Citation
Improving Screening for Mild Cognitive Impairment and Dementia among Racial and Ethnic Minorities Using the Montreal Cognitive Assessment

Material Information

Title:
Improving Screening for Mild Cognitive Impairment and Dementia among Racial and Ethnic Minorities Using the Montreal Cognitive Assessment
Creator:
Milani, Sadaf Arefi
Publisher:
University of Florida
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Epidemiology
Committee Chair:
STRILEY,CATHERINE L
Committee Co-Chair:
COTTLER,LINDA B
Committee Members:
CHEN,XINGUANG
MARSISKE,MICHAEL

Subjects

Subjects / Keywords:
dementia
epidemiology
ethnicity
hispanic
mci
race
screening

Notes

General Note:
The prevalence of Alzheimer's disease (AD), the leading cause of dementia, is increasing as the population ages. Persons with dementia are more likely to be older adults with fewer years of formal education and members of racial/ethnic minority groups. Mild cognitive impairment (MCI), cognitive impairment that is greater than expected for one's age but is not quite as severe as dementia, can be used as an early intervention point, given that 5-10% of individuals with MCI progress to develop dementia. The Montreal Cognitive Assessment (MoCA) was developed to distinguish between Mild Cognitive Impairment (MCI) and normal aging or AD in older adults, however, its applicability in minority populations has not yet been widely studied. This dissertation reports on 1) optimal race-specific cut-offs of the MoCA with the highest predictive validity for MCI and dementia, stratified by education; 2) the discriminative ability of each MoCA subtest to distinguish between normal aging , MCI, and dementia, by race/education; and 3) the effect of relevant risk factors on overall MoCA scores and sub-scores by race/ethnicity. Data come from HealthStreet and the National Alzheimer's Coordinating Center (NACC). HealthStreet is an ongoing community engagement program at the University of Florida which assesses the health concerns of community residents in Florida. Additionally, older adults are asked if they are willing to complete a MoCA (n=241). NACC maintains a database of information, including MoCA scores, collected through 39 past and present Alzheimer's Disease Centers (n=3,985). Results from Chapter 3 show that overall, optimal cutoffs vary by race/ethnicity and education level. Moreover, MoCA subtests do not show clinical utility in distinguishing MCI from normal aging or dementia (Chapter 4). Additionally, in both samples, age and education were the most important predictors of MoCA score (Chapter 5). These finding have implications for future test adaptation in minority populations which will allow health care professionals to appropriately distinguish between normal aging, MCI, and dementia.

Record Information

Source Institution:
UFRGP
Rights Management:
All applicable rights reserved by the source institution and holding location.
Embargo Date:
8/31/2019

Downloads

This item has the following downloads:


Full Text

PAGE 1

IMPROVING SCREENING FOR MILD COGNITIVE IMPAIRMENT AND DEMENTIA AMONG RACIAL AND ETHNIC MINORITIES USING THE MONTREAL COGNITIVE ASSESSMENT By SADAF A REFI MILANI A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2018

PAGE 2

2018 SADAF A REFI MILANI

PAGE 3

To my parents and my sister

PAGE 4

4 ACKNOWLEDGMENTS I would like to acknowledge my mentor, Dr. Catherine Striley, who provided me with unwa vering support throughout my Ph D program I would also like to thank my committee members Dr. Linda B. Cottler, Dr. Xinguang Chen, and Dr. Michael Marsiske for their invaluable feedback. This work would not be possible without their support. I would like to acknowledge the administrative assistanc e of Becca Pieters, Tamara Millay, and Abigail Hummel and the University of Florida Graduate School and Department of Epidemiology for funding my PhD. I would like to thank my parents for emphasizing the importance of education and providing me with every opportunity that they could I also thank my sister, Sahar, for proofreading all of my essays at the last minute throughout my schooling and teaching me the importance of commas. I would also like to acknowledge both old and new friends who have supported me throughout this process Elizabeth Gray, Mirsada Serdarevic, Georg Scheutz, Andrew J. Garcia, Romain Gloaguen, Faith Cayobit, Tessa Frohe, and Denny Fe Agana Garcia. You all ha ve made my experience in Gainesville a very memorable one.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 11 ABSTRA CT ................................ ................................ ................................ ................... 12 1 INTRODUCTION ................................ ................................ ................................ .... 14 Epidemiology of Dementia ................................ ................................ ...................... 14 Racial and Ethnic Differences ................................ ................................ ................. 15 Educational Attainment ................................ ................................ ........................... 16 The Montreal C ognitive Assessment ................................ ................................ ...... 17 Conceptual Framework ................................ ................................ ........................... 19 2 GENERAL METHODS ................................ ................................ ............................ 23 Data Sourc es and Study Sample ................................ ................................ ............ 23 ................................ ................. 23 HealthStreet ................................ ................................ ................................ ..... 23 Specific Aims ................................ ................................ ................................ .......... 24 Overall Study Measures ................................ ................................ .......................... 25 NACC Measures ................................ ................................ ................................ ..... 25 Clinician Diagnosis ................................ ................................ ........................... 25 MoCA Score and Subscore ................................ ................................ .............. 25 Covariates ................................ ................................ ................................ ........ 26 Sociodemographics ................................ ................................ ................... 26 Health variables ................................ ................................ ......................... 26 HealthStreet Measures ................................ ................................ ........................... 27 MoCA S core and Subscore ................................ ................................ .............. 27 Covariates ................................ ................................ ................................ ........ 27 Sociodemographics ................................ ................................ ................... 27 Dementia risk factors ................................ ................................ ................. 27 Methods of Analysis ................................ ................................ ................................ 28 Chapter 3 ................................ ................................ ................................ .......... 28 Chapter 4 ................................ ................................ ................................ .......... 29 Chapter 5 ................................ ................................ ................................ .......... 30 3 OPTIMAL CUTOFFS FOR THE MONTREAL COGNITIVE ASSESSMENT VARY BY RACE AND ETHNICITY AMO NG NACC PARTICIPANTS .................... 37

PAGE 6

6 Background ................................ ................................ ................................ ............. 37 Methods ................................ ................................ ................................ .................. 40 Population ................................ ................................ ................................ ........ 40 Measures ................................ ................................ ................................ .......... 41 Data Analysis ................................ ................................ ................................ ... 41 Results ................................ ................................ ................................ .................... 42 Discussion ................................ ................................ ................................ .............. 45 4 SUBTEST ANALYSIS OF THE MONTREAL COGNITIVE ASSESSMENT, BY RACE AND ETHNICITY, AMONG NACC PARTICIPANTS ................................ .... 62 Background ................................ ................................ ................................ ............. 62 Methods ................................ ................................ ................................ .................. 64 Population ................................ ................................ ................................ ........ 64 Measures ................................ ................................ ................................ .......... 64 Data Analysis ................................ ................................ ................................ ... 66 Results ................................ ................................ ................................ .................... 66 Discussion ................................ ................................ ................................ .............. 69 5 INFLUENCE OF RELEVANT RISK FACTORS FOR DEMENTIA ON MONTREAL COGNITIVE ASSESSMENT SCORES AMONG NACC AND HEALTHSTREET PARTICIPANTS ................................ ................................ ......... 82 Background ................................ ................................ ................................ ............. 82 Methods ................................ ................................ ................................ .................. 84 Population ................................ ................................ ................................ ........ 84 HealthStreet ................................ ................................ ............................... 84 ................................ .... 85 Measures ................................ ................................ ................................ .......... 86 HealthStreet ................................ ................................ ............................... 86 NACC ................................ ................................ ................................ ......... 87 Data Analysis ................................ ................................ ................................ ... 88 Results ................................ ................................ ................................ .................... 88 HealthStreet ................................ ................................ ................................ ..... 88 NACC ................................ ................................ ................................ ............... 89 Discussion ................................ ................................ ................................ .............. 90 6 CONCLUSIONS ................................ ................................ ................................ ..... 97 Main Findings ................................ ................................ ................................ ......... 97 Conceptual Framework ................................ ................................ ......................... 103 Strengths and Limitations ................................ ................................ ..................... 104 NACC ................................ ................................ ................................ ............. 104 HealthStreet ................................ ................................ ................................ ... 105 Public Health Consequence and Future Research ................................ ............... 106 Future research ................................ ................................ ................................ .... 108

PAGE 7

7 LIST OF REFERENCES ................................ ................................ ............................. 110 BIOGRAPH ICAL SKETCH ................................ ................................ .......................... 117

PAGE 8

8 LIST OF TABLES Table page 2 1 Overall total MoCA score (out of 30 total) and official individual subtest and item total possible scores ................................ ................................ ................... 31 2 2 Questions used in analyses ................................ ................................ ................ 32 3 1 Characteristics of NACC study participants, by race/ethnicity (n=3,895). ........... 51 3 2 Test characteristics of the Montreal Cognitive Assessment, using the current cutoff of 26 for detection of MCI or dementia, by race/ethnicity, among NACC study participants. ................................ ................................ ............................... 52 3 3 Optimal MoCA cutoffs to distinguish between normal aging (NA) and MCI by race/ethnicity and education level, a mong NACC st udy participants with normal aging or MCI ................................ ................................ ........................... 53 3 4 Optimal MoCA cutoffs to distinguish between MCI and dementia by race/ethnicity and education level, among NACC study participants with MCI or dementia (DM) ................................ ................................ ................................ 56 3 5 Average MoCA score among non Hispanic White NA CC participants, by decade of age and education level ................................ ................................ ..... 59 3 6 Average MoCA score among non Hispanic Black NACC participants, by de cade of age and education level ................................ ................................ ..... 60 3 7 Average MoCA score among Hispanic NACC participants, by decade of age and education level ................................ ................................ ............................. 61 4 1 Characteristics of NACC study participants included in these analyses, by race/ethnicity ................................ ................................ ................................ ...... 75 4 2 Summary of Stepwise multinomial logistic regression for MoCA subtests and clinician diagnosis of normal aging (NA) or dementia (DM) among overall populati on vs. Non Hispanic Whites ................................ ................................ ... 76 4 3 Summary of Stepwise multinomial logistic regression for MoCA subtests and clinician diagnosis of normal aging ( NA ) or dementia (DM) among overall population vs. Non Hispanic Blacks ................................ ................................ ... 77 4 4 Summary of Stepwise multinomial logistic regression for MoCA subtests and clinician diagnosis of normal aging (NA) or dementia (DM) among overall population vs. Hispanics ................................ ................................ ..................... 78

PAGE 9

9 4 5 Adjusted odds ratios and 95% confidence intervals for the association between MoCA items and clinician diagnosis of normal aging or dementia (compared to MCI), adjusted for age and education, by race/ethnicity. .............. 79 5 1 Sociodemographic characteristics and frequency of risk factors among HealthStreet participants who have compl eted a MoCA ................................ ..... 94 5 2 Linear regression coefficients for risk factors associated with MoCA scores among HealthStreet participants, ranked by standardized betas ........................ 94 5 3 Linear regression coefficients for risk factors associated with MoCA scores among HealthStreet participants from stepwise regression, ranked by standardized betas ................................ ................................ ............................. 95 5 4 Sociodemographic characteristics and frequency of risk factors among NACC participants who hav e completed a MoCA at their baseline visit (n=3180) ................................ ................................ ................................ ............. 95 5 5 Linear regression coefficients for risk factors associated with MoCA s cores among NACC participants, ranked by standardized betas ................................ 96 5 6 Linear regression coefficients for risk factors associated wit h MoCA scores among NACC participants from stepwise regression, ranked by standardized betas ................................ ................................ ................................ ................... 96

PAGE 10

10 LIST OF FIGURES Figure page 3 1 Classification accuracy of optimal MoCA cutoffs for MCI, by race/ethnicity among NACC study participants with normal aging or MCI ................................ 54 3 2 Optimal MoCA cutoffs to distinguish between normal aging and MCI by race/ethnicity and education level, among NACC study participants with normal aging or MCI ................................ ................................ ........................... 55 3 3 Classification accuracy of optimal MoCA cutoffs for dementia, by race/ethnicity, among NACC study participants with MCI or dementia ............... 57 3 4 Optimal MoCA cutoffs to distinguish between MCI and dementia by race/ethnicity and education level, among NACC s tudy participants with MCI or dementia ................................ ................................ ................................ ......... 58

PAGE 11

11 LIST OF ABBREVIATIONS AD D isease ADC AOR Adjusted Odds Ratio APOE Apolipoprotein E CHW Community Health Worker CI Confidence Interval D M Dementia HSt HealthStreet MCI Mild Cognitive Impairment MoCA Montreal Cognitive Assessment MMSE Mini Mental State Examination NA Normal Aging NACC Coordinating Centers NH Non Hispanic UDS Uniform Data Set UF University of Florida US United States

PAGE 12

12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy IMPROVING SCREENING FOR MILD COGNITIVE IMPAIRMENT AND DEMENTIA AMONG RACIAL AND ETHNIC MINORITIES USING THE MONTREAL COGNITIVE ASSESSMENT By Sadaf Arefi Milani August 2018 Chair: Catherine W. Striley Major: Epidemiology (AD) the leading cause of dementia, is increasing as the population ages. Persons with d ementia are more likely to be older adults with fewer years of formal education and members of racial/ethnic minori ty groups. Mild cognitive impairment (MCI), cognitive impairment that is greater than intervention point, given that 5 10% of individuals with MCI progress to develop de mentia. The Montreal Cognitive Assessment (MoCA) was developed to distinguish between Mild Cognitive Impairment (MCI) and normal aging or AD in older adults, however, its applicability in minority populations has not yet been widely studied. This dissertat ion reports on 1) optimal race specific cut offs of the MoCA with the highest predictive validity for MCI and dementia, stratified by education; 2 ) the discriminative ability of each MoCA subtest to distinguish between normal aging MCI, and dementia, by race/education; and 3) the effect of relevant risk factors on overall MoCA scores and sub scores by race/ethnicity.

PAGE 13

13 (NACC) HealthStreet is an ongoing community engagement program at the University of Florida which assesses the health concerns of community residents in Florida. Additionally, older adults are asked if they are willing to complete a MoCA (n=241). NACC maintains a database of information, including MoCA scores, collec ted through 39 past and present Alzh show that overall, optimal cutoffs vary by race/ethnicity and education level. Moreover MoCA subtests do not show clinical utility in distinguishing MCI from nor mal aging or dementia (Chapter 4). Additionally, in both samples, age and education were the most important predictors of MoCA score (Chapter 5). These finding have implications for future test adaptation in minority populations which will allow health car e professionals to appropriately distinguish between normal aging MCI, and dementia

PAGE 14

14 CHAPTER 1 INTRODUCTION Epidemiology of Dementia Dementia is a syndrome, or group of symptoms, primarily characterized by loss of cognitive function [1] Symptoms include problems with memory loss, language, and disease, the most common form of dementia, accounts for about 60 80% of dementia cases [1 ] AD is characterized by the presence of beta amyloid plaques and neurofibrillary tangles that result in damage and eventual destruction of neurons in the brain [1] Other forms of dementia include vascular dementia, dementia with Lewy bodies, mixed demen [1] Risk factors for dementia include both potentially modifiable and non modifiable factors. Non modifiable risk factors include older age, familial history of dementia, and the presence of the Apolipoprotein E (APOE) e4 gene [1] Potentially m odifiable risk factors include low economic status, educational attainment, social and cognitive engagement, physical activity, traumatic brain injury, depression, and vascular risk factors such as hype rtension, smoking, and diabetes [1] Mild cognitive impairment less severe than dementia, is an intermediate stage between normal aging and dementia [2] A review of 41 studies found that less than half of individuals with MCI will develop dementia [3] T he prevalence of dementia is expected to increase both globally and within the U.S. as the number of individuals aged 60 and over increase [4] ; however, some

PAGE 15

15 literature suggests that there may be a decrease in the prevalence of dementia in the U.S. partially due to an increase in educational attainment and better treatment of vascular risk factors [5] As of 2017 about 5.5 million people in the U.S. were living with AD a nd this number is projected to rise [1] This increase in individuals living with AD and dementia will lead to a large social and economic burden on individuals living with dementia, their families, health care systems, and society as a whole [6] Screening for MCI and dementia will improve quality of life for individuals who are living with these disorders. About half of individuals with AD have been diagnosed, and of these, only 33% are aware of their diagnosis [7] When including caregivers, awar eness of the diagnosis only increased to 45% [7] This may be due to physician concerns about stigma or the usefulness of an early diagnosis as well as a lack of documentation of the diagnosis in medical records [7] Early detection of MCI and dementia giv es individuals time to express their wishes, build a care team, make legal and financial arrangements, and create advance health directions before their impairment becomes too severe [7] Moreover, early detection helps physicians manage conditions that often occur with old age [7] Racial and Ethnic Differences A large racial/ethnic disparity exists in the prevalence of dementia ; African Americans are two times as likely as older non Hispanic Caucasians to develop dementias, including AD while older Hispanics are about one and a half times as likely as older non Hispanic Caucasians to develop dementia [1,8] This difference is likely attributable to the higher burden of risk factors for dementia, such as hypertension and diabetes, and di fferences in socioeconomic factors in minority populations [1,8 10] However, measurement bias may exist in the quantification of dementia burden and risk

PAGE 16

16 factors among minority groups, since racial bias have been found to exist in screening and diagnosing dementia [9] The number of racial/ethnic minority group members has increased in the U.S. and is projected to continue to increase, representing a large group of individuals who are at a h igh risk for developing dementia [9] Educational A ttainment Indi viduals with fewer years of formal education have a greater risk of dementia and AD compared to individuals with more years of formal education [11 14] Contador and colleagues found that the risk of dementia decreased 7% per year of education [12] This a ssociation has consistently been found in the literature, functioning through several potential mechanisms. Education is associated with other risk factors for dementia, including socioeconomic status, lifestyle factors, such as smoking and physical activi ty, vascular risk factors, and cognitive engagement [15] Overtime educational attainment of the population has increased. However, racial differences still exist; most older African Americans in the U.S. experienced education during the Jim Crow period (until 1954) in the south, where schools were segregated by race During this time, African Americans spent fewer days in school compared to whites but the large gap in education has lessened over time [16] Educational attainment also can affect brain biology, which can impact the pathology of AD. Previous literature has found that higher years of educational attainment are associate d with fewer beta amyloid plaques [17] and the creation of fewer tau peptide neurofibrillary tangles, both hallmarks of AD pathology [18] More years of formal education are also associated with increased cognitive reserve [19 21] The cognitive reserve theory states that individuals with more cognitive reserve are better equipped to handle damage to their brain, delaying the manifestation of

PAGE 17

17 symptoms of dementia [19 21] More years of education are also associated with increased regional cortical thickness [22] and increased cerebral volume [23] both of which are positively associated with cognitive reserve. However, individ uals with more cognitive reserve experience more rapid cognitive decline once their symptoms present compared to individuals with less reserve and earlier onset of symptoms [19] Although years of education has consistently been used as a proxy for cogniti ve reserve, previous work has found that literacy level better predicts cognitive decline, rather than years of education, especially among ethnically diverse populations [24] The Montreal Cognitive Assessment The Montreal Cognitive Assessment (MoCA), dev eloped in 2005 as a screening tool to detect Mild Cognitive Impairment (MCI), has a high sensitivity for detection of MCI in patients who would score as normal on the Mini Mental State Examination (MMSE) [25] It takes approximately 10 minutes to complete and has a total of 30 points [25] The MoCA has subtests on orientation, attention, verbal memory, language, visuospatial, and executive function [25] Using a cutoff score of 26, the MMSE had a sensitivity of 18% for detecting MCI, while the MoCA had a se nsitivity of 90% [25] Although it is widely accepted that the MoCA is a better test for MCI compared to the Mini Mental State Examination (MMSE) [25 27] some studies suggest that a cutoff of 26 is not the best cutoff. A study by Luis and colleagues found that while using the recommended cut off of 26 the MoCA had a sensitivity of 97% and specificity of 35% but when the cut off was lowered to 23 its performance improved with a sensitivity and specificity of 96% [28] Cut off performance has also been foun d to differ depending on the setting in which the MoCA is administered. Since the ROC analysis of the MoCA found that the recommended cutoff score has a high sensitivity at the expense of

PAGE 18

18 specificity, this threshold may be more appropriate in a primary car e setting, while a lower threshold would be more appropriate in a memory disorder clinic [26] A recent systematic review found that a MoCA c ut off of 23, rather than 26, improves overall performance and reduces the false positive rate [29] Additionally, the psychometric properties of MoCA subtests and items have not been widely studied [30] Cecato and colleagues examined MoCA subtests and items individually and found that word repetition, inverse digits, serial 7, phrases, verbal fluency, abstraction, and word recall were markers of early impairment (MCI) and the clock drawing, rhino naming, delayed recall of five words, and orientation were markers of severe impairment (AD) [31] Moreover, Roalf and colleagues found that the clock drawing serial 7s, orientation place, delayed recall, abstraction, naming (rhinoceros), trails, and fluency discriminated between healthy controls and between affected [32] Another stu dy looking for the three MoCA items with the largest effect sizes, found that serial 7s and delayed recall were the best items for distinguishing MCI from normal a ging while serial 7s, delayed recall, and orientation were the best items at discriminating A D from MCI [30] Moreover, its application in diverse populations has not been widely studied. Most of the studies examining the MoCA and its subtests have been limited to Caucasian samples Those that have include d minorities have found that the cutoffs o riginally established may not be appropriate for these groups [33 35] Part of the reason for differences in performance may be due to differences is underlying educational attainment. To account for differences in educational attainment t hat may

PAGE 19

19 affect Mo CA scores, a one point education adjustment is added to the total score of individuals with 12 or less years of education [25] However, previous research has found that the one point education adjustment can affect the reliability of the MoCA in some grou ps [36] Additionally, in one of the very few studies with a diverse sample, Rossetti and colleagues found that even after applying the education adjustment to MoCA scores, over half of the sample (62%) scored below the cutoff score of 26 from the original ly published study suggesting that this cutoff and education adjustment may not be appropriate [37] One study found that the MMSE had an optimal specificity when the cutoff was lowered and modified by education [38] however this should be more closely ex amined with the MoCA in minority groups The se results suggest a need to further examine and understand the performance of the MoCA in diverse populations, inc luding in minority individuals. Conceptual Framework lifecourse when interpreting performance on neuropsychological tests, Glymour and Manly created a model to explain the pathways linking ra ce/ethnicity and cognitive aging [16] Rel evant to this dissertation, this model explains that performance on circumstances alone, the focus of most literature R acial/ethnic disparities persist because of an uneven distribution of risk factors for cognitive decline, as well as other health c onditions, throughout the lifec ourse of racial/ethnic minority group members. Distally socially patterned mediators include geographic exposures/immigration, socioeconomic position, discrimination, group resou rces/ties, and group identity. Individual mediators including material adversity, psychological or psychosocial

PAGE 20

20 stressors, cognitive engagement, and test taking skills. Proximal mediators include physical health, health behaviors, and medical access. In this dissertation, informed by the original Glymour model, we will test v ariables available in the data sets proposed for this study including race/ethnicity and education (distal social patterned mediators), individual mediators (family history) and proximal mediators (health history and health behaviors: high blood pressure, diabetes, traumatic brain injury, depression, and smoki ng status) Distal social patterned mediators include immigration and socioeconomic position Immigration is important to consider when examining cognitive status; international immigrants tend to have better health than comparable non immigrants, especially among blacks [16] Moreover, international immigration is likely cognitively beneficially because individuals who immigrate usually have to navigate a new environment in a second language and will gain bilingual skills [16] Socioeconomic position, including factors such as education and occupation, differs greatly across racial/ethnic groups and is associated with a range of cognitive outcomes. Low educational attainment is associated with dementia [11 14] ; occupation p redicts cognitive decline in the elderly [16] These two factors are especially important because the cognitive reserve theory suggests that individuals with more educational, occupational, cognitive, and social engagement are better equipped to handle damage to their brain, delaying the development of functional deficits [19,20] Individual mediators, including material adversit y, link social disadvantages to individual level factors that affect cognitive aging. Poor nutrition, especially in early life, is one factor that affects cognitive impairment in later years [16] Although less widely studied, the uggests that the cumulative effect of poor nutrition and other

PAGE 21

21 adverse conditions throughout the lifecourse is important when examining cognitive aging [16,39] Proximal mediators included in this model are medical access, physical health, and health behav iors. Disparities exist in access to medical care across races/ethnicities due to many barriers including financial and transportation difficulties, geographic access, and cultural and language barriers [16] A large number of risk factors of dementia delineate this group of proximal mediators within the physical health and health behaviors grouping, including hypertension, diabetes, depression and smoking. The accumulation in cognitive chan ge and, potentially, a diagnosis of cognitive impairment. T his dissertation will focus on race/ethnicity and diagnosis of cognitive impairment, along with risk factors that are mediators on the pathway between race/ethnicity and cognitive aging. We will fo cus on screening for dementia and MCI using the MoCA, with a focus on low educational attainment and race/ethnicity, s ince the highest population attributable risk of potentially modifiable risk factors for AD worldwide was low educational attainment (lowe r secondary education or less) [13] These analyses will examine these factors, focusing on race/ethnicity and educational attainment, to understand if MoCA overall scores and sub scores differ when considering these sociodemographic factors. Additionally, this dissertation will focus on the effect of risk factors on overall MoCA score in two unique samples. By understanding the performance of the MoCA and its subtests among minority members, health professionals can appropriately distinguish between normal aging MCI, and dementia in these populations. When appropriate norms are applied, the correct

PAGE 22

22 intervention or care can be prescribed. Differences seen across populations may also point to different prevention strategies.

PAGE 23

23 CHAPTER 2 GENERAL METHODS Data Sources and Study Sample Coordinating Center and HealthStreet cohort to examine the use of the Montreal Cognitive Assessment (MoCA) in minority populations. information collected through 39 past and present National Institute on Aging funded formation on sociodemographic characteristics, family history, dementia history, and clinical evaluation findings are collected. The data set we used, the Uniform D ata S et (UDS), includes subjects enrolled at ADCs since 2005 wi th a range of cognitive statu s as measured through clinician assessment and neuropsychological tests. Typically, subjects are recruited through clinician referral, self referral by patients or family members, or recruitment through community organizations. Written informed consent is colle cted from all participants and co participants, usually a close family member or friend of the participant. Data is collected by clinicians and their trained staff through a standardized evaluation of subjects enrolled in ADCs and their family members [40] Subjects are followed stage condition keeps them from participating they are followed up for autopsy after they discontinue their annual visits [41,42] HealthStreet HealthStreet, a community engagement program at th e University of Florida started in 2011, has engaged over 10,000 community members using the Community

PAGE 24

24 Health Worker (CHW) model. CHWs are trained and trusted members of the communities they serve. They spend every day engaging community members and assess ing their health conditions and health concerns [43,44] CHWs interview individuals using the HealthStreet Intake Form, an in person questionnaire that takes about 20 minutes to complete. This form contains questions on sociodemographic characteristics, re search interest s and attitudes towards research health conditions, health concerns, as well as contact information so that partici pants can be reached for follow up calls. Older adults aged 65 and over are also asked if th ey are willing to complete a MoCA. If consent is given, the information collected through this interview can be shared with other researchers so that the community members can be informed about and linked to opportunities to participate in health research relevant to them Specific Aims The NACC and HealthStreet datasets provide comprehensive information with which we will examine the performance of the MoCA among racial/ethnic minority members through the following aims: 1. Identify race specific optimal cutoff values of the MoCA when stratified by diagnosis as the gold standard (NACC) a) H1A: M inority groups will require lower cutoffs to distinguish between normal aging and MCI, and further, between MCI and dementia b) H2A: C utoffs will need to decrease as educational attainment decreases to ensure optimal MoCA performance. 2. Analyze the discriminat ive ability of each MoCA subtest and item to distinguish between 1) normal aging and M CI and 2) MCI and dementia by race/ethnicity, when controlling for age and education level (NACC)

PAGE 25

25 a) H2: MoCA subtest scores of people who identify as a minority will differentially discriminate between normal aging, MCI, and dementia. 3. Evaluate the effect of risk factors, including family history, high blood pressure, diabetes, traumatic brain injury, depression, and smoki ng status on overall MoCA score by race/ethnicity in a community dwelling sample (HSt) and a large AD research database (NACC). a) H3 : Relevan t risk factors for dementia will be associated with a lower overall MoCA score. Overall Study Measures Race and ethnicity variables were combined to create a three level race/ethnicity variable categorized as non Hispanic White, non Hispanic Black, and Hispanics. Total MoCA score was recoded as a continuous variable 0 30, with lower values signifying more i mpairment. MoCA subscores were also recorded as quantitative variables. Table 2 1 shows the points associated with each MoCA subscore and item. Education was categorized as high ), some college (13 to 16 years), and more than colle ge (>16 years). NACC Measures Table 2 2 lists NACC variable name, question, original coding, and coding used for each measure in these analyses. Clinician Diagnosis Clinician diagnosis was categorized as normal cognition impaired not MCI, MCI, or dementi a. Participants who were diagnosed as impaired not MCI were dropped from these analyses due to uncertainty over the source of their impairment. MoCA Score and S ubscore Overall MoCA score was recorded as a continuous variable with values from 0 30. Raw MoCA scores were used. MoCA subscores were also recorded as quantitative

PAGE 26

26 variables. Lower scores signify more impairment. Table 2 1 shows how many points are associated with each subtest and item. Covariates Sociodemographics Sociodemographic information for t hese analyses include age, sex, race, ethnicity, and education. Age was left as a continuous variable but was also categorized by decade of age. Sex was categorized as female or male. Race and ethnicity were combined into non Hispanic White, non Hispanic B lack, and Hispanic. Education was originally recorded as years of education but was categorized into high school or less ), some college (13 to 16 years), and more than college (>16 years) for these analyses. Health variables Dementia risk facto rs include high blood pressure, diabetes, traumatic brain injury, depression, smoking status, and family history of AD. Presence of hypertension, judgement, following a medical history interview with the subject and co participant. Hypertension, diabetes, traumatic brain injury, and depression were originally recorded as absent, recent/active, remote/inactive, and unknown; however, they were combined into a two level lifetime (r ecent/active and remote/inactive) history of condition or no lifetime history of condition variable. The basis of our smoking status variable came this was coded as yes participants were categorized as lifetime smokers. Family history was coded as yes if participants reported at least one first degree family member with cognitive impairment.

PAGE 27

27 HealthStreet Measures Table 2 2 lists HealthStreet variable name, question, ori ginal coding, and coding used for each measure in these analyses. MoCA Score and Subscore Overall MoCA score was recorded as a continuous variable with values from 0 30. Raw MoCA scores were used. MoCA subscores were also recorded as quantitative variables. Lower scores signify more impairment. Table 2 1 shows how many points are associated with each subtest and item. Covariates Sociodemographics Sociodemographic information for these analyses include age, sex, r ace, ethnicity, education, and food insecurity. Age was left as a continuous variable. Sex was categorized as female or male. Race and ethnicity were categorized as non Hispanic White, non Hispanic Black, and Hispanic. Education was originally recorded as last grade completed but was categorized it into high ), some college (13 to 16 years), and more than college (>16 years) for these analyses. Food e there been times in the last 12 months when you did not have enough money to buy insecure. Dementia risk factors Dementia risk factors assessed include history of high blood pressure, diabetes, depression, traumatic brain injury, smoking, and family history of AD or other memory

PAGE 28

28 ever been told you had, or have you ever had a problem with (co blood pressure, diabetes, depression, and serious head injury, those who reported yes were coded as having a lifetime history of either high blood pressure, diabetes, ever smoked AD or a memory disorder were coded as having a family history of AD or other memory disorder. Methods of Analysis Analyses in this dissertation were conducted used SAS software version 9.4 of the SAS System for Windows (SAS Institute Inc., Cary, NC, US A) and MedCalc, version 18.5 (MedCalc Software bvba, Ostend, Belgium; https://www.medcalc.org; 2016) P values less than 0.05 were considered to be statistically significant for all analyses in this dissertation. Chapter 3 Individuals from NACC who complet ed a MoCA at their baseline visit and were included in the March 2018 data freeze were included in these analyses. Moreover, only individuals who reported either being White or Black, reported their ethnicity (Hispanic/Latino or non Hispanic/Latino), and r eported their years of education were included in these analyses. Our final sample was 3,895. A NACC data request form was submitted by the author, and a personalized dataset including only NACC participants who have completed a MoCA was generated by NACC and provided to the author.

PAGE 29

29 Descriptive statistics were calculated for sociodemographic characteristics. Chi square Tests of Independence were used to test differences in sociodemographic characteristics by race/ethnicity. Test measures including specifici ty, sensitivity, positive predictive value, and negative predictive value were calculated for the current MoCA cutoff of 26. Optimal cutoffs of the MoCA and areas under the ROC curve (AUC) were equal to the sensitivity plu s specificity minus one [45] Cutoffs were calculated for distinguishing 1) MCI from normal aging and 2) MCI from dementia. Cutoffs were stratified by education. Confidence intervals for the AUCs were calculated and bootstrapped confidence intervals for cutoffs were generated using 100,000 iterations. Chapter 4 Individuals from NACC who completed a Mo CA at their baseline visit, were included in the March 2018 data freeze, reported either being White or Black, reported their ethnicity (Hispanic/Latino or non Hispanic/Latino), and reported their years of education were included in these analyses. Our fin al sample included 3,895 people, the same sample used in Chapter 3 Descriptive statistics were calculated for sociodemographic variables and Chi Square Tests of Independence were used to test differences in sociodemographic characteristics by race/ethnic ity. Stepwise multinomial logistic regression was used to establish which subtests best predict clinician diagnosis, by race/ethnicity. Subtests were deconstructed into items and multinomial logistic regression was used to calculate the odds ratios of the association between each item and clinician diagnosis, adjusting for age and education. Analyses were conducted among the whole sample and then by race/ethnicity.

PAGE 30

30 Chapter 5 Individuals from NACC who completed a MoCA at their baseline visit, were included i n the March 2018 data freeze, reported either being White or Black, reported their ethnicity (Hispanic/Latino or non Hispanic/Latino), and who were not missing values for relevant risk factors were included in these analyses. Our final sample from NACC inc luded 3,180 participants Participants from HealthStreet who completed a MoCA reported being White or Black, reported their ethnicity (Hispanic/Latino or non Hispanic/Latino), and who were not missing values for relevant risk factors were included in thes e analyses. Our final samp le from HealthStreet included 21 1 participants Descriptive statistics were calculated for sociodemographic characteristics in each sample. Linear regression was used to determine the effect of each dementia r isk factors on total MoCA score among 1) the HealthStreet sample and 2) the NACC sample. Then, stepwise regression was used to establish the most important sociodemographic characteristics and risk factors for prediction of MoCA score.

PAGE 31

31 T able 2 1 Overall total MoCA scor e (out of 30 total) and official individual subtest and item total possible scores Subtest/Item Points Visuospatial/Executive Trail 1 Cube 1 Clock Contour 1 Clock Number 1 Clock Hands 1 Naming 3 Attention Digits 2 Letters 1 Serial 7s 3 Language Repetition 2 Fluency 1 Abstraction 2 Delayed Recall 5 Orientation Date 1 Month 1 Year 1 Day 1 Place 1 City 1 Overall Score 30

PAGE 32

32 Table 2 2. Q uestions used in analyses Variable Dataset Question Original Coding New Coding Age NACC Age Continuous < 40 40 49 50 59 60 69 70 79 80 89 >= 90 And continuous HSt Age Continuous Continuous Sex NACC 1=Male 2=Female Male Female HSt Sex 1= Male 2= Female 3= Transgender Male Female Race/ Ethnicity NACC Race/ethnicity: 1= American Indian/Alaskan Native 3= Asian 4= Black/African American 6= Native Hawaiian/Pacific Islander 7= White 9= Other White/non Hispanic Black/non Hispanic Hispanic Are you Hispanic or Latino? 1= No 5= Yes

PAGE 33

33 Table 2 2. Continued Variable Dataset Question Original Coding New Coding HSt Race/ethnicity: 1= American Indian/ Alaskan Native 3= Asian 4= Black/African American 6= Native Hawaiian/Pacific Islander 7= White 9= Other White/non Hispanic Black/non Hispanic Hispanic Are you Hispanic or Latino? 1= No 5= Yes Education NACC Last grade completed: Continuous 13 16 years >16 years And continuous HSt Last grade completed: Continuous Continuous Clinician Diagnosis NACC Does the subject have normal cognition and normal behavior? Does the subject meet the criteria for dementia? If the subject does not have normal cognition or behavior and is not clinically demented, indicated the type of cognitive impairment. Various skip patterns for clinician evaluation of cognitive status: Normal Impaired, not MCI MCI Dementia Normal Aging MCI Dementia HSt Not available

PAGE 34

34 Table 2 2. Continued Variable Dataset Question Original Coding New Coding Food Insecurity NACC Not available HSt Have there been times in the last 12 months when you did not have enough money to buy food that you or your family needed? 1= No 5= Yes No Yes High Blood Pressure NACC Record the presence or absence of a history of these conditions at this visit, as determined by the the medical history interview with the subject and co participant: high blood pressure 0= Absent 1= Recent/active 2= Remote/inactive No= Absent Yes= Recent/Active and Remo te/Inactive HSt Have you ever been told you had, or have you ever had a problem with high blood pressure? 1= No 5= Yes No Yes Diabetes NACC Record the presence or absence of a history of these conditions at this visit, as determined by the the medical history interview with the subject and co participant: diabetes 0= Absent 1= Recent/active 2= Remote/inactive No= Absent Yes= Recent/Active and Remote/Inactive HSt Have you ever been told you had, or have you ever had a problem with diabetes? 1= No 5= Yes No Yes

PAGE 35

35 Table 2 2. Continued Variable Dataset Question Original Coding New Coding Traumatic Brain/Serious Head Injury NACC Record the presence or absence of a history of these conditions at this visit, as determined by the the medical history interview with the subject and co participant: traumatic brain injury 0= Absent 1= Recent/active 2= Remote/inactive No= Absent Yes= Recent/Active and Remote/Inactive HSt Have you ever been told you had, or have you ever had a problem with serious head injury? 1= No 5= Yes No Yes Depression NACC Indicate the presence of any diagnoses: active depression. 0= No 1= Yes Depression: No Yes HSt Have you ever been told you had, or have you ever had a problem with depression? 1= No 5= Yes No Yes Smoking NACC Has the subject smoked more than 100 cigarettes in his/her life? 1= No 5= Yes Current smoking: No Yes HSt Have you ever smoked cigarettes? 1= No 5= Yes Lifetime smoking: No Yes

PAGE 36

36 Table 2 2. Continued Variable Dataset Question Original Coding New Coding Family History NACC Are there affected first degree relatives (biological parents, full siblings, or biological children)? 0= No 1= Yes Family History: No Yes HSt Have you or your parents been or another memory disorder? If yes, aged when diagnosed? A1= Self, AD A2= Self, dementia A3= Self, other B1= Mother, AD B2= Mother, dementia B3= Mother, other C1= Father, AD C2= Father, dementia C3= Father, other Age of diagnosis recorded as a continuous variable Family History: No Yes

PAGE 37

37 CHAPTER 3 OPTIMAL CUTOFFS FOR THE MONTREAL COGNITIVE ASSESSMENT VARY BY RACE AND ETHNICITY AMONG NACC PARTICIPANTS Background the presence of beta amyloid plaques and neurofibrillary tangles that damage neurons in the brain [1] By 2017, AD affected approximately 5.5 million Americans; this number is projected to increase to as high as 16 million by 2050 [1] Risk factors for AD include non modifiable factors, such as older age, family history, and presence of the Apolipoprotein E (APOE) e4 gene, and potentially modif iable risk factors, including low educational attainment, low socioeconomic status, hypertension, smoking, diabetes, depression, and low social and cognitive engagement [1,46,47] Mild cognitive impairment (MCI), the stage between normal aging and dementia is not as severe as dementia, but individuals have greater cognitive impairment than is expected for their age [2,48] MCI can be used for early detection and prevention of progression to dementia. By diagnosing MCI in older people, health care professio nals can address factors such as controlling cardiovascular risk factors, increasing exercise, and initiating cognitive training, that may reduce progression from MCI to AD [49] Racial/ethnic minorities are disproportionately at risk for dementia; African dementias than their non Hispanic White counterparts [1,8] This difference is likely due to disparities in risk factors, such as hypertension, diabetes, and socioecon omic factors that occur disproportionally among minority populations compared to the majority population [1,8 10] However, one must consider the potential of measurement bias when examining the burden of dementia among minority populations, since racial

PAGE 38

38 b iases exist in screening and diagnosis of dementia [9] This is a timely topic because the number of racial/ethnic minority group members has increased and will continue to increase in the United States [9] Reflecting on the conceptual framework of this d issertation, educational attainment is important to consider because it is known to affect a life course [16] Among minority populations, low e ducational attainment is a particularly important risk factor for dementia; individuals with fewer years of formal education have a greater risk for developing dementia [11 14] The benefits of education tend to initiate a particular trajectory of conditions which may in turn impact brain health [50] Racial and ethnic disparities in e ducational attainm ent and quality persist even though Jim Crow laws have been overturned Educational attainment is also strongly associated with cognitive reserve [19 21] The cognitive reserve theory states that individuals with more educational, occupational, and cogn itive engagement are more resilient to damage to their brain, delaying the presentation of symptoms of dementia [19 21] However, Manly and colleagues found that literacy level, rather than years of education, better predicts cognitive decline regardless of race/ethnicity; literacy and years of education are not concordant [24] Estimates are that only half of individuals with AD have been diagnosed, and, of those diagnosed, only 33% are aware of their diagnosis [7] Yet early detection of AD is important be cause it gives individuals time to express their wishes, build a care team, create advance health directives, make legal and financial arrangements, and enroll in clinical trials before their disease progresses to an advanced stage [7] Early detection

PAGE 39

39 als o helps physicians deliver better care and help patients better manage the comorbid conditions that often occur with greater age [7] Screening is essential to early detection of MCI and AD. The Montreal Cognitive Assessment (MoCA), was developed as a scre ening tool to distinguish between normal aging and mild cognitive impairment [25] Administered by a trained psychometrist, geriatrician or other health professional, the MoCA takes approximately 10 minutes to complete and contains subtests on orientation, attention, verbal memory, language, visuospatial function, and executive function [25] Scored out of 30 points, it has a one point educational adjustment (addition) for individuals with 12 years of education or less [25] The MoCA has a high sensitivity for detection of MCI in patients who would score as normal on the Mini Mental State Examination (MMSE), another widely used tool to assess cognitive impairment; using a cutoff of 26 for mild cognitive impairment the MoCA had a sensitivity of 90% while the MMSE had a sensitivity of 18% [25] The MoCA is superior to the MMSE because its memory testing involves fewer learning trials, more words, and a longer delayed recall period compared to the MMSE and it has more challenging items that assess mild impairmen t in executive functions, visuospatial processing, and language abilities that may be present in individuals with MCI than the MMSE [25] Although it is widely accepted that the MoCA is better than the MMSE for detection of MCI [25 27] some studies sugges t that the cutoff of 26 results in a high proportion of normal individuals being classified as cognitively impaired [28,29] The application of the MoCA in minority groups has not been widely studied; most of the studies examining the MoCA and its subtests have been limited to Caucasian samples. Studies that included minority participants have found that the

PAGE 40

40 originally established cutoff of 26 may result in a high likelihood of classifying normal minority individuals as cognitively impaired [33 35] For ins tance, using the recommended cutoff of 26, Sink and colleagues found that over 90% of their cohort of African Americans with type 2 diabetes participating in the study of type 2 diabetes and t [35] Rossetti and colleagues found that about 80% of their cohort of community dwelling African Americans fell below the threshold [34] These findings suggest a need to reevaluate MoCA cutoffs and their performance in minority populations; they result in too many false positive results. In these analyses, we aim to identify race specific optimal cutoff values of the MoCA when stratified by education and age, using clinician diagnosis as the gold standard. We hypothesize that minority groups will requir e lower cutoffs to distinguish between normal aging and MCI or dementia, and further, between MCI and dementia. We also hypothesize that cutoffs will need to decrease as educational attainment decreases to ensure optimal MoCA performance. Methods Populatio n (ADC). The data set used in these analyses, the uniform data set (UDS), includes subjects en rolled at ADCs since 2005 with a range of cognitive status, measured through neuropsychological tests and clinician assessment. Each ADC recruits subjects according to its own protocol; recruitment methods include clinician referral, self referral by patie nts or family members, and/or active recruitment through community

PAGE 41

41 organizations [40] Participants who completed a MoCA at their baseline visit, who reported being either non Hispanic White, non Hispanic Black, or Hispanic, who reported their education, a nd were included in the March 2018 data freeze were included in these analyses (n=3,895). Measures Race and ethnicity variables were combined to create a three level race/ethnicity variable categorized as non Hispanic White, non Hispanic Black, and Hispan ic. The MoCA score was recorded as a continuous variable from 0 to 30. Clinician diagnosis was recorded as either normal cognition impaired not MCI, MCI, or dementia. Only participants who were diagnosed as having normal cognition MCI, or dementia were i ncluded in these analyses ; however, normal cognition will be referred to as normal aging herein We dropped those coded as impaired, not MCI, due to uncertainty over their diagnosis and whether or not it was relevant to either MCI or dementia Dementia enc ompassed multiple etiologic diagnoses including AD, Lewy body disease, Frontotemporal lobar degeneration. Therefore, the term dementia will be used herein. and more tha n college (>16 years). Because the suggested educational adjustment has been found to affect the validity of the MoCA and may not be appropriate in all groups [37] we used raw MoCA scores for these analyses. We accounted for differences due to education b y stratifying cutoffs by education within each racial/ethnic group. Age was categorized by decade of age less than 50, 50 59, 60 69, 70 79, and 80 or more. Data Analysis Using SAS 9.4 software, descriptive statistics were calculated for population char acteristics including race/ethnicity, age, and education. Test measures including

PAGE 42

42 specificity, sensitivity, positive predictive value, and negative predictive value were calculated. The SAS 9.4 ROCPLOT macro was used to calculate ideal cutoffs for the MoCA to calculate corresponding areas under the ROC curve (AUC) when 1) identifying MCI and 2) distinguishing between MCI and dementia, by race/ethnicity when stratified by education. The Youden index, equal to the sensitivity plus specificity minus one, is a valued way to summarize the performance of a diagnostic test [45,51] MedCalc was used to calculate confidence intervals for AUCs and to generate bootstrapped confidence intervals for cutoffs using 100,000 iterations [52] Resul ts Of the 3,895 participants, 80.7% were non Hispanic White, 15.0% were non Hispanic Black, and 4.2% were Hispanic; over half (56.8%) were female with an average age of 69.7 (range 19 to 101; median: 70.0; IQR: 65.0 76.0) years. A majority of participants reported 13 to 16 years of education (43.8%) or more than 16 years of education (40.5%). Relatively few reported 12 or less years of education (15.7%). Over half of participants (62.5%), screened as cognitively impaired, with a MoCA score of less than 26 Most participants had normal aging as judged by a clinician (48.4%); the remaining had dementia (27.6%) or MCI (24.0%). Age, sex, years of education, clinician judged cognition, and MoCA score all significantly differed by race/ethnicity (Table 3 1). MoC A scores had a negative skew in both impaired and unimpaired groups, potentially driven by the individuals with fewer years of education (Figure 3 1). Using the originally established score cutoff of 26 to detect MCI or dementia, the sensitivity of the MoC A was 89.5% in non Hispanic Whites, 92.6% in non Hispanic Blacks, and 94.6% in Hispanics. The specificity of the MoCA with the established cutoff

PAGE 43

43 was 74.3% in non Hispanic Whites, 40.3% in non Hispanic Blacks, and 43.1% in Hispanics. Given these, the posit ive predictive value of the MoCA was 79.9% in non Hispanic Whites, 52.0% in non Hispanic Blacks, and 68.0% in Hispanics and the negative predictive value of the MoCA was 86.1% in non Hispanic Whites, 88.5% in non Hispanic blacks, and 86.1% in Hispanics (Ta ble 3 2). With this cutoff, 25.7% of non Hispanic Whites, 59.7% of non Hispanic Blacks, and 56.9% of Hispanics without clinician judged impairment screened as impaired on the MoCA (data not shown). The Youden Index was used to identify an optimal cutoff fo r detecting the presence of MCI, among individuals who had normal aging or MCI. Overall, 25 was the optimal cutoff for identifying either MCI (sensitivity=0.816, specificity= 0.669). When stratified by race/ethnicity, identified cutoffs were: 25 among non Hispanic Whites, 23 among non Hispanic Blacks, and 24 among Hispanics (Figure 3 1 ). Further stratification by education identified a decrease in optimal MoCA cutoffs for detecting either MCI or AD among those with fewer years of education (Table 3 3). Amon g non Hispanic Whites the optimal cutoff was 22 among those with 12 of education, 24 among those with 12 to 16 years of education, and 25 among those with >12 years of education. Among non Hispanic Blacks the optimal cutoff was 19 among those with 12 of education, 23 among those with 12 to 16 years of education, and 23 among those with >12 years of education. Among Hispanics the optimal cutoff was 23 among those with 12 of education, 24 among those with 12 to 16 years of education, and 24 among those wit h >12 years of education (Table 3 3, Figure 3 2 ). Table 3 3 includes confidence intervals for these calculated cutoffs. A few significant differences should be noted. The overall cutoff for non Hispanic Whites is

PAGE 44

44 significantly different from the overall c utoff for non Hispanic Blacks. Among non rom the cutoff for individuals with >16 years of education. Among non Hispanic Blacks, the on is significantly different from the cutoff for both individuals with 13 16 and >16 years of education. Lastly, the cutoff for non Hispanic Hispanic ion. Overall, 19 was the optimal cutoff for further distinguishing between MCI and dementia (sensitivity=0.727, specificity= 0.799). When stratified by race/ethnicity, identified cutoffs were: 19 among non Hispanic Whites, 16 among non Hispanic Blacks, and 16 among Hispanics (Figure 3 3 ). Further stratification by education identified a decrease in optimal MoCA cutoffs for distinguishing between MCI and dementia among those with fewer years of education (Table 3 4). Among non Hispanic Whites the optimal cut off was 17 among those with 12 of education, 19 among those with 12 to 16 years of education, and 20 among those with >12 years of education. Among non Hispanic Blacks the optimal cutoff was 13 among those with 12 of education, 17 among those with 12 to 16 years of education, and 19 among those with >12 years of education. Among Hispanics the optimal cutoff was 15 among those with 12 of education, 16 among those with 12 to 16 years of education, and 19 among those with >12 years of education (Table 3 4, Figure 3 4). Table 3 4 includes confidence intervals for these calculated cutoffs. A few significant differences should be noted. Again, t he overall cutoff for non Hispanic Whites is significantly different from the overall cutoff for non Hispanic Blacks. Among non

PAGE 45

45 Hispanic Whites, the cutoff for those with is significantly different for the cutoff for individuals with >16 years of education. Among non Hispanic Blacks, the cutoff for those with is significantly different f rom the cutoff for both individuals with 13 16 and >16 years of education. Lastly, the cutoff for non Hispanic Hispanic We considered stratifying cutoffs by age, in addition to race/ethnicity and education; however, in the non Hispanic Black and Hispanic groups the cell counts became too small. Average MoCA scores mostly increased as years of education increased and decreased as age increased in all three groups (Table 3 5, 3 6, 3 7 ). Discussion We found that the MoCA had different optimal cutoffs among different racial/ethnic minorities. These cutoffs were lower than the standardized cutoff and lowered as years of education decreased. While the original paper establishing the cutoff of 26 had alm ost 300 participants [25], we established these differing cutoffs using a highly regarded national sample of almost 4,000 individuals. Uniquely, the NACC database includes clinical diagnosis and MoCA scores for both affected individuals and normal controls This allowed us to calculate cutoffs that maximized sensitivity and specificity to delineate between health aging, MCI and dementia. Overall, we identified 25 as the optimal cutoff for detecting MCI among the total sample. This is consistent with previo us findings. A Cochrane Review from 2015 concluded that it is likely that cutoffs lower than 26 would be more useful for correctly identifying MCI or AD [53] A few studies have found a lower optimal cutoff would be more specific and sensitive for screenin g. Luis, Keegan, and Mullan administered the

PAGE 46

46 MoCA to community dwelling older adults in the Southeastern United States [28] They found that the MoCA had much better sensitivity and specificity (96% and 95%, respectively) to identify cognitive impairment v ersus normal aging with a lower cutoff of 23 compared to the recommended cutoff of 26 (97% and 35%, respectively) [28] However, their study focused on community dwelling residents in the Southeast United States. Moreover, a systematic review by Carson and colleagues found that lowering the MoCA cutoff to 23 reduces the false positive rate, resulting in improved diagnostic accuracy [29] While we too found a lower cutoff performed better than the recommended 26, our findings can be contrasted with these tea additionally stratified scores by race/ethnicity and education. This stratification resulted in a need for even lower cutoff values for individuals with lower years of education. When stratifying by race/ethnicity, we found the optim al cutoff value should be lower among minority groups. The optimal cutoff for MCI remained 25 among Whites; however, sensitivity and specificity was maximized when the cutoffs were lowered to 23 among non Hispanic Blacks and 24 among Hispanics. Previous l iterature has found that using lower cutoffs of the MoCA would be beneficial in African American populations by improving diagnostic accuracy [33 35] ; however, there is a gap of information on what should be the normative score for racial and ethnic popula tions. The literature that does include Hispanic populations has found that, among community dwelling adults, minorities performed more poorly on the MoCA compared to their White counterparts, and found that Hispanics scored the lowest on the MoCA among al l groups measured [54] These observed differences may be due to a lack of cultural equivalence of test items or to differential variability in MoCA score by race/ethnicity. African Americans

PAGE 47

47 have previously been found to perform worse on majority normed c ognitive tests compared to their White counterparts, making race specific data crucial for clinical assessments [55] especially given our findings that optimal cutoffs differ by race/ethnicity. When we further stratified by education within each racial/e thnic group, we found different optimal cutoffs for MCI and dementia. When distinguishing MCI from dementia the confidence intervals for overall cutoffs among non Hispanic Whites and Blacks did not overlap, suggesting that these cutoffs are significantly d ifferent. Within each racial/ethnic group, the cutoff decreased as years of education decreased. This is to be expected, especially since the MoCA developers recommend an educational adjustment. These patterns were also consistent when distinguishing betwe en MCI and dementia. The observed differences in optimal cutoff when stratifying by race/ethnicity and education level may be attributable to differences in education, socioeconomic status, and comorbid medical conditions, all of which are risk factors for dementia [7,46,47] and all of which may be reflected in the education variable. However, it is important to note that even though we observed a strong racial difference in optimal cutoffs, non Hispanic Blacks included in the NACC sample were diagnosed as cognitively normal by a clinician more frequently than their non Hispanic White and Hispanic counterparts (58.9% vs. 46.7% and 43.9%, respectively). Further, the MoCA adds a one point education adjustment for individuals who wever, our results suggest that the one point educational adjustment may not be adequate when considering differences in test performance due to years of education. We observed that not only do optimal cutoffs

PAGE 48

48 differ between on and those with >12 years of education, but also between those with 13 to 16 years of education and >16 years of education. This was particularly evident among non Hispanic Blacks; when distinguishing between normal aging and MCI, the cutoffs for the ove education and those with 12 to 16 years of education also significantly differed. Moreover, again among Non Hispanic Blacks, whe n distinguishing between MCI and 16 years significantly differed by four points. Surprisingly, few studies have examined the validity of the educational adjustment; however, studies have examined normative MoCA scores when stratified by education. Rossetti and colleagues found that overall, the mean MoCA score of their ethnically diverse participants was 24; when stratified by education, the mean MoCA score was 21 among those with <1 2 years of education, 23 among those with 12 years of education, and 25 among those with > 12 years of education [37] When they reduced their sample to community dwelling African Americans alone, they found that the mean MoCA score of their cognitively nor mal participants was 23 [34] When stratified by education, the mean MoCA score decreased to 20 among those with <12 years of education, 22 among those with 12 years of education, and 24 among those with > 1 2 years of education. Together, our findings sugge st that the one point educational adjustment is not enough given differences in educational attainment. Additional support comes from Malek Ahmadi and colleagues who found that younger and more educated

PAGE 49

49 individuals performed better on the MoCA than their c ounterparts [56] However, their study differs because it only included non Hispanic Whites [56] Racial/ethnic differences observed in the performance of the MoCA suggests that the current cutoff is not appropriate for use among minority groups, particularly among non Hispanic Blacks. Given that the non Hispanic Blacks included in these analyses are mostly cognitively normal, future research is needed to identify the source of variability in MoCA performance by race/ethnicity. We plan to examine t he role of MoCA subtests in the variability of overall score by race/ethnicity and education in future work. Some limitations should be considered when interpreting the results of these analyses. Subjects in the NACC database may not represent the United S tates population, since they are referral or volunteer based participants and tend to be more educated than the general public. Moreover, normal aging controls in ADC samples tend to be highly educated and not representative of the United States populatio n. The participants to consent to autopsy before their baseline visit, thus excluding individuals who may not agree to autopsy [41] This may have resulted in more homogeneity in the sample than expected regardless of race/ethnic differences. There is also potential bias due to the timing of the MoCA; the MoCA is administered before clinician evaluation and may influence clinicians to agree with the MoCA score. This would artifici ally increase sensitivity and specificity of our calculated MoCA cutoff scores. Nevertheless, clinicians normally see screening results and neurological testing results when making their diagnosis. This is the norm in the field. Additionally, the number o f non Hispanic Black and Hispanic individuals in the NACC database is relatively small compared to White

PAGE 50

50 participants. Our distribution of Hispanics is sparse (Figure 3 1 and 3 2) potentially underpowering our analysis. This is reflected in the large confi dence intervals calculated for our Hispanic group. However, our findings are still novel since many other data sets almost solely focus on Whites. The inclusion of multiple race/ethnicities in the NACC data is unique, especially since data is collected fro m subjects whose ages fall in a wide age range: from less than 40 years to older than 90 years. Although previous work has found that literacy is a better predictor of cognitive decline [24] our data only captures education. Future work to address the eff ect of literacy on MoCA scores is warranted. NACC should consider adding a literacy level to the NACC required data capture. Racial/ethnic specific cutoffs may become increasingly important to correctly identify MCI and dementia in minority populations w ho are at high risk of developing dementia, given the growing diversity of the United States. More accurate cutoffs for the MoCA should be used among minority populations. By stratifying by race/ethnicity and education level before applying a cutoff value for MoCA score, we can correctly identify individuals in need of more in depth screening. Given the importance of early detection and diagnosis of dementia, these new cutoffs might aid healthcare professionals in early diagnosis and treatment of MCI and de mentia.

PAGE 51

51 Table 3 1. Characteristics of NACC study participants, by race/ethnicity (n=3,895 ). Characteristic Overall (n=3,895) Non Hispanic Whites (n=3,145; 80.7%) Non Hispanic Blacks (n=586;15.0%) Hispanics (n= 164; 4.2%) P Value Mean Age (SD) 69.7 (9.8) 69.6 (10.0) 70.1 (8.7) 68.9 (9.9) 0.2905 Decade of Age <50 50 59 60 69 70 79 0 108 (2.8%) 400 (10.3%) 1300 (33.4%) 1531 (39.3%) 556 (14.3%) 97 (3.1%) 330 (10.5%) 1017 (32.3%) 1257 (40.0%) 444 (14.1%) 5 (0.9%) 52 (8.9%) 228 (38.9%) 213 (36.4%) 88 (15.0%) 6 (3.7%) 18 (11.0%) 55 (33.5%) 61 (37.2%) 24 (14.6%) 0.0137 Sex Male Female 1683 (43.2%) 2212 (56.8%) 1461 (46.5%) 1684 (53.6%) 163 (27.8%) 423 (72.2%) 59 (36.0%) 105 (64.0%) <0.0001 Years of Education 13 16 >16 611(15.7%) 1706 (43.8%) 1578 (40.5%) 404 (12.9%) 1372 (43.6%) 1369 (43.5%) 153 (26.1%) 263 (44.9%) 170 (29.0%) 54 (32.9%) 71 (43.3%) 39 (23.8%) <0.0001 Cognitive Status Normal MCI Dementia 1886 (48.4%) 936 (24.0%) 1073 (27.6%) 1469 (46.7%) 733 (23.3%) 943 (30.0%) 345 (58.9%) 158 (27.0%) 83 (14.2%) 72 (43.9%) 45 (27.4%) 47 (28.7%) <0.0001 MoCA Score <26 1461 (37.5%) 2434 (62.5%) 1268 (40.3%) 1877 (59.7%) 157 (26.8%) 429 (73.3%) 36 (22.0%) 128 (78.1%) <0.0001

PAGE 52

52 Table 3 2. Test characteristics of the Montreal Cognitive Assessment, using the current cutoff of 26 for detection of MCI or dementia, by race/ethnicity, among NACC study participants. Race/Ethnicity Sen sitivity Specificity PPV NPV Non Hispanic White 89.50% 74.34% 79.91% 86.12% Non Hispanic Black 92.53% 40.29% 51.98% 88.54% Hispanic 94. 57% 43.06% 67.97% 86.11% Overall 90.10% 66.91% 74.36% 86.38%

PAGE 53

53 Table 3 3. Optimal MoCA cutoffs to distinguish between normal aging ( NA ) and MCI by race/ethnicity and education level, a mong NACC study participants with normal aging or MCI ; MoCA scores range from 0 30 and lower scores signify more impairment (n=2,822) Years of Education Number of Participants Cutoff (95%CI) b Sensitivity (%) Specificity (%) AUC (95%CI) NH Whites n=2,202; 78.0% Overall NA = 1469 MCI= 733 79.51 74.32 0.843 [0.827, 0.858] 12 a NA =116 MCI=97 71.13 83.62 0.828 [0.770, 0.876] 13 16 NA =623 MCI=315 72.38 83.28 0.859 [0.835, 0.880] >16 H C=730 MCI=321 72.50 80.96 0.832 [0.808, 0.854] NH Blacks n=503;17.8% Overall NA =345 MCI=158 71.52 71.59 0.769 [0.730, 0.805] 12 a NA =67 MCI=50 56.00 83.58 0.721 [0.630, 0.800] 13 16 NA =165 MCI=71 73.24 74.55 0.765 [0.706, 0.818] >16 NA =113 MCI=37 64.86 82.30 0.791 [0.717, 0.853] Hispanics n=117; 4.2% Overall NA =72 MCI=45 84.44 55.56 0.727 [0.637, 0.805] 12 a NA =20 MCI=12 91.67 30.00 0.569 [0.383, 0.742] 13 16 NA =37 MCI=18 83.33 62.16 0.773 [0.640, 0.875] >16 NA =15 MCI=15 80.00 93.33 0.927 [0.770, 0.990] a education or less; b bootstrap confidence interval (100000 iterations; random number seed: 978)

PAGE 54

54 Figure 3 1 Classification ac curacy of optimal MoCA cutoffs for MCI, by race/ethnicity among NACC study participants with normal aging or MCI

PAGE 55

55 Figure 3 2. Optimal MoCA cutoffs to distinguish between normal aging and MCI by race/ethni city and education level, among NACC study participants with normal aging or MCI; MoCA scores range from 0 30 and lower scores signify more impairment (n=2,822) *MoCA scoring rules adjust for of education, adding one point to the total score of individuals with 12 years of educati on or less

PAGE 56

56 Table 3 4. Optimal MoCA cutoffs to distinguish between MCI and dementia by race/ethnicity and education level, among NACC study participants with MCI or dementia (DM); MoCA scores range from 0 30 and lower scores si gnify more impairment (n=2,009 ) Years of Education Number of Participants Cutoff (95%CI) b Sensitivity (%) Specificity (%) AUC (95%CI) NH Whites n=1,676; 83.4% Overall MCI=733 DM=943 70.31 83.06 0.844 [0.826, 0.861] MCI=97 DM=191 70.16 80.41 0.840 [0.793, 0.881] 13 16 MCI=315 DM=434 72.12 81.27 0.844 [0.816, 0.869] >16 MCI=321 DM=318 68.24 84.37 0.838 [0.807, 0.866] NH Blacks n=241;12.0% Overall MCI=158 DM= 83 78.31 88.61 0.898 [0.853, 0.933] MCI=50 DM=36 77.78 96.00 0.919 [0.839, 0.967] 13 16 MCI=71 DM=27 77.78 92.96 0.870 [0.787, 0.930] >16 MCI=37 DM=20 90.00 83.78 0.944 [0.849, 0.987] Hispanics n=92; 4.6% Overall MCI=45 DM=47 72.34 97.78 0.886 [0.803, 0.943] MCI=12 DM=22 90.91 100.00 0.966 [0.839, 0.999] 13 16 MCI=18 DM=16 75.00 100.00 0.858 [0.695, 0.953] >16 MCI=15 DM=9 77.78 73.33 0.726 [0.507, 0.886] a education or less; b bootstrap confidence interval (100000 iterations; random number seed: 978)

PAGE 57

57 Figure 3 3. Classification accuracy of optimal MoCA cutoffs for dementia, by race/ethnicity among NACC study participants with MCI or dementia

PAGE 58

58 Figure 3 4. Optimal MoCA cutoffs to distinguish between MCI and dementia by race/ethni city and education level, among NACC study partic ipants with MCI or dementia; MoCA scores range from 0 30 and lower scores signify more impairment (n=2,009) individuals with 12 years of education or less.

PAGE 59

59 Tab le 3 5. Average MoCA score among non Hispanic White NACC participants by decade of age and education level (n=3,145) Years of Education 12 (n= 404 ; 1 2.9 %) 13 16 (n= 1,372 ; 43.6 %) >16 (n= 1,369; n=43.5 %) Age N Mean (SD) Median N Mean (SD) Median N Mean (SD) Median <5 0 6 26.2 ( 3 .8) 27.5 58 24.4 ( 6.9 ) 27.0 33 27.8 (2.1) 28.0 50 59 5 7 17.9 ( 7.4) 19.0 163 21.7 (7. 5) 24.0 110 23.8 (6. 8 ) 2 6.0 60 69 95 19.8 ( 6.0) 21.0 453 22.2 (6.7) 24.0 469 24.5 (5. 6) 26.0 70 79 168 18.6 (6. 3) 19.0 513 22.0 (6. 1) 23 .0 576 23.7 (5. 2) 25.0 8 0 78 18.0 (6. 7) 19.0 185 20. 4 (6.1 ) 22.0 181 22.9 (5. 2) 24 .0 Total by Education 404 18.8 ( 6.5) 20.0 1372 2 1.9 (6. 6) 24.0 1369 24.0 (5. 5) 26.0

PAGE 60

60 Table 3 6. Average MoCA score among non Hispanic Black NACC participants by decade of age and education level (n=586) Years of Education 12 (n= 153; 26.1 %) 13 16 (n= 263 ; 44. 9 %) >16 (n= 170 ; n=2 9 0 %) Age N Mean (SD) Median N Mean (SD) Median N Mean (SD) Median < 5 0 1 26.0 ( ) 26.0 3 25.3 (1.2) 26.0 1 20.0 ( ) 20.0 50 59 1 0 19.5 ( 5.1) 20.0 27 24.7 (3. 8 ) 25.0 1 5 26.5 (2. 4) 27.0 60 69 54 21.3 (5. 2) 23.0 107 24.1 (2.9) 25.0 67 24.3 ( 4.5 ) 25.0 70 79 52 19.2 (5.5) 20 .5 94 21.6 (4. 5 ) 23.0 67 23.1 (4. 4 ) 24.0 8 0 36 14.9 ( 7.2) 15.5 32 18. 9 ( 6.8) 20.0 20 21.3 ( 4.5 ) 22.0 Total by Education 153 19.0 (6. 3) 20.0 263 22. 7 (4.6 ) 24.0 170 23.6 (4. 5 ) 24.0

PAGE 61

61 Table 3 7. Average MoCA score among Hispanic NACC participants by decade of age and education level (n=164) Years of Education 12 (n= 54 ; 3 2.9 %) 13 16 (n=71; 43.3 %) >16 (n= 39; n=23.8 %) Age N Mean (SD) Median N Mean (SD) Median N Mean (SD) Median < 5 0 2 14.0 (7.1) 14.0 3 24.3 ( 7.2) 28.0 1 30.0 ( ) 30.0 50 59 6 12.7 ( 7.5) 11.5 9 16.3 (8.3) 18.0 3 23.3 ( 6.1) 22 .0 60 69 17 17.9 (6. 9) 20.0 24 24.0 ( 4.2) 25.0 14 24.2 (3. 7 ) 26.0 70 79 17 18.2 (6. 4 ) 19.0 29 22.1 (4. 1) 22.0 15 22. 2 (5. 4 ) 24.0 8 0 12 14.4 ( 6.0) 14.5 6 16.5 (7 .8) 17.0 6 20.3 (2.6 ) 1 9 5 Total by Education 54 16.5 (6.7) 18.0 71 21.6 (5.9 ) 23.0 39 22.9 (4.7) 24 .0

PAGE 62

62 CHAPTER 4 SUBTEST ANALYSIS OF THE MONTREAL COGNITIVE ASSESSMENT, BY RACE AND ETHNICITY AMONG NACC PARTICIPANTS Background Dementia, a broad category of neurodegenerative disorders, affects approximately 8.8% of the population aged 65 and older [5]. As the population ages, the number o f older adults at risk for developing dementia will increase [4]. Mild cognitive impairment (MCI) is defined as cognitive impairment that is greater than expected for d evelop dementia; a review of 41 studies found that less than half of individuals with MCI develop dementia [3]. Diagnosis of MCI is difficult to ascertain [2,57] however, early detection is crucial to prevent further impairment, manage patients comorbid c onditions, and allow patients to express their directives before impairment becomes too severe [58] Racial and ethnic minority groups have disproportionally high rates of dementia [1,8]. African Americans have double the risk of developing late onset Alzh disease (AD), the most common type of dementia, compared to their white counterparts [1]. Hispanics also have a high burden of AD, with one and a half times the risk of developing late onset AD compared to their white counterparts [1]. This dispari ty in dementia burden is potentially attributed to a higher burden of risk factors, especially socioeconomic status for dementia including among minority populations [1,9,10]. However, we must consider if these are true differences in dementia incidence an d distribution of risk factors or a result of measurement bias in these populations when examining cognitive impairment as well as associated risk factors.

PAGE 63

63 The Montreal Cognitive Assessment (MoCA) is widely accepted to be a better test for detecting MCI th en the Mini Mental State Examination (MMSE), one of the most frequently used neuropsychologic tests [25 27]. The MoCA is one page, 30 point assessment, covering eight cognitive domains, with a suggested cutoff of 26 for cognitive impairment [25]. The valid ation study of the MoCA found that with a cutoff of 26,the MoCA had a much higher sensitivity compared to the MMSE (90% vs 18%, respectively) [25]. However, the utility of this cutoff has been widely questioned [28,29], especially for use in minority popul ations [33 35]. The MoCA is an important tool for detection of MCI and dementia, however, the psychometric properties of its subtests and items have not been widely studied [30]. A study that did examine MoCA subtests and items individually found that word repetition, inverse digits, serial 7, phrases, verbal fluency, abstraction, and word recall discriminated between MCI and normal aging and that the clock drawing, rhino ceros naming, delayed recall of five words, and orientation discriminated between MCI a nd AD [31]. Additionally, Roalf and colleagues examined the ability of MoCA subtests and items to discriminate between healthy controls and affected individuals (MCI, AD, pati ents with the goal of creating a short version of the examination [32]. They found that the clock drawing, serial 7s, orientation place, delayed recall, abstraction, rhinoceros naming trails, and fluency subtests discriminated between the two groups [32] Another study looking to create an abbreviated MoCA with the three items with the largest effect sizes, found that serial 7s and delayed recall were the best items for

PAGE 64

64 distinguishing between normal aging and MCI while the serial 7s, delayed recall, and o rientation items were the best at discriminating between MCI and AD [30]. However, these studies primarily focused on populations as a whole and have not examined racial/ethnic differences. Performance on cognitive testing is affected by exposures througho ut the lifecourse [16] ; these exposures are unequally distributed by race and ethnicity and can explain differences in the burden of dementia as well as performance on cognitive tests. Therefore, it is imperative to consider race when interpreting performa nce on cognitive tests. W e aimed to analyze the discriminative ability of each MoCA subtest/item to distinguish between 1) normal aging and MCI and 2) MCI and dementia by race/ethnicity, when controlling for age and education level. Methods Population Dat Center (NACC) Uniform Data Set (UDS). NACC maintains a database of information Each center enrolls subj ects according to their own protocol; however, data is collected through a standardized evaluation, administer by clinicians and their trained staff, including a range of cognitive measures, neuropsychological tests, and clinician assessment [40] Measure s Clinician diagnosis was recorded as normal, impaired but not MCI, MCI, and dementia. Only individuals who were judged as normal, MCI, or dementia were included in these analyses. Individuals who were impaired but did not have MCI were not included due to the unclear etiology of their impairment.

PAGE 65

65 Overall raw MoCA score was recorded from 0 30. MoCA subtest scores [visuospatial/executive (trails, cube copy, clocks), naming, attention (digits, letters, serial 7s), language (repetition, fluency), abstraction, delayed recall, orientation (date, month, year, day, place, and city)] were used to predict clinician diagnosis. Trails, cube, clock contour, clock numbers, clock hands, letter A, fluency, and orientation (date, month, year, day, place, and city) were all coded as binary variables (incorrect/correct). The remaining variables were coded as ordinal variables. Naming was scored from zero to three, with one point given for each animal (of three) correctly named. The digits item was scored from zero to two, with one point given for being able to correctly read the first set of digits in forward order and one point given for being able to correctly read the second set of digits in a backwards order. Serial 7s was scored from zero to three; zero points for no corre ct subtractions, one point for one correct subtraction, two points for two or three correct subtractions, and three points for four or five correct subtractions. Repetition was scored from zero to two, with one point given for each sentence that was correc tly repeated. Abstraction was scored from zero to two, with one point given for each similarity that was correctly identified. Lastly, delayed recall was scored from zero to give, with one point given for each word (of five) that was recalled without a cue Race and ethnicity were categorized into a three level race/ethnicity variable: non Hispanic White, non Hispanic Black, and Hispanic. Education was categorized as high school or less (12 years or less), college (13 to 16 years), and more than college (>16 years). Age was recorded as a continuous variable.

PAGE 66

66 Data Analysis Individuals who completed a MoCA at their baseline visit and were included in the March 2018 data freeze were included in these analyses. Moreover, only individuals who reported either bein g White or Black, reported their ethnicity (Hispanic/Latino or non Hispanic/Latino), and reported their years of education were included in these analyses. Our final sample included 3,895 people. D escriptive statistics were calculated for population charac teristics including race/ethnicity, age, and education. Stepwise multinomial logistic regression was used to determine which subtests best predicted clinician diagnosis. Finally, using multinomial logistic regression, we deconstructed subtests into items a nd analyzed the adjusted odds ratios of the association between each item and clinician diagnosis, adjusting for age and education. Separate analyses were conducted among the 1) whole sample, 2) only among non Hispanic Whites, 3) only among non Hispanic Bl acks, and 4) only among Hispanics. P values less than 0.05 were considered to be statistically significant. All analyses were conducted using SAS software, version 9.4. Results Overall, our final sample of 3,895 participants was mostly female (56.8%) wit h a mean age of 69.7. 80.7% were non Hispanic White, 15.0% were non Hispanic Black, and 4.2% were Hispanic. They mostly had some college or more than a college education (43.8% and 40.5%, respectively) while few reported less than a high school education ( 15.7%). Almost half (48.4%) were judged to be cognitively normal by a clinician. The remaining had either MCI (24.0%) or dementia (27.6%). Sex, years of education, cognitive status, and MoCA score all significantly differed by race/ethnicity (Table 4 1). T he non Hispanic Black group had significantly more females than males

PAGE 67

67 while Hispanics had more females than the Non Hispanic White group but less than non Hispanic Blacks. Non Hispanic Blacks and Hispanics comprised a higher proportion of the sample with l ess than or equal to 12 years of education and smaller proportion of those with over 16 years of education. Non Hispanic Blacks had more individuals with clinician diagnosed normal aging but less individuals with normal MoCA scores ( ) than the Non Hispa nic White group (58.9% vs 46.7% and 73.3% vs 59.7%, respectively). Tables 4 2, 4 3, and 4 4 show the results of the stepwise multinomial logistic regression. When looking at the overall sample, all subtests, education, and age were significant in predictin g clinician diagnosis. This held true for non Hispanic Whites (Table 4 2), however, when only looking at non Hispanic Blacks or Hispanics, fewer subtests predicted clinician diagnosis. Among non Hispanic blacks, visuospatial/executive, attention, language, delayed recall, and orientation subtests were significant predictors of clinician diagnosis (Table 4 3). Among Hispanics, visuospatial/executive, delayed recall, and orientation subtests and education were significant predictors of clinician diagnosis (Ta ble 4 4). When deconstructing subtests into items, results were mostly consistent with results from the stepwise regression (Table 4 5). In our overall sample, items from all subtests were significant in predicting MCI from normal aging and dementia, consi stent with the inclusion of all subtests in the final model from Table 4 2. Within the visuospatial/executive subtest the clock number and clock hand item predicted MCI from both normal aging and dementia while the cube item only predicted MCI from normal aging Naming was significant in predicting MCI from both normal aging and

PAGE 68

68 dementia. Within the attention subtest, the letters and serial 7s items were significant in predicting MCI from both normal aging and dementia. Within the language subtest, repetiti on and fluency significantly discriminated MCI from both normal aging and dementia. Moreover, abstraction significantly predicted MCI from dementia and delayed recall significantly predicted MCI from both normal aging and dementia. Within the orientation s ubtests, date, day, and place, significantly predicted MCI from normal aging and dementia, while month and year only discriminated between MCI and dementia. The city item did not predict clinician diagnosis. Results from non Hispanic Whites were similar to that of the overall sample (Table 4 5), given that the previous stepwise regression conducted included all subtests in both models (overall and non Hispanic Whites). In addition to all items that significantly predicted MCI from either normal aging or dem entia in the overall sample, a few more items were significant among non Hispanic Whites. In this group, trails also predicted MCI from both normal aging and MCI, while digits only predicted MCI from dementia. Abstraction predicted MCI from normal aging T his is consistent with the previous stepwise regression; in the overall model abstraction only significantly predicts MCI from dementia, while in the non Hispanic White model it predicts MCI from both normal aging and dementia. Less items were significant among non Hispanic Blacks (Table 4 5). Significant items were mostly part of subtests included in the final stepwise model. The visuospatial/executive subtest was included in the final stepwise model, however no individual item was significant in predictin g MCI from normal aging or dementia, potentially suggesting an additive effect of this subtest with other subtests to predict

PAGE 69

69 clinician diagnosis. Alone, naming was significant in predicting MCI from aging; however, when included with all other test items, naming was not significant and not part of the final stepwise model. Within the attention subtest, serial 7s significantly predicted MCI from dementia, consistent with the stepwise model. Within the language subtest, both repetition and fluency significan tly predicted MCI from normal aging Delayed recall was significant in predicting MCI from normal aging Within the orientation subtest, place was the only item that predicted clinician diagnosis; it predicted MCI from both normal aging and dementia. Even less items were significant among Hispanics, potentially due to the small number of Hispanics included in these analyses (Table 4 5). All significant items were consistent with the results of the stepwise regression. Within the visuospatial/executive subte sts, the clock number item was significant in distinguishing MCI from dementia. Naming, attention, language, and abstraction items were not significant in predicting clinician diagnosis, however, delayed recall significantly predicted MCI from normal aging Within the orientation subtest, the day item significantly predicted MCI from dementia. No other orientation subtests were significant in predicting clinician diagnosis. Discussion Our findings suggest that not all MoCA subtests demonstrate clinical util ity, especially in minority populations. While most subtests did significantly differ between either normal aging and MCI or MCI and dementia in our total sample, when the sample was restricted to our minority groups, these differences did not persist. Wh en conducting stepwise analyses to determine which subtests best predict clinician diagnosis of either normal aging or dementia, compared to MCI, we found that while all subtests included in the MoCA were important for the total sample and Non

PAGE 70

70 Hispanic Whi tes, fewer subtests were significant in non Hispanic Blacks and Hispanics. Moreover, education was not included in the final stepwise model among non Hispanic Blacks, though it was included in the model for Hispanics, suggesting that education does not exp lain the variability in diagnosis among non Hispanic Blacks. This demonstrates that the MoCA and its subtests may have less diagnostic utility among minority populations. When examining item performance among the total sample, we found several markers of early impairment that predicted MCI from normal aging The cube, clock number, clock hands, naming, letters, serial 7s, repetition, fluency, delayed recall, orientation date, orientation day, and orientation place items significantly differed between the n ormal aging and MCI group. All three previously mentioned studies found delayed recall to discriminate between normal aging and MCI [30 32] Moreover, Morton et al. and Cecato et al. found that serial 7s discriminated between normal aging and MCI [30,31] Similar to our analyses, in addition to these two items, in their analysis of 136 older adults from Brazil, Cecato and colleagues also found that phrases (which we call repetition), and fluency discriminated between normal aging and MCI [31] However, th ey also found that repetition (the unscored part of delayed recall), digits, and abstraction discriminated between normal aging and MCI [31] In our total sample these subtests did not discriminate between these two groups. We did not look at word repetit ion alone, the unscored part of the memory subtest. However, we did examine it in the context of delayed recall where participants are asked to recall the words at the end of the MoCA. Differences due to coding of variables may account for why they did not observe a difference for orientation items. We looked at each orientation item

PAGE 71

71 separately and as a whole, while they looked at the six orientation items as a whole. When only examining differences among non Hispanic Whites, the abstraction item also discr iminates between normal aging and MCI. We also found markers of severe impairment that predicted MCI from dementia in our sample as a whole. The clock number, clock hands, naming, letters, serial 7s, repetition, fluency, abstraction, delayed recall, orient ation date, orientation month, orientation year, orientation day, and orientation place items predicted dementia among our total sample. Similarly Cecato and colleagues found that the clock number, clock hands, delayed recall, and orientation items discrim inated between MCI and dementia [31] They also found that the rhino naming was associated with more impairment [31] however, due to all three naming items being combined in our data set, we were not able to examine this difference. Moreover, Horton and c olleagues had similar findings. They found that serial 7s, delayed recall, and orientation discriminated between MCI and dementia [30] When restricting the sample to only non Hispanic Blacks or only Hispanics, the ability of items to discriminate between normal aging and MCI or MCI and dementia diminished drastically. Only the naming, repetition, delayed recall, and orientation place items significantly discriminated between normal aging and MCI among non Hispanic blacks, while only the delayed recall disc riminated between these two groups. Among non Hispanic Blacks only the serial 7s, orientation date, and orientation place items discriminated between MCI and dementia. Among Hispanics, even fewer items discriminated between MCI and dementia; only the cloc k number and orientation day subtests.

PAGE 72

72 It is important to note that delayed recall was the only subtest which was a significant predictor of normal aging and MCI among our total sample as well as each race/ethnic group. This subtest was also important in literature reviewed in distinguishing between 1) normal aging and any impairment, 2) normal aging and MCI, and 3) MCI and dementia [30 32] Previous work found that abstraction discriminated between normal aging and MCI [31] In our total sample, abstract ion discriminated between MCI and dementia only, however, when only examining non Hispanic Whites, abstraction significantly predicted clinician diagnosis. Abstraction was not a significant predictor of clinician diagnosis among non Hispanic Blacks and His panics. Clock contour did not discriminate between MCI and normal aging or dementia in any of the models, likely because this item has a ceiling effect. Previous work has found that most individuals get full points on the clock contour, with 100% of indivi duals in the normal aging group and MCI group scoring full points, while 90.4% of individuals with AD scored full points [31] Within the orientation subtest, the city item also did not discriminate between MCI and health aging or dementia in any in any of the four models. Much of the normative work involving the MoCA and validation studies have been limited to non Hispanic White populations, which may explain why these subtests and items do not perform as well in minority groups. Moreover, studies examinin g the performance of the MoCA subtests have concluded that the cutoff of 26 is too high for use in minority populations [33 35] The ability of the MoCA subtests to discriminate between MCI and normal aging or dementia among minority populations demonstrat ed

PAGE 73

73 in these analyses may explain why the total cutoff score should be lower to correctly identify impairment in minority populations. These analyses have a few limitations to consider. I ndividuals included in the NACC database tend to be more educated than the general public in the United States population. However, we did control for years of education in these analyses. Moreover, recruitment is referral based and some ADCs require participants to consent to autopsy [40] which may introduce selection bias into our sample, excluding individuals who may have different views on research and autopsy. Our sample also has a relatively small number of non Hispanic Blacks and even fewer Hispanics, potentially underpowering the analyses focusing solely on these gro ups. Nevertheless, the NACC data set is still unique given that most studies focus on majority populations and do not include many minority members. Additionally, due to the nature of NACC coding, we could not distinguish performance on the individual nami ng items. NACC coding of the MoCA naming subtest is coded as a quantitative variable from zero to three, rather than focusing on the rhino, camel, and lion items individually. Previous work did find that the rhino naming item discriminates between MCI and AD [31] however we could not establish this using our sample. Early detection of MCI is crucial to improve health outcomes of individuals who develop MCI By understanding the psychometric properties of the MoCA subtests, we can focus on the subtests that have more diagnostic utility in distinguishing MCI from either normal aging or dementia. In our analyses we found that while most of the subtests work well in the total sample and in non Hispanic white populations, few of them discriminated between MCI an d either normal aging or dementia in minority

PAGE 74

74 populations. This suggests a need to further evaluate subtests for use in screening of non Hispanic Blacks and Hispanics.

PAGE 75

75 Table 4 1 Characteristics of NACC study participants included in these analys es, by r ace/ethnicity (n=3,895) Characteristic Total (n=3,895) Non Hispanic Whites (n=3,145; 80.7%) Non Hispanic Blacks (n=586; 15.0%) Hispanics (n=164; 4.2%) Mean Age (SD) 69.7 (9.8) 69.6 (10.0) 70.1 (8.7) 68.9 (9.9) Sex** Male Female 1683 (43.2%) 2212 (56.8%) 1461 (46.5%) 1684 (53.6%) 163 (27.8%) 423 (72.2%) 59 (36.0%) 105 (64.0%) Years of Education** 13 16 >16 611 (15.7%) 1706 (43.8%) 1578 (40.5%) 404 (12.9%) 1372 (43.6%) 1369 (43.5%) 153 (26.1%) 263 (44.9%) 170 (29.0%) 54 (32.9%) 71 (43.3%) 39 (23.8%) Cognitive Status** Normal MCI Dementia 1886 (48.4%) 936 (24.0%) 1073 (27.6%) 1469 (46.7%) 733 (23.3%) 943 (30.0%) 345 (58.9%) 158 (27.0%) 83 (14.2%) 72 (43.9%) 45 (27.4%) 47 (28.7%) MoCA Score** <26 1461 (37.5%) 2434 (62.5%) 1268 (40.3%) 1877 (59.7%) 157 (26.8%) 429 (73.3%) 36 (22.0%) 128 (78.1%) p <0.05 ** p <0.0001

PAGE 76

76 Table 4 2. Summary of Stepwise multinomial logistic regression for MoCA subtests and clinician diagnosis of normal aging ( NA ) or dementia ( DM ) among overall population vs. Non Hispanic Whites Overall Non Hispanic Whites Variable Dx Order B SE B Order B SE B Visuospatial/Executive NA 5 0.316 0.723 0.241** 4 0.414 0.060 0.320** D M 0.255 0.050 0.195** 0.257 0.054 0.198** Naming NA 8 0.348 0.107 0.1223* 7 0.449 0.126 0.156** D M 0.215 0.096 0.076* 0.320 0.110 0.112* Attention NA 3 0.236 0.059 0.193** 2 0.394 0.078 0.324** D M 0.230 0.051 0.245** 0.327 0.058 0.269** Language NA 4 0.442 0.062 0.245** 5 0.450 0.074 0.249** D M 0.203 0.064 0.112* 0.226 0.071 0.125* Abstraction NA 9 0.141 0.081 0.057 9 0.206 0.101 0.082* D M 0.287 0.083 0.117** 0.355 0.093 0.142** Delayed Recall NA 1 0.612 0.032 0.627** 1 0.666 0.037 0.693** D M 0.218 0.043 0.223** 0.169 0.046 0.176** Orientation NA 2 0.849 0.118 0.641** 3 0.919 0.143 0.703** D M 0.849 0.065 0.641** 0.806 0.072 0.616** Age NA 7 0.0002 0.005 0.001 6 0.006 0.006 0.033 D M 0.035 0.006 0.192** 0.035 0.007 0.193** Education NA 6 0.104 0.019 0.162** 8 0.092 0.023 0.135** D M 0.087 0.021 0.3147** 0.092 0.024 0.086* R 2 0.6677 0.6949 p < 0.05;** p <0.001

PAGE 77

77 Table 4 3. Summary of Stepwise multinomial logistic regression for MoCA subtests and clinician diagnosis of normal aging ( NA ) or dementia ( D M ) among overall population vs. Non Hispanic Blacks Overall Non Hispanic Blacks Variable Dx Order B SE B Order B SE B Visuospatial/Executive NA 5 0.316 0.723 0.241** 3 0.249 0.112 0.170* D M 0.255 0.050 0.195** 0.629 0.192 0.429* Naming NA 8 0.348 0.107 0.1223* D M 0.215 0.096 0.076* Attention NA 3 0.236 0.059 0.193** 5 0.105 0.111 0.080 D M 0.230 0.051 0.245** 0.482 0.144 0.365** Language NA 4 0.442 0.062 0.245** 4 0.520 0.130 0.275** D M 0.203 0.064 0.112* 0.110 0.207 0.058 Abstraction NA 9 0.141 0.081 0.057 D M 0.287 0.083 0.117** Delayed Recall NA 1 0.612 0.032 0.627** 2 0.481 0.073 0.447** D M 0.218 0.043 0.223** 0.365 0.152 0.339* Orientation NA 2 0.849 0.118 0.641** 1 0.658 0.240 0.422* D M 0.849 0.065 0.641** 0.876 0.199 0.562** Age NA 7 0.0002 0.005 0.001 D M 0.035 0.006 0.192** Education NA 6 0.104 0.019 0.162** D M 0.087 0.021 0.3147** R 2 0.6677 0.5903 p < 0.05;** p <0.001

PAGE 78

78 Table 4 4. Summary of Stepwise multinomial logistic regression for MoCA subtests and clinician diagnosis of normal aging ( NA ) or dementia ( D M ) among overall population vs. Hispanics Overall Hispanics Variable Dx Order B SE B Order B SE B Visuospatial/Executive NA 5 0.316 0.723 0.241** 3 0.352 0.201 0.275 D M 0.255 0.050 0.195** 0.729 0.284 0.570** Naming NA 8 0.348 0.107 0.1223* D M 0.215 0.096 0.076* Attention NA 3 0.236 0.059 0.193** D M 0.230 0.051 0.245** Language NA 4 0.442 0.062 0.245** D M 0.203 0.064 0.112* Abstraction NA 9 0.141 0.081 0.057 D M 0.287 0.083 0.117** Delayed Recall NA 1 0.612 0.032 0.627** 2 0.462 0.137 0.436** D M 0.218 0.043 0.223** 0.147 0.224 0.148 Orientation NA 2 0.849 0.118 0.641** 1 1.019 0.559 0.906 D M 0.849 0.065 0.641** 1.527 0.437 1.357** Age NA 7 0.0002 0.005 0.001 D M 0.035 0.006 0.192** Education NA 6 0.104 0.019 0.162** 4 0.162 0.062 0.358* D M 0.087 0.021 0.3147** 0.018 0.087 0.041 R 2 0.6677 0.6506 p < 0.05;** p <0.001

PAGE 79

79 Table 4 5. Adjusted odds ratios and 95% confidence intervals for the association between MoCA items and clinician diagnosis of normal aging or dementia (compared to MCI), adjusted for age and education, by race/ethnicity. Four different models are presented below: to tal population, NH Whites, NH Blacks, and Hispanics. Reference groups for binary test items is 0, scoring incorrectly Subtest/Item Adjusted odds ratios for normal aging (vs. MCI) Adjusted odds ratios for dementia (vs. MCI) AOR 95% CI AOR 95% CI Visuospatial/Executive Trail Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.15 1.34 0.88 1.27 0.91, 1.46 1.00, 1.80* 0.54, 1.44 0.45, 3.61 0.79 0.75 0.46 1.37 0.61, 1.03 0.56, 1.00* 0.18, 1.17 0.27, 7.05 Cube Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.37 1.50 1.32 1.44 1.12, 1.67* 1.18, 1.90* 0.81, 2.16 0.50, 4.18 0.96 0.88 1.15 0.71 0.75, 1.24 0.67, 1.15 0.40, 3.32 0.08, 6.41 Clock Contour Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 2.04 2.11 1.48 0.58 0.98, 4.27 0.90, 4.98 0.19, 11.75 0.02, 14.30 1.10 1.42 0.33 0.10 0.62, 1.97 0.75, 2.67 0.04, 2.93 0.001, 8.55 Clock Number Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.76 2.31 1.12 0.77 1.27, 2.44* 1.55, 3.46** 0.58, 2.16 0.15, 4.04 0.60 0.61 0.44 0.08 0.45, 0.80* 0.44, 0.84* 0.18, 1.10 0.01, 0.63* Clock Hands Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.35 1.34 1.43 1.41 1.09, 1.68* 1.04, 1.72* 0.90, 2.29 0.47, 4.21 0.71 0.75 0.47 0.31 0.55, 0.91* 0.57, 0.99* 0.18, 1.24 0.05, 2.08 Naming Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.45 1.57 1.55 1.90 1.17, 1.77* 1.20, 2.05* 1.03, 2.34* 0.69, 5.25 0.80 0.72 0.81 1.13 0.65, 0.96* 0.58, 0.89* 0.42, 1.56 0.21, 5.93 p <0.05 ** p <0.0001

PAGE 80

80 Table 4 5. Continued Subtest/Item Adjusted odds ratios for normal aging (vs. MCI) Adjusted odds ratios for dementia (vs. MCI) AOR 95% CI AOR 95% CI Attention Digits Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.02 1.17 0.79 1.29 0.81, 1.28 0.88, 1.57 0.49, 1.29 0.49, 3.42 0.80 0.75 1.21 2.43 0.64, 1.01 0.58, 0.97* 0.55, 2.66 0.45, 13.13 Letters Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.93 1.76 2.01 4.39 1.29, 2.89* 1.07, 2.90* 0.89, 4.46 0.53, 36.58 0.50 0.54 0.46 0.07 0.36, 0.68** 0.38, 0.77* 0.17, 1.22 0.004, 1.25 Serial 7s Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.33 1.63 1.18 1.56 1.13, 1.55* 1.32, 2.01** 0.88, 1.59 0.86, 2.81 0.78 0.77 0.54 1.02 0.69, 0.90* 0.66, 0.89* 0.36, 0.83* 0.43, 2.42 Language Repetition Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.54 1.59 1.68 1.36 1.32, 1.81** 1.32, 1.92** 1.17, 2.40* 0.70, 2.66 0.84 0.82 1.06 0.65 0.72, 0.99* 0.68, 0.98* 0.59, 1.90 0.19, 2.23 Fluency Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.63 1.60 1.90 1.12 1.31, 2.04** 1.22, 2.08* 1.19, 3.05* 0.41, 3.05 0.77 0.75 1.11 0.19 0.60, 0.98* 0.57, 0.98* 0.45, 2.73 0.03, 1.14 Abstraction Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.14 1.24 1.16 1.50 0.97, 1.34 1.01, 1.51* 0.83, 1.61 0.73, 3.08 0.75 0.71 0.70 1.17 0.64, 0.89* 0.59, 0.85* 0.39, 1.23 0.31, 4.38 Delayed Recall Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.84 1.94 1.57 1.43 1.73, 1.96** 1.81, 2.09** 1.35, 1.83* 1.04, 1.96* 0.81 0.85 0.78 0.81 0.75, 0.88** 0.77, 0.93* 0.56, 1.08 0.42, 1.55 p <0.05 ** p <0.0001

PAGE 81

81 Table 4 5. Continued Subtest/Item Adjusted odds ratios for normal aging (vs. MCI) Adjusted odds ratios for dementia (vs. MCI) AOR 95% CI AOR Orientation Orientation Date Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 2.42 2.51 1.50 6.86 1.59, 3.70** 1.54, 4.09* 0.51, 4.43 0.56, 84.47 0.51 0.58 0.24 0.54 0.38, 0.68** 0.42, 0.79* 0.07, 0.78* 0.07, 4.46 Orientation Month Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.19 1.42 0.27 >99.99 0.46, 3.08 0.47, 4.29 0.02, 3.52 <0.01, >99.99 0.36 0.35 0.82 0.09 0.20, 0.64* 0.18, 0.65* 0.07, 9.22 0.001, 6.80 Orientation Year Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.43 2.52 1.30 <0.01 0.50, 4.07 0.71, 8.95 0.16, 10.83 <0.01, >99.99 0.38 0.35 0.64 <0.01 0.21, 0.67* 0.20, 0.70* 0.09, 4.76 <0.01, >99.99 Orientation Day Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 3.38 4.01 2.72 4.32 1.74, 6.53* 1.77, 9.08* 0.62, 12.01 0.21, 87.82 0.36 0.37 0.32 0.02 0.25, 0.52** 0.25, 0.55** 0.08, 1.21 0.001, 0.452* Orientation Place Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 2.71 2.56 3.56 6.23 1.74, 4.21** 1.44, 4.55* 1.68, 7.55* 0.24, 159.24 0.44 0.42 0.38 0.40 0.32, 0.59** 0.30, 0.59** 0.15, 0.96* 0.01, 12.60 Orientation City Model 1: Total Model 2: NH White Model 3: NH Black Model 4: Hispanic 1.57 1.75 1.13 0.99 0.64, 3.86 0.61, 5.02 0.11, 11.66 0.03, 38.26 0.61 0.67 1.86 0.07 0.30, 1.23 0.31, 1.46 0.13, 27.37 <0.01, 5.99 p <0.05 ** p <0.0001

PAGE 82

82 CHAPTER 5 INFLUENCE OF RELEVANT RISK FACTORS FOR DEMENTIA ON MONTREAL COGNITIVE ASSESSMENT SCORES AMONG NACC AND HEALTHSTREET PARTICIPANTS Background The global prevalence of dementia in individuals aged 60 and over ranges from 5 to 7%, and the prevalence is expected to increase as the population ages [4] As of 2017, approximately 5.5 million indiv iduals in th e United States [1] Since there is no disease modifying treatment or cure for dementia or AD, it is important to focus on reducing an [59] By addressing risk factors, especially the modifiable ones, we can reduce the burden of dementia. Norton and colleagues found that about one third of the worldwide cases of dementia are attributable to seven potentially modifiable risk factors (diabetes, midlife hypertension, midlife obesity, physical inactivity, depression, sm oking, and low educational attainment) [13] Reducing the prevalence of e ach of these risk factors by 10 to 20% per decade would result in a reduction of between 8 and 15% in the worldwide prevalence of dementia [13] Moreover, according to Association there is enough evidence to conclude that regular physical activity and management of cardiovascular risk factors can reduce the risk of dementia [59] Mild cognitive impairment (MCI) is the stage between normal aging and dementia in which an individual has greater cognitive impairment than is expected for their age, however, it is not as severe as dementia [48,60] A workgroup in the National Institute disting uish between MCI and AD. These criteria include concern about change in an

PAGE 83

83 memory), and preservation of independence in functional abilities, but no dementia [57] Risk factors for MCI are the same as those for AD, however t he most common predictors for progression from MCI to AD include clinical severity of MCI, APOE e4 carrier status, brain atrophy measured through MRI, and several neuroimaging biomarkers [60] A revie w of 41 studies found that less than half of individuals with MCI develop [3] The Montreal Cognitive Assessment (MoCA) was developed as a screening tool for MCI [25] It has a high sensitivity for detection of MCI for patients who would screen as normal on the Mini Mental State Examination (MMSE), another commonly used tool for assessing cognitive impairment [25] The MoCA is scored out of 30 points with a current cut off of 26 for impairment [25] It is easy to administer and contains items on visuospatial function, executive function, verbal memory, language, attention, and orientation [25] Previous studies have found that sociodemographic and health variables can affect performance on cognitive assessments [61] however, this effect has not been widely studied on MoCA scores It has however, been widely established that normative MoCA data differ by decade of age and education [37] and also by race/ethnicity [33 35] but relatively few studies have examined the influence of risk factors for dementia, aside from age and education, on MoCA scores. One study by Freitas and colleagues that did examine the influence of sociodemographic and health variables on MoCA sc ores found that age, educational level, and subjective memory complaints predicted MoCA scores, among their sample

PAGE 84

84 of community based Portuguese cognitively healthy adults [61] Some studies have found that gender predicts MoCA score [62 64] however this has not been consistently observed [61] Moreover, poor performance on cognitive screening tests in the presence of depression has consistently been observed on other cognitive tests, but has not been tested on the MoCA score [65,66] Additionally, accordi ng to the lifecourse perspective, i t is important to examine risk factors as they shape an individual s health throughout their life. All of these exposures differ by race/ethnicity and relevant risk factors for dementia differentially a ffect people throug hout their lives. By examining the effect of these risk factors on MoCA score, we can establish if any of these risk factors are predictive of MoCA score and are therefore especially important in ascertaining a diagnosis of MCI or dementia These analyses aimed to evaluate the effect of relevant risk factors on overall MoCA score in 1) a community dwelling sample and 2) a large research database. Methods Population HealthStreet Data from the first part of this analyses comes from HealthStreet, a community engagement program at the University of Florida. Community Health Workers (CHWs) are trained and trusted members of the communities they serve and they spend every day engaging community members [43,44] CHWs interview individuals using the HealthStreet Intake Form, an in person interview that takes approximately 20 minutes to complete. The Intake F orm contains questions on participants sociodemographic characteristics, health conditions, health concerns, trust in resear ch and research interest, as well as contact information so that participants can be followed up. Older

PAGE 85

85 adults are also asked if they are willing to complete a MoCA. Moreover, if consent is given, the information collected through this process can be sha red with other researchers to potentially link community members to opportunities to participate in research. HealthStreet participants who completed a MoCA, reported their race as White or Black, reported their ethnicity (non Hispanic/Latino or Hispanic/L atino) and were not missing values for dementia predictors were included in these analyses (n=211) Data from the second part of this analyses comes from the NACC, which maintains a database of information f data from the Uniform Data Set (UDS) which has participant information on sociodemographic characteristics, family history, dementia history, and clinician evaluation findings and also includes individuals wit h a wide range of cognitive statuses, measured through clinician assessment and neuropsychological tests [40] Subjects for enrollment at ADCs are typically recruited through clinicians, self referral, family members, or community organizations. All partic ipants and their coparticipants, who are usually a close family member or friend of the participant, must give written informed consent. Data is collected through a standardized evaluation of subjects enrolled in ADCs and their family members [40] Subject s are followed up with annually, however, if their late stage condition hinders them from participating they are followed up for autopsy [41,42] NACC participants who completed a MoCA at their baseline visit, reported their race as White or Black, reporte d their ethnicity (non Hispanic/Latino or Hispanic/Latino), and were not missing values for dementia predictors were included in these analyses (n=3,180).

PAGE 86

86 Measures HealthStreet Overall MoCA score was recorded as a continuous variable from 0 30, with lower scores signifying more impairment. Raw MoCA scores were used; the one point educational adjustment was not included. Sociodemographic information included age, sex, race, ethnicity, education, and food insecurity. Age was recorded as a continuous variable. Sex was coded as female or male. Race and ethnicity were combined into a three level race/ethnicity variable, categorized as non Hispanic White, non Hispanic Black, and Hispanic. Information for our education variable was elicited through the food insecurity variable was the question Those who respon ded yes were coded as food insecure, a proxy for socioeconomic status. Dementia risk factors assessed through HealthStreet include hypertension, diabetes, depression, serious head injury, smoking and family history. Information on these is collected using serious head injury, those who reported yes were coded as having a lifetime history of the respective condition. L y disorder? Those who report a parent

PAGE 87

87 history of AD or another memory disorder were coded as having a family history dementia or other memory disorder. NACC Overall MoCA score was recorded as a continuous score, as described above. Sociodemographic informa tion included age, sex, race, ethnicity, and education. Age was left as a continuous variable and sex was categorized as female or male. Race and ethnicity were combined into a three level race/ethnicity variable, categorized as non Hispanic White, non His panic Black, and Hispanic. Information on education was Dementia risk factors assessed through NACC include hypertension, diabetes, traumatic brain injury, depression, smoking, an d family history of AD. Presence of hypertension, diabetes, traumatic brain injury, and depression are recorded based on participant. Hypertension, diabetes, traumati c brain injury, and depression were originally recorded as absent, recent/active, remote/inactive, and unknown. This was combined into a binary variable with those who were coded with recent/active or remote/inactive as having either lifetime hypertension, lifetime diabetes, a lifetime brain injury, or lifetime depression or not for each disease Information for our smoking ipants were categorized as endorsing lifetime smoking. Lastly, family history of dementia or another memory disorder was coded as yes if participants reported at least one first degree family member with cognitive impairment.

PAGE 88

88 Data Analysis Descriptive stat istics and Chi Square Tests of Independence were calculated for sociodemographic characteristics in each sample. Linear regression was used to determine the effect of each dementia risk factor on total MoCA score in separately 1) the HealthStreet populatio n and 2) the NACC population. Then, stepwise linear regression was used to determine the most important sociodemographic characteristics and risk factors for prediction of MoCA score. All analyses were conducted using SAS software, version 9.4. P values l ess than 0.05 were considered to be statistically significant. Results HealthStreet The 211 individuals incl uded in these analyses were 71.0 years old on average, 58.3% were non Hispanic White, 38.9 % were non Hispanic Black, and 2.8 % were Hispa nic. Over half were female (60.7 %) and participants had a mean of 14.3 years of e ducation. Almost one third (29.9 %) reported food insecurity. More than half report ed lifetime hypertension (61.6 %), one in three lifetime depression (28.9 %), and o ne in four lifetime di abetes (22 .8%). Lifetime serious head injury was reported by 6.2% percent of participants. Over half of participants reported a history of lifetime smoki ng (60.7%) while 20.4 % reported a family history of dementia or another memory disorder (Table 5 1). He althStreet participants included in these analyses had an average MoCA score of 24.2 (range 9.0 to 30.0 ; median: 25.0; IQR: 21.0 27.0 ). Table 5 2 shows the results of linear regression examining risk factors associated with MoCA scores among HealthStreet. Education, a ge, race/ethnicity, and sex were the most important factors (p<0.05) in predicting total MoCA score among this

PAGE 89

89 sample. Results from stepwise regression included education, age race/ethnicity, sex, smoking and serious head injury in the final model (Table 5 3). In this model, education had the strongest standardized beta coefficient, followed by age, sex, race, smoking and serious head injury Serious head injury was included in the final model, however it was not significant (p=0.145) For ea ch year increase in education, the MoCA score increased by 0.448 points. For each year increase in age the MoCA score decreased 0. 1 06 points. Males, non Hispanic Blacks and Hispanics, and lifetime smokers have lower scores, and thus more impairment, than females, Non Hispanic Whites and lifetime non smokers. NACC The 3, 180 individuals in included in these analyses, were 69. 6 years old on the average, 81.8 % were non Hispanic Whit e, 14.1 % were non Hispanic Black, and 4.2% were Hispanic. Over half were fem ale (5 7.8%) and participants had 16.0 years of education on average. Almost half repo rted lifetime hypertension (45.8 %), while fewer participants reported lifetime diabetes (12.3%), lifetime depression (16.0%), or lif etime traumatic brain injury (16 6%). About 40.8 % repo rted lifetime smoking while 62.0 % reported a family history of dementia or another memory disorder (Table 5 4). NACC participants included in these analyses had an average MoCA score of 22.3 (range 0.0 to 30.0; median: 24.0; IQR: 19.0 27.0) Table 5 5 shows the results of a linear regression examining risk factors associated with MoCA scores among NACC participants. Education, a ge, depression sex, family history, smoking history, and hypertension, were the most important factors (p<0.05) in predicting overall MoCA score (Table 5 6). These were the variables included in the final model when using stepwise regression. In this model, education

PAGE 90

90 had the strongest standardized beta coefficie nt followed by age, history of depression, sex, family history, smoking, and hypertension. For each year increase in education, the MoCA score increased by 0.625 points. For each year increase in age, the MoCA score decreased 0.080 points. Males, individu als with lifetime hypertension or depression, individuals with no family history of dementia or other memory disorder, and lifetime non smokers have lower scores, and thus more impairment, than females, individuals with no lifetime depression or hypertensi on, individuals with a family history of dementia or other memory disorder, and lifetime smokers. Discussion Our study showed some consistent relationships in our community dwelling and AD research based sample. In both of our samples, education was the st rongest predictor of MoCA scores. This is to be expected given that dementia is more common among individuals with fewer years of education [12] Moreover, previous work has consistently found that education affects performance on MoCA score [37,61] as wel l as cognitive screening tests in general [64,66] The second strongest predictor in both samples was age. Again, this is in line with the literature. Dementia is more common among older adults and older adults also tend to perform more poorly on the MoCA [37,61] Sex was another important predictor in both samples, with women scoring better than men in both samples. L iterature has consistently found that women outperform men on cognitive screening tests [63] Lastly, lifetime smoking was another important predictor of MoCA scores in both samples, however the direction of the association was different in each sample. In HealthStreet lifetime smokers scoring more poorly on the MoCA which is to be expected given that smoking is associated with cognitive decli ne and dementia [67] In NACC, lifetime smokers scored better on the

PAGE 91

91 MoCA. This may be due to differences in how the smoking variables were categorized; we coded as HealthStreet participants as lifetime smokers if they have ever smoked a cigarette, while w e coded NACC participants as lifetime smokers if they have smoked more than 100 cigarettes in their life, due to the nature of how the data were originally collected. Additionally, in NACC, smoking was highly correlated with a family history of dementia (r influence on overall MoCA score. However, some differences also emerged between our two samples. Among our community dwelling participants (HealthStreet), race/ethnicity emerged as an important predictor of MoCA scores with minority groups scoring lower on the MoCA. This is consistent wit h previous literature that has found that non Hispanic Blacks and Hispanics generally score lower on the MoCA compared to their non Hispanic White counterparts [33 35,54] HealthStreet has a high number of racial/ethnic minority members which may have allo wed for this difference to emerge. Our NACC sample is mostly non Hispanic White (80.7%) so effects of race/ethnicity may be masked due to the small number of minority group members. In our national research sample (NACC), hypertension and depression were also significant predictors of MoCA scores. Hypertension and depression are both documented risk factors for dementia [13] so this finding is consistent with the current literature. Additionally, depression and hypertension were correlated in the NACC sam ple (r=0.64). However, these may not have emerged as important in the HealthStreet sample due to differences in the nature of data collection. All risk factors are determined from self report in HealthStreet while they are determined by self report,

PAGE 92

92 health history, and clinician diagnosis in NACC. Smoking was also highly correlated with depression and hypertension in the HealthStreet sample (r=0.897 and r=0.804, respectively), potentially masking the true effect of depression and hypertension on overall MoC A scores. Family history was another factor that was important among the NACC sample but not the HealthStreet sample. This may be due to the higher prevalence of family history in NACC or may be due to higher salience of the question for NACC participants. Over 60% of participants in the NACC reported a family history of dementia or another memory disorder while only about 20% of participants in HealthStreet reported a family history of dementia or another memory disorder. Some important limitations should be considered for each data source. T he NACC data may not be representative of the general United States population since participation is referral or volunteer based. Additionally, those with normal aging tend to be more educated than the general public. There is a potential for selection bias since some ADCs require participants to consent to autopsy before they can participate. Additionally, the number of non Hispanic Blacks and Hispanics is relatively small compared to the number of non Hispanic Whites included in the sample. However, the NACC data is still a rich source of information with MoCA scores, clinician assessed risk factors, and clinician diagnosis. The HealthStreet data also may not be representative of the general United States population, since convenience sampling is used and non Hispanic Blacks are oversampled. HealthStreet excludes the participation of more impaired individuals since it does not recruit from populations requiring legally authorized representatives (LARs). All of the info rmation on health condition collected through HealthStreet is determined through self report, although, previous literature has

PAGE 93

93 found that there is generally an agreement between self report and medical records [68,69] In spite of these limitations, Healt hStreet is a diverse sample with information on a wide variety of variables. In two very different samples, similar patterns of predictors of total MoCA score emerged. Education was the top predictor of MoCA score in both samples, followed by age. Gender a nd lifetime smoking were important predictors of MoCA score in both samples. Two of these four factors (education and smoking) are considered to be potentially modifiable, emphasizing the importance of addressing risk factors for dementia.

PAGE 94

94 Table 5 1 Sociodemographic characteristics and frequency of risk factors among HealthStreet participants who have completed a MoCA (n=2 11 ) Characteristics HealthStreet (n=2 11 ) Age (SD) 71.0 (6.8) Race/ethnicity NH White NH Black Hispanic 123 (58.3 %) 82 (38.9 %) 6 (2.8% ) Sex Male Female 8 3 (39 3 %) 128 (60.7 %) Education (SD) 14.3 (3. 3 ) Food Insecurity 6 3 (29.9 %) Hypertension 13 0 (61. 6 %) Diabetes 4 8 (22.8 %) Depression 61 ( 28.9 %) Serious Head Injury 13 (6.2%) Smoking History 128 (60.7 %) F amily H istory + 43 (20.4 %) Table 5 2 Linear regression coefficients for risk factors associated with MoCA scores among HealthStreet participants ranked by standardized betas ; MoCA scores range from 0 30 and lower scores signify more impairment, therefore, negative betas represent risk factors for a lower score Variable B SE B P Value Education* 0.445 0.082 0.351 <0.0001 Age* 0.119 0.041 0.190 0.004 Race/Ethnicity* 1.433 0.501 0.188 0.005 Sex* 1.473 0.545 0.171 0.007 Smoking History + 0.946 0.564 0.110 0.095 Serious Head Injury 1.599 1.087 0.091 0.143 Family History + 0.572 0.649 0.055 0.379 Depression 0.505 0.607 0.054 0.406 Hypertension 0.444 0.559 0.051 0.428 Food Insecurity 0.354 0.591 0.038 0.550 Diabetes 0.251 0.637 0.025 0.694 R 2 0.2295 *p<0.05

PAGE 95

95 Table 5 3. Linear regression coefficients for risk factors associated with MoCA scores among HealthStreet participants from stepwise regression ranked by standardized betas ; MoCA scores range from 0 30 and lower scores signify more impairment, therefore, negative betas represent risk factors for a lower score Variable B SE B P Value Education* 0.448 0.080 0.353 <0.0001 Age* 0.106 0.039 0.170 0.007 Race/Ethnicity* 1.360 0.489 0.178 0.006 Sex* 1.423 0.536 0.165 0.009 Smoking History + 1.081 0.535 0.125 0.045 Serious Head Injury 1.561 1.068 0.089 0.145 R 2 0.2385 *p<0.05 Table 5 4. Sociodemographic characteristics and frequency of risk factors among NACC participants who have completed a MoC A at their baseline visit (n=3180 ) Characteristic NACC (n=3180 ) Age (SD) 69.6 (9. 6 ) Race/ethnicity (n=3944) NH White NH Black Hispanic 2600 (81.8 %) 448 (14 1 %) 132 (4.2%) Sex Male Female 1341 (42 .2 %) 1839 (57.8 %) Education (SD) 16.0 (2.8 ) Hypertension 1457 (4 5.8 %) Diabetes 3 84 (12. 1 %) Depression 508 (16.0%) TBI 529 (1 6.6 %) Smoking History 1297 (40. 8 %) Family H istory + 1971 (6 2.0 %)

PAGE 96

96 Table 5 5. Linear regression coefficients for risk factors associated with MoCA scores among NACC participants ranked by standardized betas ; MoCA scores range from 0 30 and lower scores signify more impairment, therefore, negative betas represent risk factors for a lower score Variable B SE B P Value Education* 0.619 0.039 0.276 <0.0001 Age* 0.080 0.011 0.125 <0.0001 Depression* 1.877 0.284 0.110 <0.0001 Sex* 1.284 0.217 0.102 <0.0001 Family History +* 0.804 0.215 0.063 0.0002 Smoking History + 0.612 0.214 0.048 0.004 Hypertension 0.377 0.221 0.039 0.088 Traumatic Brain Injury 0.430 0.282 0.026 0.127 Diabetes 0.273 0.331 0.014 0.409 Race/Ethnicity 0.102 0.216 0.008 0.276 R 2 0.1176 *p<0.05 Table 5 6. Linear regression coefficients for risk factors associated with MoCA scores among NACC participants from stepwise regression ranked by standardized betas ; MoCA scores range from 0 30 and lower scores signify more impairment, therefore, negative betas rep resent risk factors for a lower score Variable B SE B P Value Education* 0.625 0.038 0.279 <0.0001 Age* 0.080 0.011 0.124 <0.0001 Depression* 1.877 0.284 0.110 <0.0001 Sex* 1.277 0.216 0.101 <0.0001 Family History +* 0.807 0.215 0.063 0.0002 Smoking History + 0.609 0.214 0.048 0.005 Hypertension* 0.431 0.215 0.034 0.045 R 2 0.1179 *p<0.05

PAGE 97

97 CHAPTER 6 CONCLUSIONS As the population ages, the prevalence of dementia will increase in the United States [1] Dementia is more common among older adults with fewer years of formal education, as well as among minority group members [1] Early detection is important to allow patients to express their wishes and to help clinicians provide more specialized care, giv en that comorbid diseases are common among older adults [7] .The Montreal Cognitive Assessment (MoCA) was developed to detect Mild Cognitive Assessment [25] an early stage of impairment, however, its applicability in minority populations has not been widel y studied. Studies that have examined its use in minorities have found that the cutoff of 26 is not appropriate [33 35] This dissertation provides a more in depth examination of the use of the MoCA and its subtests, as well as the effect of risk factors o n overall MoCA scores, by race/ethnicity. Main Findings Results from Chapter 3 showed that the current MoCA cutoff of 26 is not the used statistics that calculates th e optimal value which maximizes sensitivity and specificity, found that a cutoff of 25 would improve the accuracy of the MoCA in screening for MCI. Consistent with our hypotheses, s tratification by race/ethnicity resulted in differing cutoffs for MCI among these groups; the optimal cutoff remained 25 among non Hispanic Whites, however, it lowered to 23 and 24 among non Hispanic Blacks and Hispanics, respectively. When cutoffs were further stratified by education between each race/ethnic group, more differen ces emerged. In most cases, as years of education increased so did optimal cutoffs for detection of MCI. Among non Hispanic

PAGE 98

98 Whites, cutoffs by years of education were 22 among those with a high school education ( 12 years) 24 among those with a college ed ucation (13 16 years), and 25 among those with more than a college education (>16 years). Among non Hispanic Blacks, cutoffs by years of education were 19 among those with a high school education ( 12 years) 23 among those with a college education (13 16 years), and 23 among those with more than a college education (>16 years). Among those with an Hispanic ethnicity, cutoffs by years of education were 24 among those with a high school education ( 12 years) 23 among those with a college education (13 16 ye ars), and 24 among those with more than a college education (>16 years). Moreover, in Chapter 3 we also calculated cutoffs to further distinguish between distinguishing betw een MCI and dementia among all of our participants. Stratification by race/ethnicity resulted in differing cutoffs for MCI among these groups; the optimal cutoff remained 19 among non Hispanic Whites, however, it lowered to 16 among non Hispanic Blacks and Hispanics. Further stratification by education among race/ethnic groups reveled some differences in optimal cutoffs. In most cases, as years of education increased, so did optimal cutoffs for detection of dementia. Among non Hispanic Whites, cutoffs by ye ars of education were 17 among those with a high school education ( 12 years) 19 among those with a college education (13 16 years), and 20 among those with more than a college education (>16 years). Among those who reported non Hispanic Black race/ethnic ity, cutoffs by years of education were 13 among those with a high school education ( 12 years) 17 among those with a college education (13 16 years), and 19 among those with more than a college education (>16

PAGE 99

99 years). Among those who reported a n Hispanic ethnicity, cutoffs by years of education were 15 among those with a high school education ( 12 years) 16 among those with a college education (13 16 years), and 19 among those with more than a college education (>16 years). All of the cutoffs calculated w ere lower than the currently used cutoff of 26. Moreover, t hese cutoffs suggest that not only is a cutoff of 26 not appropriate, especially in minority groups, but the one point educational adjustment is also not sufficient to account for differences parti education. We see that in most cases, optimal cutoffs differ more than one point for those with less than a high school education (who receive the adjustment) from the cutoffs in the two other education groups. These findings are consistent with previous work that found that lowering the cutoff of 26 would be clinically useful [29] especially in minority populations [33 35] and that cutoffs should be stratified by educational attainment [37] In Chapter 4 we further examined the MoCA and analyzed performance of the subtests and items by race and ethnic group. Using stepwise regression, we predicted which subtests of the MoCA were significant in predicting clinician diagnosis. Consistent with ou r hypotheses, we found that subtests and items differentially discriminated between MCI and normal aging or dementia, and these subtests and items differed in each race/ethnic group. The ability to discriminate between levels of cognitive impairment is cli nically useful. We found that not all subtests and items demonstrated clinical utility. Among the overall sample and among non Hispanic Whites, all subtests were significant predictors; age and education continued to be important predictors of

PAGE 100

100 score. Howe ver, only the visuospatial/executive, attention, language, delayed recall, and orientation subtests predicted clinician diagnosis among the group of non Hispanic Blacks while only visuospatial/executive, delayed recall, and orientation subtests, as well as education, predicted clinician diagnosis among the group of Hispanics. Analyses such as this one will be useful for the field as additional screening tests are developed, some of which only cover one type of item, such as the use of a digital pen for cloc k drawing, Further deconstruction of the subtests into items was consistent with results from the previous stepwise regressions. In our overall model, items from all subtests were significant in predicting clinician diagnosis consistent with the results of the final stepwise regression that included all subtests. Results from the model among non Hispanic Whites were similar to those of the overall model. Among non Hispanic Whites the trails, clock number, clock hand, naming, letters, serial 7s, repetition fluency items, abstraction, delayed recall, orientation date, orientation day, and orientation place items predicted MCI from both normal aging and dementia while the c ube item only predicted MCI from normal aging and the digit, orientation month, and or ientation year items only predicted MCI from dementia. However, less items were significant among the non Hispanic Black and Hispanic groups, which was to be expected, given that the final non Hispanic Black model and the final Hispanic model included fewe r subtests than the overall model. Among non Hispanic Blacks the naming, repetition, fluency, delayed recall, and orientation place items discriminated between normal aging and MCI while the serial 7s and orientation date items discriminated between MCI and dementia. Among Hispanics,

PAGE 101

101 delayed recall discriminated between normal aging and MCI while the clock numbers and orientation day discriminated items between MCI and dementia. Moreover, delayed recall was significant in predict ing normal aging from MCI in all four models. The clock contour and orientation city items did not discriminate between normal aging and MCI or MCI and dementia in either of the four models; in other words, these items did not improve screening accuracy in any population tested. These findings were consistent with previous literature. Reviewed literature found that delayed recall was significant in distinguishing between normal aging and any impairment, normal aging and MCI, and MCI and dementia [30 32] Mo reover, clock contour likely has a ceiling effect. Previous work has found that most individuals get full points on this item. One study found that 100 individuals in the healthy cognition group scored full points while 90.4% of individuals in the AD group scored full points [31] Not at all items may be culturally appropriate and future studies need to further evaluate items for use in screening of dementia in these minority groups. Finally, in Chapter 5 we examined the effect of relevant risk factors fo r dementia on MoCA scores in a community dwelling sample and a national research database to determine which risk factors are predictive of MoCA score and especially important in ascertaining a clinician diagnosis. Moreover, we wanted to rank the risk fact ors in each sample to compare across samples, especially given that the NACC sample is not representative of the general population. Consistent with our hypotheses, we found that known dementia risk factors were associated with lower MOCA scores. We found that among both the HealthStreet and NACC samples, low education al attainment and age were the top two risk factors for predicting MoCA scores. The literature has consistently

PAGE 102

102 found that educational attainment and age predict MoCA score and our findings ad d to this literature [37,61] In both samples, sex and lifetime smoking were also significant predictors. Women generally outperform men on cognitive tests [63] However, lifetime smoking was a risk factor among HealthStreet participants and a protective f actor among NACC participants. This may be due to differences in how the se variables were defined. In HealthStreet lifetime smoking was defined as ever smoking a cigarette while in NACC this was defined as having smoked at least 100 cigarette s lifetime, Others have consistently found smoking to be a risk factor related to dementia [70] Additionally, smoking was highly correlated with a positive family history of dementia, which may explain this unexpected finding. Future research may clarify these unexpected and inconsistent findings. Other differences in risk factors associated with the MoCA score s emerged. In HealthStreet race/ethnicity was a significant predictor of MoCA scores while in NACC hypertension, depression, and family history of dementi a or another memory disorder were important in predicting total MoCA score. T he fact that hypertension and depression were significant predictor s of the MoCA score in NACC may be attributable to the fact that hypertension and depression were clinically no t subjectively, evaluated in NACC. I n HealthStreet, people may identify themselves as having high blood pressure when their clinician would consider it pre hypertension. While HealthStreet CHWs do measure hypertension at the time of intake, all variables i ncluded in these analyses were self Additionally, this may result in more severe depression being diagnosed in NACC, while self report of depression in

PAGE 103

103 HealthStreet may include milder depressive symptoms. Thus differences in findings on hypertension and depression may be more a function of how they were measured that differences in the risk associated with each. Other research has found that hyperten sion and depression are associated with dementia [46,71] T he t emporality of these associations w as not measured in th e s e cross sectional analyses. It is clinically important to find risks factors associated with lower MoCA scores as we have done in Chapter 5. Older adults fear dementia and are motivated to reduce their risk for this health outcome. Clinicians can help people focus on reducing key, modifiable risk factors for dementia, such as preventing and controlling hypertension and quitting smok ing. They can also provide their patients with plans to increase their physical and cognitive exercise [59,72] Conceptual Framework Revisiting the theoretical model that framed this dissertation, a lifecourse view helps explain our findings. Factors thro ughout the lifecourse must be considered when examining cognitive test performance. In Chapter 3 we found that not only did optimal cutoffs for detection of MCI and dementia differ by race/ethnicity, but additionally by educational attainment. Low educati onal attainment is a risk factor for dementia with individuals with fewer years of formal education at a greater risk for developing dementia [11 14] life experience and may initiate a particular trajec tory of life conditions which may shape how one cognitively ages [50] Additionally, r acial/ethnic differences persist in educational attainment and quality. In Chapter 4 we found that not all MoCA subtests and items were clinically useful in racial/ethn ic minority groups. This is consistent with our lifecourse framework,

PAGE 104

104 given that this framework explains that performance on cognitive testing is affected by among racial an d ethnic minority group members. This burden of exposures may affect performance on cognitive testing. Therefore, it is imperative that we test the applicability of cognitive screening tools in racial/ethnic groups to ensure that they are appropriate for u se in all groups in which they will be used. Lastly, in Chapter 5 the effect of lifetime risk factors on MoCA scores were examined. According to the lifecourse perspective of cognitive aging, the presence of life affects their health outcomes. These risk factors are unevenly distributed among racial/ethnic minority group members and can explain the larger burden of dementia as well as the lower MoCA scores among these populations. The factors examined in this dissertation, and the consequences of their accumulation throughout the lifecourse may explain the di fference in dementia burden and in MoCA performance among racial/ethnic minority group members. Strengths and Limitations Overall this dissertation provides an in depth examination of the MoCA, MoCA subtests, and the effect of risk factors on total MoCA score, by race/ethnicity. However, there are some limitations to consider. NACC The NACC cohort is not a representative sample of the U.S. population because subjects with dementia are referral or volunteer based participants and normal aging controls tend to be highly educated. There also is the potential for selection bias given can

PAGE 105

105 participat e This may result in a more homogenous sample then if we included individuals who would not consent to autopsy. There is also potential bias due to the timing of evaluation. The MoCA is administered before clinician examination, potentially informing the clinician diagnosis to agree with the MoCA score. However, this is the norm in the field; clinicians usually see screening results before they make their diagnosis. The naming subtest was coded from zero to three rather than each naming item so we could not establish if any of the naming items discriminate between MCI and normal aging or dementia. Additionally, there is no information on literacy level which has been proven to explain cognitive decline better than years of education There i s also no information on participant immigration status which is an important predictor of cognitive health ; immigrants tend to have better health than comparable non immigrants However, this is still a unique data source. It contains baseline MoCAs on a lmost 4,000 subjects from less than 40 years old to over 90 years old with a wide range of cognitive statuses. Additionally, there is a clinician diagnosis for every participant who has completed a MoCA allowing us to calculate optimal cutoffs Even thoug h the number of racial/ethnic minority members included is small, the fact that they are included is still notable, given that many studies focus mostly on non Hispanic Whites. HealthStreet The HealthStreet cohort may not be a representative sample of the U.S. population given that the H ealthStreet cohort is recruited using convenience sampling and non Hispanic Blacks are oversampled Moreover the HealthStreet cohort does not include more impaired individuals who may need a legally authorized representativ e (LAR) to consent for them. The health conditions collected through the HealthStreet

PAGE 106

106 interview are determined through self report, however, this has previously been found to generally be in agreement with medical records [68,69] Additionally, we could no t examine immigration status with HealthStreet data because the vast majority of individuals who completed the MoCA were born in the United States (96.2%). Despite these limitations, the HealthStreet cohort is still a diverse community based sample with a large proportion of minority members. HealthStreet CHWs also collect real time information on a wide variety of variables, including sociodemographic characteristics, health concerns and conditions, and AD history. Public Health Consequence and Future Research As the population ages, more individuals will be at risk for developing dementia. The number of individuals living with AD in the United States is projected to increase to as high as 16 million by 205 0 [1] In the absence of a cure, e arly detection is critical to detect cases of MCI and dementia and improve quality of life for those living with dementia as well as their loved ones and caregivers. Early detection allows patients to express their wishes before their impairment becomes too severe, make legal and financial arrangements, make advanced health directives, and enroll in clinical trials, if they are interested [7] Moreover, timely diagnosis helps health care professionals better care for patien ts and manage their comorbid conditions that usually occur with greater age [7] Not only is timely diagnosis important to improve patient outcomes, but it also can have an economic impact. It is projected that dementia will cost the U.S. up to 1.1 trillio n dollars in 2050 [1] A recent study modeled the impact of a dementia screening, diagnosis, and management program in primary care setting s and found that implementing this type of program could save Medicare & Medicaid up to $22 billion in

PAGE 107

107 2025 and $29 b illion in 2050 [73] Additional l y appropriate screening will allow resources to be invested where they are needed. The MoCA is a valuable tool that can be used to detect MCI and dementia earlier than other screening tools [25] Given that initial screeni ng with the MoCA is relatively inexpensive, refining its use and increasing its use in primary care settings would identify individuals who truly need further evaluation and care and could reduce the resources spent on false positives. This is especially i mportant among racial/ethnic minorities, given that the current cutoff of the MoCA results in a high proportion of false positives. This dissertation is among one of the first studies to c alculate optimal MoCA cutoffs in a diverse national sample by race/e thnicity and education These analyses suggest that the current MoCA cutoff may not be appropriate for use in racial/ethnic minorities. Our results show that optimal cutoffs for detection of MCI and dementia among non Hispanic Whites are higher than those for non Hispanic Blacks and Hispanics. Additionally, optimal cutoffs decreased as years of education decreased. These findings suggest that cutoffs should be stratified by race/ethnicity and then education to properly screen minority group members for MCI and dementia. Use of these cutoffs should be used going forward to help identify individuals in need of more rigorous screening and to aid clinicians in early diagnosis and treatment of MCI and dementia. Additionally, this dissertation is among one of the first studies to examine MoCA subtests and items by race/ethnicity; subtests analyses show that not all subtests are crucial in predicting clinician diagnosis among non Hispanic Blacks and Hispanics. This is important to address given that these subtests assess a wide variety of cognitive domains. Items may not be culturally relevant to all groups. Lastly, this dissertation is

PAGE 108

108 among one of the first studies to explore which risk factors are important in prediction of MoCA score in a national research based sample and a local community sample. A ge, education, gender, and smoking status consistently predicted MoCA score in both of our samples. Education and smoking are both potentially modifiable risk factors and should be addressed. Future research Future research should examine optimal cutoffs and the ability of MoCA subtests to predict clinician diagnosis in a sample with more minority group members. Given that over half of the NACC sample has a positive family history for dementia or other memory disorder, future research should examine optimal cutoffs in a population of individuals with no family history and compare their findings to ours. Additionally, the role of literacy should be examined to see if it better expla ins differences in cutoffs and subtest performance compared to years of education. Lastly, future research should examine the effect of country of origin in differences in MoCA performance among Hispanic individuals, given the heterogeneity of this ethnic group. Association that will affect future research in the field and in this area. The Imaging Dementia Evidence for Amyloid Scanning (IDEAS) study will help provide information on the consequences of early MCI and dementia screening using PET scans [74] Its findings are intended to help shape public policy on cognitive screening. Positive consequences should also be applicable to other early screening efforts. In addition, the US Study to Protect Brain Health Through Lifestyle Intervention to Reduce Risk (US POINTER) study, one of many international replications of the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER) and part of

PAGE 109

109 the World FINGER study is investigating the effects of interven ing in modifiable lifestyle and vascular risk factors [75] These studies will help refine our understanding of what risk factors should be targeted to reduce the prevalence of dementia in the future. While study sites should provide enough diversity to test its application in other populations, efforts to include minority populations in the US in AD research must continue. Future research efforts will also need to refine recruitment and retention eff orts for minority populations in AD research so that findings are generalizable to the population as a whole not specific groups

PAGE 110

110 LIST OF REFERENCES [1] Dement J Alzheim ers Assoc 2017;13:325 73. [2] Petersen RC, Caracciolo B, Brayne C, Gauthier S, Jelic V, Fratiglioni L. Mild cognitive impairment: a concept in evolution. J Intern Med 2014;275:214 28. doi:10.1111/joim.12190. [3] Mitchell AJ, Shiri Feshki M. Rate of progres sion of mild cognitive impairment to dementia -meta analysis of 41 robust inception cohort studies. Acta Psychiatr Scand 2009;119:252 65. doi:10.1111/j.1600 0447.2008.01326.x. [4] Prince M, Bryce R, Albanese E, Wimo A, Ribeiro W, Ferri CP. The global preva lence of dementia: A systematic review and metaanalysis. Alzheimers Dement 2013;9:63 75.e2. doi:10.1016/j.jalz.2012.11.007. [5] Langa KM, Larson EB, Crimmins EM, Faul JD, Levine DA, Kabeto MU, et al. A Comparison of the Prevalence of Dementia in the United States in 2000 and 2012. JAMA Intern Med 2017;177:51 8. doi:10.1001/jamainternmed.2016.6807. [6] Hurd MD, Martorell P, Delavande A, Mullen KJ, Langa KM. Monetary Costs of Dementia in the United States. N Engl J Med 2013;368:1326 34. doi:10.1056/NEJMsa1204629. [7] 2017. https://alzimpact.org/img/Policy_Brief_Early_Detection_and_Diagnosis_Brief_AIM. pdf (accessed March 26, 2018). [8] Mayeda ER, Glymour MM, Quesenberry CP, Whitmer RA. Inequalities in dementia incidence between six racial and ethnic groups over 14 years. Alz heimers Dement J Alzheimers Assoc 2016;12:216 24. doi:10.1016/j.jalz.2015.12.007. [9] Chin AL, Negash S, Hamilton R. Diversity and disparity in dementia: the impact of ethnoracial differences in Alzheimer disease. Alzheimer Dis Assoc Disord 2011;25:187 95. doi:10.1097/WAD.0b013e318211c6c9. [10] Yaffe K, Falvey C, Harris TB, Newman A, Satterfield S, Koster A, et al. Effect of socioeconomic disparities on incidence of dementia among biracial older adults: prospective study. BMJ 2013;347:f7051. doi:10.1136/bmj .f7051. [11] Beydoun MA, Beydoun HA, Gamaldo AA, Teel A, Zonderman AB, Wang Y. Epidemiologic studies of modifiable factors associated with cognition and dementia: systematic review and meta analysis. BMC Public Health 2014;14:643. doi:10.1186/1471 2458 14 643.

PAGE 111

111 [12] Contador I, Ser T del, Llamas S, Villarejo A, Benito Len J, Bermejo Pareja F. Impact of literacy and years of education on the diagnosis of dementia: A population based study. J Clin Exp Neuropsychol 2017;39:112 9. doi:10.1080/13803395.2016.1204 992. [13] Norton S, Matthews FE, Barnes DE, Yaffe K, Brayne C. Potential for primary based data. Lancet Neurol 2014;13:788 94. doi:10.1016/S1474 4422(14)70136 X. [14] Then FS, Luck T, Angermeyer MC, Riedel Heller SG. Education as protector against dementia, but what exactly do we mean by education? Age Ageing 2016;45:523 8. doi:10.1093/ageing/afw049. [15] Xu W, Tan L, Wang H F, Tan M S, Tan L, Li J Q, et al. Education and Risk of Dementia: Dose Re sponse Meta Analysis of Prospective Cohort Studies. Mol Neurobiol 2016;53:3113 23. doi:10.1007/s12035 015 9211 5. [16] Glymour MM, Manly JJ. Lifecourse social conditions and racial and ethnic patterns of cognitive aging. Neuropsychol Rev 2008;18:223 54. do i:10.1007/s11065 008 9064 z. [17] Bennett DA, Schneider JA, Wilson RS, Bienias JL, Arnold SE. Education modifies the association of amyloid but not tangles with cognitive function. Neurology 2005;65:953 5. doi:10.1212/01.wnl.0000176286.17192.69. [18] Rolst ad S, Nordlund A, Eckerstrm C, Gustavsson MH, Blennow K, Olesen PJ, et al. High education may offer protection against tauopathy in patients with mild cognitive impairment. J Alzheimers Dis JAD 2010;21:221 8. doi:10.3233/JAD 2010 091012. [19] Stern Y. Cog 2012;11:1006 12. doi:10.1016/S1474 4422(12)70191 6. [20] Stern Y. What is cognitive reserve? Theory and research application of the reserve concept. J Int Neuropsychol Soc JINS 2002;8:448 60. [21] Stern Y. Cognitive reserve. Neuropsychologia 2009;47:2015 28. doi:10.1016/j.neuropsychologia.2009.03.004. [22] Liu Y, Julkunen V, Paajanen T, Westman E, Wahlund L O, Aitken A, et al. -evidence fr om structural MRI analysis. Neuroradiology 2012;54:929 38. doi:10.1007/s00234 012 1005 0. [23] Foubert Samier A, Catheline G, Amieva H, Dilharreguy B, Helmer C, Allard M, et al. Education, occupation, leisure activities, and brain reserve: a population bas ed study. Neurobiol Aging 2012;33:423.e15 25. doi:10.1016/j.neurobiolaging.2010.09.023.

PAGE 112

112 [24] Manly JJ, Schupf N, Tang M X, Stern Y. Cognitive decline and literacy among ethnically diverse elders. J Geriatr Psychiatry Neurol 2005;18:213 7. doi:10.1177/08919 88705281868. [25] Nasreddine ZS, Phillips NA, Bdirian V, Charbonneau S, Whitehead V, Collin I, et al. The Montreal Cognitive Assessment, MoCA: A Brief Screening Tool For Mild Cognitive Impairment. J Am Geriatr Soc 2005;53:695 9. doi:10.1111/j.1532 5415.20 05.53221.x. [26] Damian AM, Jacobson SA, Hentz JG, Belden CM, Shill HA, Sabbagh MN, et al. The Montreal Cognitive Assessment and the Mini Mental State Examination as Screening Instruments for Cognitive Impairment: Item Analyses and Threshold Scores. Dement Geriatr Cogn Disord Basel 2011;31:126 31. doi:http://dx.doi.org/10.1159/000323867. [27] Markwick A, Zamboni G, de Jager CA. Profiles of cognitive subtest impairment in the Montreal Cognitive Assessment (MoCA) in a research cohort with normal Mini Mental S tate Examination (MMSE) scores. J Clin Exp Neuropsychol 2012;34:750 7. doi:10.1080/13803395.2012.672966. [28] Luis CA, Keegan AP, Mullan M. Cross validation of the Montreal Cognitive Assessment in community dwelling older adults residing in the Southeaster n US. Int J Geriatr Psychiatry 2009;24:197 201. doi:10.1002/gps.2101. [29] Carson N, Leach L, Murphy KJ. A re examination of Montreal Cognitive Assessment (MoCA) cutoff scores. Int J Geriatr Psychiatry 2017. doi:10.1002/gps.4756. [30] Horton DK, Hynan LS, Lacritz LH, Rossetti HC, Weiner MF, Cullum CM. An Abbreviated Montreal Cognitive Assessment (MoCA) for Dementia Screening. Clin Neuropsychol 2015;29:413 25. doi:10.1080/13854046.2015.1043349. [31] Cecato JF, Martinelli JE, Izbicki R, Yassuda MS, Aprahamian I. A subtest analysis of the Montreal cognitive assessment (MoCA): which subtests can best disease? Int Psychogeriatr 2016;28:825 32. doi:10.1017/S1041610215001982. [32] Roal f DR, Moore TM, Wolk DA, Arnold SE, Mechanic Hamilton D, Rick J, et al. Defining and validating a short form Montreal Cognitive Assessment (s MoCA) for use in neurodegenerative disease. J Neurol Neurosurg Psychiatry 2016;87:1303 10. doi:10.1136/jnnp 2015 3 12723. [33] Goldstein FC, Ashley AV, Miller E, Alexeeva O, Zanders L, King V. Validity of the montreal cognitive assessment as a screen for mild cognitive impairment and dementia in African Americans. J Geriatr Psychiatry Neurol 2014;27:199 203. doi:10.117 7/0891988714524630.

PAGE 113

113 [34] Rossetti HC, Lacritz LH, Hynan LS, Cullum CM, Van Wright A, Weiner MF. Montreal Cognitive Assessment Performance among Community Dwelling African Americans. Arch Clin Neuropsychol Off J Natl Acad Neuropsychol 2017;32:238 44. doi:10.1093/arclin/acw095. [35] Sink KM, Craft S, Smith SC, Maldjian JA, Bowden DW, Xu J, et al. Montreal Cognitive Assessment and Modified Mini Mental State Examination in African Americans. J Aging Res 2015;2015:872018. doi:10.1155/2015/ 872018. [36] Bernstein IH, Lacritz L, Barlow CE, Weiner MF, DeFina LF. Psychometric evaluation of the Montreal Cognitive Assessment (MoCA) in three diverse samples. Clin Neuropsychol 2011;25:119 26. doi:10.1080/13854046.2010.533196. [37] Rossetti HC, Lacri tz LH, Cullum CM, Weiner MF. Normative data for the Montreal Cognitive Assessment (MoCA) in a population based sample. Neurology 2011;77:1272 5. doi:10.1212/WNL.0b013e318230208a. [38] Hogervorst E, Rahardjo TB, Bandelow S. Cross cultural validation of a de mentia screening test. Alzheimers Dement J Alzheimers Assoc 2011;7:S160 1. doi:10.1016/j.jalz.2011.05.433. [39] Allostatic Load Scores Among Blacks and Whites in the United States. Am J Public Health 2006;96:826 33. doi:10.2105/AJPH.2004.060749. [40] Beekly DL, Ramos EM, Lee WW, Deitrich WD, Jacka ME, Wu J, et al. The National Alzheimer Dis Assoc Disord 2007;21:24 9 58. doi:10.1097/WAD.0b013e318142774e. [41] Morris JC, Weintraub S, Chui HC, Cummings J, Decarli C, Ferris S, et al. The Uniform Data Set (UDS): clinical and cognitive variables and descriptive data from Alzheimer Disease Centers. Alzheimer Dis Assoc Diso rd 2006;20:210 6. doi:10.1097/01.wad.0000213865.09806.92. [42] Weintraub S, Salmon D, Mercaldo N, Ferris S, Graff Radford NR, Chui H, et al. Neuropsychological Test Battery. Alzheimer Dis Assoc D isord 2009;23:91 101. doi:10.1097/WAD.0b013e318191c7dd. [43] Years. Oxford University Press; 2011. [44] ng the Community in Research with the HealthStreet Model: National and International Perspectives 2013;3:98 109. doi:10.1159/000343144.

PAGE 114

114 [45] Youden WJ. Index for rating diagnostic tests. Cancer 1950;3:32 5. doi:10.1002/1097 0142(1950)3:1<32::AID CNCR282003 0106>3.0.CO;2 3. [46] Richard E, Reitz C, Honig LH, Schupf N, Tang MX, Manly JJ, et al. Late Life Depression, Mild Cognitive Impairment, and Dementia. JAMA Neurol 2013;70:383 9. doi:10.1001/jamaneurol.2013.603. [47] van der Flier WM, Scheltens P. Epidemiol ogy and risk factors of dementia. J Neurol Neurosurg Psychiatry 2005;76:v2 7. doi:10.1136/jnnp.2005.082867. [48] Roberts R, Knopman DS. Classification and Epidemiology of MCI. Clin Geriatr Med 2013;29. doi:10.1016/j.cger.2013.07.003. [49] Langa KM, Levine DA. The Diagnosis and Management of Mild Cognitive Impairment: A Clinical Review. JAMA 2014;312:2551 61. doi:10.1001/jama.2014.13806. [50] Mirowsky J, Ross C. Education, Social Status and Health. Soc Forces 2005;84:609 10. doi:10.1353/sof.2005.0097. [51] P using two criteria based on the receiver operating characteristic curve. Am J Epidemiol 2006;163:670 5. doi:10.1093/aje/kwj063. [52] Sc hoonjans F. MedCalc. MedCalc 2018 https://www.medcalc.org/manual/howtociteMedCalc.php (accessed June 2, 2018). [53] Davis DH, Creavin ST, Yip JL, Noel Storr AH, Brayne C, Cullum S. Montreal Cognitive Assessment for the diagnosis of dementias. Cochrane Database Syst. Rev., John Wiley & Sons, Ltd; 2015. [54] Galvin JE, Tolea MI, Chrisphonte S, Saravia K. RACIAL AND ETHNIC DISPARITIES IN MEASURES OF COGNITIVE FUNCTION: IMPACT OF SOCIOECONOMIC STATUS, PHYSI CAL FUNCTION, AND MOOD. Alzheimers Dement J Alzheimers Assoc 2016;12:P392 3. doi:10.1016/j.jalz.2016.06.741. [55] Manly JJ, Jacobs DM, Sano M, Bell K, Merchant CA, Small SA, et al. Cognitive test performance among nondemented elderly African Americans and whites. Neurology 1998;50:1238 45. [56] Malek Age and education adjusted normative data for the Montreal Cognitive Assessment (MoCA) in older adults age 70 99. Aging Neuropsychol Cogn 20 15;22:755 61. doi:10.1080/13825585.2015.1041449.

PAGE 115

115 [57] Albert MS, DeKosky ST, Dickson D, Dubois B, Feldman HH, Fox NC, et al. The Recommendations from the National Institute on Aging Alzhe Alzheimers Assoc 2011;7:270 9. doi:10.1016/j.jalz.2011.03.008. [58] 72. doi:10.1056/NEJM p1107288. [59] Baumgart M, Snyder HM, Carrillo MC, Fazio S, Kim H, Johns H. Summary of the evidence on modifiable risk factors for cognitive decline and dementia: A population based perspective. Alzheimers Dement 2015;11:718 26. doi:10.1016/j.jalz.2015.05.016. [60] Petersen RC, Roberts RO, Knopman DS, Boeve BF, Geda YE, Ivnik RJ, et al. Mild cognitive impairment: ten years later. Arch Neurol 2009;66:1447 55. doi:10.1001/archneurol.2009.266. [61] Freitas S, Simes MR, Alves L, Santana I. Montreal Cognitive Assessment: influence of sociodemographic and health variables. Arch Clin Neuropsychol Off J Natl Acad Neuropsychol 2012;27:165 75. doi:10.1093/arclin/acr116. [62] Ribeiro PCC, Oliveira BHD, Cupertino APFB, Neri AL, Yassuda MS. Performance of the elderly in the CERAD cognitive battery: relations with socio demographic variables and perceived health. Psicol Reflex E Crtica 2010;23:102 9. doi:10.1590/S0102 79722010000100013. [63] Scazufca M, Almeida OP, Valla da HP, Tasse WA, Menezes PR. Limitations of the Mini Mental State Examination for screening dementia in a community with low socioeconomic status: results from the Sao Paulo Ageing & Health Study. Eur Arch Psychiatry Clin Neurosci 2009;259:8 15. doi:10.100 7/s00406 008 0827 6. [64] Anderson TM, Sachdev PS, Brodaty H, Trollor JN, Andrews G. Effects of sociodemographic and health variables on Mini Mental State Exam scores in older Australians. Am J Geriatr Psychiatry Off J Am Assoc Geriatr Psychiatry 2007;15:4 67 76. doi:10.1097/JGP.0b013e3180547053. [65] Gallacher JE, Elwood PC, Hopkinson C, Rabbitt PM, Stollery BT, Sweetnam PM, et al. Cognitive function in the Caerphilly study: associations with age social class, education and mood. Eur J Epidemiol 1999;15:161 9. [66] Moraes C, Pinto JA, Lopes MA, Litvoc J, Bottino CMC. Impact of sociodemographic and health variables on mini mental state examination in a community based sample of older people. Eur Arch Psychiatry Clin Neurosci 2010;260:535 42. doi:10.1007/s0040 6 010 0104 3.

PAGE 116

116 [67] Ott A, Andersen K, Dewey ME, Letenneur L, Brayne C, Copeland JRM, et al. Effect of smoking on global cognitive function in nondemented elderly. Neurology 2004;62:920 4. [68] Johansson J, Hellnius ML, Elofsson S, Krakau I. Self report as a selection instrument in screening for cardiovascular disease risk. Am J Prev Med 1999;16:322 4. [69] Okura Y, Urban LH, Mahoney DW, Jacobsen SJ, Rodeheffer RJ. Agreement between self report questionnaires and medical record data was substantial for diab etes, hypertension, myocardial infarction and stroke but not for heart failure. J Clin Epidemiol 2004;57:1096 103. doi:10.1016/j.jclinepi.2004.04.005. [70] Zhong G, Wang Y, Zhang Y, Guo JJ, Zhao Y. Smoking Is Associated with an Increased Risk of Dementia: A Meta Analysis of Prospective Cohort Studies with Investigation of Potential Effect Modifiers. PLoS ONE 2015;10. doi:10.1371/journal.pone.0118333. [71] Rnnemaa E, Zethelius B, Lannfelt L, Kilander L. Vascular risk factors and dementia: 40 year follow up of a population based cohort. Dement Geriatr Cogn Disord 2011;31:460 6. doi:10.1159/000330020. [72] National Academies of Sciences, Engineering, and Medicine, Health and Medicine Division, Board on Health Sciences Policy, Committee on Preventing Dementia a nd Cognitive Impairment. Preventing Cognitive Decline and Dementia: A Way Forward. Washington (DC): National Academies Press (US); 2017. [73] Borson S, Scanlan J, Hummel J, Gibbs K, Lessig M, Zuhr E. Implementing Routine Cognitive Screening of Older Adults in Primary Care: Process and Impact on Physician Behavior. J Gen Intern Med 2007;22:811 7. doi:10.1007/s11606 007 0202 8. [74] Rabinovici GD, Hillner B, Whitmer RA, Carrillo M, Gatsonis C, Siegel B. Imaging dementia: Evidence for amyloid scanning (IDEAS) a national study to evaluate the clinical utility of amyloid PET. Alzheimers Dement J Alzheimers Assoc 2015;11:P263 4. doi:10.1016/j.jalz.2015.07.341. [75] http://alz.org/us pointer/overview.asp (accessed June 11, 2018).

PAGE 117

117 BIOGRAPHICAL SKETCH Sadaf Are fi Milani received her Bachelor of Science in biology, with a specialization in biotechnology, from the Univers ity of Florida in May 2013. The following August she began working on her Master of Public Health (MPH) with a concentration on epidemiology at the University of Florida During her master s she worked as a research assistant at HealthStreet, a community e ngagement program at the University of Florida. For her final master s project, she completed an internship at the Global Malaria Programme at the World Health Organization headquarters, where she examined insecticide treated net distribution campaigns in Sub Saharan Africa. S he finished her master s in May 2015 and received the MPH Exemplary Student Award as well as the Best All Around MPH E pidemiology Student Award. Sadaf was awarded a Graduate School Fellowship and began her Doctor of Philosophy (PhD) training at the University of Florida in August 2015, under the mentorship of Dr. Catherine Striley. During her PhD she was involved on cam pus as the President of the College of Public Health and Health Professions Doctoral Student Council and as an active member of Alpha Epsilon Lambda, the graduate student honor society. She worked as a research assistant on the Preventing Hospital Readmiss ion through CHWs Study and the Medication Use, Safety, and Evidence Study. She also worked as a teaching assistant for six semesters for Public Health Concepts and Principles of Epidemiology. Sadaf has several publi cations and has presented her work at national and international University of Florida Graduate School Council, and University of Florida Department of Fund. Sadaf completed her PhD in August of 2018.