48) or no impairment (n = 53), were asked to complete self- and informant-rated versions of the Prospective and Retrospective Memory Questionnaire (PRMQ). Participants also completed behavioural measures of global cognitive function [the Mini-Mental State Examination (MMSE)], as well as a behavioural measure of prospective memory (Virtual Week). Results: Self-reported impairments were equivalent across the three groups, and informant reports of impairment, while higher for those with dementia, did not distinguish MCI from controls. For the combined sample and for all three groups separately, both self- and informant reports of prospective memory showed poor convergent validity, at best correlating only weakly with Virtual Week. Self-reported prospective memory was correlated with informant report only in the dementia group, not in the control or MCI groups. Convergent and discriminant validity were poor, with self- and informant-rated prospective memory more strongly related to self- and informant-rated retrospective memory than to scores on Virtual Week. Conclusion: These data indicate that self-report and informant report may neither accurately measure prospective memory of older people, nor be sensitive to objective prospective memory difficulties in people with MCI and dementia. These data have potentially important implications for clinical practice. © 2015 S. Karger AG, Basel
Claire Thompson Department of Clinical Psychology James Cook University, Singapore Campus 149 Sims Drive, Singapore 387380 (Singapore) E-Mail claire.thompson @ jcu.edu.au
Downloaded by: Univ. of California Santa Barbara 198.143.33.33 - 7/28/2015 1:09:57 PM
Key Words Subjective memory complaint · Self-report · Informant report · Prospective memory · Mild cognitive impairment · Dementia · Prospective and Retrospective Memory Questionnaire · Virtual Week
The diagnosis of dementia or mild cognitive impairment (MCI), or major or minor neurocognitive disorder using DSM-5 nomenclature, requires the concern of an individual, his/her informant or the clinician that there has been a significant decline in cognitive function [1]. This requirement is noteworthy because it assumes that some degree of insight exists into one’s own cognition, or that of a care recipient or patient, at the developing stage of MCI [1]. However, clinicians may not have a longitudinal knowledge of the patient’s cognitive abilities, and recent research does not support the assumption of subjective insight of the patient or informant into cognitive change in late adulthood. For example, the incidence of subjective memory complaints (as assessed in a recent epidemiological study of ageing in people without dementia) was 89.5% for participants’ self-reports and 64.2% for informant-reported complaints [2]. This rate of subjective cognitive complaints clearly exceeds the incidence of MCI and dementia in this cohort. It potentially reflects a high prevalence of ‘worried well’ individuals and suggests that self- or informant-reported cognitive complaints may not provide a valid source of clinical information on which to base a diagnosis of MCI or dementia [3]. To date, most studies that have assessed both self- and informant-rated cognition, or the incidence of anosognosia, have tended to focus on retrospective memory (RM) or on functional ability relating to memory performance. For example, Buckley et al. [4] found a weak positive association between self-ratings of RM and related function in participants with no cognitive impairment but an inverse association for those with dementia. Thus, for the latter group, those with greater cognitive decline were more likely to report no change or even an improvement in their retrospective memory and function. In another study, Farias et al. [5] found that self-reports of function did not differentiate between groups with dementia, MCI or controls, whereas informants were more likely to report functional decline in those with dementia than in MCI or control groups, a finding which implies poorer awareness of functional loss in those with dementia. Finally, Chung and Man [6] assessed participants with MCI and age-matched controls. The results revealed a significant positive association between cognitive function (including memory) and self-reported memory satisfaction or ability in the control group, but no relationship between these constructs for those with MCI. 252
Gerontology 2015;61:251–257 DOI: 10.1159/000371347
Taken together, these studies suggest poor validity of self-report and at best only weak validity of informant report of memory and cognition. They also indicate that the validity of self-report may be weakest for patients with the greatest degree of cognitive impairment. However, for all the studies reported here, the assessment of memory focussed on memory for past events or RM. Increasingly, the importance of prospective memory (PM) is being recognised, particularly in late adulthood. This is because PM refers to memory for future intentions, such as remembering to take medication or turn off appliances, and is therefore crucial to the maintenance of functional independence and well-being more broadly. One of the most widely used measures of self- and informant-rated memory difficulties is the Prospective and Retrospective Memory Questionnaire (PRMQ) [7]. The PRMQ is a 16-item questionnaire that measures both PM (e.g., deciding to do something and then forgetting to do it) and RM (e.g., forgetting something you were told). Additionally, the PRMQ distinguishes between short-term memory (e.g., mislaying something you just put down) and long-term memory (e.g., failure to recognise a place previously visited), as well as between self-cued recall (e.g., forgetting what was watched on television the previous day) and environmentally cued recall (e.g., forgetting to buy a birthday card, even when you see the shop). The PRMQ was designed to measure these constructs using an 8-category structure (i.e., self-cued, short-term PM; self-cued, short term RM; environmentally cued, shortterm PM; environmentally cued, short-term RM; selfcued, long-term PM; self-cued, long-term RM; environmentally cued long-term PM, and environmentally cued, long-term RM). In their original validation study, Smith et al. [7] used the PRMQ to assess older adults with and without dementia. Those with dementia were rated by an informant (usually a spouse or an adult child), while those with no cognitive impairment completed a self-rated questionnaire. The results indicated that PM failures were more frequently reported than RM failures for both groups, with all responses for the dementia group close to the impaired end of the scale. In order to test the possibility that completing a questionnaire on behalf of another per se is associated with greater ratings of impairment, a subgroup of cognitively healthy, married couples were asked to rate both themselves and their spouses in order to assess influences of completing a questionnaire about somebody else. No significant differences emerged between ratings of self and other, suggesting that the two measurements Thompson/Henry/Rendell/Withall/ Brodaty
Downloaded by: Univ. of California Santa Barbara 198.143.33.33 - 7/28/2015 1:09:57 PM
Self-Report and Informant Report of Prospective and Retrospective Memory Impairment
did cohere, although actual correlations between self- and informant reports were not provided. However, although Smith et al. [7] used informant report to control for the possibility of unreliable self-ratings by those with dementia (which was expected due to the association between dementia and anosognosia), they did not explore actual self-ratings of those with dementia, nor did they include any ratings of individuals with milder cognitive impairment. Most importantly, the question of how self-report or informant report data related to objective performance on a behavioural measure of PM was not tested. Eschen et al. [8] partially addressed the limitations of the study by Smith et al. [7] by asking participants with MCI or mild Alzheimer’s disease (AD), as well as nonclinical controls, to self-rate their PM performance on the PRMQ. The results indicated that self-rated PM difficulties were equivalent across the three groups [8]. The findings of Eschen et al. therefore suggest that the validity of self-report assessment may be limited. However, informant ratings were not collected in this study, nor were the self-report assessments compared with performance on a behavioural measure of PM. The present study was designed to address these gaps by asking participants from both clinical and nonclinical older adult populations (MCI, dementia and healthy controls) to complete self- and informant-rated versions of the PRMQ, as well as an objective measure of PM (Virtual Week) [9]. Virtual Week has previously been shown to be a valid and sensitive indicator of PM in each of these groups [10]. The degree to which each of the questionnaire-based measures relates to global cognitive performance on the Mini-Mental State Examination (MMSE) [11] was also considered to evaluate the specificity of any observed associations with Virtual Week. We predicted that there would be group differences in both self- and informant-rated PM and RM, with failures more commonly reported in the dementia than in the MCI group, who would in turn report more failures than controls. It was further hypothesised that, consistent with Slavin et al. [2], self-reported memory complaints would be higher than informant-reported memory complaints in all three groups. Less clear was whether selfand informant-rated measures of PM would be positively correlated with one another, or related to performance on the behavioural measures (Virtual Week and the MMSE). While theoretically self- and informant-rated reports should cohere, as noted, prior research has indicated that these measures are not reliably associated with objective
Participants Participants who had been recruited from the University of New South Wales Memory and Ageing Study [12] and the Prince of Wales Hospital, Aged Care Psychiatry, Memory Disorders Clinic [13] for a study of PM [10] were asked to complete this study. Approval by the institutional ethics committees was obtained prior to any data collection. The PRMQ was completed by 138 of the 140 participants approached. Participants were aged 64–92 years (mean = 78.6, SD = 5.12), and 51.4% were females. Of the 138 participants, 53 were cognitively healthy, 48 met the criteria for MCI (14 amnestic single-domain, 6 amnestic multi-domain and 28 nonamnestic cases) and 37 met the criteria for dementia. Diagnoses of dementia subtype were not known in all cases but included 10 cases of AD, 2 mixed AD and vascular dementia, 1 Lewy body dementia and 1 dementia pugilistica; however, these diagnostic groupings are not explored in this study as the groups would be too small for meaningful analysis. Additionally, 117 informants took part in this study (although all participants were asked to identify informants, there were 21 informants who were not located or did not consent). The demographic data for participants and informants are presented in table 1. The three groups (dementia, MCI and controls) did not differ in age [F(2, 135) = 2.47, p = 0.088, ηp2 = 0.04], education [F(2, 135) = 0.95, p = 0.910, ηp2 = 0.01] or gender [χ2(2, n = 138) = 1.77, p = 0.412, φ = 0.11], but MMSE scores differentiated the groups [F(2, 135) = 20.29, p < 0.001, ηp2 = 0.23]. Follow-up Bonferroni tests indicated that participants with dementia had lower scores (mean = 25.8, SD = 3.3) on the MMSE relative to controls (mean = 28.7, SD = 1.42, p < 0.001), and the MCI group (mean = 28.0, SD = 1.6, p < 0.001), but that the latter two groups did not differ (p = 0.233). The MMSE scores indicate that dementia participants were in the mild stage of illness. The informants for each group did not differ in age [F(2, 85) = 0.39, p = 0.681, ηp2 = 0.01], gender [χ2(2, n = 116) = 0.18, p = 0.913, φ = 0.04), relationship to the participant [χ2(10, n = 102) = 7.56, p = 0.672, φ = 0.27] or number of years that they had known the participant [F(2, 79) = 0.32, p = 0.729, ηp2 = 0.01]. There was no difference in the proportion of informants that were cohabiting with the participant [χ2(2, n = 136) = 0.98, p = 0.612, φ = 0.10)] or, for those not cohabiting, in the number of hours of weekly contact the informants had with the participants [F(2, 38) = 1.18, p = 0.317, ηp2 = 0.06].
Subjective Ratings of PM in MCI and Early Dementia
Gerontology 2015;61:251–257 DOI: 10.1159/000371347
cognitive difficulties, or with one another. Since it has been suggested that this lack of coherence may be particularly pronounced in cognitively impaired populations, one possibility is that a significant relationship between self- and informant reports with objectively measured PM may only be evident in the control group. The results of this study will therefore have important conceptual, as well as practical, implications for informing the debate on the potential use of such measures as valid indicators of PM difficulties in older adult populations, and as potential diagnostic tools specifically.
253
Downloaded by: Univ. of California Santa Barbara 198.143.33.33 - 7/28/2015 1:09:57 PM
Materials and Methods
dementia participants Controls (n = 53) Participant characteristics Age, years 77.6 ± 4.7 Education, years 11.7 ± 3.3 Male gender 41.5 Informant characteristics Age, years 63.3 ± 13.0 Male gender 31.4 Relationship to participant Spouse 41.0 Child 23.9 Other relative 6.5 Friend 26.1 Other 2.2 Years known 42.4 ± 16.8 Cohabitants 43.2 6.0 ± 5.6 Weekly contacta, h
MCI (n = 48)
Dementia (n = 37)
78.6 ± 4.9 11.7 ± 3.7 54.2
80.1 ± 5.7 11.4 ± 3.9 51.4
63.2 ± 14.2 33.3
60.1 ± 14.2 28.6
30.6 25.0 8.4 36.1 0.0 39.8 ± 18.2 41.9 9.3 ± 8.8
34.6 42.3 3.8 19.2 0.0 43.5 ± 14.0 55.0 5.9 ± 2.4
Values are means ± SD or percentages. a Measured for those not cohabiting only.
Table 2. Estimates of Cronbach’s alpha for the PRMQ by group
Self-rated PM Self-rated RM Informant-rated PM Informant-rated RM
Control (n = 53)
MCI (n = 48)
Dementia (n = 37)
0.86 0.79 0.87 0.82
0.86 0.71 0.92 0.81
0.81 0.77 0.92 0.89
PRMQ and the Informant Version of PRMQ The PRMQ is a 16-item questionnaire about memory failures. It assesses PM and RM for both short- and long-term memory as well as for self- and environmentally cued memory [7]. Participants rate the frequency with which they experience each particular memory failure on a 5-point scale as follows: never = 1; rarely = 2; sometimes = 3; quite often = 4; very often = 5. The internal consistency of the scale in the present study was high, with overall Cronbach’s alpha estimated at 0.89 for both the self-report and informant report forms. Internal consistency coefficients for the PM and RM subscales for each of the three groups are reported in table 2. If the informant was not present at the assessment, the PRMQ was sent by mail with a cover letter, information and consent form and a replypaid return envelope. The questionnaire included the participant’s subject number only, with no other identifying information; however, the cover letter and information sheet identified the participant and made it clear that the questionnaire was about the participant, not about the informant, and it was entirely confidential.
254
Gerontology 2015;61:251–257 DOI: 10.1159/000371347
Cognitive Assessments The MMSE brief assessment of general cognition [11] and the Virtual Week assessment of PM [9] were administered before the PRMQ was given to participants and informants. Virtual Week takes the structure of a board game, in which participants move through each day as they circuit the board, with a series of events occurring and tasks to be remembered each day. The possible scores for the version used range from 0 to 8. The test is to facilitate administration and data capture. Virtual Week has evidence of good reliability and validity, with reliability estimates ranging from 0.84 to 0.94 [14] and validity established through controlled studies showing group differences with poorer performance in the aged [9, 14] and in clinical populations including MCI and dementia [10]. Virtual Week has also been argued to be more ecologically valid than most laboratory-based assessments of PM in that it closely approximates the types of PM tasks that occur in daily life [15].
Results
Group Differences on Virtual Week and PRMQ A one-way ANOVA showed group differences in Virtual Week scores [F(2, 135) = 21.49, p < 0.001], with Bonferroni tests indicating that the dementia group (mean = 1.03, SD = 2.14) had poorer performance relative to both the control group (mean = 3.58, SD = 2.05, p < 0.001) and the MCI group (mean = 2.00, SD = 2.04, p = 0.050). The MCI group also had significantly poorer performance than the controls (p < 0.001). The mean PRMQ scores (number of memory errors) are presented in figure 1 as a function of source (self-report, informant report), domain (PM, RM) and group (control, MCI, dementia). These data were analysed using a three-way mixed ANOVA with the within-subject variables of source (self, informant) and memory domain (PM, RM), and the between-subjects variable of group (control, MCI, dementia). The results showed that there was a main effect of source [F(1, 114) = 18.67, p < 0.001, ηp2 = 0.14], domain [F(1, 114) = 4.09, p = 0.045, ηp2 = 0.04] and group [F(2, 114) = 8.18, p < 0.001, ηp2 = 0.13]. There was a two-way interaction between source and group [F(2, 114) = 4.51, p = 0.013, ηp2 = 0.07], but there was no two-way interaction between memory domain and group [F(2, 114) = 1.72, p = 0.184, ηp2 = 0.03] or between memory domain and source [F(1, 114) = 0.16, p = 0.686, ηp2
Gerontology 2015;61:251–257 DOI: 10.1159/000371347
How Valid Are Subjective Ratings of Prospective Memory in Mild Cognitive Impairment and Early Dementia? Claire L. Thompson a Julie D. Henry b Peter G. Rendell c Adrienne Withall d Henry Brodaty e
a
James Cook University (Australia), Singapore Campus, Singapore; b The University of Queensland, St. Lucia, Qld., Australian Catholic University, Melbourne Campus, Melbourne, Vic., and d School of Public Health and Community Medicine, Faculty of Medicine, and e Centre for Healthy Brain Ageing and Dementia Collaborative Research Centre, The University of New South Wales, Sydney, N.S.W., Australia c
Abstract Background: Prospective memory refers to memory for future intentions and is a critical predictor of functional capacity in late adulthood. For many other cognitive abilities, selfand informant-rated methods of assessment are routinely used to guide clinical decision-making. However, little is known about the validity (and consequently the clinical utility) of subjective reports of prospective memory difficulties. Objective: The aim of this study was to compare clinical [mild cognitive impairment (MCI), dementia] and nonclinical older adults (healthy controls) on self- and informant-rated versions of prospective and retrospective memory function, as well as objective measures of prospective memory. Critical here was not only the assessment of between-group differences, but also whether these different methods of assessing memory function would show appropriate convergent and discriminant validity. Methods: A total of 138 participants aged between 64 and 92 years, diagnosed with
© 2015 S. Karger AG, Basel 0304–324X/15/0613–0251$39.50/0 E-Mail [email protected] www.karger.com/ger
dementia (n = 37), MCI (n = 48) or no impairment (n = 53), were asked to complete self- and informant-rated versions of the Prospective and Retrospective Memory Questionnaire (PRMQ). Participants also completed behavioural measures of global cognitive function [the Mini-Mental State Examination (MMSE)], as well as a behavioural measure of prospective memory (Virtual Week). Results: Self-reported impairments were equivalent across the three groups, and informant reports of impairment, while higher for those with dementia, did not distinguish MCI from controls. For the combined sample and for all three groups separately, both self- and informant reports of prospective memory showed poor convergent validity, at best correlating only weakly with Virtual Week. Self-reported prospective memory was correlated with informant report only in the dementia group, not in the control or MCI groups. Convergent and discriminant validity were poor, with self- and informant-rated prospective memory more strongly related to self- and informant-rated retrospective memory than to scores on Virtual Week. Conclusion: These data indicate that self-report and informant report may neither accurately measure prospective memory of older people, nor be sensitive to objective prospective memory difficulties in people with MCI and dementia. These data have potentially important implications for clinical practice. © 2015 S. Karger AG, Basel
Claire Thompson Department of Clinical Psychology James Cook University, Singapore Campus 149 Sims Drive, Singapore 387380 (Singapore) E-Mail claire.thompson @ jcu.edu.au
Downloaded by: Univ. of California Santa Barbara 198.143.33.33 - 7/28/2015 1:09:57 PM
Key Words Subjective memory complaint · Self-report · Informant report · Prospective memory · Mild cognitive impairment · Dementia · Prospective and Retrospective Memory Questionnaire · Virtual Week
The diagnosis of dementia or mild cognitive impairment (MCI), or major or minor neurocognitive disorder using DSM-5 nomenclature, requires the concern of an individual, his/her informant or the clinician that there has been a significant decline in cognitive function [1]. This requirement is noteworthy because it assumes that some degree of insight exists into one’s own cognition, or that of a care recipient or patient, at the developing stage of MCI [1]. However, clinicians may not have a longitudinal knowledge of the patient’s cognitive abilities, and recent research does not support the assumption of subjective insight of the patient or informant into cognitive change in late adulthood. For example, the incidence of subjective memory complaints (as assessed in a recent epidemiological study of ageing in people without dementia) was 89.5% for participants’ self-reports and 64.2% for informant-reported complaints [2]. This rate of subjective cognitive complaints clearly exceeds the incidence of MCI and dementia in this cohort. It potentially reflects a high prevalence of ‘worried well’ individuals and suggests that self- or informant-reported cognitive complaints may not provide a valid source of clinical information on which to base a diagnosis of MCI or dementia [3]. To date, most studies that have assessed both self- and informant-rated cognition, or the incidence of anosognosia, have tended to focus on retrospective memory (RM) or on functional ability relating to memory performance. For example, Buckley et al. [4] found a weak positive association between self-ratings of RM and related function in participants with no cognitive impairment but an inverse association for those with dementia. Thus, for the latter group, those with greater cognitive decline were more likely to report no change or even an improvement in their retrospective memory and function. In another study, Farias et al. [5] found that self-reports of function did not differentiate between groups with dementia, MCI or controls, whereas informants were more likely to report functional decline in those with dementia than in MCI or control groups, a finding which implies poorer awareness of functional loss in those with dementia. Finally, Chung and Man [6] assessed participants with MCI and age-matched controls. The results revealed a significant positive association between cognitive function (including memory) and self-reported memory satisfaction or ability in the control group, but no relationship between these constructs for those with MCI. 252
Gerontology 2015;61:251–257 DOI: 10.1159/000371347
Taken together, these studies suggest poor validity of self-report and at best only weak validity of informant report of memory and cognition. They also indicate that the validity of self-report may be weakest for patients with the greatest degree of cognitive impairment. However, for all the studies reported here, the assessment of memory focussed on memory for past events or RM. Increasingly, the importance of prospective memory (PM) is being recognised, particularly in late adulthood. This is because PM refers to memory for future intentions, such as remembering to take medication or turn off appliances, and is therefore crucial to the maintenance of functional independence and well-being more broadly. One of the most widely used measures of self- and informant-rated memory difficulties is the Prospective and Retrospective Memory Questionnaire (PRMQ) [7]. The PRMQ is a 16-item questionnaire that measures both PM (e.g., deciding to do something and then forgetting to do it) and RM (e.g., forgetting something you were told). Additionally, the PRMQ distinguishes between short-term memory (e.g., mislaying something you just put down) and long-term memory (e.g., failure to recognise a place previously visited), as well as between self-cued recall (e.g., forgetting what was watched on television the previous day) and environmentally cued recall (e.g., forgetting to buy a birthday card, even when you see the shop). The PRMQ was designed to measure these constructs using an 8-category structure (i.e., self-cued, short-term PM; self-cued, short term RM; environmentally cued, shortterm PM; environmentally cued, short-term RM; selfcued, long-term PM; self-cued, long-term RM; environmentally cued long-term PM, and environmentally cued, long-term RM). In their original validation study, Smith et al. [7] used the PRMQ to assess older adults with and without dementia. Those with dementia were rated by an informant (usually a spouse or an adult child), while those with no cognitive impairment completed a self-rated questionnaire. The results indicated that PM failures were more frequently reported than RM failures for both groups, with all responses for the dementia group close to the impaired end of the scale. In order to test the possibility that completing a questionnaire on behalf of another per se is associated with greater ratings of impairment, a subgroup of cognitively healthy, married couples were asked to rate both themselves and their spouses in order to assess influences of completing a questionnaire about somebody else. No significant differences emerged between ratings of self and other, suggesting that the two measurements Thompson/Henry/Rendell/Withall/ Brodaty
Downloaded by: Univ. of California Santa Barbara 198.143.33.33 - 7/28/2015 1:09:57 PM
Self-Report and Informant Report of Prospective and Retrospective Memory Impairment
did cohere, although actual correlations between self- and informant reports were not provided. However, although Smith et al. [7] used informant report to control for the possibility of unreliable self-ratings by those with dementia (which was expected due to the association between dementia and anosognosia), they did not explore actual self-ratings of those with dementia, nor did they include any ratings of individuals with milder cognitive impairment. Most importantly, the question of how self-report or informant report data related to objective performance on a behavioural measure of PM was not tested. Eschen et al. [8] partially addressed the limitations of the study by Smith et al. [7] by asking participants with MCI or mild Alzheimer’s disease (AD), as well as nonclinical controls, to self-rate their PM performance on the PRMQ. The results indicated that self-rated PM difficulties were equivalent across the three groups [8]. The findings of Eschen et al. therefore suggest that the validity of self-report assessment may be limited. However, informant ratings were not collected in this study, nor were the self-report assessments compared with performance on a behavioural measure of PM. The present study was designed to address these gaps by asking participants from both clinical and nonclinical older adult populations (MCI, dementia and healthy controls) to complete self- and informant-rated versions of the PRMQ, as well as an objective measure of PM (Virtual Week) [9]. Virtual Week has previously been shown to be a valid and sensitive indicator of PM in each of these groups [10]. The degree to which each of the questionnaire-based measures relates to global cognitive performance on the Mini-Mental State Examination (MMSE) [11] was also considered to evaluate the specificity of any observed associations with Virtual Week. We predicted that there would be group differences in both self- and informant-rated PM and RM, with failures more commonly reported in the dementia than in the MCI group, who would in turn report more failures than controls. It was further hypothesised that, consistent with Slavin et al. [2], self-reported memory complaints would be higher than informant-reported memory complaints in all three groups. Less clear was whether selfand informant-rated measures of PM would be positively correlated with one another, or related to performance on the behavioural measures (Virtual Week and the MMSE). While theoretically self- and informant-rated reports should cohere, as noted, prior research has indicated that these measures are not reliably associated with objective
Participants Participants who had been recruited from the University of New South Wales Memory and Ageing Study [12] and the Prince of Wales Hospital, Aged Care Psychiatry, Memory Disorders Clinic [13] for a study of PM [10] were asked to complete this study. Approval by the institutional ethics committees was obtained prior to any data collection. The PRMQ was completed by 138 of the 140 participants approached. Participants were aged 64–92 years (mean = 78.6, SD = 5.12), and 51.4% were females. Of the 138 participants, 53 were cognitively healthy, 48 met the criteria for MCI (14 amnestic single-domain, 6 amnestic multi-domain and 28 nonamnestic cases) and 37 met the criteria for dementia. Diagnoses of dementia subtype were not known in all cases but included 10 cases of AD, 2 mixed AD and vascular dementia, 1 Lewy body dementia and 1 dementia pugilistica; however, these diagnostic groupings are not explored in this study as the groups would be too small for meaningful analysis. Additionally, 117 informants took part in this study (although all participants were asked to identify informants, there were 21 informants who were not located or did not consent). The demographic data for participants and informants are presented in table 1. The three groups (dementia, MCI and controls) did not differ in age [F(2, 135) = 2.47, p = 0.088, ηp2 = 0.04], education [F(2, 135) = 0.95, p = 0.910, ηp2 = 0.01] or gender [χ2(2, n = 138) = 1.77, p = 0.412, φ = 0.11], but MMSE scores differentiated the groups [F(2, 135) = 20.29, p < 0.001, ηp2 = 0.23]. Follow-up Bonferroni tests indicated that participants with dementia had lower scores (mean = 25.8, SD = 3.3) on the MMSE relative to controls (mean = 28.7, SD = 1.42, p < 0.001), and the MCI group (mean = 28.0, SD = 1.6, p < 0.001), but that the latter two groups did not differ (p = 0.233). The MMSE scores indicate that dementia participants were in the mild stage of illness. The informants for each group did not differ in age [F(2, 85) = 0.39, p = 0.681, ηp2 = 0.01], gender [χ2(2, n = 116) = 0.18, p = 0.913, φ = 0.04), relationship to the participant [χ2(10, n = 102) = 7.56, p = 0.672, φ = 0.27] or number of years that they had known the participant [F(2, 79) = 0.32, p = 0.729, ηp2 = 0.01]. There was no difference in the proportion of informants that were cohabiting with the participant [χ2(2, n = 136) = 0.98, p = 0.612, φ = 0.10)] or, for those not cohabiting, in the number of hours of weekly contact the informants had with the participants [F(2, 38) = 1.18, p = 0.317, ηp2 = 0.06].
Subjective Ratings of PM in MCI and Early Dementia
Gerontology 2015;61:251–257 DOI: 10.1159/000371347
cognitive difficulties, or with one another. Since it has been suggested that this lack of coherence may be particularly pronounced in cognitively impaired populations, one possibility is that a significant relationship between self- and informant reports with objectively measured PM may only be evident in the control group. The results of this study will therefore have important conceptual, as well as practical, implications for informing the debate on the potential use of such measures as valid indicators of PM difficulties in older adult populations, and as potential diagnostic tools specifically.
253
Downloaded by: Univ. of California Santa Barbara 198.143.33.33 - 7/28/2015 1:09:57 PM
Materials and Methods
dementia participants Controls (n = 53) Participant characteristics Age, years 77.6 ± 4.7 Education, years 11.7 ± 3.3 Male gender 41.5 Informant characteristics Age, years 63.3 ± 13.0 Male gender 31.4 Relationship to participant Spouse 41.0 Child 23.9 Other relative 6.5 Friend 26.1 Other 2.2 Years known 42.4 ± 16.8 Cohabitants 43.2 6.0 ± 5.6 Weekly contacta, h
MCI (n = 48)
Dementia (n = 37)
78.6 ± 4.9 11.7 ± 3.7 54.2
80.1 ± 5.7 11.4 ± 3.9 51.4
63.2 ± 14.2 33.3
60.1 ± 14.2 28.6
30.6 25.0 8.4 36.1 0.0 39.8 ± 18.2 41.9 9.3 ± 8.8
34.6 42.3 3.8 19.2 0.0 43.5 ± 14.0 55.0 5.9 ± 2.4
Values are means ± SD or percentages. a Measured for those not cohabiting only.
Table 2. Estimates of Cronbach’s alpha for the PRMQ by group
Self-rated PM Self-rated RM Informant-rated PM Informant-rated RM
Control (n = 53)
MCI (n = 48)
Dementia (n = 37)
0.86 0.79 0.87 0.82
0.86 0.71 0.92 0.81
0.81 0.77 0.92 0.89
PRMQ and the Informant Version of PRMQ The PRMQ is a 16-item questionnaire about memory failures. It assesses PM and RM for both short- and long-term memory as well as for self- and environmentally cued memory [7]. Participants rate the frequency with which they experience each particular memory failure on a 5-point scale as follows: never = 1; rarely = 2; sometimes = 3; quite often = 4; very often = 5. The internal consistency of the scale in the present study was high, with overall Cronbach’s alpha estimated at 0.89 for both the self-report and informant report forms. Internal consistency coefficients for the PM and RM subscales for each of the three groups are reported in table 2. If the informant was not present at the assessment, the PRMQ was sent by mail with a cover letter, information and consent form and a replypaid return envelope. The questionnaire included the participant’s subject number only, with no other identifying information; however, the cover letter and information sheet identified the participant and made it clear that the questionnaire was about the participant, not about the informant, and it was entirely confidential.
254
Gerontology 2015;61:251–257 DOI: 10.1159/000371347
Cognitive Assessments The MMSE brief assessment of general cognition [11] and the Virtual Week assessment of PM [9] were administered before the PRMQ was given to participants and informants. Virtual Week takes the structure of a board game, in which participants move through each day as they circuit the board, with a series of events occurring and tasks to be remembered each day. The possible scores for the version used range from 0 to 8. The test is to facilitate administration and data capture. Virtual Week has evidence of good reliability and validity, with reliability estimates ranging from 0.84 to 0.94 [14] and validity established through controlled studies showing group differences with poorer performance in the aged [9, 14] and in clinical populations including MCI and dementia [10]. Virtual Week has also been argued to be more ecologically valid than most laboratory-based assessments of PM in that it closely approximates the types of PM tasks that occur in daily life [15].
Results
Group Differences on Virtual Week and PRMQ A one-way ANOVA showed group differences in Virtual Week scores [F(2, 135) = 21.49, p < 0.001], with Bonferroni tests indicating that the dementia group (mean = 1.03, SD = 2.14) had poorer performance relative to both the control group (mean = 3.58, SD = 2.05, p < 0.001) and the MCI group (mean = 2.00, SD = 2.04, p = 0.050). The MCI group also had significantly poorer performance than the controls (p < 0.001). The mean PRMQ scores (number of memory errors) are presented in figure 1 as a function of source (self-report, informant report), domain (PM, RM) and group (control, MCI, dementia). These data were analysed using a three-way mixed ANOVA with the within-subject variables of source (self, informant) and memory domain (PM, RM), and the between-subjects variable of group (control, MCI, dementia). The results showed that there was a main effect of source [F(1, 114) = 18.67, p < 0.001, ηp2 = 0.14], domain [F(1, 114) = 4.09, p = 0.045, ηp2 = 0.04] and group [F(2, 114) = 8.18, p < 0.001, ηp2 = 0.13]. There was a two-way interaction between source and group [F(2, 114) = 4.51, p = 0.013, ηp2 = 0.07], but there was no two-way interaction between memory domain and group [F(2, 114) = 1.72, p = 0.184, ηp2 = 0.03] or between memory domain and source [F(1, 114) = 0.16, p = 0.686, ηp2
