DEMOGRAPHICS AND COGNITIVE IMPAIRMENT AS DEFINED BY THE MONTREAL COGNITIVE ASSESSMENT IN A PHOENIX COMMUNITY MEMORY SCREEN A Thesis submitted to the University of Arizona College of Medicine ‐‐ Phoenix in partial fulfillment of the requirements for the Degree of Doctor of Medicine Christine Parsons Class of 2015 Mentors: Roy Yaari, MD, MAS and Jan Dougherty, RN, MS Dedication: I dedicate this work to my loving parents – you are my greatest support. Thank you for raising me with a passion for knowledge and, more importantly, a passion to care for others. Acknowledgements: I am very grateful for the time and dedication of those who helped me complete this thesis: Dr. Roy Yaari and Jan Dougherty for their mentorship; Dr. Chengcheng Hu for his great contribution of time as a statistical resource and advisor; Dr. Jessica Langbaum and Dr. Yui Napatkamon for their statistical and project development guidance; Dr. Sandra Jacobson for her outside review of the work; and Jan Dougherty and Liz Benyo, for providing the Phoenix community memory screen data. Abstract: Memory screening in the community promotes early detection of memory problems, as well as Alzheimer’s disease (AD) and related illnesses, and encourages appropriate intervention. The Montreal Cognitive Assessment (MoCA) is a rapid and sensitive screening tool for cognitive impairment that can be readily employed at the clinical level, but little is known about its utility as a community screening tool. Also, little is known regarding the demographics of the population that presents for a community screen. The research aims to evaluate the demographics of the participants that attended community memory screens in the greater Phoenix metropolitan area and to evaluate the prevalence of screen positives using the MoCA. It is hypothesized that cognitive impairment will be significantly prevalent in the screened population and that age and family history of dementia will correlate with the presence of cognitive impairment. The study methods involve descriptive analysis and application of statistical tests to evaluate for significant relationships between demographic variables and MoCA scores. The population (n=346) had a mean age of 72 (SD =10.7), was primarily female (70%), primarily Caucasian (68%) and 86% had greater than a high school education. A 58% prevalence of cognitive impairment was found in the population as defined by the MoCA. Increased age, male gender, and non‐Caucasian race correlated with lower MoCA scores. Lower education correlated with lower MoCA scores despite the inherent educational correction in the MoCA. Diabetes and a family history of AD were not significant factors. Although the number of true positives following methodical diagnosis is unknown, given the validity of the MoCA in discerning cognitive impairment, the screen was likely worthwhile and supports more routine use of community memory screens. Variables identified that were associated with increased cognitive impairment better describe the population at risk and can be utilized to focus future screening efforts. Table of Contents: Introduction…………………............................................…………...…………………………………………………...1 Background and Significance…………………………………………...………….………………………………...1 Aims……………………………………………………………………..………...………….………………………………...4 Expected Results………………………………..…………………………...………….………………………………...5 Innovation and Impact...………………...………………………………….……….………………………………...5 Research Materials and Methods………………………………………………………………………………………………7 Design………………………………………………………………………………..…………………………………..……..7 Method of Analysis………………….……………………………………………………………………………..……..8 Results…….…………………...…………...…………………………………………………………………………………………...12 Discussion.…………………...…………............................................…………………………………………………...27 Future Directions…………...…………...………………………………………….……………………………………………...36 Conclusion.…………………...…………...………………………………………………………..………………………………...37 References.…………………...…………...…………………………………………………………………………..……………...40 Appendix A: Regression analysis of individual sites…………………………………………………………..……..44 List of Tables: Table 1 Power to detect between‐group difference for any continuous variable ..................... 9 Table 2 Power to detect association between cognitive impairment status (MoCA <26) and another binary variable ........................................................................................ 10 Table 3 Descriptive statistics for continuous study variables ................................................... 13 Table 4 Descriptive statistics for categorical study variables ................................................... 14 Table 5 Univariate analysis for the association between age and MoCA score by combined data set and by individual testing sites ....................................................................... 16 Table 6 Univariate analysis for the association between education and MoCA score by combined data set and by individual testing sites ..................................................... 17 Table 7 Univariate analysis for the association between gender and MoCA score by combined data set and by individual testing sites ....................................................................... 18 Table 8 Univariate analysis for the association between past medical history of diabetes mellitus and MoCA score by combined data set and by individual testing sites ....... 20 Table 9 Univariate analysis for the association between family history of Alzheimer’s and MoCA score by combined data set and by individual testing sites ............................ 21 Table 10 Univariate analysis for the association between race and MoCA score by combined data set and by individual testing sites ....................................................................... 22 Table 11 Linear regression analysis of raw MoCA score by combined data set ........................ 25 Table 12 Logistic regression analysis of impairment status by combined data set ................... 26 Table 13 Linear regression analysis of MoCA score by the Chinese Senior Center data set ................................................................................................................................ 44 Table 14 Linear regression analysis of MoCA score by the Adam Diaz Senior Center data set ................................................................................................................................ 44 Table 15 Logistic regression analysis of MoCA score by the Adam Diaz Senior Center data set ................................................................................................................................ 44 Table 16 Linear regression analysis of MoCA score by the Goelet Beuf Community Center data set ....................................................................................................................... 44 Table 17 Logistic regression analysis of MoCA score by the Goelet Beuf Community Center data set ....................................................................................................................... 45 Table 18 Linear regression analysis of MoCA score by the Paradise Valley Community Center data set ....................................................................................................................... 45 Table 19 Logistic regression analysis of MoCA score by the Paradise Valley Community Center data set ....................................................................................................................... 45 Table 20 Linear regression analysis of MoCA score by the Pecos Community Center data set ................................................................................................................................ 45 Table 21 Linear regression analysis of MoCA score by the Shadow Mountain Community Center data set ............................................................................................................ 45 Table 22 Linear regression analysis of MoCA score by the Sunnyslope Community Center data set ....................................................................................................................... 46 Introduction: Background and Significance: Cognitive Impairment: Mild Cognitive Impairment (MCI) is often the transitional state between the cognitive changes associated with normal aging and early dementia. In most clinic‐based studies of MCI, there was conversion to Alzheimer’s disease (AD) in 40% to 80% of patients at five years, with an annual conversion rate of 10% to 15%.1,2,3 The criteria for amnestic MCI, according to Petersen 2004 are: memory impairment described by the patient, relatives, or both; cognitive impairment objectified by a neuropsychological test battery; no impairment of activities of daily living; and absence of dementia as defined by the DSM‐IV criteria.4,5 Dementia is a general term that describes a group of symptoms – such as loss of memory, judgment, language, complex motor skills and other intellectual function – caused by the permanent damage or death of the brain’s neurons over a prolonged period. One or more of several diseases, including AD, can cause dementia. AD is the most common cause. According to the National Institute on Aging, recent estimates of the prevalence of AD in the United States differ, ranging from 2.4 million to 4.5 million, depending on how the disease is measured.6 Age is the greatest known risk factor for dementia, as the incidence about doubles approximately every five years in individuals between the ages 65 and 95.7 A 2005 U.S. Census Bureau report on aging notes that the population age 65 and older in 2030 is expected to be twice as large as in 2000, growing from 35 million to 72 million and representing nearly 20 percent of the U.S. population.8 Type two diabetes mellitus is another risk factor for dementia that is supported by numerous studies9 and requires increasing concern as its prevalence increases with the growing population. Cognitive Impairment Screening: Barriers to recognition of cognitive impairment include that the patients’ families often fail to appreciate the signs of early cognitive symptoms10 and that patients and their loved ones may minimize the symptoms, or are concerned about stigma. Additionally, many people believe that 1
memory loss and cognitive decline are a normal part of aging,11,12 particularly those in certain minority groups such as African Americans and Hispanics.13 Given these barriers, it may fall to the primary care physician to recognize cognitive impairment. However, MCI and early dementia are not routinely screened for in the clinical setting. The aforementioned factors may partially account for the multitude of missed dementia cases, which are estimated to be greater than 25%.14 People with MCI or early stages of dementia are even more likely to be unrecognized than people with moderate to severe dementia.15 Therefore, there has been a growing interest in cognitive screening assessments by independent organizations such as the Alzheimer’s Foundation of America (AFA), the Alzheimer’s Association, and the Banner Alzheimer’s Institute (BAI). It is reasonable to screen for MCI given the increased risk of progression to dementia or AD in those with MCI when compared with similarly aged individuals in the general population. Screening is not a diagnosis, but can lead to the referral of appropriate individuals for further evaluation. Currently, there is no standardized methodology for performing these screens and no national consensus on dementia screening.10 When screening is performed in clinical practice, typical screening tools employed such as the Mini‐Mental State Examination (MMSE) or Mini‐Cog are not sensitive to the milder forms of cognitive impairment.3,16,17,18 While in‐depth clinician evaluation is more sensitive, it can be time consuming. The Montreal Cognitive Assessment (MoCA) is a newer screening test that is more sensitive than the MMSE in detecting early stages of memory loss or cognitive dysfunction,3,16,17 but it is not as widely used. The United States Preventative Services Task Force (USPSTF) systematic review to support its recommendation on screening for cognitive impairment in community‐dwelling adults found that in three fair quality studies (n=1,235) with high prevalence of MCI (>40%), the sensitivity of the MMSE in detecting MCI (excluding people with dementia) ranged from 45 to 60% and the specificity ranged from 65 to 90% using a cut‐off score of 27 or 28. The review found that in one good quality study (n=99) with 20% prevalence of MCI and in one fair quality study (n=120) with 24% prevalence of MCI, the sensitivity of the MoCA in detecting MCI (excluding people with dementia) was 80% and 100% and the specificity was 76% and 50%, respectively.19 In the good quality study described, the MoCA detected more participants with cognitive impairment than did the MMSE, with the MoCA detecting true 2
cognitive impairment in 30–40% of participants achieving adequate MMSE scores at two different cutoffs.16 This finding mirrored the trend of multiple other studies.17,20,21 The MoCA assesses different cognitive domains: attention and concentration, executive functions, memory, language, visuoconstructional skills, conceptual thinking, calculations, and orientation. It takes less than 10 minutes to administer the MoCA. The total possible score is 30 points, where a score of 26 and above is considered normal and less than 26 is indicative of cognitive impairment. Although the MoCA question archetypes were designed such that scores would not be influenced by educational differences among test subjects, one point is added for an individual who has 12 years or fewer of formal education, with a possible maximum of 30 points.5 Value of Community Memory Screens: Community memory screening with the MoCA could be used to help identify those with cognitive impairment and direct them to seek further attention with their physician. Although treatment options for cognitive impairment are limited to short‐term effectiveness at this time19, detection offers other benefits. Diagnosis of MCI and early dementia provides greater opportunities for necessary social, legal, and financial planning and can reduce the incidence of accidental injury due to unrecognized dementia. Diagnosis of MCI could reduce psychosocial distress associated with a sudden diagnosis of dementia. It has been shown that greater knowledge about AD symptoms is associated with increased intentions to seek help from professional sources.22 Accordingly, community screening for cognitive impairment and increased patient understanding of its implications may promote earlier help‐seeking behaviors. Given the established genetic component of AD and heritability implications, screening and subsequent diagnosis provides knowledge that is important to the family of the affected individual in this respect as well. In addition to the associated benefits of detecting cognitive impairment using community memory screening, the demographic data gleaned can be analyzed to characterize the particular population that attends these screens. The data can also be used to describe populations with MCI and dementia. The body of literature concerning MCI is somewhat limited 3
given the relative newness of this concept and there is relatively little in the literature regarding memory screening in the community as opposed clinic populations. Also, although associations between various demographic variables and cognitive impairment are described in the literature, this analysis in broader community populations such as that of this study is more limited. Such analysis could be utilized to predict those in the community who may be cognitively impaired or at risk given their demographic variables. Accordingly, the results could be used to focus screening efforts towards a particular population that has a higher incidence or may have a higher risk of cognitive impairment. Aims: 1) To evaluate the demographics of the population that attended community memory screens in the greater Phoenix metropolitan area. The independent variables of age, education, gender, family history of AD, history of diabetes mellitus, and race were examined relative to MoCA scores. These variables were selected because they either are key demographic descriptors of the population that presents for a community memory screen or they are variables that have been explored as risk factors for dementia in the literature. Through this analysis, potential confounding variables in the screened population can also be identified that may necessitate score adjustment. Demographic analysis is important to better understand the population that attends a community memory screen. This information can also be utilized to improve recruitment. The population that was less represented in attendance can be identified in order to extend recruitment efforts. Analysis can also help describe the most at risk population for cognitive impairment by characterizing the screen positives in the study. This would help elucidate a target population to focus advertising efforts for future community memory screens, optimizing the pretest probability. 4
2) To evaluate the prevalence and degree of screen positives and negatives in the screened population using the Montreal Cognitive Assessment. The data was analyzed to determine the prevalence of participants who screened positive for cognitive impairment as defined by the MoCA in the tested population and were referred for further evaluation. This prevalence implicates the utility of community memory screening given the high sensitivity and specificity of the MoCA for detecting cognitive impairment established in the literature. Data regarding patient follow‐up, including diagnostic assessment for cognitive impairment, was not available for analysis in this study. Expected Results: It is hypothesized that family history will positively correlate with cognitive impairment given the partially genetic nature of AD, and recognized correlation of dementia and family history. It is hypothesized that age will correlate with cognitive impairment as dementia is predominantly a disease of aging. Cognitive impairment is expected to vary with a medical history of diabetes as studies have correlated type two diabetes with higher dementia risk.9 It is hypothesized that education will correlate somewhat with cognitive impairment as this finding has been demonstrated with the MoCA in the literature.23 It is hypothesized that the prevalence of cognitive impairment will be higher in the screened population than in the general population. It might be assumed that some participants had memory concerns or a family history of memory problems prompting them to attend the screen. Thus, screen participants may have a higher prevalence of risk factors for cognitive impairment than that of the general Phoenix population. Innovation and Impact: Memory screens targeting a broader portion of the community than those in clinic populations are not well documented in the literature based on PubMed searches. Particularly, published data from screening with the MoCA in the general community is relatively sparse due to its less frequent use than the MMSE. Furthermore, there is no literature on such screening in the Phoenix area based on PubMed searches. The incidence of screen positives in the study could 5
help elucidate the utility of screening in the community setting versus reserving screening for use in primary care physicians’ practices. This utility would be better described if a gold standard was applied to the data to determine true screen positives; however, the high sensitivity of the MoCA and relatively high specificity as described in the literature19,20 enables some inferred commentary on efficacy. The research also assesses demographic variables, including the association of race and gender in relation to cognitive impairment as assessed by the MoCA. Statistically significant findings may indicate that cultural and sex differences affect screening outcomes. This may help focus screening efforts towards a higher risk population and may also provide commentary on the effect of human experiences on memory or suggest possible inherent biases in the MoCA. Thus the project addresses a unique area of research with likely impact. 6
Research Materials and Methods: Design: The Banner Alzheimer’s Institute, in partnership with the City of Phoenix, Arizona State University College of Nursing & Healthcare Innovation and the BHHS Legacy Foundation conducted free memory screens on November 15, 2011 at ten locations across Phoenix. The screens were conducted on National Memory Screening Day and involved administration of the MoCA version 7.1, tips for brain health and information about where patients can find help for memory concerns. The screen was advertised through the Arizona Republic, Banner’s internal newsletter, a local TV station, and flyers which were posted at the senior and community centers where the screens were later conducted. Patients called the locations directly to sign up. The screens were conducted by volunteer nurse practitioner students or registered nurses who had received formal training from a single trainer to administer the test using standardized verbal instructions (from mocatest.org). The test administrators were reviewed for consistency (observed by the trainer while role‐playing administering the test for adherence to the standardized instructions) prior to proctoring the tests. The same version of the MoCA was available in English, Chinese and Spanish and multilingual test administrators administered the test to participants whose primary language was Spanish or Mandarin in order to limit skew due to language and potentially cultural barriers. However, it is important to note that the Spanish and Chinese versions have not been fully validated in the literature as equivalent to the English version of the MoCA. For instance, several studies have found good sensitivity, but fair specificity of the Chinese24,25 and Spanish26,27 versions of the MoCA. The following variables were collected for each patient: test location, gender, family history of AD, history of DM, age, years of education, race, and MoCA score. Analysis was conducted using a combined data set from ten screening site locations in Phoenix as well as by the individual location samples. Patients were questioned on DM history because of the large Hispanic and Native American population in Phoenix and the high prevalence of this disease in the population. No further patient information was collected and thus no patient identifiers were recorded. No information collected was on the list of HIPAA identifiers. After completion and 7
submission of the F309 Human Research Determination form, the Office for Human Research Protections (OHRP) confirmed that the work is not human research. Because the research is not human research, no participant consent for the research was necessary, the research is HIPAA compliant, and no specific security of the data was necessary. Method of Analysis: The sample includes 346 subjects: 200 with MoCA scores < 26 (defined as cognitive impairment by the MoCA) and 146 with MoCA scores ≥ 26. The sample size provides satisfactory power to detect any relevant difference in demographic characteristics between these two groups. For a continuous variable such as age (in years) and education (in years), Table 1 gives the power of the nonparametric Wilcoxon rank‐sum test when the true difference in the group means ranges from 30% to 60% of the within‐group standard deviation (SD). A difference in this range would be reasonable to expect. For example, the overall SD of age is 10.7 years, and the within‐group SD is expected to be below that level. Hence 50% of the SD is about 5 years. To test for association between the cognitive impairment status and another binary variable (e.g. gender), the Fisher’s exact test was used, as follows: if P1 is the proportion of a certain level of the binary characteristic in the cognitively impaired subjects, and P2 is the corresponding proportion in the cognitively unimpaired subjects, when the true difference between P1 and P2 is 0.15, the study has good power (75% to 90%) to detect this (see Table 2). When the true difference is 0.2, the power is very high (95% or above). Data was organized in a data schema to facilitate analysis. Basic quality assessment of the data was conducted such as verifying complete data sets for each subject and assessing the reasonableness of ranges. Stata® was employed to execute statistical tests. Descriptive analysis of the data including calculation of the mean, median, standard deviation, minimum, and maximum was performed. 8
Table 1. Power to detect between‐group difference for any continuous variable. Between‐group difference Power 30% SD 76.4% 40% SD 94.7% 50% SD 99.4% 60% SD > 99.9% 9
Table 2. Power to detect association between cognitive impairment status (MoCA <26) and another binary variable. N1 MoCA<26 200 200 200 200 200 200 200 P1 0.25 0.35 0.45 0.55 0.65 0.75 0.85 N2 MoCA≥26 146 146 146 146 146 146 146 P2 P1‐P2 Power 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.15 0.15 0.15 0.15 0.15 0.15 0.15 93.9% 84.3% 78.1% 75.7% 76.8% 81.3% 89.3% 10
Non‐parametric tests of significance were employed as the MoCA scores and other continuous variables were not normally distributed. Based on the distribution, no MoCA scores were classified as outliers and thus no data was removed from the sample. Univariate analysis was performed for each variable. The Fisher’s exact test was used to assess contingency tables where a categorical variable (e.g. gender) was compared to the proportion of cognitive impairment that is present for each gender (MoCA score <26) and the proportion in which it is not present (MoCA score ≥26). The Wilcoxon rank‐sum test was used to evaluate a binary categorical variable compared to MoCA score, a continuous variable. For non‐binary categorical variables (e.g. location and race), the Krushkal‐Wallis test was used to evaluate the association with MoCA score. For significant Krushkal‐Wallis results, Tukey‐Kramer pairwise comparisons were evaluated. The Wilcoxon rank‐sum test was used to evaluate a continuous variable (e.g. age) compared to whether cognitive impairment was present (positive) or not present (negative) as assessed by the MoCA. Spearman’s rank correlation was used to measure the association between two continuous variables such as age and MoCA score. Stepwise linear regression was performed to evaluate the effect of all variables on MoCA score. Non‐binary categorical variables that were found to be significant by this method were then evaluated with the likelihood ratio test to ascertain the overall significance of the variable (i.e. the significance of the race variable versus Caucasian or African American, etc.). Stepwise logistic regression was performed to evaluate the significant odds ratios of all variables in regards to cognitive impairment (MoCA score ≤26). The log likelihood ratio test was then used to evaluate significant non‐binary categorical variables. Stepwise linear and logistic regression 11
Results: 58% of the sample population is considered cognitively impaired by the MoCA. Participants in this group had a greater mean age of approximately 3 years in comparison to those who had MoCA scores greater than or equal to 26. This group also had fewer mean years of education by approximately 1 year (See Table 3). The female participants outnumbered the male participants 241 to 105. 62% of the males and 56% of the females were considered cognitively impaired by the MoCA. 88 participants had a positive family history of Alzheimer’s disease (FH of AD), with 53% of these participants having MoCA scores less than 26 compared to 59% of those with no FH of AD. 109 participants had a past medical history positive for diabetes mellitus (PMH of DM), with 65% of these participants having MoCA scores less than 26 compared to 54% of those with no PMH of DM. Caucasian, African American, Hispanic, and Asian participants had 53%, 83%, 80%, and 58% respective prevalences of cognitive impairment according to the MoCA. The prevalence of cognitive impairment also varied by the location of the test site (See Table 4). 12
Table 3. Descriptive statistics for continuous study variables. MoCA MoCA<26 MoCA≥26 Age MoCA<26 (years) MoCA≥26 Education MoCA<26 (years) MoCA≥26 n 346 200 146 346 200 146 346 200 146 mean 23.728
21.060
27.384
72.208
73.450
70.507
13.624
13.073
14.380
median 25
22
27
73
74
71.5
14
13
14
sd 4.357
3.864
1.188
10.669
10.833
10.233
3.470
3.365
3.479
min 6 6 26 25 25 32 0 1 0 max p‐value 30 25 30 97 <0.0001
97 0.012
90 22 <0.0001
22 0.000
21 13
Table 4. Descriptive statistics for categorical study variables. n male 105
Gender female 241
positive 88
FH of AD negative 258
109
PMH of positive DM negative 237
Caucasian 236
African American 24
Race Hispanic 36
Asian 43
Other 7
Chinese SC 41
Deer Valley CC 30
Adam Diaz SC 35
Goelet Beuf CC 38
Devonshire CC 35
Location Paradise Valley CC 38
Pecos CC 39
Shadow Mountain CC 33
South Mountain CC 33
Sunnyslope CC 24
Community center (CC), senior center (SC) *Other includes Native American and Indian mean median
23.781
25
23.705
25
24.682
25
23.403
25
23.606
24
23.785
25
24.483
25
20.708
21
20.472
22
23.884
25
24.429
26
24.341
25
23.700
25
22.829
24
22.289
25
24.686
25
24.421
25
24.051
25
23.909
25
21.030
21
26.750
28
sd 3.895
4.552
3.780
4.498
3.856
4.576
3.656
5.353
5.358
4.640
4.117
4.357
4.036
4.668
5.695
3.385
3.621
3.043
3.357
5.423
3.040
min 9 6 13 6 9 6 11 9 6 11 16 11 13 9 9 16 15 16 15 6 17 p‐
value 0.637
max n, MoCA<26 n, MoCA≥26
30
65
40
30
135
106
30
47
41 0.028
30
153
105
30
71
38 0.216
30
129
108
30
124
112 0.0001
29
20
4
28
29
7
30
25
18
27
2
5
30
22
19 0.363
28
16
14
29
23
12
30
23
15
30
20
15
29
22
16
29
24
15
29
21
12
29
25
8
30
4
20
14 The results of a two‐sample Wilcoxon rank‐sum (Mann‐Whitney) test suggest that there is a statistically significant difference between the underlying distributions of the age of individuals with MoCA scores less than 26 and the age of individuals with MoCA scores greater than or equal to 26 (z = 2.508, p = 0.012). The positive z score indicates that MoCA scores less than 26 are more prevalent as age increases. The mean ages for participants with MoCA scores less than 26 and for participants with MoCA scores greater than or equal to 26 are 73.5 and 70.5 years, respectively. The results of a Spearman’s Rho Rank Order Correlation suggest that the relationship between age and MoCA score is statistically significant (rho = ‐0.238, p < 0.0001). The negative Spearman’s rho indicates that MoCA scores decrease as age increases. All of the significant findings from the individual locations support this trend (See Table 5). The results of a two‐sample Wilcoxon rank‐sum (Mann‐Whitney) test suggest that there is a statistically significant difference between the underlying distributions of the education of individuals with MoCA scores less than 26 and the education of individuals with MoCA scores greater than or equal to 26 (z = ‐3.783, p = 0.0002). The negative z score indicates that MoCA scores less than 26 are less prevalent as education increases. The mean education for participants with MoCA scores less than 26 and for participants with MoCA scores greater than or equal to 26 are 13.1 and 14.4 years, respectively. The results of a Spearman’s Rho Rank Order Correlation suggest that the relationship between education and MoCA score is statistically significant (rho = 0.325, p < 0.0001). The positive Spearman’s rho indicates that MoCA scores increase as education increases. All of the significant findings by the Spearman’s Rho Rank Order Correlation from the individual locations support this trend (See Table 6). The results of a Fisher’s exact test suggest that there is not a statistically significant relationship between gender and MoCA scores less than 26 (p = 0.344). The results of a Wilcoxon rank‐sum test suggest that there is not a statistically significant difference between the underlying distributions of the MoCA scores of females and the MoCA scores of males (z = 0.472, p = 0.6371). These tests also yielded no significant results in analysis of 9 out of 10 of the individual sites (See Table 7). 15 Table 5. Univariate analysis for the association between age and MoCA score by combined data set and by individual testing sites. p‐value*, association between age (years) and cognitive impairment (MoCA <26) 0.012
0.339
0.307
0.109
0.127
0.828
0.287
0.427
0.195
0.785
0.755
Combined Chinese SC Deer Valley CC Adam Diaz SC Goelet Beuf CC Devonshire CC Paradise Valley CC Pecos CC Shadow Mountain CC South Mountain CC Sunnyslope CC *Wilcoxon Rank‐Sum test †Spearman’s Rho Rank Order Correla on rho†, association between age (years) and MoCA score ‐0.238
‐0.270
‐0.229
‐0.385
‐0.422
‐0.013
‐0.246
‐0.234
‐0.286
‐0.305
‐0.204
p‐value† <0.0001
0.088
0.224
0.022
0.008
0.942
0.136
0.151
0.107
0.084
0.338
16 Table 6. Univariate analysis for the association between education and MoCA score by combined data set and by individual testing sites. p‐value*, association between education rho†, association between education (years) and cognitive impairment (MoCA <26) (years) and MoCA score 0.000
0.325
0.674
0.025
0.159
0.336
0.065
0.448
0.023
0.457
0.611
0.243
0.427
‐0.022
0.266
0.503
0.027
0.497
0.580
0.299
0.009
0.205
Combined Chinese SC Deer Valley CC Adam Diaz SC Goelet Beuf CC Devonshire CC Paradise Valley CC Pecos CC Shadow Mountain CC South Mountain CC Sunnyslope CC *Wilcoxon Rank‐Sum test †Spearman’s Rho Rank Order Correla on p‐value† <0.0001
0.877
0.069
0.007
0.004
0.159
0.897
0.001
0.003
0.091
0.338
17 Table 7. Univariate analysis for the association between gender and MoCA score by combined data set and by individual testing sites. Combined Chinese SC Deer Valley CC Adam Diaz SC Goelet Beuf CC Devonshire CC Paradise Valley CC Pecos CC Shadow Mountain CC South Mountain CC Sunnyslope CC *Fisher’s Exact test †Wilcoxon Rank‐Sum test p‐value*, association between gender and cognitive impairment (MoCA <26) 0.344 0.746 1.000 1.000 0.728 0.700 0.016 1.000 1.000 0.241 0.615 p‐value†, association between gender and MoCA score 0.637
0.978
0.638
0.748
0.236
0.747
0.030
0.431
0.714
0.128
0.179
18 The results of a Fisher’s exact test suggest that there is not a statistically significant relationship between a positive PMH of DM and MoCA scores less than 26 (p = 0.079). The results of a Wilcoxon rank‐sum test suggest that there is not a statistically significant difference between the underlying distributions of the MoCA scores of participants without a PMH of DM and the MoCA scores of participants with a PMH of DM (z = 1.239, p = 0.216). These tests also did not yield significant results in any of the individual screening site samples (See Table 8). The results of a Fisher’s exact test suggest that there is not a statistically significant relationship between a positive FH of AD and MoCA scores less than 26 (p = 0.382). The results of a Wilcoxon rank‐sum test suggest that there is a statistically significant difference between the underlying distributions of the MoCA scores of participants without a FH of AD and the MoCA scores of participants with a FH of AD (z = ‐2.198, p = 0.028). The negative z score indicates that MoCA scores are lower for patients with a FH of AD. Significant findings from the individual locations support this trend (See Table 9). The results of a Fisher’s exact test suggest that there is a statistically significant relationship between race and MoCA scores less than 26 (p < 0.001). The results of a Kruskal‐Wallis test indicate that there is a statistically significant difference among the medians of the represented race distributions of MoCA scores (H = 30.378, 4 degrees of freedom, p=0.0001). Post‐Hoc Analysis with the Tukey‐Kramer method indicates that there is a statistically significant difference between the MoCA scores of certain races in the following pairwise comparisons: Caucasian compared to African American, Caucasian compared to Hispanic, African American compared to Asian, and Hispanic compared to Asian. Race was found to be significant in univariate analysis in only one of the individual screening site samples (See Table 10). The results of a Kruskal‐Wallis test indicate that there is a statistically significant difference among the medians of the testing site specific distributions of MoCA scores (H = 32.730, 9 degrees of freedom, p=0.0001). Post‐Hoc Analysis with the Tukey‐Kramer method indicates that there is a statistically significant difference between the MoCA scores of certain testing sites. 19 Table 8. Univariate analysis for the association between past medical history of diabetes mellitus (PMH of DM) and MoCA score by combined data set and by individual testing sites. Combined Chinese SC Deer Valley CC Adam Diaz SC Goelet Beuf CC Devonshire CC Paradise Valley CC Pecos CC Shadow Mountain CC South Mountain CC Sunnyslope CC *Fisher’s Exact test †Wilcoxon Rank‐Sum test p‐value*, association between PMH of DM and cognitive impairment (MoCA <26) 0.079 0.685 1.000 1.000 0.061 1.000 0.504 0.477 0.145 0.699 0.300 p‐value†, association between PMH of DM and MoCA score 0.216
0.945
0.912
0.956
0.133
0.762
0.776
0.459
0.203
0.651
0.767
20 Table 9. Univariate analysis for the association between family history of Alzheimer’s disease (FH of AD) and MoCA score by combined data set and by individual testing sites. Combined Chinese SC Deer Valley CC Adam Diaz SC Goelet Beuf CC Devonshire CC Paradise Valley CC Pecos CC Shadow Mountain CC South Mountain CC Sunnyslope CC *Fisher’s Exact test †Wilcoxon Rank‐Sum test p‐value*, association between FH of AD and p‐value†, association between FH of cognitive impairment (MoCA <26) AD and MoCA score 0.382 0.028
0.321 0.031
0.466 0.423
0.059 0.192
1.000 0.448
0.419 0.984
0.187 0.103
0.734 0.830
1.000 0.743
1.000 0.098
0.578 0.802
21 Table 10. Univariate analysis for the association between race and MoCA score by combined data set and by individual testing sites. p‐value*, association between race and cognitive impairment (MoCA <26) Combined <0.001
Chinese SC 0.444
Deer Valley CC 0.129
Adam Diaz SC 0.710
Goelet Beuf CC 0.031
Devonshire CC 1.000
Paradise Valley CC§ N/A
Pecos CC 0.641
Shadow Mountain CC 0.523
South Mountain CC 0.190
Sunnyslope CC 0.635
White (W), African American (AA), Hispanic (H), Asian (A) §Only Caucasian participants were in this sample. *Fisher’s Exact test †Krushkal‐Wallis test ‡Tukey‐Kramer
H†, association between race and MoCA 30.378 3.216 2.970 2.889 14.134 1.659 N/A 3.421 1.211 2.360 1.635 d.f. p‐value 4
3
3
2
4
2
N/A
4
1
2
4
0.0001
0.359
0.396
0.236
0.007
0.436
N/A
0.490
0.271
0.307
0.803
Significant Pairwise Comparisons‡ W vs. AA, W vs. H, AA vs. A, H vs. A
N/A
N/A
N/A
W vs. AA
N/A
N/A
N/A
N/A
N/A
N/A
22 Variables that have a statistically significant effect on MoCA scores by linear regression include: age (effect = ‐0.097, p<0.001), gender (male is the reference, effect= ‐0.952, p=0.039), education (effect=0.291, p<0.001), race (overall p<0.0001) and testing site (location, overall p=0.018). The results indicate that MoCA scores decrease as age increases, where an increase in 10 years of age yields approximately a one point drop in MoCA score for the sample population. The results indicate that MoCA scores are approximately one point less for males compared to females in the sample. Education has the effect of increasing MoCA scores by approximately one third of a point for each year increase in a participant’s education. Race is significant overall, and the effect on MoCA scores of the individual races of African American, Hispanic and Asian can be compared in reference to the Caucasian race. The results indicate that for this sample, MoCA scores are approximately 3 points less for participants who are African American, Hispanic and Asian when compared to Caucasian. Results may be somewhat confounded by interaction between age and gender, with the likelihood ratio test yielding a significant result comparing the linear model and the model with an interaction term (p=0.039). There was no interaction between education and gender by the likelihood ratio test (p=0.2018). The variables with significant effects as discerned by linear regression performed on individual screening site samples varied from site to site (See Appendix A). Variables that have statistically significant odds ratios in regards to MoCA scores less than 26 include: age (OR=1.026, 95% CI 1.003 to 1.050), gender (male is the reference, OR=1.782, 95% CI 1.068 to 2.975), education (OR=0.908, 95% CI 0.841 to 0.980), and race (overall p<0.001). For each year increase in age there is a 2.6% increase in the odds of having cognitive impairment. For a 10 year increase in age, the odds of having cognitive impairment is 29% greater. In the sample, men had a 78% increase in the odds of having cognitive impairment (men were cognitively impaired 1.8 times more often than women). For each year increase in education there is a 9.2% decrease in the odds of having cognitive impairment. 23 The variables with significant odds ratios as discerned by logistic regression performed on individual screening site samples varied from site to site (See Appendix A). 24 Table 11. Linear regression analysis of raw MoCA score by combined data set. Variable Effect age ‐0.097
gender ‐0.952
education 0.291
race Caucasian 0
African American ‐3.403
Hispanic ‐3.252
Asian ‐3.080
Other ‐0.675
location Chinese SC 0
Deer Valley CC ‐3.727
Adam Diaz SC ‐4.351
Goelet Beuf CC ‐4.300
Devonshire CC ‐3.128
Paradise Valley CC ‐3.540
Pecos CC ‐3.104
Shadow Mountain CC ‐3.275
South Mountain CC ‐3.777
Sunnyslope CC ‐1.084
*Overall p‐value for significance of race †Overall p‐value for significance of location 95% CI ‐0.136
‐0.058 ‐0.046 ‐1.857
0.420 0.162
referent ‐5.150
‐1.656 ‐4.802
‐1.703 ‐5.835
‐0.325 ‐3.580
2.231 referent ‐6.817
‐0.637 ‐7.410
‐1.292 ‐7.202
‐1.398 ‐6.140
‐0.116 ‐6.591
‐0.489 ‐6.066
‐0.142 ‐6.350
‐0.201 ‐6.921
‐0.634 ‐4.154
1.985 p‐value <0.001
0.039
<0.001
<0.0001*
<0.001
<0.001
0.029
0.648
0.018†
0.018
0.005
0.004
0.042
0.023
0.040
0.037
0.019
0.488
25 Table 12. Logistic regression analysis of impairment status by combined data set. Variable Odds Ratio age 1.026
gender 1.782
education 0.908
race Caucasian 0
African American 5.364
Hispanic 3.523
Asian 1.086
Other 0.339
*Overall p‐value for significance of race
95% CI 1.003
1.050 2.975 1.068
0.980 0.841
referent 1.711
16.819 1.402
8.850 0.543
2.173 0.060
1.896 p‐value 0.027
0.027
0.013
0.001*
0.004
0.007
0.816
0.218
26 Discussion: The high prevalence of cognitive impairment as defined by the MoCA in the screened population is of interest. The study found that 58% of the sample population was cognitively impaired by the MoCA. This result seems quite high, but there is little in the literature on a similar screening population from which to compare the results. The prevalence of cognitive impairment as assessed by the MoCA was similarly high at 62% in a large, ethnically diverse population‐based study (n= 2,653; mean age 50.3 years, range 18‐85; Caucasian 34%, African American 52%, Hispanic 11%, other 2%).23 The high percentage of African American participants could potentially be skewing those results, as it was found in the Phoenix study that African American race in comparison to the Caucasian control was statistically associated with lower MoCA scores. The prevalence of true cognitive impairment in the Phoenix metropolitan area is not known, however, it is likely not as high as that in the screened population as the screening was done on a volunteer basis rather than as a random sample. For example, some participants may have had a specific reason for attending the screen, such as subjective memory loss or a family history of dementia, and thus might have been at a higher risk for cognitive impairment compared to the general population. A recent systematic review found that the prevalence of mild cognitive impairment in the general population ranged from 3% to 42%, where the mean ages for the different studies ranged from 65 to 80.5 years, with the majority of mean ages in the mid 70s.28 This population is comparable in age to the Phoenix community memory screen sample (mean age of study participants was 72.2, with SD=10.7). The review found variation in the prevalence depending on the operational definition of cognitive impairment and the authors noted that the wide differences in prevalence estimates pose a significant challenge to understanding the social burden of MCI. This variation also hinders interpretation of the Phoenix study results, but it can still be stated that the prevalence of MCI by the MoCA in the screened population is seemingly much higher than that of the general population. The magnitude of the difference and the consequential implication for the utility of community memory screens is even more considerable when the comparison is made with conservative prevalence estimates of MCI in the general community. 27 Although the data is not ascertainable, determining the true screen positives in the Phoenix study would be of interest. The sensitivity and specificity of the MoCA in a similar population are not well described. A study of an English‐speaking, community‐dwelling population in Florida (n=118) with a fraction of the participants recruited from the community via advertising efforts, found that the MoCA had a 97% sensitivity and 35% specificity.29 This specificity is much lower than that described by Nasreddine et al20 or the USPSTF review. It is noteworthy that the sample size in the Florida study is relatively small and a fraction of the sample was recruited from a memory clinic as opposed to a true community memory screen where the sample is composed entirely of individuals responding spontaneously to community advertising efforts. These conflicting findings suggest that the MoCA needs to be validated in community screening populations to best assess its utility in this setting. It is possible that cognitive impairment as currently defined by the MoCA does not apply as accurately to community screen populations as the samples that were used to construct the cut off values for determining cognitive impairment. More insight could be achieved from follow up on the screen positives to determine the percent that were true positives after diagnostic testing. There is a lack of literature on community memory screens with well delineated follow‐up that substantiate their efficacy. A Virginia study of community‐based persons aged 44 to 91 concerned about or interested in screening their memory (n=999), found that 44.3% of the total sample received follow‐up recommendations after a neuropsychologist interviewed and a neuropsychometrist assessed each participant using subtests of the Wechsler Memory Scale III.30 Of the 106 participants who followed up with their primary care physician, 50.9% reported one or more conditions or diagnoses were identified that were thought to be contributing to the difficulties noted during their screenings. The Virginia study remarks somewhat on the utility of community cognitive screening, however, more definitive data is needed. Although the Phoenix study does not provide measured efficacy of community memory screens, it does provide worthwhile commentary. Given the high prevalence of cognitive impairment in the Phoenix study sample and the majority finding that the MoCA has a high sensitivity and specificity in detecting MCI, a high level of cognitive impairment likely exists in the screen positives. Thus the screen likely provided substantial utility. 28 It was expected that age would correlate directly with the presence of cognitive impairment as dementia is primarily a disease of aging. Evaluation of a well‐characterized cohort showed that age‐associated decline in health status, assessed by a frailty index made up only of attributes that are not known as cognitive risk factors, predicts the incidence of AD and dementia.31 Apart from age being an independent risk factor, many age‐related health problems, such as heart disease, hypertension, stroke, and DM, are recognized as AD risk factors. It is even possible that the effect of age is not fully represented as there may be a survivor effect where older persons might have had more age‐associated comorbid conditions and been more likely to die or unable to present to a voluntary memory screen secondary to such comorbid conditions. The results of statistical analysis align with the literature in regards to the general population. Participants with MoCA scores less than 26 had a greater mean age of approximately three years in comparison to those who had MoCA scores greater than or equal to 26 (73.5 and 70.5 years, respectively). Linear regression indicates that an increase in ten years of age yields approximately a one point drop in MoCA score. Logistic regression suggests that for each year increase in age there is a 2.6% increase in the odds of having cognitive impairment and for a ten year increase in age, the odds of having cognitive impairment is 29% greater. Thus to improve screening efficacy, the screens should target older adults. A definitive age to recommend commencement of screening is beyond the scope of this study. Despite the inherent educational correction in the MoCA, it was expected that education would correlate with cognitive impairment given that this relationship has been described with the MoCA in the literature.23 Participants with MoCA scores less than 26 had fewer mean years of education by approximately one year in comparison to those who had MoCA scores greater than or equal to 26 (13.1 and 14.4 years, respectively). Linear regression indicates that MoCA scores increase by approximately one third of a point for each year increase in a participant’s education. Logistic regression suggests that for each year increase in education there is a 9.2% decrease in the odds of having cognitive impairment. Given the results, it is possible that a low level of education is not fully accounted for by the MoCA. Johns E.K. et al. recently suggested that to better adjust the MoCA for lower educated subjects, two points should be added to the total score for subjects with 4‐9 years of education, and one point for 10‐12 years of 29 education.32 In the Phoenix sample, 39.6% of the participants had 12 or fewer years of education. In addition, literacy was not assessed yet could be an important factor as a Manhattan study (n=1002) found that literacy level was a better predictor of decline in memory, executive function, and language skills than was years of education.33 The education variable may also be confounded by test taking ability. There might be both a disparity in years of education and test taking ability between the lower and higher ends of the education spectrum. For example, those at the lower end of the spectrum may lack the test taking skills that participants at the higher end of the spectrum possess through greater experience. Test taking anxiety and unfamiliarity with the style of test may also play a role in determining the outcomes of the MoCA. Still, it is also possible that the magnitude of association between education and MoCA score is partly accounted for by a true relationship between cognitive decline and years of education. Several studies of normal aging have reported more rapid cognitive and functional decline among persons with lower educational attainment.34,35,36 It is of note that 14.5% of the Phoenix participants self‐reported less than 12 years of formal education, which underrepresents the true proportion of the Phoenix population with less than a high school education. According to the United States Census Bureau based on data from 2008‐2009, 19.9% of adults aged 25 and greater in Phoenix had less than a high school education.37 Thus, recruitment efforts could be adjusted to better reach this population. For example, literacy and language barriers could be addressed. Although less substantiated than the relationship between some of the other tested variables and MCI, gender differences were expected based on the literature. In a study of 757 non‐
demented, community‐dwelling elderly individuals from an English‐speaking background categorized as younger (70‐79 years) or older (80‐90 years), the prevalence of MCI as determined by the Petersen criteria was lowest in younger women (32.3%) and similar across men and older women (41.9%‐43.6%).38 The participants in the Phoenix sample are comparable in age to the younger group in the aforementioned study, and the Phoenix study results parallel the findings that the females had a lower prevalence of MCI in this age group and that gender had a significant effect on outcomes. In the Phoenix sample, 62% of the males and 56% of the 30 females were considered cognitively impaired by the MoCA and linear regression indicates that MoCA scores are approximately one point less for males compared to females. Females may have fewer risk factors that were not accounted for in this study when compared to men. For example, the population‐based Sydney Memory and Aging Study found statistically significant interactions that reflected the differences between sex groups in risk factor profiles for MCI including sociodemographic, lifestyle, and cardiac, physical, mental, and general health factors.38 Linear regression results may be somewhat confounded by a significant interaction between age and gender. Females had a mean age of approximately one year less than men in the sample (71.9 compared to 72.9 years). It is possible that being female makes one inherently more likely to go to a memory screen at a younger age. Women may be more likely to pursue health maintenance and screening earlier on in the development of cognitive decline or before cognitive decline is present. It is also interesting that the female participants outnumbered the male participants 241 to 105. In a survey of those who participate in memory screenings (n=2,562 or 13% of those screened) conducted by MetLife Mature Market Institute, the Alzheimer’s Foundation of America and the Center for Productive Aging at Towson University, a much larger percentage of women (74%) than men (29%) expressed concerns about their memory of the 73% of survey participants that reported they had memory concerns.39 Also, surveyed men were twice as likely as women to report that they were encouraged by family or a friend to attend the screening. This may indicate that women are more concerned and proactive about their health and thus more likely to present to community memory screens. The much higher percentage of women in the Phoenix sample indicates that recruitment efforts could be improved to increase the number of males presenting for a community memory screen to better represent the general community. Increasing male participants could also improve screening efficacy as logistic regression suggests that men had a 78% increase in the odds of having cognitive impairment compared to women. The study’s finding of increased prevalence of MCI in men compared to women as assessed by the MoCA could be advertised to encourage men to present for evaluation. 31 Cognitive impairment was expected to vary with a medical history of DM. Studies have consistently shown a relation of type two diabetes with higher dementia risk, where the association is stronger for vascular dementia compared to AD.9 109 of the Phoenix participants had a positive PMH of DM, with 65% of these participants having MoCA scores less than 26 compared to 54% of those with no PMH of DM. The study did not find a statistically significant relationship between a positive PMH of DM and MoCA score. Exploration of other factors such as if the patients’ DM was well controlled could account for the discrepancy. Those with a history of consistently well controlled DM may have skewed results for instance. Participants who answered that they had a positive PMH of DM were not asked to specify if this was type 1 or 2, whether it was controlled or uncontrolled, or the duration of their history. Thus, the degree to which a PMH of DM affected cognition in this study may not be as significant as that described in the literature. It is mentionable that the results were in line with the findings of a recent systematic review of longitudinal population‐based studies. The review found that more recent studies did not find a significant relationship between diabetes and dementia and that these studies only observed a significant relationship in subgroups of patients, for example in patients with undiagnosed diabetes or in those that did not have an apolipoprotein E4 allele.40 This illustrates that there may be other factors that influence the effect of diabetes on cognitive function. It was expected that a FH of AD would positively correlate with cognitive impairment given the partially genetic nature of AD, and recognized correlation of dementia and family history. 88 participants had a positive FH of AD, with 53% of these participants having MoCA scores less than 26 compared to 59% of those with no FH of AD. The results indicate that there is not a statistically significant relationship between a FH of AD and MoCA scores. This result can be explained as the participants with a FH of AD were not compared to those without a FH of AD in the general population, but rather were compared to individuals without a FH of AD in a unique population of volunteer participants that may have some increased risk for cognitive impairment as discussed previously. Furthermore, patients who answered that they had a positive FH of AD were not asked to specify if this was in a first degree relative. Last, until fairly recently, dementia often was not recognized for what it was and was considered a natural part 32 of aging.10 Accordingly, participants may have been unaware of a positive FH of AD when one did in fact exist. The effect of race on MoCA score in this study population was interesting. Caucasian, African American, Hispanic, and Asian participants had a 53%, 83%, 80%, and 58% prevalence of cognitive impairment according to the MoCA, respectively. The results indicate that there is a statistically significant relationship between race and MoCA scores and linear regression found that MoCA scores are approximately three points less for participants who are African American, Hispanic and Asian when compared to Caucasian. This may indicate that the MoCA or its protocol may not be appropriately translated or free of cultural influences. The two non‐
English versions of the MoCA employed in the Phoenix study have not received the same validation as the English version.24,25,26,27 Different languages and cultures employ possible differences in thought processes and world views requiring a culture fair (conceptual equivalency) communicational approach.41 Although race was found to significantly affect MoCA scores and the effect varied by individual races, factors such as socioeconomic status, quality of education, and acculturation would have to be explored to better elucidate these results. For example, a relationship was found between Hispanic acculturation (as measured by the Acculturation Scale for Mexican Americans) and performance on selected tests of the Halstead‐Reitan Battery among college students42 and several studies show that African American acculturation (as measured by the African American Acculturation Scale) is related to cognitive test performance, even after accounting for age, years of education, and sex.43,44,45,46 When quality of education is included as a covariate, the predictive power of acculturation is weakened.47 Early education is now thought to play a large role in many of the differences in cognitive function that were previously attributed to race.48 To account for race differences in MoCA scores, racial norms may need to be established to interpret test results and deeper study of the test itself may be required to unearth and correct any areas of misunderstanding from a cultural perspective. It is also possible that a single tool may not be equivalent across different cultures. It is important to account for the aforementioned factors not only when analyzing the information relative to ethnic minority groups but also when making culturally focused interventions. 33 Furthermore, the sample populations for races other than Caucasian were small in comparison (Caucasian sample size was 236, while the sample sizes for the other three races ranged from 24 to 43 participants). Thus, recruiting could be improved to better reach races other than Caucasian to be more representative of the race distribution in the general Phoenix community. A study that conducted 19 focus groups (n=177) of ethnically diverse participants 50 years and older (six groups of African American, four of Chinese, three of Vietnamese, four of non‐
Hispanic White, and two of American Indian participants) found that many participants did not recall reading or hearing about brain health in the media.49 This was particularly true among Vietnamese groups, and some African American participants. A perceived barrier to seeking brain health information by participants, particularly among African Americans, American Indians, and Caucasians, was confusion caused by conflicting and changing media messages about both general and brain health. This implicates a lack of knowledge and confusion concerning the need for memory screening that may be more prevalent among certain ethnic groups. Participants recommended a multimedia approach to inform others about brain health. Importantly, Chinese and Vietnamese participants noted that media in their native languages were key information sources. All groups suggested targeting preexisting social groups. For instance, African Americans and American Indians suggested presenting brain health information at church. Based on the study, to best reach potential participants, information needs to be presented clearly and in the native language though multiple media outlets including television, radio and print media. Relevant social groups and community organizations such as churches, alumni associations and senior centers should be involved for better communication. The ten screening locations had sample sizes ranging from 24 to 41 participants. The prevalence of cognitive impairment as assessed by the MoCA varied by the location of the screening site and linear regression found that location has a significant effect overall on MoCA scores. This heterogeneity may explain why the results of statistical tests by location do not completely mirror those of the combined data set. However, the relationships between the tested variables and MoCA scores in the individual screening site samples generally align with those in the combined data set. Thus, no one location is skewing the combined results and the sample 34 is relatively homogenous as a whole. This enables more robust conclusions to be drawn from the combined analysis. 35 Future Directions: The prevalence of cognitive impairment as measured by the MoCA was extremely high in the sample population. Given the lack of literature on community memory screens in a similar study population, it is difficult to know if these results are typical. Thus, further exploration of the prevalence of cognitive impairment discerned by community memory screens is warranted. Follow up on the screen positives to determine the percent that were true positives after diagnostic testing would be valuable. This further research would better comment on the utility of community memory screens. Research addressing whether the MoCA is an equivalent memory test in non‐Caucasian races is of interest. Without such validation, it is difficult to comment on its screening efficacy in any ethnically diverse sample. Further research is required to elucidate any confounding effect of education on MoCA scores and discern if new cut‐off values or test format changes are required for MoCA scores to correlate better with true cognitive impairment at lower levels of education. Addressing these factors is important to ensure adequate sensitivity and specificity when employing the MoCA in the community screen setting, were the population is expected 36 Conclusion: The study results of demographic analysis can be utilized to better characterize the population that attends a community memory screen as well as characterize groups with cognitive impairment as measured by the MoCA in similar populations. The results found that there was a high prevalence of cognitive impairment as defined by the MoCA in the population that attends a metropolitan community memory screen with a mean age of 72. The results also found that increased age correlated with increased prevalence of cognitive impairment. Thus, screening at around age 70 in this type of community appears beneficial. Given that the occurrence was so high, screening at an earlier age would likely still uncover significant MCI and may be justified. Cognitive impairment was more abundant in males as compared to females and this relationship was found to be significant via regression analysis. This finding substantiates the less established MCI risk factor of sex. Cognitive impairment was not found to vary with a PMH of DM. As DM is a multifaceted variable, subpopulations within this category may need to be investigated to determine its true effect. Although univariate analysis found that MoCA scores were lower for patients with a FH of AD, a FH of AD was not found to be significant in regression analysis. This may reflect the uniqueness of a community memory screen population and suggest the value of screening even in the absence of traditional risk factors. Education correlated with cognitive impairment despite the inherent educational correction in the MoCA. Further research is needed to ascertain if this confounding factor can be sufficiently rectified with a different correction such as that proposed by Johns E.K. et al. Cognitive impairment was more prevalent in certain races and significant relationships were observed. It is unclear if cognitive impairment was more prevalent in certain races independent of other variables or if confounding factors such as cultural ones necessitate MoCA score adjustment or adjustment in the test itself. Following the screen described by this study, the Banner Alzheimer’s Institute has conducted annual memory screens in the Phoenix metropolitan area using the MoCA. The study results of demographic analysis can be used to focus and improve recruitment in these screens and may 37 be applicable to other similar communities. The considerable majority of patients in the screen were female. Thus more advertisement towards men can be employed to improve their awareness and accessibility to the screen. Also, the diversity in the Phoenix area was underrepresented as there were few participants from certain races such as Native American. The Phoenix population with less than a high school education was underrepresented as well. Culturally sensitive and appropriate advertising efforts could be employed and the addition of different testing site locations could be considered to better reach these groups. Although the number of true positives following methodical diagnosis is not known, given the high prevalence of cognitive impairment in the study sample and the high sensitivity and specificity of the MoCA, it may be proposed that the community screen was a worthwhile effort in discerning cognitive impairment in the community. Given the implications of discerning such considerable cognitive impairment in the Phoenix sample, further study of community memory screen populations is necessary to better establish the results and utility of community memory screening so that formal recommendations can be put forth. Nonetheless, the Phoenix community memory screen helped educate and create awareness of memory issues and brain health in a considerable number of the population. The need for this benefit was demonstrated in the MetLife survey of memory screen participants, where 84% had seen their doctor in the previous 6 months yet only 24% of those who were concerned about their memory had discussed these concerns with their doctor, and 30% had not discussed their concerns with anyone. Given the relatively large sample size of 346 participants, large deviations from the literature for many of the variables evaluated were not anticipated and in general were not seen. Variations from the expected results may be the result of confounding variables such as socioeconomic status and geographic region in addition to the factors previously mentioned. As elaborated, the demographic analysis better characterizes the population that attends a community memory screen. This is also useful in order to identify less represented groups from the general population and to identify populations at increased risk of cognitive impairment. 38 This enables both improved recruitment and more targeted advertising efforts. The higher prevalence of cognitive impairment as evaluated by the MoCA in the Phoenix sample relative to the general community is also of great interest. The population that attends a community memory screen may be uniquely different than the populations studied in the literature. 39 References: 1. Petersen RC, Negash S. Mild cognitive impairment: An overview. CNS Spectr. 2008;13(1):45‐
53. 2. Schmidtke K, Hermeneit S. High rate of conversion to alzheimer's disease in a cohort of amnestic MCI patients. IntPsychogeriatr. 2008;20(1):96‐108. 3. Chertkow H, Massoud F, Nasreddine Z, et al. Diagnosis and treatment of dementia: 3. mild cognitive impairment and cognitive impairment without dementia. CMAJ. 2008;178(10):1273‐1285. 4. Petersen RC. Mild cognitive impairment as a diagnostic entity. J Intern Med. 2004;256(3):183‐194. 5. Nasreddine Z. The montreal cognitive assessment. www.mocatest.org. Updated 2012. Accessed 03/10/2012. 6. Rodgers AB. Alzheimer’s disease: unraveling the mystery. National Institute on Aging. National Institutes of Health.U.S. Department of Health and Human Services. NIH Publication Number: 08‐3782. Sept 2008. 7. Evans DA, Funkenstein HH, Albert MS, et al. Prevalence of Alzheimer’s disease in a community population of older persons. Higher than previously reported. JAMA. 1989;262(18):2551‐6. 8. Wan H, Sengupta M, Velkoff VA, DeBarros KA. 65+ in the United States. 2005 U.S. Census Bureau, Current Population Reports. P23‐209. U.S. Government Printing Office. Washington, DC, 2005. 9. Luchsinger JA. Type 2 diabetes, related conditions, in relation and dementia: An opportunity for prevention? J Alzheimers Dis. 2010;20(3):723‐736. 10. Powers R, Ashford W, Peschin S. Memory matters: Treatment of Alzheimer’s disease and related dementias and recommendations for national policy. Alzheimer’s Foundation of America. www.alzfdn.org. Published December 2008. Accessed 03/10, 2012. 11. Solomon PR, Murphy CA. Should we screen for Alzheimer’s disease? A review of the evidence for and against screening for Alzheimer’s disease in primary care practice. Geriatrics. 2005, 60(Nov): 26‐31. 12. Knopman D, Donohue JA, Gutterman EM. Patterns of care in the early stages of Alzheimer’s disease: impediments to timely diagnosis. J Am Geriatr Soc. 2000 Mar;48(3):300‐4. 13. ICAN2: Investigating Caregivers’ Attitudes and Needs. Harris Interactive for the Alzheimer’s Foundation of America. February 9, 2007, www.alzfdn.org. 14. Freund B. Office‐based evaluation of the older driver. J Am Geriatr Soc. 2006 Dec;54(12):1943‐4. 15. Borson S, Scanlan JM, Watanabe J, Tu Sp, Lessig M. Improving identification of cognitive impairment in primary care. Int. J Geriatr Psychiatry. 2006 Apr;21(4):349‐55. 40 16. 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 state examination (MMSE) scores. J Clin Exp Neuropsychol. 2012;34(7):750‐757. 17. Lonie JA, Tierney KM, Ebmeier KP. Screening for mild cognitive impairment: A systematic review. Int J Geriatr Psychiatry. 2009;24(9):902‐915. 18. Kaufer DI, Williams CS, Braaten AJ, Gill K, Zimmerman S, Sloane PD. Cognitive screening for dementia and mild cognitive impairment in assisted living: Comparison of 3 tests. J Am Med Dir Assoc. 2008;9(8):586‐593. 19. Lin JS, O'Connor E, Rossom RC, et al. Screening for Cognitive Impairment in Older Adults: An Evidence Update for the U.S. Preventive Services Task Force [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2013 Nov. (Evidence Syntheses, No. 107.) 20. Nasreddine ZS, Phillips NA, Bedirian V, et al. The montreal cognitive assessment, MoCA: A brief screening tool for mild cognitive impairment. J Am Geriatr Soc. 2005;53(4):695‐699. 21. Pendlebury ST, Cuthbertson FC, Welch SJ, Mehta Z, Rothwell PM. Underestimation of cognitive impairment by mini‐mental state examination versus the montreal cognitive assessment in patients with transient ischemic attack and stroke: A population‐based study. Stroke. 2010;41(6):1290‐1293. 22. Werner P. Knowledge about symptoms of alzheimer's disease: Correlates and relationship to help‐seeking behavior. Int J Geriatr Psychiatry. 2003;18(11):1029‐1036. 23. Yu J, Li J, Huang X. The beijing version of the montreal cognitive assessment as a brief screening tool for mild cognitive impairment: A community‐based study. BMC Psychiatry. 2012;12:156‐244X‐12‐156. 24. Hu JB, Zhou WH, Hu SH, et al. Cross‐cultural difference and validation of the chinese version of montreal cognitive assessment in older adults residing in eastern china: Preliminary findings. Arch Gerontol Geriatr. 2013;56(1):38‐43. 25. Munoz‐Neira C, Henriquez Chaparro F, Delgado C, Brown J, Slachevsky A. Test your memory‐spanish version (TYM‐S): A validation study of a self‐administered cognitive screening test. Int J Geriatr Psychiatry. 2014. 26. Gomez F, Zunzunegui M, Lord C, Alvarado B, Garcia A. Applicability of the MoCA‐S test in populations with little education in colombia. Int J Geriatr Psychiatry. 2013;28(8):813‐820. 27. Rossetti HC, Lacritz LH, Cullum CM, Weiner MF. Normative data for the montreal cognitive assessment (MoCA) in a population‐based sample. Neurology. 2011;77(13):1272‐1275. 28. Ward A, Arrighi HM, Michels S, Cedarbaum JM. Mild cognitive impairment: Disparity of incidence and prevalence estimates. Alzheimers Dement. 2012;8(1):14‐21. 29. Luis CA, Keegan AP, Mullan M. Cross validation of the montreal cognitive assessment in community dwelling older adults residing in the southeastern US. Int J Geriatr Psychiatry. 2009;24(2):197‐201. 41 30. Crews WD,Jr, Harrison DW, Keiser AM, Kunze CM. The memory screening outreach program: Findings from a large community‐based sample of middle‐aged and older adults. J Am Geriatr Soc. 2009;57(9):1697‐1703. 31. Song X, Mitnitski A, Rockwood K. Nontraditional risk factors combine to predict alzheimer disease and dementia. Neurology. 2011;77(3):227‐234. 32. Johns EK, et al. Level of education and performance on the Montreal Cognitive Assessment: New recommendations for education corrections. Presented at the Cognitive AgingConference 2010, Atlanta, Georgia, April 15‐18th, 2010. Retrieved from: http://www.mocatest.org/ 33. Manly JJ, Schupf N, Tang MX, Stern Y. Cognitive decline and literacy among ethnically diverse elders. J Geriatr Psychiatry Neurol. 2005;18(4):213‐217. 34. Christensen H, Korten AE, Jorm AF, et al. Education and decline in cognitive performance: Compensatory but not protective. Int J Geriatr Psychiatry. 1997;12(3):323‐330. 35. Albert MS, Jones K, Savage CR, et al. Predictors of cognitive change in older persons: MacArthur studies of successful aging. Psychol Aging. 1995;10(4):578‐589. 36. Chodosh J, Reuben DB, Albert MS, Seeman TE. Predicting cognitive impairment in high‐
functioning community‐dwelling older persons: MacArthur studies of successful aging. J Am Geriatr Soc. 2002;50(6):1051‐1060. 37. U.S. Census Bureau. State and County QuickFacts. www.census.gov. Updated Mar 2014. Accessed 04/07/2014. 38. Sachdev PS, Lipnicki DM, Crawford J, et al. Risk profiles for mild cognitive impairment vary by age and sex: The sydney memory and ageing study. Am J Geriatr Psychiatry. 2012;20(10):854‐865. 39. MetLife Foundation Alzheimer’s Survey: What America Thinks. MetLife Foundation. www.metlife.org. 2006. Accessed 5/3/2014. 40. Biessels GJ. Caffeine, diabetes, cognition, and dementia. J Alzheimers Dis. 2010;20 Suppl 1:S143‐50. 41. Valle R. Outreach to ethnic minorities with alzheimer's disease: The challenge to the community. Health Matrix. 1988;6(4):13‐27. 42. Arnold BR, Montgomery GT, Castaneda I, Longoria R. Acculturation and performance of Hispanics on selected Halstead‐Reitan neuropsychological tests. Assessment. 1994;1(3):239‐
248. 43. Kennepohl S, Shore D, Nabors N, Hanks R. African american acculturation and neuropsychological test performance following traumatic brain injury. J Int Neuropsychol Soc. 2004;10(4):566‐577. 44. Lucas JA. Acculturation and neuropsychological test performance in elderly African Americans. Journal of the International Neuropsychological Society. 1998; 4(77). 42 45. Manly JJ, et al. African American acculturation and neuropsychological test performance among nondemented community elders. Journal of the International Neuropsychological Society. 1998; 4(7). 46. Manly JJ, Miller SW, Heaton RK, et al. The effect of african‐american acculturation on neuropsychological test performance in normal and HIV‐positive individuals. the HIV neurobehavioral research center (HNRC) group. J Int Neuropsychol Soc. 1998;4(3):291‐302. 47. Manly JJ, Byrd DA, Touradji P, Stern Y. Acculturation, reading level, and neuropsychological test performance among african american elders. Appl Neuropsychol. 2004;11(1):37‐46. 48. Manly J. Race, culture, education, and cognitive test performance among older adults. In: Scott SM, Alwin DF, ed. Handbook of Cognitive Aging: Interdisciplinary Perspectives. Thousand Oaks, CA: Sage Publications; 2008:398–417. 49. Friedman DB, Laditka JN, Hunter R, et al. Getting the message out about cognitive health: A cross‐cultural comparison of older adults' media awareness and communication needs on how to maintain a healthy brain. Gerontologist. 2009;49 Suppl 1:S50‐60. 43 Appendix A: Regression analysis of individual sites Table 13. Linear regression analysis of MoCA score by the Chinese Senior Center data set. Variable age FH of AD Effect ‐0.153
4.354
95% CI ‐0.280
0.126
‐0.026 8.582 p‐value 0.020
0.044
Logistic regression of the Chinese Senior Center data set yielded no significant variables. Linear and logistic regression of the Deer Valley Community Center data set yielded no significant variables. Table 14. Linear regression analysis of MoCA score by the Adam Diaz Senior Center data set. Variable age education Effect ‐0.136
0.756
95% CI ‐0.246
0.419
‐0.027 1.093 p‐value 0.017
<0.001
Table 15. Logistic regression analysis of MoCA score by the Adam Diaz Senior Center data set. Variable FH of AD Odds Ratio 0.176
95% CI 0.033
0.937 p‐value 0.042
Table 16. Linear regression analysis of MoCA score by the Goelet Beuf Community Center data set. Variable Effect age ‐0.169
race Caucasian 0
African American ‐6.934
Hispanic ‐7.839
Asian ‐6.124
Other ‐2.216
*Overall p‐value for significance of race 95% CI ‐0.319
‐0.019 referent ‐11.239
‐2.630 ‐14.471
‐1.206 ‐12.901
0.653 ‐8.876
4.443 p‐value 0.028
0.003*
0.002
0.022
0.075
0.503
44 Table 17. Logistic regression analysis of MoCA score by the Goelet Beuf Community Center data set. Variable PMH of DM education Odds Ratio 39.075
0.585
95% CI 1.247
1223.928 0.958 0.357
p‐value 0.037
0.033
Linear and logistic regression of the Devonshire Community Center data set yielded no significant variables. Table 18. Linear regression analysis of MoCA score by the Paradise Valley Community Center data set. Variable gender (male) Effect ‐2.476
95% CI ‐4.836
‐0.116 p‐value 0.040
Table 19. Logistic regression analysis of MoCA score by the Paradise Valley Community Center data set. Variable gender (male) Odds Ratio 8.400
95% CI 1.530
46.108 p‐value 0.014
Table 20. Linear regression analysis of MoCA score by the Pecos Community Center data set. Variable education Effect 0.530
95% CI 0.275
0.785 p‐value <0.001
Logistic regression of the Pecos Community Center data set yielded no significant variables. Table 21. Linear regression analysis of MoCA score by the Shadow Mountain Community Center data set. Variable age education Effect ‐0.152
0.499
95% CI ‐0.294
0.171
‐0.011 0.828 p‐value 0.036
0.004
45 Logistic regression of the Shadow Mountain Community Center data set yielded no significant variables. Linear and logistic regression of the South Mountain Community Center data set yielded no significant variables. Table 22. Linear regression analysis of MoCA score by the Sunnyslope Community Center data set. Variable education Effect 0.388
95% CI 0.125
0.651 p‐value 0.006
Logistic regression of the Shadow Mountain Community Center data set did not converge. 46