Differential Benefits of Memory Training for Minority Older Adults in the SeniorWISE Study

Abstract

Purpose: Cognitive training improves mental abilities in older adults, but the benefit to minority elders is unclear. We conducted a subgroup analysis of subjects in the SeniorWISE (Wisdom Is Simply Exploration) trial to examine this issue. Design and Methods: SeniorWISE was a Phase 3 randomized trial that enrolled 265 nondemented community-dwelling older adults aged 65 years and older between 2001 and 2006. Participants were randomly assigned to 12 hr of either memory or health training. Results: The sample was 79% female, 71% Caucasian, 17% Hispanic, and 12% African American. On the Rivermead Behavioural Memory Test (RBMT), 28% of the sample scored normal, 47% scored poor, and 25% impaired. Memory performance changed differently over time depending on the demographic characteristics of participants. Both Hispanics and Blacks performed better than Whites on visual memory, and Blacks performed better over time on instrumental activities of daily living. On all performance measures, lower pretest scores were associated with relatively greater improvements over time. Implications: Our analyses suggested that minority participants received differential benefits from the memory training; however, this remains speculative because the 3 ethnic groups in the sample were not equivalent in size. The question of why Black and Hispanic participants often made greater improvements needs further exploration.

Longitudinal studies have demonstrated the predictability of cognitive deficits with age, particularly in episodic memory (Salthouse, 2003; Schaie, 1989). However, in recent years, the view of cognitive aging as an inevitable decremental process has shifted and psychosocial interventions have begun to be used to prevent or remediate cognitive decline. The “Disuse Hypothesis” (Salthouse, Berish, & Miles, 2002; Salthouse & Mitchell, 1990) and the “Use It or Lose It” Learning Theory of Mental Discipline (Bigee, 1999) are popular models guiding cognitive intervention research, even though they were first tested in animal models. Exercises for the brain are hypothesized to assist with the maintenance of cognitive function. This can be viewed as a self-help strategy that individuals value, believing that their efforts will pay off in reduced incidence of cognitive impairment; however, the mental stimulation hypothesis has not been completely supported empirically (Salthouse, 2006).

Cognitive impairments and dementia are disproportionately prevalent in some groups of older adults (Clark et al., 2005; Espino, Lichtenstein, Palmer, & Hazuda, 2001; Haan, Mungas, Gonzalez, Ortiz, Acharya, & Jagust, 2003; Parker & Philp, 2004). For example, older adults at risk of impaired episodic memory and executive function often have depressive symptoms (Butters et al., 2008; Carlson, Xue, Zhou, & Fried, 2009; Herrmann, Goodwin, & Ebmeler, 2007). Furthermore, complaints about declining memory performance are often associated with anxiety and depression (Dux et al., 2008; Fiocco, Wan, Weekes, Pim, & Lupien, 2006; McDougall, Strauss, Holston, & Martin, 1999). High-risk groups, many of whom are linked by both low income and low literacy, include racial and ethnic minorities, women, individuals with chronic and disabling conditions, and the elderly. Not only are these older adults at greater risk of developing cognitive impairments and Alzheimer’s disease (AD), but they also experience disparities in both treatment and research on cognitive disorders (Glymour, Weuve, & Chen, 2008; Sloan & Wang, 2005; Wilson et al., 2002; Zuckerman et al., 2008).

In Hispanic culture, there is a generalized stigma associated with mental disorders: any afflicted individual is considered to be “crazy” or have “bad blood” (Taussig & Trejo, 1992). Memory problems and fear of developing AD are particularly acute among Hispanics, but little is known about memory in this population (Gallagher-Thompson et al., 1997). Older Latinos show Alzheimer’s symptom onset at a mean of 6.8 years earlier than Anglos. In one study (Espino et al., 2002), 89 Mexican American elders who were seen in an outpatient memory evaluation clinic had greater than expected memory impairment, 83% had impairments in instrumental activities of daily living (IADLs), and 63% had high levels of depressive symptoms.

The Duke Established Populations for the Epidemiologic Studies of the Elderly (EPESE) study noted, however, that cognitive problems were difficult to recognize in Hispanic families (Watson, Lewis, & Fillenbaum, 2005). Among informants who reported memory loss, 30% were found not to have cognitive impairment. Yet among participants for whom family informants reported no memory loss, 75% were diagnosed with cognitive impairment or dementia. Mexican Americans are 2.2 times more likely than European Americans to have Mini-Mental State Examination (MMSE) scores less than 24, indicating cognitive impairment (Henderson & Gutierrez-Mayka, 1997). In the Hispanic EPESE study, only 35.6% of the Mexican American elderly passed both the MMSE and the executive clock-drawing task (Black et al., 1999; Royall, Espino, Polk, Palmer, & Markides, 2004).

Even using uniform detection methods and controlling for reported duration of dementia symptoms, cognitive impairment is significantly more severe when AD is recognized in Blacks than it is in Whites (Shadlen, Larson, Gibbons, McCormick, & Teri, 1999). In a sample of Black and White community elderly, age and race were significant predictors of memory performance, even accounting for education, depression, gender, and memory complaints (McDougall, 2002). One study found that 35% of African Americans compared with only 15% of Whites scored low enough to diagnose mild cognitive impairment (MCI). However, when new racially sensitive scoring methods were applied, the difference in MCI rates disappeared: 17% of European Americans and only 12% of African Americans scored low enough for a diagnosis (Marenberg, 2004). Greater social resources, as defined by social networks and social engagement, have been associated with less cognitive decline in old age among Black elderly (Barnes et al., 2005).

The notion of adaptation of older individuals to age-related cognitive decline, known as compensation, has been useful in examining cognitive aging, although “plasticity” may be more useful because this reflects a dynamic ability to adjust and change that involves both cellular and synaptic mechanisms (Burke & Barnes, 2006; Elias & Wagster, 2007; Greenwood, 2007). Mental stimulation, social engagement, and confidence building may be mechanisms to maintain cognitive health, particularly for minority elders (Bassuk, Glass, & Berkman, 1999; Fried et al., 2004; Mahncke, Bronstone, & Merzenich, 2006; Saczynski et al., 2006; Wilson et al., 2003). In a tristate sample of older adults with varying levels of memory performance, adults aged 60 years and older had low memory self-efficacy (McDougall, 2009). There was also a consistent pattern of decline in their self-efficacy beliefs with age. Therefore, interventions designed to improve self-efficacy may be as important as teaching mnemonic strategies, particularly for minority elders (McDougall, 2004). However, when minorities have been included in memory improvement studies, the findings have rarely noted the differential benefits of training by race and/or ethnicity (Ball, Berch, Helmers, Jobe, Leveck, Marsiske et al., 2002; Unverzagt et al., 2007; Willis, Tennstedt, Marsiske, Ball, Elias, Koepke et al., 2006).

The SeniorWISE (Wisdom Is Simply Exploration) study was a Phase 3 randomized clinical trial that examined the effectiveness of memory versus health training in improving everyday memory performance. The main study findings have been reported elsewhere (McDougall, Becker, Pituch, Vaughan, Acee, & Delville, 2010). The Cognitive Behavioral Model of Everyday Memory (CBMEM) was used to deliver the memory training content. This model was derived from self-efficacy theory and was operationalized with its four theoretical components: enactive mastery experience, vicarious experience, verbal persuasion, and physiologic arousal (McDougall, 2009). The CBMEM includes a number of interrelated theoretical components. The first components are antecedent factors, which reside with the individual. They include age, cognition, education, gender, and health. The CBMEM intervention is the second component of the model. The intervention is a psychosocial intervention that emphasizes cognitive and behavioral aspects and combines a unique package of cognitive skill development in exposure, repeated practice, relevant modeling, self-modeling, cognitive skill modeling, exhortation, suggestion, and desensitization. The intervention integrates both didactic content and practical application. The third components are mediators of memory performance, which are known to affect memory performance. They include anxiety, depression, metamemory, and memory self-efficacy. The next component is the proximal outcome of the intervention, everyday memory performance, and finally, the distal outcome is function, specifically IADLs. Theoretically, the intent is to transfer learning from the classroom to everyday experience.

The major hypothesis tested in the SeniorWISE study was that at-risk older adults who received a memory training intervention would show significantly better memory self-efficacy and memory performance and better function in IADLs than subjects in the comparison condition at postclass (2 months after baseline), and this benefit would be maintained over the remaining assessments, at postbooster (6 months), postclassroom follow-up (14 months), and end of study (26 months).

Cognitive training has been found to improve mental abilities in older adults, but the benefit to minority elders is unclear. We therefore conducted a subgroup analysis of subjects in the SeniorWISE trial to examine this issue.

Methods

Participants

Independent adults were recruited from a metropolitan area in Central Texas via print and TV media, as well as directly from city-run senior activity centers, churches, health fairs, and festivals. The memory training was implemented at seven sites in the community: four senior centers, a university-based wellness clinic, and two apartment complexes for low-income older adults.

A total of 346 independent adults were recruited, but 81 individuals were excluded: 21 did not meet inclusion criteria, 38 declined, and 22 were couples, and in each case, the female spouse was excluded from the testing so that adequate numbers of males would be represented in the sample. The final sample (N = 265) was 72% non-Hispanic White, 17% Hispanic, and 11% African American (Table 1). The average age was 75 years, and average MMSE score was 26. The majority of the participants were female (77%).

Table 1.

Baseline Demographic Variables of Sample in Memory and Health Training Groups

		Comparison group, n = 130	Intervention group, n = 135
Variables	Score range	M ± SD or N (%)	M ± SD or N (%)
Age	65–94	74.83 ± 6.22	74.69 ± 5.74
Education	0–22	13.81 ± 3.71	13.39 ± 3.9
Gender
Male		30 (23.1)	30 (22.2)
Female		100 (76.9)	105 (77.8)
Race
White		93 (71.5)	96 (71.1)
Black		15 (11.5)	15 (11.1)
Hispanic		22 (16.9)	24 (9.1)
Marital status
Married		37 (28.5)	45 (33.8)
Never married		3 (2.3)	5 (3.8)
Divorced		26 (20)	25 (18.8)
Widowed		64 (49.2)	58 (43.6)

Note: Age and education are in years; Hispanic was coded as 1 = Hispanic and 0 = not Hispanic (White is the reference group); Black was coded as 1 = Black and 0 = not Black (White is the reference group); class attendance refers to the number of intervention classes attended by the participant not including booster sessions; pretest = the baseline score on the dependent variable of interest.

At the final testing, there were 108 individuals in the memory intervention condition and 101 in the health training condition. According to t-test analyses, the memory and health training conditions did not differ significantly at baseline on either the study variables or demographics. In this secondary analysis, we were interested in whether participants received differential benefits of memory training by ethnicity.

Measures

Screening.—

To be eligible, elderly had to have adequate hearing and vision and ability to communicate. Other eligibility criteria included age (≥65 years), ability to speak and understand English, and willingness to participate in the study for 24 months. Diagnoses that precluded people from participating included AD or other dementia; Hodgkin’s disease; neuroblastoma; or cancer of the liver, lung, or brain. Potential participants were initially screened for sensory loss using a self-report evaluation of hearing and vision, either over the telephone or in person. Visual and hearing acuity were subsequently evaluated at the “in-person” eligibility screening by evaluator observation and by a self-report checklist developed for the study. Communication was assessed using a seven-item checklist designed for the study and completed by a member of the team at the initial screening. Medical eligibility was assessed verbally using a health status checklist designed for the project.

The MMSE was used to screen for cognitive impairment. Initially, those with scores higher than 23 were eligible to participate. However, in order to recruit Hispanics with low education, the eligibility score was changed to 20 or higher at one senior center. Four females and one male were included in this adjusted screening protocol. The Controlled Oral Word Association Test (COWAT) and the Trail-Making Test were administered to rule out dementia. On the COWAT, participants were required to achieve an age- and education-corrected score of 24 or higher to be eligible for the study. Participants were required to pass Trails A and/or Trails B at or above the 10th percentile for their age group to be included in the study.

Procedure

Once participants were screened and consented, Excel was used to randomly assign participants’ numeric codes to either the experimental or a comparison condition health training. Because there were multiple sites, randomization was done separately for each site. Participants were randomly assigned to 11 different cohorts, which met at seven different sites in the community.

Regardless of treatment condition assignment, memory or health training, each cohort participated in 12 hr of classroom sessions. One hundred and thirty-five individuals were assigned to the memory training intervention and 130 to the health training intervention. No class was smaller than 4 individuals or larger than 15; we strove for an average class size of 12 individuals.

Intervention Design.—

Memory training sessions were presented in a small group format twice a week for a month. Participants received 12 hr of structured training in topics, including memory and health, memory functions and mechanisms, actions affecting memory for people of all ages, memory beliefs and aging, and use of internal and external memory strategies. The memory training intervention was planned and implemented to enhance treatment fidelity in the three areas of delivery, receipt, and enactment as recommended by the Consortium guidelines proposed by Bellg and colleagues (2004).

At each site, all participants were trained in stress inoculation techniques during the first 30 min of class. Although the amount of practice varied, participants had an opportunity to engage in the use of the skills, either in imagination or in reality. A stressful situation was presented as a set of small units, each of which could be seen as a manageable problem that the participant had appropriate skills for confronting. The application phase of each memory training class enabled participants to consolidate the skills covered in earlier stages and to identify any problems, which the instructor addressed. Thirty minutes of lecture were presented, with time allowed for questions and answers; then, an additional 30 min were allocated to strengthen enactive mastery experience. A female septuagenarian role model taught the memory training classes.

At each session, homework was reviewed and a list of common problems and concerns was made. The thoughts and feelings that occurred when subjects forgot were discussed, and group solutions to problems and reactions to problems were identified. Participants were asked to describe memory successes and failures and to write out specific questions that they would like to be answered by the group. Feedback on performance accomplishments and verbal persuasion was given continually throughout the session. There also was an observer in the classroom to track each individual’s performance and mastery of practice activities. The observer debriefed with the instructor and pointed out areas that need to be reviewed in the next session to prevent the development of poor habits. After each class, each participant wrote down some aspect of the learning or an essential point for the day, such as a memory strategy. The individual participant was scored with either a plus or a minus (0, 1) for the class session. Between the eight-session CBMEM, and four booster sessions, a participant could have a possible score between 0 and 12. To meet this aspect of treatment fidelity, we included those participants in the analyses who attended six of the eight sessions or received 75% of the dosage.

Four booster sessions were delivered over 3 months to enhance perceived self-efficacy and provide a transitional experience between the theoretical and practical concerns of memory improvement. Four 2-hr sessions were held weekly for 1 month. The emphasis in these sessions was on practical everyday memory strategies needed for maintenance of IADLs.

This health training condition consisted of 18 topics that emphasized successful aging (McDougall, Becker, Acee, Vaughan, Pituch, & Delville, 2009). They included exercise, spirituality and health, alternative medicine, weight management, getting the most from your physician visit, caring for the caretaker, healing foods, drug interactions, osteoporosis, maintaining relationships, health myths, consumer fraud, nutrition, leisure activities, writing family stories, health monitoring tests for home use, useful web sites, and buying drugs in foreign countries. The topics were delivered via PowerPoint presentations in 90-min classes, and the facilitator, a middle-aged Caucasian male, was a graduate nursing student. He did not follow a structured curricular format as in the memory training but allowed class discussion to occur based on the needs and interests of participants. Participants received the same number of classroom hours as in the memory training condition, as well as four 2-hr booster sessions over the 3 months following the training.

Results

Outcome Measures

We included outcome measures that matched the content of the training. All cognitive and self-report measures were administered at baseline, then at postclass (2 months after baseline), again at postbooster (6 months), then at postclassroom follow-up (14 months), and at the end of the study (26 months).

Hopkins Verbal Learning Test–Revised.—

Verbal memory was tested with the Hopkins Verbal Learning Test–Revised (HVLT-R), which assesses immediate recall, delayed recall, and recognition memory (Benedict, 1997; Brandt, 1991). Test–retest correlation was .66 for the delayed recall subscale, which was used in these analyses.

Brief Visuospatial Memory Test–Revised.—

Visual memory performance was determined with the Brief Visuospatial Memory Test–Revised (BVMT-R), which asks the individual to reproduce a series of geometric designs (Cockburn & Smith, 1989; Gale, Baxter, Connor, Herring, & Comer, 2007). It was administered with three learning trials and a 25-min delayed recall trial. Test–retest correlation was .72.

Rivermead Behavioral Memory Test.—

Everyday memory was measured with the Rivermead Behavioral Memory Test (Loewenstein, Amigo, Duara, Guterman, Hurwitz, Berkowitz et al., 1989; Wilson, Cockburn, Baddeley, Hiorns, 1989). The standard profile score has a possible range from 0 to 24 and is sometimes interpreted using cutoff points for four categories of memory function: normal (22–24), poor memory (17–21), moderately impaired (10–16), and severely impaired memory (0–9). Cronbach’s alpha in this sample was .73.

Direct Assessment of Functional Status.—

IADLs were tested with a performance-based measure, the Direct Assessment of Functional Status (DAFS), which contains 85 items (Loewenstein, Ardila, Rosselli, Hayden, Duara, Berkowitz et al., 1992). The DAFS measures a broad spectrum of behaviors in each of seven functional domains. A modified version of the DAFS was used in this study to take into account the lack of kitchen and bathroom facilities in the evaluation rooms at the study sites. The domains measured included time orientation, communication abilities, transportation, financial skills, and shopping skills. A fifth domain, medication knowledge, was added for this study. Alpha reliability was .79.

Memory Self-Efficacy Questionnaire.—

Memory self-efficacy was measured with the Memory Self-Efficacy Questionnaire (MSEQ; Berry, West & Dennehey, 1989), on which respondents predicted their performance level and estimate their strength and confidence in performing 10 everyday tasks. In this study, the 35-item version of the MSEQ was used; alpha reliability was .95.

Spielberger State–Trait Anxiety Inventory.—

The Spielberger State–Trait Anxiety Inventory (STAI) was used to measure anxiety. The STAI was developed for the expressed purpose of detecting anxiety in adults (Spielberger, Gorsuch, & Lushene, 1970) and is used to differentiate between the temporary condition of “state anxiety” and chronic “trait anxiety.” Alpha reliabilities were .85–.89.

Center for Epidemiological Studies–Depression Scale.—

The Center for Epidemiological Studies–Depression scale (CES-D) evaluated depressive symptoms (Radloff & Teri, 1986). Somatic complaints are emphasized on this measure that individuals respond to on a 4-point Likert scale. Reliability coefficients from .85 to .91 have been obtained when the CES-D was used with older adults. Alpha reliability in this study was .78.

Statistical Analysis

The multivariate approach to repeated measures analyses was used to avoid the more restrictive assumptions underlying the univariate model. The predictor variables were ethnicity, group assignment, time, and education. Interaction effects were also analyzed. These analyses were restricted to those participants who attended at least six (75%) of the classes. This complies with the design of the grant, but it also restricts the range of the dosage variable and the Dosage × Group interaction.

These findings are shown in Figures 1^–9. Tables 1 and 2 show changes over time by intervention condition for Black and Hispanic participants. With a few exceptions, the results generally suggested modest gains in performance and self-report measures. The pattern was often different for Black and Hispanic participants.

Figure 1.

Mini-Mental State Examination (MMSE) treatment group scores.

Figure 2.

Hopkins Verbal Learning Test–Revised treatment group scores.

Figure 3.

Brief Visuospatial Memory Test–Revised treatment group scores.

Figure 4.

Rivermead Behavioural Memory Performance Test standard profile treatment group scores.

Figure 5.

Direct Assessment of Functional Status treatment groups total score.

Figure 6.

Memory self-efficacy treatment group scores.

Figure 7.

Center for Epidemiological Studies–Depression scale (CES-D) treatment group scores.

Figure 8.

State anxiety treatment group scores.

Figure 9.

Trait anxiety treatment group scores.

Table 2.

Descriptive Statistics and Relevant p Values for Performance Outcomes of Minority Participants

	Group		Time 1	Time 2	Time 5	p, Time T1–T2	p, Time × Group T1–T2	p, Time T1–T5	p, Time × Group T1–T5
Cognition
Hispanic	Health	M	26.13	25.53	26.33	.52	.30	.42	.73
		SD	2.72	2.33	2.85
	Memory	M	26.17	26.44	26.67
		SD	2.62	2.36	2.70
Black	Health	M	26.46	26.46	26.31	.48	.64	.29	.16
		SD	2.37	3.38	2.43
	Memory	M	26.75	26.75	27.75
		SD	1.83	0.71	1.98
Total sample	Health	M	27.98	27.77	27.95
		SD	2.27	2.33	2.15
	Memory	M	28.05	28.31	28.04
		SD	2.06	1.82	2.27
Everyday memory
Hispanic	Health	M	13.94	14.94	13.50	.00	.33	.66	.84
		SD	5.51	5.59	5.94
	Memory	M	17.28	19.78	17.11
		SD	4.24	3.62	3.36
Black	Health	M	14.69	14.54	15.77	.11	.11	.33	.33
		SD	5.78	6.01	6.51
	Memory	M	16.22	17.89	16.22
		SD	2.99	4.01	2.95
Total sample	Health	M	18.31	19.07	18.47
		SD	4.60	4.57	4.78
	Memory	M	19.08	20.08	19.13
		SD	3.44	3.18	3.93
Instrumental activities
Hispanic	Health	M	76.13	78.06	73.75	.09	.72	.12	.12
		SD	7.29	5.81	9.16
	Memory	M	79.71	80.35	79.71
		SD	5.93	7.05	6.79
Black	Health	M	77.08	77.31	77.23	.50	.43	.33	.40
		SD	10.45	10.76	9.26
	Memory	M	80.56	81.89	82.56
		SD	3.64	4.48	5.05
Total sample	Health	M	81.95	83.11	82.08
		SD	6.30	6.00	6.93
	Memory	M	83.05	83.81	82.57
		SD	4.13	4.30	5.50
Verbal memory
Hispanic	Health	M	41.44	38.44	40.31	.74	.04	.64	.24
		SD	10.63	11.78	13.48
	Memory	M	42.74	44.68	45.37
		SD	11.39	8.79	7.57
Black	Health	M	37.92	38.00	37.77	.12	.43	.06	.05
		SD	10.74	11.07	15.59
	Memory	M	38.78	42.44	46.44
		SD	10.17	8.32	9.48
Total sample	Health	M	47.45	48.23	48.28
		SD	11.27	11.98	13.28
	Memory	M	49.64	50.31	50.48
		SD	11.12	10.08	10.87
Visual memory
Hispanic	Health	M	33.25	31.50	32.25	.55	.75	.52	.88
		SD	13.52	13.53	14.36
	Memory	M	33.32	32.05	31.74
		SD	10.90	10.59	12.32
Black	Health	M	32.85	34.85	36.54	.66	.21	.36	.60
		SD	12.97	12.12	17.33
	Memory	M	34.89	28.67	35.89
		SD	11.20	6.38	12.31
Total sample	Health	M	44.03	43.15	42.90
		SD	15.04	14.78	16.57
	Memory	M	44.07	43.97	41.46
		SD	13.50	13.79	14.05

Note: Cognition = Mini-Mental State Examination; everyday memory = standard profile score from the Rivermead Behavioral Memory Performance Test; visual memory = delayed recall T score from the Brief Visuospatial Memory Test–Revised; instrumental activities = Direct Assessment of Functional Status; verbal memory = delayed recall T score from the Hopkins Verbal Learning Test–Revised.

Performance Measures

The results of all performance measures are included in Table 2. On cognition scores, we found significant effects for education, F(1, 196) = 36.70, p = .000, η² = .16; ethnic group, F(2, 196) = 17.82, p = .000, η² = .15; and treatment condition, F(1, 196) = 3.89, p = .050, η² = .02. Whites’ MMSE scores were stable across time, whereas the Hispanic and African American groups had more variance (Figure 1). There were no group differences at baseline in MMSE scores.

On verbal memory, less educated Black and Hispanic participants scored lower than Whites. The HVLT showed significant effects for ethnic group, F(2, 189) = 17.57, p = .000, η² = .06; education, F(1, 189) = 11.66, p = .001, η² = .01; and treatment condition, F(1, 189) = 3.97, p = .048, η² = .05. However, Whites scored higher than the two ethnic groups (Figure 2).

On visual memory at baseline, BVMT-R scores of older, less educated Black and Hispanic participants had significantly lower scores than Whites. There were significant effects for ethnic group, F(2, 198) = 11.66, p = .000, η² = .11; and education, F(1, 198) = 9.97, p = .002, η² = .05. However, both Hispanics and Blacks performed better than Whites on visual memory (Figure 3).

On everyday memory performance, older less educated individuals scored lower, as did Hispanic and Black participants (Figure 4). The RBMT measure showed statistically significant changes by ethnic group, F(2, 198) = 17.81, p = .000, η² = .15; education, F(1, 198) = 14.70, p = .000, η² = .07; treatment condition, F(1, 198) = 9.17, p = .003, η² = .04; and an interaction of Ethnicity × Treatment Condition, F(2, 198) = 6.29, p = .002, η² = .06. There was more variability over time for all ethnic groups in the memory training condition. For Hispanics, scores in both conditions significantly increased from baseline to end of class (p < .001). However, for Blacks, there was an interaction between time and group, suggesting that those in the memory treatment condition were more likely to have higher memory performance scores than those in the health treatment condition. However, for non-Blacks (i.e., Caucasians and Hispanics combined), there was not a meaningful difference in memory performance scores on the Rivermead between the treatment and comparison conditions.

On the DAFS, older, less educated Hispanic and Black participants had lower scores at baseline (Figure 5). We found significant effects for education, F(1, 198) = 21.96, p = .000, η² = .10; ethnicity, F(2, 198) = 17.03, p = .001, η² = .15; treatment condition, F(1, 198) = 9.17, p = .003, η² = .04; and an interaction of Treatment Condition × Ethnic Group, F(2, 198) = 5.56, p = .004, η² = 05. Participants with initially higher DAFS scores and those who attended more classes improved more from baseline to the end of the study than other participants (p < .01 and p < .05). Those with fewer years of education improved more than others (p < .05) from baseline to end of the study, and Blacks improved more than Whites (p < .01).

Self-Report Measures

The results of all self-report measures are included in Table 3. On memory self-efficacy, participants in the memory training condition reported significantly higher scores than those in the health training condition at baseline (p < .05). In addition, younger participants rated their memory self-efficacy higher than older participants (p < .01). We found significant effects for treatment condition, F(1, 188) = 13.02, p = .000, η² = .07; education, F(1, 188) = 8.11, p = .005, η² = .04; and an interaction between ethnicity and treatment condition, F(2, 188) = 3.76, p = .025, η² = .04. We also found variability between groups across time, with Whites scoring below other ethnic groups in the memory training condition and above them in the health training condition (Figure 6).

Table 3.

Descriptive Statistics and Relevant p Values for Self-Report Outcomes of Minority Participants

Memory self-efficacy
Hispanic	Health	M	40.09	34.63	38.00	.56	.68	.14	.35
		SD	12.97	19.65	20.13
	Memory	M	55.19	54.97	46.13
		SD	23.33	25.08	19.29
Black	Health	M	42.11	36.68	38.39	.34	.53	.22	.93
		SD	10.05	13.99	10.65
	Memory	M	54.71	52.71	50.43
		SD	26.60	17.91	20.78
Total sample	Health	M	46.02	45.35	47.38
		SD	16.28	16.88	15.65
	Memory	M	49.44	51.60	52.04
		SD	18.82	18.29	18.54
State anxiety
Hispanic	Health	M	30.50	30.63	36.00	.54	.50	.13	.44
		SD	8.40	11.94	12.81
	Memory	M	31.00	28.17	32.83
		SD	9.52	9.23	9.62
Black	Health	M	29.69	27.08	26.54	.35	.62	.60	.36
		SD	9.40	7.77	6.65
	Memory	M	30.78	30.33	31.67
		SD	9.93	9.47	13.35
Total sample	Health	M	31.36	32.90	32.54
		SD	8.18	8.27	9.75
	Memory	M	32.32	31.37	31.62
		SD	9.32	8.39	9.88
Trait anxiety
Hispanic	Health	M	30.94	34.63	34.31	.48	.24	.31	.67
		SD	9.36	9.09	13.43
	Memory	M	31.89	30.56	33.28
		SD	10.49	9.43	13.75
Black	Health	M	30.23	34.69	30.08	.27	.25	.31	.37
		SD	10.84	10.24	9.38
	Memory	M	32.50	33.75	29.75
		SD	9.61	8.36	8.53
Total sample	Health	M	31.36	32.90	32.54
		SD	8.18	8.27	9.75
	Memory	M	32.32	31.37	31.62
		SD	9.32	8.39	9.88
Depressive symptoms
Hispanic	Health	M	9.06	9.81	12.69	.87	.77	.09	.43
		SD	7.64	8.21	12.13
	Memory	M	9.94	9.06	11.28
		SD	7.30	6.39	11.92
Black	Health	M	9.31	12.15	8.69	.46	.35	.39	.67
		SD	7.05	9.06	9.47
	Memory	M	12.89	11.33	11.11
		SD	9.06	10.59	6.23
Total sample	Health	M	9.29	9.88	9.71
		SD	6.63	7.63	8.77
	Memory	M	8.94	8.77	9.24
		SD	6.83	7.08	8.81

Note: Memory self-efficacy = Memory Self-Efficacy Questionnaire; state and trait anxiety = Spielberger State–Trait Anxiety Inventory; depressive symptoms = Center for Epidemiological Studies–Depression scale.

On anxiety, we found significant effects for education, F(1, 196) = 5.55, p = .020, η² = .03. There were no other significant differences on anxiety scores across groups (Figures 7 and 8).

On depression (Figure 9), we found significant effects for education, F(1, 197) = 5.66, p = .018, η² = .03. There were no other significant differences on depression scores across groups.

Discussion

Those who enrolled in the study and attended the eight training sessions and the booster sessions and completed all five testing interviews were on average 74 years of age and more likely to be females than males (183 vs. 60). Of the 265 people who were eligible and enrolled, 189 were Caucasian, 30 were African American, and 46 were Hispanic.

On global cognition, participants in the intervention group outperformed the health training participants at postclass. Whites scored significantly higher than Hispanics or Blacks. The intervention produced a proximate improvement in cognition, verbal memory, and everyday memory performance.

Participants improved on everyday memory performance through the booster sessions. Blacks increased by almost 14 points postbooster sessions and made the largest gains and Hispanics improved an average of 5 points. At study completion, Blacks maintained their gains of approximately 10 points and Hispanics of approximately 8 points on verbal memory. Those with less education made the greatest gains on cognition (p < .05), as did those who were Black, and this trend continued to the end of the study, controlling for other factors.

Depression scores tended to rise and fall over the 26-month enrollment period. However, most people (96%) did not report high levels of depressive symptoms at any time in the study. At all four measurement points, only 4% of the sample scored above the cutoff, indicating high symptoms of depression. Anxiety was particularly apparent among individuals who did not complete the study (19% dropped out of the memory training and 17% dropped out of the health training). Interestingly, those in the memory training condition appeared to decrease in trait anxiety from baseline to post-test, whereas those in the health training group increased.

Our analyses suggested that minority participants received differential benefits from the memory training; however, this remains speculative because there have been few published findings of cognitive interventions with minority elders (McDougall, 1998). Although the intent of this intervention project was to examine the impact of memory training on the cognitive functioning of seniors, fortuitously, participants in both the memory and the health training conditions improved on many of the outcome measures. Some of the results differed based on the demographic characteristics of subjects (e.g., older subjects made greater gains in visual memory over the course of the study). It is also worth noting that some of the improvements persisted only through the end of classes, for example, improvements in IADLs.

Testing with minority elders was often done over many days to accommodate their frustration with the complex battery of measures. Practice effects from repeated testing may have accounted for some of the differential gains for Blacks and Hispanics, or the intervention may have been more effective for minority participants (Lowe & Rabbitt, 1998; Wilson et al., 2009).

One study tested visual memory with the BVMT-R in a sample of community-residing elderly, and consequently, little normative data exist for comparison purposes (Gale et al., 2007). This is the first memory training study to provide data for the BVMT-R with a sample of minority elderly. Our participants scored much higher on delayed recall (33 vs. 6) than older adults in another study with the same average age but with a 3-point lower MMSE score. The HVLT-R and the BVMT-R were the most difficult performance measures in our study. Each requires three learning trials and a delayed recall. Whites appeared to perform better overall, but minority groups showed the greatest improvement.

The everyday memory scores of participants did not change, although our participants showed great variability in performance. At baseline, almost half (47%) of the participants had poor memory; 75 individuals (28%) scored in the normal range on the RBMT (McDougall, Becker, & Arheart, 2006) and the remaining 67 individuals (25%) scored in the mild and severely impaired categories. Individuals with normal memory function were younger, had more years of school, and had higher MMSE scores. We were surprised at the variability in performance because individuals with dementia were excluded from participation.

There was substantial variability on memory self-efficacy scores. Participants in the memory intervention condition had greater memory self-efficacy at baseline than those in the health promotion condition; younger participants also began the study with a more positive view. Caucasian and Hispanic participants in the memory training condition significantly improved their self-efficacy beliefs at postclassroom. African Americans’ scores dropped on average approximately 3 points postclassroom; however, they increased substantially postbooster sessions by 12 points, making the largest gains. Over time, the treatment gains diminished, and those Blacks in the memory training condition (n = 8) who completed the study compared with those in health training (n = 11) had substantially lower memory self-efficacy scores. For Hispanic completers, there were equal numbers in both conditions (n = 16). Those in the memory training groups declined on average by 5 points, whereas those in the health training groups maintained their gains over time. Not surprisingly, as in other studies, those Whites with lower memory self-efficacy at baseline made greater gains over the course of the study than those who began the study with higher scores in both training conditions (McDougall, 2002; West, Bagwell, & Dark-Freudeman, 2008). Whites made relatively greater gains by approximately 4 points on memory self-efficacy scores from beginning to end of the study.

One aspect of the study that cannot be overlooked is the social engagement that was built into the intervention. The small group format in both the memory and the health training conditions enhanced performance accomplishment and verbal persuasion and facilitated reducing negative beliefs related to cognitive aging, all components that build self-efficacy. We did not include a measure of social engagement; however, anecdotal comments from participants validated that this was a major reason for study retention in both conditions. Other studies have also documented the benefits of social engagement related to the maintenance of cognitive function and everyday skills (Barnes, Mendes de Leon, Wilson, Bienias, & Evans, 2004; Barnes et al., 2005; Bassuk et al., 1999; Fried et al., 2004; Saczynski et al., 2006; Wilson et al., 2003).

The DAFS was a reliable and valid measure of the performance of basic IADLs. These participants may have learned how to improve their DAFS scores over time, but we observed ceiling effects at the first administration before practice could have affected performance. The measure was not sensitive to somewhat more complex activities of daily living, such as medication administration. Our participants, however, had high scores, without much variability. Deficits observed in the complex IADLs might signal early cognitive decline and the need for additional testing of cognitive functioning. A more challenging measure of IADLs might also be more sensitive to changes following interventions designed to improve memory functioning in the elderly. In a pilot study, we have revised the DAFS by adding greater complexity to the medication subscale (McDougall, Becker, Vaughan, Acee, & Delville, 2009).

Limitations

The lack of a no-treatment control condition was a limitation in study design. Furthermore, the generalizability of the findings from the small sample of African Americans and Hispanics is not clear. A major limitation of this study was our inability to determine how the participants used the information gained in the classes. After each class, each participant wrote down some aspect of the learning or an essential point for the day, such as a memory strategy. The individual participant was scored with either a plus or a minus (0, 1) for the class session. Between the eight-session CBMEM, and four booster sessions, a participant could have a possible score between 0 and 12. Data from the classroom tests were used to examine the relationships between classroom performance and memory mastery. The minority participants were unable to follow through on this aspect of treatment outcomes and we therefore chose not to include these scores from the Caucasian subsample. To meet this aspect of treatment fidelity, we included those participants in the analyses who attended six of the eight sessions or received 75% of the dosage.

Conclusions

The results suggest that simply combining all individuals into one group would mask changes between minority group participants. Given the unequal sizes of the three groups of participants, to speculate about clinical significance would be premature at this time. The question of why Black and Hispanic participants often made greater improvements needs further exploration. Future research should examine ethnic and racial groups separately, because different patterns were observed for Blacks and Hispanics, and should consider the moderating effects of race and ethnicity on memory training. Despite those questions, the study contributes new knowledge to the cognitive aging and intervention literature and provides a foundation for further examination of the effects of memory training on aging racial and ethnic minorities.

Funding

Support for this research was provided by the National Institute of Aging (NIA) grant R01 AG15384. Trial Registration at ClinicalTrials.gov (NCT00094731, http://www.clinicaltrials.gov/ct/show/NCT00094731).