Multiple Types of Memory and Everyday Functional Assessment in Older Adults

Abstract

Objective

Current proxy measures for assessing everyday functioning (e.g., questionnaires, performance-based measures, and direct observation) show discrepancies in their rating of functional status. The present study investigated the relationship between multiple proxy measures of functional status and content memory (i.e., memory for information), temporal order memory, and prospective memory in an older adult sample.

Method

A total of 197 community-dwelling older adults who did (n = 45) or did not meet (n = 152) criteria for mild cognitive impairment (MCI), completed six different assessments of functional status (two questionnaires, two performance-based tasks, and two direct observation tasks) as well as experimental measures of content memory, prospective memory, and temporal order memory.

Results

After controlling for demographics and content memory, the temporal order and prospective memory measures explained a significant amount of variance in all proxy functional status measures. When all variables were entered into the regression analyses, content memory and prospective memory were found to be significant predictors of all measures of functional status, whereas temporal order memory was a significant predictor for the questionnaire and direct observation measures, but not performance-based measures.

Conclusion

The results suggest that direct observation and questionnaire measures may be able to capture components of everyday functioning that require context and temporal sequencing abilities, such as multi-tasking, that are not as well captured in many current laboratory performance-based measures of functional status. Future research should aim to inform the development and use of maximally effective and valid proxy measures of functional ability.

Introduction

Functional ability generally refers to an individual's ability to care for oneself, manage one's affairs, and live relatively independently in the community (Willis, 1996). Accurate and efficient assessment of an older adult's functional ability is important for evaluating and addressing safety concerns, informing clinical diagnoses and treatment, and developing a more comprehensive understanding of the aging process. Common methods used to operationalize and quantify Instrumental Activities of Daily Livings (IADLs), include questionnaires, performance-based measures, and direct observation methods. However, these methods do not always correlate highly with each other and can show significant discrepancies in their ratings of functional ability (e.g., Burton, Strauss, Bunce, Hunter, & Hultsch, 2009; Kivinen, Sulkava, Halonen, & Nissinen, 1998; Schmitter-Edgecombe, Parsey, & Cook, 2011; Schmitter-Edgecombe & Parsey, 2014b), indicating that the methods are not assessing completely overlapping constructs. Understanding differences between proxy measures of functional ability is important so that proper assessment techniques can be chosen (or a new one developed) to optimize the process of getting a valid assessment of an older adult's everyday functional ability level.

Methods for measuring functional status each have unique strengths and weaknesses. For example, although self- and informant-report questionnaires are inexpensive, relatively easy to administer, and capture ability levels across a variety of situations, they are subject to considerable reporter bias (Kemp, Brodaty, Pond, & Luscombe, 2002; Martyr, Nelis, & Clare, 2014). Questionnaire methods are also vulnerable to differential interpretation of questions or answer options, as well as to interference by factors such as culture, language, and level of education. Additionally, as cognitive decline progresses, individuals may become less reliable assessors of their own functional ability (Clare, Whitaker, & Nelis, 2010; Graham, Kunik, Doody, & Snow, 2005; Onor, Trevisiol, Negro, & Aguglia, 2006), and for these individuals, informant-report of ability has been found to relate more closely to objective cognitive assessment data than self-report (Farias, Mungas, & Jagust, 2005).

Performance-based tasks involve having the individual complete activities that approximate IADLs in the presence of a trained observer. These measures are more objective than questionnaires in that they are usually highly scripted and have pre-determined coding systems. However, individuals completing a performance-based task are in a controlled environment and are being cued and prompted by an examiner; thus, these individuals are not required to self-initiate and multi-task in the same ways that they would be in a more realistic setting (Moore, Palmer, Patterson, & Jeste, 2007).

In contrast, direct observation measures require an individual to carry out certain activities either in their own home or in a naturalistic environment (e.g., Fisher & Jones, 2012; Schmitter-Edgecombe, McAlister, & Weakley, 2012). Direct observation measures may be the most ecologically valid type of functional ability measures, as they allow researchers to make use of more open-ended tasks and to operationalize and analyze subtle behavioral changes and processes that are not easily measured in questionnaire or performance-based measures (Griffith et al., 2003; Wadley et al., 2009). However, these measures may still be subject to certain limitations that threaten ecological validity, including influences like unfamiliarity with the simulated environment or assigned activity, decreased motivation, and assessment anxiety. These measures can also be difficult to annotate and score, making them often time-consuming and inconvenient, and vulnerable to subjectivity.

Few prior studies have examined questionnaires, performance-based and direct observation measures as proxies of functional status in a single study (e.g., Schmitter-Edgecombe, Parsey, & Cook, 2011). By exploring the relationship between cognitive constructs thought to play an important role in supporting everyday activities of daily living and these functional status measures, we may bfe able to identify a potential source of difference between functional outcome measures. More specifically, memory is a significant cognitive factor that influences functional ability levels in older adults (Farias et al., 2005; Stuck et al., 1999; Tuokko, Morris, & Ebert, 2005). While memory has often been measured as a unidimensional construct (i.e., content memory), there are multiple unique types of memory that may affect everyday functional abilities and may be differentially captured by these functional status measures.

Content memory refers to the information that is presented (Craik & Jennings, 1992). For example, recalling a word list or remembering the information “I am out of milk” are examples of content memory. Contrastingly, temporal order memory and prospective memory more significantly capture the context in which the information is presented and/or retrieved rather than the information itself. Temporal order memory refers to memory for the order in which events occurred (Janowsky, Shimamura, & Squire, 1989; Underwood, 1977), whereas prospective memory refers to the ability to remember to do something in the future (Dobbs & Rule, 1987; McDaniel & Einstein, 1996). For example, “I noticed before I left for work that I am out of milk” captures temporal order memory, whereas remembering to get milk when passing the grocery store on the way home from work is an example of prospective memory (event-based). Content memory is traditionally linked with the temporal lobes of the brain (Gordon, Rissman, Kiani, & Wagner, 2014; Squire & Zola-Morgan, 1991; Wixted & Squire, 2011), whereas the frontal lobes have been shown to play a key role in temporal order and prospective memory (DeVito & Eichenbaum, 2011; Gilbert, 2011; Rubens & Zanto, 2011). As described subsequently, these three types of memory (content, temporal order, and prospective) all contribute to an individual's ability to complete everyday activities, and deficits in any one of these areas can interfere with successful everyday functioning. It is important to note that, although temporal order, prospective and content memory are considered distinct types of memory, they also interact. For example, the success of prospective memory relies on the individual also successfully recalling content information (i.e., what they are to do) from memory.

Content memory (typically assessed using list and prose memory tasks) has emerged as an important contributor to functional ability in older adults across several studies (Farias et al., 2005; Stuck et al., 1999; Tuokko et al., 2005). For example, content memory was found to provide a unique contribution to informant-reported IADL ability of older adults at various points on the spectrum from healthy aging to dementia longitudinally over the course of 5 years (Farias et al., 2009). There is also evidence for content memory's role in contributing to functional ability level both cross-sectionally and longitudinally within a community sample of older adults who did not have dementia (Koehler et al., 2011). However, the research on content memory's influence on IADL ability is still variable, as some studies have shown that the link between content memory and functional ability may not be strong, especially when predicting long-term functional ability (Cahn-Weiner, Malloy, Boyle, Marran, & Salloway, 2000; Monaci & Morris, 2012; van Hooren et al., 2005).

A few studies have demonstrated that prospective memory and temporal order memory play an important role in supporting everyday activities of daily living. In healthy older adults, prospective memory was uniquely associated with IADL ability, as assessed by questionnaire measures (Woods, Weinborn, Velnoweth, Rooney, & Bucks, 2012). Jones, Livner, and Bäckman (2006) found that prospective memory made an independent contribution to the prediction of the development of Alzheimer's disease in pre-clinical Alzheimer's patients over and above the contribution of retrospective (content) memory, although functional ability was not directly assessed. Temporal order memory, source memory and prospective memory were also found to make an independent contribution to informant-rated IADL performance over and above that of content memory in a sample of community dwelling older adults (Schmitter-Edgecombe, Woo, & Greeley, 2009). Given that most current laboratory performance-based task are structured and afford scaffolding and task isolation, such tasks may be less likely than questionnaires and direct observation measures to capture the influence that poor prospective and temporal order memory abilities may have on an individual's ability to complete everyday tasks.

In this study, we expand prior research by examining the relationships between multiple types of memory (measured using an experimental memory paradigm) and three different proxy measures of functional abilities (i.e., questionnaires, performance-based measures, and direct observation measures) in a community-dwelling sample of older adults. Examining these relationships may help to increase understanding of factors that may contribute to the common lack of correlations and discrepancies found across functional status measures in older individuals (Burton et al., 2009; Kivinen et al., 1998; Schmitter-Edgecombe et al., 2011; Schmitter-Edgecombe & Parsey, 2014b). Participants who ranged from healthy to having mild cognitive impairment (MCI) completed memory tasks that assessed content memory, temporal order memory, and prospective memory. It was hypothesized that temporal order and prospective memory would make an independent contribution over and above content memory to performance on the questionnaire and direct observation measures, but not the performance-based measures. Specifically, individuals with higher prospective memory and temporal order memory scores were expected to perform better (i.e., show a higher level of independent ability) on questionnaire and direct observation measures of functional ability. This is because direct observation and the real-life tasks assessed by questionnaires require individuals to self-initiate and multi-task and to rely on situational and environmental context where prospective and temporal order memory would often be utilized to support performance. In contrast, most laboratory performance-based tasks are structured, afford scaffolding and do not require multi-tasking, and are therefore likely to be less reliant on prospective memory and temporal order memory for task completion.

Methods

Participants and Procedure

Participants were 197 community-dwelling older adults between the ages of 60 and 94 (M = 71.89, SD = 7.53). Forty-five of the participants met criteria for MCI. Tables 1 and 2 provide demographic and relevant comparison information for the MCI and cognitively healthier older adults in addition to the full sample. For the purposes of this study, rather than separating the MCI group, the two groups were combined and treated as a community-dwelling older adult sample along a continuum of functional ability levels. It is important to study this continuum of functional ability levels as deficits in IADLs have been found to be stronger predictors than an MCI diagnosis of progression to dementia (Pérès et al., 2008; Purser, Fillenbaum, Pieper, & Wallace, 2005). Furthermore, examining older adults across a continuum of cognitive impairment allows for better evaluation of a wide range of cognitive and functional abilities (Seligman, Giovannetti, Sestito, & Libon, 2014).

Table 1.

Predictor means and standard deviations by group

Variable or test	HOA n = 152		MCI n = 45		F-valuea	Total N = 197
	M	SD	M	SD		M	SD	Range
Age	71.68	7.55	72.60	7.48	t = −0.72	71.89	7.53	60–94
Education	16.35	2.92	14.89	3.21	t = 2.87*	16.01	3.04	8–20
Activity content memory	12.38c	1.91	11.06d	2.10	6.49*	12.10	2.02	6–16
MAS prose memory	6.01e	1.35	4.69f	1.31	16.71**	5.71	1.45	2–9
Prospective memory	11.76g	3.83	9.30h	5.68	4.82*	11.22	4.40	0–16
Temporal order memoryb	10.35i	5.35	14.65h	6.43	8.65**	11.23	5.84	0–36

Note: MAS = Memory Assessment Scale.

^aF-value with education as a covariate.

^bHigher scores represent poorer performance.

^cn = 136.

^dn = 36.

^en = 152.

^fn = 45.

^gn = 133.

^hn = 37.

ⁱn = 143.

*p < .01. **p < .001.

Table 2.

Outcome means and standard deviations by group

Variable or test	HOA n = 152		MCI n = 45		F-valuea	M	Total N = 197
	M	SD	M	SD			SD	Range
Questionnaire
IADL-C selfb	38.99c	12.72	49.08d	24.43	6.66*	41.13	16.38	27–133
IADL-C KIb	33.33e	8.18	47.40f	25.72	12.98**	36.95	15.94	27–125
Direct observation
Eight Activitiesb	14.27g	3.55	17.38h	5.87	9.53**	14.91	4.30	8–29
Day Out Taskb	13.04i	2.99	15.00j	4.08	6.84**	13.43	3.32	8–27
Performance-based
EPT	21.04k	2.39	18.82l	3.89	19.86**	20.55	2.92	12–24
OTDL-R	22.59m	3.06	20.16n	3.38	16.32**	22.05	3.29	10–28

Note: IADL-C = Instrumental Activities of Daily Living: Compensation Scale (KI = knowledgeable informant); EPT = Everyday Problems Test; OTDL-R = Revised Observed Test of Daily Living.

^aF-value with education as a covariate.

^bHigher scores represent poorer performance.

^cn = 145.

^dn = 39.

^en = 110.

^fn = 38.

^gn = 144.

^hn = 37.

ⁱn = 114.

^jn = 29.

^kn = 139.

^ln = 39.

^mn = 151.

ⁿn = 43.

*p < .01. **p < .001.

Participants were recruited through community advertisements, health and wellness fairs, physician referrals, and previous studies in the laboratory. The current study was part of a larger research study that investigated the relationship between cognition and everyday functioning in the aging population (Masked Reference 1). Participants completed a medical screening over the phone to rule out head trauma with permanent brain lesion, history of cerebrovascular accidents, current or recent (within the past year) psychoactive substance abuse, and known neurological, medical, or psychiatric causes of cognitive dysfunction (e.g., epilepsy and schizophrenia). The initial phone screening also included the Clinical Dementia Rating instrument to screen for the presence of dementia (Hughes, Berg, Danziger, Coben, & Martin, 1982; Morris, 1993), and the Telephone Interview of Cognitive Status (Brandt & Folstein, 2003) to obtain an initial estimate of cognitive functioning. Individuals with a Clinical Dementia Rating score greater than or equal to 1 (indicating possible mild dementia), a Telephone Interview of Cognitive Status score below 20 (indicating moderate to severe cognitive impairment), or who endorsed more than 10 symptoms of depression on the short-form of the Geriatric Depression Scale, indicating a potential presence of significant depression (Yesavage & Sheikh, 1986), were excluded from participation.

Participants were classified into the MCI category using established criteria (Petersen & Morris, 2005; Petersen et al., 2001). These criteria included self- or informant-report of subjective impairment in memory for at least 6 months, objective evidence of impairment as determined by a score of 1.5 or more standard deviations below age-matched norms or relative to prior testing scores on a measure in one or more cognitive domains (i.e., memory, language, executive functioning, and/or speeded processing; see Schmitter-Edgecombe and colleagues (2012) for list of standardized neuropsychological test used in each domain for diagnosis), generally preserved functional abilities as indicated by Clinical Dementia Rating score of 0.5 or less, and failure to meet Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision criteria for dementia. To determine diagnostic category, two experienced neuropsychologists reviewed informant and participants interview data, neuropsychological assessment results, CDR ratings, and available medical information (e.g., prior neuropsychological assessments, brain imaging). The memory measures used as predictors and the functional status measures examined in this study were not used in diagnosis. Both participants with amnestic and nonamnestic and both single and multidomain MCI were included in the current sample. The Institutional Review Board at Washington State University approved the research protocol and all participants provided informed consent prior to participation in the study.

Participants who met the initial screening criteria were invited to complete a series of standardized and experimental neuropsychological assessments across two sessions that lasted approximately 3 hr each. The first session occurred in the laboratory setting, where the standardized assessment of content memory and the performance-based tasks occurred. The second session occurred in an apartment located on the (masked location) campus where participants completed several naturalistic tasks (e.g., sweeping, filling a pill holder with correct medications). During the second session, participants were administered the direct observation measures and the activity memory paradigm used to obtain scores of content, prospective, and temporal order memory. Sessions were scheduled approximately two weeks apart (M = 13.83 days, SD = 26.98). In between sessions, participants were given a packet of questionnaires to complete and knowledgeable informants were interviewed by phone. As compensation for their time, participants received a report that detailed their performance on the standardized neuropsychological measures administered. They were also given pre-paid parking passes for the sessions and a travel stipend.

Measures

Memory Measures

Activity memory paradigm

This activity memory paradigm (Schmitter-Edgecombe, Woo & Greeley, 2009) consisted of the participant completing eight different everyday activities, each lasting approximately 1–8 min, and provided measures of content memory, temporal order memory, and prospective memory. The paradigm was repeated twice within the session (16 distinct everyday activities were performed), with one iteration measuring incidental memory and the other measuring intentional memory. Thus, 16 total tasks were completed. The eight IADL-like activities for the incidental memory iteration included sweeping and dusting, filling a medication dispenser, writing out a birthday card and check, watching a DVD, watering houseplants, answering a phone call, cooking soup, and selecting a work outfit. The eight IADL-like activities for the intentional memory iteration included changing a light bulb, washing hands, cleaning kitchen countertop, using a telephone and phone book, sorting and folding laundry, filing mail, and providing game instructions. Initially, participants were not informed that they would need to remember each of the eight activities or the order in which they completed the activities, thus assessing incidental memory. For the second set of eight activities, participants were informed that they would need to remember the tasks and their order.

Prospective memory

Before beginning the event-based prospective memory paradigm, participants were informed that the examiner wanted to see how well they could remember to do something in the future without being reminded. Participants were told that they needed to remind the experimenter to record the clock time several times over the course of the next hour. Participants were informed that after they completed each activity, they would be asked by the examiner to rate how much mental effort they felt the activity required on a scale from 0% (little mental effort) to 100% (significant mental effort). Participants were told that completing the mental effort rating would be their cue to tell the experimenter to record the current clock time. The paradigm was not started until the participant clearly understood the instructions for the task. Once the task began, there was no additional reference to the prospective memory task, but the mental effort rating scale remained within view of the participant. Participants who did not tell the experimenter to record the clock time after the final mental effort rating were queried to determine whether or not they remembered the instructions for the prospective memory task. If the participant did not recall the prospective memory instructions (indicating a primarily retrospective, rather than prospective, memory failure), the participant's data was excluded from analysis for this study. This paradigm was completed twice within the session. A score for prospective memory was obtained by summing the total number of times the participant correctly remembered to tell the experimenter to record the clock time after completing the mental effort rating across both sets of eight activities. Cronbach's alpha for the prospective memory score was 0.70.

Content memory (“Memory for Activities”)

After each set of eight activities was completed, participants were asked to free-recall, in any order, the tasks that they just completed. Participants were instructed to provide enough of a description of each task so that the examiner knew which task they were referring to, and the examiner asked for clarification when necessary. The number of the 16 tasks that the participant was able to free-recall was used as one measure of content memory (hereafter referred to as memory for activities). Cronbach's alpha for the content memory score was 0.53.

Temporal order memory

After the free-recall task was complete, eight cards with descriptions of the tasks were placed in front of the participant to measure temporal order memory. Participants were told to arrange the cards in the order that they completed the eight tasks. This procedure was repeated for both iterations of the paradigm. The temporal order memory score was calculated by summing the absolute differences between the participant's order for each task and the actual order in which the task occurred (Mangels, 1997; for example, if the participant recalled the first task as occurring third, an absolute difference of two is calculated). The total temporal order memory score from each set of eight activities was added together to obtain a summed temporal order memory score. Cronbach's alpha for the temporal order memory score was 0.66.

Memory Assessment Scale: prose memory subtest

For a second measure of content memory, participants were administered the prose memory subtest from the MAS battery (Williams, 1991). For this subtest, participants were read a short story consisting of three sentences once. Participants were asked to recall the story immediately after it was read and to answer nine questions about the story. After a time delay during which other tasks were completed, participants were again asked to free recall details from the story, and then were asked the same nine questions about the story again. The total of the nine questions answered correctly after delayed recall was used as a second measure of content memory, hereafter referred to as prose memory. Test–retest reliability was reported to be >.70 for all subtests of the MAS (Williams, 1993).

Functional Status Measures

Questionnaires

Instrumental Activities of Daily Living: Compensation Scale (IADL-C): self- and informant-report

The IADL-C (Schmitter-Edgecombe, Parsey, & Lamb, 2014) is a 27-item questionnaire that assesses functional ability and compensatory strategy use in older adults. There are both self-rated and informant-rated forms of the measure, and both of these forms were used in this study. Each item consists of a description of an activity, followed by a Likert scale for the individual or informant to indicate the individual's current ability level in completing this activity (1: independent, no aid to 8: not able to complete activity anymore; there are also options to indicate that the individual does not need to complete the activity or that the informant has no basis for judgment). The questionnaire measures four specific domains of IADLs (money and self-management, home daily living, travel and event memory, and social skills) and produces sub-scores for each domain as well as a total overall score. In the current study, the total overall score from the self-rated and informant-rated IADL-C was used in the analyses. This questionnaire has demonstrated good convergent and discriminant validity, as well as good internal consistency and test–retest reliability (see Schmitter-Edgecombe et al., 2014 for complete psychometric information).

Performance-based measures

The Revised Observed Test of Daily Living (OTDL-R)

The OTDL-R (Diehl et al., 2005) is a performance-based test of everyday problem solving abilities. Participants are presented with everyday scenarios, questions to be answered about each scenario, and simulated materials (e.g., page from a phone book). Participants are asked to perform the steps necessary to answer the questions using the provided materials (e.g., balance a checkbook using a list of transactions). There are a total of nine tasks (three tasks each in the domains of medication use, telephone use, and financial management; each task includes multiple individual question items relating to the overall task) with 28 items total (maximum score = 28).

The Everyday Problems Test (EPT)

The EPT (Willis & Marsiske, 1993) is a paper-and-pencil performance-based measure of everyday cognitive competence. Participants complete problems based on tasks of everyday living using printed materials (e.g., calculating nutritional information from a food product label). For the purposes of this study, four of the seven total domains in this measure were assessed (shopping, transportation, household, and meal preparation), because the remaining three domains were assessed by the OTDL-R. Questions are presented in multiple-choice format, and each correctly selected answer earns one point, with a maximum possible score of 24.

Direct observation measures

Simple/isolated activities (“Eight Activities”)

Participants completed eight activities of daily living (Schmitter-Edgecombe & Parsey, 2014a; Schmitter-Edgecombe et al., 2011) that represented different types of daily activities in a campus apartment (e.g., sweeping and dusting, filling a medication dispenser, and cooking soup) in order to provide a direct observation measure of functional ability. Prior to beginning each activity, participants were given verbal instructions by test administrators about what to do for the activity. Participants were asked to return all materials to their original positions after each activity was completed. After receiving verbal instructions, participants carried out each activity on the main floor of the apartment whereas two test administrators independently observed via camera system, annotated, and coded participant actions from a separate room upstairs. Actions coded included critical and non-critical omissions and substitutions, irrelevant actions, and inefficient actions. Based on these coded actions, a rubric was used to obtain a score for each participant for each task, which ranged from one (task completed without any errors) to five (less than 50% of task completed). Scores for the eight activities are then summed to derive a total score for this measure (see Schmitter-Edgecombe et al., 2011 for complete scoring rules). Two independent raters coded each participant's performance. Discrepancies were discussed between raters (and a third party when necessary) when they arose, and the agreed upon coding of the error was added to the working document of errors for the particular task step. A prior study using this direct observation paradigm found that agreement between raters ranged from 95.45% to 97.97% (Schmitter-Edgecombe et al., 2011).

Multi-tasking/integrated activities (“Day Out Task”)

During the Day Out Task (DOT) (McAlister & Schmitter-Edgecombe, 2013; Schmitter-Edgecombe et al., 2012), which is a direct observation measure of functional ability, participants were asked to imagine that they were preparing for a day out, which included meeting a friend at a museum and later going to the friend's house to prepare and eat dinner. Participants were provided with a written list of eight subtasks to be completed to prepare for the day out. Participants were instructed by the experimenter to multi-task and interweave the subtasks in a way that felt natural and efficient. Each of the eight subtasks was coded according to completeness and efficiency, based on a scale from one (complete/efficient) to four (never attempted). An overall DOT accuracy score was then computed by summing the accuracy scores from each of the eight subtasks (see Schmitter-Edgecombe et al., 2012 for complete scoring rules). Similarly to the Eight Activities scoring procedures, two coders scored each participant's performance and discrepancies were discussed when they arose. A previous study examining interrater agreement in this measure found 96.92% agreement between raters for the subtask accuracy scores (Schmitter-Edgecombe et al., 2012).

Statistical Analysis

Pearson correlations between predictor and outcome variables were used to examine relationships among variables in the total sample. All measures were approximately normally distributed, with the exception of the informant-report questionnaire measure (skewness = 3.160, kurtosis = 11.914). This variable was not transformed for normality, in order to maintain the interpretability of the variable (Osborne, 2002) and because past research has suggested that multiple regression analyses with sufficiently large sample sizes, such as the current one, are robust to violations of normality (Casson & Farmer, 2014). Hierarchical linear multiple regression analyses were conducted to investigate the relationship between the memory measure predictors (i.e., content memory, prospective memory, and temporal order memory) and functional ability as measured by the six functional status measures (i.e., two questionnaires, two direct observation assessments, and two performance-based assessments).

Prior to running the analysis, participants who exhibited retrospective, rather than prospective, memory failure in the activity memory paradigm measure for prospective memory were excluded from the dataset (n = 7; 5 MCI). There was no significant difference in temporal order memory performance for the incidental (M = 5.67, SD = 3.81) and intentional (M = 5.57, SD = 3.80) iterations of the paradigm, t = 0.27, which were totaled. As expected, free recall for the completed activities was significantly greater in the intentional iteration (M = 6.40, SD = 1.10) compared to the incidental iteration (M = 5.70, SD = 1.47), t = 5.65, p < .001. These two measures were, however, correlated (r = 0.22, p < .005) and the study findings were similar when the incidental and intentional memory for activities measures were entered separately or combined in the regression analyses. Therefore, to remain consistent with the prospective memory and temporal order memory variables, the combined total score was used.

Six separate hierarchical regression analyses were run, one for each of the six functional ability outcome measures. In each analysis, age and education were entered into the first block to control for these two variables in the remainder of the analysis. The two measures of content memory (i.e., MAS prose memory score and memory for activities completed score) were entered into the second block. Prospective and temporal order memory measures were entered into the third block. Entering the content memory measures in block two and prospective and temporal order memory measures (i.e., non-content memory measures) in block three allowed for investigation of whether these other types of memory would account for a significant amount of variance over and above that accounted for by content memory measures.

Results

Tables 1 and 2 display demographic information and average scores on predictor and outcome variables, respectively. Table 3 displays Pearson correlations among all predictor and outcome variables. Generally, there were few significant correlations among predictor variables, suggesting that these aspects of memory are dissociable. The exception was that the memory for activities measure was significantly correlated with the temporal order memory measure (r = −.28, p < .01), although the magnitude of this relationship is relatively weak. This relationship is not surprising because these two measures assess processes regarding similar information (free-recall for the activities completed versus memory for the order of completion of those same activities). With the exception of the self-report questionnaire measure, which correlated only with the informant-report questionnaire measure, the measures of functional ability generally showed significant correlations with each other. Consistent with prior work (Farias et al., 2005; Rueda et al., 2015), when self- and informant-report data are concurrently compared, informant-report has been found to be more strongly associated with objective markers of disease progression and more strongly correlated with cognitive testing.

Table 3.

Pearson intercorrelations among predictor and outcome variables

	1	2	3	4	5	6	7	8	9	10	11	12
Demographics
1. Age	—
2. Years of education	−0.01	—
Memory
3. MAS prose mem.	−0.14	0.09	—
4. Activity content mem.	−0.38*	0.05	0.15	—
5. Prospective mem.	−0.11	0.07	0.18	0.18	—
6. Temporal order mem.a	0.33*	−0.06	−0.19	−0.28*	−0.10	—
Functional ability
7. IADL-C selfa	−0.04	0.01	−0.10	−0.20	−0.25*	0.18	—
8. IADL-C informanta	0.08	−0.13	−0.42*	−0.15	−0.36*	0.38*	0.25*	—
9. Eight Activitiesa	0.34*	−0.16	−0.22*	−0.22*	−0.31*	0.31*	0.18	0.24*	—
10. Day Out Taska	0.36*	−0.23*	−0.29*	−0.16	−0.33*	0.33*	0.08	0.30*	0.37*	—
11. EPT	−0.24*	0.37*	0.27*	0.20	0.38*	−0.23*	−0.11	−0.36*	−0.45*	−0.43*	—
12. OTDL-R	−0.38*	0.29*	0.26*	0.31*	0.28*	−0.29*	−0.11	−0.27*	−0.34*	−0.43*	0.40*	—

Note: MAS = Memory Assessment Scale; IADL-C = Instrumental Activities of Daily Living: Compensation Scale; EPT = Everyday Problems Test; OTDL-R = Revised Observed Test of Daily Living.

^aHigher scores represent poorer performance.

*p < .01.

Questionnaire Measures

Table 4 displays hierarchical regression data for all analyses. Analysis of the regression models for the two questionnaire measures revealed that content memory measures accounted for a significant amount of variance above and beyond age and education for both the self-report measure [ΔR² = 0.064, ΔF(2,156) = 5.370, p < .01] and the informant-report measure [ΔR² = 0.192, ΔF(2,125) = 15.054, p < .01]. Age and education did not account for a significant amount of variance in the first model of either analysis. As hypothesized, prospective and temporal order memory accounted for a significant amount of variance over and above demographics and content memory for both the self-report measure [ΔR² = 0.084, ΔF(2,154) = 7.654, p < .01] and the informant-report measure [ΔR² = 0.126, ΔF(2,123) = 11.561, p < .01]. After controlling for all other variables, age (t = −2.632, p < .01), memory for activities (content memory measure; t = −2.075, p < .05), prospective memory (t = −2.735, p < .01), and temporal order memory (t = 2.777, p < .01) emerged as significant predictors of the self-report functional ability measure. Age (t = −2.344, p < .05), prose memory (content memory measure; t = −3.793, p < .01), prospective memory (t = −3.683, p < .01), and temporal order memory (t = 3.154, p < .01) emerged as significant predictors for the informant-report functional ability measure. The R² values for each of these analyses showed that, overall, the set of predictors accounted for 15.1% total variance in the self-report questionnaire measure, and 32.9% total variance in the informant-report questionnaire measure. In summary, the content, temporal order and prospective memory measures were found be significant predictors for the questionnaire measures of functional ability.

Table 4.

Summary of hierarchical regression results for demographic and memory predictors of different measures of functional ability

Variables	Functional ability measures
	Questionnaire		Direct observation		Performance-based
	IADL-C selfa	IADL-C informanta	Eight Activitiesa	Day Out Taska	EPT	OTDL-R
Model 1
Age	−0.04	0.03	0.33**	0.38**	−0.17*	−0.40**
Education	−0.01	−0.10	−0.21**	−0.21*	0.36**	0.29**
Total R2	0.00	0.01	0.15	0.19	0.17	0.24
F for R2	0.16	0.70	14.43**	15.24**	15.01**	25.45**
Model 2
Age	−0.15	−0.09	0.26**	0.35**	−0.11	−0.31**
Education	0.01	−0.09	−0.20**	−0.21**	0.34**	0.27**
Activity content mem.	−0.24**	−0.13	−0.09	0.04	0.11	0.16*
MAS prose mem.	−0.10	−0.42**	−0.24**	−0.24**	0.23**	0.19**
Change in R2	0.06**	0.19**	0.07**	0.06*	0.06**	0.06**
Total R2	0.07	0.20	0.22	0.25	0.23	0.30
F for R2	2.77*	7.96**	11.14**	10.37**	11.04**	17.11**
Model 3
Age	−0.22**	−0.20*	0.21**	0.26**	−0.07	−0.27**
Education	0.03	−0.05	−0.18**	−0.20**	0.31**	0.25**
Activity content mem.	−0.17*	−0.06	−0.04	0.12	0.07	0.13
MAS prose mem.	−0.04	−0.30**	−0.18*	−0.21**	0.18*	0.16*
Prospective mem.	−0.21**	−0.29**	−0.22**	−0.33**	0.30**	0.17*
Temporal order mem.a	0.23**	0.26**	0.16*	0.20*	−0.07	−0.11
Change in R2	0.08**	0.13**	0.07**	0.13**	0.09**	0.04*
Total R2	0.15	0.33	0.28	0.38	0.32	0.34
F for R2	4.53**	10.05**	10.37**	12.75**	11.38**	13.24**

Note: Standardized beta coefficients presented for predictors. MAS = Memory Assessment Scale; IADL-C = Instrumental Activities of Daily Living: Compensation Scale; EPT = Everyday Problems Test; OTDL-R = Revised Observed Test of Daily Living.

^aHigher scores represent poorer performance.

*p < .05. **p < .01.

Direct Observation Measures

The content memory measures accounted for a significant amount of variance over and above age and education alone for both the Eight Activities isolated direct observation task score [ΔR² = 0.067, ΔF(2,160) = 6.810, p < .01] and the DOT integrated direct observation task score [ΔR² = 0.055, ΔF(2,126) = 4.624, p < .05]. As hypothesized, prospective and temporal order memory measures accounted for additional significant variance over and above the demographic and content memory measures for both the Eight Activities score [ΔR² = 0.065, ΔF(2,158) = 7.129, p < .01] and the DOT score [ΔR² = 0.134, ΔF(2,124) = 13.434, p < .01]. After controlling for all other variables, age (t = 2.711, p < .01), education (t = −2.723, p < .01), prose memory (content memory measure; t = −2.549, p < .05), prospective memory (t = −3.135, p < .01), and temporal order memory (t = 2.110, p < .05) emerged as significant predictors of the Eight Activities score. Age (t = 3.059, p < .01), education (t = −2.747, p < .01), prose memory (content memory measure; t = −2.820, p < .01), prospective memory (t = −4.577, p < .01), and temporal order memory (t = 2.483, p < .05) also emerged as significant predictors of the DOT score. Thus, content, temporal order and prospective memory were all significant predictors of the direct observation measures of functional ability. Overall, the predictors accounted for 28.3% of the variance in the Eight Activities score and 38.2% of the variance in the DOT score.

Performance-based Measures

Content memory measures accounted for a significant amount of variance over and above age and education alone for both the Everyday Problems Test [ΔR² = 0.063, ΔF(2,148) = 6.058, p < .01] and Revised Observed Test of Daily Living [ΔR² = 0.061, ΔF(2,158) = 6.898, p < .01]. Inconsistent with our hypothesis, the non-content memory measures (i.e., temporal order and prospective memory) accounted for significant variance over and above the demographic and content memory measures for both the EPT [ΔR² = 0.089, ΔF(2,146) = 9.525, p < .01] and the OTDL-R [ΔR² = 0.035, ΔF(2,156) = 4.140, p < .05]. For the EPT, education (t = 4.402, p < .01), prose memory (t = 2.602, p < .05), and prospective memory (t = 4.294, p < .01) emerged as significant predictors after controlling for all other variables. For the OTDL-R, age (t = −3.610, p < .01), education (t = 3.838, p < .01), prose memory (t = 2.280, p < .05), and prospective memory (t = 2.488, p < .05) all emerged as significant predictors after controlling for other variables. The set of predictors accounted for 31.9% of the variance in EPT score and 33.7% of the variance in OTDL-R score. While content and non-content memory accounted for a significant amount of the variance in performance-based tasks, only the content and prospective memory measures but not the temporal order memory measures were significant predictors.

An additional set of analyses was run where the predictors in blocks two and three were switched. That is, the prospective and temporal order memory measures were entered in block two and the content memory measures were entered in block three. This allowed for examination of whether content memory would still accounted for a significant amount of the variance after controlling for the non-content memory measures. The significance patterns of the R²-change values among each step of each analysis remained identical to the primary analysis (that is, models two and three always exhibited a significant R²-change value relative to the previous model), with the exception that content memory measures did not account for a significant amount of variance over and above prospective and temporal order memory in the IADL-C self-rated questionnaire analysis [ΔR²-change = 0.026, ΔF(2,154) = 2.315, p = .102]. Despite this, memory for activities, a content memory measure, did show up as a significant predictor in this analysis after controlling for all other variables.

Discussion

Prior research indicates that current proxy measures of functional status (e.g., questionnaires, performance-based measures, and direct observation) show discrepancies in their rating of functional abilities. In this study, we examined the relationships between multiple types of memory derived from an activity memory paradigm and different proxy measures of functional abilities. Consistent with our hypothesis, content memory was found to be a significant predictor for each of the six functional status measures, such that higher content memory ability was related to greater independent functional ability. Furthermore, content memory continued to explain significant variance even after controlling for demographics and temporal order and prospective memory. These findings suggest that variance related to content memory was captured by each of the proxy measures of functional status. This extends prior research by providing evidence for a relationship between content memory and everyday functioning across multiple proxy measures of functional status within the same study. The results are also consistent with prior work which has found that content memory is important for successful performance of IADLs in older adults with and without cognitive decline, both when examining individuals cross-sectionally and when aiming to predict future decline (Farias et al., 2009; Tuokko et al., 2005).

Of note, our measure of prose memory was more often a significant predictor than activity memory in the current study. To support functional independence in the real-world environment, both recalling the content of verbal information as well as activities completed is important. There has been debate in the literature about whether the content of activities is retained differently, that is less strategically and more automatically than verbal information, given the multi-modal nature of activity memories (Helstrup, 1986; Kausler & Lichty, 1988). The potentially greater effortful nature of the prose recall task may have partly contributed to the greater sensitivity of the prose memory measure, along with the fact that data from the incidental and intentional iterations of the activity memory paradigm were combined leading to poor internal consistency reliability for the activity content memory measure. A similar pattern of the prose memory measure being the more significant predictor was also evident when the incidental and intentional measures of activity content memory were entered separately in the regression analyses in place of the combined data.

The primary focus of this study was to examine for sources of discrepancy among functional status measures by focusing on types of memory outside of content memory, specifically prospective and temporal order memory that are important to successful everyday activity completion. Direct observation measures require a degree of self-initiation, multi-tasking, and contextual awareness to complete everyday tasks. Similarly, questionnaire measures aim to gather information about real-life activities where individuals may be required to sequence, multi-tasks and remember to carry out activities in the future. Consistent with expectations, the prospective and temporal order memory measures accounted for an additional significant 7%–13% of the variance in the direct observation and questionnaire functional status measures over and above demographics and content memory (which included memory for the content of verbal information and activities), supporting the role of non-content memory in functional ability. These findings strengthen prior research, which was conducted with questionnaire measures only, in suggesting that types of memory outside of content memory are important for supporting everyday functioning in older adults (e.g., Schmitter-Edgecombe et al., 2009; Woods et al., 2012).

Although it was hypothesized that temporal order and prospective memory would not account for significant variance in performance-based measures due to the isolated and structured nature of these laboratory tasks, prospective memory, but not temporal order memory, emerged as a significant predictor for both performance-based measures in this study. Presently, there has been no prior research to suggest that prospective memory would be important for completion of performance-based tasks specifically, although research has suggested that prospective memory is important for functional ability overall (Schmitter-Edgecombe et al., 2009; Woods et al., 2012). In the current study, individuals were not required to carry out delayed intentions for the performance-based tasks: the stimuli and the tasks or questions to be completed remained in view of participants throughout task completion. One possibility is that the nature of prospective memory measured by the event-based prospective memory task in this study (i.e., recognizing a visual stimulus, appropriately recalling relevant information about what the cue is and what action must be carried out) may be related to the processing of visual stimuli required for completion of the two performance-based tasks (i.e., Revised Observed Test of Daily Living and the Everyday Problems Test). For these tasks, individuals are presented with simulated materials, such as a nutrition label or utility bill, and are asked to complete relevant tasks pertaining to these materials. Perhaps the ability to bring to mind relevant information and knowledge about what these stimuli are generally used for in everyday life (e.g., seeing a utility bill and searching to locate the total charge and due date) is related to the stimulus recognition/intended action recall process that is employed in prospective memory. Although prospective memory and schematic information processing are not identical concepts, it is possible that skills in one domain would transfer over to improve performance in the other. By eliciting important relevant skills and information relating to the visual stimuli, overall performance may improve. Similarly, it is possible that other commonalities in the skills necessary to complete performance-based and event-based prospective memory tasks may be contributing to this finding. For example, the ability to maintain instructions throughout a task and follow them appropriately is necessary both for completion of a prospective memory task and for completion of the performance-based tasks. Notably, prospective memory was the strongest predictor out of all the predictor variables in each analysis (except the self-report questionnaire, for which temporal order memory had the largest beta weight), suggesting that prospective memory is important for successful everyday functioning.

The results also indicate that memory for temporal order may be an important cognitive ability for successful real-life activity completion in older adults, as temporal order memory was a significant predictor of both questionnaire and direct observation scores. While there is a paucity of research on the role of temporal order memory in everyday functioning, previous studies have suggested that temporal order memory declines with age (Allen, Morris, Stark, Fortin, & Stark, 2015; Schmitter-Edgecombe & Simpson, 2001). One prior study found that temporal order memory was a significant predictor of everyday activity completion, but it's importance was inconsistent depending on the specific activity domain being assessed (e.g., financial management, social functioning, and medication management; Schmitter-Edgecombe et al., 2009). Findings from this study are consistent with the findings from the study by Schmitter-Edgecombe and colleagues in suggesting that temporal order memory is important for successful completion of tasks in the everyday environment. The findings extend this prior work by suggesting that some types of functional status assessments may be more capable of capturing limitations associated with temporal order memory difficulties than other measures. Specifically, because temporal order memory was not a significant predictor of the performance-based measures in this study, it appears that current laboratory performance-based measures are more limited in their ability than questionnaires and direct observation measures to adequately capture sequencing abilities that are important in everyday activity completion. This may limit the ability of many current performance-based tests to measure functional capacity in the most ecologically valid way, as such tests may not capture cognitive processes like temporal order memory that play an important role in successful everyday activity completion. Future performance-based laboratory tests could be created to better capture everyday sequencing abilities.

This study also examined the role of demographics (i.e., chronological age and number of years of education). Consistent with previous literature that has suggested that functional status declines with age (e.g., Burton, Strauss, Hultsch, & Hunter, 2006; Schmitter-Edgecombe et al., 2011), age was a significant predictor for five of the six outcome measures in this study. Education emerged as a significant predictor for the direct observation and performance-based tasks, but not the self- and informant-report questionnaires. This finding is consistent with the psychometric properties of the IADL-C as items that comprise the IADL-C behaved similarly when education was dichotomized as some college versus none (Schmitter-Edgecombe et al., 2014). These results indicate that some proxy measures of functional status may be less susceptible to the influence of education than other measures and as such may have important implications for the selection of an appropriate measure of functional status depending on the person being assessed.

Although this study contributes to the literature by simultaneously examining multiple techniques for measuring functional ability in older adults and by examining lesser-studied constructs of memory, this research is limited in several ways. The highly educated, predominantly female nature of our sample limits the generalizability of the findings to other, more heterogeneous groups. Additionally, although the current study examined a spectrum of cognitive ability levels by including both healthy older adults and older adults with MCI, further research is needed to determine if the findings of this study would generalize to a more cognitively impaired population, or individuals with other neurological conditions. Although six different functional status measures were included, the direct observation measures in this study are experimental paradigms, and further research with additional functional status measures, as well as those that take place in entirely naturalistic environments, like the individual's own home instead of a simulated home environment, is warranted. Furthermore, the differential aspects of memory were measured by performance on an experimental memory task, and the internal consistency reliability of these measures was low. In addition, we used an event-based prospective memory task. Future research using well validated clinical tests that measures these memory constructs and employing time-based prospective memory tasks (remembering to perform a task at a particular point in time rather than in response to a particular event) is also warranted. Finally, although the pattern of significant predictors yielded results that were generally consistent with the hypotheses and with prior research, it is worth noting that the magnitude of these significant predictors was still relatively small (the largest beta weight, prospective memory in predicting the DOT direct observation measure, would equate to a prediction of only 11% of the total variance). However, this magnitude should be viewed within the context of several prior meta-analyses conducted with neurologic and MCI populations which suggest that cognition only accounts about 20%–25% of the total variance in functional status (McAlister, Schmitter-Edgecombe, & Lamb, 2016; Royall et al., 2007).

In the present study, the relationship between multiple types of memory and functional status measures in a community-dwelling older adult sample was evaluated to identify potential sources of discrepancy among functional status measures. The findings suggest that direct observation and questionnaire measures may be able to capture components of everyday functioning that require context and temporal sequencing abilities, such as multi-tasking, that many current laboratory performance-based measures of functional status may not capture as well. However, replication with well-validated measures of these memory constructs is needed. Given that there are unique benefits and drawbacks for each type of functional ability measure, future research should focus on developing a more specific “gold standard” for measuring functional ability in the most accurate, reliable, and efficient way.

Funding

Jenna Beaver, M.S., and Maureen Schmitter-Edgecombe, Ph.D., Department of Psychology, Washington State University, Pullman, Washington. This work was supported by grants from the Life Science Discovery Fund of Washington State and the National Institute of Biomedical Imaging and Bioengineering [Grant #R01 EB009675]. This project was completed in partial fulfillment of Jenna Beaver's master's degree in psychology at Washington State University.

Conflict of Interest

None declared.