Abstract
Although racial residential segregation and interpersonal racial discrimination are associated with cardiovascular disease, few studies have examined their link with diabetes risk or management. We used longitudinal data from 2,175 black participants in the Coronary Artery Risk Development in Young Adults (CARDIA) Study to examine associations of racial residential segregation (Gi* statistic) and experiences of racial discrimination with diabetes incidence and management. Multivariable Cox models estimated associations for incident diabetes and GEE logistic regression estimated associations with diabetes management (meeting targets for HbA1c, systolic blood pressure, and LDL cholesterol). Neither segregation nor discrimination were associated with diabetes incidence or management.
Keywords: Diabetes, Segregation, Racial Discrimination, Epidemiology, Neighborhood
1. Introduction
In the United States, the prevalence of type 2 diabetes is nearly twice as high among non-Hispanic black adults as among non-Hispanic whites, a disparity that has widened over the past 30 years (Menke et al., 2015). Black-white disparities have also been reported in diabetes incidence (Bancks et al., 2017; Geiss et al., 2014) and mortality (Rosenstock et al., 2014). In addition, black adults with diagnosed diabetes have been found to have lower rates of HbA1c testing compared white adults (Meng et al., 2016). Given these well-described disparities, identification of the determinants of poor diabetes outcomes among non-Hispanic black individuals is of significant public health importance.
Racial disparities in diabetes outcomes are likely due to a combination of biological, social, and environmental factors including socioeconomic status, chronic stress, and exposure to neighborhood-level risks and resources (Kershaw and Pender, 2016). Underlying many of these factors is racial residential segregation, the physical separation of blacks by residence from other racial/ethnic groups, which has been described as a fundamental cause of health disparities (Williams and Collins, 2001). Segregation is believed to influence chronic disease outcomes such as diabetes by limiting access to health-promoting resources and through increasing exposure to health-harming and stressful environmental conditions such as economic disinvestment, crime, and unhealthy food outlets (Kramer and Hogue, 2009; Williams and Collins, 2001). While segregation is a structural factor influencing racial disp.arities in health, interpersonal racial discrimination is a psychosocial stressor believed to influence health outcomes through chronic stress (Williams and Mohammed, 2009) and/or through adoption of unhealthy behaviors as a coping mechanism (Borrell et al., 2010) Exposure to racial residential segregation and interpersonal racial discrimination have been linked to adverse chronic conditions including obesity, hypertension, and cardiovascular disease (Bower et al., 2015; Kershaw and Albrecht, 2015; Sims et al., 2012; Stepanikova et al., 2017). However, few studies have examined the relationship of these factors with diabetes.
Two prior studies examined associations between interpersonal racial discrimination and higher incident diabetes, both reporting a positive association (Bacon et al., 2017; Whitaker et al., 2017). To our knowledge, no studies to date have examined associations between racial residential segregation and incident diabetes among non-Hispanic blacks. Three cross-sectional studies have assessed associations with diabetes prevalence among black adults, with mixed findings. Two studies reported no association (Piccolo et al., 2015; Gaskin et al., 2014) whereas a recent study that linked electronic health records to a spatial index of racial isolation (the extent to which black residents are only exposed to members of their own racial group) found segregation to be associated with greater diabetes prevalence (Bravo et al., 2018). Further, two ecological studies found that age-adjusted diabetes mortality was higher in more segregated neighborhoods (Hunt et al., 2014; Rosenstock et al., 2014). These prior studies suggest an association between residential segregation and diabetes is plausible, but prospective studies are needed (Kershaw and Pender, 2016).
One potential explanation for the disparate findings regarding the association of segregation with diabetes prevalence versus mortality is that segregation may be more closely related to diabetes management and severity rather than development. Consistent with this idea, previous studies have shown interpersonal racial discrimination was associated with worse health behaviors and poorer glycemic control in patients with type 2 diabetes (Assari et al., 2017; Dawson et al., 2015). It is unknown whether segregation, a structural form of racial discrimination, is associated with poorer diabetes management outcomes such as hemoglobin A1c, systolic blood pressure, or low-density lipoprotein cholesterol levels. However, it is plausible that residents of more segregated neighborhoods may have worse diabetes management outcomes through pathways including chronic stress, limited access to healthy food and physical activity resources (Goodman et al., 2018), or reduced access to healthcare (Gaskin et al., 2012). Alternatively, since the findings relating segregation to diabetes mortality were based on an ecological study design, these associations might be due to ecological fallacy (Hunt et al., 2014; Rosenstock et al., 2014). Further study is needed to understand whether structural and interpersonal racial discrimination (e.g. racial residential segregation and experiences of racial discrimination) influence diabetes incidence and management.
Our objective was to estimate associations of racial residential segregation and interpersonal racial discrimination with incident diabetes and diabetes management among black participants of the Coronary Artery Risk Development in Young Adults (CARDIA) Study. We hypothesized that participants who 1) lived in more highly segregated neighborhoods or 2) reported a greater number of experiences of interpersonal racial discrimination would have increased diabetes incidence and would be less likely to meet diabetes management targets.
2. Materials and Methods
2.1. Study Population
CARDIA is a prospective multi-center cohort study of the development and determinants of cardiovascular disease and related risk factors (Friedman et al., 1988). CARDIA used population-based sampling to enroll 5,115 black and white adults aged 18–30 years in 1985–1986 from 4 locations: Birmingham, AL; Chicago, IL; Minneapolis, MN; and Oakland, CA. The cohort was recruited to be balanced across age, education, race, and sex. Participants were re-examined after 2, 5, 7, 10, 15, 20, 25, and 30 years, with retention rates of 91%, 86%, 81%, 79%, 74%, 72%, 72%, and 71%, respectively. Participants’ home addresses were geocoded at baseline and in years 7–25 in order to link census tract-level exposures (e.g. segregation). In this analysis, we included data from the baseline exam and years 7–30. We focused on black participants due to a lack of overlap in segregation scores between black and white participants and the unique historical context of segregation of black Americans (Kramer and Hogue, 2009). Among 2,637 black participants at baseline, we excluded 21 with prevalent diabetes at baseline, and 441 who were missing follow-up information at later exams needed to classify diabetes, for a total sample size of 2,175. Excluded participants were on average slightly younger, more likely to be male, smoke, report heavy alcohol consumption, and to have a high school degree or less. Exposure to racial residential segregation and interpersonal racial discrimination did not differ between included and excluded participants. CARDIA was approved by each field center’s Institutional Review Board, and all participants provided written informed consent.
2.2. Racial Residential Segregation
Data on census tract-level racial composition was linked to CARDIA participants at exams 0, 7, 10, 15, 20, and 25 using the U.S. Census or American Community Survey data that aligned most closely temporally (Kershaw et al., 2017; Pool et al., 2018). Racial residential segregation was then measured using the Getis-Ord Local Statistic (Gi*)(Getis and Ord, 1992). The Gi* statistic is a spatial autocorrelation measure that estimates the extent to which the racial composition (i.e., percent non-Hispanic black race) of a given census tract, and that of its neighboring tracts, is similar to that of the larger areal unit surrounding the tract. The Gi* statistic is a measure of clustering, one of the 5 dimensions of segregation defined by Massey and Denton, which quantifies the extent to which racially similar neighborhoods group together in space (Massey and Denton, 1988). We chose this approach over other potential measures of segregation (e.g. census tract racial/ethnic composition) because the Gi* statistic incorporates information on the racial composition of the larger metropolitan area, and thus better reflects the contextual and spatial elements of segregation. For example, a neighborhood where 30% of residents are non-Hispanic black would indicate over-representation in a city that is 5% black, but would indicate under-representation in a city that is 50% black. In CARDIA, metropolitan statistical areas were used as the larger areal unit for participants living in metropolitan areas and counties were used for participants not living in metropolitan areas. Neighboring tracts were defined as those that shared a border with the index tract. This approach was chosen over alternatives (e.g. tracts within a 1-mile radius around the index tract centroid) to be consistent across tracts of varying sizes, and has been used in prior examinations of segregation and health outcomes in CARDIA (Kershaw et al., 2017; Pool et al., 2018).
The Gi* statistic returns a z-score with higher positive values indicating the tract and its neighbors have a higher than expected percentage of black residents and higher negative values indicating the tract and its neighbors have a lower than expected percentage of black residents. As in prior studies (Kershaw et al., 2017; Mayne et al., 2019; Pool et al., 2018), segregation was categorized as high (Gi*>1.96), medium (Gi* 0 to 1.96), and low (Gi*<0) based on the critical z-score values for a 95% confidence interval (−1.96 to 1.96). As no participants lived in areas with Gi*<−1.96, the cutpoint for the low segregation category was set to 0, which indicates a z-score at the mean racial composition for the surrounding metropolitan area or county. As participants moved during the follow-up period, we examined both baseline and time-varying segregation.
2.3. Experiences of Racial Discrimination
As described previously (Cunningham et al., 2013; Krieger and Sidney, 1996; Krieger et al., 1998), in exam years 7, 15, 25, and 30, self-reported experiences of racial/ethnic discrimination were assessed using the validated Experiences of Discrimination Index (Krieger et al., 2005). Participants were asked whether they had “ever experienced discrimination, been prevented from doing something, or been hassled or made to feel inferior…because of your race or color” in seven situations. Six situations were consistent across exams: at school, getting a job, at work, getting housing, getting medical care, and on the street/in a public setting. The last situation was assessed as “from police or in the courts” in year 7 and “at home” in years 15, 25, and 30. At each exam, the 7 items were combined into an index score and categorized as no experiences, 1–2 experiences, or ≥3 experiences, as in prior studies (Cunningham et al., 2012; Krieger and Sidney, 1996). The year 7 racial discrimination category was treated as the baseline value in our analyses.
2.4. Diabetes Incidence and Management
Serum glucose was measured at each exam using the hexokinase ultraviolet method (baseline) and using hexokinase coupled to glucose-6-phosphate dehydrogenase (subsequent exams). Values were standardized across exams. Hemoglobin A1c (HbA1c) was assessed at exam years 20 and 25 using Tosoh G7 (variant mode) high-performance liquid chromatography (national Glycohemoglobin Standardization Program-certified assays). Incident diabetes was determined at each exam if any of the following criteria were met: fasting glucose ≥126 mg/dL, 2-hour glucose during the oral glucose tolerance test of ≥200 mg/dL, hemoglobin A1c ≥6.5%, or current use of medication for the treatment of diabetes mellitus (see Supplemental Methods).
Among participants who 1) developed diabetes during the study period according to the clinical definition above and 2) responded “yes” to the question: “Has a doctor or nurse ever said that you have diabetes?”, optimal diabetes management was assessed using three targets: hemoglobin A1c (HbA1c) <7.0%, systolic blood pressure (SBP) <130 mmHg, and low density lipoprotein (LDL) cholesterol <100 mg/dL (based on diabetes treatment guidelines and prior studies) (American Diabetes Association, 2018; Ikramuddin et al., 2013; Jellinger et al., 2017; McFarlane et al., 2002). We restricted to participants who self-reported a diabetes diagnosis in order to examine diabetes management among participants who knew their diabetes status. This analysis included exam years 20 and 25 only, as those are the only years in which HbA1c was measured in the entire population. Resting SBP was measured at each exam by trained technicians 3 times at 1-minute intervals using a random-zero sphygmomanometer (years 0–15) or an oscillometer (years 20–25), and the second and third measurements were averaged. Values from years 20–25 were calibrated to the random-zero readings. LDL cholesterol concentration was estimated from plasma blood samples using the Friedewald equation for individuals with fasting triglyceride values <400 mg/dL (Friedewald et al., 1972).
2.5. Covariates
Covariates included individual-level sociodemographics (age, sex, study center, education, marital status), self-reported health behaviors (alcohol use, smoking, physical activity, diet), other diabetes risk factors (parental history of diabetes, waist circumference, depressive symptoms), and census tract characteristics (poverty, population density). Details of how these covariates were measured are provided in Supplemental Table 1.
2.6. Statistical Analysis
Missing data based on exam non-attendance and item non-response are shown in Supplemental Table 2. Participants who did not attend later exams (15–30) were slightly younger on average, more likely to be male, smoke, have higher baseline physical activity, and to have a high school degree or less compared to participants who did attend. They did not differ based on segregation or discrimination exposures, or based on other baseline covariates. We imputed missing covariate values by using Markov chain Monte Carlo multiple imputation to create 10 imputed datasets. Subsequent regression coefficients were pooled across imputed datasets using SAS MI Analyze (SAS version 9.4).
To estimate associations of segregation and racial discrimination with incident diabetes, we used multivariable Cox proportional hazards regression. Time at risk was defined as time from the baseline exam until the examination at which the participant first met the definition for diabetes or the last attended exam (administrative censoring). We ran two sets of models: one estimating associations of baseline segregation/discrimination and adjusting for baseline values of the covariates listed above, and a second estimating associations of time-dependent segregation/discrimination and adjusting for time-dependent covariate values updated at each exam. For baseline models, we tested the proportional hazards assumption by adding interaction terms reflecting the product of segregation and discrimination, respectively, and the natural log of time at risk. We detected no violations of the proportional hazards assumption. In a sensitivity analysis, we calculated time at risk starting at the year 7 exam for the racial discrimination models, as discrimination was first assessed in year 7.
Models were adjusted progressively as follows: Model 1: unadjusted; Model 2: adjusted for individual sociodemographics (baseline age, sex, study center, parental history of diabetes, education, and marital status); Model 3: additionally adjusted for individual health factors (smoking status, alcohol use, physical activity, diet score waist circumference, and depressive symptoms); Model 4 additionally adjusted for neighborhood-level covariates (census tract poverty and population density).
Among the subset of participants who developed diabetes over follow-up, attended the year 20 or 25 exam and reported they had been diagnosed with diabetes, we assessed associations of time-varying segregation and discrimination with diabetes management, defined as meeting pre-specified targets for management of HbA1c, SBP, LDL cholesterol, or all 3 in exam years 20 and 25. We used multivariable generalized estimating equations (GEE) logistic regression with an exchangeable correlation structure and clustering by participant to compare the odds of meeting diabetes management targets (versus not meeting targets) by time-varying segregation and discrimination categories. Models were adjusted for exam year and progressively adjusted for covariates as described above. Diabetes management models also adjusted for diabetes duration (time in years since diabetes was first identified) and indicators for whether participants were on diabetes medications, antihypertensive medications, and cholesterol-lowering medications at each exam. We then used multivariable GEE linear regression models to estimate associations with continuous values of HbA1c, SBP, and LDL cholesterol.
3. Results
Description of Study Population
Among 2175 black participants, 1774 (82%) lived in highly segregated census tracts, 262 (12%) lived in medium segregation tracts, and 139 (6%) lived in low segregation tracts at baseline (Table 1). Over 40% of participants (n=934) reported experiencing racial discrimination in ≥3 areas at baseline. The percent of participants living in highly segregated tracts declined over time to 55%, and the percent reporting racial discrimination in ≥3 areas declined to 37% (Supplemental Table 3).
Table 1.
Distribution of Baseline Characteristics by Baseline Racial Residential Segregation and Racial Discrimination Categories
| Baseline Characteristics | Baseline Segregation Category1 | Experiences of Racial Discrimination2 | ||||
|---|---|---|---|---|---|---|
| High % | Medium % | Low % | None % | 1–2 % | 3+ % | |
| N | 1774 | 262 | 139 | 345 | 409 | 934 |
| Mean follow-up time (years) | 21.5 | 21.7 | 21.1 | 21.2 | 21.4 | 21.4 |
| N incident diabetes events | 364 | 55 | 23 | 74 | 93 | 200 |
| Percent developing incident diabetes | 20.5% | 21.0% | 16.6% | 21.5% | 18.7% | 21.4% |
| Age- Mean (SD) | 24.3 (3.9) | 24.9 (3.6) | 24.8 (3.3) | 24.2 (3.8) | 24.2 (3.8) | 24.6 (3.8) |
| Female gender | 57.7% | 55.7% | 54.0% | 66.4% | 58.4% | 53.0% |
| Field center: | ||||||
| Birmingham, AL | 26.1% | 26.7% | 10.1% | 30.5% | 28.5% | 21.0% |
| Chicago, IL | 20.3% | 14.5% | 40.3% | 24.9% | 20.8% | 19.4% |
| Minneapolis, MN | 22.9% | 20.6% | 15.1% | 15.4% | 18.7% | 26.9% |
| Oakland, CA | 30.7% | 38.2% | 34.5% | 29.2% | 32.0% | 32.7% |
| Parental history of diabetes | 17.4% | 13.0% | 18.0% | 16.0% | 18.0% | 16.5% |
| Married/married-like relationship | 28.1% | 29.4% | 33.8% | 26.9% | 25.8% | 31.1% |
| Educational attainment: | ||||||
| <High school | 14.2% | 8.0% | 8.6% | 17.6% | 11.6% | 12.3% |
| High school diploma | 38.2% | 33.2% | 29.5% | 41.6% | 40.6% | 33.1% |
| Some college or more | 47.6% | 58.8% | 61.9% | 40.8% | 47.8% | 54.6% |
| Current smoker | 34.3% | 24.0% | 27.9% | 33.2% | 30.2% | 34.0% |
| Alcohol use3 | ||||||
| None | 47.6% | 50.6% | 38.9% | 51.3% | 49.1% | 44.9% |
| Moderate | 22.5% | 20.7% | 30.9% | 22.3% | 23.8% | 22.3% |
| Heavy | 29.9% | 28.7% | 30.2% | 26.4% | 28.1% | 32.8% |
| Total physical activity score- Mean (SD)4 | 377.0 (305.5) | 421.6 (292.2) | 441.3 (312.3) | 339.9 (286.8) | 352.0 (270.5) | 425.9 (326.6) |
| Alternative healthy eating index score- Mean (SD)5 | 30.2 (11.0) | 31.7 (11.7) | 30.7 (11.1) | 29.0 (11.1) | 29.8 (11.0) | 31.4 (11.1) |
| Waist circumference (cm) - Mean (SD) | 78.4 (12.1) | 77.9 (11.7) | 77.4 (11.0) | 78.0 (12.7) | 78.4 (12.4) | 78.3 (11.4) |
| Census tract poverty6- Mean (SD) | 25.8 (12.6) | 17.1 (10.0) | 11.7 (10.0) | 25.5 (13.2) | 24.2 (12.9) | 23.0 (12.6) |
| Tract population density- Mean (SD) | 13410.1 (10707.0) | 11747.3 (8546.8) | 15635.3 (16731.6) | 13469.4 (10753.8) | 13144.9 (10518.4) | 13439.9 (11368.4) |
Racial residential segregation categorized into high, medium, and low use the value of the local Gi* statistic, which measures the deviation of the census tract’s racial composition from that of the larger metropolitan area. A Gi* statistic z-score <0 was categorized as low, 0 to 1.96 as medium, and >1.96 as high segregation.
Racial discrimination was assessed via survey. Participants reported whether they had experienced discrimination in any of 7 domains (e.g. school, getting a job, work, getting housing, getting medical care, on the street or in a public setting, from the police or in the courts). Discrimination was categorized as no experiences, 1–2 experiences, or 3+ experiences. Racial discrimination was first assessed in year 7, so this value is used for baseline discrimination. As such, baseline discrimination was missing for 398 participants.
Alcohol use was categorized as usually consuming: 0 drinks per week (none), 1–7 drinks per week for women or 1–14 drinks per week for men (moderate), or >7 drinks per week for women or >14 drinks per week for men (heavy).
Total physical activity was measured using the CARDIA physical activity questionnaire, which assessed the frequency of participation in the past 12 months in 13 moderate and vigorous intensity activities. A total physical activity score was calculated by multiplying the frequency of participation by intensity of the activity and summing across activities to create a continuous exercise unit score.
Dietary quality was assessed in years 0, 7 and 20 using a diet history questionnaire on dietary intake for the previous 28 days. Diet quality was quantified using the Alternative Healthy Eating Index, which includes vegetables, fruits, whole grains, nuts/legumes, red/processed meat, sugar-sweetened beverages/fruit juice, trans fat, long-chain fats, polyunsaturated fatty acids, sodium, and alcohol consumption.
Census tract poverty was the percent of the participant’s residential census tract population living below the federal poverty threshold (range 0–100), using U.S. Census data
Table 1 presents the distribution of diabetes events and baseline covariates by categories of baseline racial residential segregation and baseline racial discrimination experiences. Over a mean follow-up of 21.5 years, 442 cases of incident diabetes were identified in the overall study population. Participants living in low segregation neighborhoods and those who reported perceived racial discrimination in 1–2 areas were least likely to develop diabetes (Table 1).
Diabetes Incidence Models: Baseline Segregation and Discrimination as Exposures
In unadjusted models, hazard ratios (HRs) for incident diabetes comparing participants in medium and low segregation neighborhoods to those in high segregation neighborhoods were 0.94 (95% CI: 0.71, 1.25) and 0.79 (0.52, 1.21), respectively. Point estimates moved toward the null with progressive covariate adjustment (Table 2). For baseline racial discrimination, HRs for incident diabetes among participants who reported 1–2 and 3+ experiences of racial discrimination versus no experiences were 0.95 (0.70, 1.30) and 1.03 (0.79, 1.35), respectively. Progressive covariate adjustment minimally changed results (Table 2).
Table 2.
Associations of Racial Residential Segregation1 and Racial Discrimination2 with Incident Diabetes3,4
| HR for Incident Diabetes (95% CI) | |||||
| Model Type | Segregation Category | Model 1 | Model 2 | Model 3 | Model 4 |
| Baseline Values | High | Reference | Reference | Reference | Reference |
| Medium | 0.94 (0.71, 1.25) | 0.97 (0.72, 1.30) | 0.96 (0.71, 1.29) | 1.01 (0.75, 1.37) | |
| Low | 0.79 (0.52, 1.21) | 0.75 (0.49, 1.15) | 0.85 (0.55, 1.31) | 0.93 (0.59, 1.46) | |
| Time-Varying Values | High | Reference | Reference | Reference | Reference |
| Medium | 0.75 (0.59, 0.97) | 0.79 (0.61, 1.02) | 1.04 (0.80, 1.35) | 0.97 (0.74, 1.28) | |
| Low | 0.97 (0.76, 1.22) | 0.98 (0.77, 1.24) | 1.20 (0.94, 1.53) | 1.09 (0.82, 1.43) | |
| HR for Incident Diabetes (95% CI) | |||||
| Model Type | Discrimination Category | Model 1 | Model 2 | Model 3 | Model 3 |
| Baseline Values | No Experiences | Reference | Reference | Reference | Reference |
| 1–2 Experiences | 0.95 (0.70, 1.30) | 0.94 (0.69, 1.29) | 0.96 (0.69, 1.33) | 0.96 (0.69, 1.34) | |
| 3+ Experiences | 1.03 (0.79, 1.35) | 1.00 (0.77, 1.31) | 1.06 (0.81, 1.39) | 1.07 (0.82, 1.41) | |
| Time-Varying Values | No Experiences | Reference | Reference | Reference | Reference |
| 1–2 Experiences | 0.97 (0.74, 1.27) | 0.97 (0.74, 1.28) | 0.89 (0.68, 1.17) | 0.91 (0.69, 1.19) | |
| 3+ Experiences | 0.92 (0.72, 1.19) | 0.94 (0.73, 1.21) | 0.87 (0.66, 1.14) | 0.87 (0.67, 1.14) | |
HR: hazard ratio
Racial residential segregation categorized into high, medium, and low use the value of the local Gi* statistic, which measures the deviation of the census tract’s racial composition from that of the larger metropolitan area. A Gi* statistic z-score <0 was categorized as low, 0 to 1.96 as medium, and >1.96 as high segregation.
Racial discrimination was assessed via survey. Participants reported whether they had experienced discrimination in any of 7 domains (e.g. school, getting a job, work, getting housing, getting medical care, on the street or in a public setting, from the police or in the courts). Discrimination was categorized as no experiences, 1–2 experiences, or 3+ experiences.
Diabetes defined as fasting glucose ≥ 126 mg/dL, self-reported use of medications for the treatment of diabetes mellitus, 2-hour glucose during the oral glucose tolerance test of ≥ 200 mg/dL, or HbA1c ≥ 6.5%.
Results are from Cox proportional hazards regression models and include the entire study population (n=2175). Baseline value models estimated associations for baseline segregation category (or baseline racial discrimination category) and adjusted for baseline covariates only. Time-varying value models estimated associations for time-varying segregation category (or time-varying racial discrimination) and adjusted for time-varying covariates. Model 1 was unadjusted. Model 2 adjusted for baseline age, sex, study center, parental history of diabetes, education, and marital status; Model 3 additionally adjusted for smoking status, alcohol use, physical activity, diet score, waist circumference, and depressive symptoms. Model 4 additionally adjusted for census tract poverty and population density.
Diabetes Incidence Models: Time-varying Segregation and Discrimination as Exposures
In unadjusted models with time-varying exposures, HRs for incident diabetes comparing participants in medium and low segregation neighborhoods to those in high segregation neighborhoods were 0.75 (0.59, 0.97) and 0.97 (0.76 1.22), respectively. Adjustment for sociodemographic covariates slightly attenuated the estimate for medium segregation (HR: 0.79, CI: 0.61, 1.02), and adjustment for individual-level health factors and neighborhood-level covariates further attenuated the association (Table 2). For time-varying discrimination, HRs for incident diabetes among participants who reported 1–2 and 3+ experiences of racial discrimination versus no experiences were 0.97 (0.74, 1.27) and 0.92 (0.72, 1.19), respectively. Results changed little upon adjustment for covariates. Results were similar in a sensitivity analysis where we calculated time at risk starting in year 7 instead of year 0 for the discrimination models (Supplemental Table 4).
Diabetes Management Models
Among the 442 participants who developed diabetes based on our clinical definition during the study period, 321 developed diabetes by year 25 and attended either the year 20 or year 25 exam. Of these participants, 208 self-reported that a doctor or nurse had diagnosed them with diabetes, and were included in the diabetes management analysis (305 exam-years included in total). Table 3 presents the distribution of diabetes management measures (HbA1c, SBP, and LDL) and the percent of participants meeting targets for each measure, overall and by segregation and discrimination categories. Patterns varied by outcome and exposure and were not consistent between the two exams. For example, in year 25, participants living in low segregation neighborhoods were more likely to meet the SBP target of <130 mmHg (80.7% compared to 70.6% among participants in medium segregation neighborhoods and 70.7% among participants in high segregation neighborhoods). However, this pattern was not observed in exam year 20 (Table 3). In contrast to our hypothesis, participants living in low segregation neighborhoods in year 20 were the least likely to meet targets for all three measures (5.2% compared to 23.2% in medium and 20.6% in high segregation neighborhoods), although the percent meeting targets was similar between groups in year 25. Participants experience 1–2 forms of racial discrimination were least likely to meet all three targets in year 20, but most likely to meet all three targets in year 25 (Table 3).
Table 3.
Diabetes Management Among Participants After Diagnosis by Racial Residential Segregation and Racial Discrimination Category1,2,3,4
| Year 20 | Year 25 | |||
| N participants who developed diabetes by exam5 | 226 | 338 | ||
| N (%) participants with diabetes attending exam | 202 | 291 | ||
| N (%) participants with diabetes attending exam who self-reported a diabetes diagnosis6 | 124 | 181 | ||
| Mean (SD) | % Controlled | Mean (SD) | % Controlled | |
| Denominator for % Controlled | 124 | 181 | ||
| HbA1c (<7.0% defined as controlled) | ||||
| Overall | 7.7 (1.7) | 38.4% | 7.7 (1.8) | 45.6% |
| By Segregation Category: | ||||
| Low | 8.2 (1.6) | 23.7% | 7.6 (1.5) | 41.3% |
| Medium | 7.2 (1.5) | 48.0% | 8.1 (1.9) | 38.5% |
| High | 7.7 (1.8) | 40.6% | 7.6 (1.8) | 48.8% |
| By Racial Discrimination Category: | ||||
| No experiences | 7.8 (1.6) | 33.1% | 7.6 (1.7) | 48.2% |
| 1–2 experiences | 7.6 (1.6) | 40.5% | 7.6 (1.8) | 44.5% |
| 3+ experiences | 7.7 (1.9) | 39.9% | 7.8 (1.8) | 43.5% |
| SBP (<130 mmHg defined as controlled) | ||||
| Overall | 120.1 (15.7) | 78.2% | 123.6 (16.9) | 72.4% |
| By Segregation Category: | ||||
| Low | 120.2 (14.5) | 77.8% | 120.1 (12.3) | 80.7% |
| Medium | 116.7 (14.9) | 76.0% | 125.0 (21.1) | 70.6% |
| High | 121.2 (16.3) | 79.2% | 124.2 (16.6) | 70.7% |
| By Racial Discrimination Category: | ||||
| No experiences | 120.1 (13.4) | 83.0% | 124.4 (18.0) | 71.6% |
| 1–2 experiences | 123.4 (18.1) | 66.2% | 122.4 (15.8) | 73.3% |
| 3+ experiences | 117.4 (14.3) | 85.3% | 123.6 (16.6) | 72.6% |
| LDL cholesterol (<100 mg/dL defined as controlled) | ||||
| Overall | 102.9 (33.6) | 52.8% | 97.3 (32.8) | 55.9% |
| By Segregation Category: | ||||
| Low | 97.7 (25.1) | 52.6% | 95.0 (33.0) | 60.0% |
| Medium | 100.0 (37.4) | 47.2% | 91.3 (32.4) | 60.9% |
| High | 105.9 (34.8) | 54.9% | 99.7 (32.6) | 53.4% |
| By Racial Discrimination Category: | ||||
| No experiences | 98.5 (28.7) | 54.0% | 100.1 (33.1) | 50.6% |
| 1–2 experiences | 111.5 (38.2) | 42.6% | 95.7 (29.6) | 59.2% |
| 3+ experiences | 98.6 (31.0) | 60.5% | 95.5 (34.3) | 59.3% |
| All 3 controlled | ||||
| Overall | -- | 17.7% | -- | 21.2% |
| By Segregation Category: | ||||
| Low | -- | 5.2% | -- | 22.6% |
| Medium | -- | 23.2% | -- | 23.8% |
| High | -- | 20.6% | -- | 20.0% |
| By Racial Discrimination Category: | ||||
| No experiences | -- | 21.9% | -- | 17.9% |
| 1–2 experiences | -- | 10.7% | -- | 27.6% |
| 3+ experiences | -- | 21.0% | -- | 20.2% |
SD: standard deviation; SBP: systolic blood pressure; LDL: low density lipoprotein
Control defined as follows: HbA1c <7.0%, SBP <130 mm Hg, LDL <100 mg/dL.
Missing control values were imputed, so percentages reflect the total percent across all 10 imputed datasets. Mean values reflect averages across all 10 imputed datasets.
Racial residential segregation categorized into high, medium, and low use the value of the local Gi* statistic, which measures the deviation of the census tract’s racial composition from that of the larger metropolitan area. A Gi* statistic z-score <0 was categorized as low, 0 to 1.96 as medium, and >1.96 as high segregation.
Racial discrimination was assessed via survey. Participants reported whether they had experienced discrimination in any of 7 domains (e.g. school, getting a job, work, getting housing, getting medical care, on the street or in a public setting, from the police or in the courts). Discrimination was categorized as no experiences, 1–2 experiences, or 3+ experiences.
”Developing diabetes” was based on our definition of meeting any of the following criteria: fasting glucose ≥126 mg/dL, 2-hour glucose during the oral glucose tolerance test of ≥200 mg/dL, hemoglobin A1c ≥6.5%, or current use of medication for the treatment of diabetes mellitus. 442 participants developed diabetes during the follow-up period, and 321 of them developed diabetes by the year 25 exam and attended either the year 20 or year 25 exam.
Based on whether or not participants self-reported that a doctor or nurse ever said that they had diabetes. We conducted this analysis only among the subset who self-reported a diabetes diagnosis at the time of the exam in order to focus on participants who were aware they had diabetes prior to the exam. In total, 208 participants who developed diabetes over follow-up self-reported a diabetes diagnosis at the year 20 or year 25 exam.
In multivariable GEE regression models (Tables 4–5), patterns were similarly inconsistent, and confidence intervals were generally wide. In contrast to our study hypothesis, participants living in low segregation neighborhoods had lower odds of meeting the HBA1c management target compared to participants in high segregation neighborhoods (odds ratio (OR): 0.49, 95% CI: 0.25, 0.97 after adjustment for individual-level sociodemographics), although subsequent adjustment for neighborhood-level covariates attenuated the association (Table 4).
Table 4.
Associations of Racial Residential Segregation and Racial Discrimination with Dichotomous Measures of Diabetes Management After Diagnosis1,2,3
| OR for Meeting Management Target (95% CI) | |||||
| Outcome | Segregation Category | Model 1 | Model 2 | Model 3 | Model 4 |
| HbA1c controlled (<7.0%) | High | Reference | Reference | Reference | Reference |
| Medium | 0.82 (0.44, 1.53) | 0.84 (0.41, 1.74) | 0.85 (0.40, 1.81) | 1.29 (0.59, 2.81) | |
| Low | 0.59 (0.31, 1.14) | 0.49 (0.25, 0.97) | 0.50 (0.25, 1.00) | 0.87 (0.37, 2.08) | |
| SBP controlled (<130 mm Hg) | High | Reference | Reference | Reference | Reference |
| Medium | 0.99 (0.50, 1.96) | 1.03 (0.50, 2.13) | 1.10 (0.55, 2.21) | 1.16 (0.54, 2.47) | |
| Low | 1.20 (0.58, 2.50) | 1.07 (0.50, 2.31) | 1.01 (0.44, 2.34) | 1.15 (0.42, 3.14) | |
| LDL controlled (<100 mg/dL) | High | Reference | Reference | Reference | Reference |
| Medium | 1.10 (0.63, 1.91) | 0.96 (0.54, 1.71) | 1.00 (0.56, 1.78) | 0.91 (0.50, 1.63) | |
| Low | 1.15 (0.65, 2.04) | 1.14 (0.63, 2.07) | 1.12 (0.59, 2.13) | 1.02 (0.48, 2.13) | |
| All 3 controlled | High | Reference | Reference | Reference | Reference |
| Medium | 1.19 (0.59, 2.41) | 1.34 (0.65, 2.78) | 1.38 (0.65, 2.94) | 1.51 (0.66. 3.45) | |
| Low | 0.73 (0.32, 1.63) | 0.57 (0.25, 1.28) | 0.61 (0.27, 1.38) | 0.68 (0.25, 1.84) | |
| OR for Meeting Management Target (95% CI) | |||||
| Outcome | Racial Discrimination Category | Model 1 | Model 2 | Model 3 | Model 4 |
| HbA1c controlled (<7.0%) | No Experiences | Reference | Reference | Reference | Reference |
| 1–2 Experiences | 0.97 (0.54, 1.74) | 0.95 (0.48, 1.86) | 0.96 (0.48, 1.95) | 0.93 (0.45, 1.93) | |
| 3+ Experiences | 0.98 (0.55, 1.76) | 0.91 (0.46, 1.78) | 0.95 (0.48, 1.88) | 0.90 (0.43, 1.87) | |
| SBP controlled (<130 mm Hg) | No Experiences | Reference | Reference | Reference | Reference |
| 1–2 Experiences | 0.71 (0.36, 1.37) | 0.69 (0.33, 1.42) | 0.60 (0.29, 1.28) | 0.59 (0.28, 1.26) | |
| 3+ Experiences | 1.12 (0.57, 2.18) | 1.24 (0.59, 2.61) | 1.15 (0.53, 2.50) | 1.12 (0.51, 2.43) | |
| LDL controlled (<100 mg/dL) | No Experiences | Reference | Reference | Reference | Reference |
| 1–2 Experiences | 1.02 (0.58, 1.79) | 1.14 (0.63, 2.07) | 1.13 (0.61, 2.07) | 1.09 (0.58, 2.05) | |
| 3+ Experiences | 1.35 (0.75, 2.45) | 1.52 (0.79, 2.91) | 1.58 (0.81, 3.05) | 1.53 (0.79, 2.98) | |
| All 3 controlled | No Experiences | Reference | Reference | Reference | Reference |
| 1–2 Experiences | 1.04 (0.45, 2.42) | 1.04 (0.42, 2.57) | 1.03 (0.40, 2.62) | 1.00 (0.39, 2.58) | |
| 3+ Experiences | 1.11 (0.53, 2.30) | 1.02 (0.46, 2.26) | 1.01 (0.44, 2.28) | 0.98 (0.43, 2.25) | |
OR: odds ratio; SBP: systolic blood pressure; LDL: low density lipoprotein cholesterol
Racial residential segregation categorized into high, medium, and low use the value of the local Gi* statistic, which measures the deviation of the census tract’s racial composition from that of the larger metropolitan area. A Gi* statistic z-score <0 was categorized as low, 0 to 1.96 as medium, and >1.96 as high segregation. Segregation category was treated as time-varying in these models.
Racial discrimination was assessed via survey. Participants reported whether they had experienced discrimination in any of 7 domains (e.g. school, getting a job, work, getting housing, getting medical care, on the street or in a public setting, from the police or in the courts). Discrimination was categorized as no experiences, 1–2 experiences, or 3+ experiences. Discrimination was treated as time-varying in these models.
Results are from generalized estimating equations (GEE) logistic regression models with an exchangeable covariance structure, accounting for clustering by participant across 2 exams (year 20 and year 25). Models included participant exam-years after diabetes development at which participants indicated a medical professional had diagnosed them with diabetes (n=305 exam-years total from 208 participants). Model 1 was unadjusted. Model 2 adjusted for baseline age, sex, field center, parental history of diabetes, time-varying education, marital status, diabetes duration, and use of diabetes medication, antihypertensive medication, and cholesterol-lowering medication. Model 3 additionally adjusted for time-varying smoking status, alcohol use, physical activity, diet score, waist circumference, and depressive symptoms. Model 4 additionally adjusted for census tract poverty and population density.
Table 5.
Associations of Racial Residential Segregation and Racial Discrimination with Continuous Measures of Diabetes Management After Diagnosis1,2,3
| Mean Difference (95% CI) | |||||
| Outcome | Segregation Category | Model 1 | Model 2 | Model 3 | Model 4 |
| Continuous HbA1c (%) | High | Reference | Reference | Reference | Reference |
| Medium | 0.21 (−0.32, 0.74) | 0.27 (−0.26, 0.79) | 0.23 (−0.27, 0.73) | 0.06 (−0.50, 0.62) | |
| Low | 0.25 (−0.31, 0.81) | 0.22 (−0.35, 0.79) | 0.18 (−0.37, 0.74) | −0.09 (−0.75, 0.58) | |
| Continuous SBP (in mmHg) | High | Reference | Reference | Reference | Reference |
| Medium | −0.54 (−5.13, 4.06) | −0.64 (−5.79, 4.51) | −0.34 (−6.04, 5.36) | −0.60 (−7.67, 6.48) | |
| Low | −0.87 (−5.64, 4.00) | −0.10 (−5.04, 4.85) | 1.05 (−7.17, 7.28) | 1.76 (−7.13, 10.66) | |
| Continuous LDL cholesterol (in mg/dL) | High | Reference | Reference | Reference | Reference |
| Medium | −9.93 (−19.03, −0.83) | −8.63 (−17.65, 0.38) | −9.41 (−18.40, −0.41) | −8.15 (−18.28, 1.99) | |
| Low | −6.86 (−16.36, 2.64) | −5.45 (−16.00, 5.09) | −6.00 (−16.85, 4.84) | −7.71 (−22.28, 6.87) | |
| Mean Difference (95% CI) | |||||
| Outcome | Racial Discrimination Category | Model 1 | Model 2 | Model 3 | Model 4 |
| Continuous HbA1c (%) | No Experiences | Reference | Reference | Reference | Reference |
| 1–2 Experiences | −0.11 (−0.64, 0.43) | −0.10 (−0.63, 0.43) | −0.12 (−0.63, 0.39) | −0.11 (−0.63, 0.41) | |
| 3+ Experiences | 0.09 (−0.42, 0.61) | 0.19 (−0.30, 0.69) | 0.17 (−0.30, 0.64) | 0.19 (−0.29, 0.67) | |
| Continuous SBP (in mmHg) | No Experiences | Reference | Reference | Reference | Reference |
| 1–2 Experiences | 0.98 (−3.50, 5.47) | 1.43 (−3.09, 5.96) | 3.81 (−3.13, 10.76) | 5.52 (−1.93, 12.97) | |
| 3+ Experiences | −0.95 (−5.15, 3.26) | −1.03 (−5.41, 3.34) | 2.02 (−5.13, 9.18) | 4.16 (−3.56, 11.88) | |
| Continuous LDL cholesterol (in mg/dL) | No Experiences | Reference | Reference | Reference | Reference |
| 1–2 Experiences | 3.54 (−4.82, 11.90) | 2.13 (−5.91, 10.17) | 2.87 (−5.35, 11.09) | 4.73 (−6.26, 15.71) | |
| 3+ Experiences | 0.24 (−8.72, 9.21) | 0.86 (−7.97, 9.69) | 1.18 (−8.24, 10.60) | 4.39 (−11.88, 20.66) | |
SBP: systolic blood pressure; LDL: low density lipoprotein
Racial residential segregation categorized into high, medium, and low use the value of the local Gi* statistic, which measures the deviation of the census tract’s racial composition from that of the larger metropolitan area. A Gi* statistic z-score <0 was categorized as low, 0 to 1.96 as medium, and >1.96 as high segregation. Segregation category was treated as time-varying in these models.
Racial discrimination was assessed via survey. Participants reported whether they had experienced discrimination in any of 7 domains (e.g. school, getting a job, work, getting housing, getting medical care, on the street or in a public setting, from the police or in the courts). Discrimination was categorized as no experiences, 1–2 experiences, or 3+ experiences. Discrimination was treated as time-varying in these models.
Results are from generalized estimating equations (GEE) linear regression models with an exchangeable covariance structure, accounting for clustering by participant across 2 exams (year 20 and year 25). Models included participant exam-years after diabetes development at which participants indicated a medical professional had diagnosed them with diabetes (n=305 exam-years total from 208 participants). Model 1 was unadjusted. Model 2 adjusted for baseline age, sex, field center, parental history of diabetes, time-varying education, marital status, diabetes duration, and use of diabetes medication, antihypertensive medication, and cholesterol-lowering medication. Model 3 additionally adjusted for time-varying smoking status, alcohol use, physical activity, diet score, waist circumference, and depressive symptoms. Model 4 additionally adjusted for census tract poverty and population density.
Participants living in medium segregation neighborhoods had lower levels of LDL cholesterol compared to participants in high segregation neighborhoods, with a mean difference of −9.41 mg/dL (−18.40, −0.41) after adjustment for individual-level covariates. This difference was reduced to −8.15 mg/dL (−18.28, 1.99) after subsequent adjustment for neighborhood-level covariates. Patterns were similar for participants in low segregation neighborhoods, although confidence intervals were wide and crossed the null (Table 5). Point estimates were in the hypothesized direction for racial discrimination and SBP and LDL, with higher mean values among participants reporting experiences of racial discrimination, but confidence intervals were wide- for example, a relative difference in SBP of 5.52 mmHg (−1.93, 12.97) between participants reporting 1–2 versus 0 experience of discrimination in fully adjusted models).
Results were largely similar in a sensitivity analysis in which we did not restrict to participants who self-reported a diabetes diagnosis (n=321 participants, Supplemental Tables 5–6). We found no evidence of interaction by sex for either diabetes incidence or management (p-interaction >0.1 for all models).
4. Discussion
Given persistent racial disparities in diabetes outcomes (Menke et al., 2015), identification of the environmental determinants of diabetes among non-Hispanic black adults is needed. In this longitudinal cohort of young to middle-aged black U.S. adults, racial residential segregation and experiences of interpersonal racial discrimination were not associated with diabetes incidence after adjustment for individual- and neighborhood-level covariates. Consistent with our findings, a recent review of 6 prior studies on segregation and diabetes found limited evidence supporting a relationship between residential segregation and diabetes prevalence (Kershaw and Pender, 2016). However, two prior studies found higher diabetes mortality rates in areas with a higher percentage of black residents (Hunt et al., 2014; Rosenstock et al., 2014). These contrasting results in the literature led us to examine whether segregation was associated with diabetes management after diagnosis. We hypothesized that segregation would negatively influence participants’ ability to achieve target levels of HbA2c, SBP, and LDL cholesterol after diabetes diagnosis as a result of the adverse environmental consequences of segregation (e.g. limiting access to health promoting resources and increasing exposure to stressors like crime). However, we found limited evidence of an association between either segregation or racial discrimination and diabetes management.
The inconsistent results previously reported for segregation and diabetes may be a result of the use of varied measures of segregation in past studies. For example, of the three prior cross-sectional examinations of segregation and diabetes prevalence, two defined segregation based on census tract racial composition (percent black) (Piccolo et al., 2015; Gaskin et al., 2014), a crude measures that does not account for how individuals are distributed in space (Kramer and Hogue, 2009). A third study, the only one reporting higher diabetes prevalence among blacks in more segregated areas, used a spatial racial isolation index that accounted for racial composition within an index block as well as neighboring blocks (Bravo et al., 2018). The two ecological studies examining diabetes mortality defined segregation using neighborhood-level racial composition (Hunt et al., 2014) and a city-level isolation index, respectively (Rosenstock et al., 2014). The variation in how segregation is measured makes it difficult to directly compare between studies. While segregation is a multi-dimensional construct (Massey and Denton, 1988), we chose to use the Gi*statistic because this approach, by comparing the racial composition of a census tract and its neighbors to that of the larger metropolitan area, better reflects the spatial and contextual aspects of segregation. The effect of segregation on socioeconomic opportunity and environmental exposures among residents may differ depending on how a neighborhood’s composition compares to the larger area in which the neighborhood is situated.
Our finding that experiences of interpersonal racial discrimination were not associated with diabetes incidence is in contrast to results from the Black Women’s Health Study, in which perceived discrimination was associated with greater risk of incident diabetes (Bacon et al., 2017). In the Multi-Ethnic Study of Atherosclerosis, perceived discrimination was associated with incident diabetes in the full population of 5,310 black, white, Hispanic, and Chinese participants (Whitaker et al., 2017). However, in race-stratified results that focused on experiences of discrimination attributed to race/ethnicity in particular, perceived racial discrimination was not associated with incident diabetes among black participants (Whitaker et al., 2017), which is consistent with our findings. Both studies assessed experiences of discrimination using a similar approach to CARDIA, by assessing whether participants had ever experienced discrimination across a range of situations, although the specific set of situations differed between the three studies (Bacon et al., 2017; Krieger and Sidney, 1996; Whitaker et al., 2017).
No prior studies have examined associations of segregation with diabetes management among people that have been diagnosed with type 2 diabetes (Kershaw and Pender, 2016). Findings have been mixed regarding associations between interpersonal racial discrimination and diabetes management among black participants. For example, frequency of racial discrimination in the past 12 months was associated with higher SBP among black diabetes patients in the southeastern United States, but not with HbA1c or LDL (Dawson et al., 2015). Racial discrimination was also not associated with HbA1c among black women with diabetes in Connecticut (Wagner et al., 2013), and was associated with higher levels of HbA1c only among black men in a study of 163 black diabetes patients from the midwestern United States (Assari et al., 2017). Our results do not support an association of either interpersonal or structural racial discrimination with diabetes management, and we did not detect evidence of an interaction by sex. It should be noted that only 208 participants were available for this analysis based on data limitations for HbA1c, loss to follow-up, and because many participants who met our clinical definition for diabetes were unaware that they had diabetes prior to the exam. This limited our sample size and potentially our ability to detect associations- for example, between interpersonal racial discrimination and systolic blood pressure. Past studies have also generally had small sample sizes (range 77–391 participants), and typically enrolled patients from a single geographic area (Assari et al., 2017; Dawson et al., 2015; Wagner et al., 2013). CARDIA participants were enrolled from 4 urban areas in different regions of the U.S. and lived in all 50 states by year 25 (Funkhouser et al., 2018). However, the small sample size included in the diabetes management analysis may limit generalizability.
This study had several limitations in addition to the small sample size available for the diabetes management analysis. First, not all components of our diabetes definition (e.g. oral glucose tolerance test or HbA1C) were available at every exam, which may have led to misclassification of participants with underlying diabetes that was not detected using fasting glucose or medication use alone. Second, while we controlled for neighborhood characteristics and a variety of individual-level risk factors, there is still the potential for residual confounding due to unmeasured factors. Third, over 80% of participants lived in high segregation census tracts at baseline. While segregation levels varied to a greater extent in later years after some participants moved, there may not have been sufficient variability in segregation to observe an association with diabetes incidence or management. Future work is needed in other populations. In addition, it is possible that racial residential segregation and racial discrimination are more relevant to other outcomes such as diabetes mortality or complications. We were unable to examine these outcomes due to the young age of the study population (limiting assessment of cause-specific mortality) and due to lack of availability on diabetes complications (e.g. retinopathy). However, assessing the associations of segregation and discrimination with these outcomes in longitudinal studies is an important area for future research. Finally, as noted above, the small number of participants included in the diabetes management analysis may limit generalizability of our findings.
4.1. Conclusion
We found limited evidence of an association between racial residential segregation, interpersonal racial discrimination, and diabetes incidence or management among a cohort of black young to middle-aged U.S. adults. Areas for future research include longitudinal studies investigating whether racial residential segregation is associated with diabetes mortality and complications.
Supplementary Material
Highlights.
The link between racial segregation or discrimination and diabetes is understudied
We assessed associations with diabetes incidence and management in black adults
Racial segregation was not associated with diabetes incidence or management
Neither were experiences of interpersonal racial discrimination
Acknowledgements
The Coronary Artery Risk Development in Young Adults (CARDIA) study is supported by contracts HHSN268201300025C, HHSN268201300026C, HHSN268201300027C, HHSN268201300028C, HHSN268201300029C, and HHSN268200900041C from the National Heart, Lung, and Blood Institute (NHLBI); the Intramural Research Program of the National Institute on Aging (NIA); and intra-agency agreement AG0005 between the NIA and NHLBI. The neighborhood measures used in this research were developed with grants R01HL104580 and R01HL114091 from the NHLBI. SLM and MPB were supported by the National Heart, Lung, and Blood Institute (NHLBI) of the National Institutes of Health under Award Number T32HL069771 to conduct the current work. NIH had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript. This manuscript has been reviewed by CARDIA for scientific content. Data are available from the CARDIA Study, whose authors and paper proposal forms may be contacted at: http://www.cardia.dopm.uab.edu/. The contact person for paper proposals is Linda Sellers: lsellers@uabmc.edu. This dataset is not able to be provided as the data are controlled by the CARDIA investigators to ensure that there are no overlapping CARDIA paper and presentation publications. Any author who wishes to obtain the data may do so after having a manuscript proposal approved by the CARDIA steering committee.
Footnotes
Declarations of Interest: None
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- American Diabetes Association, 2018. Standards of Medical Care in Diabetes-2018 Abridged for Primary Care Providers. Clin Diabetes 36, 14–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Assari S, Lee DB, Nicklett EJ, Moghani Lankarani M, Piette JD, Aikens JE, 2017. Racial Discrimination in Health Care Is Associated with Worse Glycemic Control among Black Men but Not Black Women with Type 2 Diabetes. Front Public Health 5, 235. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bacon KL, Stuver SO, Cozier YC, Palmer JR, Rosenberg L, Ruiz-Narvaez EA, 2017. Perceived racism and incident diabetes in the Black Women’s Health Study. Diabetologia 60, 2221–2225. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bancks MP, Kershaw K, Carson AP, Gordon-Larsen P, Schreiner PJ, Carnethon MR, 2017. Association of Modifiable Risk Factors in Young Adulthood With Racial Disparity in Incident Type 2 Diabetes During Middle Adulthood. Jama 318, 2457–2465. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Borrell LN, Diez Roux AV, Jacobs DR Jr., Shea S, Jackson SA, Shrager S, Blumenthal RS, 2010. Perceived racial/ethnic discrimination, smoking and alcohol consumption in the Multi-Ethnic Study of Atherosclerosis (MESA). Preventive Medicine 51, 307–312. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bower KM, Thorpe RJ Jr., Yenokyan G, McGinty EE, Dubay L, Gaskin DJ, 2015. Racial Residential Segregation and Disparities in Obesity among Women. J Urban Health 92, 843–852. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bravo MA, Anthopolos R, Kimbro RT, Miranda ML, 2018. Residential Racial Isolation and Spatial Patterning of Type 2 Diabetes Mellitus in Durham, North Carolina. Am J Epidemiol. [DOI] [PubMed] [Google Scholar]
- Cunningham TJ, Berkman LF, Kawachi I, Jacobs DR Jr., Seeman TE, Kiefe CI, Gortmaker SL, 2013. Changes in waist circumference and body mass index in the US CARDIA cohort: fixed-effects associations with self-reported experiences of racial/ethnic discrimination. J Biosoc Sci 45, 267–278. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Cunningham TJ, Seeman TE, Kawachi I, Gortmaker SL, Jacobs DR, Kiefe CI, Berkman LF, 2012. Racial/ethnic and gender differences in the association between self-reported experiences of racial/ethnic discrimination and inflammation in the CARDIA cohort of 4 US communities. Soc Sci Med 75, 922–931. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Dawson AZ, Walker RJ, Campbell JA, Egede LE, 2015. Effect of perceived racial discrimination on self-care behaviors, glycemic control, and quality of life in adults with type 2 diabetes. Endocrine 49, 422–428. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Friedewald WT, Levy RI, Fredrickson DS, 1972. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem 18, 499–502. [PubMed] [Google Scholar]
- Friedman GD, Cutter GR, Donahue RP, Hughes GH, Hulley SB, Jacobs DR Jr., Liu K, Savage PJ, 1988. CARDIA: study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol 41, 1105–1116. [DOI] [PubMed] [Google Scholar]
- Funkhouser E, Wammack J, Roche C, Reis J, Sidney S, Schreiner P, 2018. Where are they now? Retention strategies over 25 years in the Coronary Artery Risk Development in Young Adults (CARDIA) Study. Contemp Clin Trials Commun 9, 64–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gaskin DJ, Dinwiddie GY, Chan KS, McCleary RR, 2012. Residential segregation and the availability of primary care physicians. Health Serv Res 47, 2353–2376. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gaskin DJ, Thorpe RJ Jr., McGinty EE, Bower K, Rohde C, Young JH, LaVeist TA, Dubay L, 2014. Disparities in diabetes: the nexus of race, poverty, and place. Am J Public Health 104, 2147–2155. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Geiss LS, Wang J, Cheng YJ, Thompson TJ, Barker L, Li Y, Albright AL, Gregg EW, 2014. Prevalence and incidence trends for diagnosed diabetes among adults aged 20 to 79 years, United States, 1980–2012. Jama 312, 1218–1226. [DOI] [PubMed] [Google Scholar]
- Getis A, Ord JK, 1992. The Analysis of Spatial Association by Use of Distance Statistics. Geographical Analysis 24, 189–206. [Google Scholar]
- Goodman M, Lyons S, Dean LT, Arroyo C, Hipp JA, 2018. How Segregation Makes Us Fat: Food Behaviors and Food Environment as Mediators of the Relationship Between Residential Segregation and Individual Body Mass Index. Front Public Health 6, 92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hunt BR, Whitman S, Henry CA, 2014. Age-adjusted diabetes mortality rates vary in local communities in a metropolitan area: racial and spatial disparities and correlates. Diabetes Care 37, 1279–1286. [DOI] [PubMed] [Google Scholar]
- Ikramuddin S, Korner J, Lee WJ, Connett JE, Inabnet WB, Billington CJ, Thomas AJ, Leslie DB, Chong K, Jeffery RW, Ahmed L, Vella A, Chuang LM, Bessler M, Sarr MG, Swain JM, Laqua P, Jensen MD, Bantle JP, 2013. Roux-en-Y gastric bypass vs intensive medical management for the control of type 2 diabetes, hypertension, and hyperlipidemia: the Diabetes Surgery Study randomized clinical trial. Jama 309, 2240–2249. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Jellinger PS, Handelsman Y, Rosenblit PD, Bloomgarden ZT, Fonseca VA, Garber AJ, Grunberger G, Guerin CK, Bell DSH, Mechanick JI, Pessah-Pollack R, Wyne K, Smith D, Brinton EA, Fazio S, Davidson M, 2017. AMERICAN ASSOCIATION OF CLINICAL ENDOCRINOLOGISTS AND AMERICAN COLLEGE OF ENDOCRINOLOGY GUIDELINES FOR MANAGEMENT OF DYSLIPIDEMIA AND PREVENTION OF CARDIOVASCULAR DISEASE. Endocr Pract 23, 1–87. [DOI] [PubMed] [Google Scholar]
- Kershaw KN, Albrecht SS, 2015. Racial/ethnic residential segregation and cardiovascular disease risk. Curr Cardiovasc Risk Rep 9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kershaw KN, Pender AE, 2016. Racial/Ethnic Residential Segregation, Obesity, and Diabetes Mellitus. Curr Diab Rep 16, 108. [DOI] [PubMed] [Google Scholar]
- Kershaw KN, Robinson WR, Gordon-Larsen P, Hicken MT, Goff DC Jr., Carnethon MR, Kiefe CI, Sidney S, Diez Roux AV, 2017. Association of Changes in Neighborhood-Level Racial Residential Segregation With Changes in Blood Pressure Among Black Adults: The CARDIA Study. JAMA Intern Med 177, 996–1002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kramer MR, Hogue CR, 2009. Is Segregation Bad for Your Health? Epidemiologic Reviews 31, 178–194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Krieger N, Sidney S, 1996. Racial discrimination and blood pressure: the CARDIA Study of young black and white adults. Am J Public Health 86, 1370–1378. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Krieger N, Sidney S, Coakley E, 1998. Racial discrimination and skin color in the CARDIA study: implications for public health research. Coronary Artery Risk Development in Young Adults. Am J Public Health 88, 1308–1313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Krieger N, Smith K, Naishadham D, Hartman C, Barbeau EM, 2005. Experiences of discrimination: validity and reliability of a self-report measure for population health research on racism and health. Social Science & Medicine 61, 1576–1596. [DOI] [PubMed] [Google Scholar]
- Massey DS, Denton NA, 1988. The Dimensions of Residential Segregation. Social Forces 67, 281–315. [Google Scholar]
- Mayne SL, Hicken MT, Merkin SS, Seeman TE, Kershaw KN, Do DP, Hajat A, Diez Roux AV, 2019. Neighbourhood racial/ethnic residential segregation and cardiometabolic risk: the multiethnic study of atherosclerosis. J Epidemiol Community Health 73, 26–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- McFarlane SI, Jacober SJ, Winer N, Kaur J, Castro JP, Wui MA, Gliwa A, Von Gizycki H, Sowers JR, 2002. Control of cardiovascular risk factors in patients with diabetes and hypertension at urban academic medical centers. Diabetes Care 25, 718–723. [DOI] [PubMed] [Google Scholar]
- Meng YY, Diamant A, Jones J, Lin W, Chen X, Wu SH, Pourat N, Roby D, Kominski GF, 2016. Racial and Ethnic Disparities in Diabetes Care and Impact of Vendor-Based Disease Management Programs. Diabetes Care 39, 743–749. [DOI] [PubMed] [Google Scholar]
- Menke A, Casagrande S, Geiss L, Cowie CC, 2015. Prevalence of and Trends in Diabetes Among Adults in the United States, 1988–2012. Jama 314, 1021–1029. [DOI] [PubMed] [Google Scholar]
- Piccolo RS, Duncan DT, Pearce N, McKinlay JB, 2015. The role of neighborhood characteristics in racial/ethnic disparities in type 2 diabetes: results from the Boston Area Community Health (BACH) Survey. Social Science & Medicine 130, 79–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Pool LR, Carnethon MR, Goff DC Jr., Gordon-Larsen P, Robinson WR, Kershaw KN, 2018. Longitudinal Associations of Neighborhood-level Racial Residential Segregation with Obesity Among Blacks. Epidemiology 29, 207–214. [DOI] [PubMed] [Google Scholar]
- Rosenstock S, Whitman S, West JF, Balkin M, 2014. Racial disparities in diabetes mortality in the 50 most populous US cities. J Urban Health 91, 873–885. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sims M, Diez-Roux AV, Dudley A, Gebreab S, Wyatt SB, Bruce MA, James SA, Robinson JC, Williams DR, Taylor HA, 2012. Perceived discrimination and hypertension among African Americans in the Jackson Heart Study. Am J Public Health 102 Suppl 2, S258–265. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Stepanikova I, Baker EH, Simoni ZR, Zhu A, Rutland SB, Sims M, Wilkinson LL, 2017. The Role of Perceived Discrimination in Obesity Among African Americans. Am J Prev Med 52, S77–S85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wagner JA, Tennen H, Feinn R, Finan PH, 2013. Racial discrimination and metabolic control in women with type 2 diabetes. Ethn Dis 23, 421–427. [PubMed] [Google Scholar]
- Whitaker KM, Everson-Rose SA, Pankow JS, Rodriguez CJ, Lewis TT, Kershaw KN, Diez Roux AV, Lutsey PL, 2017. Experiences of Discrimination and Incident Type 2 Diabetes Mellitus: The Multi-Ethnic Study of Atherosclerosis (MESA). American Journal of Epidemiology 186, 445–455. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Williams DR, Collins C, 2001. Racial residential segregation: A fundamental cause of racial disparities in health. Public Health Reports 116, 404–416. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Williams DR, Mohammed SA, 2009. Discrimination and racial disparities in health: evidence and needed research. J Behav Med 32, 20–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.