Comparing cognitive function in white Mexican & non-Hispanic white Americans with/without diabetes

Abstract

Objectives

An association between type-2 diabetes mellitus and cognitive decline is well known. Additionally, type 2 diabetes is known to be more physically burdensome for minorities. However, the combined impact of both ethnicity and diabetes on cognition is still not clear.

Methods

Data from the 2014 Health and Retirement Study (HRS) was used in this study to compare the cognitive functioning of non-Hispanic White (n = 10,658) and White Mexican/Mexican American (n = 847) individuals, age 50+ years, with or without type 2 diabetes. Cognitive functioning was measured by a composite of three constructs (serial 7 s, immediate, delayed recall). Ethnic groups and diabetes status were examined concerning cognitive functioning.

Results

A Multivariate Analysis of Covariance (MANCOVA) indicated significant main effects for ethnicity (F(3,11,496) = 11.15, p < .001) and diabetes status (F(3,11,496) = 3.15, p < .024), with Mexican Americans and those with diabetes exhibiting worse cognitive performance than non-Hispanic Whites and individuals without diabetes. There were significant effects for all covariates. A step-wise multiple regression indicated that education, age, depression, ethnicity and diabetes status accounted for a combined 28.4% variance in the cognitive performance composite.

Conclusions

Results found that education contributes significantly to variation of cognitive performance. The impact of education could be related to various possibilities. However, the impact of health literacy is a likely component, which has a positive relationship with level of education. Individuals with higher health literacy are more conscientious in health actions (e.g., exhibit regular self-care, glucose monitoring, and foot care). Therefore, the study results indicate it is likely that the duration of diabetes, and diabetes management (e.g., effective control of blood glucose, blood pressure, and lipids), contributing to cognitive decline. Cognitive screening at routine doctor visits is encouraged, particularly for Mexican/Mexican Americans, as the current study found support for ethnic minority vulnerability to the negative impacts of diabetes.

Diabetes is a widespread disease in the United States, with prevalence continuing to rise. The global prevalence in 2014 of both types of diabetes was 422 million, according to the World Health Organization [15], and the number is expected to increase to 552 million by 2030 [13]. Ethnic minorities are disproportionally represented within individuals with diabetes. Prevalence is highest for American Indians/Alaska Natives (14.7%), non-Hispanic Blacks (11.7%), and people of Hispanic ethnicity (12.5%) than it is for Asians (9.2%) and non-Hispanics (7.5%) [3].

Minorities, particularly Mexican/Mexican Americans, are at increased risk for type 2 diabetes. According to the Department of Health and Human Services, Hispanics are approximately twice as likely to be diagnosed with diabetes as non-Hispanic Whites, and they are 40% more likely to die from diabetes [19]. Research offers evidence that obesity, hypertension, prediabetes, alcohol consumption, immigration, urbanization, and lowered consumption of fruits and vegetables contribute to increased vulnerability [25].

Evidence has shown that the longer a person has type 2 diabetes, the more significant and rapid their cognitive decline is likely to be [10, 17, 18]. Research results regarding cognitive decline vary, with hyperglycemia, hyperinsulinemia, and hypoglycemia being central for cognitive decline in older adults [2]. Additionally, individuals with poorly controlled diabetes (as measured by HbA1c levels) over 20 years experienced more significant cognitive decline than those with better-controlled diabetes [18]. However, there is extensive evidence that cognitive decline occurs as early as mid-life among those with diabetes. In the oldest adults (age 80 and older) with type 2 diabetes, a relationship between hyperglycemia and cognitive function was not found [8]. Previous research has supported higher HbA1c levels being associated with lower cognitive function [12], so cognitive decline likely occurs before 80 years of age, with the lack of relationship for the oldest adults (80+) being due to a restriction of range. Further studies have also supported a younger age range for cognitive decline for individuals with diabetes [1]. While participants with diabetes (age 65 and older) had significantly lower baseline cognitive scores, there was not a significant difference between individuals with and without diabetes for the rate of cognitive change [1]. Therefore, it is likely that cognitive declines occur at an earlier age for individuals with diabetes.

Diabetes is a condition that affects many aspects of life and physiological systems, but especially noticeable within a psychological health domain is its effect on cognition. Cognitive decline for those with type 2 diabetes is associated with lower quality of life and poorer diabetes disease management [4]; therefore, understanding the mechanisms that precedes it is imperative. The current study aimed to investigate the impact of type 2 diabetes, as well as ethnicity, on cognitive functioning for independent living adults 50 years of age and older. This age criteria was set given that most research studies involving cognitive decline in adults with type 2 diabetes focused on adults 65 and older [1, 10] or even 80 years of age and older [8]. Multiple studies recommended a lower age criterion for participants to better understand the relationship of cognition and diabetes [1, 8], given that both studies concluded that cognitive changes likely occurred younger then the age of participants in their studies. Taking these previous research findings into consideration, an age criterion of 50 years and older for this study’s participants was decided to further investigate how cognitive functioning is influenced.

Methods

Participants

The total sample with all variables being used was 11,505 individuals who were age 50 and older. Only participants who were given the measures of interest were included. An inclusive criterion of either non-Hispanic White or White Mexican/Mexican American for ethnic groups was also used. Non-white individuals were excluded to limit any potentially confounding effect of race.

Individuals with type 1 diabetes were eliminated from the sample based on age of diagnosis, given that type 1 diabetes is peaked diagnosed at approximately 14 years old for the vast majority of cases, and almost always begins prior to age 40 [16]. Thus, this study excluded participants with diabetes who were diagnosed at age 40 or younger. Using this method, the final sample size was: 318 White Mexican/Mexican Americans and 2219 non-Hispanic Whites with type 2 diabetes, and 529 White Mexican/Mexican Americans and 8439 Whites without type 2 diabetes (see Table 1). Following informed consent, a brief battery of measures was administered.

Table 1.

Sample sizes for all groups

	Total Sample		Non-Hispanic White		Mexican/Mexican American White
	N	%	n	%	n	%
Sample Size:	11,505	100	10,658	92.6	847	7.4
Gender:
Male	4774	41.5	4440	41.7	334	39.4
Female	6731	58.5	6218	58.3	513	60.6
Type 2 Diabetes Status	2537	22.1	2219	20.8	318	37.5

†Type 2 Diabetes Status is based on 1) self-report of a physician diagnosing participant with diabetes and 2) age elimination of type 1 diabetes

Measures

Demographics and health information

This data included: gender, age at time of interview, self-reported race and ethnicity (including Hispanic type), years of education, and self-reported diabetes status. Only participants who identified their race as “White/Caucasian” were used to eliminate confounding effects of race. In addition, Hispanic type is assessed and only those who identified as either “Hispanic, Mexican” or “non-Hispanic” were included. Other Hispanic, non-Mexican, individuals were not included to reduce variability due to Hispanic ethnic diversity.

Cognitive function

These measures include immediate and delayed recall memory and mental status. For the two memory measures, four lists of 10 equivalently difficult words were used and each participant was randomly assigned one list. The total number of correct words recalled immediately after being read to the participant was their immediate recall score. After a delay of approximately ten minutes, the participant was then asked to recall as many of the ten words that they could. The number of words correctly recalled was their delayed recall score.

Mental status, measured by serial 7 s, was adapted from the Mini-Mental State Examination (MMSE). Past research has provided evidence for satisfactory reliability and construct validity for the MMSE [26]. Specifically, participants were asked to subtract seven from 100 and then seven from that resulting number (i.e., 93). They were asked to do this a total of five times. The total score for this measure was five, with zero being the lowest possible score and five indicating they got all subtractions correct.

Depression

Previous research has also demonstrated that depression can influence cognitive functioning [22], therefore depressive symptoms was taken into account for this study. The Health and Retirement Study (HRS) inquiries about depression in an 8-question form adapted from the Center for Epidemiological Studies Depression Scale (CES-D). The CES-D, including this short form, has been shown to have good reliability and validity with older adults [5, 11]. The scale has also been tested with different populations, which demonstrated acceptable validity. The very few participants who did not answer at least five questions were excluded. Alpha internal consistency reliabilities were within acceptable limits (α >0.70) for all participants and each of the four groups, those with and without diabetes crossed with Hispanic ethnicity, yes/no.

Procedure

Ethical approval for the Health and Retirement Study (HRS) Study was granted by the University of Michigan Institutional Review Board. Data for this cross-sectional study was drawn from the HRS 2014 data collection wave, gathered between March 2014 and April 2015 via in-person or telephone interview. The HRS is a population probability sample of the 48 contiguous United States of those 50 and older who are living independently; it oversamples Hispanic and Black residents. Data used included demographics, physical health, and cognition.

Planned analyses

A 2 × 2 between-subjects Multivariate Analysis of Covariance (MANCOVA) was planned to test two hypotheses: (1) being diagnosed with diabetes, as reported in the physical health data, would be associated with poorer performance of measures of cognitive functioning, and (2) White Mexican/Mexican American individuals with diabetes would demonstrate lower cognitive performance compared to non-Hispanic White individuals with diabetes. The independent variables were type 2 diabetes status (yes/no) and ethnicity (White Mexican/Mexican American/non-Hispanic White). The dependent variables were immediate recall, delayed recall, and serial 7 s performance. Age, years of education, and depression (measured by the CES-D) were covariates due to the known significant association of these variables to cognition [14, 22, 27].

Results

Hypothesis testing

Based on Pillai’s criterion, three covariates, years of education (F(3,11,496) = 507.72, p < .001; Pillai’s V = .117, partial η2 = .117), age (F(3,11,496) = 693.18, p < .001; Pillai’s V = .153, partial η2 = .153), and depression (F(3,11,496) = 86.69, p < .001; Pillai’s V = .022, partial η2 = .022) were significantly related to worse cognitive performance for older age and more depression, with an opposite relationship for more years of education. The predicated main effects of diabetes status (F(3,11,496) = 3.15, p < .024; Pillai’s V = .001, partial η2 = .001), as well as the predicated main effect for ethnicity (F(3,11,496) = 11.15, p < .001; Pillai’s V = .003, partial η2 = .003), were both significant in the expected direction controlling for the three covariates. However, the interaction between diabetes status and ethnicity was not significant (F(3,11,496) = .734, p = .532; Pillai’s V = .000, partial η2 = .000).

Overall the study predications were confirmed by significant results for the main effects of diabetes status and ethnicity. Mexican/Mexican Americans (Serial 7 s M = 2.70, SD = 1.87; Immediate recall M = 4.79, SD = 1.75; Delayed recall M = 3.86, SD = 2.00) had lower scores than non-Hispanic Whites (Serial 7 s M = 3.80, SD = 1.58; Immediate recall M = 5.49, SD = 1.73; Delayed recall M = 4.51, SD = 2.03) on all cognitive measures, without taking diabetes status into account. In addition, while the interaction was not significant, Mexican/Mexican American individuals with diabetes had lower scores on all cognitive measures (Serial 7 s M = 2.42, SD = 1.93; Immediate recall M = 4.48, SD = 1.70; Delayed recall M = 3.50, SD = 1.99) than non-Hispanic individuals with diabetes (Serial 7 s M = 3.65, SD = 1.64; Immediate recall M = 5.21, SD = 1.72; Delayed recall M = 4.17, SD = 1.96). While there is overlapping variance in both distributions, the variance between groups is smaller when examining only ethnicity groups’ performance.

The MANCOVA analysis also indicated significant effect sizes for all three covariates and both independent variables (see Table 2). The independent variables accounted for less variance in the dependent variable, cognitive function. To further examine the relationship between the covariates, independent, and dependent variables, a multiple regression step-wise analysis was therefore conducted.

Table 2.

Multivariate analysis of covariance predicting cognitive status: tests of ethnicity, diabetes status, and covariates

	Pillai’s V	df	F	ηp2
Main Effect
Ethnicity	.003	311,496	11.15**	.003
Type 2 Diabetes status	.001	311,496	3.15*	.001
Covariates
Age	.153	311,496	693.18**	.153
Education	.117	311,496	507.72**	.117
Depressive Sx	.022	311,496	86.69**	.022
Interaction effects
Ethnicity x Diabetes	.000	311,496	.734	.000

†* Significant at .05 level

‡** Significant at .01 level

§ Sx = symptoms

Secondary analyses

A composite cognition dependent variable was created for the forward multiple stepwise regression. This composite variable took into account all three of the cognitive measures. The single serial 7 s score was double weighted to ensure all constructs (serial 7 s and the two recall measure) contributed equally to the composite variable. An alpha was calculated for this three-item cognitive composite scale and was within acceptable limits for all groups, including those with and without diabetes and both ethnic groups. This composite cognition variable was used as the dependent variable in this analysis. The independent (diabetes status and ethnicity) and covariate (age, years of education, depressive symptoms) variables from the MANCOVA analysis were all used as predictor variables.

A forward stepwise multiple regression was conducted to investigate whether age, years of education, depressive symptoms, ethnicity, and diabetes status were necessary to predict cognitive performance. A stepwise regression includes the order of entry of variables based solely on statistical criteria [24]. For step 1, years of education was entered into the regression and was significantly related to cognitive performance, F(1,11,503) = 2443.41, p < .001, accounting for 17.5% of the variance in cognition. Years of education and age were entered in step 2. These two predicator variables were also significantly related to cognition, FΔ (111502) = 1417.89, p < .001, accounting for 26.6% of the variance in cognition. For step 3, depressive symptoms (CES-D) was entered along with the previous two predicator variables. These predicator variables were also significant, FΔ (111501) = 265.03, p < .001, and accounted for 28.2% of the variance in cognition. Depressive symptoms, once added to the regression in step 3, explained an additional 1.7% of the variance in cognitive performance (ΔR² = .017).

The next variable added was ethnicity, which was also significant, FΔ (111500) = 23.86, p < .001, and together all these variables accounted for 28.3% of the variance in cognition. Lastly, diabetes status was entered, along with the previous four predicator variables, FΔ (111499) = 13.36, p < .001, which altogether explained 28.4% of the variance in cognition.

It is worth noting that ethnicity has a larger standardized beta coefficient (β = −.041) than diabetes status (β = .029) when all predicator variables were entered into the regression, model 5. Further details, including standardized and unstandardized coefficients for each model, can be found in Tables 3 and 4.

Table 3.

Stepwise multiple regression analysis of cognitive performance

					Change Statistics
Step	Determinants	R	R2	Adjusted R2	R2 Δ	F Δ	Df2	Sig. F Δ
1	Years of Education	.419	.175	.175	.175	2443.41	11,503	.000
2	Years of Education, Age	.515	.266	.266	.091	1417.89	11,502	.000
3	Years of Education, Age, Depressive Sx	.531	.282	.282	.017	265.03	11,501	.000
4	Years of Education, Age, Depressive Sx, Ethnicity	.533	.284	.283	.001	23.86	11,500	.000
5	Years of Education, Age, Depressive Sx, Ethnicity, Diabetes Status	.533	.285	.284	.001	12.36	11,499	.000

†Sx = symptoms

Table 4.

Stepwise multiple regression coefficients

Variable	B	SE B	Β	t	p
Model 1
Years of Education	.791	.016	.419	49.43	.000
Model 2
Years of Education	.736	.015	.390	48.55	.000
Age	−.160	.004	−.302	−37.66	.000
Model 3
Years of Education	.690	.015	.365	45.21	.000
Age	−.161	.004	−.304	−38.32	.000
Depressive Sx	−3.02	.186	−.131	−16.28	.000
Model 4
Years of Education	.655	.017	.347	38.84	.000
Age	−.165	.004	−.311	−38.64	.000
Depressive Sx	−3.02	.185	−.131	−16.26	.000
Ethnicity	−.920	.188	−.043	−4.89	.000
Model 5
Years of Education	.651	.017	.344	38.52	.000
Age	−.164	.004	−.309	−38.40	.000
Depressive Sx	−2.96	.186	−.128	−15.93	.000
Ethnicity	−.875	.189	−.041	−4.64	.000
Diabetes Status	.099	.027	.029	3.66	.000

† Sx = symptoms

Discussion

The current study aimed to investigate the impact of type 2 diabetes and Mexican American ethnicity on cognitive functioning for individuals 50 years of age and older, experimentally controlling for race. The age criteria for the current study were set given previous literature [1, 10].

The first hypothesis, based on previous studies, was that participants with diabetes would exhibit lower performance for cognitive functioning, controlling for education, age, and depressive symptoms, than participants without diabetes [6, 18].

Previous research has provided evidence that that cognitive decline is greater for those with diabetes and that the effect of diabetes is more devastating, as well as being more prevalent, for ethnic minority individuals, per the U.S. department of Health and Human Services [1, 10, 19]. Therefore, the current study’s second hypothesis was that White Mexican/Mexican Americans with diabetes would exhibit lower performance for cognitive functioning compared to non-Hispanic White Americans with diabetes. Results from the MANCOVA analysis indicated that ethnicity and diabetes status were both significantly related to cognitive performance. Both ethnicity and diabetes status were related to a combination of their serial 7, immediate recall, and delayed recall performance and elicited significant main effects. Thus, the study’s first hypothesis was supported – participants with diabetes had lower scores on all cognitive measures than participants without diabetes.

Previous research has also found that individuals with diabetes age 80 and older experienced a greater decline (than those without diabetes) over a six-year period in multiple neuropsychological tests, including serial 7 s [6]. It is worth noting that the current study’s participants with diabetes, at age 50 and older, demonstrated similar decreased serial 7 s performance compared to those without diabetes. Importantly, this finding offers insight that diminished attention and concentration is present even at a much lower age range than previously found, as previous research did not include individuals as young as age 50.

Results also indicated that White Mexican/Mexican Americans exhibited lower performance than non-Hispanic White individuals on all cognitive measures. Given previous literature that diabetes is more prevalent and lethal for Mexican/Mexican Americans [9, 19], this is not surprising. While the current study indicated that White Mexican/Mexican American with diabetes exhibited lower performance on all cognitive measures compared to non-Hispanic Whites, the interaction term was not significant.

In addition, age, years of education, and depressive symptoms (covariate variables) all had a significant relation to cognitive performance. Given the significant relation of covariates and the previous two findings, a forward step-wise multiple regression was performed to investigate which constructs contribute more to cognitive performance. A stepwise multiple regression gives information on which predicator is contributing most to the outcome (e.g., cognitive performance) and also allows all predicators to be compared.

The stepwise multiple regression indicated that education was the strongest predicator, followed by age, depression, ethnicity, and finally diabetes status. All predicator variables together accounted for a total of 28.4% of the variance in cognitive performance. Education alone accounted for 17.5% of variance in step 1. Depression, once added to the regression in step 3, explained an additional 1.7% of the variance in cognitive performance. Besides education and age, depression was the most statistically relevant to predict cognitive performance. Depression (unlike education or age) could be lessened with talk therapy, increased social involvement and increased activity level, and/or medication. This is an interesting finding which should be taken into account during routine doctor visits. Additionally, when all predicators were included in the regression model, ethnic status elicited a stronger relationship to cognitive performance than diabetes status.

The current study offers evidence that the effects of education, age, and depression contribute significantly to mental status and memory recall. Additionally, ethnicity and diabetes status also contribute above and beyond to the composite measure of cognition, with ethnicity being the stronger component. These results offer the insight that diabetes status (e.g., simply having diabetes or not) is not as influential on cognition as originally proposed. Rather, it is likely the duration of diabetes illness that is contributing to cognitive decline [18]. According to the 2020 National Diabetes Statistics Report, type 2 diabetes is typically diagnosed between the ages of 45–64 [3]. Participants with diabetes 50 years of age and older (the age criteria for the current study) would likely have recently received a type 2 diabetes diagnosis. This fact could be contributing to the results of diabetes status on cognition. It is likely that older age participants in other research studies have also experienced a longer duration of diabetes, which could be a stronger influence on factors, such as cognitive functioning, than simply having a diabetes diagnosis.

In addition, diabetes management—measured by glycemic control – may be more influential on cognitive performance than diabetes status. This would also help explain the results of this study. The current study found that education accounted for more than 17% of variance in cognitive performance. Socioeconomic status (SES), including amount of education, has a negative relationship with risk of mortality from diabetes [23]. This relationship is related to a variety of possibilities but is likely related to perception of their disease, coping strategies for their illness, diet, and depressive symptoms even when controlling for open access health care [7].

It is also well documented that ethnicity minorities have poorer diabetes management [21], as defined by less glycemic control. Poor diabetes management is also likely the result of lack of access to health care, socioeconomic status, cultural attitudes and behaviors. Additionally, rapid progression of the disease in minorities is likely related to poor diet, obesity, and sedentary lifestyle [21]. Cognitive difficulties can also negatively affect self-care in type 2 diabetes [4], which can further exacerbate the effects of the disease. Health literacy is useful to expand on this idea. Health literacy is defined by the U.S. Department of Health and Human Services as the degree to which individuals have the capacity to obtain, process, and understand basic health information needed to make appropriate health decisions [20]. In a nationally representative sample, individuals without a high school diploma exhibited a low health literacy. The current study offered limitations by failing to include a variety of measures of diabetes management. Future research should expand this study by including measures of diabetes management, in addition to diabetes status.

In terms of treatment, cognitive screening for individuals with diabetes is encouraged. The current study involved basic tests of cognitive functioning that could be quickly administered by a medical professional (e.g., vocational nurse, nurse practitioner, general physician). These assessments, such as the three-item composite used here, could also be administered while gathering a patient’s recent medical history or preliminary body functioning tests (e.g., blood pressure, pulse, etc.). Additionally, if once a year, such as at an annual wellness visit, cognitive functioning could be screened then identifying decline and offering treatment at onset would be beneficial.

Overall, the current study has provided evidence in the use of immediate and delayed recall and serial 7 s, as a quick screening assessment, in detecting decreased cognitive performance for those with diabetes. Additionally, the importance of cognitive screening for Mexican/Mexican Americans during regular check-ups/wellness visits should be especially enforced, given their increased vulnerability for decreased cognitive functioning.

Declarations

Conflict of interest

The Authors declare that there are no conflict of interest.