Multiple marginalisation and unintended pregnancy among racial/ethnic and sexual minority college women

Background Both sexual minority and racial/ethnic minority women are at increased risk of unintended pregnancy compared to their heterosexual and non-Hispanic white peers, respectively. A recent study suggests the presence of negative interaction between sexual orientation and race/ethnicity, although it was not explicitly assessed. Objectives To characterise the burden of unintended pregnancy across groups defined by intersecting sexual orientation and race/ethnicity and to explore for potential intersectional interaction between social identities. Methods We analysed cross-sectional survey data from the National College Health Assessment collected between Fall 2015 and Spring 2018. The relative risk (RR) of unintended pregnancy among college-attending women aged 18-25 was estimated using multivariable log-binomial regression (n = 177 592). We estimated multiplicative- and additive-scale (relative risk due to interaction, RERI) interaction between race/ethnicity and sexual orientation on the risk of unintended pregnancy; 95% confidence intervals (CI) were estimated using 500 bootstrap replicates. Results Unintended pregnancy was rare across all intersecting identity groups (<= 1.5%). Compared to heterosexual non-Hispanic white women, all groups of sexual minorities and/or racial/ethnic minorities had a higher risk of unintended pregnancy. This was consistent within strata of sexual orientation and race/ethnicity. We did not find strong evidence of additive-scale interaction between sexual orientation and race/ethnicity (RERI -0.19, 95% CI -0.67, 0.28), but we did find negative multiplicative-scale interaction (interaction RR 0.76, 95% CI 0.59, 0.96). Conclusions These findings illustrate that interactions, which are typically used to quantify intersectional effects, are scale-dependent. This has implications regarding how risk estimates for multiply marginalised groups are interpreted. More broadly, our findings suggest the need to move beyond multiple jeopardy approaches to quantitative intersectional research and consider the upstream factors that may uniquely shape the health status of groups at the nexus of particular social identities.