Improving Prospective Memory in Healthy Older Adults and Very Mild Alzheimer’s Disease Patients

Abstract

Objectives

Prospective memory (PM) is the ability to remember to execute future goals (e.g., remembering to take medications), which plays an important role in independent living. Failures in PM are particularly prevalent in older adults with very mild Alzheimer’s disease (AD), and deficits are also observed in healthy older adults for certain PM tasks. This study tested the utility of a memory encoding strategy for improving PM in healthy older adults and those with very mild AD.

Design

Participants were randomly assigned to either an encoding strategy condition or a standard encoding condition.

Setting

All participants were community-dwelling individuals enrolled in a longitudinal study conducted at an Alzheimer’s Disease Research Center. Testing took place at the center and in a university testing room.

Participants

A sample of healthy older adults (Clinical Dementia Rating = 0.0, n = 38) and those who are classified to be in the very mild stage of AD (Clinical Dementia Rating = 0.5, n = 34) participated.

Intervention

We used a simple strategy (“If I see Cue X, then I will perform Intention Y”) to strengthen PM encoding and reduce the probability of forgetting to execute one’s future plans.

Measurements

PM was assessed with Virtual Week, which is a laboratory task that requires the simulation of naturalistic PM tasks (e.g., take medication at breakfast).

Results

The encoding strategy significantly reduced PM failures in healthy older adults and those with very mild AD. Importantly, the encoding strategy was effective regardless of the individual’s episodic memory ability.

Conclusion

This encoding strategy was successful in reducing PM errors in healthy and very mildly demented older adults with a range of memory abilities.

INTRODUCTION

Prospective memory (PM) refers to remembering to perform an intended action in the future. There are important real-world implications associated with failures in PM, including maintaining health (e.g., forgetting to take medication) and safety (e.g., forgetting to turn off an oven). Sixty-two percent of adults attending a memory clinic reported that PM errors were one of the most important memory failures they experienced.¹ Furthermore, PM ability predicted self-reported medication adherence in older adults.² PM deficits are particularly pronounced in patients with Alzheimer’s disease (AD),^3–5 even in its very mild or preclinical stages.^6–7 Thus, it is critical to develop strategies for reducing PM errors within AD populations.⁸ The present study investigated the utility of an encoding strategy for boosting PM in healthy older adults and patients with very mild AD.

Minimal research has tested whether encoding strategies improve PM within clinical populations.^9–11 In healthy populations, implementation intention (II) encoding has produced benefits to PM in laboratory and naturalistic settings.^12–15 An II is a simple encoding strategy in which individuals verbalize a retrieval cue and link the cue to their plan (e.g., stating “When I sit down for dinner, I will take my blood pressure medication.”). II’s increase the likelihood that healthy older adults will remember to perform a variety of important future intentions, including remembering to monitor blood glucose levels.¹⁴

Related to the present experiment, Lee and colleagues ⁸ reported that II encoding improved PM for very mild AD individuals on a simple laboratory task. Though encouraging, this initial experiment used PM tasks that were designed to be minimally challenging so that AD participants would perform at relatively high levels. Accordingly, it is important to investigate whether II’s improve PM over a range of more realistic and challenging tasks.

The present study tested for II benefits to healthy older adults and patients with very mild AD using the Virtual Week task. Virtual Week is a highly reliable, laboratory-based PM task¹⁶ that has successfully been used in a range of clinical settings, including patients with mild cognitive impairment (MCI) and early dementia¹⁷ and is sensitive to II encoding benefits.¹⁸ In Virtual Week, participants simulate the execution of intended actions within the context of a board game with each circuit representing a virtual “day.” One strength of this task is that it creates a naturalistic context by having participants complete intentions that would plausibly be completed in daily life. Another advantage is the inclusion of a variety of PM task types (regular or irregular) and different PM cue types (time-based or event-based), which are described in the Method section. Healthy older adults typically perform regular tasks and event-based tasks at a higher level than irregular tasks and time-based tasks,^{16, 19} but minimal data on this question are available for very mild AD patients.¹⁷

METHODS

Design and Participants

The experiment was a mixed-factor design with Clinical Dementia Rating (CDR; healthy/very mild AD) and Encoding Condition (standard/II) as between-participants factors. PM Task Type (regular/irregular) and PM Cue Type (event/time) were manipulated within-participants.

A notable strength of this study is the well-defined characterization of our sample. All participants were classified using the CDR scale that consists of a 90 minute clinical interview with the patient and collateral source,²⁰ conducted by clinicians at an Alzheimer’s Disease Research Center (ADRC). A CDR rating of 0.0 reflects no dementia and a CDR rating of 0.5 reflects very early stages of AD. The CDR scale has demonstrated excellent validity and reliability (93% diagnostic accuracy). ²¹ A CDR rating of 0.5 is associated with small changes in one’s daily functioning across a variety of domains. A CDR rating of 0.5 is similar to diagnoses of MCI, and most individuals classified in either category progress to having AD.²²

Seventy-two participants (38 with CDR of 0.0 and 34 with CDR of 0.5) were recruited from an ADRC where they were enrolled in a larger longitudinal study. Participants (aged 65–85) had normal or corrected vision and were screened for depression and reversible dementias. Table 1 presents the demographic and neuropsychological characteristics of the sample. Mean age was higher for CDR 0.5 than CDR 0.0 participants; thus, age was statistically controlled in the analyses. Though we employed random assignment to encoding conditions using a random number generator, participants in the II condition had significantly higher scores on the Associative Memory test. To ensure that this baseline difference did not impact the main outcomes, we conducted a follow-up to the primary analysis using associative memory as a covariate in an analysis of covariance (ANCOVA) and the results were unchanged.

Table 1.

Demographic and Neuropsychological Characteristics

	CDR 0.0		CDR 0.5		p value
	Standard Encoding	II Encoding	Standard Encoding	II Encoding	CDR Status	Encoding Condition
N (Female)	19(8)	17(10)	17(6)	14(4)
Age (Year)	74.8(±6.5)	73.2 (±6.2)	78.6(±6.4)	80.5(±6.1)	<.01	.89
Education	14.7(±2.9)	14.9(±2.5)	14.9(±2.5)	15.2(±2.8)	.72	.72
Mini Mental State Exam	28.8(±1.3)	29.1(±1.2)	26.5(±3.2)	24.9(±4.0)	<.01	.31
Associate Memory	12.9(±4.8)	16.3(±2.4)	6.5(±4.8)	9.7(±4.5)	<.01	<.01
Selective Reminding Test	47.9(±0.3)	47.7(±0.6)	45.4(±3.2)	44.2(±6.8)	<.01	.39
Forward Digit Span	6.8(±1.0)	7.0(±1.0)	6.4(±1.1)	6.5(±0.9)	.05	.54
Backward Digit Span	4.8(±1.4)	5.1(±1.3)	4.7(±1.1)	3.6(±0.8)	.01	.22
Trail Making Test-Part A	36.4(±14.3)	31.4(±9.4)	44.7(±17.1)	55.5(±19.5)	<.01	.45
Trail Making Test-Part B	92.7(±40.9)	77.2(±22.2)	123.9(±46.3)	144.0(±37.7)	<.01	.77
WMS Logical Memory	13.6(±3.2)	14.2(±4.1)	7.2(±4.3)	7.7(±5.3)	<.01	.59
Letter Number Sequencing	8.2(±3.5)	9.3(±2.0)	6.7(±2.7)	6.1(±3.3)	<.01	.74

Note: CDR refers to Clinical Dementia Rating (0.0 = healthy, 0.5 very mild Dementia) and II refers to implementation intention. Means and Standard Deviations (in parentheses) are presented for demographic and neuropsychological tests. Some participants did not have measures of Trail Making–Part B (1), Selective Reminding Test (3), and Letter Number Sequencing (5)

Participants provided informed consent in accordance with the Institutional Review Board at Washington University in St. Louis, and were compensated $10 per hour. Five participants (all CDR 0.5) were excluded from analyses due to illness (1), poor comprehension of instructions (1) or failure to complete the entire experiment due to fatigue (3).

Materials and Procedure

A computerized version of the board game, Virtual Week, was used to measure PM performance.^{17, 19} Participants completed one practice and two experimental “virtual” days, with four PM tasks per day. Participants rolled the die to progress the token around the board using a mouse click.

Each participant was tested on a computer with a touch-screen monitor in a testing room with an experimenter. After completing two laboratory memory tasks⁸, participants were told they would play a computerized game called Virtual Week where they would simulate daily activities and remember to perform various tasks. Participants completed a trial day in which the game was explained with detailed pop-up help messages and the experimenter ensured that they understood the procedures. Monday and Tuesday, however, were completed without any assistance. At the beginning of each day, participants were asked to touch the start card on the screen to receive instructions for four different PM tasks. Both days had the same two regular PM tasks (take medication at 12 noon and at the dinner event card) and two different irregular PM tasks each day (e.g., get haircut at 2 pm, drop in dry cleaning while out shopping); half of the regular and irregular tasks had event-based cues and the other half had time-based cues. Event-based cues were present in the title of the target event card (e.g., “dinner” was in the title of the dinner event card). The time-based tasks required participants to monitor a virtual clock located in the center of the screen that was calibrated to the position of the token on the board.^{23, 26}

To perform each PM task, participants were instructed to press the perform task button on the screen and select the appropriate action from the drop-down menu consisting of a list of target and distractor actions. Participants were encouraged to perform each PM task on time (i.e., within a virtual hour for time-based tasks and before reaching the next event card for event-based tasks) but were told to perform the task even if they were late.

In the standard encoding condition, participants read the instructions out loud (e.g., “phone plumber at 4pm”) for each PM task and proceeded when they were ready. Participants reviewed the instructions for each of the four PM tasks separately each day. In the II encoding condition, the parameters were identical to the standard condition except that participants were explicitly told to read a statement three times out loud for each PM task: “When I see the [PM cue], then I will press the perform task button and select the [PM task].” They were also told to mentally imagine themselves performing that task. Each PM instruction was presented for sixty seconds to allow sufficient time to complete the II encoding.

After completing Virtual Week, participants were administered a retrospective memory questionnaire in which all of the PM tasks and cues were presented in a random order on a sheet of paper. Participants were asked to match each PM task to its corresponding cue. This enabled the assessment of the retrospective memory component of the PM task (i.e., which intention needed to be remembered). Participants were then thanked, compensated, and debriefed.

Data Analysis

The dependent measure for the primary analysis was the proportion of missed responses for each of the four categories of PM tasks: regular event-based, regular time-based, irregular event-based and irregular time-based. A mixed-factor ANCOVA was conducted with centered age as the covariate, CDR status (0.0/0.5) and Encoding Condition (standard/II) as between-participants factors, and PM Task Type (regular/irregular) and PM Cue Type (event/time) as within-participants factors. A similar ANCOVA was conducted with proportion correct on the retrospective memory test as the dependent measure.

The neuropsychological assessment included measures of episodic memory, such as logical memory story A - immediate, associative memory, reading span, and digit forward and backward. An episodic memory composite score was created using performance on these five measures: based on their average z-scores. A hierarchical linear model was evaluated to determine the predictive utility of episodic memory for PM errors, above and beyond CDR status, age, education, and Encoding Condition. The interaction between episodic memory and Encoding Condition was evaluated to determine if II encoding would be useful for individuals with varying episodic memory ability.

Analyses were conducted using the Statistical Package for the Social Sciences (Chicago, IL), Version 18.²⁴ Statistical tests were two-tailed with an alpha level of .05. Effect sizes were estimated using partial eta squared (η²_p).

RESULTS

Analysis of missed PM responses

CDR 0.5 participants (M=.73, SD=±.32) committed significantly more PM errors than did CDR 0.0 participants (M=.23, SD=±.32); see Table 2 for ANCOVA results. The participants in the II encoding condition (M=.35, SD=±.39) made significantly fewer PM errors relative to those in the standard encoding condition (M=.60, SD=±.39) (see Figure 1 for means of each CDR group). There were no significant effects of PM Task, Cue, or age on the number of PM errors. There was, however, a significant interaction between CDR status and PM Task Type, such that CDR 0.0 participants missed significantly fewer regular tasks (M=.16, SD=±.32) than irregular tasks (M=.30, SD=±.37), whereas CDR 0.5 participants missed a similar amount of regular (M=.75, SD=±.35) and irregular tasks (M=.71, SD=±.34). There were no other significant interaction effects, (all Fs ≤ 3.14, ps .08, η²_p s ≤ .08).

Table 2.

Relevant Statistics for Prospective and Retrospective Memory Errors^a

	Fb	CIs	p-value	η2pc
Prospective Memory Errors
CDR Status	45.8	0.35, 0.65	<.01	.43
Encoding Condition	11.1	0.10, 0.40	<.01	.15
Task Type	2.5	−0.11, 0.01	.12	.04
Cue	1.4	−0.08, 0.02	.25	.02
Age	1.2		.27	.02
CDR X Task Type	7.8		< .01	.11
CDR X Encoding condition	1.1		.30	.02
Retrospective Memory Errors
CDR Status	27.0	0.18, 0.41	< .01	.30
Encoding Condition	0.6	−0.07, 0.16	.46	.01
Age	5.9		.02	.09
CDR X Encoding Condition	0.1		.79	<.01

^aThe statistical tests were derived from ANCOVAs, with age as the covariate.

^bThe degrees of freedom were 1 in the numerator and 62 in the denominator for all F values listed in the table.

^cη²_p = partial eta squared, an index of effect size.

Figure 1.

Mean proportion of missed prospective memory responses collapsed across regular and irregular prospective memory task types and event and time-based cues for the II and standard encoding conditions, for the CDR 0.0 and 0.5 groups The error bars indicate ± 1 standard error of the mean.

Impact of Encoding Condition on Retrospective Memory

Retrospective memory performance (see Table 2 for ANCOVA results) was comparable across the standard encoding (Ms(SDs) = .88(±.16) and .58(±.30) for CDR 0.0 and 0.5, respectively) and II encoding conditions (Ms(SDs) = .92(±.11) and .62(±.35) for CDR 0.0 and 0.5, respectively). Thus, the II encoding benefit was driven by improved prospective, rather than retrospective, memory of the intentions. As expected, retrospective memory declined with increasing age and CDR status.

Episodic Memory Ability and Strategy Effectiveness

Episodic memory ability was a significant predictor of PM errors (β=−.25, p=.02) such that individuals with better episodic memory committed fewer PM errors. Importantly, the interaction between Encoding Condition and episodic memory (β=.10, p=.38), did not account for significant variability in PM errors, suggesting the encoding strategy was effective regardless of episodic memory ability.

DISCUSSION

We investigated whether an II encoding strategy could improve PM in healthy older adults (CDR=0.0) and patients with very mild AD (CDR=0.5). This is a critical question because PM is pertinent to maintaining independent functioning and many patients may believe they cannot buffer against AD-related memory losses. The II encoding strategy was successful in restoring some PM function in this clinical group (and healthy older adults) in the present study. The interaction between CDR status and the type of PM task (regular/irregular) is notable because it speaks to the robust PM deficit observed even in the very earliest stages of AD. CDR 0.0 participants missed fewer regular tasks than irregular tasks.¹⁶ By contrast, CDR 0.5 participants did not differ across regular and irregular tasks. This is consistent with the idea that automatic-like PM processes are impaired in very mild AD but not in healthy aging.^{7, 16} The disruption of more habitual prospective remembering (e.g., remembering daily medications) could be particularly problematic for sustaining independent living in very early AD. Critically, we found that strengthening the encoding of PM intentions through an II strategy effectively reduced errors, regardless of the type of PM task.

We contend that the observed benefits in the II group are due to the use of more elaborative encoding of the PM intention. Random assignment disfavors alternative interpretations for the II effect, such as participants in II conditions being more conscientious or motivated. Additionally, the level of social interaction was held constant as participants from both groups spent the same amount of time interacting with the experimenter.

It is important to investigate the potential efficacy of behavioural interventions for AD patients. The most widely studied behavioural strategy for improving PM in AD patients is spaced-retrieval.^{11, 25–26} This technique restores some PM function in AD patients; however, the II strategy may have several advantages over spaced retrieval (e.g., IIs are easier to use and take less time to encode a given intention). Future research is needed, however, to compare the utility of different behavioural strategies. Notably, it is unclear whether behavioural interventions would provide similar benefits to cognitive performance as those observed in pharmacological treatments.

Limitations and Future Directions

There was potential for some training effects in this study; namely, participants completed two PM tasks prior to Virtual Week, and they were given similar encoding instructions in these tasks⁸. Though training effects have been observed within PM,²⁷ we would not expect such effects to differ across experimental conditions or to be fundamentally different from natural training effects that accrue in everyday settings in which patients repeatedly employ the II strategy.

Another potential limitation was the use of a laboratory-based PM task, which may not reflect everyday PM behaviors. To mitigate this concern we chose a laboratory task, Virtual Week, that was created to simulate real-world PM, and performance on this task was found to be indicative of real-world PM in terms of activities of daily living.^27–29 Furthermore, IIs have produced benefits in real-world PM tasks for healthy older adults.¹⁴ Regardless, the findings from the present study only speak to the utility of IIs in a laboratory setting, and future research is critical for determining if IIs will lead to reduced PM errors in the everyday lives of AD patients. Finally, our study focused on patients in the very early stage of dementia, and it is unclear whether observed II benefits would also be present for individuals in later stages of AD.

Conclusions

A simple encoding strategy was effective in restoring some PM function in older adults, regardless of episodic memory ability and CDR status. PM failures compromise one’s quality of life, particularly for individuals struggling with everyday activities³⁰, and developing simple behavioural interventions could reduce the concerns that are observed in older adults.¹ Furthermore, future research should investigate whether these interventions could reduce some of the potential burden felt by caregivers of AD patients. These results indicate the need for clinical trials that investigate the utility of strengthening the encoding of PM intentions that are carried out in patients’ home settings.