Comparing substance use outcomes by sexual identity among women: Differences using propensity score methods

Background: Differences in alcohol, tobacco, and other drug (ATOD) use by sexual identity vary across samples of women recruited using different sampling methods. We used propensity score (PS) weighting methods to address two methodological questions: (1) Do disparities between sexual minority women (SMW) and heterosexual women persist when differences in risk and protective factors are similarly distributed between groups, and (2) Does accounting for SMW-specific resiliency factors impact differences between non-probability samples of SMW? Methods: Four samples included SMW from a longitudinal study with a nonprobability sample (n = 373), a national general population panel sample (n = 373), and a national LGBTQ-specific panel sample (n = 311), as well as a national probability sample of heterosexual women (n = 446). Between-groups analyses using double -robust PS weighted models estimated differences in ATOD use under hypothetical conditions in which samples have similar risk and protective factors. Results: After PS weighting, imbalance in confounders between SMW and heterosexual samples was substantially reduced, but not eliminated. In double-robust PS weighted models, SMW samples consistently had significantly greater odds of drug use than heterosexuals, with odds from 8.8 to 5.6 times greater for frequent marijuana use and 4.8-3.2 greater for other drug use. Few differences between SMW samples in ATOD outcomes or other variables remained after PS weighting. Conclusion: Relative to het-erosexual women, disparities in marijuana and other drug use among SMW are evident regardless of sampling strategy. The results provide some reassurance about the validity of large nonprobability samples, which remain an important recruitment strategy in research with SMW.