How Much of the Sexual-Orientation-Related Disparities in Youth Mental Health Can be Explained By Experience of Victimization? a Mediation Analysis with Causally Defined Effects

Introduction

This study assessed to what extent sexual-orientation-related disparities in youth mental health are associated with disproportionate exposure to violence. We evaluated Pearl/VanderWeele’s causally defined direct and indirect effects, extending Muthen’s ordinal-mediator-binary-outcome method to allow multiple mediators.

Methods

We used Massachusetts Youth Risk Behavior Survey pooled 2003-2011 data to assess the contribution of violence to mental health differences between sexual minority (lesbian/gay, bisexual, or unsure of sexual identity) and heterosexual youth, separately for males and females. Mediation analysis was conducted via a structural probit model with sexual orientation, binary/ordinal mental health outcomes (depression, non-suicidal self-injury, suicidal thought, plan, attempt, and attempt with injury) and binary/ordinal mediators (threatened with weapon, physical fights, injured in fights, hurt on a date, forced sex), controlling for age, race/ethnicity and year.

Direct, indirect and total effects (DE, IE and TE) were defined from a causal inference perspective: DE compares risk of outcome given a minority status and given heterosexual status, both with mediators at heterosexual group levels [P_X0 = P(Y | minority, M(heterosexual)); P₀₀ = P(Y | heterosexual, M(heterosexual))]. IE compares risk given minority status and mediators at minority group levels [P_XX = P(Y | minority, M(minority))] and at heterosexual group levels [P_X0]. TE compares P_XX and P_X0. To compute P₀₀, P_X0 and P_XX, we adapted Muthen’s formula for the Y*-rescaling factor (for the single mediator case) to incorporate the mediator residual correlation matrix. We calculated DE, IE and TE on multiplicative (risk ratio – RR) and additive (risk difference – RD) scales: RR_TE = RR_DE * RR_IE; RD_TE = RD_DE + RD_IE. We computed the proportion of RD explained by mediation PE = RD_IE/RD_TE. Confidence intervals (CI) were bootstrapped.

Results

RR_TE are statistically significantly greater than 1, indicating greater risk for sexual minorities. For females, RR_DE (indicating differential risk when “fixing” violence exposure at heterosexual group levels) are non-significant for depression comparing lesbian and unsure youth to heterosexuals. For males, RR_DE are non-significant for all outcomes for unsure youth, and for suicide attempt with injury for all minority groups. RR_IE (indicating risk increase for sexual minorities due to differential violence exposure) range from 1.2 to 1.7 for females, 1.4 to 4 for males, all 95% CI greater than 1. PE range from 32% to 78% for females, 41% to 100% for males, all 95% CI not including zero.

Conclusion

Sexual orientation disparities in depression and suicidality are partially due to physical and sexual violence. Effective interventions are needed.