Suspected undiagnosed ADRD among Middle Eastern and North African Americans

Background ADRD underdiagnosis among minority populations is well-established and known to be more prevalent among women. Yet, it remains unclear if these patterns exist among Middle Eastern and North African (MENA) adults. We estimated ADRD underdiagnosis among MENA and other US- and foreign-born non-Hispanic Whites and compared sex-stratified results. Methods We linked 2000–2017 National Health Interview Survey and 2001 –2018 Medical Expenditure Panel Survey data (ages > = 65 years, n = 23,981). Undiagnosed ADRD was suspected if participants reported cognitive limitations without corresponding ADRD diagnosis. Results Undiagnosed ADRD was highest among MENA adults (15.8%) compared to non-Hispanic Whites (US-born = 8.1%; foreign-born = 11.8%). MENA women had 2.52 times greater odds (95% CI = 1.31 –4.84) of undiagnosed ADRD compared to US-born White women after adjusting for risk factors. Discussion This study contributes the first national estimates of undiagnosed ADRD among MENA adults. Continued research is needed to facilitate policy changes that more comprehensively address health disparities and related resource allocation.


Introduction
The prevalence of Alzheimer's disease and related dementias (ADRD) is increasing worldwide, with 6.5 million adults living with ADRD in the United States (US) (1). This gure is likely an underestimate due to lack of clinical diagnosis. Previous research suggests only half of individuals with ADRD have been diagnosed (2). Early diagnosis and awareness can be important for mitigating safety risks, preventing and treating complications, improving quality of life, and planning for future care (2).
Several studies have established disparities in ADRD incidence, prevalence, and underdiagnosis among minority populations (1)(2)(3)(4). The National Institutes of Health recognizes American Indian/Alaskan Native, Asian, Black/African American, and Native Hawaiian or Other Paci c Islanders for race and Hispanic/Latinx for ethnicity as the minimum required categories for reporting health estimates based on O ce of Management and Budget (OMB) guidelines (5). Studies have demonstrated that non-Hispanic Black and Hispanic individuals are more likely to have ADRD and be undiagnosed than non-Hispanic Whites (hereafter referred as Whites) (2)(3)(4). Fewer studies have examined ADRD disparities among Middle Eastern and North African (MENA) adults, who are included in the White classi cation by OMB federal guidelines despite a growing body of research suggesting MENA health differs signi cantly from Whites (6).
The MENA population includes Arab and non-Arab individuals who were born in or trace their ancestry to countries located in the Middle East or North Africa. The Arab population in the US grew by 72% to over 3.7 million between 2000 and 2010 (7). Data from the American Community Survey estimates that 8% of older adults (ages 65 + years) are MENA, which is greater than several other represented minority populations, including Paci c Islanders (5.8%) and Hispanics (5.5%) (8, 9). Despite representing an increasing proportion of the aging population, MENA adults are often underrepresented in studies on cognitive aging because they are di cult to identify in national datasets. Few studies, therefore, have examined the national prevalence of ADRD among MENA adults, regardless of diagnosis. Two studies found that MENA adults were more likely to report cognitive limitations, a potential indicator of ADRD, when compared to US-born Whites (10,11). Other research demonstrated foreign-born MENA adults had higher estimates of cognitive limitations compared to US-born MENA adults (12). The US prevalence of undiagnosed ADRD among MENA adults remains unknown.
The lack of ethnic identi er to recognize MENA populations contributes to an additional layer of underrepresentation overlooking the health of MENA women. ADRD is known to be more prevalent among women in general, yet it remains unclear if this pattern exists among MENA women. Studies on the physical health of MENA adults in the US have highlighted how MENA men report better health outcomes than MENA women regardless of immigrant status (13). In a study evaluating hypertension and self-reported health among foreign-born MENA, foreign-born Mexican, and US-born Whites, Read & Reynolds found that 19.63% of MENA women reported having hypertension compared to 13.03% of MENA men (13). Despite reporting higher use of healthcare (74% MENA women sought care in last 12 months compared to 58.38% MENA men) and a usual place for care, 17.45% of MENA women selfreported fair/poor health, compared to 9.07% of MENA men, and higher than both foreign-born Mexican and US-born White men and women. How sex shapes the observed differences of ADRD at the intersection of nativity status remains unknown among MENA adults.
To ll this gap in the literature, our aims were to: 1) calculate and compare estimates of suspected undiagnosed ADRD among foreign-born MENA adults compared to US-and foreign-born Whites and 2) compare estimates strati ed by sex.

Data Sources
Page 4/20 The sample was created by linking 2000-2017 NHIS person level and 2001-2018 MEPS household and medical condition les. Details of both surveys are provided elsewhere (14,15). Brie y, the NHIS collects individual health information from a sample of households every year. A subsample of participants is selected to participate in the MEPS the following year. The MEPS collects additional information on current medical conditions and health care visits in the last 12 months. Medical conditions provided to interviewers by participants are coded based on the International Classi cation of Disease (ICD) diagnostic guidance by professional coders and included in speci c medical condition les. From 2001-2015, ICD-9 codes were created. From 2016-2018, ICD-10 codes were created. To ensure con dentiality, only the rst three digits of diagnostic codes were reported in public-use les. Household and medical condition MEPS les were merged with NHIS person les. More information has been reported previously (10).

Independent Variable
The independent variable was created using the region of birth variable from NHIS. The NHIS includes two questions of country of birth. Participants are asked whether they were born in the US or a US territory overseas. Foreign-born participants were asked their country of birth. To protect con dentially, the NHIS provides a region of birth variable which combines responses into 10 world regions, including the US, Europe, Russia/former USSR countries, Middle East, and Africa, among others. Two foreign-born White groups were created. Responses from White adults born in Europe and Russia/former USSR countries were combined (hereafter, referred to as Europe). Responses from White adults born in the Middle East and Africa were combined based on previous research (10). By including foreign-born White Africans, our study aligns with 2015 US Census content testing for race and ethnicity (6, 19) currently proposed for the 2030 Census (20). The nal independent variable included US-born Whites, foreign-born Whites from Europe and foreign-born MENA adults.

Dependent Variable
The dependent variable (suspected undiagnosed ADRD) was created using the cognitive limitation variable from MEPS. Undiagnosed ADRD was suspected if a participant reported having a cognitive limitation but did not have a corresponding ADRD diagnosis. The MEPS classi es participants as having any cognitive limitation (0 = no; 1 = yes) if they self-report (or reported by proxy) having problems making decisions, requires supervisions for their own safety, or experiences confusion or memory loss.

Statistical Analysis
Weighted column percentages and standard errors were used to describe sample characteristics of USborn Whites, foreign-born Whites from Europe, and foreign-born MENA adults. Chi square tests were used to examine differences for all covariates. For two-group comparisons, post-hoc chi square analyses were conducted. Among all adults, age-and sex-adjusted prevalence estimates of suspected undiagnosed ADRD were calculated for all groups. Age-adjusted and sex-strati ed results were also calculated for all groups.
Logistic regression procedures were used to determine associations between each region of birth among Whites (US, Europe, MENA) before and after controlling for risk factors. US-born Whites were treated as the reference group based on their historical classi cation as the majority group and previous research on MENA health (22). Each covariate was included in the regression models during the model building process. Model 2 included age and sex as demographic risk factors. Model 3 included education as a potentially modi able risk for dementia during early life. In model 4, other potentially modi able risk factors were evaluated (hearing loss, hypertension, obesity, smoking, depressive symptoms, social isolation, physical inactivity and diabetes). Statistically signi cant risk factors (P < .05) were included in multivariable models. Ultimately, model 2 adjusted for age and sex, model 3 further adjusted for education, and model 4 further adjusted for depressive symptoms. Usual source of care and health insurance were also evaluated as potential mediators; however, they were not included in the nal models (p > .05). Data were analyzed using STATA 17.2 (SVYSET procedures). We conducted a sensitivity analysis limiting the sample to 2000-2014 NHIS and 2001-2015 MEPS data to account for changes in ICD-9 and ICD-10 classi cation codes.
The study was approved by the Agency for Healthcare Research and Quality. Data were analyzed at a local Federal Statistical Research Data Center. The study was deemed not subject to review or approval by the institutional review board because it used de-identi ed secondary data that does not meet the federal de nition for human subjects' research.

Descriptive Results
Descriptive characteristics are presented in Table 1. Among all older adults, when all groups were compared, statistically signi cant differences were found by age, education, hearing loss, depressive symptoms, and health insurance coverage (P < .05). When foreign-born MENA adults were compared to foreign-born Whites from Europe, they were less likely to be female (48.5% vs. 58.6%) and younger (mean age 73.6 years vs. 75.2 years). Foreign-born MENA adults were more likely to be depressed (37.3%) compared to US-(25.7%) and foreign-born Whites from Europe (33.5%). More foreign-born MENA adults (2.2%) did not have any health insurance coverage compared to other groups (US-born Whites = 0.2%; foreign-born Whites from Europe = 0.7%). MENA men were younger (mean age 70 years vs. 74.8 years) and less likely to have lower levels of education (3.4% vs. 10.2%) compared to foreign-born Whites (P < .05). MENA men were less likely to be married (18.5%) than both US-(27.6%) and foreign-born (27.8%) Whites. MENA women were more likely to have lower education (18.65) than US-born White women (5.1%). MENA women were also more likely to report depressive symptoms (40.8%) and no usual source of care (14.5%) than US-born White women (26.8% and 6.4%, respectively). Fewer MENA women had private health insurance coverage (34.9%) compared to 57.9% of US-born White women. Among foreignborn women, 79.7% of MENA women lived in the US for 15 years or longer compared to 91.5% of White women born in Europe (P < .05).  *p < .05, chi square test for comparisons between US-born non-Hispanic Whites, foreign-born non-Hispanic Whites from Europe/Russia, and foreign-born MENA Americans. † Hearing loss (yes or no) was determined by self-report of any hearing di culty, including some or serious di culty. ‡ Obesity was determined by self-reported body mass index (BMI) of ≥ 30 kg/m 2 (yes or no).

Prevalence Estimates
Prevalence estimates of suspected undiagnosed ADRD are presented in Table 2. Suspected undiagnosed ADRD prevalence estimates were highest among foreign-born MENA adults (15.8%) compared to US-(8.1%) and foreign-born (11.8%) Whites from Europe. When strati ed by sex, the age-adjusted prevalence of suspected undiagnosed ADRD was highest among foreign-born MENA women (22.5%) and lowest among US-born White (7.6%) men.

Sensitivity Analysis
Results from our sensitivity analysis are presented in Supplementary Table 1. Findings were similar to our primary results when we limited our analysis to individuals with ICD-9 classi cation codes (MEPS 2001(MEPS -2015.

Discussion
This study evaluated the prevalence of suspected undiagnosed ADRD among foreign-born MENA adults compared to US-and foreign-born Whites from Europe and compared the odds of suspected undiagnosed ADRD before and after adjusting for covariates.
Prevalence estimates were higher among foreign-born MENA adults compared to US-born Whites. MENA adults had higher odds of having suspected undiagnosed ADRD; however, sex-strati ed results were only signi cant among women. An overview of sex-speci c differences and the implications that education and depressive symptoms have as potentially modi able risk factors for ADRD among MENA adults are discussed below.
The magnitude of the higher odds of suspected undiagnosed ADRD among MENA older adults and MENA women was attenuated when education was taken into account. Previous studies have shown that adults with undiagnosed ADRD are more likely to have lower levels of education (2). Less education is not only a risk for ADRD, but a key factor that drives underdiagnosis. Older adults with lower levels of education are less likely to receive preventive health care (23), which limits their ability to be screened during primary care visits if signs and symptoms of cognitive impairment are recognized by the healthcare provider (24). Since this is the rst study to explore undiagnosed ADRD among MENA adults, we are unable to compare our results to other studies that accounted for education as a risk factor for ADRD underdiagnosis. However, our nding that education levels were lower among MENA adults compared to the US-born Whites in our sample is consistent with previous research. Studies using national data sources have reported low levels of education among foreign-born MENA adults ranging from 18-32.8% (11,13,(25)(26)(27)(28)(29)(30)(31). When strati ed by sex, studies showed that 22.7-26.7% of foreign-born MENA women (13,32) and 11% of foreign-born MENA men had less than high school education (13,33).
In the current study, we found that 10.5% of foreign-born MENA adults had less than a ninth-grade level of education in our bivariate analysis, which was not statistically signi cantly different than US-(5.9%) or foreign-born Whites (14.0%). Yet, we found a large disparity in education level when we strati ed by sex. Only 3.4% of MENA men had less than a ninth-grade education compared to 6.8% of US-and 10.2% of foreign-born White men (P = .0385). Among women, 18.6% of MENA women had less than ninth-grade level of education compared to only 5.1% of US-born and 16.6% of foreign-born White women (p < .0001).
Our results show a similar educational pattern as community-based studies in Michigan with large concentrations of older MENA/Arab populations, which found that the average years of formal education among MENA women was 2 years, with 68% reporting no formal education (34).
Depressive symptoms during late life (ages 65 and older) have also been identi ed as a potentially modi able risk factor for ADRD (35). Yet, it remains unclear whether depressive symptoms are an independent risk factor for ADRD or a prodromal symptom. Self-reported cognitive limitations could re ect symptoms of depression rather than a dementia syndrome. Despite this limitation, we still found higher odds of suspected undiagnosed ADRD among MENA women even after controlling for depressive symptoms. Although depression has not been explicitly linked to ADRD underdiagnosis, previous studies have shown that older adults are less likely to receive mental health services (36) and may prefer to receive treatment from their primary care provider (37). With most older adults having other comorbid chronic conditions that need to be addressed during primary care visits (38), there may be limited opportunities for early detection and screening for ADRD while addressing other health concerns (39). Existing literature has identi ed a wide range (5.5%-60%) of prevalence estimates for depressive symptoms among MENA individuals using community-based convenience samples (40)(41)(42), electronic health record data (43,44), and nationally representative samples (10,45). For example, using linked NHIS and MEPS data, Kindratt and colleagues reported that 38.2% of MENA adults ages 65 and older had depressive symptoms (10). The prevalence was slightly higher than what we found in the current study (37.3%) that was limited to adults ages 65 and older without an ADRD diagnosis. In the current study, we found that older MENA women had a higher prevalence of depressive symptoms than men (40.8% and 34.0%, respectively). This nding is consistent with other community-based studies that demonstrate depressive symptoms were higher among MENA women compared to MENA men (34,41,45).

Strengths and Limitations
A strength was the use of two nationally representative data sources to uncover MENA health outcomes while no ethnic identi er is available on a national scale. The NHIS is the only nationally representative health survey that allows for MENA individuals to be disaggregated from other foreign-born adults. Several studies have used NHIS data to uncover health outcomes among foreign-born Middle Eastern and Arab American immigrant populations (11,13,25,26,(29)(30)(31)(32)(33). By linking with MEPS, we accessed additional health information from a subsample of participants who completed the previous year's NHIS.
To our knowledge, this methodology has only been used once for measuring MENA health. This study replicates methods used by Kindratt and colleagues (2022) to link NHIS and MEPS data and broaden the sample to be inclusive of MENA individuals by also including White Africans (10). We added MEPS medical condition data les to evaluate ADRD underdiagnosis. Although our sample size included individuals with both ICD-9 and ICD-10 codes to determine ADRD diagnosis, a sensitivity analysis was conducted to ensure our ndings were not biased by the change in classi cation system. Removing individuals with and an ADRD diagnosis from the sample allowed us to gain a better understanding of individuals suspected to be undiagnosed.
Despite the strengths, various limitations may have affected our results. The rst limitation was survey language. Both NHIS and MEPS are only collected in English and Spanish; therefore, MENA adults who only speak Arabic may have been excluded from the sample. This limitation may have resulted in an underestimate of the disparity considering that monolingual Arabic speakers may be even less likely to access high-quality medical care and be diagnosed. However, because data are collected from one key representative from each household, responses from Arabic speakers may have been given by a bilingual family member. Another limitation is our inability to disaggregate US-born MENA individuals from other US-born Whites. We may have found an even larger disparity if we were able to make these comparisons without accounting for nativity status. The cross-sectional nature of the NHIS and MEPS are a limitation in that causal links could not be ascertained. Finally, since the sample was limited to adults who reported a medical condition that was directly linked to ICD-9 or ICD-10 billing code, individuals may not have shared this information during the interview. MEPS medical condition codes were designed to measure medical expenditures, not national prevalence estimates. Despite this limitation, using MEPS medical condition data allows us to expand research on MENA health while no ethnic identi er is available.

Conclusions
This is one of the rst studies to evaluate suspected undiagnosed ADRD among MENA adults. While no other studies have examined undiagnosed ADRD among MENA adults, our ndings are consistent with recent research reporting that Arab and MENA adults have a higher prevalence of cognitive limitations than other Whites (10)(11)(12). Furthermore, our ndings are similar to previous studies which have shown that Black and Hispanic adults have greater risk of ADRD underdiagnosis compared to Whites (2, 3). By including MENA Americans with other Whites, it not only masks their own health outcomes, but may lessen the true disparity between privileged and other marginalized groups (6). Our results highlight that MENA health outcomes are more similar to other racial and ethnic minority groups than Whites. Without an ethnic identi er for Arab/MENA adults, cognitive health outcomes among this group will remain unnoticed and planning for future care and improving quality of life among MENA adults with suspected undiagnosed ADRD will continue to be overlooked.