Experiences of Discrimination and Incident Type 2 Diabetes Mellitus

The Multi-Ethnic Study of Atherosclerosis (MESA)

Kara M. Whitaker; Susan A. Everson-Rose; James S. Pankow; Carlos J. Rodriguez; Tené T. Lewis; Kiarri N. Kershaw; Ana V. Diez Roux; Pamela L. Lutsey

Disclosures

Am J Epidemiol. 2017;186(4):445-455. 

In This Article

Methods

Study Population

The Multi-Ethnic Study of Atherosclerosis (MESA) is an ongoing cohort study of adults who were initially recruited in 2000–2002 (examination 1) from 6 different field centers in the United States (Baltimore, Maryland; Chicago, Illinois; St. Paul, Minnesota; Los Angeles, California; New York, New York; and Forsyth County, North Carolina). Additional examinations occurred at approximately 2-year intervals (examination 2: 2002–2004, examination 3: 2004–2005, examination 4: 2005–2007, and examination 5: 2010–2013). At recruitment, the 6,814 participants—who self-identified as African American (28%), Chinese (12%), Hispanic (22%), or non-Hispanic white (38%)—were aged 45–84 years and did not have known cardiovascular disease. For the present study, participants were excluded if they had baseline type 1 (n = 10) or type 2 diabetes (n = 849), defined by fasting blood glucose ≥126 mg/dL or use of glucose-lowering medications, or were missing data on diabetes status (n = 24), did not complete both discrimination questionnaires at baseline (n = 129), had no follow-up visits (n = 319), or were missing data on family income (n = 173). The final sample comprised 5,310 persons. Institutional review board approval was obtained at each study site, and written informed consent was obtained from all participants.

Discrimination

Discrimination was assessed at baseline using a modified version of the Major Experiences of Discrimination Scale and the Everyday Discrimination Scale.[25] See Web Table 1 (available at https://academic.oup.com/aje) for complete questionnaires. These scales were originally developed for the Detroit Area Study in Michigan and have demonstrated reliability and validity.[20,25–28] Major experiences of discrimination were assessed by asking participants to report whether they had ever been treated unfairly (yes/no) in 6 situations. A summary score was created by assigning a value of 0 for each negative response and 1 for each affirmative response (range, 0–6). Few participants reported more than 2 experiences of discrimination; therefore this summary score was categorized into 3 groups (0, 1, or ≥2 situations), with those reporting no discrimination (score = 0) serving as the reference group. For each affirmative response, participants were asked to select 1 reason why they believed they were treated unfairly (race/ethnicity, gender, age, religion, physical appearance, sexual orientation, income level/social class, or other). Similar summary scores were created for discrimination specifically attributed to race/ethnicity and to other sources by summing the number of times a respondent selected each attribute. Both scores have ranges of 0–6, and for analyses they were categorized as 0, 1, or ≥2 situations. Other sources of discrimination not attributed to race/ethnicity were combined due to small sample sizes.

Day-to-day experiences of discrimination, such as being treated with less respect than others, were assessed using the 9-item Everyday Discrimination Scale.[25] Participants were asked how often they experienced discriminatory events in their everyday life without reference to why they believed they were treated unfairly. Response options were on a 6-point scale: 1 = almost every day; 2 = at least once a week; 3 = a few times a month; 4 = a few times a year; 5 = less than once a year; and 6 = never. Items were reverse coded and summed, with higher scores indicating greater frequency of day-to-day discrimination. Summed scores were then averaged across the 9 items to create an everyday discrimination score (range, 1–6). After examining the distribution of scores, they were categorized into approximate quartiles: none (score = 1; 25.4%), low (score >1 and <1.5; 27.3%), moderate (score 1.5–2; 25.5%), or high (score >2; 21.8%).

Diabetes

The primary outcome was incident diabetes as defined by the 2003 American Diabetes Association criteria: fasting glucose ≥126 mg/dL, use of oral hypoglycemic medication and/or insulin, or self-reported physician diagnosis of diabetes.[29] Incident diabetes was identified at one of 4 follow-up examinations that took place in 2002–2003, 2004–2005, 2005–2007, and 2010–2011. Follow-up time was defined as time between the baseline visit and either an incident diabetes event or the last clinic visit attended, whichever occurred first.

Covariates

Baseline age, field center, race/ethnicity (African American, Chinese, Hispanic, non-Hispanic white), sex, marital status (married/living with partner, widowed, divorced or separated, never married), education (less than high school diploma, high school diploma or some college/technical school, college degree or more), and annual family income (<$20,000, $20,000–$49,999, ≥$50,000) were included in all models.

Psychosocial covariates included chronic stress, assessed using the 5-item Chronic Burden Scale,[30] and depressive symptoms were measured with the 20-item Center for Epidemiologic Studies Depression Scale.[31] Behavioral covariates included alcohol use (none or light drinker, moderate drinker, heavy drinker), smoking (current smoker, former smoker, never smoker), physical activity (metabolic equivalents of physical activities in minutes per week) as measured by the MESA Typical Week Physical Activity Survey, and diet quality score as calculated in prior MESA reports (9 food groups assessed using a food frequency questionnaire).[32,33] Measures of obesity included BMI and waist circumference. Weight and height were assessed using a balance beam scale and stadiometer, respectively. BMI was calculated as weight in kilograms divided by height in meters squared. Waist circumference was measured at the level of the umbilicus.

Statistical Analyses

Time to incident diabetes was modeled with Cox proportional hazards models. Major experiences of discrimination and everyday discrimination were included as predictors in separate models as both continuous and categorical variables. Initial models adjusted for baseline sociodemographic characteristics because these could be potential confounders of the discrimination-diabetes association. Based on findings from previous research indicating a relationship between discrimination and mental health outcomes, adverse health behaviors, and obesity, we consider these characteristics potential mediators of the relationship between discrimination and diabetes. Our theoretical model can be summarized as follows: discrimination → stress/depression → adverse health behaviors → obesity → diabetes. Therefore, in model 2 we also adjusted for chronic stress and depressive symptoms, assessed at baseline. In model 3 we further adjusted for behavioral covariates, including baseline alcohol use, smoking, physical activity and diet quality. Finally, in model 4 we additionally adjusted for obesity, using BMI and waist circumference. Interpretation of results follows from our theoretical model and accordingly differs by the statistical model being evaluated. For example, attenuation of the effect estimate after adjustment for the covariates in model 1 is interpreted as evidence of substantial confounding by 1 or more of the adjustment variables. By contrast, attenuation of the effect estimate after adjustment for the covariates in models 2, 3, or 4 is interpreted as evidence of mediation by 1 or more of the adjustment variables.

We also examined interactions with race/ethnicity, age, and sex by including cross-product terms in the models. For all models, the proportional hazards assumption was tested by inspection of log-minus-log survival versus log survival curves by including time-covariate interactions for the discrimination measures in the final models. All statistical analyses were conducted using SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina). All tests were 2-sided, with statistical significance set at P < 0.05.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....