Prospective Memory Partially Mediates the Association Between Aging and Everyday Functioning

Abstract

Objective:

Older adults commonly experience declines in everyday functioning, the reasons for which are multifactorial. Prospective memory (PM), or remembering to carry out intended actions, can be an executively demanding cognitive process that declines with older age and is independently associated with a variety of everyday functions (e.g., taking medication). This study examined the hypothesis that PM mediates the relationship between older age and poorer everyday functioning.

Method:

468 community-dwelling adults (ages 18-75) with a range of medical comorbidities (e.g., viral infection) were classified as dependent on four well-validated measures of manifest everyday functioning: activities of daily living, employment status, the Karnofsky Scale of Performance Status, and self-reported cognitive symptoms. Participants completed the Memory for Intentions Test (MIsT) to measure PM, alongside clinical tests of executive functions and retrospective memory.

Results:

Controlling for education and comorbidities, bootstrap analysis revealed a significant direct effect of age on everyday functioning and a significant mediated effect of age on everyday functioning through the indirect effect of time-based b=.006 [.003, .010] and event-based PM (b=.005, [.002, .009]), as well as slightly smaller effects for executive functions (b=.003, [.001, .005]) and retrospective memory (b=.002, [.001, .005]).

Conclusions:

These cross-sectional data suggest that executively demanding aspects of declarative memory play an important partial mediating role between an individual factor (i.e., age) and daily life activities, and highlight the possible benefit of targeting PM for improving everyday functioning in older adults.

Even among typically aging adults, older age is commonly associated with subtle declines in everyday functioning (Tucker-Drob, 2011). Examples of everyday functions that are commonly affected among older adults include medication management, handling finances, and meal preparation (Farias et al., 2018; Schmitter-Edgecombe & Parsey, 2014; Tucker-Drob, 2011). Declines in such everyday activities contribute to the level of independence that older adults maintain and are important predictors of health and well-being outcomes, including quality of life (Garre-Olmo, Vilalta-Franch, Calvó-Perxas, & López-Pousa, 2017), health status (Naseer, Forsell, & Fagerström, 2016), and even mortality (Hoogendijk et al., in press). Given the growing population of older adults (U.S. Census Bureau, 2018), there is increasing attention being paid to the identification and characterization of modifiable predictors of everyday functioning declines in aging. While contributing factors to everyday functioning success among older adults are many and complex, those that show strongest and most reliable associations include demographics (e.g., socioeconomic status), mood, medical comorbidity burden, and neurocognitive functioning (Stuck et al., 1999). The current study focuses on the role that some specific, higher-order neurocognitive abilities may play in the daily lives of older adults.

At face value, it is easy to appreciate the role that declines in higher-order neurocognitive abilities—such as executive functions and memory—could play in everyday functioning. For example, an older adult might forget whether they took their prescribed medication, leading to either a missed or extra dose, which could adversely affect their disease management and health status. It is well-established that many aspects of neurocognition are vulnerable to the effects of aging (Blazer, 2017). Episodic memory and executive functions are among the most commonly affected neurocognitive domains in older age (Deary et al., 2009) and track closely with neural network changes in the prefrontal and temporal cortices (Martinelli et al., 2013). These higher-order neurocognitive abilities also show fairly reliable associations with everyday functioning among older adults (Bell-McGinty, Podell, Franzen, Baird, & Williams, 2002; Nguyen, Copeland, Lowe, Heyanka, & Linck, 2019; Farias et al., 2009). For example, in a sample of 698 older adults, Tucker-Drob (2011) showed that longitudinal changes in memory and executive functions strongly correlated (r range .34-.92) with declines in both manifest and performance-based measures of everyday functioning.

One neurocognitive ability that is related to everyday functioning is prospective memory (PM; Twamley et al., 2008; Woods, Iudicello, et al., 2008), or remembering to carry out an intended action in the future. From a theoretical perspective, PM itself can vary from relatively executive/strategic to relatively spontaneous automatic processes, depending on task demands (McDaniel & Einstein, 2000). However, PM as a whole exists on the executive strategic end of the spectrum of declarative memory processing (Craik, 1986) and involves: (1) forming an intention to complete a task at a later occasion (e.g., attend a doctor’s appointment) following a particular cue, (2) maintaining that intention in the context of other ongoing activities (e.g., taking care of grandchildren), (3) monitoring for and identifying the cue that prompts the intended action (e.g., recognizing it is the day of the doctor’s appointment), and (4) successfully executing the intention (e.g., attending the appointment). Thus, the neurocognitive resources required to successfully execute a PM intention are multifaceted and complex; in fact, some conceptual models consider PM to be an aspect of executive functions (Suchy, 2015). Higher-order domains of executive functions and retrospective memory are particularly important for executing PM tasks (Gupta et al., 2010; McDaniel & Einstein, 2000), and the extent to which these domains are drawn upon varies depending on the characteristics of the PM task being completed. For example, event-based PM tasks with focal or salient cues (e.g., a reminder call right before an appointment) rely more heavily on automatic processes, and have been shown to correlate with medial temporal networks (e.g., Gordon, Shelton, Bugg, McDaniel, & Head, 2011) and retrospective memory (Kamat et al., 2014). In contrast, time-based PM tasks (e.g., attending an appointment on a busy Wednesday at 10:00am) require executive/strategic monitoring supported by prefronto-parietal networks (Burgess, Scott, & Frith, 2003; Oksanen, Waldum, McDaniel, & Braver, 2014) and associated executive functions, including planning and cognitive flexibility (Zogg et al., 2011).

Given the known effects of aging on the structure and function of both medial temporal and prefrontal networks (e.g., Maillet & Rajah, 2013), it is not surprising that older adults perform more poorly on laboratory PM tasks compared to their younger counterparts (e.g., Henry, MacLeod, Phillips, & Crawford, 2004). Although age can affect both event- and time-based laboratory PM, it is the PM tasks that impose higher levels of strategic demands (e.g., non-focal cues) that are more sensitive to aging (Henry et al., 2004; Kamat et al., 2014; Park, Hertzog, Kider, Morrell, & Mayhorn, 1997), perhaps by way of their shared reliance on prefrontal systems (Cona, Scarpazza, Sartori, Moscovitch, & Bisiacchi, 2015). PM is strongly related to everyday functioning among older adults, independent of general neurocognitive deficits, psychiatric factors, demographics, and medical comorbidity (Hering, Kliegel, Rendell, Craik, & Rose, 2018; Woods, Weinborn et al., 2014; Woods, Weinborn, Velnoweth, Rooney, & Bucks, 2012). The jury is still out on whether time- or event-based PM is superior in predicting everyday functioning in older adults. Some studies show that time-based PM is more strongly associated with manifest everyday functioning (e.g., instrumental ADLs in Tierney, Bucks, Weinborn, Hodgson, & Woods, 2016), while other studies suggest that event-based PM shows stronger associations (e.g., medication management in Woods, Weinborn et al., 2014). Nevertheless, it is clear that PM is a strategically demanding aspect of declarative memory (Craik, 1986) that declines with age (Henry et al., 2004) and is a reliable predictor of manifest everyday functioning in older adults (Woods et al., 2012).

No previous studies have specifically asked whether PM modulates the relationship between older age and declines in manifest everyday functioning. Mediation analysis is important because it can explain important connecting mechanisms between two variables, and it specifically describes how some causal agent (predictor variable) transmits its effect on a criterion (Hayes, 2013). To this end, the primary aim of the current study was to determine whether PM mediates the effect of older age on manifest everyday functioning. Specifically, we hypothesized that both strategically-demanding time-based PM as well as event-based PM would mediate the effect of age and manifest everyday functioning, with time-based PM possibly having a stronger indirect mediating effect as indicated by a larger effect size compared to that of event-based PM. Next, in order to determine the possible unique mediating role of PM between age and everyday functioning, we also set a secondary aim of investigating whether executive functions or retrospective memory also mediate the relationship between increasing age and manifest everyday functioning. We predicted that both executive functions and retrospective memory would each serve as partial mediators of the effect between age and manifest everyday functioning.

Materials and Methods

Participants

The study procedures were approved by the human research ethics office at the University of California San Diego (UCSD). The sample was comprised of 468 individuals aged 18–75 years (M =42.9, SD = 12.0) enrolled in the UCSD HIV Neurobehavioral Research Program, which recruits from community-based organizations, local clinics, and regional advertisements. Baseline exclusion criteria for this study included an estimated verbal IQ score less than 70 on the Wechsler Test of Adult Reading (WTAR; Psychological Corporation, 2001), or prior diagnosis of any of the following: (1) severe psychiatric disorder (e.g., schizophrenia); (2) central nervous system opportunistic infection; (3) seizure disorder; (4) head injury with loss of consciousness for more than 30 min; (5) stroke with neurological sequelae; or (6) presence of a non-HIV major neurocognitive disorder. Individuals were also excluded if they had current substance dependence or tested positive on a breathalyzer or urine toxicology screen for illicit drugs (except marijuana) on the day of testing. The demographic and clinical characteristics of the sample are shown in Table 1.

Table 1.

Demographic and clinical characteristics of the study sample (N = 468).

Variable	Data
Age (years)	43.9 (12.0)
Gender (% [n] men)	79.3 (371)
Education (years)	13.7 (2.6)
Ethnicity (% [n])
Caucasian	56.4 (264)
African American	24.1 (113)
Hispanic	16.7 (78)
Other	2.7 (13)
Estimated Verbal IQ (WTAR)	102.7 (11.3)
Memory for Intentions Test (MIsT)
Time-based (raw score of 8)	5.7 (1.6)
Event-based (raw score of 8)	6.6 (1.6)
Everyday Functioning
Activities of Daily Living (no. domains impaired, of 14)	1.5 (2.5)
Karnofsky Performance Status Scale (of 100)	93.3 (9.6)
Employment Status (% [n] unemployed or disabled)	52.4 (245)
POMS Confusion/ Bewilderment Scale (of 28)	6.6 (5.2)
Retrospective Memory Measures
WMS-III Logical Memory II (raw score of 50)	25.7 (9.2)
CVLT-II Long Delay Free Recall (raw score of 16)	10.4 (3.6)
Executive Functions Measures
Action (Verb) Fluency (no. words generated)	16.6 (5.1)
WAIS-III Backward Digit Span (raw score of 14)	6.9 (2.3)
Tower of London (raw total moves)	32.5 (19.4)
Trail Making Test, Part B – Part A (seconds)	45.2 (36.7)
Comorbidity Burdena (no. conditions, range 0 – 4)	1.8 (1.1)
Hepatitis C Virus (% [n])	15.4 (72)
Human Immunodeficiency Virus (% [n])	65.0 (304)
Lifetime Substance Dependence (% [n])	50.4 (236)
Lifetime Affective Disorder (% [n])	51.0 (239)
Lifetime Major Depression (% [n])	49.6 (232)
Current Major Depression (% [n])	9.0 (42)
Lifetime Generalized Anxiety (% [n])	10.3 (48)
Current Generalized Anxiety (% [n])	3.0 (14)

Note. Data represent M (SD) or valid population % values. IQ = intelligence quotient; WTAR = Wechsler Test of Adult Reading; MIsT = Memory for Intentions Screening Test; POMS = Profile of mood states; WMS-III = Wechsler Memory Scales, 3^rd Edition; CVLT-II = California Verbal Learning Test, 2^nd Edition; WAIS-III = Wechsler Adult Intelligence Scale, 3^rd Edition;

^aComorbidity Index is constructed from Hepatitis C virus, HIV, Substance Dependence, and Lifetime Affective Disorder.

Materials

Everyday Functioning

Participants completed four well-validated indirect measures of manifest everyday functioning. These measures were considered as a single continuous manifest everyday functioning variable ranging from 0 (none of the 4 manifest everyday functioning variables considered “dependent”) to 4 (all 4 manifest everyday functioning variables “dependent”). Instrumental and basic activities of daily living (iADLs and bADLs) were assessed using the Heaton et al. (2004) version of the Lawton and Brody (1969) ADL scale. Individuals were considered ‘dependent’ within either ADL domain if they reported two or more areas in which there had been a decline from ‘best’ to ‘now.’ Self-reported unemployment was included in the composite of everyday functioning since individuals reporting unemployment or disability are reporting an observable (e.g., objective) indicator of manifest everyday functioning (Heaton et al., 2004). Participants reported themselves as: (1) employed, (2) unemployed, (3) disabled, or (4) retired, but were not asked to provide additional information regarding reasons for being unemployed or disabled. For this study, participants who reported their employment status as “unemployed” or “disabled” were classified as ‘dependent’, while those who reported being ‘employed’ or ‘retired’ were classified as ‘independent’. The Karnofsky Performance Status Scale (KPSS; Karnofsky & Burchenal, 1949) was administered by a research nurse blind to neuropsychological results who assigned a rating of overall everyday functioning that ranged from zero (dead) to 100 (normal, no evidence of disease). KPSS “dependence’ was operationalized as a score < 90, which is consistent with operationalized definitions of functional impairment for neurocognitive disorders (e.g., DSM-5; American Psychiatric Association, 2013). In addition, this cutpoint is widely used in functional studies of clinical populations (e.g., HIV; Sheppard, Iudicello, et al., 2015). Finally, participants completed the Profile of Mood States (POMS; McNair, Lorr, & Droppleman, 1981) confusion/bewilderment scale, which is a 7-item scale (score range 0–28) that assesses cognitive symptoms. For this scale, impairment was operationalized as a score ≥ 1.0 standard deviations above the mean of age- and gender-adjusted normative standards (Nyenhuis, Yamamoto, Luchetta, Terrien, & Parmentier, 1999). Self-reported cognitive symptoms were included since they are a major component of functional decline according to the DSM-5 criteria for syndromic neurocognitive disorders (e.g., American Psychiatric Association, 2013), as well as disorders such as mild cognitive impairment (Bondi et al., 2014) and subjective cognitive impairment (Jessen et al., 2014). Moreover, studies show that these cognitive symptoms are strongly associated with functional deficits (Blackstone et al., 2012; Heaton et al., 2004).

Using this approach, “dependence” rates across the 4 everyday functioning variables were as follows: POMS confusion-bewilderment scale = 26% (n=122), ADLs = 28% (n=129), unemployment = 48% (n=223), and Karnofsky = 17% (n=79). For the continuous everyday functioning outcome variable, the median (interquartile range) number of everyday functioning domains dependent was 1 [0, 2] across the entire sample.

Neuropsychological evaluation

All participants were administered a comprehensive neuropsychological test battery by certified research assistants.

Prospective memory

We assessed PM with a standardized, performance-based measure administered in the laboratory. Participants completed the research version (Woods, Moran et al., 2008) of the Memory for Intentions Screening Test (MIsT; Raskin, 2009) during which they completed an ongoing word-search distractor task while performing eight total PM intentions. The MIsT includes four time-based (e.g., “In 15 minutes, tell me that it is time to take a break”) and four event-based (e.g., “When I show you a postcard, self-address it”) cues that are balanced on response modality and delay. There are two possible points for each MIsT trial: one for responding at the correct time (i.e., within 15% of the target time for time-based trials) or to the appropriate cue (for event-based trials) and one for a correct response. These trials are then summed across cue type to create time-based and event-based subscales (ranges = 0-8). Spearman r_s between time-based MIsT and event-based MIsT scores in the current sample was .49.

Retrospective Memory

Participants completed two measures assessing delayed retrospective memory: the Logical Memory II subtest from the Wechsler Memory Scale, 3^rd Edition (WMS-III; Wechsler, 1997), and the Long Delay Free Recall score from the California Verbal Learning Test, 2^nd Edition (CVLT-II; Delis, Kramer, Kaplan, & Ober, 2000). Raw scores from these two measures were converted into sample-based z-scores and then averaged to create a retrospective memory composite. The mean r_s between these scores in the current sample was .54.

Executive functions

Participants completed the following measures of executive functions: Part B time minus Part A time from the Trailmaking Test (Army Individual Test Battery, 1944), Total Moves score from the Tower of London – Drexel Version (Culbertson & Zillmer, 1999), Backwards Digit Span from the Wechsler Adult Intelligence Scale, 3^rd Edition (WAIS-III; Wechsler, 1997), and total words generated on Action (verb) fluency (Woods et al., 2005). Raw scores were converted into sample-based z-scores then averaged to create an executive functions composite. The mean r_s for executive functions was .30 (range .13 to .47).

Clinical Characterization

All participants underwent a neuromedical exam, history, and phlebotomy/labs, which provided information on comorbid conditions (e.g., HIV disease, Hepatitis C virus). Lifetime mood and substance dependence was operationalized by the Composite International Diagnostic Interview (Version 2.1; World Health Organization, 1998). From these variables, a single composite comorbidity index was computed by summing the number of the following conditions: (1) HIV disease, (2) Hepatitis C virus, (3) lifetime history of affective disorder (i.e., Major Depressive Disorder, Generalized Anxiety Disorder), and (4) lifetime history of substance dependence. Thus, the comorbidity scale had a range of 0 to 4, with 0 being no comorbid conditions and 4 meaning the individual was HIV+, HCV+, a lifetime history of depression or anxiety, and a lifetime history of substance dependence.

Data Analyses

The primary analyses included 4 separate mediation models with the continuous age predicting the continuous everyday functioning impairments variable. These models included either: (1) MIsT time-based PM scores, (2) MIsT event-based PM scores, (3) the executive functions composite, or (4) the retrospective memory composite as mediators of the relationship between age and everyday functioning. Model covariates were selected from the variables in Table 1 using a data-driven approach. We included all potential covariates that were associated with any two of the three variables in each mediation model (i.e., age, mediator, and everyday functioning impairments) at a critical alpha of 0.10. Only years of education and comorbidities met these criteria and were included as covariates across all 4 models. All models were constructed in Mplus version 8 using maximum likelihood estimation and 95^th percentile bootstrap confidence intervals. All findings were interpreted through examining whether the bootstrapped 95% CI of unstandardized estimates from mediation models contained zero. Considering each PM variable as a possible moderator, we also conducted separate moderation models for both MIsT time-based PM and MIsT event-based PM scores considered as moderators. In these models, age was included the focal predictor while PM score (time- or event-based) and the PM by age interaction term were each included as predictors of the continuous everyday functioning outcome.

Results

Correlations Between Variables in Mediation Models

Table 2 shows the correlation (r_s) values between age, everyday functioning impairments, time-based MIsT, event-based MIsT, executive functions, retrospective memory, education, and comorbidities. Figure 1 shows the univariate (unadjusted) association between age and everyday functioning domains impaired (r_s = .23, p < .001). All variables included in mediation models were significantly related to each other at the bivariate level (all ps < .05) with effect sizes ranging from small to large (r_s range .09 to .52). A separate single multiple regression including time-based MIsT, event-based MIsT, executive functions, and retrospective memory (along with age, education, and comorbidities as covariates) predicting everyday functioning as a continuous variable was significant (Χ²[7] = 142.93, p <.001). Among the predictors, only time-based MIsT (b = −.092 [−.168, −.015]) and comorbidities (b = .393 [.301, .485]) were significant independent predictors of everyday functioning impairments. Age, education, event-based MIsT, executive functions, nor retrospective memory were significant independent predictors of everyday functioning impairment (all 95% CIs contained 0).

Table 2.

Correlation (r_s) values between the clinicodemographic, everyday functioning, and neuropsychological variables included in the mediation models (N = 468).

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
(1) Age	--
(2) Everyday Functioning	.23**	--
(3) Time-based Prospective Memory	−.34**	−.26**	--
(4) Event-based Prospective Memory	−.35**	−.27**	.49**	--
(5) Executive Functions	−.17**	−.24**	.40**	.30**	--
(6) Retrospective Memory	−.17**	−.29**	.45**	.32**	.52**	--
(7) Education	.16**	−.18**	.16**	.13**	.32**	.33**	--
(8) Comorbidity Index	.17**	.45**	−.17**	−.12*	−.18**	−.28**	−.26**	--

^**p < .001

^*p < .05

Figure 1.

Univariate (unadjusted) r_s association between age and everyday functioning domains impaired.

Time- and Event-based Prospective Memory Mediation Models

As shown in Figure 2(A), time-based MIsT scores significantly partially mediated the relationship between age and everyday functioning impairments after controlling for education and comorbidities. The indirect effect of age on everyday functioning impairments was significant, b = .006, 95% CI [.003, .010], p < .01, while the direct effect of age on everyday functioning impairment also was significant, b = .011, 95% CI [.003, .020], p < .05, when time-based MIsT was used as a mediator. Examination of the bootstrap estimate indicates that controlling for education level and comorbidities, a participant who is 10 years older would be expected to have 0.06 more everyday functioning impairments, through the influence of time-based MIsT scores.

Figure 2.

Mediation model with unstandardized bootstrap estimates and 95% confidence interval values for age and everyday functioning mediated by (A) time-based PM scores and (B) event-based PM scores.

Similarly, Figure 2(B) shows that event-based MIsT scores significantly partially mediated the relationship between age and everyday functioning impairments after controlling for education and comorbidities. The indirect effect of age on everyday functioning impairments was significant, b = .005, 95% CI [.002, .009], p < .01, while the direct effect of age on everyday functioning impairments also was significant, b = .013, 95% CI [.004, .021], p < .01, with event-based MIsT as a mediator. Examination of the bootstrap estimate indicates that controlling for education level and comorbidities, a participant who is 10 years older would be expected to have 0.05 more everyday functioning impairments, through the influence of event-based MIsT scores.

Executive Functions and Retrospective Memory Mediation Models

Examining executive functions as a possible mediator revealed a significant indirect effect of age on everyday functioning impairments, b = .003, 95% CI [.001, .005], p < .01, while the direct effect of age on everyday functioning impairments also was significant, b = .015, 95% CI [.006, .023], p < .01. Examination of the bootstrap estimate indicates that controlling for education level and comorbidities, a participant who is 10 years older would be expected to have 0.03 more everyday functioning impairments, through the influence of executive functions.

Similarly, in a separate model there was a significant indirect effect of age on everyday functioning impairments through retrospective memory, b = .002, 95% CI [.001, .005], p < .05, while the direct effect of age on everyday functioning impairments also was significant, b = .016, 95% CI [.007, .024], p < .01. Examination of the bootstrap estimate indicates that controlling for education level and comorbidities, a participant who is 10 years older would be expected to have 0.02 more everyday functioning impairments, through the influence of retrospective memory.

As suggested by an anonymous reviewer, we have included a post hoc figure for descriptive purposes showing of the strength of the mediating effects (and 95% CIs) across mediators for each individual everyday functioning outcome (see Figure 3). Mediated effect sizes between age and everyday functioning outcomes through currently examined mediators ranged from 0.02 to 0.06 standard deviations (fully standardized with age and each everyday functioning outcome), and these mediated effects were broadly comparable across all everyday functioning outcomes.

Figure 3.

Fully standardized mediated effect sizes between age and four separate everyday functioning variables (activities of daily living, Karnofsky scale of performance status, Profile of Mood States confusion-bewilderment scale, and employment status) across different mediators (time-based prospective memory, event-based prospective memory, executive functions, retrospective memory).

Note. ADL = Lawton & Brody Activities of Daily Living number of domains impaired; Karnofsky = Karnofsky Scale of Performance Status raw scores; POMS C/B = Profile of Mood States confusion/bewilderment scale age- and gender-adjusted z-scores; Employment = employment status as “unemployed” or “disabled.” Effect size for employment was computed using mean and SD of a point biserial variable in order to compare across everyday functioning measures. Error bars indicate 95% confidence interval for each effect size.

Prospective Memory and Executive Functions as Possible Moderators

In separate moderation models with age as the focal predictor and time-based and event-based MisT as separate moderators, controlling for education and comorbidities, there were no significant interactions between age and either time- or event-based MIsT or executive functions in predicting everyday functioning (all 95% CIs included 0). In a separate moderation model with age as the focal predictor and executive functions as the moderator, controlling for education and comorbidities there was no significant interaction between age and executive functions in predicting everyday functioning, b = .004, 95% CI [−.008, .014], p > .10.

Discussion

The present study examined whether executively demanding aspects of declarative memory play a mediating role between an individual factor (i.e., age) and daily life activities. Specifically, we sought to determine whether PM mediates the effect of older age on everyday functioning. Results indicated that both time-based and event-based PM were partial mediators of the effect of older age on everyday functioning impairments. In other words, PM plays an important, but not exclusive explanatory role in everyday functioning problems that commonly accompany older age. In parallel, both executive functions and retrospective memory were significant individual mediators of the relationship between age and everyday functioning impairments, but they demonstrated mediating effect sizes that were weaker than those of PM mediators. Taken together, these findings highlight the importance of considering PM as path through which aging affects everyday functioning, and highlights a neurocognitive function that could be addressed to improve daily functioning among older adults.

Findings from the primary mediation models revealed that PM served as a partial mediator between increasing age and everyday functioning impairments. As expected, older age was associated with lower PM and a higher risk of everyday functioning impairments, which in this study was defined as a composite of vocational status, ADL declines, clinician-rated functioning, and self-reported cognitive symptoms. Importantly, however, the indirect path of age predicting everyday functioning impairments through both time- and event-based PM also were significant. This means that both relatively strategic and automatic PM processes at least partially explain the role that aging plays in daily living difficulties. Moreover, the fact that the direct path remained significant suggests that the association between age and everyday functioning is robust and not fully accounted for by PM. The effect sizes associated with the PM mediators indicated that as an adult ages by 10 years they expected to have 0.06 more everyday functioning domains impaired through the influence of time-based PM and 0.05 domains through the influence of event-based PM. Put another way, compared to a 20-year-old, a 60-year-old would be expected to have 0.24 (through time-based PM) and 0.20 (through event-based PM) more everyday functioning domains impaired. The small effect is of both conceptual and clinical relevance because it describes the magnitude of the explanatory role of PM in the well-established association of older age and risk of functional dependence (Farias et al., 2018; Schmitter-Edgecombe & Parsey, 2014; Tucker-Drob, 2011); in addition, it is independent of the effects of comorbidities (e.g., viral infections, mood and substance use disorders) and education. Taken together, these findings suggest that PM is of incremental importance to daily life in the context of aging above these important factors, which themselves can increase with older age. Moreover, the clinical importance of the currently studied everyday functioning outcomes (e.g., declines in activities of daily living such as medication management and handling finances) cannot be understated, especially since functional dependence is often cited as one of the most concerning aspects of aging reported by older adults (Prince & Butler, 2007) and is a key feature of major neurocognitive disorders (American Psychiatric Association, 2013) that has downstream effects on quality of life (Woods et al., 2015). These data highlight the incremental importance that PM plays in explaining how age exerts its effects on daily life.

Contrary to our hypotheses, findings revealed comparable effect sizes for time-based and event-based PM as mediators of the effect of aging on everyday functioning, although the effect size for time-based PM was nominally larger than that of event-based PM (.006 for time- and .005 for event-based PM). It is widely agreed that relatively more executively demanding PM tasks (e.g., time-based tasks, non-focal event-based tasks) are more vulnerable to the effects of aging (Henry et al., 2004; Kamat et al., 2014; Park et al., 1997); yet, the findings linking PM and everyday functioning in older adults are somewhat mixed. Measures of both time- and event-based PM have shown strong associations with everyday functioning outcomes (Tierney et al., 2016; Woods, Weinborn et al., 2014). Our findings showing relatively similar mediating effect sizes between time- and event-based PM in the context of aging and global everyday functioning (i.e., the ‘b’ path) stand in parallel to this literature. In our data, we also observed comparable associations at the univariate level between age and both time- (r_s = −.34) and event-based PM (r_s = −.35), even though time-based PM showed relatively lower mean performance scores (d = 0.54; thus suggesting it was more difficult). However, examining each variable considered as a mediator (time-based PM, event-based PM, executive functions, retrospective memory) as predictors in a regression model revealed that time-based PM has an independent direct effect on everyday functioning over and above event-based PM (as well as executive functions and retrospective memory). This suggests that although the total indirect path between age and everyday functioning may be explained with similar magnitudes of effects for both time- and event-based PM, the direct effect of PM on everyday functioning may be more uniquely associated with time-based PM difficulties. Nevertheless, a possible explanation of time-based PM and event-based PM having comparable mediation effect sizes using a continuous age predictor is that the research version of the MIsT uses subtle event-based cues that are non-focal to the ongoing task and therefore are not entirely automatic or spontaneous. Moreover, from the perspective of PM theory, irrespective of the relatively strategic or automatic demands of the cues, PM tasks themselves lie on the executive/strategically demanding end of the declarative memory continuum (Craik, 1986; McDaniel & Einstein, 2003). Future studies may wish to examine the contributions of highly automatic PM to everyday functioning in older adults, or perhaps whether supporting strategic processing (e.g., elaborative encoding, monitoring training) can enhance everyday functions.

Interestingly our analyses revealed that compared to PM, executive functions and retrospective memory served as significant, but weaker, mediators of the relationship between older age and everyday functioning impairments. Thus, executive functions (as indexed by verbal fluency, auditory working memory, planning, and complex attention) and retrospective memory (as indexed by delayed recall of words and stories) also play an explanatory role in the constellation of everyday functioning difficulties in older age. These findings are consistent with previous investigations implicating executive functions as an important predictor of everyday functioning among older adults (McAlister & Schmitter-Edgecombe, 2016; Royall et al., 2004; Grigsby et al., 1998) and clinical populations such as those with MCI and dementia (Tomaszewski Farias et al., 2009).

However, the mediating effects of executive functions and retrospective memory were weaker than for PM, as evidenced by: (1) PM having larger partially mediating effect sizes compared to those of executive functions and retrospective memory, (2) time-based PM being the only independent predictor of everyday functioning impairments when included in a multiple regression with event-based PM, executive functions, and retrospective memory, and (3) retrospective memory being a significant partial mediator of only one of the four everyday funcitoning outcomes (i.e., Karnofsky Scale of Performance Status; see Figure 3). Our findings also highlight the specific role that PM, as opposed to traditional neurocognitive components that contribute to PM functioning, might play in explaining how aging affects everyday functioning. While the cognitive architecture of PM includes aspects of both retrospective memory and executive functions, PM is considered a translational construct for which the whole is greater than the sum of its parts (Avci et al., 2018). In parallel, the current findings could indicate that our measure of PM (i.e., the MIsT) more closely parallels the cognitive architecture required for the activities individuals perform in their daily life. For example, list-learning and memory for prose passages (e.g., retrospective memory) may not as closely parallel the ecological, self-initiated demands for taking a medication or completing daily tasks to maintain employment as does PM, which requiring monitoring for cues in the face of a distractor task (Craik, 1986; Einstein & McDaniel, 2000). The specificity of PM serving as a mediator underscores the importance for future studies to assess specific cognitive skills such as PM in addition to traditional neurocognitive domains that are affected by age and contribute to everyday functioning changes.

There are multiple limitations of the current study that are important to consider. First, the present analyses utilized cross-sectional data in the context of a mediation design, which precludes inferences regarding temporal causality between aging, prospective memory declines, and subsequent everyday functioning impairment. However, the theoretical framework for aging affecting PM and then subsequently causing everyday functioning declines is congruent with both the typical theoretical trajectory of normal aging to dementia (e.g., full dependence of activities of daily living) and longitudinal studies showing that PM declines are associated with declines in everyday functions (Tucker-Drob, 2011). Nevertheless, future studies should examine whether the effect of age and everyday functioning through PM can be observed in prospective, longitudinal studies. In fact, previous studies have shown that PM performance predicts incidence neurocognitive disorders (Sheppard, Woods, et al., 2015); therefore, future studies should determine whether the effect of age on everyday functioning through PM might be useful for predicting broader clinical declines in neurocognitive and everyday functioning (e.g., syndromic neurocognitive disorders). Second, the present sample included a group of individuals across the lifespan, but future studies will need to examine and replicate findings showing the mediating role of PM between aging and everyday functioning within a subgroup of older adults. For example, while only 7% of the sample was age 60+, using effect size estimates it would be estimated that a 60-year-old would be expected to have 0.24 more everyday functioning domains impaired compared to a 20-year-old, through the influence of time-based prospective memory. As such, future studies will need to determine the strength of mediating effects of prospective memory within older adult samples. Another related future direction is to examine the role of aging within clinical populations. For example, previous studies of HIV+ individuals have shown possible accelerated neurocognitive aging and higher rates of mild cognitive impairment (Sheppard, Iudicello, et al., 2015; Sheppard et al., 2017), which is a risk factor for syndromic neurocognitive disorders, future studies should determine whether mediating effects of PM between aging and everyday functioning are more pronounced among those with infectious disease (e.g., HIV, HCV). Another limitation of the current study is that measures of everyday functioning were indirect, such as employment status being self-reported (e.g., not verified) and not including reasoning for being unemployed or disabled. Despite this measurement issue, it is interesting to note that the effect sizes across all 4 cognitive mediators for this employment variable were commensurate with the effect sizes of the remaining everyday functioning variables (see Figure 3). Future studies should also set forth to confirm that individuals are unemployed using external verification, and consider whether difficulties with employment are due to personal difficulties (e.g., neurocognitive impairment, disability) opposed to other external reasons (e.g., layoffs). Finally, the present study utilized a sample that was comprised of high rates of medical comorbidities, including Hepatitis C virus, HIV, and lifetime histories of mood and substance use disorders that might limit generalizability of the currents findings to healthier samples.

Notwithstanding these limitations, the present data underscore the important role that PM plays in aging and everyday functioning, and point towards the possible benefit of targeting PM processes for improving everyday functioning outcomes. Although there are no currently validated restorative interventions for PM, compensatory approaches that integrate cognitive psychology and behavioral methods may be effective. For example, at the level of encoding, prior studies have shown that implementation intentions, visualization, calendaring, and increasing the semantic relatedness of the cue-intention paring can improve PM performance (Altgassen et al., 2015; Faytell et al., 2017) and PM-based health outcomes (Pennar et al., in press). After the cue-intention is encoded, PM accuracy can be improved by supporting effective monitoring strategies, for example introducing a delay between cue onset and enacting the intention (Loft et al., 2014) or emphasizing the importance of PM task versus the ongoing task (Woods, Doyle, et al., 2014). Finally, increasing the salience of the cue, for example by increasing its focality to the ongoing task (Kliegel, Jäger, & Phillips, 2008) and/or setting well-timed alarms can improve PM (Faytell et al., 2018). In all cases, the PM interventions should be highly conscious of the PM demands of the everyday activity. Drawing back to the example of an older adult intending to attend a healthcare appointment at 3pm, one might: (1) briefly verbalize a “when-then” implementation intention and visualize successfully attending the appointment at the right time, (2) set an alarm with a specific reminder of the time at which one would leave for the appointment, and/or (3) emphasize the importance of attending the appointment above other daily activities. Taken together, the present study provides a foundation for future studies to examine how such interventions might ameliorate the mediating effect of PM between aging and everyday functioning.