! EXECUTIVE FUNCTION , HEALTH BEHAVIOR, AND CARDIOVASCULAR MORTALITY RISK b y TIMOTHY BRUNELLE B.S., University of Colorado Denver , 2007 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Arts Clinical Health Psychology Program 2017 11 01
"" ! ! ! ! ! ! ! ! This thesis for the Master of Arts degree by Timothy Brunelle Has been approved for the Clinical Health Psychology Program By James Grigsby, Chair Kevin Masters Krista Ranby Date: December 16, 2017
""" ! ! ! ! ! Brunelle , Timothy (M.A., Clinical Health Psychology Program ) Executive Function, Health Behavior, and Cardiovascular Mortality Risk Thesis directed by Professor James Grigsby. ABSTRACT The finding that cognitive impairment is a risk factor for mortality risk has been previously established in the cognitive epidemiology literature. However, the use of non specific measures of cognition (mental status or IQ) and non specific mortality outcom e variable s (all cause mortality) has made it difficult to hypothesize about specific variables that may mediate these relationships. Hispanic and Non Hispanic White, elderly (> 60 years old) participants (N = 706) of the San Luis Valley Health and Aging Study were included in an analysis that investigated the relationship between a specific domain of cognitive functioning ( executive function ) (EF) and a specific cause of death variable ( cardiovascular mortality ) (CVM) . We found that EF is a significant predictor of CVM risk in a binary logistic regression model that included age, education, ethnicity, sex, and comorbidities (Exp(B) = .878, 95% CI = .824 to .935) and higher EF was associated with a decreased likelihood of dying from CVD. This model was similar to a model that included General Mental Status (MMSE) as the predictor variable (Exp(B) = .895, 95% CI = .846 to .946) . We did not find clear support for our hypotheses that level of engagement in health behaviors (physical activity , diet, and smoking behavior) would function as mediators of this rela tionship . Our conceptualization of EF suggests that the construct may be more related to ability to engage in behavior change, which was likely not sufficiently captured based on how we operationalized our proposed mediator
"# ! ! ! ! ! ! ! ! variables . This limitation may have additionally affected our ability to demonstrate how EF and Mental status are dissociable cognitive constructs. Future investigations of the relationship between cognitive functi oning and mortality would benefit from addressing this and some of the other methodological limitations identified in our discussion . Con siderations for a prospective experimental protocol that would better enable us to study this relationship are offered. The form and content of this abstract are approved. I recommend its publication. Approved: James Grigsby
# ! ! ! ! ! ! ! ! TABLE OF CONTENTS CHAPTER I. BACKGROUND Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰.1 Introduction and Significance ................................ ................................ .................. 1 Measurement of Cognition ................................ ................................ ...................... 3 Measurement of Mortality ................................ ................................ ....................... 7 Objectives ................................ ................................ ................................ .............. 11 Hypotheses ................................ ................................ ................................ ............ 1 2 II. METHOD ..Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰15 Introduction ................................ ................................ ................................ ......... 1 5 Sampling ................................ ................................ ................................ .............. 1 5 Procedures ................................ ................................ ................................ ........... 1 7 Measures ................................ ................................ ................................ ............... 18 Data Analysis ................................ ................................ ................................ ........ 2 3 III. RESULTS .. Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰ Ã‰ 25 Descriptive Statistics ................................ ................................ ............................ 25 Evaluation of Assumptions ................................ ................................ ................... 28 Correlations Among Variables ................................ ................................ ............. 28 Evaluation of Objectives ................................ ................................ ...................... 30 IV. DISCUSSION ...Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰ 34 Implications ................................ ................................ ................................ .......... 34 L imitations ................................ ................................ ................................ ............ 37
#" ! ! ! ! ! ! ! ! Future Directions ................................ ................................ ............................. Ã‰. 42 Conclusions . Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰. .. 4 5 REFERENCES Ã‰ Ã‰Ã‰Ã‰ Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰.. 47 APPENDIX Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰Ã‰ .. Ã‰. 54
#"" ! ! ! ! ! LIST OF TABLES TABLE 1. Items on the Behavioral Dyscontrol Scale ................................ .............................. 6 2. International Classification of Disease (ICD 9 and ICD 10) codes for determining underlying cause of death due to cardiovascular disease .................. 19 3. Number of participants with comorbid health conditions by mortality status ..... 23 4. Descriptive statistics for proposed IV, DV, Mediators and Covariates ................ 26 5. Evaluation of Ethnic Group Differences . ................................ ............................. 27 6. Bivariate Correla tions Among Independent, Dependent , Mediator, and Covariates Variables ................................ ................................ ................................ ................ 2 9 7. Summary of Logistic Regression Analyses for BDS and Covariates Predicting Cardiovascular Mortality Risk ................................ ................................ ............. 30 8. Summary of Logistic Reg ression Analyses for MMSE and Covariates Predicting Cardiovascular Mortality Risk ................................ ................................ ............. 3 3
#""" ! ! ! ! ! LIST OF FIGURES FIGURE 1. Hypothesized models of how physical activity, vegetable consumption, and smoking behavior may mediate the relationship between executive functioning and cardiovascular mortality risk ................................ ................................ .......... 1 4 2. San Luis Valley Health and Aging Study Flow Chart ................................ ......... 1 7
$ ! ! ! ! ! CHAPTER I BACKGROUND Introduction and Significance IQ tests have bee n in existence since the early 1900s but active investigation of IQ score as a risk factor for health related outcomes, including mortality, is a relatively new area of research. Beginning in the mid 1990s, studies establishing an association between general cognitive ability ( and mental status ) and mortality risk started to appear regularly in the literature and the term Cognitive Epidemiology was coined for this line of research (Deary & Batty, 2007). It is likely that cognitive deficits (or cognitive decline) observed at different points in an individual's life can interact with or be caused by any number of other factors to influence one's risk of death. A number of different hypotheses have been proposed in an attempt to understand the nature of these relationships . It has been suggested that measures of general cognitive ability (e.g., IQ) reflect efficiency of information processing, an d that cognitive impairment may be the result of suboptimal brain development, (reflecting factors such as a deficient prenatal environment), which in turn confers an increased risk of mortality (Whalley, 2000; Snowdo n, 1996). Others have suggested that b etter cognitive abilities may allow an individual to attain higher levels of education and social class, both of which are factors associated with better health and decreased risk of mortality (Niesser et al., 1996, Smith et al., 1998). A third hypothesis is that cognition is positively associated with the number of healthy behaviors in which a person engages, including healthy eating, exercise, not smoking, and consuming no more than moderate amounts of alcohol (Friedman et al., 1995; Hart et al., 2003; S hipley et al.,
% ! ! ! ! ! 2006). These hypotheses are not mutually exclusive, and all may reflect pathways that mediate the relationship between cognitive ability and mortality risk . The previous studies in this field can be grossly divided into two separate lines of research. The first line includes studies that have investigated cognitive ability (usually IQ ) in children and have found a n association with mortality risk , where higher scores seem to confer a decreased risk of early death (Otoole & Stankov, 1992; Whalley & Dreary, 2001; Hart et al., 2003). These studies are ostensibly capturing premorbid levels of cognitive functioning that do not reflect the effects of chronic disease comorbidity. The second line of research, conducted with adults and the elderly has also demonstrated that cognitive impairment or decline measured at various points across the adult lifespan may also be associated with an increased risk of mortality (Wilson et al., 2003, Wilson et al., 2007, Bassuk et al., 2000; Schupf et al. , 2005; Gale et al., 1996; Gusekloo et al., 1997; Nguyen et al., 2003; Kelman et al., 1994; Neale et al., 2001; Korten et al., 1997; Berg, 1996). Given that chronic disease comorbidity (e.g. , diabetes and cardiovascular disease) becomes more prevalent with older age, research conducted with older individuals suggests that the effect s of chronic disease must be considered to understand how cognitive function is related to mortality risk. While aging is an obvious risk factor for mortality , and is associated with decline in some cognitive abilities (speed of processing, working memory, and delayed recall) (Park et al., 2002; Salthouse, 2004), the confounding of aging with chronic disease comorbidity makes it difficult to distinguish which de ficits are caused by pathological disease processes and which might be the result of ag ing per se. Identifying a completely "healthy" elderly population is extremely difficult as these individuals are the rare
& ! ! ! ! ! exceptions rather than the norm. It follows that studying "normal" aging necessitates the inclusion of chronic diseases. Epidemiological research suggests that 85% of individuals over 6 5 have at least one comorbid condition and 62% have two or more (Anders on & Horvath , 2004) . As the elderly populati on continues to increase in our country, u nderstanding the relationship between cognitive dysfunction, chronic disease, and mortality is important from a public health standpoint. Measurement of C ognition The development of brief, general cognitive screening instruments that reliably distinguish normal persons from those who are impaired has made the assessment of cogniti on more efficient and therefore more practical . This is p articularly relevant in epidemiological research studie s , in which researchers often attempt to collect a great deal of information from relatively large samples . Historically, the instrument that has been most frequently utilized for this purpose has been the Mini Mental State Exam (MMSE) (Folstein, 1975) . Due to its brief and relatively simple administration protocol, as well as its ability to reliably distinguish between normal and impaired mental status , it has been widely adopted. However, the MMSE's non specific nature , low ceiling, sensitivity to language and ethnicity ( e.g., Escobar, et al., 1986), and failure to assess certain aspects of cognition make it difficult to study specific mechanisms or pathways that could shed light on the nature of its relationship to mortality . In addition to studie s that have utilize d measures of overall mental status or general cognitive fu n ctioning , other research has suggested that specific domains of cognitive functioning may be differentially associated with mortality risk. Specific domains include processing speed ( Smits et al., 1999; Shipley et al., 2006 ) , memory
' ! ! ! ! ! (Anstey et al., 2001; Hassing 2002) , language and visuospatial reasoning abilities (Schupf et al., 2005) . There is accumulating evidence that executive functioning (EF) may be a particularly relevan t set of cognitive abilities to consider when trying to understand the relationship between cognitive impairment and increased risk of mortality. Just as Amirian et al. (2010) found that EF (and decline in EF) were stronger predictors of mortality compared to MMSE, Johnson et al. (2007) also found that tests of EF may be better predictors of mortality than the MMSE. Similarly, Sabia (2008) found that performance on tests of EF were better predictors of mortality than a general intelligence factor ( "g"). To further our understanding of how EF may function as an important predictor of mortality requires additional elaboration of this complex cognitive construct. While there is no single agreed upon definition of EF, there does appear to be a consensu s that EF is not a unitary cognitive process, but rather a set of relatively dissociable but related functional subcomponents that subserve the underlying common function of enabling an individual to engage in purposeful, goal directed behavior. Specific f unctional subcomponents of EF that contribute to this broad ability include planning, initiating, monitoring, shifting attention, inhibition, problem solving, abstract reasoning, and working memory (Lezak, 2012; Miyake, 2000). A particularly useful conceptualization, emerg ing from the work of Fuster and Luria, broadly describes EF as the capacity for autonomous regulation of one's attention and behavior. A person's ability to regulate her/his own attention and behavior is important for successful in itiation and maintenance of new behavioral repertoires . However, intervention research in health behaviors has taught us that adoption of new behaviors and breaking of bad habits is a challenging process for many individuals .
( ! ! ! ! ! Adoption of new healthy beha viors is an effortful process because it is often in conflict with well established habitual behaviors that we engage in on a rather automatic and non conscious level. Disrupting these automatic processes as well as resisting other negative behaviors and i nfluences requires deliberate, conscious control that is applied in a consistent manner, often over long periods of time (Fuster, 1997; Fuster, 1999; Quintana & Fuster, 1999; Fuster, 2000; Luria, 1980). Well learned tasks may require little or no EF contro l for their activation, but EF is required to establish new habits (e.g., regular exercise), and for disrupting old habits (e.g . , smoking behavior). Fuster (2000) contends that EF involves the integration of working memory (and current intentions held the rein), mental representations of the environment and of a goal, and a prospective "motor short term memory" that supports planning for action to achieve the goal. EF allows intentions to be maintained across time until appropriate behavior is initiated an d carried out to completion. A final component of Fuster's model is the capacity for inhibition, in that one must maintain a specific goal across time ( eating healthier foods ) while resisting impulses to engage in inappropriate or irrelevant behavior (e.g. , eating sweets). For persons with impaired EF , this process breaks down, and an inability to regulate behavior is the result. The Behavioral Dyscontrol Scale (BDS) is a measure that assesses a specific aspect of EF, the capacity for behavioral and attentional self regulation , and repres ents a formalization of Luria's (1980) approach to the assessment of frontal lobe dysfunction , or what is currently thought of as dysexecutive syndrome . The fro ntal lobes are largely involved in the organization of voluntary activity , and it stands to reason that deficits in a person's ability to execute simple motor behaviors may be an indication of frontal lobe
) ! ! ! ! ! (executive) dysfunction. Based on this paradigm, the items on the BDS are made up of items assessing for the conscious control of motor functioning and the ability to use intentions to guide behavior. The BDS has 9 items and a maximum score of 19 points. A summary of the items is provided in Table 1 (G rigsby, 1992) . Table 1 . Items on the Behavioral Dyscontrol Scale 1. Tap twice with the dominant hand and once with the non dominant hand repetitively. 2. Tap twice with the non dominant hand and once with the dominant hand repetitively. 3. Patient squeezes the examiners hand when he says, "red" and does nothing when he says "green." 4. If the examiner taps twice the patient taps once . If the examiner taps once, the patient taps twice. 5. Alternate touching the thumb and each finger of the dominan t hand, in succession to the top of the table. 6. Place the dominant hand in the sequence of positions: fist edge palm. 7. Adaptation of Heads test (Head, 1920)*. 8. Alternate counting with letter of the alphabet through the letter "L." (i.e., 1a2bÃ‰) 9 . Rating of insight into performance * Facing the examiner, the patient is asked to duplicate the position of the examiners hand, using the same hand as the examiner (that is, not mirroring). The capacity for behavioral self regulation is necessary for the effective execution of instrumental activities of daily living (IADLs, i.e., medication management, financial management, meal preparation, etc.), and if compromised, the resulting dysexecuti ve syndrome is likely to impair a person's ability to engage in self care and live independently. Just as Kaye et al. (1990) found EF (BDS) to be a better predictor of functional status when compared to the MMSE, G rigsby et al. (1998) reported that while the MMSE and the BDS both predicted observed performances on simulated IADL tasks such as managing money and medications, EF was more strongly related with IADLs and self reported ADLs than was mental status. These results suggest that general mental
* ! ! ! ! ! statu s and EF contribute differently to the capacity to engage in purposeful activity such as ADLs and IADLs, and that EF is a more important predictor for these activities . In addition to its association with impaired ADLs / IADLs, defective EF would also like ly compromise one's ability to initiate or maintain certain instrumental health behaviors. Impaired EF (as measured by the BDS) has been associated with decreased rates of successful smoking cessation among older individuals (Brega et al., 2008), and with decreased ability to engage in instrumental health behaviors that may lead to increased health care utilization among older type 2 diabetics (Tran et al., 2013). Theoretically, engaging in healthy behaviors and avoiding unhealthy behaviors would be expecte d to mitigate mortality risk. Sabia (2008) found that the relationship between EF and mortality was better accounted for by health behaviors (e.g., smoking status, alcohol consumption, consumption of fruits and vegetables, physical activity) rather than so cioeconomic variables or comorbid health conditions. The se findings suggest that the association between EF and mortality risk may be mediated by engagement in health behaviors. Thus, considered in this light, executive functioning may help bridge the gap in our understanding of the association between impaired cognitive functioning and increased risk of mortality. Measurement of Mortality Similar to the difficulty of interpreting the results of studies that utilize general measures of cognitive functionin g , the use of all cause mortality as the outcome variable of interest is also problematic. It is difficult to speculate with any degree of confidence about specific factors that might mediate the relationship between impairment on a measure of general men tal status and a non specific cause of death. The use of a more
+ ! ! ! ! ! specific cause of death variable (along with a specific domain of cognitive impairment) may help narrow the number of risk factors required to understand the nature of the observed relationship between cognitive impairment and a disease specific mortality risk . We are particularly interested in cardiovascular disease (CVD) mortality as it is the leading cause of death worldwide and is strongly associated with aging. CVD account s for 1 in every 4 deaths in the United States (Centers for Disease Control and Prevention, 2015) as well as 31% of all global deaths (World Health Organization, 2012 ) . CVD is a g eneral term used to describe any number of conditions that adversely affect the functioning of the circulatory system Ã‘ the heart and blood vessels. Conditions that fall under this CVD umbrella include coronary heart disease, cerebrovascular disease, peripheral artery disease, rheumatic heart disease, congenital heart disease, deep vein thrombosis and pulmonary embolism (World Health Organization, 201 4 ). Events ( sometimes fatal) and disorders associated with these conditions include myocardial infarction (MI), stroke, congestive heart failure (CHF), cardiac arrest, arrhythmia , and angina pectoris. Risk factors that have been identified for CVD include " unmodifiable " risk factors ( age, gender, family history ) as well as "modifiable" risk factors , including comorbid health/medical conditions (diabetes, hypertension, high blood cholest erol), psychosocial factors (depression, life stress, type A behavior, social isolation) and lifestyle/behavioral factors (tobacco use, sedentary lifestyle, unhealthy diet, obesity, alcohol use) (Allan & Fisher, 2012; Centers for Disease Control and Preven tion, 2012) . All of the above " modifiable " risk factors are possible targets for intervention , but regardless of which specific targets are selected, i ntervention strategies typically require
, ! ! ! ! ! some level of behavior modification and self regulation . This could include cessation of a bad habit such as smoking , or acquisition of habitual healthy behaviors such as sticking with a nutritional diet or starting and maintaining a regular exercise program. The degree to which someone is able to engage in pro healt h related behaviors should theoretically mitigate the ir risk of dying from CVD . An individual may not engage in such behavior in spite of an intact capacity for EF, but absent EF, s/he is unlikely to initiate health related activities appropriately and independently. A review of some of the individual risk factors for CVD indicate s that many of them are also commonly as sociated with cognitive decline and dysfunction . Hypertension has been associated with cognitive decrements in EF, learning and memory, attention, psychomotor speed, motor function, visual spatial abilities, and visual construction abilities (Elias et al., 2015) . Type 2 diabetes mellitus (T2DM) has been associated with deficits in memory, processing speed, attention, and cognitive flexibi lity (Dey et al., 1997; Ryan & Geckle 2000a; Ryan & Geckle 2000b) as well as an increased risk for the development of dementia (Brands et al., 2015). The relationship between high blood cholesterol and cognitive functioning is less clear. Hypercholesterolemia in midlife has been associated with an increased risk of dementia but conversely, low serum cholesterol after age 60 has also bee n associated with increased risk of dementia. Smoking has been associated with cognitive defi cits in EF, verbal memory, processing speed as well as an increased risk of Alzheimer's disease and vascular dementia (Swan & Lessov Schlaggar, 2015). Interventions that seek to improve cardiovascular fitness, replacing a sedentary lifestyle with increase d physical activity , also seem to be associated with improved cogniti on . The greatest impact of physical activity appears to be in the
$! ! ! ! ! domain of EF , but improvements are also often observed in global cognition and memory (Carlson & Varma, 2015). Dietary guidelines for "healthy eating" have been modified over the years, and the specific foods that might or might not confer an increased risk of cardiovascular disease have been recently questioned. Government recommendations are reviewed every five y ears, and significant changes have been made since the original "Food Pyramid" was developed in 1985. The current guidelines, in place since 2010, will likely undergo significant changes with the current review as there are proposed changes to dietary cho lesterol, fat, and sugar consumption recommendations. However, the increased consumption of fruits and vegetables has remained a consistent recommendation for cardiovascular health. Eating more fruits and vegetables adds nutrients to diets, reduces the ri sk for heart disease, stroke, and some cancers, and helps manage body weight when consumed in place of more energy dense foods (US Department of Agriculture, 2010). Additionally, c onsuming a plant based diet remains a consistent recommendation for maintai ning heart health (Willet et al.,1995; Sofi et al., 2010). The consumption of additional fruits and vegetables has been one of the hallmarks of government food recommendations but has also been one of the recommendation s that seems hardest to follow. A n analysis of the 2007 2010 National Health and Nutrition Examination Survey (NHANES) reported that just 25% of Americans consumed the daily amount of fruit recommended for their age and sex while just 13% of Americans consumed the daily recommended amount of vegetables for their age and sex (National Cancer Institute, 2015). A similar analysis of 2013 data from the Centers for Disease Control and Prevention (CDC) provided slightly lower numbers suggesting that just
$$ ! ! ! ! ! 13.1% of Americans met the daily recomme nded amount of fruits, and 8.9% met the daily recommended amount of vegetables (Moore et al., 2015). While both fruit and vegetable consumption are well below recommended daily amounts, it appears that the daily benchmark for vegetable consumption is sli ghtly more difficult (and therefore more effortful) to meet. This would theoretically require use of additional EF resources to meet this prescribed goal. Because vegetables may not be as "rewarding" to eat, their consumption likely requires engagement of EF to resist the temptation of eating sugary or fatty snacks instead of choosing healthier options. Fruits, in contrast, usually have a natural sweetness that makes them more rewarding to eat. In addition, we often cook the vegetables we consume while m ost fruits are consumed without cooking which is another factor that makes the consumption of vegetables more effortful. In addition, there is research suggesting that consumption of vegetables (but not fruit) is associated with prevention of cognitive de cline in elderly populations (Kang et al., 2005; Morris et al., 2006). The relationship between c ardiovascular risk factors, cognitive functioning , and cardiovascular mortality (CVM) risk is not clear. The overall aim of this research project is to invest igate the relationship between executive functioning (behavioral and attentional self regulation) and CVM risk and to evaluate whether engagement in instrumental health behaviors may mediate this relationship while accounting for other known risk factors. Objectives The overall objective of this study is to address several questions concerning the association between EF and a number of health behaviors that might mediate CVM in the
$% ! ! ! ! ! context of demographic, and comorbid risk factors . These questions include t he following: 1) Is there a relationship between EF and CVM risk ? 2) Do physical activity , vegetable consumption, and smoking behavior mediate the relationship between EF and risk of CVM? See Figure 1 for visual representations of the proposed mediators 3) Is EF a better predictor of CV M , physical activity, vegetable consumption, or smoking behavior compared to mental status? Hypotheses 1. There is a significant association between EF and CVM . a. Better EF will be associated with decreased risk of CVM . 2. Physical Activity , vegetable consumption, and smoking behavior are significant mediator s of the relationship between EF and CVM risk. a. Greater levels of engagement in physical activity (METs) , greater levels of vegetable consumption (servings per day), and lower le vels of lifetime smoking (packyears) will be associated with better EF and decreased risk of CVM . 3. EF will be a better predictor of CV M , physical activity, vegetable consumption, and smoking behavior compared to mental status a. The amount of variance accounted for ( R 2 and pseudo R 2 ) in linear and logistic regression models that utilize EF as the predictor variable will be greater compared to models that utilize MMSE as the predictor variable.
$& ! ! ! ! ! b. The correct classification rate (in l ogistic regression models) will be higher in models that utilize EF as the predictor variable compared to models that utilize MMSE as the predictor variable.
$' ! ! ! ! ! Figure 1. Hypothesized models of how physical activity, vegetable consumption, and smoking behavior may mediate the relationship between executive functioning and cardiovascular mortality risk . ! EF EF EF /012"345 ! 637"#"71 ! 89:974;59 ! <=>2?@AB ! C@=D">: ! <8 ! E=F745"71 ! G"2D ! <8 ! E=F745"71 ! G"2D ! <8 ! E=F745"71 ! G"2D ! 4 ! 4 ! 4 ! ; ! ; ! ; ! 3!H ! 3!H ! 3!H !
$( ! ! ! ! ! CHAPTER II METHODS Introduction The data were obtained in the course of the San Luis Valley Health and Aging Study (SLVHAS) , a population based longitudinal study of chronic illness and disability among Hispanic and Non Hispanic White (NHW) individuals residing in Alamosa and Conejos counties in Southern Colorado, who were at least 60 years old at recruitment. The area is predominantly rural, and most residents of the study area live in relatively stable small communities , or on farms or ranches . The Hispanic population in th e area i dentified themselves as being "Other Spanish/Hispanic (68%) or "Mexican/Mexican American (31%) on the 2000 census. Sampling The SLVHAS was reviewed and approved by the Colorado Multiple Institutional Review Board at the University of Colorado Denver. Household enumeration was conducted in the study counties between 1992 and 1993 (n=6482; response rate: 97.2%), and all Hispanic residents ! 60 years of age were invited to participate, along with a sample of NHW residents. Attempts were made to capture some basic information from eligible individuals who refused to participate in the full protocol in an effort to quantify any vari ables that might distinguish this group from those who agreed to participate. Eligible individuals who refused to participate were more likely to be NHW and to have a history of smoking, and less likely to have diabetes. A total of 2067 individuals initia lly agreed to participate. A total of 310 of these became ineligible prior to the initial assessment interview: 171 died, 125 moved out of the area, and 14 became ineligible
$) ! ! ! ! ! because of inaccurate age or ethnicity. Of the remaining 1757 eligible individu als, approximately 82% (n=1433) completed their first study interview between 1993 and 1995. There were 75 nursing home residents included in the initial study visits but the analyses reported here include only community dwelling individuals (n=1358). Att empts were made to administer the MMSE to all participants at the baseline interview to establish general mental status prior to administering the remaining protocol. A total of 39 community dwelling participants did not complete the MMSE at baseline. In addition to the MMSE, the Behavioral Dyscontrol Scale (BDS), a measure of executive functioning, was administered at the baseline interview. There were 51 community dwelling respondents who did not complete the BDS at baseline. Decisions to administer the full research protocol or a limited protocol (with or without a surrogate) were made largely based on participant performance on the MMSE. At the baseline interview, 179 participants required surrogate respondents for one of the following reasons: 1) MMSE score <18 points; 2) illiteracy; 3) lack of time to complete the interview; 4) access to eligible elder refused by surrogate. Surrogates who knew the participants well acted as respondents, providing selected information on behalf of the participant. It w as assumed that these individuals would not be privy to certain data that interviewers obtained by self report (e.g., routine diet and physical activity). As of March 2015, 1035 (of the original 1358) had died; of these 383 had CVD related causes listed as the underlying cause of death on their death certificates and the other 652 deceased individuals died of other causes . A total of 323 participants were thought to be still alive at the time of the death certificate search. See Figure 2 for a Study Flow Chart. Research questions were
$* ! ! ! ! ! tested as comparisons between those who were still alive (N = 323) and those who had died of CVD (N = 383) resulting in a sample for the current study of N = 706. Figure 2. San Luis Valley Health and Aging Study Flow Chart Procedure Participants were interviewed in English or Spanish, according to their preference, usually in their homes, unless they chose to be interviewed at the SLVHAS study office. Data collectors were all bilingual residents of the study area. All measures were av ailable in both English and Spanish. Originally developed in English, they were translated into Spanish by fluent speakers of both languages, then back translated into English to ensure fidelity.
$+ ! ! ! ! ! Measures Mortality Colorado death certificate searches were conducted through March 2015 to ascertain vital status. The underlying cause of death l isted on death certificates was used to identify our sample of individuals who had died from cardiovascular disease . T able 2 provides a list of the specific underl ying causes of death along with the ICD 9/ICD 10 codes that were used to determine our sample of individuals who died from cardiovascular disease. Executive Function We assessed executive functioning with the Behavioral Dyscontrol Scale (BDS), which assesses the capacity for behavioral and attentional self regulation. Specifically, it assesses a person's ability to initiate goal directed activity based on intentions, to monitor one's performance, and to inhibit irrelevant or inappropriate responses (Grigsby, et al., 1992). Possible total scores range from 0 19 points, with scores > 14 considered to be within normal limits. The BDS has been validated as a measure of ex ecutive function, and inter rater reliability, retest reliability (8 week and 6 month follow up), and internal consistency are all greater than 0.85 (Diesfeldt, 2004; Grigsby et al. 1992; Leahy et al., 2003; Suchy et al., 1997). Mental Status. The MMSE (F olstein, Folstein and McHugh, 1975) is a brief measure of general mental status with possible scores ranging from 0 30. The MMSE been shown to have adequate internal consistency (.77) (Holzer et al., 1987), test retest (8 week interval = .64)(O'Connor et a l., 1989), and interrater reliability (Berg, Franzen & Wedding, 1987).
$, ! ! ! ! ! Table 2 . International Classification of Disease (ICD 9 and ICD 10) codes for determining underlying cause of death due to cardiovascular disease Cause of Death ICD 9 code ICD 10 code Acute rheumatic fever and chronic rheumatic heart diseases 390 398 I00 I09 Hypertensive heart disease 402 I11 Hypertensive heart and renal disease 404 I13 Acute myocardial infarction 410 I21 I22 Other acute ischemic heart diseases 411 I24 Atherosclerotic cardiovascular disease, so described 429.2 I250 All other forms of chronic ischemic heart disease 412 414 I20, I25.1 125.9 Acute and subacute endocarditis 421 I33 Diseases of pericardium and acute myocarditis 420, 422 423 I30 I31, I40 Heart failure 428 I50 All other forms of heart disease 415 417, 424 427, 429.0 429.1, 429.3 429.9 I26 I28, I34 I38, I42 I49, I51 Essential hypertension 401, 403 I10, I12 Cerebrovascular disease (Stroke) 430 434, 436 438 I60 I69 Atherosclerosis 440 I70 Other major cardiovascular diseases 441 448 I71 I78 Aortic aneurysm and dissection 441 I71 Other diseases of arteries, arterioles and capillaries 442 448 I72 I78 Other disorders of circulatory system 451 459 I80 I99
%! ! ! ! ! Physical Activity Physical activity was assessed by asking participants to recall their participation in any of a number of common activities from a list of possibilities over the last 12 months, and to estimate the intensity, duration , and frequency with which they engaged in each . These questions were based on a questionnaire used in the CARDIA and Minnesota Heart Health Program. Test retest reliability (two week interval) ranged from .77 .84 for the moderate, vigorous and total amou nt of physical activity summary scores. Validity was assessed indirectly by investigating the correlations between level of total activity and skinfold thickness (r = .15 , p < .001 for women; r = .12, p < .001 for men), total caloric intake (r = .07, p < .001 for women; r = .21, p < .001 for men) , duration on a self limited maximal exercise test (r = .36, p < .001 for women; r = .25, p < .001 for men) , and a high density lipoprotein cholesterol (r = .13, p < .001 for women; r = .11, p < .001 for men) (J acobs et al., 1989). To obtain the desired health benefits, r ecommendations for physical activity traditionally prescribe that activities be performed at moderate levels of intensity for at least 150 min utes or at vigorous levels o f intensity for at least 75 minutes a week (Physical Activity Guidelines Advisory Committee, 2008). Accordingly, t he interview questions attempted to quantify physical activity in which participants engaged in at a moderate to vigorous level. Moderate intensity activities have METs values between 3 and 6 METs, while vigorous activities have METs values of 6 or greater (Physical Activity Guidelines Advisory Committee, 2008). Individuals who are following recommended guidelines are expending at least 450 METs on a weekly ba sis. An estimated number of metabolic equivalent of task units ( METs ) was
%$ ! ! ! ! ! assigned to each of the activities assessed in our interview based on published values from the Compendium of Physical Activities Tracking Guide ( Ainsworth et al. , 2011). A summary of METs values assigned to each activity is provided in Appendix . The total number of METs expended over the twelve months immediately prior to the baseline assessment was estimated from these values. Diet Diet was assessed by asking participants to recall , from a list of foods, which they had eaten , and to estimate the frequency and amount of each food they had consumed over the previous 12 months . This survey was based on the Willet's Food Frequency Questionnaire (FFQ) . Reproducibi lity (test retest reliability for an interval of one year) for all of the foods included in the FFQ ranged from .31 .92 with a mean of .59. Validity assessed by comparison to one week diet records ranged from .17 to .95 with a mean of .63 (Feskanich et al., 1993). The individual items were categorized into 5 main groups (protein, dairy, fruits, vegetables, bread/grains) , and a n estimate of servings per day was calculated for each food group. For this analysis we chose to investigate the servings per da y of vegetables as a possible diet mediator for the relationship between EF and card iovascular mortality risk. Smoking Behavior Cigarette s moking behavior was assessed by asking participants to estimate both the average number of cigarettes that they smoked on a daily basis , a nd the number of years that they smoked. Cheadle et al. (1994) conducted a metanalysis of 51 studies that compared self reported smoking behavior to biochemical measures of smoking behavior. These comparison s indicted that the s elf report measures included in their analysis had a
%% ! ! ! ! ! range of sensitivity from 6% to 100 % with an average of 87.5 %, and a range of specificity from 33% to 100% with a n average specificity of 89.2. Participant responses were converted into number of pack year s , by m ultiplying the number of packs of cigarettes (20 cigarettes per pack) consumed each year by the total number of years of smoking . Ca l culation of p ack years quantifies intensity and duration of cigarette con sumption , and allows us to i nvestigate smoking behavior as one continuous mediator variable . Covariates Covariates in the analysis include demographic variables that could potentially influence both a participant's BDS score a nd cardiovascular mortality risk. These include age, highest level of education, ethnicity ( Hispanic or non Hispanic White ), and sex . Total number of comorbid health conditions was also used as a covariate . Participants were asked whether they had ever been diagnosed by a physician as having a ny of several condition s (e.g., angina ) . Table 3 displays the frequency of these disorders based on vital status (alive or dead) .
%& ! ! ! ! ! Table 3 . Number of participants with comorbid health conditions by mortality status Condition Alive Percentage Dead due to CVD Percentage Arthritis 207 64 278 73 Cancer 29 9 55 14 Heart Attack 25 8 77 20 Mini Stroke 19 6 39 10 Major or Severe Stroke 5 2 22 6 Angina 21 7 40 10 Parkinson's Disease 2 1 9 2 High Blood Pressure 122 38 203 53 Enlarged Heart of Heart Failure 12 4 35 9 Emphysema, Chronic Bronchitis or COPD 26 8 38 10 Cirrhosis of the Liver 3 1 2 1 Kidney Failure 2 1 8 2 Osteoporosis 21 7 25 7 Seizure Disorder 6 2 4 1 Migraines of Persistent Recurrent Headaches 55 17 54 14 Depression 48 15 45 12 Diabetes 58 18 87 23 Blood Vessel Surgeries or Angioplasty 15 5 40 10 Mean number of comorbidities 2.09 (1.63) 2.77 (1.8) Data Analysis Preliminary Analyses: D ata analyses w er e conducted using SPSS version 22 (IBM Corp., 2013) . Prior to conducting our main analyses, descriptive statistics and bivariate correlations were obtained in order to characterize the structure of the dataset and identify patterns of
%' ! ! ! ! ! associations among the variables. We evaluated the linearity and normality of individual variables that we planned to include in the analyses to determine if assumptions for use of linear and logistic regression were met. The linearity assumption was examined by looking at scatterplots of all continuous predictors/covariates plo tted against the continuous DVs (DV=physical activity, DV=vegetables, DV=packyears). D ichotomous predictors (ethnicity, sex) were not evaluated for linearity. Transformations were considered for any non normal variables and an evaluation of outliers was c onducted. Because we were seeking to combine the results of two different ethnic groups for analysis, we ran preliminary analyses to determine if ethnic group differences existed for any of our variables. Hypothesis Testing: Our first objective w as evalu ated with a logistic regression model that include d BDS score as a predictor of CVM, along with age, education, ethnicity, sex, and comorbidity as covariates . We also conducted a n analysis that used all cause mortality as the outcome variable (instead of CVM) to determine how the BDS functioned as predictor of a non specific cause of death variable. Our second objective was evaluated with a 4 step process to determine whether there was support for mediation for each of our proposed mediator va riables ( Baron and Kenny , 1986) . In step 1 , for each of our three separate mediation analyses, a l ogistic regression was conducted to evaluate if a significant relationship exist ed between the independent variable (baseline BDS score) and the binary mortality variable ( "alive" or "dead due to CVD") after adjustment for covariates. In step 2, w e calculated three separate linear reg ression model s to determine whether the a paths were significant. The a path relates the predictor variable (X) to the
%( ! ! ! ! ! proposed mediator variable. In these regression equations, BDS score w as entered as the predictor variable , and the proposed mediators (physical activity, vegetable consumption, and smoking behavior ) were entered as the outcome variable. We also included our list of covariates in these equations. In step 3 , we conduct ed logistic regression s to determine whether a significant relationship exist ed between each of the proposed mediators and the binary mortality variable ( b portion of the in direct path ) , controlling for BDS score and other covariates. In step 4, we evaluate d whether there was support for full or partial mediation for each of the 3 proposed mediators , by observing whether BDS continue d to be a significant predictor of cardiovascular mortality after control ling for the mediators. Evaluation of interaction terms with our covariates was conducted to elaborate our understanding of the relationships that we were investigating . We limited t he number of these analyses to avoid inflating our chances of committing type one errors . To evaluat e our third objective , we ra n a series of regression equations using MMSE score as a predict or of CV M , physical activity, vegetable consumption, and smoking behavior, for the purpose of comparing the results to previous regression equations that utilized BDS as a predictor . CHAPTER III RESULTS Descriptive Statistics Table 4 summarizes t he descriptive statistics for all variables included in this analysis. Evaluation of missing data indicated that most of the missing data were purposefully missing, as participants were assigned to a data collection protocol in which, typically because the y had obtained MMSE scores in the demented range (i.e., < 1 8 ) , a
%) ! ! ! ! ! surrogate respondent was asked to provide inform ation on behalf of the participant . There was some information that surrogate responders were not able to provide. For example, 70 of the 75 individuals missing data for the physical activity variable; 87 of the 95 individuals missing data for the vegetable consumption variable; 16 of the 19 missing data for the smoking variable; and 6 of the 7 missing data for the edu cation variable were due to surrogate omission. In our mediation analyses, a listwise deletion procedure for missing data was utilized . Table 4 . Descriptive statistics for proposed IV, DV, Mediators and Covariates Variable N Missing Mean (SD) BDS 683 23 15.02 (4.42) Physical Activity 631 75 3219.90 (377.46) Vegetables 611 95 2.09 (1.06) Smoking 687 19 4578.03 (8817.62) Non Smoker 593 0 3602.93 (7627.04) Smoker 94 0 10729.4 (12605.87) Age 706 0 73.2 (7.88) Education 699 7 10.41 (3.99) Comorbidity 706 0 2.46 (1.76) MMSE 686 20 24.93(5.41) Variable N Missing Percentage Mortality 706 0 100 Dead CVD 383 0 54.25 Alive 323 0 55.75 Ethnicity 706 0 100 Hispanic 386 0 54.67 NHW 320 0 55.33 Sex 706 0 100 Fem ale 412 0 58.4 M ale 294 0 51.6
%* ! ! ! ! ! Notes: BDS (0 19 scale) . Physical activity (METs non transformed values) . Vegetables (servings per day) . Smoking (packyears non transformed values ). Age (age at assessment), Education (highest year of education attained) . Comorbidity (number of comorbid conditions) . MMSE (0 30 scale) . Mortality (Death due to CVD or Alive). Ethnicity (Hispanic or Non Hispanic White). Sex (Female or Male). Table 5 summarizes ethnic group differences for all variables included in our analyses . Compared to the NH W group, the Hispanic group reported lower mean scores on the BDS, physical activity, vegetable consumption , and education . Table 5 . Evaluation of Ethnic Group Differences Variable Non Hispanic White Mean (SD) Hispanic Mean (SD) t SEM BDS 16.45 (3.6) 13.86 (4.7) 8.0* .36 Physical Activity 3576.6 (3995.3) 2912.2 (3531.2) 2.2* 300.2 Vegetables 2.3 (1.1) 1.9 (1.0) 3.9** .09 Smoking 4554.5 (8424.7) 4598.1 (9152.2) .06 675.3 Age 73.4 (8.28) 73.0 (7.53) .68 .60 Education 12.4 (2.94) 8.8 (4.03) 13.26* .27 Comorbidity 2.5 (1.7) 2.3 (1.8) 1.20 .13 Variable Non Hispanic White (n = 320) N (%) Hispanic (n = 386) N (%) ! 2 Total Mortality .72 706 Alive 152 (47.5) 171 ( 44.3) 323 Dead CVD 168 (52.5) 215 (55.7) 383 Sex .177 706 Female 184 (57.5) 228 (59) 412 Male 136 (42.5) 158 (41) 294 N otes: * " # 0.05 level ** " # 0.01 level BDS (0 19 scale). Physical activity ( METs non transformed values ). Vegetables (S ervings per day ) . Smoking ( packyears non transformed values) . Age (age at assessment), Education (highest year of education attained). Comorbidity (number of comorbid conditions). MMSE (0 30 scale). Mortality (Death due to CVD or Alive). Ethnicity (Hispanic or Non Hispanic White). Sex (Female or Male).
%+ ! ! ! ! ! Evaluation of Assumption s The values of skewness and kurto sis indicate that most variables assessed at baseline were adequately normally distributed with the exception of physical activity and smoking which were successfully transformed with square root and log transformation pro cedures respectively . Linear relationships appeared to be present among the variables in our analyses . Correlations A mong V ariables Bivariate co rrelations among the independent, dependent , mediator and covariate variables are presented in Table 6 . The variable that most strongly correlated with BDS was the MMSE ( r = 0 .8) . Physical activity and education were also strongly correlated ( r > 0 .5) with BDS while age and CVD mortality were moderately correlated (3 < r < 5) with BDS . A small correlation was observed between BDS and vegetable consumption. Age and MMSE were moderately correlated with CVD mortality . P hysical activity, comorbidity, education and smoking behavior were observed to have a small correlation with CVD mortality .
%, ! ! ! ! !
&! ! ! ! ! Evaluation of Objectives Objective 1: BDS as a Predictor of CVM A binary logistic regression analysis was conducted to evaluate the relationship between BDS score and CV, controlling for age, education, ethnicity, sex, and comorbidities. A test of the full model against a constant only model was statistically significant ($ 2 = 239.51, p < 0 .001, with df = 6 ). The model explained 39.7% of the variance (Nagelkerke's R 2 ) in CVM and correctly classified 74.2% of cases. Higher BDS score was a ssociated with a decreased likelihood of dying from CVD (Exp(B) = 0 .878, 95% CI = 0 .824 to 0 .935) . Age, Sex, and Comorbidities were also significant predictors in the model wh ereas Education and Ethnicity were not. See Table 7 for a summary of results. Additionally we evaluated a logistic regression model that used all cause mortality as the outcome variable ( instead of CVM to ) to determine how well the BDS functioned as predictor of a less specific cause of death variable. B DS was significantly associated with decreased likelihood of dying (all causes), (Exp (B) = 0 .903, 95% CI = 0 .856 to 0 .952) . Table 7 . Summary of Logistic Regression Analyses for BDS and Covariates Predicting Cardiovascular Mortality Risk Variable B S.E Wald Df Exp(B) 95% Confidence Interval Lower Bound 95% Confidence Interval Upper Bound BDS .131* .032 16.330 1 .878 .824 .935 Age .147* .016 90.307 1 1.159 1.124 1.195 Education .015 .031 .232 1 1.015 .956 1.077 Ethnicity .008 .212 .001 1 .992 .655 1.504 Sex .838* .192 19.01 1 .433 .297 .631
&$ ! ! ! ! ! Comorbidities .271* .057 22.987 1 1.312 1.174 1.465 Notes: * " # 0. 0 0 1 level Objective 2: Physical Activity Mediation Analyses : Step 1. A logistic regression model that controlled for age, education, ethnicity, sex, and comorbidities, indicated that higher BDS score s were significantly associated with decreased likelihood of C VM ( Exp(B) = 0 .896 , 95% CI = 0 .835 to 0 .962 ) . Step 2 . A linear regression model that controlled for age education, ethnicity, sex, and comorbidities, indicated that BDS was a significant predictor of sqrt transformed physical activity (PAsqrt) ( B = 1.359 , 95% CI = ( 0 .634 to 2.083 ) . Step 3. A logistic regression model that controlled for age, education, ethnicity, sex, comorbiditi es, and BDS, indicated that PA sqrt was not significantly associated with likelihood of CV M ( Exp(B) = 0 .995, 95% CI = 0 .988 to 1.002 ) . Step 4. BDS continues to be a significant predictor of CV M after controlling for all covariates and PA sqrt (Exp(B) = 0 .902 , 95% CI = ( 0 .840 to 0 .968 ) . Evaluation of interactions : W e investigated the b path of this analysis to determine whether there were any significant interaction terms between any of the covariates and the physical activity variable. The only significant interaction term was with sex x physical activity (Exp(B) = 0 .986, 95% CI = 0 .972 to 0 .999) . Objective 2: Vegetable C onsumption M ediation A nalyses : Step 1. A logistic regression model that controlled for age, education, ethnicity, sex, and comorbidities, indicated that higher BDS score was significantly associated with decreased likelihood of CV M ( Exp(B) = 0 .896, 95% CI = 0 .833 to 0 .965 )
&% ! ! ! ! ! Step 2. A linear regression model that controlled for age education, ethnicity, sex, and comorbidities, indicated that BDS was not a significant predictor of vegetable consumption ( B = 0 .0 23 , 95% CI = 0 .008 to 0 .54). Step 3. A logistic regression model that controlled for age, education, ethnicity, sex, comorbidities, and BDS, indicated that vegetable consumption was not significantly associated with likelihood of CV M (Exp(B) = 0 .933, 95% CI = 0 .777 to 1.120). Step 4. In a logistic regression model that controlled for age, education, ethnicity, sex, comorbidities, and vegetable consumption, BDS continues to be a significant predictor of CVD mortality (Exp(B) = 0 .898, 95% CI = 0 .834 to 0 .966 ) . Objective 2: Smoking Behavior M ediation A nalyses : Step 1. A logistic regression model that controlled for age, education, ethnicity, sex, and comorbidities, indicated that higher BDS score s were significantly associated with decreased likelihood of CV M ( Exp(B) = 0 .884, 95% CI = 0 .828 to 0 .944 ) . Step 2. A linear regression model that controlled for age education, ethnicity, sex, and comorbidities, indicated that BDS was not a significant predictor of log transformed pack years (logPack) ( B = 0 . 039 , 95% CI = 0 .001 to 0 .079). Step 3. A logistic regression model that controlled for age, education, ethnicity, sex, comorbidities, and BDS, indicated that log Pack was significantly associated with likelihood o f CV M ( Exp(B) = 1.310, 95% CI=1.170 to 1.466 ) . Step 4. In a logistic regression model that controlled for age, education, ethnicity, sex, comorbidities, and log Pack , BDS continues to be a significant predictor of CV M ( Exp(B) = 0 .873, 95% CI = 0 .817 to 0 .934 ) .
&& ! ! ! ! ! Objective 3: MMSE as a P redictor of C ardiovascular Mortality A binary logistic regression analysis was conducted to evaluate the relationship between MMSE score and CV M , controlling for age, education, ethnicity, sex, and comorbidities. A test of the full model against a constant only model was statistically significant ($ 2 = 239.79, p < 0 .001, with df = 6). The model explained 39.6% of the variance (Nagelkerke's R 2 ) in CVM and correctly classified 75.1% of cases. Higher MMSE score was associated with a decreased likelihood of dying from CVD (Exp(B) = 0 .895, 95% CI = 0 .846 to 0 .946). Age, Sex, and Comorb idities were also significant predictors in the model wh ereas Education and Ethnicity were not. See Table 8 for a summary of results. Table 8 . Summary of Logistic Regression Analyses for MMSE and Covariates Predicting Cardiovascular Mortality Risk Variable B S.E Wald Df Exp(B) 95% Confidence Interval Lower Bound 95% Confidence Interval Upper Bound MMSE .1 11 * .0 28 1 5 . 219 1 .8 95 .8 46 .9 46 Age .14 3 * .016 82 .07 0 1 1.15 3 1.1 18 1. 189 Education .01 6 .031 .2 53 1 1.01 6 .956 1.07 9 Ethnicity .0 23 .21 3 .0 12 1 .9 77 .6 44 1. 483 Sex .8 04 * .19 1 1 7 . 735 1 .4 47 . 308 .6 50 Comorbidities .27 8 * .05 6 2 4 . 511 1 1.3 21 1.1 83 1.4 74 Notes: * " # 0.0 0 1 level The MMSE and the BDS accounted for nearly identical amount s of variance (Nagelkerke's R 2 ) in CVM risk, 39.6% compared to 39.7% respectively. The classification rates were also nearly identical with BDS correctly classifying 74.2% participants and the MMSE classifying 75.1% of participants.
&' ! ! ! ! ! Objective 3: MMSE as a P redictor of Physical Activity A linear regression model that controlled for age education, ethnicity, sex, and c omorbidities, indicated that MMSE was a significant predictor of sqrt transformed physical activity (PAsqrt) ( B = 1.004 , 95% CI = ( 0 .349 to 1.659 ). The fully adjusted model accounted for 20.3% of variance compared to a model that used BDS instead of the MMSE , which accounted for 19.7% of variance. Objective 3: MMSE as a P redictor of Vegetable Consumption A linear regression model that controlled for age education, ethnicity, sex, and comorbidities, indicated that MMSE was not a significant predictor of vegetable consumption ( B = 0 .014 , 95% CI = 0 .015 to 0 .044) . The fully adjusted model accounted for 4.5 % of variance in vegetable consumption , c om pared to a model that used the BDS instead of the MMSE , which accounted for 4.8% of variance. Objective 3: MMSE as a P redictor of Smoking B e havior A linear regression model that controlled for age education, ethnicity, sex, and comorbidities, indicated that MMSE was not a significant predictor of log transformed pack years (logPack) ( B = 0 .0 33 , 95% CI = 0 .001 to 0 .068). The fully a djusted model accounted for 19.4 % of variance in smoking behavior compared to a model that used the BDS instead of the MMSE , which accounted for 19.3% of variance. CHAPTER IV DISCUSSION Implications
&( ! ! ! ! ! The results of our analyses have provided support for our first hypothesis, that there was a significant association between EF and cardiovascular mortality risk where higher EF scores were associated with a decreased risk of dying from cardiovascular disease. These results a dd to the previous findings that have established a link between EF and all cause mortality (Amirian, 2010) . Additionally, EF was a slightly stronger predictor in a model that had CVM as the outcome variable compared to when it was used as a predictor in a model that had all cause mortality as the outcome variable. By specifying a particular domain of cognitive functioning (EF) that can be theoretically related to engagement in health behaviors , along with a more specific cause of death variable (CVD mortality) that can also theoretically be related to the health behaviors, we hoped to minimize the number of explanatory factors that we would have to consider to evaluate this relationship. However, the nature of this relationship is still not clear as none of the mediation analyses that we conducted found clear support for any of our proposed health behaviors functioning as medi ators of the observed relationship between EF and CV M . Among the three proposed mediator variables, EF was a significant predictor only of physical activity . In addition, packyears was the only proposed mediator that was a significant predict or of CV M . The causal effect of EF on CVD mortality through health behavior is likely more complex than proposed in our simple mediation mo dels. Given that the mediator variables are conceptually related (viz., they are health behaviors), it may make theoretical sens e to consider them together as a composite variable that combines all three of our proposed mediators into one latent "health behavior" variable. In addition, t he indirect effect of EF on CVD mortality through physical activity, vegetable
&) ! ! ! ! ! consumption, or smoking behavior may be conditional on some fourth variable that was not captured in our simple mediation models . Preacher, Rucker, & Hayes (2007) discuss the concept of conditional mediation (moderated mediation) and its potential value in elabor ating com plex mediation models. In an effort t o understand why we failed to fi nd clear support f or the mediation models, we conducted a limited number of secondary interaction analyses to investigate how the covariates function as moderators of the indirect effects that we hypothesized would exist. Although none of the three mediation models was significant, the physical activity model had th e most support. In the mediation model for physical activity, the a path ( i.e., the re lationship between EF and physical activity) was significant, but the b path ( the relationship between physical activity and CVD mortality) fell short of significan ce . In analyzing the b path, we found that sex x physical activity was the only significant interaction term. When we repeated the same 4 step me diation analysis with only female participants, there was support for the hypothesis that physical activity act s as a partial mediator of the relationship between EF and cardio vascular mortality . When the procedure was repeated with only male participants , the relationship was not significant . A n explanation for this gender difference observed in our sample is not evident. It is noteworthy that for the physical activity variable, the mean and standard deviation for males w ere higher than for females . The large amount of variability overall within our physical activity variable makes it difficult to speculate about or interpret these findings , as th ey may reflect true variability, or may just be a reflect ion of unreliability in our measurement of physical activity . It is also noteworthy that d espite significant ethnic differences in physical activity, ethnicity was not a significant moderator of phys ical
&* ! ! ! ! ! activity , and ethnicity was similarly distributed for males and females . A more detailed analysis of possible moderators may be in order for future analyses. T he results associated with our third objective did not support the hypothesis that EF (relative to mental status) would be a stronger predictor of CVM, physical activity, vegetable consumption , and smoking behavior. The two variables were remarkably similar in terms of their ability to predict CVM , as indicated by comparing pseudo R 2 va lues and classification rates across the two models . Comparison of R 2 values produced in separate linear regression models did not suggest a difference in their relative abilities to predict level of engagement in any of our proposed mediators either. Gi ven that the two measures were so highly correlated, this finding is perhaps unsurprising. Limitations The way we conceptualized EF suggests that it may be most relevant as a predictor of behavior change. However, we did not have any data regarding whether participants were attempting to engage in any behavior change , or attempting to meet any prescribed goal (e.g., dietary recommendations ) , and neither were there data concerning awareness of daily recommendations for vegetable consumption or physica l activity. Hence, some individual s may have had intact EF , but not engage in health behaviors because they lack information about the potential benefits. Although we propose that intact EF is necessary for engagement in novel health behaviors, intact EF may be necessary but not sufficient for engaging in these health behaviors. Moreover, maintenance of any behavior, whether it is beneficial for health (exercise, healthy eating) or negative for health (smoking) can occur independent ly of EF if it is alre ady being carried out in a habitual/automatic manner. It is possible that the sample might have
& + ! ! ! ! ! included individuals who had acquired an impairment in EF after they had already established good (or negative) health behaviors. In this group of participant s, engagement in healthy behaviors occurs in a habitual/automatic manner and does not require EF . By the same token, habitual smoking is a behavior that takes place in an automatic manner , and therefore its maintenance occurs independent ly of EF. Th o se participants who smoke may be accumulating higher packyears despite intact EF. The presence of these groups of individuals might obscure the directional relationships that we hyp othesized would exist between impaired EF and failure to engage in health beha viors . This operationalization of our behavioral variables may have contributed to why the BDS and MMSE were indistinguishable as predictors in the regression models. Although the two measures are significantly correlated , it is nevertheless possible for an individual to be cognitively intact (as measured by MMSE) , but unable to initiate novel or complex tasks , as in adopting new health behaviors. Examining health behaviors in the context of an attempt to change one's behavior might allow for a better demo nstration of how the two constructs are also dissociable. Among the participants, there was significant variability in how long they survived after their baseline assessment. Some participants might have lived for an additional 20 years after their asse ssment wh ereas others may have died very soon after their information was collected. This introduces a number of potential complicating factors into our analysis. It is questionable how well a baseline assessment of EF may be related to a person's risk o f mortality 20 years later. In addition we have no indication of how their EF may have changed in the intervening years. The youngest members of the original SLVHAS cohort would now be 82 years old, and are likely to have developed a
&, ! ! ! ! ! number of comorbid co nditions in the intervening years. A longitudinal study with multiple follow up points would be necessary to be able to truly understand any observed relationships. In a similar manner, a participant who dies very soon after his/her data were obtained may complicate the interpretation of our analysis. It is possible that these participants may have been in the advanced stages of CVD disease when assessed, and their engagement in health behaviors may not have had an appreciable benefit to their health or mitigated their risk of death. Additionally, CVD may have already been so advanced in these individuals that it was limiting t heir ability to engage in health behaviors, and may also have been causing cognitive deficits, thereby introducing the problem of reverse causality. Another limitation to our analyses may have been our definition of cardiovascular mortality. Examination o f Table 2 indicates that a large number of participants that were part of the "Alive" group had a history of cardiovascular disease , and therefore probably shared characteristics that overlapped with th os e who had already died from CVD. I n all likelihood, a number of these individuals will eventually also die from CVD. This factor might have obscured or weakened any relationships with CVM that were investigated in our analyses , especially as the EF data were obtained about 22 years ago . Given that CVD and other chronic disease s w ere already present within many of the participants of the sample at the baseline assessment of EF, it is not possible to reliably infer the direction of causation for the relationship between EF and CVM that was observed in our sample. Perhaps a decline in EF is caused by cardiovascular disease, or associated with a physiological decline associated with such disease. In that case there could be a positive feedback loop, with advancing CV disease causing executive
'! ! ! ! ! impairment, an d increasingly deficient EF making it more difficult to manage one's illness. Capturing premorbid levels of EF would enable us to have a stronger theoretical rationale for inferring a causal pathway from impaired EF to increased risk of CVM that works thr ough engagement in health behaviors. Additionally, when defining our mortality variable, we used only the underlying cause of death. However, death certificates allow for the indication of other factors that may have contributed to an individual's death , in addition to a proximal cause of mortality . It is possible that a more detailed evaluation of factors that contributed to a participant's death may have allowed us to better define two distinct groups for our mortality variable. Death certificate data are frequently somewhat unreliable in this regard. Table 4 indicates that the NHW and Hispanic groups differed significantly with respect to BDS score , physical activity, vegetable consumption , and education . Although the relationship s between BDS and physical activity , and BDS and vegetable consumption , were in the expected direction ( i.e., lower BDS associated with lower physical activity and lower vegetable consumption ) , the groups did not differ in terms of CVD mortality risk. Addit ionally, although Hispanic s had significantly l ess education than NHWs (means = 8. 8 and 12. 4 , respectively) , Hispanics did not have a significantly higher number of comorbid health conditions , as would be predicted based on previous research (Winkleby et a l., 1992) . T hese findings could be indicative of the "Hispanic paradox , " where by increased risk factors do not necessarily confer an increased risk of mortality/morbidity (Markides & Coreil, 1986) . The reasons for this epidemiologic phenomenon are still unclear , although a number of a lternative explanations for this
'$ ! ! ! ! ! paradox have been proposed (Crimm i ns et al., 2007; Markides & Eschbach, 2005) . Evaluation of these explanations is beyond the scope of this analysis. However, g iven that we combined the ethn ic groups for our analyses, and more than half of our sample was Hispanic, it is possible that this phenomenon may also have obscured or weakened some of the relationships that we were investigating in our overall sample . Self report instruments, including the structured interview used to capture the physical activity , diet , and smoking variables in these analys e s , allow researchers to capture information from large groups of people efficiently, and do not have the effect of changing the behavior u nder study. However, there are also obvious limitations to this method. Social desirability bias may lead to over reporting physical activity or healthy eating behavior , and perhaps under reporting smoking behavior . In addition, the ability to recall pa st behavior can be a cognitively demanding task, and there is significant interindividual variability in a person's ability to report past physical activity , smoking behavior, and eating behavior accurately, particularly if a great deal of time has passed (Sallis, 2000; Prince et al., 2008). If we look at the mean s and standard deviations for the physical activity and smoking variables in particular (Table 3), we see that the standard deviations are in both cases larger than the means . Though it is possible that this may be a true reflection of the range of differences observable in an elderly population, it may also reflect unreliability in the self report instruments used in the SLVHAS . The ability to detect an effect in the context of this amount of variability is significantly diminished. The total METs expended over the entire year for the average participant in this study was 3219, this amounts to approximately 62 METs on a weekly basis. To meet the daily recommended amount of physical activity, an individual would need to expend at
'% ! ! ! ! ! least 450 METs per week. As a group, participants in this study were far below the recommended daily amount of physical activity. Because of advanced age the SLVHAS sample, it is not surprising that their l evel of physical activity was low, but there is a possibility that not enough participants achieved a level of physical activity sufficient to mitigate cardiovascular mortality risk to allow us to detect differences . Pack years is a convenient and simple way of quantifying intensity and duration of smoking behavior , and is useful in epidemiological research that does not have a primary goal of creating a specific model of cigarette smoking risk (Lubin & Caborasco, 2013). Howeve r, it does not consider the differential predictive effects of duration or intensity based on age or years since starting/quitting smoking (Peto, 2012). Although our analysis confirmed that higher number of pack years was associated with increased mortali ty risk , the association with BDS was not significant. Given that our conceptualization of the link between EF and health behavior would be strongest when o bserved in the context of attempted behavior change (e.g., cessation of a bad habit) , investigating attempts to stop smoking may have had a stronger association with EF . Although consumption of vegetables is associated with better health, and is a relatively good proxy for a healthy diet, it is only one aspect of nutrition . I ndividual s may consume the recommended daily amount of vegetables , but may not meet other dietary guidelines , therefore increasing their risk of CVD mortality. A lthough difficult to define, a more compr e hensive "healthy eating" index m a y be a better predictor of CVD mortality , as it may b e a more accurate indicator of overall eating habits . Future Directions
'& ! ! ! ! ! O ur analyses provide additional support for the previously demonstrated relationship between EF and mortality , and also extend s this line of research by finding an association between EF and a specific cause of death variable (CVM) . While there were sev eral methodological limitations of our analysis , there was evidence that EF may be more strongly associated with CVM than with all cause mortality. The use of disease specific causes of mortality is an area tha t warrants further exploration. In addition, the continued investigation of EF or other specific aspect s of cogniti on, and their association with mortality , is warranted given that the use of gene ral cognitive ability or general mental status has not provided us with particularly useful insights for how to understand these relationships. While our analyses have not provided compelling evidence that engagement in health behaviors is a causal pathway in the relationship between EF and CVM, additional analys es of our current data set may be useful, after addressing some of the methodological limitations discussed above . Modifying our current mortality variable by excluding participants with a history of CVD from the alive group (or by combining them in a group that includes those who have already died from CVD ) may also improve our ability to detect an effect , although the exclusion of participants would reduce statistical power . To ad dress the problems introduced by the variability in timing of assessments in relation to time to death, we might want to consider evaluating the hypotheses based on how long participants survived after their baseline assessment. This would allow us to und erstand how the relationship between EF and CVM may differ as a function of time until death. Modifying the smoking behavior variable to reflect attempts to stop smoking may be more in line with our conceptualization of EF a s a predictor of behavior change.
'' ! ! ! ! ! In addition, combining our proposed health behavior variables into one composite/latent variable that also considers moderators might improve the theoretical reasoning for our proposed mediation model , and enhance our abilit y to detect an effect , assuming one does exist . The SLVHAS was a community wide, interview based epidemiologic study that sought to understand differences in chronic illness and functional status as they affect Hispanic and non Hispanic White older persons. A prospective, experimental protocol designed specifically to examine this model of behavioral regulation would allow more thorough study of the relationship between EF and engagement in health behaviors , and of how engagement in health behaviors might be related to the risk of developing, and of dyi ng from, CVD. Recruiting participants to engage in an intervention study aimed at increasing a particular health behavior (e.g., increasing physical activity) would ensure that one was measuring individuals who were attempting to acquire a novel health beh avior as a habit . Comprehensive data collection at baseline (before the start of the intervention), with regular follow up data collection time points , would also be essential to understand how EF may be related to success (or failure) in initiating and maintaining a regular exercise program. Baseline measurements should at least include EF and biological markers for CVD so that one could ensure that o ne was capturing premorbid levels of EF in at least some participants . Additional measurements of EF at different follow up time points may allow us to investigate how the intervention itself may be affecting EF and how maintenance of a n established behavi or may be related to EF. The use of monitoring devices such as accelerometers might allow the capture of objective data concerning physical activity , eliminating the need to rely entirely on self report. If it
'( ! ! ! ! ! were possible to follow such a cohort for a relatively long period of time (e.g., at least 6 10 years), we could evaluate how EF over time is related to engagement in physical activity and the risk of developing , and dying from, CVD. A n additional layer of analysis that could be added to such a lon gitudinal research protocol might involve the use of an intervention with the potential to improve executive function (e.g., meditation) , allowing one to determine if a change (improvement) in EF is associated with increased likelihood of initiating or maintaining an exercise program. Conclusion The link between cognition and mortality is a consistent finding in the field of cognitive epidemiology . However, the complexity of the relationship make s it difficult to explain , and there are any number of potential factors that could interact to affect this relationship. We attempted to narrow the number of variables to consider by choosing a specific cognitive domain (EF) as well as a more specific ca use of death variable (cardiovascular mortality) and proposing a causal pathway that acted through engagement in health behaviors. However, the way in which we operationalized the outcome and mediator variables likely contributed a n unknown amount of error into the analyses. In addition , the simple mediation models that were proposed did not fully capture the complexity of these relationships. These limitations make detection of a significant effect less likely , and make it somewhat more difficult to interpret significant findings. There are certainly methodological changes that could greatly improve this analysis but there are also limitation s imposed by the data collected in this observational epidemiological study . Designing a n experimental st udy that fully integrates our
') ! ! ! ! ! conceptualization of EF and how it is related to engagement in novel health behaviors would allow us better to elucidate a causal pathway between cognition and mortality.
'* ! ! ! ! ! References Ainsworth BE, Haskell WL, Herrmann SD, Meckes N, Bassett Jr DR, Tudor Locke C, Greer JL, Vezina J, Whitt Glover MC, Leon AS. 2011 Compendium of Physical Activities: a second update of codes and MET values. Medicine and Science in Sports and Exercise , 2 011;43(8):1575 1 581. Allan, Robert. The evolution of cardiac psychology. In: Allan R, Fisher J, editors. Heart and Mind: The Practice of Cardiac Psychology. 2nd ed. Washington, DC: American Psychological Association; 2012. pp. 5 Ã 8. Amirian, E., Baxter, J., Grigsby, J., Curran Everett, D., Hokanson, J. E., & Bryant, L. L. (2010). Executive function (capacity for behavioral self regulation) and decline predicted mortality in a longitudinal study in Southern Colorado. Journal of clinica l epidemiology , 63 (3), 307 314. Anderson, G., & Horvath, J. (2004). The growing burden of chronic disease in America. Public Health Reports , 119 (3), 263 Ã 270. doi:10.1016/j.phr.2004.04.005 Anstey, K. J., Luszcz, M. A., Giles, L. C., & Andrews, G. R. (2001 ). Demographic, health, cognitive, and sensory variables as predictors of mortality in very old adults. Psychology and aging , 16 (1), 3. Baron, R. M., & Kenny, D. A. (1986). The moderator Ã mediator variable distinction in social psychological research: Conc eptual, strategic, and statistical considerations. Journal of personality and social psychology , 51 (6), 1173. Brega, A. G., Grigsby, J., Kooken, R., Hamman, R. F., & Baxter, J. (2008). The impact of executive cognitive functioning on rates of smoking cess ation in the San Luis Valley Health and Aging Study. Age and ageing , 37 (5), 521 525. Bassuk, S. S., Wypij, D., & Berkmann, L. F. (2000). Cognitive impairment and mortality in the community dwelling elderly. American journal of epidemiology , 151 (7), 676 688. Berg, S. (1996). Aging, behavior, and terminal decline. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (4th ed.,pp. 323 337). New York:Academic Press. Brands, A.M.A., van den Berg., E., Kessels, R.P., & Biessels, G.J. (2015). Cognition in diabetes and pre diabetes stages. In Waldstein, S. R. & Elias, M.F. (Eds.), Neuropsychology of Cardiovascular Disease (184 218). New York: Psychology Press. Carlson, M.C. & Varma, V.R. (2015). Activity and neurocognitive health in older adults. In Waldstein, S. R. & Elias, M.F. (Eds.), Neuropsychology of Cardiovascular Disease (79 108). New York: Psychology Press.
'+ ! ! ! ! ! Centers for Disease Control and Prevention. (2015). Heart disease risk factors . Retrieved from http://www.cdc.gov/heartdisease/risk_factors.htm Crimmins, E. M., Kim, J. K., Alley, D. E., Karlamangla, A., & Seeman, T. (2007). Hispanic Paradox in Biological Risk Profiles. American Journal of Public Health , 97 (7), 1305 Ã 1310. http://doi.org/10.2105/AJPH.2006.091892 Deary, I. J., & Batty, G. D. (2007). Cognitive epidemiology. Journal of Epidemiology and Community Health , 61 (5), 378 384. Dey, J., Misra, A., Desai, N. G., Mahapatra, A. K., & Padma, M. V. (1997). Cognitive function in younger type II diabetes. Diabetes Care , 20 (1), 32 35. Diesfeldt HFA: (2004) Executive functioning in psychogeriatric patients: Scalability and construct validity of the Behavio ral Dyscontrol Scale (BDS). International Journal of Geriatric Psychiatry, 19, 1065 1073. Diesfeldt HFA: (2004) Executive functioning in psychogeriatric patients: Scalability and construct validity of the Behavioral Dyscontrol Scale (BDS). International Journal of Geriatric Psychiatry, 19, 1065 1073. Elias, M. F., Goodell, A.L., & Robbins, M.A. (2015). Hypertension, blood pressure, and cognitive functioning. In Waldstein, S. R. & Elias, M.F. (Eds.), Neuropsychology of Cardiovascular Disease (109 149). Ne w York: Psychology Press . Escobar, J. I., Burnam, A., Karno, M., Forsythe, A., Landsverk, J., & Golding, J. M. (1986). Use of the Mini Mental State Examination (MMSE) in a community population of mixed ethnicity: Cultural and linguistic artifacts. The Jou rnal of nervous and mental disease , 174 (10), 607 614. Feskanich, D., Rimm, E. B., Giovannucci, E. L., Colditz, G. A., Stampfer, M. J., Litin, L. B., & Willett, W. C. (1993). Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire. Journal of the American Dietetic Association , 93 (7), 790 796. Friedman, H. S., Tucker, J. S., Schwartz, J. E., Tomlinson Keasey, C., Martin, L. R., Wingard, D. L., & Criqui, M. H. (1995). Psychosocial and behavioral predictors of longevity: The aging and death of the" Termites.". American Psychologist , 50 (2), 69. Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). "Mini mental state": a practical method for grading the cognitive state of patients for the clinic ian. Journal of psychiatric research , 12 (3), 189 198. Fuster JM: (1997) The prefrontal cortex: Anatomy, physiology, and neuropsychology of the frontal lobe. 3rd ed. Philadelphia: Lippincott Raven.
', ! ! ! ! ! Fuster JM: (1999) Cognitive functions of the frontal lobes. In BL Miller & JL Cummings (eds.), The human frontal lobes. New York: Guilford, 187 195. Fuster JM: (2000) Executive frontal functions. Experimental Brain Research, 133, 66 70. Gale, C. R., Martyn, C. N., & Cooper, C. (1996). Cognitive impairment and mortality in a cohort of elderly people. Bmj , 312 (7031), 608 611. Grigsby, J., Kaye, K., Baxter, J., Shetterly, S. M., & Hamman, R. F. (1998). Executive cognitive abilities and functional status among com munity ! dwelling older persons in the San Luis Valley Health and Aging Study. Journal of the American Geriatrics Society , 46 (5), 590 596. Grigsby, J., Kaye, K., & Robbins, L. J. (1992). Reliabilities, norms and factor structure of the Behavioral Dyscontrol Scale. Perceptual and motor skills , 74 (3), 883 892. Gussekloo, J., Westendorp, R. G. J., Remarque, E. J., Lagaay, A. M., Heeren, T. J., & Knook, D. L. (1997). Impact of mild cognitive impairment on survival in very elderly people: cohort study. BMJ , 315 ( 7115), 1053 1054. Hart, C. L., Taylor, M. D., Smith, G. D., Whalley, L. J., Starr, J. M., Hole, D. J., ... & Deary, I. J. (2003). Childhood IQ, social class, deprivation, and their relationships with mortality and morbidity risk in later life: prospective observational study linking the Scottish Mental Survey 1932 and the Midspan studies. Psychosomatic Medicine , 65 (5), 877 883. Hassing, L. B., Small, B. J., von Strauss, E., Fratiglioni, L., & BÂŠckman, L. (2002). Mortality related differences and changes i n episodic memory among the oldest old: Evidence from a population based sample of nonagenarians. Aging, Neuropsychology, and Cognition , 9 (1), 11 20. IBM Corp. Released 2013. IBM SPSS Statistics for Macintosh, Version 22.0. Armonk, NY: IBM Corp. Patrick D L, Cheadle A, Thompson DC, et al. The validity of self reported smoking: a review and meta analysis. Am J Public Health 1994;84:1086 Ã 93. Jacobs, D. R., Hahn, L. P., Haskell, W. L., Pirie, P., & Sidney, S. (1989). Validity and reliability of short physical activity history: CARDIA and the Minnesota Heart Health Program. Journal of Cardiopulmonary Rehabilitation and Prevention , 9 (11), 448 459. Johnson, J. K., Lui, L. Y., & Yaffe, K. (2007). Execut ive function, more than global cognition, predicts functional decline and mortality in elderly women. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences , 62 (10), 1134 1141.
(! ! ! ! ! Kang, J. H., Ascherio, A., & Grodstein, F. (2005). Fruit and vegetable consumption and cognitive decline in aging women. Annals of neurology , 57 (5), 713 720. Kaye K, Grigsby J, Robbins LJ, Korzun B: (1990) Prediction of independent functioning and behavior problems in geriatric patients. Journal of the Am erican Geriatrics Society, 38, 1304 1310. Kelman, H. R., Thomas, C., Kennedy, G. J., & Cheng, J. (1994). Cognitive impairment and mortality in older community residents. American Journal of Public Health , 84 (8), 1255 1260. Korten , A. E., Henderson, A. S., Christensen, H., Jorm, A. F., Rodgers, B., Jacomb, P., & Mackinnon, A. J. (1997). A prospective study of cognitive function in the elderly. Psychological medicine , 27 (04), 919 930. Leahy B, Suchy Y, Sweet JJ, Lam CS: (2003) Beha vioral Dyscontrol Scale deficits among traumatic brain injury patients, Part I: Validation with nongeriatric patients. The Clinical Neuropsychologist, 17, 474 491. Lezak MD, Howieson DB, Bigler ED, Tranel D: (2012) Neuropsychological assessment, 5 th ed. N ew York: Oxford University Press. Lubin, J. H., & Caporaso, N. E. (2013). Misunderstandings in the misconception on the use of pack years in analysis of smoking. Br J Cancer , 108 (5), 1218 1220. Luria AR: (1980) Higher cortical functions in man. 2d editio n. New York: Basic Books. Markides, K. S.; Coreil, J. (1986). The health of Hispanics in the Southwestern United States: An epidemiological paradox. Public Health Reports, 101, 253 265. Markides, K. S., & Eschbach , K. (2005). Aging, migration, and mortality: current status of research on the Hispanic paradox. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences , 60 (Special Issue 2), S68 S75. Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D. (2000). The unity and diversity of executive functions and their contributions to complex "frontal lobe" tasks: A latent variable analysis. Cognitive psychology , 41 (1), 49 100. Moore LV, Dodd KW, Thompson FE, Gr imm KA, Kim SA, Scanlon KS. Using Behavioral Risk Factor Surveillance System data to estimate the percentage of the population meeting US Department of Agriculture Food Patterns fruit and vegetable intake recommendations. Am J Epidemiol 2015;181:979 Ã 88. M orris, M. C., Evans, D. A., Tangney, C. C., Bienias, J. L., & Wilson, R. S. (2006). Associations of vegetable and fruit consumption with age related cognitive change. Neurology , 67 (8), 1370 1376.
($ ! ! ! ! ! Neale, R., Brayne, C., & Johnson, A. L. (2001). Cognition and survival: an exploration in a large multicentre study of the population aged 65 years and over. International Journal of Epidemiology , 30 (6), 1383 1388. Neisser, U., Boodoo, G., Bouchard Jr, T. J., Boykin, A. W., Brod y, N., Ceci, S. J., ... & Urbina, S. (1996). Intelligence: knowns and unknowns. American psychologist , 51 (2), 77. Nguyen, H. T., Black, S. A., Ray, L. A., Espino, D. V., & Markides, K. S. (2003). Cognitive impairment and mortality in older Mexican America ns. Journal of the American Geriatrics Society , 51 (2), 178 183. O'Connor, D. W., Pollitt, P. A., Hyde, J. B., Fellows, J. L., Miller, N. D., Brook, C. P. B., & Reiss, B. B. (1989). The reliability and validity of the Mini Mental State in a British communi ty survey. Journal of psychiatric research , 23 (1), 87 96. O'Toole, B. I., & Stankov, L. (1992). Ultimate validity of psychological tests. Personality and individual differences , 13 (6), 699 716. Park, D. C., Lautenschlager, G., Hedden, T., Davidson, N. S. , Smith, A. D., & Smith, P. K. (2002). Models of visuospatial and verbal memory across the adult life span. Psychology and aging , 17 (2), 299. Peto, J. (2012). That the effects of smoking should be measured in pack years: misconceptions 4. British Journal of Cancer , 107 (3), 406 Ã 407. doi:10.1038/bjc.2012.97 Physical Activity Guidelines Advisory Committee. (2008). Physical activity guidelines for Americans. Washington, DC: US Department of Health and Human Services , 15 34. Preacher, K. J., Rucker, D. D., & Hayes, A. F. (2007). Addressing moderated mediation hypotheses: Theory, methods, and prescriptions. Multivariate behavioral research , 42 (1), 185 227. Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior research methods , 40 (3), 879 891. Prince, S. A., Adamo, K. B., Hamel, M. E., Hardt, J., Gorber, S. C., & Tremblay, M. (2008). A comparison of direct versus self report measures for assessing physical activity in adults: a systematic review. International Journal of Behavioral Nutrition and Physical Activity , 5 (1), 56. Quintana J, Fuster JM: (1999) From perception to action: Temporal integrative functions of prefrontal and parietal neurons. Cerebral Co rtex, 9, 213 221.
(% ! ! ! ! ! Ryan, C. M., & Geckle, M. (2000). Why is learning and memory dysfunction in Type 2 diabetes limited to older adults?. Diabetes/metabolism research and reviews , 16 (5), 308 315. Ryan, C. M., & Geckle, M. O. (2000). Circumscribed cognitive dysfunction in middle aged adults with type 2 diabetes. Diabetes care , 23 (10), 1486 1493. Sabia, S., GuÂŽguen, A., Marmot, M. G., Shipley, M. J., Ankri, J., & Singh Manoux, A. (2010). Does cognition predict mortality in midlife? Results from the Whitehall II cohort study. Neurobiology of aging , 31 (4), 688 695. Sallis, J. F., & Saelens, B. E. (2000). Assessment of physical activity by self report: status, limitations, and future directions. Research quarterly for exercise and sport , 71 (sup2), 1 14. Salthouse, T. A. (2004). What and when of cognitive aging. Current directions in psychological science , 13 (4), 140 144. Schupf, N., Tang, M. X., Albert, S. M., Costa, R., Andrews, H., Lee, J. H., & Mayeux, R. (2005). Decline in cognitive and functional sk ills increases mortality risk in nondemented elderly. Neurology , 65 (8), 1218 1226. Shipley, B.A., Der, G., Taylor, M.D., Deary, I.J., 2006. Cognition and all cause mortality across the entire adult age range: health and lifestyle survey. Psychosom. Med. 68, 17 Ã 24. Smith, G. D., Hart, C., Hole, D., MacKinnon, P., Gillis, C., Watt, G., ... & Hawthorne, V. (1998). Education and occupational social class: which is the more important indicator of mortality risk?. Journal of epidemiology and community health , 52 (3), 153 160. Smits, C. H., Deeg, D. J., Kriegsman, D. M., & Schmand, B. (1999). Cognitive functioning and health as determinants of mortality in an older population. American journal of epidemiology , 150 (9), 978 986. Snowdon, D. A., Kemper, S. J., Mor timer, J. A., Greiner, L. H., Wekstein, D. R., & Markesbery, W. R. (1996). Linguistic ability in early life and cognitive function and Alzheimer's disease in late life: findings from the Nun Study. Jama , 275 (7), 528 532. Sofi, F., Abbate, R., Gensini , G. F., & Casini, A. (2010). Accruing evidence on benefits of adherence to the Mediterranean diet on health: an updated systematic review and meta analysis. The American journal of clinical nutrition , 92 (5), 1189 1196. Suchy Y, Blint, A, Osmon DC: (1997) Behavioral Dyscontrol Scale: Criterion and predictive validity in an inpatient rehabilitation unit population. The Clinical Neuropsychologist, 11, 258 265.
(& ! ! ! ! ! Swan, G.E. & Lessov Schlaggar, C.N. (2015). The effects of tobacco smoke on cognition and the brai n. In Waldstein, S. R. & Elias, M.F. (Eds.), Neuropsychology of Cardiovascular Disease (3 34). New York: Psychology Press. Tran, D., Baxter, J., Hamman, R. F., & Grigsby, J. (2014). Impairment of executive cognitive control in type 2 diabetes, and its ef fects on health related behavior and use of health services. Journal of behavioral medicine , 37 (3), 414 422. US Department of Agriculture, US Department of Health and Human Services. Dietary guidelines for Americans, 2010. Washington, DC: US Government Printing Office; 2012. Available at http://www.health.gov/dietaryguidelines/2010.asp Usual Dietary Intakes: Food Intakes, U.S. Population, 2007 10. Epidemiolog y Research Program Web site. National Cancer Institute. http://epi.grants.cancer.gov/diet/usualintakes/pop/2007 10/. Updated May 20, 2015 . Accessed August 24, 2015 . Whalley, L. J., Starr, J. M., Athawes, R., Hunter, D., Pattie, A., & Deary, I. J. (2000). Childhood mental ability and dementia. Neurology , 55 (10), 1455 1459. Whalley, L. J., & Deary, I. J. (2001). Longitudinal cohort study of childhood IQ and survival up to age 76. Bmj , 322 (7290), 819. Willett, W. C., Sacks, F., Trichopoulou, A., Drescher, G ., Ferro Luzzi, A., Helsing, E., & Trichopoulos, D. (1995). Mediterranean diet pyramid: a cultural model for healthy eating. The American journal of clinical nutrition , 61 (6), 1402S 1406S. Wilson, R. S., Beckett, L. A., Bienias, J. L., Evans, D. A., & Bennett, D. A. (2003). Terminal decline in cognitive function. Neurology , 60 (11), 1782 1787. Wilson RS, Beck TL, Bienias JL, Bennett DA. Terminal cognitive decline: accelerated loss of cognition in the last years of life. Psychosom Med 2007;69:131 Ã 7. W inkleby, M. A., Jatulis, D. E., Frank, E., & Fortmann, S. P. (1992). Socioeconomic status and health: how education, income, and occupation contribute to risk factors for cardiovascular disease. American journal of public health , 82 (6), 816 820. World Hea lth Organization. (2014). Cardiovascular diseases [Fact sheet]. Retrieved from http://www.who.int/mediacentre/factsheets/fs317/en
(' ! ! ! ! ! APPENDIX Estimated metabolic equivalent of task units (METs) for different activities Activity Moderate METs Vigorous METs Vigorous home chores/activities (chopping wood, shoveling snow/sand/gravel, moving furniture, carrying a child >20lbs) 6 Vigorous work activities (carrying over 20lbs, digging ditches, loading trucks, stacking lumber, heavy carpentry, heavy construction, shoveling coal, firefighting, moving or pushing objects >75lbs, horseback riding including trotting, galloping or roping, branding, pulling chicos , shearing sheep, calving, lambing, haying, irrigating) 8 Vigorous jogging, running, hiking (vigorous backpacking, mountain climbing, cross country skiing) 8 Vigorous racquet sports (tennis, badminton, paddleball, racquetball, handball, squash) 8 Vig orous bicycling (riding faster than 10mph, riding an exercise cycle hard, rowing machine, bicycling to work, jumping rope) 8 Vigorous swimming (swimming in a pool, lake or ocean, snorkeling, scuba diving, kayaking) 6 Vigorous dancing (vigorous exercise class, jazzercise, Jane Fonda workout, vigorous disco class, vigorous break dancing, aerobic dancing, intermediate or advanced ballet, jazz dance, vigorous ballroom dancing) 6 Vigorous sports (basketball, football, ice skating, roller skating, water ski ing, alpine/downhill skiing, vigorous martial arts, soccer/rugby, field hockey, ice hockey) 7 Vigorous conditioning exercises (calisthenics, weight lifting, free weight training, nautilus or universal work out) 6 Non vigorous home maintenance (outdoor painting, carpentry, raking, mowing, weeding, gardening, washing car) 5
(( ! ! ! ! ! Non vigorous indoor household chores (mopping floors, scrubbing tile, vacuuming, window cleaning, wallpapering, indoor painting, pushing a stroller with a child, carpentry , lifting and carrying <20lbs including groceries, playing with children moderate effort, caring for dependent adult) 5 Non vigorous work activities (walking and carrying <20lbs, waitressing, household maid services, nursing, auto repair, general carpent ry, packing boxes, stocking, assembly, machined tooling, horseback riding general, feeding livestock) 5 Non vigorous home exercise or calisthenics (non vigorous exercise cycle or rowing, low impact aerobics, water aerobics) 4 Non vigorous sports (softball/baseball, shooting baskets, table tennis, volleyball, horseback riding non galloping, archery, non vigorous canoeing/rowing/sailing, non vigorous dancing, biking <10mph, Tai Chi) 4 Bowling or golf 4 Non vigorous walking or hiking (pleasure wa lking on level ground, walking to/during work, using stairs, non vigorous cross country hiking, stream fishing, hunting) 4 Notes: Interviewers described vigorous activities as activities that increase your heart rate, make you sweat, make you breathe hard or raise your body temperature. Non vigorous activities were described as activities that take some effort, but do not make you sweat as much or raise your heart rate as much as the vigorous activities.