Bayesian estimation methods for survey data with potential applications to health disparities research.

Understanding how and why certain communities bear a disproportionate burden of disease is challenging due to the scarcity of data on these communities. Surveys provide a useful avenue for accessing hard-to-reach populations, as many surveys specifically oversample understudied and vulnerable populations. When survey data is used for analysis, it is important to account for the complex survey design that gave rise to the data, in order to avoid biased conclusions. The field of Bayesian survey statistics aims to account for such survey design while leveraging the advantages of Bayesian models, which can flexibly handle sparsity through borrowing of information and provide a coherent inferential framework to easily obtain variances for complex models and data types. For these reasons, Bayesian survey methods seem uniquely well-poised for health disparities research, where heterogeneity and sparsity are frequent considerations. This review discusses three main approaches found in the Bayesian survey methodology literature: 1) multilevel regression and post-stratification, 2) weighted pseudolikelihood-based methods, and 3) synthetic population generation. We discuss advantages and disadvantages of each approach, examine recent applications and extensions, and consider how these approaches may be leveraged to improve research in population health equity.

Differences in Serum miRNA Profiles by Race, Ethnicity, and Socioeconomic Status: Implications for Developing an Equitable Ovarian Cancer Screening Test.

Serum miRNAs are promising biomarkers for several clinical conditions, including ovarian cancer. To inform equitable implementation of these tests, we investigated the effects of race, ethnicity, and socioeconomic status on serum miRNA profiles. Serum samples from a large institutional biobank were analyzed using a custom panel of 179 miRNA species highly expressed in human serum, measured using the Abcam Fireplex assay via flow cytometry. Data were log-transformed prior to analysis. Differences in miRNA by race and ethnicity were assessed using logistic regression. Pairwise t tests analyzed racial and ethnic differences among eight miRNAs previously associated with ovarian cancer risk. Pearson correlations determined the relationship between mean miRNA expression and the social deprivation index (SDI) for Massachusetts residents. Of 1,586 patients (76.9% white, non-Hispanic), compared with white, non-Hispanic patients, those from other racial and ethnic groups were younger (41.9 years ± 13.2 vs. 51.3 ± 15.1, P < 0.01) and had fewer comorbidities (3.5 comorbidities ± 2.7 vs. 4.6 ± 2.8, P < 0.01). On logistic regression, miRNAs predicted race and ethnicity at an AUC of 0.69 (95% confidence interval, 0.66-0.72), which remained consistent when stratified by most comorbidities. Among eight miRNAs previously associated with ovarian cancer risk, seven significantly varied by race and ethnicity (all P < 0.01). There were no significant differences in SDI for any of these eight miRNAs. miRNA expression is significantly influenced by race and ethnicity, which remained consistent after controlling for confounders. Understanding baseline differences in biomarker test characteristics prior to clinical implementation is essential to ensure instruments perform comparably across diverse populations. PREVENTION RELEVANCE: This study aimed to understand factors affecting miRNA expression, to ensure we create equitable screening tests for ovarian cancer that perform well in diverse populations. The goal is to ensure that we are detecting ovarian cancer cases earlier (secondary prevention) in women of all races, ethnic backgrounds, and socioeconomic means.

The measurement of racism in health inequities research.

There is limited literature on the measures and metrics used to examine racism in the health inequities literature. Health inequities research is continuously evolving, with the number of publications increasing over time. However, there is limited knowledge on the best measures and methods to examine the impact of different levels of racism (institutionalized, personally mediated, and internalized) on health inequities. Advanced statistical methods have the potential to be used in new ways to examine the relationship between racism and health inequities. In this review, we conduct a descriptive examination of the measurement of racism in the health inequities epidemiologic literature. We examine the study design, methods used for analysis, types of measures used (e.g., composite, absolute, relative), number of measures used, phase of research (detect, understand, solutions), viewpoint (oppressor, oppressed), and components of structural racism measures (historical context, geographical context, multifaceted nature). We discuss methods (e.g., Peters-Belson, latent class analysis, difference in differences) that have demonstrated potential for future work. The articles reviewed were limited to the detect (25%) and understand (75%) phases, with no studies in the solutions phase. Although the majority (56%) of studies had cross-sectional designs, many authors pointed to the need for longitudinal and multilevel data for further exploration. We examined study design features as mutually exclusive elements. However, racism is a multifaceted system and the measurement of racism in many studies does not fit into a single category. As the literature grows, the significance of methodological and measurement triangulation to assess racism should be investigated.

Racial and ethnic heterogeneity in diets of low-income adult females in the United States: results from National Health and Nutrition Examination Surveys from 2011 to 2018.

BACKGROUND: Poor diet is a major risk factor of cardiovascular and chronic diseases, particularly for low-income female adults. However, the pathways by which race and ethnicity plays a role in this risk factor have not been fully explored. OBJECTIVES: This observational study aimed to identify dietary consumption differences by race and ethnicity of US female adults living at or below the 130% poverty income level from 2011 to 2018. METHODS: A total of 2917 adult females aged 20 to 80 years from the National Health and Nutrition Examination Survey (2011-2018) living at or below the 130% poverty income level with at least one complete 24-hour dietary recall were classified into 5 self-identified racial and ethnic subgroups (Mexican, other Hispanic, non-Hispanic [NH]-White, NH-Black, and NH-Asian). Dietary consumption patterns were defined by 28 major food groups summarized from the Food Pattern Equivalents Database and derived via a robust profile clustering model, which identifies foods that share consumption patterns across all low-income female adults and foods that differ in consumption patterns based on the racial and ethnic subgroups. RESULTS: All food consumption patterns were identified at the local level, defined by racial and ethnic subgroups. Legumes and cured meats were the most differentiating foods identified across all racial and ethnic subgroups. Higher consumption levels of legumes were observed among Mexican-American and other Hispanic females. Higher consumption levels of cured meat were observed among NH-White and Black females. NH-Asian females had the most uniquely characterized patterns with a higher consumption of prudent foods (fruits, vegetables, and whole grains). CONCLUSIONS: Differences among the consumption behaviors of low-income female adults were found along racial and ethnic lines. Efforts to improve the nutritional health of low-income female adults should consider racial and ethnic differences in diets to appropriately focus interventions.