Reproductive
Expanding our methods for studying contraceptive preferences: Application of the multilevel analysis of individual heterogeneity and discriminatory accuracy (MAIHDA) approach to the statewide Surveys of Women Mikaela Smith* Mikaela Smith Payal Chakraborty Robert Hood Autumn Kirkendall
Introduction: Most research on contraceptive access focuses on use rather than preference. Given existing reproductive coercion and medical mistrust, an intersectional framework may better capture equity gaps between preference and use. We used a new quantitative method to examine how race/ethnicity, sexual orientation, and socioeconomic status (SES) inform the social patterning of preferred contraceptive use.
Methods: We used 2021–2023 data from 7 states from the state-representative Surveys of Women (N=12,947). We conducted an intersectional multilevel analysis of individual heterogeneity and discriminatory accuracy (MAIHDA). We nested participants within 32 social strata defined by race/ethnicity, sexual orientation, and SES. We fit two multilevel logistic models via Bayesian MCMC with noninformative priors: (1) an empty model with random intercepts for social strata (simple intersectional) and (2) a model adding fixed effects for social strata (intersectional interaction). We compared MAIHDA with a traditional logistic regression approach with interaction terms.
Results: Overall, 27% of respondents (N=3,385) were not using their preferred contraceptive method. Lesbian, gay, bisexual, and other sexually minoritized (LGB+) women, racial/ethnic minoritized women, and women with lower SES had lower odds of preferred method use. Additive effects explained 91% of the between-stratum variance in preferred method use while interaction effects explained 9%. Black LGB+ women without college degrees and with incomes under $100,000 had the lowest probability of preferred method use (51%), while White heterosexual women with college degrees and incomes over $100,000 had the highest (80%). ORs from MAIHDA and the traditional approach were similar.
Conclusion: Our findings reflect structural inequities in the healthcare system. Traditional models show how isolated identities predict preferred method use but are insufficient in capturing nuances in intersectional experiences.