Table 4.

Multivariate logistic regressions of lesbian, gay, bisexual, and transgender (LGBT) young adults recruited via social media (vs intercept). Predictors include variables related to the recruitment methods in the bivariate analyses (P<.25). Variables for past 30-day cigarette and alcohol use are dichotomized to any past 30-day use vs no past 30-day use (reference category, REF). Social media use frequency variables are dichotomized to at least once a day vs less than once a day (REF). Analytic N=2945 (social media N=1183, intercept N=1762).

Variable		AORa	95% CI	P value
Age		0.76	0.72-0.80	<.001
LGBTb identity
	Gay men	REFc	REF	REF
	Lesbian or gay women	1.88	1.54-2.29	<.001
	Bisexual men and women	1.46	1.17-1.82	.001
	Gender minorities	1.68	1.26-2.25	<.001
	Other sexual minorities	2.48	1.62-3.80	<.001
Education
	High school or less	1.07	0.83-1.40	.60
	Some college	1.27	1.03-1.56	.03
	College plus	REF	REF	REF
Race or ethnicity
	White, non-Hispanic	REF	REF	REF
	Black, non-Hispanic	0.76	0.58-1.01	.05
	Hispanic	0.73	0.61-0.89	.001
	Other or multiracial, non-Hispanic	0.70	0.54-0.90	.006
Past 30-day cigarette use		0.94	0.80-1.10	.42
Past 30-day alcohol use		0.33	0.24-0.44	<.001
Past-year pride event attendance		1.31	1.06-1.64	.02
Facebook use at least once a day		1.43	1.14-1.80	.002
Instagram use at least once a day		0.73	0.62-0.86	<.001

^aAOR: adjusted odds ratio.

^bLGBT: lesbian, gay, bisexual, and transgender.

^cREF: Reference category.