Mexican-American Ethnicity and Cognitive Function: Findings from an Elderly Southwestern Sample

H ispanics now comprise the largest ethnic minority group in the United States, and the majority of Hispanics are of Mexican descent. Forty percent of the increase in the United States’ population between 1990 and 2000 has been accounted for by its 35.3 million Hispanic residents, and the number of Hispanics in the United States aged 65 and older is expected to grow at a rate five times greater than that of non-Hispanic whites in the 21 ^st century. 1 , 2 Yet, little is known concerning patterns of life-cycle cognitive function for this ethnic group. A few reports in the dementia research literature suggest that both the prevalence and incidence of Alzheimer’s type dementia is higher in Hispanics of Caribbean origin relative to non-Hispanic whites, but little cognitive epidemiology has been published for Mexican-Americans. 3 – 5 It is difficult to select the proper instruments for cognitive screening research among the Mexican-American population. The Mini-Mental State Examination (MMSE) has not been fully standardized on Mexican-American samples. 6 – 9 In addition, the MMSE correlates strongly with educational attainment, which could confer bias in studies of Mexican-American ethnicity and cognitive function. 7 – 10 Clock-drawing tests are relatively simple screening tests that access somewhat separate cognitive functional domains from the MMSE and appear less vulnerable to sociocultural influences. 11 Clock-drawing has been shown to improve the sensitivity of the MMSE to mild cognitive impairment syndromes in non-Alzheimer’s type dementias. 12 CLOXI and CLOXII tests are specific types of clock drawing tests and have separate assessments. CLOXI is sensitive to executive control function. CLOXII was designed to test constructional praxis. 13 CLOX has been translated into Spanish and validated in multiple elderly populations, including Mexican-Americans. 14 – 16 Executive control functions are cybernetic processes that control the sequencing, initiation, and monitoring of complex goal-directed activities that are mediated by the frontal lobes and their connections to other brain areas by white matter tracts. 17 CLOXI is more sensitive to executive control function than many other clock drawing tests. 14 Previous research indicates that only a small proportion of the variance in CLOX failure rates is specifically attributable to the combined effects of education, socioeconomic status, acculturation, language, and gender. 16 We designed a screening study of cognitive functions in elderly samples of noninstitutionalized Mexican-American and non-Hispanic white populations in El Paso County, Texas. We administered the MMSE, CLOXI and CLOXII to a stratified, random sample of Mexican-American and non-Hispanic whites. We sought to determine whether patterns of cognitive function in this sample are associated with education, ethnicity, and socioeconomic status.

METHOD

Subjects

In 2000 and 2001, we interviewed a stratified random sample of 1,152 noninstitutionalized elderly (65 and older) residents of El Paso County. This sample represents 83% of respondents originally contacted for interview. The survey instrument was translated into Spanish through cross-translation. In-home interviews were conducted in Spanish or English, depending on respondents’ preference. The entire sampling procedure was conducted in order to ascertain direct and moderating effects 18 of socioeconomic status, and Mexican-American versus non-Hispanic white ethnicity on a number of health and mental health phenomena. Twenty-four of El Paso County’s 94 census tracts (based on 1990 census) were randomly selected after stratification for median income and ethnic composition. Specifically, eight census tracts were randomly selected for lower, middle, and upper-middle income categories. Each of these 24 census tracts was screened through telephone interviews to identify households containing one or more members 65 years of age or older. Screening interviews also provided demographic information including respondent’s age, gender, and ethnicity. Independent stratified random sampling procedures were then established for Hispanics and non-Hispanic whites. These procedures ensured virtually equal representation of women and men within each of the three census-tract income categories. The same is true for age cohorts 65 to 74 years old, and 74 years and older. Mean and median ages for the entire sample are 74.9 and 74.0, respectively. Fifty percent of the sample is female.

Assessment of Education and Mexican-American Ethnicity

Cognitive function correlates with the level of formal education received. 19 – 21 For the purposes of this study, education was measured by years of educational attainment, including ordinal categories for “high-school graduate,” “college graduate,” and “at least some post-college education.” A high school degree represents successful completion of 12 years of formal education after preschool, both in the United States and in Mexico. The degree of Mexican-American ethnicity was defined through respondents’ childhood and adolescent developmental history. “Mexican” respondents lived in Mexico until age 16, were born in Mexico, and/or reported that they were mostly reared in Mexico. “Mexican-Americans” reported that they lived in the United States until age 16, were born in the United States and/or reported that they were mostly reared in the United States. Although we measured the degree of acculturation in our subjects, we deliberately selected a geographic and developmental criterion for ethnicity to enhance objectivity in group definition, and to give us the clearest possible view of cognitive performance variance that might be attributable to environmental and constitutional factors. Indeed, there is a high correlation between our ethnic categorization (non-Hispanic white, N=353; Mexican-American, N=414; Mexican, N=385) and scores on the Hazuda 22 Mexican acculturation scale, showing correlation coefficients of 0.86 (Pearson) and 0.85 (Spearman). The Hazuda scale itself constitutes a highly reliable acculturation measure among our research sample (Cronbach alpha=0.98).

Assessment of Demographic (Control) Variables

A demographic summary of our sample is included in Table 1 . Each ethnic group is well represented with similar numbers of men and women, and age groupings. Percentages of women in the non-Hispanic white, Mexican-American, and Mexican subsamples are 49.9%, 49%, and 52.5%, respectively. Age was measured as of the respondent’s last birthday. The sampling procedure was stratified so that half of the total sample consisted of people aged 75 and older. The three ethnic categories are almost identical in average age of respondent, at 75 years. As would be expected from these similarities, F-values for gender and age were not significant.

Enlarge table

Despite our attempt to stratify the selection of census tracts by ethnicity and median income, group disparities in educational attainment are large. Mean years of educational attainment for non-Hispanic whites, Mexican-Americans, and Mexicans are 13, 6, and 6 years, respectively. Eighty-seven percent of non-Hispanic whites possess at least a high school diploma, whereas 38% and 16% of the Mexican-American and Mexican subsamples, respectively, are high school graduates. The F-value of 403.31 for ethnic effects on education reflects this between-group disparity.

Assessment of Cognitive Function

The study’s three cognitive function measures include a global assessment of cognition through the MMSE, 6 an assessment of executive control function through CLOXI, 13 , 23 and a simple CLOXII copying task that is somewhat independent of executive skills. 13 , 15 – 22 In CLOXI, participants are asked to draw a clock according to a set of verbal instructions. This is a free-hand drawing on the blank side of a coding instrument devised by the instrument’s author. For CLOXII, the participant simply copies a clock drawn by the investigator. Testing was administered in English or Spanish, depending upon the preference of the subject. For the purpose of this analysis, “impairment” has been defined on the MMSE as a score of 24 or less (out of 30); 9 or less (out of 15) on CLOXI; and 11 or less (out of 15) on CLOXII. These cutoff criteria were established in advance of data analysis and are based upon previously reported performance results from cognitively impaired populations. 13 , 23

Ordinary Least Square Analysis

Ordinary least square regression has been used in multivariate analyses. We considered it to be preferable to logistic regression because the full arrays of scores on the MMSE, CLOXI, and CLOXII can be taken into account. Identical ordinary least square regression analyses have been conducted on MMSE, CLOXI, and CLOXII. This procedure permits us to assess the association between Mexican-American ethnicity and each cognitive function measure, after the effects of education, age, and gender are removed from the three equations. All relevant variables have been subjected to analysis for nonlinearity and collinearity.

RESULTS

The Effect of Mexican-American Ethnicity on Cognition

Table 2 shows rates of performance below cutoff levels on the cognitive tests according to ethnic group membership, and it shows that the degree of Mexican-American ethnicity correlates with poor performance on all cognitive measures. The failure rate for each measure increases steadily with degree of Mexican-American ethnicity. The percentages of non-Hispanic whites’ test failures on the MMSE, CLOXI, and CLOXII are 7.1%, 24.3%, and 6.6%, respectively. Corresponding percentages for Mexican-Americans are 27.9%, 33.2%, and 19.4%. The Mexican ethnic grouping possesses the highest failure rates for MMSE, CLOXI, and CLOXII at 37.7%, 46.5%, and 28.7%.

Enlarge table

Results from Multivariate Analyses

Multivariate analyses of potential modifier variables are presented in Table 3 . As in previous studies, 7 , 10 , 19 – 21 educational attainment significantly predicts cognitive function. Education has a larger effect on MMSE scores, however, than on either CLOXI or CLOXII. The beta value for the correlation between education and MMSE (0.40, Table 3 ) is larger than for those between education and performance on CLOX I and II (0.25, 0.21, Table 3 ). These analyses indicate that 30% of the variance in MMSE scores is attributable to Mexican-American ethnicity, education, and age ( Table 3 ). For CLOXI and CLOXII, ethnicity, education, and age explain only 15% and 16%, respectively, of the variance in performance ( Table 3 ). These data confirm previous reports that performance on CLOX drawing tasks is less subject to educational and cultural bias than is performance on other screening cognitive tests, such as the MMSE. 16

Enlarge table

Effects of Mexican-American Ethnicity

These ordinary least square results ( Table 3 ) indicate that Mexican-American ethnicity is negatively associated with cognitive function on all measures, even after effects of demographic variables and education are removed from the model. The standardized regression equations (betas) for the relationship between education and cognitive performance decrease from 0.40 for the MMSE ( Table 3 ) to 0.15 for the CLOXI ( Table 3 ). The corresponding beta values for the relationship between ethnicity and cognitive performance, however, remain stable across all three cognitive function measures (−0.08, MMSE; −0.09, CLOXI; −0.11, CLOXII). Zero-order correlation coefficients (not reported in tabular form) for relationships between cognitive performance and ethnicity and education on the MMSE are −0.33 and 0.48, respectively. After controlling for effects of the other three variables in each equation, the partial correlation coefficients for these two predictors are −0.06 and 0.30. For CLOXI, the two zero-order correlations for ethnicity and education are −0.24 and 0.33, whereas their partial correlation counterparts are −0.07 and 0.18. Zero-order coefficients for ethnicity and education on CLOXII are −0.24 and 0.31. Partial correlation coefficients for these variables are −0.09 and 0.16. The similar partial coefficients for ethnicity on all three cognitive performance measures (−0.06 for MMSE, −0.07 for CLOXI, −0.09 for CLOXII) suggest that ethnicity exerts some persisting effect on cognitive performance, whether measured by the MMSE or CLOX.

DISCUSSION

Our findings indicate that in a stratified, random sample of 1,152 Mexican-American and non-Hispanic white elderly residents of El Paso County, Mexican-American ethnicity is a significant variable associated with poor cognitive functioning on a battery of screening tests. The performance impairments occur across a variety of functional cognitive domains, including domains associated with frontal/executive functions. This finding persists after controlling for effects of education, age, and gender. There is additional evidence in this study that impairment in cognitive function has negative functional implications for activities of daily life. High rates of “nondementing” cognitive impairment have previously been reported in Hispanic populations. Gurland et al. 4 reported that cognitive impairment was more common in a group of Caribbean Hispanics in New York City than among non-Hispanic whites. Most of the cognitively impaired persons in this study did not meet formal criteria for “dementia,” but were labeled “borderzone not-demented” instead. Less than half of the study population were rated free of cognitive impairment. Similar results were reported by Fitten, Ortiz, and Ponton. 24 We found that as many as 40% of 100 community-dwelling Caribbean Hispanics age 55 and older, not previously diagnosed with dementia, developed dementia over a period of 3 years. Our report from a Mexican-American data set suggests that this phenomenon of widespread, age-related poor cognitive performance is not limited to Hispanics of Caribbean ancestry. 3 – 4 From the data reported here, we cannot be sure of the neuropsychiatric determinants which underlie the finding of disproportionately poor cognitive functioning among persons of Mexican-American ethnicity. Acculturation variables alone do not seem to be the answer. Our results do not contain clinical information sufficient to allow formal diagnoses of dementia syndromes, so we do not know whether these results are driven by a high prevalence of late-life neuropsychiatric disease in this population. At this time we must say that there seem to be unspecified determinants of later life cognitive impairment in these populations, and we do not know what they are. The data are consistent with a wide range of developmental or neurobiological operating variables, and only future studies will help us understand what these variables might be. Of course, our ultimate goal will be to intervene appropriately, either with prevention or therapeutics.

In conclusion, we here report significant associations between the degree of Mexican-American ethnicity and patterns of cognitive impairment in a large El Paso sample of subjects 65 years of age and older. This pattern of cognitive impairments appears on culturally unbiased measures and is generally independent of age and educational attainment. We are not sure of the explanation for these results, and we hope to plan further studies that will help us understand them better.