Application of intersectionality framework and area-level indicators in machine learning analysis of depression disparities in all of us research program data

Key findings from this study

• The study found that inability to work represented the strongest depression risk factor, with further elevation when combined with medium per capita income.
• The researchers demonstrate that non-binary identity carried dramatically elevated risk when intersected with older age (OR 5.14) or foreign birth status (OR 10.23).
• The authors report that protective factors differed by gender and age group, with younger adults benefiting from parenthood while military service protected older adults.

Overview

The study examines depression disparities using electronic health records from the All of Us research network, integrating area-level socioeconomic and religious indicators with individual-level demographic and social determinant variables. The analysis employed mixed-effects logistic regression with LASSO feature selection and incorporated intersectional interactions to identify depression risk and protective factors across overlapping identity categories.

Methods and approach

The cross-sectional cohort comprised 33,994 individuals who completed social determinants of health surveys and had at least one inpatient visit between 2020 and 2025, with depression diagnosis as the outcome variable. Independent variables included area-level measures from US Religious Census, American Community Survey, and All of Us (religious adherents, birthrates by living arrangement, median area income, deprivation index) and individual-level variables (age, income, gender, sexual orientation, race/ethnicity). The analysis incorporated interactions between individual-level variables and between individual and area-level variables. Mixed-effects logistic regression with LASSO-selected features and 3-digit ZIP code as a random effect estimated associations with depression.

Results

Male sex at birth demonstrated protective effects among straight (OR 0.71) and non-binary (OR 0.42) individuals. Older adults with active military service (OR 0.64) or health insurance (OR 0.80) and younger adults with one to two children (OR 0.72) showed reduced depression risk. Inability to work emerged as the strongest risk factor (OR 2.01), amplified by medium per capita income (OR 1.39). Housing concerns (OR 1.24) and frequent disrespect (OR 1.41) increased risk. Intersectional risk effects included non-binary identity in older adults (OR 5.14), non-binary individuals born outside the USA (OR 10.23), women with children (OR 1.39), and Hispanic active-duty members (OR 1.90). Area deprivation in older adults elevated risk (OR 1.31).

Implications

Clinical and public health interventions for depression prevention and treatment require strategies that account for overlapping personal identities and incorporate place-aware approaches. The substantial intersectional risk elevations—particularly among non-binary individuals and those with multiple disadvantaged status categories—suggest that standard depression prevention protocols may inadequately address specific population subgroups. Integration of both area-level socioeconomic indicators and individual identity intersections into clinical decision support systems could improve targeting of preventive and therapeutic resources.

Scope and limitations

This summary is based on the study abstract and available metadata. It does not include a full analysis of the complete paper, supplementary materials, or underlying datasets unless explicitly stated. Findings should be interpreted in the context of the original publication.