Application of Intersectionality Framework andArea-level Indicators in Machine Learning Analysisof Depression...

Depression is a complex mental health disorder influenced by various social determinants of health (SDOH) at individual and community levels. Area-level factors and intersectionality framework, which considers overlapping personal identities, are used in this paper to get a nuanced picture of depression disparities.

This cross-sectional study uses electronic health records data from the All of Us research network. Our study cohort includes 20,042 individuals who completed the SDOH surveys in All of Us and had at least one in-patient visit, with 27.3% diagnosed with depression since 2020. We used depression diagnosis as an outcome, while independent variables include US Religious Census and American urvey responses, area-level variables, sociodemographic characteristics: age group, income, gender, sexual orientation, immigration status, marital status, and race/ethnicity - and the interactions of the latter with each other and with other variables. The association between depression diagnosis and the variables is reported by fitting the logistic regression model on the subset of variables identified by LASSO method.

The analysis revealed that area-level indicators, such as religious adherence and childbirth rates, significantly influenced depression outcomes when interacting with personal identity variables: area-level religious adherence was associated with increased depression odds for women (OR 1.33, 95% CI 1.15-1.54) and non-binary individuals (OR 3.70, 95% CI 1.03-13.31). Overlapping identities, such as younger adults unemployed for less than a year and never married Middle Eastern and North African participants showed higher depression odds (OR 2.3, 95% CI 1.06-4.99, and OR 3.35, 95% CI 1.19-9.45, respectively).

The findings underscore the importance of considering all types of factors: individual, area-level, and intersectional in depression research.