Introduction Most research on the sexual health of gay, bisexual, and other men who have sex with men (MSM) relies upon convenience samples, which may not fully represent the MSM population. To investigate this form of selection bias, we concurrently sampled MSM from three distinct venue types, compared sociodemographic, behavioural, and STI testing characteristics across samples, and characterised the degree of overlap between them.
Methods MSM 18+ years of age from British Columbia, Canada completed an anonymous survey during July-December 2016. Participants were recruited from: mobile sex-seeking apps (n=662); a MSM-branded STI clinic (n=303); and in-person at the Vancouver Pride Festival (n=307). Cross-sample comparisons with >10% relative difference and p<0.05 are shown. Sample overlap was measured by self-reported use of other recruitment venue types.
Results MSM recruited from apps included more bisexual men (24%) than those recruited from the clinic (9%) or from Pride (7%), while MSM recruited from the clinic included more East Asian and Latino men (14% and 13%, respectively, vs. 13% and 6% from Pride, and vs. 5% and 3% recruited from apps). More 18–29 year-olds were recruited from the clinic (47%, vs. 29% Pride, 16% apps). A larger proportion of MSM recruited from apps reported >10 sex partners in the past year (35%, vs. 28% clinic, 21% Pride). Finally, more clinic participants tested for STI/HIV in the past year (82%) as compared with participants recruited from apps (75%) or Pride (73%). When asked which other venues they frequented, 19% of the total sample reported using all 3 recruitment venue types, while 11% of Pride participants, 7% of apps participants, and 4% of clinic participants reported only using the recruitment venue where sampled.
Conclusion We found large differences between MSM sampled from apps, an STI clinic, and Pride, with 22% unlikely to be sampled if relying on a single venue type for recruitment. Our results underscore the importance of multiple sampling strategies in MSM research and provide specific cross-sample differences.