Fig. 1. Beta-diversity is linked with ethnic differences in depressive symptom scores.

A Beta-diversity predicting PHQ9 depression. It presents results of linear regression analyses that model β-diversity as a predictor of depressive symptom levels. Panel (A) horizontal bars present ΔR² after progressive adjustments for confounders (models 1a to 3), and respectively without and with ethnicity included in each regression model. B Ethnicity predicting PHQ9 depression. It presents results of linear regression in which β-diversity is modeled as a mediator of the association between ethnicity and depressive symptom levels (see lower figure). Bars present ΔR² the prediction of PHQ9 by ethnicity after progressive adjustments (models 1a to 3). Blue bars present ΔR² without β diversity incorporated as a mediator in the model, and the orange bars present ΔR² when mediation is assumed. The % in the table (right) indicate the attenuation of the direct effect by mediation. Regression models: We used two-sided linear regression analyses, no adjustments were made for multiple comparisons. Model 1a adjusted for age and gender; Model 1b added education; Model 2 further added behavioral factors (alcohol, smoking, exercise, BMI); Model 3 added GI disease, Diabetes, PPI use, Recent antibiotics, Diarrhea. All ΔR² p ≤ 0.001, except for ethnicity-inclusive Model 3 (p = 0.023).