Racial and Ethnic Disparities in Type-2 Diabetes: A Multilevel Perspective

Abstract

In the United States (US), diabetes affects an estimated 13% of adults (25.8 million people).2, 3 A disproportionate burden of the disease is borne by US minority populations.4 Black and Hispanic Americans have higher prevalence of type 2 diabetes mellitus (T2DM),5 achieve poorer disease control,6, 7 and have more T2DM complications than their White counterparts.8, 9 Efforts to reduce these disparities are hindered by the fact that patients typically have T2DM for 4-7 years prior to diagnosis.10 There is a confluence of disadvantages: behavioural risk factors, genetic predisposition, lack of access to adequate health care, and local environmental disadvantages, all are likely to contribute to these increased burdens in a synergistic fashion. A comprehensive understanding of “upstream” factors contributing to racial/ethnic differences in T2DM therefore offers the greatest potential to reduce the “downstream” costs of T2DM faced by disadvantaged populations.11 This research investigated the roles of certain risk factors in racial/ethnic variation in T2DM using the Boston Area Community Health (BACH) Survey. The BACH Survey is a community-based, stratified random sample, epidemiologic cohort of 5,502 Boston, Massachusetts residents. Follow-up surveys were conducted approximately five (BACH II, 2008-2010, N=4,144) and seven (BACH III, 2010-2012, N=3,155) years later. The BACH III survey was designed to assess the relative contributions of (1) genetic, (2) lifestyle/behavioural, (3) psychosocial, (4) biophysiologic, (5) contextual/neighbourhood, and (6) social/economic determinants to racial/ethnic disparities in diabetes. Therefore, my analyses focused on the 3,155 participants of the third wave of the BACH survey. First, I examined the role of biogeographic ancestry (BGA) versus socioeconomic factors in racial/ethnic disparities in the incidence of T2DM over roughly seven years of follow-up. I used the excess relative risk method, the risk difference method, and g-computation to examine the direct and indirect effects of race/ethnicity on T2DM incidence. Using the g-computation method, I found that socioeconomic factors accounted for 44.7% of the excess risk of T2DM among Blacks and 54.9% among Hispanics. The findings indicated that BGA had almost no direct association with T2DM and was almost entirely mediated by self-identified race/ethnicity and socioeconomic factors. Second, I examined the role of neighbourhood contextual factors in racial/ethnic disparities. Two-level random intercepts logistic regression was applied to assess the associations between race/ethnicity, neighbourhood characteristics (census tract socioeconomic status, racial composition, property and violent crime, open space, geographic proximity to grocery stores, convenience stores, and fast food, and neighbourhood disorder) and prevalent T2DM (BACH III diabetes status). Multilevel models indicated a significant between-neighbourhood variance estimate of 0.943, providing evidence of neighbourhood variation. Individual-level demographic factors (race/ethnicity, age and gender) explained 22.3% of the neighbourhood variability in T2DM. However, the addition of neighbourhood-level variables to the model had very little effect on the magnitude of the racial/ethnic disparities and on the between-neighbourhood variability. Finally, I assessed the relative contributions of six domains of influence to racial/ethnic disparities in T2DM: (1) socioeconomic, (2), local environmental, (3) psychosocial, (4) lifestyle/behavioural, (5) biophysiologic, and (6) genetic/ancestral. I constructed risk scores for each domain of influence and used structural equation models (SEM) to evaluate the direct effects of each conceptual domain of influence on T2DM prevalence as well as the indirect effect of each conceptual domain on the magnitude of the racial/ethnic disparities in T2DM. The final models indicated that 38.9% of the total effect of Black race on T2DM prevalence was mediated by the socioeconomic, environmental, psychosocial, lifestyle/behavioural risk scores with 21.8% of the total effect of Black race being explained by socioeconomic risk. 45.7% of total effect of Hispanic ethnicity was mediated. Again, the largest mediator was the socioeconomic risk score with 26.2% of the total association explained.My analyses consistently demonstrated that social determinants contributed to racial/ethnic disparities in T2DM. My results suggest that socioeconomic factors are the largest contributors to the causation and/or amplification of these disparities. Biogeographic ancestry (an individual’s genetic race/ethnicity) had no direct effect on T2DM prevalence or incidence. Neighbourhood factors did not contribute to racial/ethnic disparities once individual socioeconomic factors were taken into account. Finally, while lifestyle/behavioural and biophysiologic characteristics had significant direct effects on T2DM prevalence, they did not appear to substantially contribute to disparities in T2DM once socioeconomic factors were taken into account. These results have national and local policy implications as they suggest that in order to reduce disparities, either wide-scale social and economic policy shifts need to occur, or interventions need to be targeted toward racial/ethnic minorities and the socially and economically disadvantaged.