|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 THE RELATIONSHIP BETWEEN BASIC C OGNITION, EVERYDAY COGNITION AND EVERYDAY FUNCTION: A LONGITUDINAL APPROACH By ANNA YAM A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORID A IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2010
2 2010 Anna Yam
3 To my Grandmothers
4 ACKNOWLEDGMENTS I thank my mentor, Dr. Michael Mars iske, whose guidance accounts for a signific ant proportion of variance in my academic functioning, and whose overall positive influence is too great to concretize. I also thank Shannon Sisco, for all her help and support, as well as Jennifer Rosado and Eile en Davis, for making me feel at home.
5 TABLE OF CONTENTS page ACKNOWLEDG MENTS .................................................................................................. 4LIST OF TABLES ............................................................................................................ 8LIST OF FIGURES .......................................................................................................... 9ABSTRACT ................................................................................................................... 10 CHA PTER 1 LITERATURE REVIEW .......................................................................................... 12Overvi ew ................................................................................................................. 12Older Adults Ever yday Function ............................................................................. 14Assessment of Older Adul ts Everyday Func tion .................................................... 15Antecedents of Everyd ay Functi oning .................................................................... 18Basic Cogni tion ....................................................................................................... 19Assessment of Ba sic Cogn ition .............................................................................. 20Everyday C ognition ................................................................................................. 20Assessments of Ever yday Cogn ition ...................................................................... 22Ecological Validity ................................................................................................... 23The Predictive Validity of M easures of Basi c Cognition .......................................... 24Cross-sectional Associations betw een Basic Cognition and Everyday Functi on ........................................................................................................ 25Longitudinal Relationships between Basic Cognition and Everyday Function .. 26The Predictive Validity of Meas ures of Everyd ay Cognition .................................... 27The Relationship between Basic Cogn ition and Everyd ay Cognition ............... 29Cross-sectional relationship betw een basic and everyday cognition ................ 29Longitudinal relationship between basic and everyd ay cognition ..................... 30The Relationship between Basic Cogniti on, Everyday Cognition and Everyday Functi on ............................................................................................................... 30Summary ................................................................................................................ 312 STATEMENT OF THE PROB LEM ......................................................................... 33Overvi ew ................................................................................................................. 33Mediation of Basic Cognition Variance in Everyday Function by Everyday Cognit ion ............................................................................................................. 34Aim 1 ................................................................................................................ 34Hypothesis 1 ..................................................................................................... 34Longitudinal Relationships between Basic Cognition and Everyday Function, and Mediation by Ever yday Cogn ition ................................................................. 35Aim 2 ................................................................................................................ 35Hypothesis 2 ..................................................................................................... 35
6 3 RESEARCH DESIGN AND ME THODS .................................................................. 36Overvi ew ................................................................................................................. 36The ACTIVE Study ................................................................................................. 36Recruitm ent ...................................................................................................... 37Eligib ility ........................................................................................................... 37Participants ............................................................................................................. 38Full ACTIVE Sample ........................................................................................ 38Present Study Sample ...................................................................................... 39Aggregate Longitudinal Retention Pattern of the Present Study Sample ......... 39Characterization of Attrition Effects .................................................................. 39Measur es ................................................................................................................ 40Basic Cognition Measures ................................................................................ 40Everyday Cognition Measures .......................................................................... 44Everyday Func tion Meas ure ............................................................................. 46Statistical Analyses ................................................................................................. 46Data Preparati on and Reduc tion ...................................................................... 47Criteria for Eval uation of Models ...................................................................... 48Analytical Framework for Aim 1: Cross-sectional Mediation ............................. 51Analytical Framework for Aim 2.1: C hange over Time in Study Constructs ...... 52Analytical Framework for Aim 2.1: Longi tudinal Mediation of Basic Cognitive Variance in Everyday Functi on by Everyday Cognition ................................. 534 RESULTS ............................................................................................................... 59Overvi ew ................................................................................................................. 59Aim 1: Baseline Mediation of Basic Cogni tion Variance in Everyday Function by Everyday C ognition ............................................................................................. 59Preliminary Analyses ........................................................................................ 59Nested Model Findings ..................................................................................... 59Aim 2.1: Change over Time in Study C onstructs .................................................... 62Aim 2.2: Longitudinal Mediation of Basic Cognitive Variance in Everyday Function by Ever yday C ognition .......................................................................... 635 DISCUSSI ON ......................................................................................................... 76Overview of Findings .............................................................................................. 76Limitations ............................................................................................................... 80Conclusi ons ............................................................................................................ 82Future Dir ections .................................................................................................... 84APPENDIX A EVERYDAY PROBLE MS TEST ............................................................................. 86B OBSERVED TASKS OF DAILY LI VING ................................................................. 87C TIMED INSTRUMENTAL ACTIVI TES OF DAIL Y LIVING ...................................... 88
7 D MINIMUM DATA SET ............................................................................................. 89LIST OF RE FERENCES ............................................................................................... 91BIOGRAPHICAL SKETCH .......................................................................................... 100
8 LIST OF TABLES Table page 3-1 Sociodemographic charac teristics of the present study sample and the full ACTIVE sa mple .................................................................................................. 553-2 Sociodemographic comparis on of the present study samp le to the rest of the ACTIVE sample at baseline ................................................................................ 553-3 Sample attrition, comparing partici pants assessed at Year 5 to those not assessed at Year 5 ............................................................................................. 563-4 Measures used ................................................................................................... 564-1 Bivariate correlations between measur es of basic and everyday cognition and everyday function ........................................................................................ 674-2 Factor loadings for the basic and everyday cogni tion factors ............................. 684-3 Baseline mediation ne sted model fi t indi ces ....................................................... 694-4 Intercorrelations between basic co gnition factors and everyday cognition factor and observed ev eryday f unction ............................................................... 694-5 Model fit indices and variance explained for models examining the effects of time in study constructs ...................................................................................... 704-6 Estimates of fixed and random effects for the fixed quadrat ic time model for each vari able ...................................................................................................... 714-7 Fit indices and variance explained for the longi tudinal mediation nested models ................................................................................................................ 724-8 Fixed and random estimates for the final longitudinal mediation model ............. 72
9 LIST OF FIGURES Figure page 1-1 Conceptual model for the present st udy, showing the hy pothesized predictive relationships between study constructs .............................................................. 323-1 ACTIVE study design. Note: booster tr aining occurred prior to the first and third follow-up assessment s. .............................................................................. 573-2 Aggregate longitudinal retention pattern of the current study sample. Note: the chart doesnt describe parti cipants entrances and ex its .............................. 584-2 Longitudinal trajectories for all study constructs. Note, negative values of basic speed and everyday function were plotted to facilitate visual comparis ons. ...................................................................................................... 75
10 Abstract of Thesis Pres ented to the Graduate School of the University of Florida in Partial Fulf illment of the Requirements for t he Degree of Master of Science THE RELATIONSHIP BETWEEN BASIC C OGNITION EVERYDAY COGNITION AND EVERYDAY FUNCTION: A LONGITUDINAL APPROACH By Anna Yam May 2010 Chair: Michael Marsiske Major: Psychology Elders ability to perform in strumental activities of da ily living (IADL), such as managing finances, telephone use, and meal prepar ation, is a key determinant of their ability to live independently. Traditional neuropsychological measures of basic cognition (e.g., memo ry, processing speed, and reasoning) are often only modest predictors of ADL. Several studies have argued that measures of everyday cognition, which simulate cognitive challenges enc ountered in everyday life, might be more predictive of IADL, and account for the effects of basic cognitive abilities in older adults everyday functioning. To date, most studies have focused on cross-sectional associations. The present study used the notreatment control group of the ACTIVE trial (N=698, 74% female, mean age = 74 years, mean education = 13 years), which had five occasions of measurement ( baseline, and follow-up assessm ents after 1-, 2-, 3-, and 5years). Isolating this group permitted us to us e ACTIVE as a large, 5-year longitudinal trial of cognitive change. Mixed effects l ongitudinal models revealed a quadratic decline trend for all constructs, with steeper decline fo r basic cognitive measures than everyday cognition or IADL functioning. Individual differences in memory, processing speed and
11 everyday cognition emerged as significant uni que predictors of IADL level. In addition, intra-individual changes in memory and everyday cognition were uniquely and significantly coupled with changes in IADLs over the 5-year period. Implications for the predictive validity of the asse ssment of older adults, and for longitudinal aging research will be discussed.
12 CHAPTER 1 LITERATURE REVIEW Overview Most of gerontological science is form ed around the question of what keeps older adults functioning well into advanc ed old age? How can we conceptualize, measure, and promote independence and cont inued self-care? Although effective functioning in daily life has many components (e.g., physica l, emotional, financial, social, spiritual) (Willis, 1996), the current proposal focuses par ticularly on cognitive contributions to everyday function in late life. In the past several decades, a growing body of literature has attempted to explicate how traditional and largely acontextual measures of cognition (measures designed for use in experimental, laboratory, and clinical applications) might be related to elders ab ility to maintain independence and well being in later life (Allaire & Marsiske, 1999; Corne lius & Caspi, 1987). One part of this line of scholarship has been the identific ation of a bridge construct, everyday cognition that may be seen as the mediator between more basic cognitive abilities and everyday functioning. That is, everyday cognition is viewed as the instantiation of basic abilities (e.g., information processing skills like memo ry, reasoning, speed, attention, working memory, etc.) in real-world contexts (Sc haie & Willis, 1986, Willis & Marsiske, 1991). In other words, everyday cognition encompasses tasks such as reading a nutrition label or understanding a medication regimen using co gnitive skills like attention and working memory. Considerable work has been done with the aim of validating these everyday approaches (see More, Palmer, Patterson, & Jeste, 2007, for a review). Validation would necessarily entail demonstr ating significant relationships with everyday function
13 and capturing at least as much variance in everyday function as basic cognitive approaches. If appropriately validated, everyday c ognition approaches would offer a more parsimonious and ecologically valid approach to the assessment and pr ediction of older adults everyday function, which in turn ma y provide a better locu s for interventions aimed at heading off functional decline. The conceptual model representing the hypothesized relationships between the constr ucts investigated in this study are illustrated in Figure 1-1. The present conceptual model is unidirectional. That is older adults engagement in everyday activities is likely to influence their performance on both basic and everyday cognitive tasks. While a number of studies su pport the idea that activity engagement serves as a protective factor against cognitive decline (Hultsch, Hertzog, Small, Dixon, 1999; Allaire & W illis, 2006), the empirical examination of the reciprocal interrelationships between study constructs is outside the scope of the present work, which focuses exclusively on t he contribution of cognitive precursors to self-reported everyday functioning. As illustrated in Figure 1-1, the purpose of this study was to investigate the relative contributions of basic and everyday cogniti on to predicting older adults everyday function. Embedded in this question was whether (a) everyday cognitive measures (which tend to brief, process-impure, and hi ghly face valid) would be as effective at predicting everyday function as basic labo ratory measures of cognition, and (b) everyday cognitive measures might be more cl osely related to everyday function than basic cognitive measures, because of their greater similarity to the target outcome (i.e., higher predictive/ecological validity). Superimposed on these questions is the issue of change over time. The current study asked not only what the nature of the association
14 between everyday cognition, basic cognition, and everyday function might be, but it also addressed how these constructs traveled toget her, i.e., whether they evinced coupled change over the 5 years of this longitudinal study. The following sections will be organized as follows: First, the concepts of everyday function, everyday cognition, and basic c ognition, as well as methods for their assessment will be defined and described. Second, the concept of ecol ogical validity as it pertains to the assessment of older adults everyday function will be discussed. Third, research examining the interrelationships among basic cognition, everyday cognition, and everyday function will be reviewed in both cross-sectional and longitudinal studies. This review will set the context for t he questions examined in this study. Older Adults Everyday Function Broadly def ined, everyday function encompasse s tasks of daily living that facilitate independence in this society (Willis, 1996). Katz, Ford, Moskowitz, Jackson, and Jafee (1963), as well as Lawton and Brody (1969) hav e distilled older adults daily activities into discrete domains of everyday functi oning. Specifically, Katz and colleagues identified a basic set of activities of daily living (ADL), such as bathing, toileting, dressing, and eating, which are related to se lf-care. On the other hand, Lawton & Brody (1969) described a set of activities such as food preparation, medication use, financial management, and transportation, which are relat ed to the management of ones affairs. These are referred to as instrumental activities of daily living (IADL). The percentage of people 65 years and older increased 12% since 1990 to 2000, and among the older population, those 85 years and over showed the highest percentage of increase (~37%) (Hetzel & Smith, 2000). Given that as a group, individuals over 65 years amount to 12.4% of the total U.S populati on (Hetzel & Smith,
15 2000), and those 85 and over constitute the gr oup most at risk for health and functional impairments (Campion, 1994), older adults functional independence is a concern not only for these individuals and their families, but also society as a whole (Willis, 1996). Specifically, impairment in older adults ever yday activities has been linked to reduced psychological well-being (Willis, 1991; Lawton, 1987), greater health care utilization, increased rates of institutionalization (Wo linsky, et al., 1983; Wolinsky, Callahan, Fitzgerald, & Johnson, 1993; Miller & Weisse rt, 2000) and higher mortality (Wolinsky et al., 1993; Ferrucci, et al., 1991; Miller & We issert, 2000). Furthermo re, loss of IADL functional competence has been shown to be predictive of dementia (Prs et al., 2008; Barberger-Gateau, Fabrigoule, Rouch, Letenneur, & Dartigues 1999). Given the numerous negative consequences of functional impairment, clinical and empirical assessments of older adults everyday functioning aim to identify individuals who already have difficulty performing these everyday tasks, and those who might be at risk for functional decline. While the former hel ps clinicians identify individuals who are already in need of assistance, the latter leaves more room for intervention designed to prevent functional decline. In either case, thorough and accura te assessment may be the first step to improvements in the daily lives of older individuals as well as the reduction of the broader negative soci o-economical impact of aging. Assessment of Older Adul ts E veryday Function In general, researchers and clinicians are unable to obtain naturalistic samples of what older adults actually do in their ever yday lives. First, older adults natural task performance, even in the context of their ho me environment, is likely to be influenced by the presence of an observer, and second, n aturalistic assessments are costly, time consuming, and are historically not carried out by professionals outside Occupational
16 Therapy. Consequently, most functional assessments pose the question Can the individual perform an activity? and as such, they likely capture the everyday functional competence of older adults, or their pot ential to perform ADL and IADL tasks (Willis, 1996; Salthouse, 1990), as opposed to what they do. There are three main types of approac hes to the assessment older adults everyday function. First, while there is no gold standard meas ure, one widely utilized approach involves self-report measures which typically consist of questions asking older individuals to rate their level of perce ived functional independenc e, and/or level of experienced difficulty perfo rming tasks in a number of ADL and IADL domains. While these questionnaire-based assessments are subj ect to bias due to their dependence on the subjective impressions of individuals they continue to be widely utilized in both research and clinical practice (Diehl, 1998). A second approach to the assessment of the everyday function of older adults involves proxy (i.e., spouse, family mem bers, and caregivers) reports. Two measures commonly utilized to obtain proxy-ratings of older adults everyday functioning include the Clinical Dementia Rating Scale (CDR; Morris, 1993) and the Blessed Dementia Rating Scale (DS; Blessed, Tomlinson, & Roth, 1968). Proxy-based assessments are generally questionnaires that are similar in content to th e self-assessment measures. While this approach benefits from being more objective, it is not without flaws and biases. Because proxies vary in their actual contact with the older adult, their capacity to provide valid information, they have been shown to misrepresent the functional competence of older adults (Roy all et al., 2007; Diehl, 1998). A third approach to the assessment of older adults everyday function involves the observation of performance of older adults on everyday tasks. Because the goal of this
17 type of approach is to obtain objective rati ngs of performance, these assessments are designed for administration by clinicians, such as Occupational Therapists. While such approaches have the str ength of being based on observations of performance, they are costly and time consuming to administer (Willis, 1996), and consequently utilized lessfrequently. Furthermore, given their structured, non-individualized, clinical nature, they necessarily only gather a small sample of a persons behavior. Thus, because of the brief and sheltered nature of the testing sessions, areas of strength or weakness may not be adequately assessed (Sbordone, 1997). The issue of nomenclature is an impor tant one, with respect to whether assessments indeed capture a persons everyday functioning (to the extent a given approach permits), or whether they are in stead capturing the c ognitive components of everyday functioning (i.e. everyday cognition to be discussed in more detail below). Because studies frequently dont draw a clear distinction between these two everyday concepts, for the purposes of this manuscript, everyday function will be concretized to represent the empirical outcome s (scores) of selfa nd/or proxyreported ADL/IADL function, and performancebased assessments when actual task performance was observed and coded by a clin ician, such as an Occupational Therapist. Such assessments, which are most frequently undertaken with patients recovering from injuries, ar e typically done in the home or a simulated setting. Irrespective of the approach utilized, the assessments of everyday function described above are better suited for the detec tion of current functional impairment, and less well-suited to detect individuals at risk for functional decline (however see the review by Miller & Weissert, 2000, showing t hat across multiple longitudinal studies, current impairments in IADL predict more fu nctional impairment over time). One reason
18 for this has to do with how late in the lifespan functional decline typically occurs. That is, studies show that decline in IADL precedes ADL decline, and usually begins after age 80 in community samples (Fillenbaum, 1985). Furthermore, assessments of older adults everyday functioning provide little information with regard to the causes of impairment or decline (Willis, 1991). For these reasons, researchers have focused their efforts on understanding the physical and cognitive proce sses that might antecede older adults everyday functioning. Antecedents of Everyday Functioning The antecedents of older adults everyday functioning include phy sical health, cognitive functioning, & psychological factors (Baltes, Mayr, Borchelt, Maas, & Wilms 1993). Other factors that influence and are in fluenced by everyday functioning are the social/interpersonal, community, & financial aspects of older adults lives (Willis, 1996). With respect to the effects of physica l health, Diehl (199 8), reviewing prior research, reported that older adults everyday function is positively related to their general physical heath and sensory functioning, with relationships between ratings of physical health and ratings of everyday func tion ranging from .30 to .54. For example, Marsiske, Klumb, and Baltes (1997) reporte d latent-level correlations of .38 and .47 between older adults everyday function and hearing and vision respectively. Overall, data from studies of the association betwe en health/sensory functioning and everyday function suggest that deterioration of health in creases the likelihood that assistance with activities of daily living becomes necessary. With respect to psychological factor s, researchers have looked at deficit awareness, mood (Caron, 1996; Barberger-G ateau et al., 1992), control and selfefficacy (e.g. Baltes & Baltes, 1986, Ba ltes et al., 1990; Duffy & MacDonald, 1990).
19 These studies have found that psychological factors influence the way older adults perceive their everyday abilities, which in turn influences their responses on questionnaires assessing their everyday func tion. For example, Caron (1996) reported that greater awareness of impai rment was highly correlated ( r =0.69) with functional scores. Baltes and colleagues (1990) found a significant positive relationship between self-efficacy beliefs and perceived level of everyday function. Finally, research has shown that subjective depression contributes to functional impairment independently of cognitive impairment (Bar berger-Gateau et al., 1992). With respect to cognitive function, ther e are two main approaches to quantifying older adults cognitive fitness: measures of basic cogniti on and measures of everyday cognition. Before reviewing these two approaches in more detail, it is important to note that, as detailed above, cognition is one of a multitude of factors that influence older adults everyday functioning. Therefore, its explanatory power is confined by its inability to capture the variance that might be attr ibuted to other relev ant human factors. Basic Cognition One widely -utilized approach to the assessment of older adults cognitive functioning involves the assessment of basic cognitive abilities such as visual and verbal shortand longterm memory, speed of information processing, visual attention, verbal fluency, and executive abilities or reasoning. These aspects of cognition have been widely and extensively studied at the behavioral level, and researchers have created behavioral measures that, to some ex tent, isolate and quantify these skills. With increasing accessibility of electroencephalo graphy and neuroimaging techniques, there is also a growing knowledge base with respect to the neural correlates of these abilities (e.g. the involvement of the hippocampus in memory). Importantly, these basic abilities
20 appear to underlie other, higher order proce sses such as problem solving, decision making, and creativity, the output of which is reflected in everyday functioning (Willis & Schaie, 1993, Willis & Marsiske, 1991). Assessment of Basic Cognition To quantify basic cognitive abi lities, researchers and clinicians utiliz e measures that were developed to assess intelligenc e, academic performance or neurological impairment. The measures used for these purposes are structur ed, paper and pencil tasks that elicit a maximum level of performance, often under timed conditions (Salthouse, 1990). Tests of intelligence (e.g. IQ tests) have a long history of being utilized to predict academic and vocational performance, while neuropsychological tests were created and historically utilized to detec t neurologic insult (e.g. brain lesions) and to quantify the resulting cognitive impairment. There is a methodological drawback to the utilization of such measures in the assessment of unimpaired older adults. These measures were not designed wit h the intention of assessing everyday functioning, and consequently lack the sensitivity to do so (Spooner & Pachana, 2006). When administered cross-sectionally, these measur es almost invariably show that younger adults outperform older adults, which contri butes little to our understanding of the everyday functioning of older adults. As Salthouse (1990) pointed out, assessments of older adults cognitive fitness need to bridge the gap between these age-related discrepancies in basic cognitive abilities, and the often successful functioning of older adults in everyday life (Salthouse, 1990). Everyday Cognition Everyday cognitio n has been conceptualized as a higher order set of abilities, representing the cognitive components of everyday functioning (Willis & Marsiske, 1991;
21 Willis & Schaie, 1993). This type of cogni tion is thought to be goal-directed and involving the manipulation of dat a or objects in everyday life in order to attain selfand home-maintenance goals (Marsiske & Marg rett, 2006). As such, these cognitive processes are contextualized, in that they resemble fa miliar, relevant, everyday challenges that rely on the individuals ex perience for an effective solution (Marsiske & Margrett, 2006). What makes them higher order is that in addition their reliance on more basic cognitive abilities like memory reasoning, and speed of processing, they also make use of accumulated experience, which represents domain specific knowledge (e.g. the persons prior experience of having prepared a meal) (Marsiske & Margrett, 2006). Thus, everyday cognitive abilities are viewed as compiled skills (Salthouse, 1990). Because everyday cognitive abilities dr aw upon this compiled, domain specific knowledge, some researchers have speculated that these abilities might be preserved against the age-related decline observed in the more basic cognitive abilities (Salthouse, 1990). In a meta-analysis of age differences in everyday cognition, Thornton and Dumke (2005) concluded that thei r results do not support theories of the preservation of these abilities. However most studies reviewed by these authors were cross-sectional, which confounds cohort and ma turational factors (Schaie, 1965) and has been shown in statistical simulation st udies to be misleading with respect to conclusions about change over time, particularly when results from correlations are used to draw these conclusions (Kraemer, Yesavage, Taylor, & Kupfer, 2000). For example, in their longitudinal study, Willis, Jay, Diehl, and Marsiske (1992) showed that 57% of their sample did not show a reliabl e decline in these abilities over a 7-year
22 period. Therefore, the question of whether these abilities are, at least to an extent, buffered against age-rela ted decline is still open. The terminology utilized in reference to this type of cognition has been variable, encompassing terms such as practical problem solving (Denney, 1989), everyday problem solving (Marsiske & Margrett, 2006, Thornton & Dumke, 2005), and practical intelligence (Sternberg & Wagner, 1986). Th is preponderance of nomenclature stems in part from the variable theoret ical orientations of the res earchers in the field (Marsiske & Margrett, 2006). Another distinction exis ts between types of everyday cognitive assessment approaches. That is, everyday probl ems might be either illstructured, or well-structured, with the former being more open ended-type problems, and the latter being problems that must be approached more systematically and having one target correct answer (Allaire & Marsiske, 2002). The present study focuses exclusively on the well-structured approach to the assessment of everyday cognition, which is grounded in cognitive and psychometric traditions, and employs the term everyday cognition (Poon, Rubin, & Wilson, 1989) to underscore this focus. Assessments of Everyday Cognition A number of measures have been developed to assess the everyday cognition of older adult s. Marsiske and Margrett (2006) conducted a focused review of multiple measures of everyday cognition, in which they summarized the theoretical underpinnings, structural properties, and availa ble evidence for valid ity, of multiple instruments designed to assess older adults performance of ever yday-type tasks. According to this review, well-structured ta sks share a substantial portion of their variance with basic cognitive tasks (50-80%), with provides support for the notion that well-structured assessments are a bridge between basic abilities and everyday
23 outcomes. Furthermore, while multiple ever yday instruments assess a single everyday domain, there is evidence to support the i dea that everyday cognition is better viewed as a multifaceted construct (e.g. Marsiske & Willis, 1995). Therefore, the present study is specifically focused on the utility of multi-domain, well-structured measures of everyday cognition in predicting everyday function. As such, these measures resemble the tasks that older adults might do in their everyday life, such as food preparati on, medication use, financial management, and transportation, where adults are given real-w orld stimuli and asked to solve novel everyday problems that have one correct an swer. These measures are structured, laboratory-based tests that, much lik e neuropsychological and psychometric assessments, elicit a maximum level of performance, however on tasks that are familiar and relevant to everyday life. Ecological Validity The question of relevance is central to the concept of ecological v alidity. Heaton and Pendleton (1981), foreseeing the increased use of neuroim aging techniques for the assessments of brain injury, in many wa ys supplanting neuropsychological measures, called for more research into the ecological validity of these measures as predictors of everyday function. Ecological validity is the functional and predictive relationship between the patients performance on a set of neuropsychological tests and the patients behavior in a variety of real world settings (Sbordone, 1996, p. 16). Two approaches to establishing ecological validity have emerged: verisimilitude and veridicality (Frazen & Wilhelm, 1996). Verisimilitude refers to the similarity between the task demands of the test and the demands of the everyday environment. Establis hing verisimilitude requires tests that are
24 comprised of everyday cognitive tasks, an approach taken by researchers who have developed measures of everyday cogni tion (Spooner & Pachana, 2006; Chaytor & Schmitter-Edgecombe, 2003). Therefore, such tests differ from traditional neuropsychological measures in their intend ed focus -identifying individuals with functional limitations -ra ther than diagnosing and describi ng the etiology of brain dysfunction (Chaytor & Sc hmitter-Edgecombe, 2003). Veridicality on the other hand, refers to the extent to which results of an assessment are related to scores on other meas ures that predict everyday functioning (Spooner & Pachana, 2006; Chaytor & Sc hmitter-Edgecombe, 2003). In this formulation, the veridicality of basic and every day cognitive assessments, particularly as it relates to change in function over time (i.e. predictive valid ity), is the central focus of the present study, and prior work in this domain is reviewed below. The Predictive Validity of Measures of Basic Cognition Research studies investigating the re lations hip between basic cognition and everyday function have attempted to answer the questions: what is the strength of the cross-sectional and longitudi nal association between basi c cognitive measures and measures of everyday function and what specific aspects of cognition might be related to everyday function? The following sections ex amine this body of work to date. First, the large body of cross-sectional work, suited to the identification of current functional limitations, and less well-suited to address pr edictive questions is examined. Second, longitudinal studies of the association betwe en basic abilities and everyday function, better-suited to address questions of predict ive utility/veridicality, are reviewed.
25 Cross-sectional Associations betw een Basic Cognition and Everyday Function While this review is not exhaustive, it nev ertheless captures the effects as they pertain to strengths of associ ation, as well as specific cognitive domains that appear to be salient predictors of everyday function. Of the studies exam ined, three examined clinical populations (e.g. Farmer & Eakman, 1995; McCue, Rogers, & Goldstein, 1990; Baird, Podell, Lovell, & McGinty, 2001) and of these, two ut ilized performance-based assessments of everyday function (e.g. Farmer & Eakman, 1995; McCue, Rogers, & Goldstein, 1990). The rest examined healthy ol der adults and utilized selfor informantreport measures of everyday function, parti cularly IADL. Despite some methodological differences, and different measures utilized, across studies memory and executive function emerged as significant cross-sectional predictors of function (e.g. Royall, Palmer, Chiodo, & Polk, 2005a; Farmer & Eakman, 1995; McCue et al., 1990; Jefferson, Paul, Ozonoff, & Cohen, 2006; T an, Hultsch, & Strauss 2009). Processing speed also emerged as a significant predictor in a subset of the studies (e.g. Tan et al., 2009; Farmer & Eakman, 1995), however this re lationship was smaller, and appeared to be associated with timeliness of IADL perfo rmance (Farmer & Eakman, 1995). Bivariate associations between specific basic abilities and function across studies were mostly in the moderate range (0.48 to 0.61). With respect to global cogni tive functioning and reviews of the literature in this domain, findings from large representative samples of older adults have shown that the relationship between everyday function and gener al cognition scores is in the order of .50-.60 (Fillenbaum, 1985), which appear to be at least as high and often higher than relationships with specific basic abilities. In a review of the cognitive correlates of functional status, Royall et al. (2007) concluded that the variance that can be
26 specifically attributed to cognition is modest and that global measures are surprisingly strong correlates of functional status. In summary, cross sectional studies ty pically show moderate relationships between basic cognitive measures and assessment s of everyday function. With respect to specific abilities, memory, executive f unction, and to a lesser extent processing speed, appear to be the best and most consis tent predictors of current levels of everyday function. Longitudinal Relationships between B asic Cognition and Ever yday Function Longitudinal assessments have the advantage of being able to capture trajectories of change and concurrent relationships between c onstructs of interest. With respect to the longitudinal trajectory of functional decline, in a stud y examining the hierarchy of functional loss (i.e. which everyday skills became impaired first) associated with decline in basic cognitive abilities in older adults Njegovan, Hing, Mitc hell, and Molnar (2001) reported that dependency in IADLs tended to o ccur at higher cognitive scores compared with ADLs. These authors also reported that a greater number of persons in their sample became dependent in at least one IA DL compared to ADLs over a 4 year period. This is consistent with prior findi ngs that IADLs evince earlier age-related decline than ADLs (Willis, 1996). Most of the studies reviewed examined longitudinal basic cognition and everyday function in healthy older adults, however some studies also included groups with mild cognitive impairment (MCI) and/or dementia (e.g. Tomaszewski Farias et al., 2008; Barberger-Gateau et al., 1999) and one study examined a geriatric sample with cardiovascular disease (Cahn-Weiner, Mall owy, Boyle, Marran & Salloway, 2007). Nearly all longitudinal studies employed self-reported ADL and/or IADL ratings as
27 measures of longitudinal everyday functi oning, with two studies employing informant reported IADL (Cahn-Weiner et al., 2007, Tomaszewski Farias et al., 2008). Across studies, changes in IADL were significantly related to changes in memory and executive function. Results of random effects regr essions revealed a -0.69 correlation between changes in memory and IADL, and -0.72 bet ween changes in executive function and IADL (Tomaszewski Farias et al., 2008). In their clinical sample, Cahn-Weiner et al. (2007) found that only executive function predicted changes in IADL. Barberger-Gateau et al. (1999) found that speed of processing wa s a significant longitudinal predictor of IADL (odds ratios 0.52 to 0. 74). In their studies using gr owth curve analyses, Royall and colleagues (2004, 2005a, 2005b) found a moderate negative relationship between change in IDAL and executive function (-0.57) and that executive function mediated the effect of memory on IADL function. In summary, there is evidence for signifi cant, moderate to large relationships between changes in basic cognitive abilities, particularly executive function, memory and processing speed, and changes in perceived levels of everyday functioning in healthy older adults as well as in pre-clinical and clinical geriatric samples. Executive functioning in particular emerges as a salient longitudinal predi ctor of everyday function. The Predictive Validity of Measu res of Ever yday Cognition Given the strict definition of everyday function posited in this manuscript, to date relatively few studies have examined the predictive validity (relationship to everyday function) of measures of everyday cognition. That is, most studies that claim to study everyday function have utilized meas ures of everyday cognition as outcome variables, using these measures as a pro xy for everyday functioning without adequate evidence of their veridicality. In their review of measure of everyday cognition, Moore,
28 Palmer, Patterson, & Jeste (2007) similarl y note that information on the predictive validity of these measures was largely unavailable. Of the studies that exam ined the relationship between everyday cognition and everyday function, one study utilized a clinical geriatric sample of individuals diagnosed with Alzheimers Disease (AD; e.g. Willis et al., 1998), while the remaining studies utilized samples of healthy older adults (e.g Allaire & Marsiske, 2002; Diehl, Willis, & Schaie,1995) or mixed samples (e.g. Tan et al., 2009). While ther e was heterogeneity in the everyday cognitive measures used, in most studies, selfand caregiverrated ADL/IADL was the selected measure of ever yday function. Results of these crosssectional analyses revealed significant moderate associations between function and everyday cognition in AD patients ( r =0.36; e.g. Willis et al., 1998), higher associations in healthy older adults ( r =0.50; e.g. Diehl et al., 1995) and still higher associations at the latent level ( r = -0.69; e.g. Allaire & Marsi ske, 2002). On the other hand, Tan and colleagues (2009) correlations between ev eryday function and everyday cognition ranged from -0.037 (n.s) to 0.43, suggesting some variability in outcomes. However, the results from Tan et al. (2009) should be inte rpreted with caution as the extent to with their measure of everyday function is refl ective of traditional ADL/IADL domains is unclear. In summary, it appears that in healthy older adults, there is evidence for a moderate to high association between every day cognition and everyday function, thus supporting the predictive validity of these measures. However, little is known about the strength of this relationship over time.
29 The Relationship between Basic C ognition and Ever yday Cognition If everyday cognition is to be conceptualized as the instantiation of basic abilities in everyday context, then it would be import ant to first examine the strength of the relationship between basic cognitive abilities and everyday cognition. As mentioned above, there has been a considerable amou nt of work done in this domain, as measures of everyday cogni tion have been utilized as proxy measures representing everyday function. The vast majority of t he studies reviewed investigated samples of healthy older adults. At the sa me time, all but one study (i.e. Willis, Jay, Diehl, & Marsiske, 1992) in this domain have been at the cross-sectional level and this body of work has been very heterogeneous with respec t to the everyday and basic measures utilized, as well as with respect to the types of basic abilities investigated (Marsiske & Margrett, 2006). Cross-sectional relationship between basic and everyd ay cognition Nevertheless, one relatively consistent fi nding from this literat ure is that of a moderate to large bivariate relationship between measures of basic and everyday cognition overall, with correlations from 0. 31 to 0.86 (e.g. Thor nton, Deria, Gelb, Shapiro, Hill, 2007; Diehl et al., 1995; Burton, Strauss, Hultsch, H unter, 2006; Allaire & Marsiske, 1999; Wood et al., 2005; Weather bee & Allaire, 2008). Results from regression analyses suggest that with demogr aphic variables accounted for, basic cognitive variables together ex plained 32% to 55% of the variance in everyday cognition (e.g. Thornton et al., 2007; Wood et al., 2005; Mitchell & Miller, 2008). As with studies investigating the relationship between bas ic cognition and everyday function, executive/reasoning abilitie s and memory, and to a lesser extent processing speed emerged as the most consist ent predictors of everyday cognitive abilities, however
30 these abilities were also the most consistent ly studied. For example, Burton et al. (2006) reported that after controlling for demographic variables, executive abilities contributed 11% unique explanatory variance to the predic tion of everyday cognition, followed by 3% unique variance contributed by speed. Wood et al. (2005) found that non-speed basic abilities uniquely explai ned 17% of the variance in timed IADL outcome, with 14% unique explained by speeded basic cognition. Longitudinal relationship between basic and everyday cognition Willis et al. (1992) examined the predict ive relationship between basic measures of fluid and crystallized intelligence, speed, and memory and a measure of everyday IADL cognition Educational Testing Service s Test of Basic Skills (ETS Basic Skills; ETS, 1977). Using data collected from two occasions of measurement, the authors reported that baseline level of fluid intelligence was predictiv e of later ETS Basic Skills performance explaining 52% of the variance, of the 70% tota l variance accounted for by all basic abilities and demographics. In summary, there does appear to be eviden ce for a significant moderate to high cross-sectional association between basic co gnition and everyday cognition in healthy older adults, with executive/reasoning/fluid in telligence emerging as the most salient predictor. Longitudinal results from W illis et al. (1992) appear to support these conclusions, further suggesting that basic fl uid intelligence can be predictive of later everyday cognitive abilities. The Relationship between Basic Cognition, Everyday Cognition and Everyday Function Willis et al. (1998) exa mined all three constructs in a sample of individuals with AD. These authors reported that the everyday cognitive tasks accounted for significant
31 additional variance in everyday function bey ond that accounted for by global cognitive measures. These results were supported by Allaire and Marsiske (2002), who also showed that everyday cognition accounted for the basic cognitive variance in everyday function, and significant uni que variance beyond that. Summary Prior work shows that both basic and everyday cognition have moderate crosssectional relationships to everyday functi on, supporting the use of these measures to assess individuals who already have some f unctional impairment. At the cross-sectional level, everyday cognition has been shown to mediate the basic cognitive variance in everyday function, and to explain additional va riance beyond that. This finding provides empirical support for the notion that basic abili ties are a part of, or involved in everyday cognition, however findings of concurrent change in these abilities would provide considerably stronger support for this notion. Of the basic abilities, memory, executive/reasoning skills, and speed of processi ng appear to be consistently related to everyday function both cross-sectionally and over time. On the other hand, little is known about whether changes in everyday cognitive abilit ies are predictive of changes in everyday function. The examination of the longitudinal relationships between basic and everyday cognition and everyday functi on would permit more informed conclusions regarding the unique predictive utility of basic ability and everyday cognition measures, facilitating a better conceptualization of t he parsimony and validity of the everyday cognition approach (as operationalized), to the assessment of everyday function.
32 Figure 1-1. Conceptual m odel for the present study, showing the hypothesized predictive relationships between study constructs
33 CHAPTER 2 STATEMENT OF THE PROBLEM Overview A large body of research lends support to the predictive relationship between basic cognition (e.g., cognitive abi lities such as memory, r easoning, and processing speed) and everyday function (e.g., older adults se lf-reported levels of difficulty and/or independence in performing activities of daily living), between bas ic cognition and everyday cognition (e.g., cognitive tasks that simulate activities of daily living), and between everyday cognition and everyday function. Furthermore, there is some evidence that everyday cognition mediates the basic ability variance in everyday function (Allaire & Marsiske, 2002; Willis et al., 1998). That is, age-related changes in basic abilities are hypothesized to influence performance on tasks assessing contextualized cognition, which it turn cont ributes to changes in the ability to perform critical household and self-maintenance tasks. In previous research these relationships were examined predominantly at the cross-sectional level. Longitudinally, most research has examined only the direct relationshi p between basic cognition and everyday function, suggesting that baseline levels of basic cognition are predictive of longitudinal change in everyday function (Willis et al., 1992) Due to their face validity and reliance on contextualized, domain specific knowledge, measures of everyday cognition have been conceptualized as more direct measures of the cognitive processes involved in the successful execution of the tasks that comprise everyday functioning. Given the research findings that both basic and everyday cognition predict everyday function, the present study investigated w hether everyday cognition vari ance might account for the basic ability variance in everyday function. The present analyses utilized longitudinal
34 data collected at five occasions of measurem ent, at baseline, annually 1-3 years postbaseline and at 5 years postbaseline, from a large sample of healthy, communitydwelling older adults age 65 years and over. There were two aims of the study. Mediation of Basic Cogni tion Variance in Ever yday Function by Everyday Cognition Aim 1 This aim addressed the extent to which ev eryday cognition measures might constitute the cognitive component of everyday function, as measured in this study. The purpose of the first aim was to replicate and extend prior cross-sectional mediation findings (e.g. Allaire & Ma rsiske, 2002; Willis et al ., 1998), and to assess baseline mediation in the present sample with the current measures, as a precursor to the assessment of mediation at the longitudinal le vel in Aim 2. Specifically, the aim was to assess whether the relationship between basic cognitive abilities (memory, reasoning, speed) and everyday function (self ratings of IADL) is mediated by individual differences in everyday cognition (Everyday Problems Te st, Observed Tasks of Daily Living, Timed IADL) at the baseline o ccasion of assessment. Hypothesis 1 Everyday cognition measures will show significant direct associations with everyday function, and they will exp l ain all or most of the basic memory/reasoning/speed related variance in ev eryday function. These findings would support the conceptualization of everyday cognition as a more proximal measure of everyday function.
35 Longitudinal Relationships between Basi c Cognition and Ever yday Function, and Mediation by Everyday Cognition Aim 2 The purpose of the second aim was to ex amine the predictive interrelationships between longitudinal changes in basic cogn ition, everyday cognition and everyday function, extending the analyses from Aim 1 to longitudinal data. Thus, the second aim was to examine if (a) longitudinal changes in everyday function are associated with concurrent time-varying changes in basic cogn ition, (b) changes in everyday cognition will serve as time-varying mediators of the basic cognition/everyday function relationship. Embedded in this aim is the inve stigation of the trajec tory of longitudinal change in all of our constructs (basic abi lities, everyday cognition and everyday function). Hypothesis 2 As basic cognition changes, (a) everyday function is expected to change in the same direction as basic cognition. With r egard to mediation, (b) changes in basic cognition are expected to be associated wit h changes in everyday cognition, and changes in everyday cognition, in turn to be associated with changes in everyday function. Based on prior research, no residual relationship between basic cognition and everyday function is expected. Furthermore, it is expected that (c) the unique path between everyday cognition and everyday func tion will explain more variance in everyday function than the unique path between basic cognition and everyday function (from (a) above). If supported, these findings would provide strong evidence to support the conceptualization of every day cognition as the instantiati on of basic abilities in tasks that better predict older adults everyday functioning.
36 CHAPTER 3 RESEARCH DESIGN AND METHODS Overview The present study represent s cross-sectional and longitudinal analyses of data from control participants enrolled in the Ad vanced Cognitive Training for Independent and Vital Elderly (ACTIVE) study. The followi ng sections describe the ACTIVE study, outlining procedure, participant recruitment, and eligibility. Ne xt the present sample, and measures used are described. The ACTIVE Study ACTIVE is a longitudinal, randomized c ontrolled single-blind clinical trial designed to determine whether the effects of cognitive training would transfer to everyday functioning, as assessed by laboratory measures. The study has a 4-group design, which includes 3 treatment groups and one no-contact control group. Participants in the treatment groups received cognitive training in memory or reasoning or speed of information processing, all of wh ich have previously been shown to improve cognitive abilities. Assessment of cognitive skills was conducted at baseline, 0.23 years following the training (post-test ), and annually at 1, 2, 3 and 5 years after post-testing. Data collection took place at six field site s. Standardization of data collection procedure was ensured through training and quality cont rol procedures. The ACTIVE study design is detailed extensively by Jobe and colleagues (2001) and illus trated in Figure 3-1. The present study was focused on subset of 698 controls, which constituted a nonintervention longitudinal sub study from the larger ACTIVE trial.
37 Recruitment The main goals of recruitment for the ACTI VE trial was to induct older adults who were at risk for loss independence due to functional dec line, but who have yet to experience such declines, and to compile a r epresentative sample of older adults, with particular emphasis on representation of Af rican-American elders (Jobe et al., 2001). Recruitment ran from Marc h 1998 through October 1999. Recruitment strategies and sources varied by site; strategies included on -site presentations, letters to interested persons, newspaper advertisements, introduc tory letters and follow-up telephone calls. (Jobe et al., 2001) The University of Alabama recruited par ticipants through the Alabama Department of Public Safety and through UAB eye clinic s. The Hebrew Rehabilitation Center for Aged recruited from congregate and senior housing sites, senior centers, and a research volunteer registry. Indiana University recruited through a network of facilities providing activities and social services to seniors, as well as local churches and senior citizens organizations. Johns Hopkins Univer sity recruited from senior community organizations and centers, churches, senior housing, and through programs offering or coordinating wellness or serv ice programs for seniors. P ennsylvania State University recruited through a state-f unded pharmaceutical assistance program for low-income elders. Wayne State University recrui ted from a large range of community organizations, churches, hospital-based seni or assessment centers, senior-housing sites, and driver registration lists. Eligibility To produce a sample of individuals that is physically and cognitive ly healthy enough to remain in the study for at least 2 y ears (initial enrolment period), not planning
38 to move away, not yet experiencing ADL impairment that would undermine potential intervention benefits, and not having a terminal condition that would make individuals unavailable for follow-up testing, the following exclusion crit eria were implemented: a) less than 65 years of age at init ial screening; b) substantial existing cognitive decline (score of 23 or less on the Mini-Mental Stat e Examination (MMSE; Folstein et al., 1975); self-reported diagnosis of Alzheimers Diseas e); c) substantial existing functional decline (self-report of 2 ADL disabilities) (Morris et al ., 1997); d) medical conditions that are likely to lead to imminent functional decline or mortality before the 2 year assessment (e.g. certain cancers); e) seve re sensory losses (sel f-report of extreme difficulty reading newspaper print or performance-based vision test scores exceeding 20/50) (Mangione et al., 1992); f) communicative difficult ies substantial enough to prevent participation in study pr otocol (interviewer-rated); g) recent cognitive training; or h) unavailability during t he phases of the study. Participants Full ACTIVE Sample A total of 2802 cognitively healthy, comm uni ty-dwelling older adults aged 65 to 94 years comprised the ACTIVE analytic samp le. Table 3-1 presents the baseline sociodemographic characteristics of the full analytic ACTIVE sample. Following baseline assessment, participants were randomly assigned to one of 4 groups: memory training, reasoning training, speed training or no-c ontact control. Following random assignment, 703 individuals were randomized to memory training, 699 to reasoning training, 702 to speed training and 698 to the no-contact contro l group. In the absence of intervention, the control sample constituted a de facto l ongitudinal study that pe rmitted us to examine natural rates of change in basic and everyday cognition, and everyday function
39 functional. Thus, only data from the no-contac t control participants was included in the present analyses. Present Study Sample The present study sample consist ed of the 698 individuals randomized to the nocontact control group and who contributed dat a on at least one of the 14 measures selected for this study (see measures table, Table 3-4). Table 31 presents the baseline socio-demographic characteristics of the present sample. Relative to the rest of the ACTIVE analytic sample at baseline (i.e. th ose randomized to intervention groups), the present study sample was .45 years older, with no significant differences in years of education, MMSE scores, gender, race, or marita l status (Table 3-2). Thus, it was quite representative of the parent sample. Aggregate Longitudinal Retention Patte rn of the Present Stud y Sample The present study assessed data collected at baseline, Year 1, Year 2, Year 3 and Year 5 of the ACTIVE clinic al trial. The longitudinal re tention pattern of the subsample selected for this study is presented in Figure 3-2, which depicts sample size and the percentage of baseline sample assessed at each occasion. It is important to emphasize that these number s are aggregate numbers. At each wave, participants could exit and re-enter the study. Characterization of Attrition Effects To characterize the selectivity of longitudi nal attrition to facilitate interpretation of the generalizability of results, study participants from the or iginal baseline sample (N698) who were assessed at Year 5 (N=452) were compared to those who were not assessed at this occasion (N=246). Results from these analyses are presented in Table 3-3. Relative to those not assessed, thos e assessed at Year 5 were younger, had more
40 years of education, hig her MMSE scores and had a higher percentage of females. White participants were more likely to be re -assessed at Year 5 than participants in other racial groups. There were no signific ant differences in marital status. These results suggest that attrition was selective, leading to the retention of younger, healthier individuals over time. Measures Table 3-4 outlines the measures employ ed in the present analyses organized by domain. The choice of measures to includ e in the present analyses stemmed from prior ACTIVE studies, which showed t hat these measures yielded sufficient variance for the present analyses (e.g. Ball et al., 2002). Basic Cognition Measures Basic memor y. Memory ability, the ability to encode and recall word lists and paragraphs, was included as a consequence of prior work illustrating its relationships to everyday cognition and function (Diehl et al., 1995; Willis et al., 1992; Willis, 1996). Furthermore, memory ability has been show n to be vulnerable to age-related decline (Schaie, 1996). Basic memory was assess ed using the Hopkins Verbal Learning Test, Related Word Lists (HVLT; Brandt, 1991); Rey Auditory-Verbal Learning Test, Unrelated Word Lists (AVLT; Rey, 1941); and the Rivermead Behavioral Memory Test, Paragraph Recall task (RBMT-PR; Wilson et al., 1985). The HVLT required participants to recall a list of 12 words, which could be grouped into semantically related categories (e.g precious stones, animals). The list was repeated and recalled for 3 trials and include d a recognition task in which participants heard the list of words from the original list, new word s from the same semantic categories as the target list, and new semant ically unrelated words. Participants were
41 asked to identify whether or not they rec ognized each word from the original list. Responses to each trial were provided in wri tten form. The test-retest reliability of the HVLT was 0.73 (Ball et al., 2002). The AVLT measure used in this study consis ted of five learning and recall trials. Participants heard a 15-item word list that was read aloud by the tester with an 8second pause between each word. They were then asked to write down in any order as many words as they could remember, in cluding any words recalled correctly on previous trials. Two minutes were allowed fo r writing down the words they recalled. This was repeated five times. The participants re sponses were then scored for total number of words correctly recalled for each of the five trials (0 for each trial), with total scores potentially ranging from 0 to 75. The te st-retest reliability of the AVLT was 0.78 (Ball et al., 2002). The RBMT-PR assessed story memory. Alternate passages were administered at each occasion. Participants listened to t he story on an audiotape, and were asked to write down everything they could remember immediately after the story was finished. The story is divided into idea units, or individual lexical items, which were used for scoring story recall. Each story consisted of 21 idea units. Participants received scores for the quality of their recall of each idea unit (0 = not recalled; 0.5 = approximately accurate; 1 = completely accurate). The su m of these for each person at each occasion provided the RBMT-PR raw score, which r epresented total recall. The test-retest reliability of the RBMT-PR was 0.60 (Ball et al., 2002). Basic reasoning. Inductive reasoning skills were targeted due to prior work showing their vulnerability to age-related dec line, and illustrating a relationship between
42 these abilities and everyday cognition and function (Schaie, 1996; Allaire & Marsiske, 2002; Royall et al., 2007). Basic reasoning was assessed using the Letter Sets, Letter Series, and Word Series tasks (Gonda & Schaie, 1985; Thurstone & Thurstone, 1949; Ekstrom et al., 1976, respectively). These measures are standardized, timed, paperand-pencil assessments. Letter Sets items consist of five sets of letter with four letters in each set. Four of the five sets are alike in some way and t he participant was asked to determine which letter set did not belong with the rest. Af ter working through several examples, the participant was given 7 minutes to work th rough 15 problems. The total number of items completed correctly in the time allowed ( possible scores of 0-15) was the score. The test-retest reliability of the Letter Sets was 0.69 (Ball et al., 2002). Letter Series items consist of series of le tter of varying lengths, and the participant was asked to determine the letter that s hould go next in the series, selected from among five choices. After working thr ough several examples, the participant was presented with a total of 30 word series and told he/she had a 6-minute time limit. The total number of items complet ed correctly in the time allowed (possible scores of 0-30) was the score. The test-retest reliability of th e Letter Series was 0.86 (Ball et al., 2002) Word Series items resembled those of t he Letter Series task, but consisted of a series of days of the week or months of the year, and the participant determined how the series progressed, in order to select the next day/month in the series from among five choices. After working through severa l examples the partici pant was presented with a total of 30 word series and told he/she had a 6-minute time limit. The total number of
43 items completed correctly in the time allo wed (possible scores of 0) was the score. The test-retest reliability of the Word Series is 0.84 Basic speed. This domain targeted visual-spatial perceptual speed, and was selected due to prior studies illustrating its relationship to everyday outcomes, such as driving (Ball, Owsley, Sloane, Roenker, & Bruni, 1993), and everyday function (Barberger-Gateau, 1999). Thus, basic speed was assessed via the Useful Field of View (UFOV; Ball, et al., 1993) tasks 2 and 3. Tasks 1 and 3 were also assessed; however they were excluded from analyses due to floor and ceiling effects that limited their variance (Ball et al., 2002). UFOV Task 2, or Divided Attention, wa s comprised of central and peripheral stimuli presented on a computer screen. The stimuli were dichromatic, two-dimensional drawings of either a car or a truck. During the task, one of two stimuli was presented centrally and at the same time, a car wa s presented to one of 8 locations around the periphery of this central stimulus. Participants were required to judge the identity of the centrally presented stimulus and to identify t he location of the peri pheral stimulus, at decreasing latencies. The minimum stimulus duration needed to perform both aspects of this task at 75% correct was the score. UFOV task 3, or Selective Attention, was identical to UFOV Task 2, with the exception of the addition of visual clu tter around between the central and peripheral stimuli, in the form of an a rray of triangles. Again, the minimum stimulus duration needed to perform both aspects of this task, which now included speeded visual search, at 75% correct was the score.
44 For both measures, higher scores represented longer latencies required for a correct response; therefore hi gher scores were indicative of worse performance, while lower scores were indicative of better perform ance on this task. The test-retest reliability of a UFOV composite of the tw o tasks used in this study, plus a third, was 0.80 (Ball et al., 2002). Everyday Cognition Measures Everyday cognition, the abili ty to solve wellstructured, everyday problems, was selected due to prior findings illustrating a relationship with self-reported everyday function (Allaire & Marsiske, 2002), as well as age-related decline in this domain (Willis et al., 1992) assessed with the Everyday Probl ems Test (EPT; Willis & Marsiske, 1993), the Observed Tasks of Daily Living (OTD L; Diehl, et al., 2005), and the Timed Instrumental Activities of Daily Living (tIA DL; Owsley, Sloane, McGwin, & Ball 2002). The EPT is a pencil-and-paper measure designed to assess adults ability to solve problems of daily living using printed mate rials. Participants were presented with 14 everyday stimuli (e.g. medication labels transportation schedules, telephone rate charts, Medicare benefits charts) and were asked to answer two questions about each stimulus (e.g. to calculate the number of days the supply of medication will last). The EPT has items representing 7 domains of da ily living: meal preparation and nutrition, medication use and health behaviors, te lephone use, shopping, financial management, household management, and transpor tation. Scores represent ed the number of correct answers generated (possible scores of 0 28). The standardized alpha for the EPT was 0.94 and that the 1-year Spearman-Brown test-retest st ability was 0.91 (Marsiske &Willis, 1994). Ball and colleagues (2002) reported a test-retest reliability of 0.87 for this measure. See Appendix A for sample items.
45 The OTDL is conceptually similar to the EPT, in that participants are required to use printed materials (e.g. ca ke mix ingredients, medicine bottles, telephone book) to solve everyday problems. The OTDL cons isted of nine tasks, with a total of 13 questions addressing three IADL domains (medications, phone usage, and financial management). An example task is allowing t he subject to examine three medication containers with pharmacy labels attached and then asking him/her how many days will a refill of a given medication last or which medications mi ght cause drowsiness. Testers were required to indicate the correctness of each response by circling yes or no and then writing verbatim responses in the spac e allowed. Furthermore, testers indicated whether or not it was necessary to prom pt the participant. Perf ormance on the OTDL measure was scored by a certified scorer (p ossible scores 0 28). Diehl et al. (2005) reported that the Kuder-Richardsons corre cted alpha for the total measure was 0.71 and the internal consistency of the total me asure was 0.81. Ball et al. (2002) report an internal consistency of 0.75. S ee Appendix B for sample items. In the tIADL, participants were asked to perform 5 everyday tasks (finding a phone number in the phonebook, counting out change, reading food can ingredients, finding items on a shelf, and reading directions on m edicine containers) using everyday stimuli (e.g. a phonebook, coins, cans of food). A ccuracy of task performance and time to complete each task was recorded with a st opwatch. For each task there was a completion time and an error code, which we re combined as follows. Subjects with a major error code on a given item were assigned the maximum time allotted on that item. For those with a minor error code, a time penalty was added to their completion time; this penalty was defined as 1 SD based on the data from all participants who had
46 completed that item without error (range of scores 46 806). Owsley et al. (2002) reported that the 7-8 week test -retest reliability for the tIADL was 0.85. Ball et al. (2002) reported a test-retest reliability of 0. 64. See Appendix C for sample items. Everyday Function Measure Everyday Function, which focused on instrumental activities of daily living (IADL) capacity, was selected due to its role as a key outcome in most aging research (Diehl, 1998). This domain was assessed with a self-r eport measure, drawn from the Minimum Data Set methodology (Morris et al., 1997). In strumental activities of daily living questions elicited self reported capacity in areas such as preparing meals, housework, managing finances, managing health care, shopping, telephone use and travel. Participants were asked, in the last 7 days, how much of the activity did you do on your own? and then asked how difficult was it (o r would it have been) to do on your own? Responses ranged from not difficult, to great difficulty, on a 5-point Likert-type scale (possible scores 19 to 57; the measure is included in Ap pendix D). According to the coding of this scale, higher scores represent ed more difficulty with IADL, and lower scores represented less difficulty, thus a lower score on this measure would be indicative of gain or improvement, while a hi gher score would be in dicative of loss or deterioration in everyday functioning. The reliability analysis of a precursor measure using a similar assessment approach yielde d a weighted Kappa of 0.76, which is indicative of excellent reliabi lity (Morris et al., 1997). In ACTIVE, Ball et al. (2002) reported an internal consistency alpha of 0.75 for the IADL Difficulty scale. Statistical Analyses The study addressed two main experimen tal hypotheses. First we aimed to replicate prior cross-sectional dat a and show that everyday cognition mediates the basic
47 cognition variance (both assessed at the lat ent level) in everyday function at the baseline occasion of measurement (Aim 1). To address these questions, analyses examined the relationships between all obs erved measures of basic and everyday cognition and everyday function, followed by analyses of latent-level mediation using nested models via Structural Equation Modeling (SEM). Second, we aimed to confirm that basic c ognitive abilities would be longitudinal (occasion to occasion) predictors of every day function, and that everyday cognition would mediate this relationship. We em ployed the Multilevel modeling (MLM; Bryk & Raudenbush, 1992) approach to assess the questi on of longitudinal mediation of the basic cognition variance in everyday function by everyday cognition. To do so we first assessed the nature of longitudinal change in t he data (Aim 2.1) followed by analyses of mediation using nested models (Aim 2.2). Data Preparation and Reduction To be conformal with the assumptions of multivariate statistics (especially normality), and to produce correct effect c odings for some model s, data were first prepared for the analyses, as described in the next sections. Aim 1: Structural Equation Models. All data were Blom transformed (Blom, 1958) by ACTIVE statisticians prior to analyse s. The Blom transformation converts each value of the item distri bution to its Z-score equivalent at the same percentile on the standard normal distribution, und er the assumption that the underlying true score is normally distributed, resulting in improved normality of the data. Aim 2.1 and 2.2: Multilevel models. For basic memory, r easoning and speed, and for everyday cognition, composite vari ables were computed by averaging across constituent measures (basic memory: RBMT -PR, HVTL, AVLT; basic reasoning: Letter
48 Series, Letter Sets, Word Series; basic speed: UFOV 2 & 3; everyday cognition: EPT, tIADL, OTDL). These composite variables we re utilized in the MLM analyses of change over time. Visual inspecti on of the longitudinal data suggested both linear and quadratic time trends. Furthermore, prior longitudinal work in aging suggests that quadratic decline in abilities is a common trajectory of change (Schaie, 1994), which is likely related to similar patterns of change observe d in brain anatomy (e.g. Raz et al., 1995). Two time variables, one representing linear ti me (coded as 0.23, 1. 23, 2.23, 3.23, 5.23, to reflect occasions of measurement and their spacing) and another representing quadratic time, computed by squaring the li near time variable and taking the residual, were computed to fit to the longitudinal dat a. All variables were then converted to Zscores, which were utilized in analyses to facilitate comparisons. Aim 2.2: Multilevel mediation models. In addition to the linear and quadratic time variables used in analyses of change over time (Aim 2.1), 8 a dditional variables were created, to be examined as predictors of everyday function. For each cognitive variable: basic memory, reasoning, speed, and everyday cognition, two variables were computed, one to represent the mean performance in each domain across occasions, and the second to capture devia tion from this mean value at each occasion. This latter variable is therefore centered on the mean level, and represents the slope of longitudinal change in performance relati ve to the mean (Singer &Willett, 2003). Criteria for Evaluation of Models Prior to considering the specific models, this section outlines the criteria used to evaluation models for the two main study aims. Aim 1: Structural Equation Models. Structural equation models were estimated as a planned series of nested models, in which each s ubsequent model added or
49 removed paths from the preceding model. For these nested models, model adequacy was evaluated using a number of criteria (e.g. Martens 2 005, Weston & Gore, 2006), including the root mean square error of appr oximation (RMSEA), a fit index indicating the discrepancy between the original and repr oduced covariance matrix divided by the degrees of freedom and for which values of 05 or lower were indicative of adequate fit; pCLOSE, or the probability of close fit a ssessed the null hypothesi s that the population RMSEA was no greater than .05; the Comparative Fit Index (CFI), for which values close to 1.0 indicate a good fit; the Norm ed Fit Index (NFI), which compared the improvement in the minimum discrepancy for the specified model to the discrepancy for the independence model, where a value below 0.90 indicated that the model can be improved; the Relative Fit Index (RFI), whic h takes the degrees of freedom of the model into account, values close to 1.0 indicated good fit; the Incremental Fit Index (IFI), for which values close to 1.0 indicated good fit (Bentler & Chou, 1987; Hu & Bentler, 1997; Marsh, Balla & McDonald, 1988; Schumaker & Lomax, 2004). Furthermore, change in Chi-Square ( 2), where significance was indicative of a notable change in relative model fit, and the Akaikes Information Cr iterion (AIC), where improvement in fit signaled lower relative values, were also examined. Aims 2.1 and 2.2: Multilevel models. MLM allows for the examination of fixed and random effects. Fixed effects refer to the aver age effects, or e ffects that hold true across all individuals. Random effects test whether there are significant individual differences in the obtained fixed effects. For example, with res pect to the effects of time, fixed effects would illustrate whether the longitudinal data across individuals can be characterized by growth, decline or a combinat ion of the two. A random effect of time
50 would illustrate whether this slope of change varies significantly between individuals (i.e. some individuals improve or decline faster or slower than others). Furthermore, MLM analyses permit the examination of predictors that interact with the dependent variable on separate levels. When data is clustered within persons, meaning that the same people are observed at multiple occasions of measurement (as in this study), Level 2 predictors answer the question: to what exte nt does a persons level on the predictor influence their outcome at each occasion. For ex ample, what is the effect of a persons mean level of memory on thei r self-reported everyday func tioning (IADL Difficulty)? On the other hand, a Level 1 predictor varies with the dependent variable at each occasion and permits the examination of how change in the predictor is related to change in the dependent variable. For example, what might participants endorse on everyday function, on those occasions when their bas ic memory performance is higher or lower than their mean value of basic memory? All longitudinal analyses were conducted via series of nested models. These models were estimated using the Maximum lik elihood (ML) method, which was selected to facilitate the examination of both fix ed and stochastic model components making it better suited for conducting nested model te sts. The fit of eac h subsequent modeling step was compared to that of the prior step. Furthermore the withinand betweenperson variance explained by subsequent modeli ng steps was also compared to that of the initial, worst-fitting model, which defined the withinand betweenperson variance to be explained (i.e. model 1 in Aim 2.1 and 2.2). For each model, relative goodness of fit was assessed via an examination of the reduction in -2LL (denoted 2), as well as via changes in AIC, & the Schwartz Bayesian Info rmation Criterion (BIC). Improvements in
51 the predictive value of a modeling step were evaluated by the extent to which the modeling step explained the withinand betweenperson va riance, relative to the criterion model (Bryk & Raude nbush, 1992). Decreases in the intercept related and residual variance, represent a proportional re duction of the prediction error, which is analogous to R2, and used as an estimate of effect size. Analytical Framework for Aim 1: Cross-sectional Mediation Structural Equation Modeling (SEM) with Amos 17 software (Arbuckle, 2008) was used to assess the relationships among bas ic cognition everyday cognition, and everyday function. Multiple steps were us ed to evaluate whether particular paths were essential to model fit, or w hether they could be eliminated. The broad goal of this section was to evaluate whether everyday cogni tion might account for some or all of the basic cognitive variance in everyday func tion, although formal mediation analyses were not conducted because of the failure to m eet certain preconditions for establishing mediation (Baron & Kenny, 1986). Explication of modeling steps. Modeling steps were as follows: 1. Measurement model: this was the best-fitting model consisting of all possible correlations between the four hypothesized latent-level factors: basic memory, basic reasoning, basic speed, every day cognition, and observed (non-latent) everyday function (IADL difficulty). Giv en that the Measurement model was the best-fitting model, its fit was used as the criterion against which subsequent models fit was assessed. 2. Fully-recursive model : this model estimated all possible regression paths between the four cognitive lat ent factors, the predictors, and everyday function, the outcome variable. This step was mathemat ically equivalent to the Measurement model; however it yielded standardized es timates that represent the unique relationships between each predictor and th e outcome. Changes in these values in subsequent nested models were evaluated in addition to changes in model fit. 3. Basic cognitive factors predict everyday function : in this step, the path from everyday cognition to everyday functi on is removed, with the resulting model
52 assessing the unique relationships between latent-level basic memory, reasoning and speed, and everyday function without ev eryday cognition as a predictor. 4. Basic memory predicts everyday function: in this model the paths from basic reasoning and speed to everyday function are removed to assess whether basic memory is a significant predictor of ever yday function. This step was based off the results from step 3, which established that Basic reasoning and speed were not significant unique predictors of Everyday function. 5. Basic memory and everyday cognition predict everyday function: in this model the path from everyday cognition to everyday function is added to model 4, to evaluate whether the addition of ever yday cognition as a predictor would diminish the unique relationship between basic memory and everyday function, which would suggest mediation of this variance by everyday cognition; 6. Direct path from basic memory to everyday function is removed : the purpose of this step is to assess the extent of the mediation of the basic memory variance in everyday function by everyday cognition By removing the direct path from basic memory to everyday function, this model allows for an assessmen t of the indirect path from basic memory via everyday cognition to everyday function. Analytical Framework for Aim 2.1: Change over Ti me in Study Constructs Prior to conducting analyses investigating whether changes in basic and everyday cognition varied systematically with changes in everyday function, a first set of analyses sought to describe the functional form of 5year cognitive change in each of the study constructs. Explication of modeling steps. The nested models to assess change over time were estimated in a series of five steps for each study construct (basic memory, reasoning, speed, everyday c ognition and everyday function). The five steps estimated were as follows: 1. Unconditional means (no predictors): this was t he worst fitting model, which defined the withinand betweenperson variance to be explained. This model served as the criterion model for evaluati ons of the predictive value of subsequent models. 2. Fixed linear time : in this model the fixed effects of the linear time variable were assessed.
53 3. Random linear time : in this model the random effe cts of the linear time variable were added to model 2. 4. Fixed quadratic time : in this model the fixed effect s of the quadratic time variable were assessed in addition to the predictors in model 3. 5. Random quadratic time : this model adds an estimate of the random effects of the Quadratic time variable to the predictors in model 4. Analytical Framework for Aim 2.1: Longitudinal Me diation of Basic Cognitive Variance in Everyday Function by Everyday Cognition The longitudinal mediation of the basic cognition variance in everyday function was also examined using the MLM approach, building on the models in Aim 2.1, with Everyday function as the dependent variable. Explication of modeling steps. The seven nested models were as follows: 1. Fixed quadratic time : this model was the final estima ble model from Aim 2.1, which explored the nature of the longitudinal change in ever yday function by fitting linear and quadratic time variables to the data (model 4 from Aim 2.1; Table 4-10). For the series of the following nested models, this model serves as the baseline, worstfitting model, which defines the wit hinand between-person variance to be explained. Furthermore, this model estab lishes the fixed effects of linear and quadratic time as well as the to-be-explai ned variance attributable to the random effects of linear time. Thus, in subsequent m odels, in addition to testing mediation of the basic cognition variance in everyday function by everyday cognition, the goal is to determine whether additional predictors a) account for the fix ed effects of linear and quadratic time, and b) account for the signi ficant variance in individual positive linear slopes. 2. Fixed mean basic cognition : in this model, the fix ed effects of mean basic memory, reasoning and speed are added to the model. 3. Fixed centered basic cognition : in this model, the fixe d effects of the centered Basic memory, reasoning and speed are added to the model. 4. Random centered basic cognition : in this model, the random effects of the centered Basic memory, reasoning and speed are added to the model. 5. Fixed mean everyday cognition : in this model, the fixed effects of mean Everyday cognition are added to the model. 6. Fixed effects of centered everyday cognition : in this model, t he fixed effects of centered Everyday cogniti on are added to the model.
54 7. Random effects of centered everyday cognition : in this model, the random effects of centered everyday cognition are added to the model.
55 Table 3-1. Sociodemographic characteristi cs of the present study sample and the full ACTIVE sample Present study sample (N=698) Full ACTIVE sample (N=2802) Mean S.D. Rang e Mean S.D. Rang e Age 74.05 6.05 65-94 73.63 5.91 65-94 Years of Education 13.37 2.71 6-20 13.53 2.7 4-20 MMSE 27.27 2.00 23-30 27.31 2 23-30 Gender % % Women 73.6 75 Men 26.4 25 Race White 71.1 72 African American 26.8 26 Other/Unknown 2.1 2 Married 37.1 36 MMSE: Mini-Mental State Exam Table 3-2. Sociodemographic co mparison of the present study sample to the rest of the ACTIVE sample at baseline Characteristic ACTIVE Trained Sample (N=2,104) Present Study Sample (N=698) t df d 2 p Age, years 73.50 74.05 2.152800 0.09 0.032 Years of Education 13.58 13.37 1.752798 0.07 0.080 MMSE 27.33 27.27 0.612800 0.02 0.543 Sex, %Female 76.60 73.60 2.540.111 Race, %White 72.80 71.10 1.910.384 Marital Status, %Married 35.00 37.10 0.550.458 MMSE: Mini-Mental State Exam
56 Table 3-3. Sample attrition, comparing participants assessed at Year 5 to those not assessed at Year 5 Characteristic Assessed at Year 5 (N=452) Not Assessed at Year 5 (N=246) t df d 2 p Age, years 73.58 74.91 2.78 696 0.22 0.006 Years of Education 13.58 13.00 -2.71 696 0.21 0.007 MMSE 27.52 26.83 -4.41 696 0.35 0.000 Sex, %Female 68.30 66.30 10.66 0.001 Race, %White 75.70 63.00 17.66 0.000 Marital Status, %Married 36.30 38.60 0.37 0.542 MMSE: Mini-Mental State Exam Table 3-4. Measures used Cognitive Domain Te st Published Source Basic Memory Hopkins Verbal Learning Test, Related words (HVLT); Rey Auditory Verbal Learning Test, Unrelated words (AVLT); Rivermead Behavioral Memory Test, Paragraph recall (RBMT-PR) Brandt, 1991; Rey, 1941; Wilson et al., 1985 Basic Reasoning Letter Sets; Letter Series; Word Series Gonda & Schaie, 1985; Thurstone & Thurstone, 1949; Ekstrom et al., 1976 Basic Speed Useful Field of View (UFOV), Tasks 2 & 3 Ball & Owsley, 1993 Everyday Cognition Everyday Problems Test (EPT); Observed Tasks of Daily Living (OTDL); Timed IADL (tIADL) Willis & Marsiske, 1993; Diehl et al., 2005; Owsley et al., 2002 Everyday Function Activities of Daily Living and IADL functioning Minimum Data Set -IADL perceived degree of difficulty Morris et al., 1997
57 Figure 3-1. ACTIVE study design. Note: boos ter training occurred prior to the first and third follow-up assessments.
58 Figure 3-2. Aggregate longitudinal retention pattern of the current study sample. Note: the chart doesnt describe participants entrances and exits
59 CHAPTER 4 RESULTS Overview This study investigated the relations hip between basic c ognition, everyday cognition and everyday function in a sample of healthy, community-dwelling older adults age 65 and over who were tested at five occa sions of measurement (baseline, Year 1, Year 2, Year 3 and Year 5 post baseline). Specifically, the study addressed the question of whether everyday cognition accounts fo r the basic ability variance in everyday function cross-sectionally, or at a specific poi nt in time, as well as longitudinally over time. Aim 1: Baseline Mediation of Basic Cogni tion Varia nce in Everyday Function by Everyday Cognition Preliminary Analyses Initial analy ses examined the interrelations hips between all observed measures of Basic and Everyday cognition, and Everyday function (see Table 3-4 in the Measures section) at baseline. Table 4-1 displays the bivariate corre lations between measures of basic memory, reasoning, and speed, everyday cognition and everyday function. As can be seen in the table, all measures were signifi cantly correlated, with the exception of the correlations between Everyday function (IADL difficulty) and the RBMT-PR, and everyday function and Letter Sets. Correla tions between measures of basic and everyday cognition were medium to large (-0.27 to 0.83), while all correlations with everyday function were small (.03 to 0.19) (Cohen, 1992). Nested Model Findings The following models assessed the baseline relationship between basic cognition, everyday cognition and everyday functi on. A comprehensive explication of these
60 modeling steps is presented in the Statistical Analyses section of Methods, above. This portion of the analyses was intended to replicat e and extend prior findings (e.g. Allaire & Marsiske, 2002), and is the modal appr oach in the literat ure to date. (1) Measurement model. The standardized four-factor solution for the Measurement model is displayed in Tabl e 4-2. All measures loaded highly and significantly on their respective factors, co nfirming the hypothesized four-factor structure (basic memory, basic reasoning, basic speed, everyday cognition). Model fit indices for this and all subsequent baseline models are displayed in Table 4-3. As Table 4-3 illustrates, the fit of the Measurem ent model was good. Standardized factor intercorrelations are presented in Table 4-4. The correlations between all basic cognitive factors and everyday cognition were high (-0.57 to 0.85), while the correlations with everyday function were low (-0.10 to -0 .15). Results from t he measurement model reflect these from the bivari ate intercorrelations between observed measures of basic and everyday cognition and everyday function, and suggest that measures of cognition have stronger relationships with each other than they do with everyday function. (2) Fully-recursive model. This model estimated all possible regression paths between the four cognitive latent factors and Everyday function. As illustrated in Table 7, the fit of this model wa s good, with no decrement in fit from the Measur ement model. The standardized loadings, and variance expl ained is presented in Figure 4-1. As Figure 4-1, panel A illustrates, only the uni que relationships between basic memory and everyday cognition (0.33), and basic reas oning and everyday cognition (0.58) were significant ( p < .05), suggesting that when added to the predictive value of memory and reasoning, speed of processing is not a signific ant predictor of everyday cognition. None
61 of the paths from basic or everyday cogni tive variables to everyday function reached significance, suggesting that when all cognitive latent variables were simultaneously included as predictors of ever yday function, their unique predictive variance, above and beyond the variance shared with one another, was trivial. The basic cognitive factors together accounted for 80% of t he variance in everyday cognition, while all the cognitive factors together accounted for only 3% of the variance in everyday function. (3) Basic cognitive factors predict everyday function. In this model, the path from everyday cognition to everyday functi on was removed, to allow only the basic cognitive factors to predict everyday functi on. The fit of this model was good, with nonsignificant decrement in fit fr om the previous, Fully recursiv e model (Table 4-3, panel A). The standardized loadings and variance ex plained in the dependent variable is presented in Figure 4-1, panel B. As illustrated by Figure 4-1, panel B, with basic reasoning and speed controlled for, only basic memory significantly predicted everyday function ( p < .05). Without everyday cognition in the model, the basic cognitive factors still accounted for 3% of the variance in everyday function. (4) Basic memory predicts everyday function. Given results from the previous model, the paths from both basic reasoning and speed were removed, leaving only basic memory as a predictor of everyday function. This model exhibited good fit and no decrement in fit from the pr evious model (Table 4-3). Figur e 4-1, panel C illustrates the standardized loadings and variance explained by this model. Basic memory alone was a significant predictor of everyday function ( p < .05), accounting for 2% of its variance. (5) Basic memory and everyday co gnition predict everyday function. In this model, everyday cognition was included as an additional direct predictor of everyday
62 function, in addition to basic memory. The fi t of this model was good, and resulted in improvement in fit over the previous model where basic memory alone predicted everyday function (Table 4-3). The standardi zed loadings and variances explained are presented in Figure 4-1, panel D. As Figure 4-1, panel D illustrates, basic memory was a significant predictor of everyday cognition ( p < .05), accounting for 80% of its variance. With everyday cognition in the model, neither basic memory nor everyday cognition was a significant unique predictor of everyday function. The change in the predictive status of basic memory, from a -weight of 0.15 ( p < .05) to 0.08, ( p > .05), suggests that everyday cognition accounted for its variance in everyday function. (6) Direct path from basic memory to everyday function is removed. To examine the extent of the mediation of the basic memory variance in everyday function by everyday cognition, the direct path from basic memory to everyday function was removed. This fit of this model was good, and resulted in no decrement in fit from the previous model (Table 4-3). The standardize d loadings, presented in Figure 4-1, panel E, show that everyday cognition became a si gnificant predictor of everyday function, thus fully accounting for the basic me mory variance in everyday function. Aim 2.1: Change over Time in Study Constructs The following set of analyses was carried out in order to describe the trajectory of change ov er time in the study constructs A comprehensive ex plication of these modeling steps is presented in the Statistical Analyses section in Methods. These descriptive analyses allowed us to characterize longitudinal change in the study constructs, the nature of which must be established prior to investigations of coupled change, which were carried out in Aim 2.2, building on these results.
63 Table 4-4 displays the fit indices and vari ances explained by the series of nested models that examined the fixed (across individuals) and random (interindividual differences) effects of linear and quadratic time for each study construct (basic memory, reasoning, and speed, everyday cognition and ev eryday function). Of the five constructs under investigation, four fa iled to achieve convergence when Random quadratic time was estimated. Subsequently, step 4, Fixed quadratic time, was the last estimable model. Table 4-5 displays the fixed and random estimates for modeling step 4, Fixed quadratic time, for each of the five variables. As illustrated in Table 4-5, each modeling step represented significant improvements in fit over the preceding m odeling step, which was evidenced by a reduction in the -2LL ( 2). All variables evidenced significant linear positive trends, indicative of gain or improvement over time, that were modified by significant negative quadratic trends, indicative of loss or decline, that were true across participants (fixed effects). Furthermore, all variables evidenced significant individual differences (random effects) in linear gain (Fi gure 4-2). These findings indicate that in general, participants evidenced (linear) im provement, and there were individual differences in this gain. Across participants, this positive linear trajectory was modified by a quadratic trend, which suggested that the overall longi tudinal trend in the study constructs was initial increase followed by a decline, which did not appear to vary between participants. Aim 2.2: Longitudinal Mediation of Basi c Cognitive Variance in Ever yday Function by Everyday Cognition Nested models were used to examine the longitudinal interrelationships between basic cognition, everyday cognition, and ever yday function as well as to determine the extent to which the cognitive predictors accounted for the effects of time. Table 4-6
64 displays the model fit characteristics as we ll as the between-person (Level 2) withinperson (Level 1), and linear time-related (Level 1) variance explained by the predictors in each of the seven nested models. Table 4-7 displays the estimates for the final model 7, Random effects of cent ered everyday cognition. As illustrated by Table 4-6, each m odeling step represented a significant improvement in fit relative to the previous model, with the exceptio n of the final model, which failed to achieve a significant improvement in fit ( 2 (1) = 0.08, p >.05). As Table 4-7 (right side) illustrates, the mean and centered basic and everyday predictors together accounted for about 29% of the between-person and 14% of the within-person variance in everyday function, above and bey ond that accounted for by the effects of time. Furthermore, these predictors acc ounted for about 42% of the individual differences in linear gain observed in everyday function. As illustrated by Table 4-7, the fixed effects of both linear and quadratic ti me, and the random effects of linear time remained significant with all the cognitive predictors in the model, suggesting that the cognitive predictors failed to fully explai n the linear gain and quadratic decline in everyday function that was observed across participants, as well as for the individual differences in linear gain. With respect to the longitudinal interrelationships between basic abilities, everyday cognition and everyday function, results fr om models 2 5 will first be summarized, and subsequently compared to those of m odel 6, illustrated in Table 4-7. In model 2, the fixed effects of all the mean cognitive variables were estimated. Mean memory and speed emerged as significant predictors ( p < .001), suggesting that older adults whose average, cross-occasion le vel of memory is higher, report fewer
65 functional difficulties Older adults whose level of speed was slower (indicative of poorer performance) reported more f unctional IADL difficulty. In model 3, the fixed effects of ce ntered memory, reasoning and speed were included to the predictors in model 2. Thes e predictors, which were coded to capture change over time in these constructs, allow ed us to examine whether change in basic abilities was predictive of change in everyday function. None of these centered basic cognitive variables emerged as significant predictors of every day function, which suggested that across individuals, occasi on-to-occasion changes in basic cognitive abilities (relative to their mean performanc e across occasions) were not associated with functional scores. In model 4, random effect s of these centered basic cognitive variables were estimated, and were found not to be signi ficant, suggesting that change in basic cognition was not a predictor of change in ev eryday function. The estimates for model 4 are presented in Table 4-8. Therefore, models 2-4 illustrated that over time, individual differences in memory and speed are predictiv e of individual differences in everyday function. In model 5, the fixed effects of mean everyday cognition were added to the predictors in model 4. The estimates for mo del 5 are presented in Table 4-8. These effects were highly significant ( p < .001), suggesting that higher mean levels of everyday cognition were associated with less self-reported difficulty in everyday function. With everyday cognition in the model, m ean memory was no longer significant. Table 4-8, right side illustrates an inferential test for the mediation of the fixed effects of
66 mean basic memory by mean everyday cognition. According to this analysis, mediation would be supported if the estimates from model 5 fell outside of the confidence intervals of the corresponding estimates from model 4. As illustrated by Table 4-8, mean everyday cognition mediated the memoryr elated variance in everyday function, mirroring the cross-sectional fi ndings from Aim 1. Interestingly, with mean everyday cognition in the model, mean reasoning emerged as a significant predictor ( p = .025), which is indicative of the presence of a suppressor effect. In model 6, the fixed effects of centered everyday cognition were assessed. These effects were significant ( p = .003), suggesting an over all occasion-to-occasion association between everyday cognition and ever yday function. In model 7, the random effects of centered everyday cognition we re estimated. These effects were not significant, suggesting that the occasion-to-occasion association between everyday cognition and everyday function did not vary in magnitude between participants, or at least that this study was not sensitive to i ndividual differences in the strength of this relationship. As illustrated in Table 4-7, ac ross participants a point increase in the level of basic reasoning was associated with 11 standard deviation units increase in selfreported functional difficulty; a point incr ease in mean level of speed was associated with a .09 unit increase in functional difficult y, and a point increase in mean level of everyday cognition was associated with 0. 23 unit decrease in functional difficulty. Furthermore, on average, on occasions when scores on everyday cognition went up, self-reported functional difficu lty decreased by 0.05 units.
67 Table 4-1.Bivariate correlations between measures of basic and everyday cognition and everyday function Measure 1 2 3 4 5 6 7 8 9 10 11 12 1. HVLT 1.00 2. AVLT 0.69 ** 1.00 3. RBMT-PR 0.49 ** 0.49 ** 1.00 4. Letter Sets 0.43 ** 0.39 ** 0.40 ** 1.00 5. Letter series 0.49 ** 0.42 ** 0.46 ** 0.67 ** 1.00 6. Word series 0.50 ** 0.45 ** 0.45 ** 0.63 ** 0.83 ** 1.00 7. UFOV2 -0.40 ** -0.33 ** -0.32 ** -0.39 ** -0.48 ** -0.47 ** 1.00 8. UFOV3 -0.38 ** -0.32 ** -0.27 ** -0.40 ** -0.45 ** -0.44 ** 0.62 ** 1.00 9. Timed IADL -0.45 ** -0.42 ** -0.34 ** -0.43 ** -0.52 ** -0.51 ** 0.36 ** 0.39 ** 1.00 10. OTDL 0.47 ** 0.42 ** 0.43 ** 0.46 ** 0.56 ** 0.55 ** -0.39 ** -0.39 ** -0.49 ** 1.00 11. EPT 0.52 ** 0.45 ** 0.52 ** 0.56 ** 0.68 ** 0.66 ** -0.44 ** -0.39 ** -0.54 ** 0.65 ** 1.00 12. IADL Difficulty -0.11 ** -0.16 ** -0.04 -0.03 -0.10 -0.08 0.09 0.10 0.19 ** -0.10 ** -0.10 1.00 p<.05 **p<.001 HVLT: Hopkins Verbal Learning Test, Related Words; AVLT: Auditory Verbal Learning Test, Unrelated Words; RBMT-PR: Rivermead Behavioral Memory Test Paragr aph Recall Task; UFOV: Useful Field of Vi ew; IADL: Instrument al Activities of Daily Living; OTDL: Observed Tasks of Daily Living; EPT: Everyday Problems Test
68 Table 4-2. Factor loadings for the basic and everyday cognition factors Factors Measure Basic Memory Basic ReasoningBasic SpeedEveryday CognitionCommunalities 1. HVLT 0.85 0.72 2. AVLT 0.80 0.64 3. RBMT-PR 0.65 0.42 4. Letter Sets 0.73 0.53 5. Letter series 0.92 0.85 6. Word series 0.89 0.80 7. UFOV2 0.80 0.65 8. UFOV3 0.77 0.59 9. Timed IADL -0.67 0.44 10. OTDL 0.74 0.55 11. EPT 0.85 0.73 All factor loadings were significant HVLT: Hopkins Verbal Learning Test, Related Words; AVLT: Auditory Verbal Learning Test, Unrelated Words; RBMT-PR: Rivermead Behavioral Memory Test Paragr aph Recall Task; UFOV: Useful Field of Vi ew; IADL: Instrument al Activities of Daily Living; OTDL: Observed Tasks of Daily Living; EPT: Everyday Problems Test
69 Table 4-3. Baseline mediatio n nested model fit indices Model Model fit characteristics Model fit comparisons 2dfRMSEApCLOSECFI NFIRFIIFI 2 df AIC 1. Measurement model 108.3845.000.05 0.770.98 0.970.960.990.000.00198.39 2. Fully recursive model 108.3845.000. 050.770.98 0.970.960.990.000.00198.39 3. Basic cognitive factors 110.6846.000. 050.770.98 0.970.960.98-2.301.00198.68 4. Basic memory 111.6848.000.040.830.98 0.970.960.99-1.002.00195.68 5. Basic memory and everyday cognition 110. 7047.000.040.810.98 0. 970.960.990.981.00196.68 6. Direct path from basic memory is removed 111.8048.000.040.830.98 0. 970.960.98-1.101.00195.80 Dependent variable: Everyday function (IADL difficulty) df: degrees of freedom; RMSEA: Root Mean Square Error of Approximation; pCLOSE: probability of Close Fit; CFI: Comparative Fit Index; NFI: Normed Fit Index ; RFI: Relative Fit Index; IFI: Incremental Fit Index; AIC: Akaike Information Criterion Table 4-4. Intercorrelations between basic cognition factors and everyday cognition fa ctor and observed everyday function Factor 1 2 3 4 5 1. Basic Memory 1.00 2. Basic Reasoning 0.66 ** 1.00 3. Basic Speed -0.57 ** -0.65 ** 1.00 4. Everyday Cognition 0.76 ** 0.85 ** -0.65 ** 1.00 5. Everyday Function (IADL Difficulty) -0.14 ** -0.10 0.12 -0.15 ** 1.00 *p <.05 **p <.001
70 Table 4-5. Model fit indices and variance explained for model s examining the effects of time in study constructs AIC: Akaike Information Criterion; BIC: Bayesi an Information Criterion
71 Table 4-6. Estimates of fixed and random effects fo r the fixed quadratic time model for each variable Dependent Variable Fixed Effect Estimate S.E df t p Random EffectEstimateS.E Wald Z p Basic memory Intercept -0.100.04696.01 -2.890.004 Residual 0.180.0125.690.00 Linear time 0.280.031752.089.760. 000 Intercept 0.810.0517.060.00 Quadratic time -0.390.031571.62-13. 110.000 Linear time 0.020.003.770.00 Basic reasoning Intercept -0.100.04702.18-2.630.009 Residual 0.100.0026.030.000 Linear time 0.300.021749.8714.430. 000 Intercept 0.890.0517.830.000 Quadratic time -0.280.021570.57-13. 120.000 Linear time 0.010.004.170.000 Basic speed Intercept 0.070.03707.711. 900.058 Residual 0.220.0126.220.000 Linear time -0.400.031734.02-12.980.000 Intercept 0.750.0416.760.000 Quadratic time 0.380.031597.3111.72 0.000 Linear time 0.010.002.610.009 Everyday cognition Intercept -0.070.04698. 96-2.060.040 Residual 0.190.0127.330.000 Linear time 0.200.031892.357.230. 000 Intercept 0.820.0517.200.000 Quadratic time -0.200.031695.47-7.070.000 Linear time 0.010.002.940.003 Everyday function Intercept 0.030.03685. 391.110.266 Residual 0.470.0227.610.000 (IADL Difficulty) Linear time -0.150.042027.47-3.410.001 Intercept 0.470.0314.150.000 Quadratic time 0.260.041746.245.950.000 Linear time 0.060.015.490.000
72 Table 4-7. Fit indices and va riance explained for the longi tudinal mediati on nested models AIC: Akaike Information Criterion; BIC: Bayesian Information Criterion. Table 4-8. Fixed and random estimates fo r the final longitudinal mediation model Fixed Effect Estimate S.E df t p Ra ndom Effect Estimate S.E Wald Z p Level 2 (Between Subject Effects) Memory mean -0.070.04679.81-1.690.092 Reasoning mean 0.110.05689.072.280.023 Speed mean 0.090.04658.812.500.013 Everyday cognition mean -0.230.05708.71-4.360.000 Intercept -0.040.03671.89-1.300.194 Between-subject residual 0.330.0312.610.00 Level 1 (Within Subject Effects) Linear time -0.150.051868.39-3.160.002 Linear time 0.030.013.430.00 Quadratic time 0.200.051753.364.010.000 Memory change -0.010.02344.02-0.970.332 Memory change 0.000.010.780.44 Reasoning change 0.000.02430.69-0.310.756 Reasoning change 0.010.010.980.33 Speed change 0.010.01427.590.630.527 Speed change 0.000.000.400.69 Everyday cognition change -0.050.02376.85-2.970. 003 Everyday cognition change 0.000.010.280.78 Within subject residual 0.410.0221.400.00 Memory/Reasoning/Speed/Everyday c ognition change: centered variables
73 Table 4-9. Inferential test for the m ediation of basic cognitive variance in ev eryday function by everyday cognition Fixed Effects 95% Confidence Interval Estimate p Lower Bound Upper Bound Estimate model 5 Estimate model 6 Intercept -0.0340.212-0.088 0.020 Linear -0.1690.001-0.264-0.073 -0.170 -0.155 Quadratic 0.2220.000 0.123 0.321 0.222 0.205 Memory mean -0.1490.000-0.225-0.074 -0.072 -0.071 Reasoning mean 0.0010.974-0.078 0.081 0.105 0.106 Speed mean 0.1210.001 0.051 0.191 0.091 0.089 Memory change -0.0180.256-0.048 0.013 -0.018 -0.015 Reasoning change -0.0090.575-0.039 0.021 -0.009 -0.005 Speed change 0.0120.433-0.017 0.040 0.011 0.009 Memory/Reasoning/Speed/Everyday c ognition change: centered variables Estimate model 5: fixed estimates from model 5, Fixed mean everyday cognition; Estimate model 6: fixed estimates from model 6, Fixed centered everyday cognition
74 A. B. C. D. E. Figure 4-1. Baseline SEM models investi gating the mediation of Basic cognition variance in Everyday function by Everyday cognition
75 Figure 4-2. Longitudinal trajectories for a ll study constructs. Note, negative values of basic speed and everyday function were plotted to facilitate visual comparisons.
76 CHAPTER 5 DISCUSSION Overview of Findings The present study investigated the extent to which bas ic and everyday cognitive abilities uniquely predicted older adults ever yday function (self-reported difficulty carrying out IADL), at a specific point in ti me (cross-sectionally) and over the course of five years (longitudinally). Mo re specifically, however, the present study assessed to what extent everyday cognition would a ccount for the basic cognitive variance in everyday function, thus allowing for the eval uation of the claim that everyday cognition is a more proximal, ecologically valid and parsimonious measure of everyday function. Cross-sectional analyses were conducted on data from the baseline occasion of measurement. This provided us with pre liminary estimates based on the largest possible sample size, and it replicates the conventional cross-sectional approach used in the extant literature in this domain. With variables assessed at the latent level, the bivariate correlation results from the Measurement model suggested that the relationship between the cognitive factors a nd everyday function was small (Table 4-2). This was likely due to the fact that at bas eline, when the mean age of the sample was around 74 years, few individuals endorsed havi ng many IADL difficulties, which is consistent with findings from previous studies citing 80 years as the modal onset of selfreported functional difficulties (Fillenbaum, 1985) Therefore, the restriction in range of the everyday function variable further re stricted its potential covariance with other variables under investigation. Neverthele ss, with basic memory, reasoning and speed alone predicting everyday function, basic memory emerged as a significant unique predictor, and as hypothesized, its variance in everyday function was fully mediated by
77 everyday cognition. At baseline, cognitive predictors accounted for only 2% of the variance in everyday function, which alludes to the likely importance of numerous other determinants of self-reported functional difficulty, including health, emotional, and social factors in addition to the restriction of r ange already mentioned. Overall, these findings replicated those of prior studies albeit with a much lower predictive effect size (Willis et al., 1995, Allaire & Marsiske, 2002). The main focus of the present study we re the analyses of the covariance of study constructs over time, which if supported, would facilitate more informed conclusions regarding meaningful interrelationships. Ov er time, across participants all study constructs evidenced significant linear positive trends, indicative of gain. As the plots in Figure 4-2 suggest, this trend was evidenced over the first three-to-four occasions of measurement, and likely represents practice effects in the cognitive measured constructs. The positive linear trend was considerably smaller for everyday cognition, suggesting that these measures we re less susceptible to practice effects. In the case of everyday function, the small initial gain obs erved might have reflected a response bias related to study involvement. For example, since everyday cognitive skills were tested at each occasion, it is possible that having the opportunity to perform these tasks favorably biased responses regarding difficult y of performing them in everyday life. Furthermore, the significant random effects of this linear trend suggested that these effects varied in magnitude bet ween participants. In contra st, the quadratic trend, which was evident beginning around year 3 post base line and was indicative of decline in study constructs, did not differ in magnitude between individuals. In sum, these findings suggested that practice effects were in evid ence as long as 3 to 4 years post study
78 onset, thus decline, if present during th is time, might have been overshadowed until participants reached an age where decrement in performance was substantial enough to outweigh these practice effects. With respect to the relationship between study constructs over time, the present analyses assessed the extent to which basic and everyday cognition might account for the linear gain and quadratic decline in everyday function, and conjointly the extent to which everyday cognition might account for the effects of basic abilities on function over time. These analyses, which examined coupled change over five occasions of assessment, were conducted to answer the ques tion: does decline in cognitive abilities relate to (i.e. translate to) loss in function? To date, no st udy has investigated whether change across cognitive systems was coupled wit h change in function, a pattern that was thus far implied by cross-sectional findings. Results from these analyses suggest that the linear gain, quadratic decline and individual di fferences in linear gain in everyday function can only in part be attributed to cognitive factors. Specifically, about 42% of the individual differences in the in itial gain can be attributed to cognition. A number of other interesting findings emerged from these l ongitudinal analyses. First, basic processing speed emerged as a si gnificant unique predictor of everyday function, in contrast to findi ngs at the cross-sectional leve l. These results augment the findings from existing longitudi nal studies, many of which did not examine speed as a separate predictor of function (e.g. Roya ll et al., 2004, 2005a, 200 5b; Barberger-Gateau et al., 1999; Tomaszewski Farias et al., 2008). Second, everyday cognition accounted for the effects of basic memory in everyday function, suggesting that memo ry abilities are utilized in solving everyday problems,
79 such as looking up a phone number or balancing the checkbook (indeed they explained 80% of the variance in everyday cognition at baseline), which in turn affects the way older adults perceive their ability to solve such problems in their everyday life. Third, basic reasoning showed evidence of a suppressor effect when in a model with everyday cognition, and emerged as a si gnificant negative predictor of function when everyday cognition was included as a predict or in the model. This was in contrast to the bivariate relationship between reasoning and function, which was positive (better reasoning associated with better function). This outcome suggests that everyday cognition was able to isolate and remove a specific type of shared variance in reasoning, thus purifyi ng a small, residual effect of reasoning that was associated with an increase in functional difficulty. While pur ely speculative, this residual variance in reasoning, not accounted for by memory, s peed, or everyday cognition, might be attributed to the cost-tradeoff of what it takes to be disproportionately good at reasoning, such as lack of physical pursu its, which would negatively affect daily functioning. Since reasoning abil ities are frequently correlat ed with levels of education and intellectual pursuits, it might be concluded that individuals who are high on these factors and therefore high on reasoning, mi ght be less likely to engage in physical activity, which would independently predispose t hem to functional difficulties over time. Finally, results suggest that everyday c ognition independently contributed to the prediction of both individual diffe rences in level as well as change in everyday function, which was not the case for the basic cognitive abilities. These results strongly support the predictive validity of everyday cogniti on as a predictor of everyday function.
80 Limitations The present study has a num ber of limitations. The sa mple under investigation was specifically and meticulous ly selected for participation in a larger longitudinal cognitive intervention study. The exclusion cr iteria, which are detailed in the methods section, were implemented to ensure parti cipants capability to comply with training regimens, and as much as possible, retent ion of participants throughout the course of the study. Therefore, participant s in the current study sample were likely to have less health-related, cognitive, and functional problems than older adults in the general population, and very possibly, participants in mo st other studies. This selection bias was likely the major reason behind the unusually sma ll proportion of variance in everyday function that could be attributed to cogniti on (basic and everyday) at baseline, which was 2%, relative to the 20% that is typically reported (Royall et al., 2007). Furthermore, selection biases extend to t he effects of attrit ion. That is, the characterization of attrition effect suggest ed that attrition was selective, for younger, more educated individuals with higher baseli ne global cognitive scores (MMSE-based). These biases have not yet been explored in the analyses. Instead, the present analyses treated missing data as though it were missing at random (MAR), however this was not directly investigated. Future research wit h this sample will explore two alternative strategies: (a) covariate dependent attrition (i.e., are there changes in parameter estimates and standard errors when covariates predictive of attrition are controlled for?) and (b) pattern mixture models (i.e., where separate estimates are derived for groups with different missing data patterns, and then the final estimates are pooled across these missing data subgroups) (Littl e, 1993; Hedeker & Gibbons, 1997),
81 Due to the magnitude of t he data-collection venture (i .e. 2802 participants retested 5 times at different sites), so me of the testing occurred in a study-optimzied format (e.g. in groups, written versus verbal responding in memory tests). While this has been the normative approach for testing in this literatur e, from a neuropsychological perspective it is difficult to speculate on the effects of these testing approaches to the pattern of individual differences observed. If, for example, non-optima l testing introduced noise to our ability to reliably capture individual differences in basic skills, this would have attenuated some of the relationships observed. The psychometrics of the individual estimate s revealed that for some of the basic cognition measures (especially memory) reli ability tended to be fairly low (as low as 0.60 for some measures). The everyday cognition measures tended to have higher reliabilities (between 0.75 and 0.80). This is re levant to the current findings, because it is not clear that the better predictive perform ance of everyday cognition with regard to everyday function is not, at least in part, due to the superior reliability of these everyday cognition measures. From a different perspective, the higher reliability of these everyday cognition measures suggests that (a) despite their relative novelty to the field, they provide good measurement of individual differenc es, and (b) their superior psychometric properties may more strongly argue for their inclusion in clinical assessment batteries. A large sample size (N=698), as well as the (unprecedented) five occasions of measurement might be regarded as a strength in this study. However, it appears that even with a sample this large the analyse s nonetheless lacked the power to detect certain types of individual differences, particularly person-to-person variability in the rate
82 of quadratic decline. Furthermore, it appears t hat 5 years post baseline in a sample of healthy older adults with a mean age of 74 year s may not be sufficient to detect real decline. Therefore, analyses of data collected 10 years after baseline may be more illuminating as to the effects of interest. Moreover, on theoretical grounds, even with 10 predict ors the longitudinal MLM presented in Table 4-7 represents an under-s pecified model. That is, to adequately explain the variance in everyday function, covariates such as age, health status, sensory functioning, mood, and others s hould be included to help distill the unique effects of cognition. Conclusions The present findings have a number of impo rtant implications. First, results from this study highlight the importance of longi tudinal assess ments with multiple occasions of measurement. As findings from Aim 2.1 i llustrate, the first 3 years post initial assessment were characterized by practice effects in the cognitive tasks and response biases in the self-reported measure. Crosssectional analyses using different cohorts and studies with only fewer occasions of m easurement likely fail to capture these effects, which might lead to misinterpretations of the trajectory of age-related cognitive and functional decline. Second, the present study illustrates the differing effects of when in the lifespan older adults are assessed, which was highlight ed by the shift from growth to decline around the 3rd year post baseline. What basic abilities might fail to capture, are the effects of more crystallized aspects of kno wledge, such as previous domain-specific experience with everyday tasks. This exper ience may render everyday abilities less
83 susceptible to decline earlier in the lifespan (i .e. prior to 78-80 years of age in healthy older adults). Fourth, this study illustrates that every day cognition contributes something unique to the prediction of everyday function, above and beyond what might be captured by measures of basic abilities, and appears to be a more proximal measure of everyday function thus supporting its predictive validity. This finding suggests that everyday cognitive measures might represent a better, more ecologically valid approach to the clinical assessment of older adults everyday function than traditional neuropsychological measures. Given the pres ent findings of coupled change over time between everyday cognition and everyday functi on, everyday cognition may also be a promising locus for intervention efforts aimed at preventing functional decline. Finally, the present findings of coupl ed change over time between everyday cognition and everyday function, as well as the significant associations between mean level of everyday cognition and everyday function over time, support the clinical utility of everyday cognitive measures. Specifically, evidence of coupled change over time supports the use of these measures as c linical assessments in which change in performance on these everyday problem solv ing measures might signal concurrent changes in everyday functioning. Thus, a clinician might administer everyday cognitive measures at each annual appoint ment as a means of monito ring the older patient for signs that changes in everyday functioning are imminent. Thus, the everyday cognition measures serve as proxies for in situ observations of the older adult. According to the present findings, everyday cognitive measures might be less suitable as proxies of current functioning in cases where little functional impairment
84 exists. Our baseline data found little associ ation between individual differences in everyday cognition and rated functional difficulty. Notably, as the sample aged and individual differences in level of everyday functioning widened, everyday cognitive measures also showed promise as singleoccasion clinical assessments of current functioning, as suggested by significant associations between mean levels of everyday cognition and everyday function over time. Future Directions Being the first of its kind (i.e., to ex amine five-year concurrent change in basic and everyday cognition, as well as ev eryday function), this study serves as a good foundation for future studies aimed at optimizing the asse ssment of older adults, and creating and delivering interventions to im prove older adults everyday functioning. Future studies would greatly enhance our under standing of the trajectory of cognitive and functional decline, as well as of the relationships between basic and everyday cognitive and functional domains through replication and extension of these findings with longer longitudinal sequences, broader assessments of bas ic cognition via larger theoretically derived batteries, and measures of function less prone to ceiling effects in samples of healthy older adults. The five year longitudinal period in this st udy is useful (in part because it seems to have captured a normative transition from young-old to old-old age, an attendant accelerated decline in cognition during the late r period of study. As Schaies (1994) and others work has shown, however, the time c ourse for late life cognitive decline is much more protracted, and appears to become manifest in late midlife and then extend for the next three-to-four decades. Thus, introduction of everyday cognition measures into longer-term sequences would provide a fuller picture of whether everyday cognition
85 declines at a slower rate than basic cognition, and whether it serves as a useful sentinel measure for functional decline. With regard to broader assessments of basic function, the current study was limited to memory, reasoning and speed because these were the clinical endpoints of the ACTIVE intervention. A broader neuropsychological battery aimed at assessing functional changes in cognition would also measure aspects of executive function, subtypes of memory (including verbal and nonverbal, attention, episodic and semantic, working memory versus secondary memory, et c.). In addition, following conventional approaches in cognitive aging research, most of the measures were speeded (confounding ability variance with speed varianc e). Ideally, tests would be given under power/untimed conditions, to provide purer esti mates of individual abilities. So too, following neuropsychological assessment pr actice, individualized testing in a one-onone situation may have provi ded purer estimates of indi viduals actual ability, unconfounded by environmental distraction and insufficient monitoring of testing performance. Lastly, it is vexing that the chief outcome in this study is a self-reported measure of functional limitation, prone to both ceiling effects and self-perception biases. As the main outcome of interest in most ger ontological research, a more fully-rounded functional assessment (including multiple me asures of self-report with a broader range of outcomes; proxy-asse ssment where possible, and possibly even behavioral observation) would be ideal. Th is would yield a functional composite score of maximal reliability and sensitivity to individual differences.
86 APPENDIX A EVERYDAY PROBLEMS TEST
87 APPENDIX B OBSERVED TASKS OF DAILY LIVING
88 APPENDIX C TIMED INSTRUMENTAL ACTIVI TES OF DAILY LIVING
89 APPENDIX D MINIMUM DATA SET
91 LIST OF REFERENCES Allaire, J. C ., & Marsiske, M. (1999). Everyday cogniti on: age and intellectual ability correlates. Psychology and Aging 14(4), 627-44. Allaire, J. C., & Marsiske, M. (2002). Welland ill-defin ed measures of everyday cognition: relationship to older adults' in tellectual ability and functional status. Psychology and Aging 17(1), 101-15. Allaire, J. C., & Willis, S. L. (2006). Competence in everyday activities as a predictor of cognitive risk and mortality. Neuropsychology, Development, and Cognition. Section B, Aging, Neuropsychology and Cognition, 13(2), 207-224. doi: 10.1080/13825580490904228. Arbuckle, J. L. (2008). AMOS 17.0 Users Guide Chicago, IL: SPSS Inc. Baird, A., Podell, K., Lovell, M., & McGinty, S. B. (2001). Complex real-world functioning and neuropsychological test per formance in older adults. The Clinical Neuropsychologist 15(3), 369-379. Ball, K., Berch, D. B., Helmers, K. F., Jobe, J. B., Leveck, M. D. Marsiske, M., et al. (2002). Effects of cognitive training interventions with older adults: a randomized controlled trial. JAMA: The Journal of the Am erican Medical Association 288 (18), 2271-81. Ball, K., Owsley, C., Sloane, M. E., Roenker, D. L., & Br uni, J. R. (1993). Visual attention problems as a pr edictor of vehicle cras hes in older drivers. Investigative Ophthalmology & Visual Science, 34, 3110. Baltes, M. M. & Baltes, P. B. (1986). The psychology of control and aging. Hillsdale, NJ: Erlbaum. Baltes, M. M., Mayer, U., Borchelt, M., Maas, L., & Wilm s, H.-U. (1993). Everyday competence in old age and very old age: An inter-disciplinary perspective. Ageing and Society, 13, 657-680. Baltes, M. M., Wahl, H. W., & Schmid-Furstoss, U. (1990). The daily life of elderly Germans: Activity patterns, perso nal control, and functional health. Journal of Gerontology: Psychological Sciences, 45 P173-P179. Barberger-Gateau, P., Chaslerie, A., Dart igues, J., Commenges, D., Salamon, R., & Gagnon, M. (1992). Health measures corre lates in a french elderly community population: The Paquid study. Journal of Gerontology 47(2), S88-97. Barberger-Gateau, P., Fabri goule, C., Rouch, I., Letenneur L., & Dartigues, J. F. (1999). Neuropsychological correlates of self-reported performance in instrumental activities of daily living and prediction of dementia. The Journals of Gerontology. Series B, Psychological Sciences and Social Sciences 54(5), P293-303.
92 Baron, R. M., & Kenny, D. A (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic and statistical considerations. Journal of Personality and Social Psychology, 51 1173-1182. Bentler, P. M. and C. P. Chou (1987). Practical issues in structural modeling. Sociological Methods and Research, 16 (1), 78-117. Blessed, G., Tomlinson, B. E. & Roth, M. (1968). The associati on between quantitative measures of dementia and senile change in the cerebral grey matter of elderly subjects. British Journal of Psychiatry, 114, 797-811. Blom, G. (1958). Statistical estimates and transformed beta variables New York, NY: John Wiley & Associates. Brandt, J. (1991). The Hopkins Verbal Learni ng Test: Development of a new memory test with six equivalent forms. Clinical Neuropsychology 5, 125. Bryk, A. S., & Rauden bush, S. W. (1992). Hierarchical linear models: Applications and data analysis methods. Thousand Oaks, CA: Sage Publications. Burton, C. L., Strauss, E., Hultsch, D. F., & Hunter, M. A. (2006). Cognitive functioning and everyday problem solving in older adults. The Clinical Neuropsychologist 20(3), 432-52. Cahn-Weiner, D. A., Malloy, P. F., Boyle, P. A., Marran, M., & Salloway, S. (2000). Prediction of functional status from neuropsychological tests in communitydwelling elderly individuals. The Clinical Neuropsychologist 14(2), 187-195. Campion, E. W. (1994). The oldest old. The New England Journal of Medicine 330 (25), 1819-1820. Caron, J. (1996). The role of cognition, depression, and awareness of deficit in predicting geriatric rehabilitation patients AD L and IADL performance Unpublished doctoral dissertation, University of Detroit, MI. Chaytor, N., & Schmitter-Edgecombe, M. (2003). The ecolog ical validity of neuropsychological tests: a review of the literature on everyday cognitive skills. Neuropsychology Review 13(4), 181-197. Cohen, M. (1992). A power primer. Psychological Bulletin, 112 (1), 155-159. Cornelius, S. W., & Caspi, A. (1987). Ev eryday problem solving in adulthood and old age. Psychology and Aging 2 (2), 144-153. Denney, N. W. (1989). Everyday problem so lving: Methodological issues, research findings, and a model. In L. W. Poon, D. C. Rubin, & B. A. Wilson (Eds.), Everyday cognition in adulthood and late life (pp.30-351). New York: Cambridge University Press.
93 Diehl, M. (1998). Everyday competence in later life: current status and future directions. The Gerontologist 38(4), 422-433. Diehl, M., Marsiske, M., Horgas A. L., Saczynski, J., Rosenb erg, A., & Willis, S. L. (2005). The Revised Observed Tasks of Daily Living: A performance-based assessment of everyday problem solving in older adult s. Journal of Applied Gerontology, 24, 211-230. Diehl, M., Willis, S. L., & Schaie, K. W. (1995) Everyday problem solving in older adults: observational assessment and cognitive correlates. Psychology and Aging 10(3), 478-491. Duffy, M. E., & MacDonald, E. (1990). Dete rminants of functiona l health of older persons. The Gerontologist 30(4), 503-509. Educational Testing Service. (1977). Basic Skills Assessment Test: Reading Princeton, NJ: Author. Ekstrom, R. B., French, J. W., Harman, H., & De rman, D. (1976). Kit of FactorReferenced Cognitive Tests rev. ed. Princeton, NJ: Educat ional Testing Service. Farmer, J. E., & Eakman, A. M. (1995). The relationship between neuropsychological functioning and instrumental activities of da ily living following acquired brain injury. Applied Neuropsychology 2 (3-4), 107-115. Ferrucci, L., Guralnik, J. M., Baroni, A., Tesi, G., Antonini E., & Marchionni, N. (1991). Value of combined assessment of physical health and functional status in community-dwelling aged: a prospective study in Florence, Italy. Journal of Gerontology 46(2), M52-56. Fillenbaum, G. G. (1985). Screening the elderly. A brief inst rumental activities of daily living measure. Journal of the American Geriatrics Society 33(10), 698-706. Folstein, M. F., Folstein, S. E., & McHugh, P.R. (1975). Mini-mental state: A practical method for grading the cognitive stat e of patients for the clinician. Journal of Psychiatric Research 12,189. Frazen, M. D., & Wilhelm, K. L. (1996). Conceptual foundations of ecological validity in neuropsychological assessment. In R. J. Sbordone & C. J. Long (Eds.), Ecological validity of neuropsychological testing (pp. 91-112). Boca Raton, FL: St Lucie Press. Gonda, J., & Schaie K. W. (1985). Schaie-Thurstone Mental Abil ities Test: Word Series Test Palo Alto, CA: Consulti ng Psychologists Press. Hedeker, D., & Gibbons, R.D. (1997). Application of random-effects pattern-mixture models for missing data in longitudinal studies. Psychological Methods, 2, 64-78
94 Heaton, R. K., & Pendleton, M. G. (1981). Use of neuropsychological tests to predict adult patients' everyday functioning. Journal of Consulting and Clinical Psychology 49(6), 807-821. Hetzel, L., & Smith (2000). The 65 Years and Over Population: 2000 U.S. Census Bureau, Washington DC. Hu, L. and P. M. Bentler (1999) Cutoff criteria for fit index es in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling 6 (1): 1-55. Hultsch, D. F., Hertzog, C., Sma ll, B. J., & Dixon, R. A. (1 999). Use it or lose it: engaged lifestyle as a buffer of cognitive decline in aging? Psychology and Aging 14(2), 245-263. Jefferson, A. L., Paul, R. H., Ozonoff, A., & Cohen, R. A. (2006). Evaluating elements of executive functioning as predi ctors of instrumental activities of daily living (IADLs). Archives of Clinical Neur opsychology: The Official Journal of the National Ac ademy of Neuropsychologists 21(4), 311-320. Jobe, J. B., Smith, D. M., Ball, K., Tennstedt, S. L., Ma rsiske, M., Willis, S. L., et al. (2001). ACTIVE: a cognitive intervention trial to promote independence in older adults. Controlled Clinical Trials 22(4), 453-79. Katz, S., Ford, A. B., Moskowitz, R. W., Jac kson, B. A., & Jaffe, M. W. (1963). Studies of Illness in the Aged: T he Index of ADL: A Standardiz ed Measure of Biological and Psychosocial Function. JAMA 185 (12), 914-919. Kraemer, H. C., Yesavage, J. A., Taylor, J. L., & Kupfer, D. (2000). How can we learn about developmental processes from cr oss-sectional studies, or can we? The American Journal of Psychiatry 157 (2), 163-171. Lawton, M. P. (1987). Cont extual perspectives: Psychosocial influences. In Handbook for clinical memory assessment of older adults (pp. 22-42). Washington, DC: American Psychological Association. Lawton, M. P., & Brody, E. M. (1969). Assessment of older people: self-maintaining and instrumental activities of daily living. The Gerontologist 9 (3), 179-186. Little, R. (1993), Pa ttern-mixture models for multiv ariate incomplete data, Journal of the American Statistical Association 88, 125. Mangione, C. M., Phillips, R. S. Seddon, J. M., Lawrence, M. G., Cook, E. F., Dailey, R., et al. (1992). Development of the 'Act ivities of Daily Vision Scale': A measure of visual functional status. Medical Care 30(12), 1111-1126. Marsh, H. W., Balla, J. R., & McD onald, R. P. (1988). Goodness of fit indexes in confirmatory factor analysis: The effect of sample size. Psychological Bulletin, 103 391-410.
95 Marsiske, M., Klumb, P., & Ba ltes, M. M. (1997). Everyday activity patterns and sensory functioning in old age. Psychology and Aging 12(3), 444-457. Marsiske, M., & Margre tt, J. A. (2006). Everyday probl em solving and decision making. In J. E. birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (6th ed., pp. 315-342). Amsterdam: Elsevier. Marsiske, M., & Willis, S. L. (1995). Dimensionality of everyday problem solving in older adults. Psychology and Aging 10 (2), 269-283. Martens, M. P. (2005). The use of structural equati on modeling in counseling psychology research. The Counseling Psychologist, 33 269-298. McCue, M., Rogers, J. C., & Goldstei n, G. (1990). Relationships between neuropsychological and functional assessment in elderly neuropsychiatric patients. Rehabilitation Psychology, 35 (2), 91-99. Miller, E. A., & Weissert, W. G. (2000). Predicting elderly people's risk for nursing home placement, hospitalization, functional im pairment, and mortality: a synthesis. Medical Care Research and Review: MCRR 57(3), 259-297. Mitchell, M., & Miller, L. S. (2008). Prediction of functional status in older adults: the ecological validity of f our Delis-Kaplan Executive Function System tests. Journal of Clinical and Experimental Neuropsychology 30(6), 683-690. Moore, D. J., Palmer, B. W. Patterson, T. L., & Jeste, D. V. (2007). A review of performance-based measures of functional living skills. Journal of Psychiatric Research, 41(1-2), 97-118. Morris, J. C. (1993). The Clini cal Dementia Rating (CDR): current version and scoring rules. Neurology 43(11), 2412-2414. Morris, J. N., Fries, B. E., Steel, K., Ikagam i, I., Bernabei, R., Carpenter, G. I., et al. (1997). Comprehensive clinical assessment in community setting: Applicability of the MDS-HC. Journal of the American Geriatric Society 45, 1017. Njegovan, V., Hing, M. M., Mitchell, S. L., & Molnar, F. J. (2001) The hierarchy of functional loss associated with cognitive decline in older persons. The Journals of Gerontology. Series A, Biologic al Sciences and Medical Sciences 56(10), M638643. Owsley, C., Sloane, M., McGwin, G., & Ball, K. (2002). Timed instrumental activities of daily living tasks: relationship to cognitive function and everyday performance assessments in older adults. Gerontology 48(4), 254-65.
96 Prs, K., Helmer, C., Amieva, H., Orgogozo, J., Rouch, I., Dartigues, J., et al. (2008). Natural history of decline in instrumental activities of daily living performance over the 10 years preceding the clinical di agnosis of dementia: a prospective population-based study. Journal of the American Geriatrics Society 56(1), 37-44. Poon, L. W., Rubin, D. C ., & Wilson, B. A. (1989). Everyday cognition in adulthood and late life. New York: Cambridge University Press. Raz, N., Lindenberger, U., Rodrigue, K. M., Kennedy, K. M., Head, D., Williamson, A., et al. (2005). Regional Brain Changes in Aging Healthy Adults: General Trends, Individual Differences and Modifiers. Cereb. Cortex 15 (11), 1676-1689. doi: 10.1093/cercor/bhi044. Rey, A. (1941) Lexamen psychologique dans les cas dencephalo pathie tramatique. Archives de Psychologie 28,21. Royall, D. R., Lauterbach, E. C. Kaufer, D., Malloy, P., Cobur n, K. L., & Black, K. J. (2007). The cognitive correlates of functional status: a review from the Committee on Research of the American Ne uropsychiatric Association. The Journal of Neuropsychiatry and Clinical Neurosciences 19(3), 249-265. Royall, D. R., Palmer, R., Ch iodo, L. K., & Polk, M. J. (2004). Declining executive control in normal aging predicts change in functional status: the Freedom House Study. Journal of the American Geriatrics Society 52(3), 346-352. Royall, D. R., Palmer, R., Chiodo, L. K ., & Polk, M. J. (2005a). Executive control mediates memory's association with change in instrumental activities of daily living: the Freedom House Study. Journal of the American Geriatrics Society 53(1), 11-17. Royall, D. R., Palmer, R., Chi odo, L. K., & Polk, M. J. ( 2005b). Normal rates of cognitive change in successful aging: the freedom house study. Journal of the International Neuropsychological Society: JINS 11(7), 899-909. Salthouse, T. A. (1990). Cogniti ve competence and expertise in the aging. In J. E. Birren & k. W. Schaie (Eds.), Handbook of the psychology of aging (3rd Ed., pp310-319). New York: Academic Press. Sbordone, R. J. (1996). Ecol ogical validity: Some critical issues for the neuropsychologist. In R. J. S bordone, & C. J. Long (Eds.), Ecological validity of neuropsychological testing ( pp. 15-41). Boca Raton, FL: St Lucie Press. Sbordone, R. J. (1997). The ecol ogical validity of neuropsycholog ical testing. In A. M. Wedding, & J. Webster (Eds.), Ecological validity of neuropsychological testing, (pp. 365392). Delray Beach, FL: GR Press/St. Lucie Press. Schumacker, R. E. & Lomax R. G. (2004). A beginner's guide to structural equation modeling, (2nd ed). Mahwah, NJ: Lawrenc e Erlbaum Associates.
97 Schaie, K. W. (1965). A general model for the study of developmental problems. Psychological Bulletin, 64, 92-107. Schaie, K. W. (1994). The course of adult intellectual development. American Psychologist 49(4), 304-313. doi: doi:10. 1037/0003-066X.49. 4.304. Schaie, K. W. (1996). Intellectual Development in Adul thood. The Seattle Longitudinal Study New York: Cambridge University Press. Schaie, K. W. and Willis, S. L. (1986). Can decline in adult intellectual functioning be reversed? Developmental Psychology, 22, 223-232. Singer, J. D., & Willett, J. B. (2003). Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence London: Oxford University Press. Spooner, D. M., & Pachana, N. A. (2006). Ecological validity in neuropsychological assessment: a case for greater consideratio n in research with neurologically intact populations. Archives of Clinical Neuropsychology : The Official Journal of the National Academy of Neuropsychologists 21(4), 327-337. Sternberg, R. J., & Wagner, R. K. (1986). Practical intelligence: Nature and origins of competence in the everyday world. Cambridge: Cambridge University Press. Tan, J. E., Hultsch, D. F., & Strauss, E. (2009). Cognitive abilities and functional capacity in older adults: results from t he modified Scales of Independent BehaviorRevised. The Clinical Neuropsychologist 23(3), 479-500. Thornton, W. J. L., & Dumke, H. A. (2005). Age differences in everyday problem-solving and decision-making effectiveness: a meta-analytic review. Psychology and Aging 20(1), 85-99. Thornton, W. L., Deria, S., Gelb, S., Shapiro, R. J., & Hill, A. (2007). Neuropsychological mediators of the links among age, chronic illness, and everyday problem solving. Psychology and Aging 22(3), 470-81. Thurstone, L. L., & T hurstone, T. G. (1949). Examiner Manual for the SRA Primary Mental Abilities Test (Form 10-14) Chicago: Science Research Associates. Tomaszewski Farias, S., Cahn-We iner, D. A., Harvey, D. J. Reed, B. R., Mungas, D., Kramer, J. H., et al. (2009). Longitudi nal changes in memory and executive functioning are associated with longitudinal change in instrumental activities of daily living in older adults. The Clinical Neuropsychologist 23(3), 446-461. Weatherbee, S. R., & Allair e, J. C. (2008). Everyday cognition and mortality: performance differences and predictive utilit y of the Everyday Cognition Battery. Psychology and Aging 23(1), 216-21.
98 Weston, R., & Gore, P.A. (2006). A brief gui de to structural equation modeling. The Counseling Psychologist, 34 719-751 Wilson, B. A., Cockburn, J., Baddeley, A. (1985) The Rivermead Behavioral Memory Test. Reading, England: Tham es Valley Test Co.; Gaylord, MI: National Rehabilitation Services. Willis, S. L. (1991). Cognition and ever yday competence. In K.W. Schaie (Ed.), Annual review of gerontology and geriatrics (Vol. 11, pp. 80-109). New York: Springer. Willis, S. L. (1996). Everyday cognitive competence in elderly persons: conceptual issues and empirical findings. The Gerontologist 36(5), 595-601. Willis, S. L., Allen-Burge, R., Do lan, M. M., Bertrand, R. M., Yesavage, J., & Taylor, J. L. (1998). Everyday problem solving among in dividuals with Alzheimer's disease. The Gerontologist 38(5), 569-577. Willis, S. L., Jay, G. M., Diehl, M., & Marsiske, M. (1992). Longitudinal Change and Prediction of Everyday Task Competence in the Elderly. Research on Aging 14(1), 68-91. Willis, S. L., & Marsiske, M. (1991). Life-span perspective on practical intelligence. In D. E. Tupper & K. D. Cicerone ( Eds.). The neuropsychology of everyday life: Issues in development and rehabilitation (pp. 183-198). Boston: Kluwer. Willis, S. L., & Marsiske, M. (1993). Manual for the Everyday Problems Test University Park, PA: Pennsylvania State University. Willis, S. L., & Schaie, K. W. (1993). Everyday cognition: Taxonomic and methodological considerations. In J. M. Puckett & H. W. Reese (Eds.) Mechanisms of everyday cognition (pp. 33-54). Hillsdale, NJ: Lawrence Erlbaum. Willis, S. L., Tennstedt, S. L., Marsiske, M., Ball, K., Elias, J., Koepke, K. M., et al. (2006). Long-term effects of cognitive traini ng on everyday functional outcomes in older adults. JAMA: The Journal of the American Medical Association 296 (23), 2805-14. Willis, S. L., Jay, G. M., Diehl, M., & Marsiske, M. (1992). Longitudinal Change and Prediction of Everyday Task Competence in the Elderly. Research on Aging 14(1), 68-91. Wolinsky, F. D., Callahan, C. M ., Fitzgerald, J. F., & Johns on, R. J. (1993). Changes in functional status and the risks of subs equent nursing home placement and death. Journal of Gerontology 48(3), S94-101. Wolinsky, F. D., Coe, R. M., Miller, D. K., Prendergast, J. M. Creel, M. J., & Chvez, M. N. (1983). Health services utilizati on among the noninstitutionalized elderly. Journal of Health and Social Behavior 24(4), 325-337.
99 Wood, K. M., Edwards, J. D., Clay, O. J., Wadley, V. G., R oenker, D. L., & Ball, K. K. (2005). Sensory and cognitive factors influenc ing functional ability in older adults. Gerontology 51(2), 131-141.
100 BIOGRAPHICAL SKETCH Anna Yam graduated Cum Laude from the Un iversity of California, San Diego (UCSD) wit h bachelors degrees in Hum an Development and Linguistics. After graduating from UCSD and until her acceptance into the docto ral program in Clinical and Health Psychology at the University of Florida in 2008, Ms. Yam has worked as a Research Assistant at the Laboratory for Cogni tive Neuroscience, at the Salk Institute for Biological Studies, studying the genetic underpinnings, brain structure and function, cognition, and behavior of individuals with Wil liams syndrome. At the University of Florida, Ms. Yam is currently pursuing her doctorate in clinical psychology, with a specialization in neuropsychology. Her research interests and focus is in cognitive aging and interventions to improve the everyday lives of older adults.