Background Serosorting, preferentially engaging in unprotected anal intercourse (UAI) with partners of the same HIV status, is practiced by some MSM as a risk reduction strategy.

Methods May through November 2010, we assessed serosorting practices and beliefs among MSM seeking STI testing at an urban LGBT health center. We compared serosorting practices and beliefs by HIV status and partner characteristics with Pearson χ² tests. Sexual behaviours, partner characteristics, and HIV status were based on self-report.

Results 705 HIV-negative and 88 HIV-positive MSM completed the assessment. Median time since last HIV test was 7.5 months; nearly a third had not been tested in the last year. Overall, 53% reported no UAI, 27% reported seroconcordant UAI, and 18% reported serodiscordant UAI with last sex partner. UAI was more common with seroconcordant partners than with serodiscordant partners (55% vs 37%, p<0.001); 23% of HIV-positive and 17% of HIV-negative men reported serodiscordant UAI. 81% of men with seroconcordant partners said they knew their partner's status by getting tested together or talking with them; 5% determined partners' status indirectly and 14% did not report how they determined their partner's status. Seroconcordant UAI was more common among HIV-positive than HIV-negative men (53% vs 24%, p<0.001). Among 288 HIV-negative men reporting UAI: seroconcordant UAI was more common with main vs casual partners (76% vs 41%, p<0.001) and within monogamous vs non-monogamous relationships (69% vs 51%, p=0.003); serodiscordant UAI was more common among men with ≥3 recent sex partners vs <3 partners (50% vs 33%, p=0.005) and those who reported any anonymous sex partners vs none (53% vs 28%, p<0.001). Among men who said that they could not be talked out of safer sex with a seroconcordant partner, 10% of HIV-negative men and 16% of HIV-positive men reported serodiscordant UAI. 19% of HIV-negative men who endorsed seroconcordant partnerships reported serodiscordant UAI.

Conclusions Discrepancies between serosorting endorsement and practice underscore the importance of assessing cognitions related to risk-taking and behaviour. High rates of partner concurrency, infrequent testing, and indirect assessment of partner serostatus may limit the effectiveness of serosorting as a risk reduction strategy, even with main partners. The contribution of serosorting to HIV transmission within primary relationships warrants further research.