Supporting Strategic Processes Can Improve Time-based Prospective Memory in the Laboratory Among Older Adults with HIV Disease

Abstract

Objective:

Older adults with HIV disease demonstrate moderate deficits in time-based prospective memory (PM), which is the strategically demanding ability of remembering to perform a task at a specific time. Using theory from the PM literature, we hypothesized that supporting strategic processes would improve time-based PM in the laboratory among HIV+ older adults.

Method:

One-hundred forty-five HIV+ participants were randomly assigned to a control condition or an experimental group in which strategic processing was supported at encoding (i.e., implementation intentions and visualization), monitoring (i.e., content-free cuing), and/or cue detection (i.e., auditory alarm). The HIV+ control group and 58 seronegative participants completed two ongoing language tasks with a time-based PM requirement. The HIV+ experimental groups underwent counterbalanced time-based PM trials under both control and strategically supported conditions.

Results:

The HIV+ cue detection group showed a large within-subjects improvement, which was strongly related to lower scores on separate clinical time-based PM measure and was accompanied by a large reduction in clock checking behavior. Results also revealed a small within-subjects improvement in time-based PM in the encoding condition.

Conclusions:

Supporting strategic encoding and cue detection processes in the laboratory can improve time-based PM deficits in older HIV+ adults, which may inform the development of more naturalistic PM-based interventions to enhance health behaviors.

HIV disease is associated with moderate deficits in time-based prospective memory (PM; Carey et al., 2006), which is a multi-faceted, strategically-demanding cognitive ability (e.g., remembering to take a prescribed medication at 2pm). Conceptual models of PM agree that there are several broad stages of time-based PM, which include: (1) forming an intention that is linked to a retrieval cue (e.g., a specific time); (2) retaining this cue-intention pairing over a delay interval during which the passage of time is monitored as one is engaged in an ongoing task(s); (3) detecting the correct PM retrieval cue (i.e., the proper time); and (4) retrieving the appropriate task action from retrospective memory and executing it (e.g., Kliegel et al., 2008; Kvavilashvil & Fisher, 2007; McDaniel & Einstein, 2000). The Multiprocess view (McDaniel & Einstein, 2000) suggests that time-based PM places greater demands on voluntary, strategic monitoring processes served by prefrontostriatal networks (e.g., Oksanen et al., 2014; Okuda et al., 2007) than does focal event-based PM for which more salient environmental cues encourage automatic/spontaneous retrieval served by mediotemporal networks (Gordon et al., 2011). Consistent with this proposition, it is generally agreed that time-based PM is highly reliant on the transient, strategic cognitive control processes required to monitor time (Huang, Loft & Humphreys, 2014; Waldum, & Sahakyan, 2013). For example, tasks that require participants to make a target response after a time duration has elapsed negatively impacts concurrent task performance, indicating that prospective timing can be resource demanding (Taatgen, van Rijn, & Anderson, 2007). In line with this, individuals can be slower to perform ongoing tasks when they have time-based PM demands compared to when performing ongoing tasks alone (referred to as PM “costs”; Hicks, Marsh, & Cook, 2005; cf. Oksanen et al., 2014).

HIV preferentially disrupts frontostriatal systems (e.g., DuPlessis et al., 2014) and thus HIV-associated PM deficits are greater on strategically-demanding time-based tasks than for most event-based tasks (Avci et al., 2018). From the Multiprocess view, HIV-associated time-based PM deficits appear to be driven by failures in strategic monitoring and cue detection, as supported by findings of increased PM omission errors that coincide with a lack of costs to ongoing tasks and normal post-test recognition of PM target times (e.g., Carey et al., 2006; Morgan et al., 2012). HIV also impacts strategies that individuals can use to control time-based PM task requirements. One strategy is to make prospective timing estimates; indeed, persons with HIV-Associated Neurocognitive Disorders (HAND) show deficits in time estimation and production, which are associated with lower time-based PM (Doyle et al., 2015a). Another strategy might be to make volitional clock checks during a time-based PM task (Harris & Wilkins, 1982), which are associated with frontoparietal activation (Oksanen et al., 2014) and can reduce the cognitive resource demands associated with time-based PM tasks (Huang et al. 2014). Individuals with HAND make fewer clock checks during time-based PM tasks, which is associated with worse time-based PM accuracy and executive dysfunction (Doyle et al., 2013).

The increasing number of older adults living with HIV (Smit et al., 2015) may be particularly vulnerable to impairment in time-based PM. HIV and aging tend to have additive effects on brain structure and function (e.g., Valcour et al., 2004). Among typically aging adults, older age is reliably associated with declines in time-based PM (Henry et al., 2004) that increase risk of poor naturalistic health compliance (Kamat et al., 2014), functional problems (Hering et al., 2018), and lower quality of life (Woods et al., 2015). In HIV disease, older age is correlated with lower PM (Poquette et al., 2013) and less frequent clock checking during PM tasks (Doyle et al., 2013). At the group-level, the combined effects of HIV and aging on time-based PM in the laboratory are additive (Weber et al., 2011), exacerbated by long delays (Avci et al., 2016; Morgan et al., 2012) and independent of global neurocognitive impairment, comorbidities, and HIV disease severity (Avci et al., 2016). This is clinically relevant because HIV-associated deficits in time-based PM are a reliable, independent predictor of everyday functioning (Woods et al., 2008), health behaviors such as medication adherence (e.g., Doyle et al., 2015b; Sheppard et al., 2016; Woods et al., 2009), and quality of life (Doyle et al., 2012).

An emerging body of work has therefore examined approaches to bolster PM performance in HIV disease, but none have focused on time-based PM or older adults. In 60 young adults with HIV, Faytell et al. (2017) observed that a brief visualization/imagery exercise at encoding improved event-based PM in the laboratory relative to a control condition, primarily in HIV+ individuals with poor clinical time-based PM. Two studies have focused on enhancing monitoring to event-based cues as a means of improving PM in HIV disease. In a sample of HIV+ young adults, Loft et al. (2014) demonstrated that introducing a short delay (600ms) between cue onset and when the individual was allowed to respond improved non-focal PM, presumably because it reduced strategic demands by providing participants additional time to examine the ongoing task stimuli for the PM cue (Heathcote, Loft & Remington, 2015). In parallel experiments, Woods et al. (2014) bolstered the non-focal event-based PM of HIV+ youth with HAND and substance use disorders by emphasizing the importance of the PM task relative to the ongoing task, which increased their attention allocation to cue detection. Faytell et al. (2018) investigated the benefits of encoding (i.e., calendaring) and cue detection (i.e., a mobile telephone alarm) to improve naturalistic time-based PM in 47 younger HIV+ adults. Results showed large effect size benefits for participants in the cue detection support condition, regardless of whether they used the calendar. A more recent study observed a clinically significant reduction in viral load in 73% of HIV+ youth 24 weeks after a brief training in implementation intentions, visualization, and salient cue-intention pairings related to medication adherence (Pennar et al., 2019). The effects of this single-arm intervention were slightly weaker, but still notable in HIV+ youth with lower clinical PM scores (57.1% reduction vs 87.5%). Taken together, these studies suggest that supporting the strategic aspects of PM in younger adults with HIV can improve event-based PM in the laboratory, semi-naturalistic time-based PM, and health behaviors that depend on PM.

The current study aimed to enhance the time-based PM functioning of older HIV+ adults in the laboratory by systematically supporting strategic processes at the encoding, monitoring, and cue detection phases. Our approach to improving laboratory-based PM was informed by Kliegel’s neuropsychological phase model of PM (Kliegel et al., 2008) and the dynamic extension of the Multiprocess view (Scullin et al., 2013). The Dynamic Multiprocess Framework posits that strategic monitoring processes and automatic/spontaneous retrieval processes dynamically interact to support PM. According to this framework, internal thoughts or attention to contextual cues in the environment associated with PM intentions can spontaneously bring intentions to mind and prompt the engagement of the strategic monitoring processes (in this case, prospective timing estimates, or checking the external clock) required for PM retrieval. As discussed earlier, older individuals and those with HIV disease typically strategically monitor time less accurately when completing time-based PM tasks (Doyle et al., 2013, 2015a); therefore we reasoned that laboratory supports designed to prompt strategic time monitoring could benefit their time-based PM. We adopted three theory-based approaches to improving PM, which are outlined below.

At the encoding phase of a typical time-based PM task, an individual forms a specific intention to be executed at a specific time, which involves elements of planning and retrospective memory binding (Kliegel et al., 2008). To support the encoding phase, we examined the benefits of implementation intentions with visualization, which traditionally involves an if-then statement (e.g., “When it is 8am tomorrow, then I will take my medicine”) accompanied by a brief visualization exercise in which one imagines successfully performing the intention at the right time. This support should strengthen the time-based cue-intention association, which may be effective in helping individuals plan and maintain the intent to strategically monitor time near the correct time for PM retrieval. In line with this, this type of intervention at encoding has been shown to provide moderate benefits to PM across the literature (Chen et al., 2015), including in HIV+ youth (Faytell et al., 2017; Pennar et al., 2019).

At the monitoring phase of a time-based PM task, which requires time perception and cognitive flexibility, an individual must make prospective timing estimates or check the clock, the frequency of which increases as the target time nears (Kliegel et al., 2008; Harris & Wilkins, 1982). To support monitoring, we examined the potential benefits of content-free cuing (Fish et al., 2007), which involves intermittently prompting individuals to strategically monitor (Scullin et al., 2013, 2019) in a manner that does not overtly specify the PM goal. For example, training individuals to “Stop, Think, Organize, and Plan” with regard to their PM intentions when prompted with only the acronym “STOP” has shown to prompt strategic monitoring and increase PM accuracy in brain injury (e.g., Fish et al., 2007) and enhance naturalistic multitasking in HIV+ substance users with executive dysfunction (Casaletto et al., 2016).

The cue detection phase is the cardinal strategic aspect of a time-based PM task and places demands on cognitive flexibility and subtle signal detection, as an individual must become aware that the proper time to fulfil the intention has arrived (Kliegel et al., 2008). To directly support this phase, we examined the potential benefits of a salient alert near the time of the PM target. The salient alert was hypothesized to greatly reduce ongoing time monitoring demands because it provides a clear indication when the individuals needs to switch from the ongoing task to the time-based PM intention (Scullin et al., 2013). In this way, the auditory alert serves the conceptual and practical purpose of converting the highly strategically demanding time-based task into a much less demanding (i.e., more automatic) event-based PM task with a low retrospective memory load. The potential benefits of a cue salience manipulation like this can be inferred from studies showing that the effects of HIV on PM are reduced, if not eliminated, when the cue is highly salient; for example, when an event-based cue is focal to the ongoing task (Loft et al., 2014) or when there is an alarm near the time an intention needs to be completed (Faytell et al., 2018). We also examined whether combining all three strategic supports at encoding, monitoring, and cue detection would further boost the beneficial effects of the single supportive interventions in the laboratory.

Method

Participants

A total of 217 participants aged 50 and older were recruited from greater San Diego county, including community-based organizations, infectious disease clinics, the general community, and cohort studies at the University of California San Diego (UCSD) HIV Neurobehavioral Research Program (HNRP). HIV serostatus was confirmed with Medmira rapid tests. We excluded participants with histories of severe psychiatric disorders (e.g., psychosis), neurological disease not due to HIV infection (e.g., Parkinson’s disease, seizure disorders, non-HIV-associated dementia), estimated verbal IQ scores < 70 (n=2), head injury with loss of consciousness > 30 min (n=2), color blindness, current substance dependence, or positive breathalyzer for alcohol or urine screens for illicit substances (n=2). Eight of the 211 otherwise eligible participants (3.8%) did not complete the primary experiment, including two participants who refused the task, one participant who was unable to engage meaningfully with the task due to low computer literacy, and five participants who did not complete the task due to technical administration errors. Thus, the final analyzable sample was 203 participants, which included 145 HIV+ persons and 58 HIV- persons whose demographic, psychiatric, neurocognitive, and medical characteristics are displayed in Table 1. Of note, the HIV- group was older than the HIV+ participants and included a higher proportion of women and persons with histories of depression and anxiety (ps<.05).

Table 1.

Demographic and clinical characteristics of the six study groups.

	HIV−		HIV+
Variable	Control (n=58)	Control (n=27)	Encoding II/Vis (n=30)	Monitoring STOP (n=30)	Cue Detect Tone (n=29)	Combination All (n=29)
Demographic
Age (yrs)*	60.6 (7.6)	57.9 (7.1)	57.0 (7.0)	58.6 (6.2)	54.7 (4.0)	57.2 (6.0)
Education (yrs)	14.6 (2.5)	13.3 (2.5)	14.2 (2.7)	14.8 (2.7)	13.9 (2.9)	14.0 (2.8)
Sex (% women)*	34.5	11.1	20.0	16.7	13.8	10.3
Ethn (% Hispanic)	17.2	14.8	26.7	23.3	13.8	20.7
Race (% White)	65.5	79.3	63.0	62.1	66.7	66.7
Psychiatric
Substance	53.5	66.7	76.7	66.7	72.4	69.0
Depression*	32.8	74.1	76.7	66.7	69.0	62.1
Anxiety*	17.2	33.3	53.3	33.3	48.3	34.5
Neurocognitive
Estimated VIQa	110 (16)	106 (15)	105 (13)	103 (14)	106 (14)	107 (16)
NCIb (%)	24.1	38.5	37.9	50.0	22.2	14.8
Medical
Comorbidities	1.6 (1.4)	2.3 (1.8)	2.1 (1.4)	2.2 (1.4)	2.0 (1.5)	2.4 (1.7)
HIV Disease
Duration (yrs)	-----	20.9 (8.22)	21.3 (8.9)	20.0 (8.9)	21.7 (8.7)	22.0 (8.4)
Nadir CD4	-----	147 (132)	174 (182)	209 (1550	146 (151)	260 (233)
Current CD4	-----	582 (228)	714 (349)	711 (226)	654 (264)	694 (392)
AIDS	-----	76.9	73.3	56.7	65.5	65.5
ART (%)	-----	96.0	90.0	93.1	100	82.8
HIV RNA (%)		4.1	0.0	7.14	11.5	0

Note. AIDS = Acquired Immune Deficiency Syndrome; ART = antiretroviral therapy; Comorbidities refers to the number of medical diagnoses per the nursing evaluation (e.g., hypertension, diabetes, hepatitis C). Ethn = ethnicity; HIV = Human immunodeficiency virus; II/Vis = Implementation intentions and visualization; NCI = Neurocognitive impairment (per NIH Toolbox or CogState); STOP = Stop, Think, Organize, and Plan (i.e., context-free cueing)

^an = 189.

^bn = 192.

^*p < .05

Study Design

This study used both between- and within-subjects design elements (see Figure 1). All participants completed three blocks of the same ongoing language task (i.e., either lexical decision-making or word categorization). The first block was always a baseline version of the language task that did not include PM instructions. In the next two blocks, a time-based PM task was added to the ongoing language task during which participants were instructed to press a specific key at prespecified times. For participants in the control conditions, strategic processing was unsupported in both PM blocks. For participants in the experimental conditions, the order of the supported and unsupported PM blocks was counterbalanced to minimize practice confounds. There was a brief delay in between all three tasks during which the participants received the instructions and engaged in practice trials to ensure their understanding.

Figure 1.

Flow chart of the study design. The administration order of the PM trials 2a and 2b were counterbalanced. Cat = Ongoing word categorization task; Impl Int = implementation intentions; Lex = Ongoing lexical decision-making task; PM = prospective memory; Vis = visualization;

Ongoing Language Tasks

All participants were randomly assigned to complete either a lexical decision-making (n=103) or a word categorization (n=100) task as the ongoing language task for all three study blocks. We used two different types of ongoing tasks to ensure that our findings regarding the PM interventions were generalizable across tasks with different stimuli-response pacing. That is, lexical decision response times are usually about twice as fast as word categorization response times, which could influence PM accuracy, and the potential impact of the interventions, because faster ongoing response times could make it more difficult to inhibit ongoing responses. The response box for the ongoing language tasks was an 11 × 8” device with four illuminated buttons. The primary response buttons used for the ongoing language tasks were 2” in diameter and colored green (left-hand side) and red (right-hand side). The two secondary response buttons used for the PM blocks were 1” in diameter and colored white (top center) and blue (bottom center). Experiments were coded using PHP and JavaScript, with data stored in real time on a locally hosted MySQL database

Lexical Decision-Making.

Participants viewed 600 English letter strings. Three-hundred of the letter strings were medium frequency words (i.e., 20–50 occurrences per million) of 4–6 letters in length that were drawn from the Sydney Morning Herald database (Dennis, 1995). Words that were uncommon in American English were replaced. The remaining 300 letter strings were 4–6 letter non-words that were obtained from the Macquarie University ARC non-word database (Rastle, Harrington, & Coltheart, 2002). One hundred words and 100 non-words were assigned to List A, B, and C. The assignment of these three lists to the baseline block, PM Unsupported block, and PM Supported block was counterbalanced. Within each list, the order of presentation of the words and non-words was random without replacement. On each lexical decision task the trial sequence was as follows: The first 250-ms display was a focus point marked by a fixation displayed in white on a black background, followed by a letter string that remained on the screen until a response was detected. The total block length was approximately 4 min and 55 sec. The intertrial interval was calculated by subtracting lexical decision response time and fixation time from 3s (duration of each trial) (Hicks et al., 2005). Participants were asked to respond as quick and accurately as possible as to whether the letter string was an English word (e.g., muffin) by pressing the left green response button or a non-word (e.g., gfftew) by pressing the right red response button. Participants underwent 10 practice trials to ensure their comprehension of the task. Accuracy and response time for word trials were recorded. When analyzing accuracy and response time data, we excluded +− 2 trials from a clock check, PM target, or stop cue. In addition, we excluded response times >3 SD above each participant’s mean.

Word Categorization.

Participants viewed 450 category-word pairs. Their task was to decide whether the word in lower case font on the left was a member of the broad semantic category word shown in upper-case font on the right (e.g., ring-JEWELRY). Two-hundred and twenty-five category-word matches were selected from the Van Overschelde Rawson, and Dunlosky (2004) category norms. The other 225 category-word pairs were non-matching (e.g., venue-BIRD). There were 75 matches and 75 non-matches assigned to List A, B, and C. For each trial the sequence was as follows: The first 250-ms display was a focus point marked by a fixation displayed in white on a black background, followed by a category-word pair that remained on the screen until a response was detected. Each trial was separated by a 100-ms fixation cross. The total block length was approximately 4 min and 55 sec. The assignment of these three lists to the baseline block, PM Unsupported block, and PM Supported block was counterbalanced. Within each list, the order of presentation of the category-word pairs was random without replacement. Participants were instructed to press the left green button for yes responses and that the right red button was for no responses. Participants underwent 10 practice trials to ensure their comprehension of the task. When analyzing accuracy and response time data, we excluded +− 2 trials from a clock check, PM target, or stop cue. In addition, we excluded response times >3 SD above each participant’s mean.

Time-based PM Tasks.

After participants completed the baseline ongoing language task block, they were told that examiner was also interested in their ability to remember to perform a future intention. Participants were instructed to press the center white response button at 2, 5, and 9 min during the ongoing language task. This irregular staggering of target times was incorporated to reduce the likelihood that this task would be habitual and thus potentially lower the demands on cognitive resources. PM responses ± 20 sec of the target time were scored as correct and are reported as percent accurate (block range = 0–100%).

Participants could press the center blue response button to reveal a clock at the center of the screen, which would display for 5 sec. Participants were encouraged to complete the task with as few clock checks as possible, since in the real-world individuals do not typically check the time with high frequency. Discouraging clock checking also increases the resource demands of the PM task (Huang et al., 2014), thereby giving us the best opportunity to find a meaningful impact of our supportive interventions. Clock checks were recorded as raw values (individual PM block range = 0–55).

Participants were asked to repeat the instructions and completed a brief language task practice trial with a 25-sec PM target to ensure their understanding of the PM task. Regardless of the PM block order, the instructions for the control (unsupported) and experimental (supported) blocks explained that the new PM task was identical to the prior PM task, but the conditions under which it would be completed (i.e., presence or absence of support) were now different. Next, participants were engaged in a non-verbal distractor task (i.e., paper-and-pencil mazes) for 1 min prior to the beginning of each PM block. Task order was not significantly associated with PM accuracy (Fs<1.0, ps>.10). The total duration of the active PM blocks was 10 min and 45 sec. After completing both PM blocks, participants were administered a two-item, four-choice recognition test for the PM response key (sample 90% correct with no between-groups differences, p=.230) and specific target times (sample 99% correct with no between-groups differences, p=.704).

Experimental Groups.

HIV+ participants were randomized into one of the five conditions detailed below. All HIV- participants were assigned to the Control condition.

Control.

Participants in this condition (HIV+ N=27; HIV- N=58) received two control unsupported PM blocks of the ongoing task as described above.

Encoding.

Participants in this counterbalanced condition (HIV+ N=30) completed one control unsupported PM block and one supported PM block in which they engaged in an implementation intention and a visualization exercise. For the implementation intention, participants repeated the following phrase three times: “WHEN I am doing the (ongoing language task) AND it has been 2, 5, and 9 minutes since the start of the task, THEN I WILL press the TOP MIDDLE (WHITE) button.” Next, participants closed their eyes for 30 sec and were asked to imagine themselves sitting in the chair in front of the computer performing the ongoing language task and pressing the top middle white button at the appropriate time.

Monitoring.

Participants in this counterbalanced condition (HIV+ N=30) completed one control unsupported PM block and one supported PM block in which they were taught to use content-free cueing (Fish et al., 2007). Specifically, participants were prompted with a 15-sec visual reminder aid (i.e., the words “Stop, Think, Organize, and Plan” bookended by stop sign graphics) that appeared at 1, 3:30, and 7 min during the ongoing task. The participants in this condition were provided with the following instructions:

“At various times during the ongoing task, you will see pictures of stop signs along with the words ‘Stop, Think, Organize, Plan.’ When you see the stop signs, I want you to stop what you are doing and reorient yourself to the task. Specifically, I want you to review in your mind the following steps as a strategy to help you remember to press the TOP MIDDLE (WHITE) button at the correct times: 1) Stop what you are doing, or take a time-out from the ongoing task to ask yourself “What am I doing and am I on track?”. Think about what you need to remember to do by asking yourself “What else do I have to do and when?’ For this task, this should prompt you to recall that you are supposed to press the TOP MIDDLE (WHITE) button at specific times while completing the ongoing task. Organize what you need to do for the task by asking ‘What can help me complete this task accurately?’ At this step you can review the times at which you are supposed to press the TOP MIDDLE (WHITE) button and recall that the BOTTOM MIDDLE (BLUE) button shows you the clock. Plan for how to correctly complete the task based on your review of the prior steps. This STOP reminder will appear on the screen for 15 seconds to allow you to use this strategy and the task will begin again.”

After a brief practice trial of using this support, the instructions were repeated to the participants, who were given the opportunity to ask for clarification prior to beginning the task.

Cue Detection.

Participants in this counterbalanced condition (HIV+ N=29) completed one control unsupported PM block and one supported PM block in which they were provided with a salient tone near each time-based target. In the supported PM trial, participants were instructed that a bell would sound during the ongoing language task and alert them 5-sec prior to each time they were supposed to make a PM response (i.e., at 1:55, 4:55, and 8:55 min).

Combination.

Participants in this counterbalanced condition (HIV+ N=29) completed one unsupported PM block and one supported PM block in which they were provided with all three of the support strategies outlined above; i.e., encoding (implementation intentions and visualization), monitoring (context-free cueing), and cue detection (salient tone).

Clinical PM Task.

All participants received the Cambridge Prospective Memory Test (CAMPROMPT; Wilson et al., 2005), which is a clinical measure of PM that includes three time-based tasks (e.g., change ongoing task in 7 min) and three event-based tasks (e.g., recall hidden objects at the end of the test). Paper and pencil puzzles serve as the ongoing task. Participants are allowed (but not required) to take notes as a compensatory strategy; 87% of the entire sample took notes during the task. Each PM task is scored from 0 (item failed even after two prompts) to 6 (item recalled spontaneously in response to the correct cue). For this study, we used only the time-based PM scale, which ranges from 0–18 (full sample M=10.8, SD=4.8).

Other Study Measures

The PM data were collected as part of a comprehensive neuropsychological study visit that included a variety of self-report and performance-based measures of mood, cognition, personality, psychosocial factors, health literacy, and everyday functioning. For the current study, we report descriptive data regarding: 1) Diagnoses of current and lifetime Substance Use, Major Depressive, and Anxiety Disorders from the Composite International Diagnostic Interview (v2); 2) Neurocognitive impairment based on Global Deficit Scores (GDS; Carey et al., 2004) derived from either the NIH Toolbox (Casaletto et al., 2015; n=89) or CogState (Woods et al., 2016; n=103); 3) Estimates of premorbid verbal IQ from either the NIH Toolbox (Casaletto et al., 2015; n=89) or Wide Range Achievement Test – 4 (Casaletto et al., 2014; n=93); and 4) HIV disease and treatment characteristics derived from standard laboratory values from a blood draw and a neuromedical examination performed by a certified research nurse.

Data Analysis

All primary hypotheses were evaluated with repeated measures multivariate analysis of variance (RMANOVA), which included both within- and between-subjects factors as detailed below. Covariates were selected from those variables in Table 1 that were associated with both the within- and between-subjects factors at p<.05 for each planned analysis separately (i.e., covariates thus differed across the primary analyses of the ongoing tasks, PM accuracy, and clock checks). Note that, most of the dependent variables were non-normal per Shapiro-Wilk W test (ps<.05). However, each main effect and post-hoc analysis was evaluated to ensure that the results did not change with non-parametric statistics. A critical alpha of .05 was used for primary null hypothesis significance testing and accompanied by Hedges g effect size estimates. Planned post-hoc analyses were conducted using a Bonferroni correction to limit Type I error risk. Data were analyzed in JMP 14.0 (Carey, NC).

Results

Ongoing Language Tasks

Table 2 shows the descriptive data for ongoing language task accuracy and response time on the lexical decision-making and word categorization across the experimental groups.

Table 2.

Ongoing task accuracy and response time data for the six study groups.

	HIV−		HIV+
Ongoing Task	Control (n=58)	Control (n=27)	Encoding II/Vis (n=30)	Monitoring STOP (n=30)	Cue Detect Tone (n=29)	Combination All (n=29)
Baseline
Category (n=100)
Accuracy (%)	92 (5)	94 (4)	95 (4)	93 (4)	96 (2)	93 (6)
RT (ms)	1596 (255)	1660 (278)	1590 (279)	1623 (225)	1429 (253)	1538 (343)
Lexical (n=103)
Word
Accuracy (%)	99 (1)	98 (4)	99 (1)	98 (5)	99 (1)	99 (2)
RT (ms)	728 (172)	701 (103)	711 (87)	736 (130)	661 (53)	712 (112)
Non-Word
Accuracy (%)	98 (3)	98 (3)	98 (2)	97 (6)	99 (3)	98 (2)
RT (ms)	786 (203)	724 (83)	797 (180)	778 (143)	753 (124)	738 (142)
Unsupported PM
Category (n=100)
Accuracy (%)	94 (5)	92 (9)	96 (3)	94 (6)	92 (17)	94 (6)
RT (ms)	1530 (243)	1625 (319)	1428 (217)	1595 (324)	1265 (361)	1448 (363)
Lexical (n=103)
Word
Accuracy (%)	98 (2)	97 (3)	99 (2)	97 (4)	99 (1)	99 (2)
RT (ms)	776 (168)	771 (84)	743 (98)	746 (110)	720 (63)	759 (133)
Non-Word
Accuracy (%)	98 (3)	98 (3)	98 (3)	98 (3)	98 (3)	99 (2)
RT (ms)	789 (169)	790 (103)	782 (134)	765 (86)	751 (67)	747 (129)
Supported PM
Category (n=100)
Accuracy (%)	94 (5)	92 (6)	96 (3)	94 (6)	97 (2)	93 (6)
RT (ms)	1468 (251)	1527 (311)	1447 (224)	1557 (280)	1293 (171)	1401 (363)
Lexical (n=103)
Word
Accuracy (%)	98 (3)	98 (2)	99 (1)	98 (3)	99 (1)	99 (1)
RT (ms)	776 (171)	732 (102)	766 (113)	761 (131)	676 (77)	749 (158)
Non-Word
Accuracy (%)	98 (2)	98 (2)	98 (3)	97 (5)	99 (2)	99 (2)
RT (ms)	769 (154)	768 (88)	797 (149)	767 (111)	719 (75)	727 (135)

Note. HIV = Human immunodeficiency virus; II/Vis = Implementation intentions and visualization; ms = milliseconds; PM = prospective memory; RT = response time; STOP = Stop, Think, Organize, and Plan (i.e., context-free cueing)

Lexical Decision-Making.

A repeated measures MANOVA examined the effects of Block (three levels, within-subjects), Experimental Condition (six levels, between-subjects), and their interaction on word accuracy. Results showed a main effect of Block (F(2,96)=5.3, p=.006), with higher accuracy in the baseline block as compared to the two PM blocks (ps<.007), which did not differ from one another (p=.279). There was no main effect of Experimental Condition or interaction between Block and Experimental Condition (ps>.10). An RMANOVA covaried for age examined the effects of Block and Experimental Condition on word RT. Results showed a main effect of older age with slower RT (F(1,96)=8.3, p=.005), but no significant main effects of Block or Experimental Condition and no interactions (ps>.10).

Word Categorization.

A repeated measures MANOVA examined the effects of Block and Experimental Condition on word categorization accuracy. Results showed no significant main effects of Block or Experimental Condition or their interaction (ps>.10). An RMANOVA covaried for age examined the effects of Block and Experimental Condition on word categorization RT. Results showed no significant main effects of Block or Experimental Condition or their interaction (ps>.05).

Time-based PM Accuracy

Figure 2 displays the PM accuracy data across the study conditions. A repeated measures MANOVA covaried for age was conducted with the Supported and Unsupported PM Blocks as within-subjects factor. The between-subjects factors were the six Experimental Conditions and the two Ongoing Tasks. Results showed no significant main effects of age, Experimental Condition, Ongoing Task, or Supported PM Block (ps>.10). There was, however, a significant interaction between Experimental Condition and Supported PM Block (F(5,195)=2.8, p=.020). Planned within-subjects post-hoc analyses of the supported experimental PM blocks as compared to the control unsupported PM blocks¹ (using a Bonferroni correction of p=.0083) showed significant enhancement of PM in the Encoding (p=.003, g=.36) and Cue Detection (p=.0002, g=1.18) study conditions. In contrast, there were no significant within-subjects effects in the HIV- Control (p=.030, g=0.19), HIV+ Control (p=.490, g=0.08), Monitoring (p=.586, g=0.09), or Combined (p=.165, g=0.33) study conditions on PM, although the latter was associated with a small-to-medium effect size. All other interaction terms were non-significant (ps>.10).

Figure 2.

Bar chart depicting the mean time-based prospective memory (PM) accuracy scores (± full sample SE) in the Unsupported (control) and Supported (experimental) prospective memory blocks across the six study conditions. * Bonferroni-adjusted p < .008 for the within-subjects comparison of time-based PM for the Unsupported (control) and Supported (experimental) PM blocks in that study condition. Detect = Detection; HIV = Human immunodeficiency virus; II/Vis = Implementation intentions and visualization; PM = prospective memory; STOP = Stop, Think, Organize, and Plan (i.e., content-free cueing).

Next, we examined the correlates of the beneficial effects of PM support using a simple difference score (i.e., Supported PM accuracy - Unsupported PM accuracy) separately in the Encoding (M=12.2±3.7) and Cue Detection (M=26.4±6.1) conditions. Results showed a large negative correlation between CAMPROMPT time-based PM scores and the benefits of PM support in the Cue detection condition (r=−.47, p=.001), but not in the Encoding condition (r=−.24, p=.198). PM support benefit was not related to global neurocognitive functioning (as measured by the Global Deficit Score) in either condition (ps>.10).

Clock Checking

Figures 3 displays the clock checking data across the study groups and experimental conditions. A repeated measures MANOVA covarying for sex was conducted. The two between-subjects effects were Experimental Condition and Ongoing Task and the two within-subjects effects were the number of clock checks in the supported and unsupported PM blocks (Supported PM Block). Results showed no main effects of Ongoing Task, Experimental Condition, or sex (Fs<3.5, ps>.05). There was a significant main effect of Supported PM Block (F(1,195)=17.0, p<.0001), which interacted with both Experimental Condition (F(5,195)=9.1, p<.0001) and Ongoing Task (F(1,195)=5.6, p<.019). Planned within-subjects post-hoc analyses of the interaction between Supported PM Block and Experimental Condition² using a Bonferroni correction (p=.0083) showed significant reductions in clock checks in the Cue Detection (p<.001, g=.72) and Combined (p<.001, g=.48) supported experimental PM blocks as compared to the unsupported control PM blocks. There were no significant within-subjects effects of Supported PM block on clock checks observed in the HIV- Control (p=.259, g=0.09), HIV+ Control (p=.398, g=0.12), Encoding (p=.578, g=.07), or Monitoring (p=.772, g=−0.03) study conditions. Within-subjects post-hoc analyses of the interaction between Supported PM Block and Experimental Condition using a Bonferroni correction (p=.025) showed significant reductions in clock checks in the supported versus unsupported PM blocks for persons in the Word Categorization ongoing task condition (p<.001, g=.36), but not for those in the Lexical Decision-making condition (p=.04, g=.13).

Figure 3.

Bar chart depicting the mean number of clock checks (± full sample SE) in the Unsupported (control) and Supported (experimental) prospective memory blocks across the six study conditions. * Bonferroni-adjusted p < .008 for the within-subjects comparison of the clock checks in the Unsupported (control) and Supported (experimental) prospective memory blocks in that study condition. Detect = Detection; HIV = Human immunodeficiency virus; II/Vis = Implementation intentions and visualization; STOP = Stop, Think, Organize, and Plan (i.e., content-free cueing).

In Figure 4, we present a post-hoc analysis of the effects of Experimental Condition and Ongoing Task on clock checking data across the early (i.e., in the first minute after the task begins or after a PM target), middle, and near (i.e., the minute immediately preceding the PM target) PM target time intervals. In the Unsupported (control) block, we observed a main effect of time (F(2,194)=40.2, p<.0001), such that clock checks increased as the PM target neared, and a main effect of Ongoing Task (F(1,195)=5.6, p=.02), whereby clock checking was higher in the Word Categorization task as compared to the Lexical Decision-making task. There were no significant main effects of Experimental Condition and no interactions (all ps>.10). For the Supported (experimental) block, we observed main effects of Experimental Condition (F(5,195)=6.6, p<.0001), time (F(2,194)=31.5, p<.0001), and their interaction (F(10,388)=1.9, p=.043). In the early phase, the HIV+ Cue Detection group checked the clock less frequently than HIV- Controls (p=.02), but there were no other between-group differences (ps>.10). In the middle and near phases, the HIV+ Cue Detection group checked the clock less frequently than all other groups (ps < .05), except the HIV+ Combined group (ps >.10).

Figure 4.

Box and whisker plot of the mean number of clock checks in the early, middle, and near PM target time intervals for all of the study groups in both the Unsupported (control; panel a) and Supported (experimental; panel b) conditions. Detect = Detection; HIV = Human immunodeficiency virus; II/Vis = Implementation intentions and visualization; STOP = Stop, Think, Organize, and Plan (i.e., content-free cueing).

Post-hoc Analyses of HIV-associated Neurocognitive Disorders

Considering prior work suggesting that the primary effects of HIV on PM are driven by individuals with HIV-associated neurocognitive disorders (HAND; see Avci et al., 2018 for a review), we examined the effects of HAND on the Unsupported (control) PM accuracy and clock checking trials. We observed significant omnibus differences between the HAND+, HAND-, and HIV- groups (p = .01) for PM accuracy. This effect was driven by the low performance in the HAND+ group (63.6+− 6.0%) who differed significantly from the HAND- group (82.3+−3.1%; p=.003, d=−.57) and showed a small, non-significant difference from the HIV- group (73.6+−4.6%; p=.107, d=−.27). The three groups did not differ in clock checking (p = .527).

Discussion

A primary finding from this experiment is that the PM functioning of older adults with HIV disease can benefit from salient auditory alerts, which supported the accurate execution of a time-based PM task in the laboratory. The salient alarm improved time-based PM accuracy for older HIV+ adults by 26.4% from baseline levels, which was a large within-subjects effect size. In a between-subjects study of HIV-infected youth (M age 23 years), Faytell and colleagues (2018) described a comparably large benefit of an audible mobile alarm to enhance naturalistic time-based PM (i.e., daily text messaging responses). Taken together, these data support the conceptual notion that HIV-associated deficits in time-based PM are driven by failures in cue detection (Avci et al., 2018; Carey et al., 2006) that can be supported by enhancing the salience of the retrieval cue. The incorporation of a salient alarm appears to have converted the strategically demanding time-based PM cue into a more automatic event-cued PM task (Shelton et al., 2013, 2019). Note that the salient alarm did not eliminate all PM requirements, as individuals were still required to perform a series of cognitive operations that could have gone awry in the setting of HIV, including verifying the match of the target with the intended target, inhibiting their performance on the ongoing task, shifting set to the PM task, accurately retrieving the prescribed intention, executing the motoric intention, and deactivating the intention (Kliegel et al., 2008; Marsh et al., 2003; Scullin et al., 2013). Thus, it is remarkable that this cue detection intervention facilitated such high levels of PM accuracy among older HIV+ adults, who can show mild deficits on even fairly automatic event-based PM tasks (see Woods et al., 2010). Given the many challenges of translating laboratory findings of cognitive improvement into effective strategies that can be reliably deployed in daily life, future work is needed to determine whether these strategic supports might be applicable to the everyday PM of older HIV+ adults who carry a disproportionate burden of HIV-associated, Non-AIDS (HANA) conditions (e.g., cardiovascular disease), including HAND (Valcour et al., 2004), which can affect retention in HIV care (Jacks et al., 2015) and health behaviors (e.g., Thames et al., 2011).

In this laboratory-based study of older HIV+ adults, the introduction of salient auditory alarms near the cue was accompanied by a sharp reduction in clock checking behavior. Under normal control conditions for a time-based PM task, participants typically monitor time by checking the clock in a “J-shaped” curve, which is characterized by: 1) an initial increase in clock checks immediately after the assignment of the intention; 2) an intermediate lag; and 3) and increase in clock checking as the cue time approaches (Harris & Wilkins, 1982). A post-hoc examination of the pattern of clock checks in relation to the PM target (see Figure 4) showed this expected J-shaped curve for all of the experimental groups in the control (unsupported); however, in the supported condition, the older HIV+ adults in the Cue Detection condition showed much a much flatter pattern of low clock checking, which was strikingly different from the control, encoding, and monitoring groups during the middle and near PM target phases. In other words, they maintained a low level of cue monitoring to achieve higher PM accuracy than the other groups, who ramped up their cue monitoring in the expected fashion. Thus, highly salient strategic supports at cue detection appear to diminish overt monitoring without dampening or enhancing ongoing task performance for older HIV+ adults.

In prior research in HIV disease, lower clock checking was associated with poorer clinical time-based PM and aspects of executive functions (Doyle et al., 2013). Post-hoc analyses of the current study data align with those prior studies by showing that more frequent clock checking was associated with higher PM accuracy in the unsupported block (r_s=.41, p=.030), but not in the supported Cue Detection block (r_s=−0.13, p=.493). Thus, participants who received the salient auditory alarm engaged in less strategic cue monitoring to achieve higher time-based PM accuracy scores relative to baseline. In line with the Dynamic Multi-Process Framework (Scullin et al., 2013, 2019), the salient auditory cue seemed to spontaneously bring the time-based PM intention to mind, after which individuals checked the clock or simply made their PM response. To that end, the magnitude of the reductions in clock checking were moderate with the ongoing Word Categorization task, which was characterized by lower accuracy and slower response times than the Lexical Decision-making ongoing task, for which only small, non-significant reductions in clock checking were observed. Although speculative, it is possible that the salient alarm reduced the need for overt strategic monitoring of the time-based cue, allowing more cognitive resources to be allocated to the ongoing task with higher demands.

Of clinical relevance, older HIV+ adults with lower time-based PM on a separate clinical measure (i.e., the CAMPROMPT) showed the greatest benefits of the salient alarm on laboratory PM accuracy. In other words, the cue salience supports were most effective for persons with the greatest clinical PM deficits. Two prior studies have reported an association between clinical PM deficits and benefit from cognitive support to bolster event-based PM (Faytell et al., 2017) and adherence (Pennar et al., 2019) in youth with HIV+. In this study, the association between clinical PM and benefits of the cue salience support was accompanied by a large effect size, was not present in the Encoding (i.e., implementation intentions/visualization) condition, and could not be better explained by global neurocognitive deficits. This finding is important because, as with other neurocognitive functions (see Heaton et al., 2010), deficits in PM are not universal in HIV disease and we did not exclude persons with “normal” time-based PM. It is possible therefore that the effect sizes observed for the benefits of cue salience might be even larger in a sample selected on the basis of clinical time-based PM impairment. One challenge for future work might be to teach older HIV+ adults with time-based PM impairment to: 1) recognize PM intentions in their daily lives that are time-based; 2) develop practical strategies to enhance the salience of the cue (e.g., convert it into a highly automatic event-based PM task); and 3) implement those strategies via trial and error to enhance PM accuracy for daily activities. Since many modern alerting strategies rely heavily on technology (e.g., smart phones and watches) that are often difficult for older adults with neurocognitive disorders (Woods et al., 2019), intervention developers should be mindful of software design and feasibility issues in this population.

Supporting strategic encoding processes in the laboratory by way of implementation intentions and visualization produced smaller improvements in time-based PM for older HIV+ adults (i.e., a 12% improvement from baseline levels). These effects were slightly lower than meta-analytic estimates of the benefits of implementation intentions and visualization on PM in healthy adults aged 60–75 (Chen et al., 2015), which may reflect the clinical population used in this study and/or the simplicity and brevity of our manipulation as compared to other studies that use more in-depth, rich imagery training methods delivered over several sessions (e.g., Potvin et al., 2011). Nevertheless, several studies in implementation intentions and PM have used comparably brief visualization exercises (e.g., Chasteen et al., 2001; McDaniel, Howard, & Butler, 2008) and our findings broadly converge with prior studies of naturalistic event-based PM (Faytell et al., 2017) and medication adherence (Pennar et al., 2019) in youth with HIV. Implementation intentions and visualization/imagery should support PM cue detection and execution by activating the mental representation of the cue-intention pairing and reducing its strategic demands on the prefrontal systems (Gollwitzer, 1999) that are required to plan and maintain the intent to strategically monitor time. The cognitive mechanism of the improvement in time-based PM associated with implementation intentions and visualization in this study is unclear, however. Our design does not allow us to dissect the relative contribution of implementation intentions and visualization to improving time-based PM, which were combined to maximize their effectiveness. Moreover, we did not measure the depth or vividness of the imagery and the benefit in performance was not reliably associated with time monitoring, ongoing task performance, retrospective memory for the cue-intention pairing, or clinical PM performance. With regard to the latter, it is plausible that this strategy might therefore produce comparable benefits to the time-based PM of older HIV+ adults across the spectrum of PM ability and independent of ongoing task demands (cf. Faytell et al., 2017). Given the relative ease with which implementation intentions and visualization are relatively straightforward to use in daily life (e.g., they require minimal training and no ancillary equipment), such strategies may therefore be an attractive intervention target for tailored, translational efforts to improve naturalistic PM, perhaps particularly for non-habitual short-term PM tasks (e.g., remembering to message a healthcare provider).

Content-free cuing as operationalized by the STOP paradigm did not significantly affect time-based PM or clock checking among older HIV+ adults. This very small effect size was not expected because there is evidence of deficient time-based PM monitoring in HIV (e.g., Doyle et al., 2013), and prior research on PM in brain injury (Fish et al., 2007) and metacognition for executive functions in HIV (Casaletto et al., 2016) suggest that this technique can be effective in improving strategically demanding cognition. It is possible that more direct monitoring cues (e.g., displaying current and target times) might have been more effective. Another potential reason for the lack of effectiveness of the STOP paradigm was that participants received only very brief training on its use (<3-min) as part of a computerized laboratory task. Qualitatively, our impression was that the older HIV+ adults in this study had difficulties learning and using the STOP paradigm. To that point, a recent study demonstrated that brief STOP training exercises may not enhance event-based PM among older adults due difficulties comprehending the instructions and using the technique (Fine et al., in revision). Given the well-documented deficits in learning associated with HIV disease (e.g., Doyle et al., 2019), future studies may therefore strive to provide more comprehensive and effective training on the use of the STOP paradigm, perhaps using retrospective memory boosters (e.g., self-generation, retrieval practice) that have shown to improve recall in HIV (e.g., Avci et al., 2017; Weber et al., 2012).

It is well established that the profile of declarative memory impairment in HIV is heterogenous across encoding and retrieval (see Doyle et al., 2019), which informed our hypothesis that combining supportive strategies at the encoding, monitoring and cue detection phases of time-based PM would be effective in improving accuracy. To that end, Faytell et al. (2018) suggested there may be additive effects of supporting encoding and cue detection for naturalistic time-based PM in HIV+ youth. Yet the older HIV+ adults in the combined condition in this study demonstrated only a 9% improvement in time-based PM from baseline. Although this benefit was not statistically significant, it was nevertheless associated with a small-to-medium effect size that was broadly comparable to that observed for the encoding condition. As such, we are cautious in interpreting this finding as “null”, particularly since the baseline time-based PM scores in this group were very high (87.4% accuracy) and we observed a moderate decline in clock checking behavior in this condition. It is also possible that participants became confused by the number of different supportive strategies they were taught in such a short time. Future studies might use a more rigorous factorial design to determine the effectiveness of using one or more strategies; for example, is the effectiveness of implementation intentions and visualization boosted by an explicit monitoring prompt?

The benefits of supporting PM in the laboratory did not vary by the type of ongoing language task. The ongoing task demands were low, limited to language tasks with no executive or concurrent memory demands. In daily life, people often have much more cognitively (and emotionally) demanding ongoing tasks, so it will be important for future research to examine the effectiveness of encoding and cue detection support with more demanding ongoing tasks (e.g., working memory). These strategies may have the best chance of succeeding under more challenging circumstances given that enhancing encoding takes place during a break in the ongoing demands (e.g., taking a few minutes to do visualization/implementation intentions before leaving home for the day) and cue detection supports the execution of the intention via an external reminder (e.g., salient alarm). There was no reliable evidence of PM costs to the ongoing tasks. This is likely because clock checking rates were relatively high (M = 11 in the control conditions), even though we discouraged clock checking. Much of the need for internal control, such as maintaining the intent to perform the PM task and prospective timing, can be externalised by performing a well-practiced task such as clock checking (Huang et al., 2014; Waldum, & Sahakyan, 2013), and further by using the knowledge that one will be externally reminded by a salient alarm (i.e., cue detection and combination conditions).

This laboratory-based experiment has several limitations that temper conclusions to be drawn and may inform future directions. One limitation is that our HIV seronegative participants only completed an unsupported PM control trial. Moreover, our seronegative group had a high rate of medical and psychiatric comorbidities (e.g., substance use disorders, depression). As such, we do not know how the benefits of these approaches in our HIV+ sample compares to that which would be observed in healthier older adults. This HIV- “comparison” sample may have also dampened the effects of HIV disease on PM accuracy and clock checking in the unsupported control condition; indeed, post-hoc analyses showed that individuals with HAND showed medium effect size deficits in PM accuracy as compared to HIV+ persons without HAND and small effects relative to the HIV- cohort. The absence of HAND effects on monitoring in this study was surprising (see Doyle et al., 2013) may be a function of the older sample, the use of ongoing tasks with somewhat low cognitive demand (cf. working memory), and/or the low retrospective memory load of the PM intention itself. It should also be noted that there was only a very brief delay (and no intervening task, aside from instructions) in between the two PM trials. Although this procedure was counterbalanced, it might have nevertheless produced mild confounds of either practice effects or carry-over benefits from the support condition to the unsupported condition; however, the absence of significant order effects on PM accuracy argues against this interpretation. Another notable limitation is that the study groups were comprised mostly of white, well-educated men from Southern California, so the external validity of the findings to under-represented racial/ethnic groups, women, and persons living in other regions of the United States (and internationally) is unknown.

Finally, it is worth highlighting that this study used a laboratory time-based PM paradigm with fairly short PM target times and a very low retrospective memory load that might not reflect everyday PM tasks. Moreover, we did not include any direct assessments to determine whether the observed laboratory effects might generalize to daily life situations. Our primary goal was to determine whether this set of particular PM supports could be beneficial to older HIV+ adults in the laboratory. Such experiments are critical as “proof of concept” and to discover mechanisms of response under controlled conditions; yet their intended application in the real world requires accurate recall and deployment of these strategies in daily environments, which often include uncontrolled contextual factors (e.g., busyness, interruptions) that can alter their effectiveness. This represents a major challenge for the field of neuroAIDS and other areas of clinical neuroscience; i.e., how does one translate mechanistic laboratory findings into effective clinical treatments that improve quality of life? In daily contexts, PM is often employed over the course of hours and days and often with more complex retrospective memory demands (e.g., remember to refill your prescription for medication X with healthcare provider Y at the end of the month). Moreover, the aging-PM paradox (Rendell & Thompson, 1999) suggests that because of these long delays and the natural environment, factors determining PM success in daily life may differ across the lifespan, including reliance on motivation, routine, and busyness. For example, age and HIV have additive effects on long, but not short delay PM tasks in the laboratory (Avci et al., 2016; Morgan et al., 2012), but the HIV effects are dampened on naturalistic PM tasks with very long delays (Avci et al., 2016; Weber et al., 2011). Thus it remains to be determined whether supporting strategic PM at the encoding and/or cue detection phases enhances naturalistic PM, symptoms of PM, or health behaviors with strong PM components (e.g., medication adherence).

Conclusion

Older adults living with HIV have difficulties “remembering to remember,” which affects their abilities to live independently and follow healthcare provider instructions. This laboratory-based experiment demonstrates that older HIV+ adults can improve their memory performance by using specific strategies that increase the salience of the PM cue and thus potentially reduce the neurocognitive demands of remembering to remember. Future studies using more naturalistic memory methods will determine whether similar improvements in remembering to remember can be achieved in the daily lives and health behaviors (e.g., medication adherence) of older adults with HIV disease.

Public Health Significance.

Older adults living with HIV have difficulties “remembering to remember,” which affects their abilities to live independently and follow healthcare provider instructions. This experiment demonstrates that older HIV+ adults can improve their memory performance in the laboratory by using specific strategies that reduce the cognitive demands of remembering to remember. Future studies using more naturalistic memory methods will determine whether similar improvements in remembering to remember can be achieved in the daily lives and health behaviors (e.g., medication adherence) of older adults with HIV disease.