“…All models were adjusted for survey mode and included the following variables, which were selected based on prior research indicating that comparable factors are predictive of substance use in other samples, excepting non-binary gender identity, which was included as a variable of interest due to the absence of prior research assessing its relationship to substance use: age (continuous; Kessler et al, 2005), higher educational attainment (yes/no; Gfroerer et al, 1997), higher income (yes/no; Bassuk et al, 1998), people of color (yes/no; Wallace and Muroff, 2002), MTF spectrum (yes/no; Herbst et al, 2008), non-binary gender (yes/no; Flentje et al, 2015), medical gender affirmation (yes/no; Wilson et al, 2015), lifetime intimate partner violence (yes/no; Reisner et al, 2013), lifetime PTSD diagnosis (yes/no; Jacobsen et al, 2001; Rowe et al, 2015), past 7-day clinically significant depression (yes/no; Davis et al, 2008), current mental health treatment (yes/no; Kessler et al, 1996), inability to access gender-transition care in past 12 months (yes/no; Wilson et al, 2015), public accommodations discrimination in past 12 months (yes/no; Nuttbrock et al, 2014b), past 12-month unstable housing (yes/no; Bassuk et al, 1998), and past 12-month sex work (yes/no; Rekart, 2005). Models were adjusted for recent substance use, as appropriate.…”