Heterogeneous associations of socioeconomic status with metabolic disease in racial and ethnic subgroups in the United States: A cross-sectional cohort study in NHANES and All Of Us

Abstract

Background Unfavorable socioeconomic status (SES) is associated with adverse health outcomes and is believed to at least partially mediate racial and ethnic health disparities such that some clinical risk algorithms now incorporate SES measures. However, whether the associations of improved SES with improved health are uniform across US racial and ethnic subpopulations is unknown. Methods and Findings Among adult participants in the National Health and Nutrition Examination Survey 1999–2018 and the All of Us cohort v7, we used logistic regression to examine the association between SES measures (education and income) and type 2 diabetes (T2D) and obesity prevalence, comparing this association in the overall population and in subgroups of self-reported race and ethnicity. modeling SES in several ways to assess for race-by-SES interactions and non-linear or threshold effects. Results Age-adjusted rates of T2D and obesity were highest among non-Hispanic Black, Mexican American, Other Hispanic, and Other/Multi-Racial participants and in those with lower SES. In stratified analyses, higher educational attainment and income were independently associated with lower rates of prevalent T2D and obesity among non-Hispanic White and Asian participants, with smaller or even reversed associations observed among other racial and ethnic groups, particularly non-Hispanic Black participants, in both datasets. Heterogeneity was confirmed by in race-by-socioeconomic interaction analyses. SES measures demonstrated variable patterns of association with disease (e.g., linear, threshold, or U-shaped associations) based on the SES measure, outcome, racial or ethnic group, and dataset used. Conclusions The direction, magnitude, and shape of the association between SES and metabolic disease are heterogeneous across US racial and ethnic groups, SES measures selection and transformation, diseases, and datasets. To prevent biased estimates in both research and clinical calculators which increasingly attempt to incorporate SES, researchers and clinicians should examine for heterogeneity of associations between groups, particularly racial and ethnic groups.