Perceptions about risk and actual behavioural risk in the ACB population converge in some areas, and diverge in others. First, while participants saw HIV risk as removed from Canada, quantitative data showed that behavioural HIV risk was higher among immigrants who had been in Canada longer and was higher among born Canadians than among immigrants. The comparatively low overall HIV prevalence in Canada may be responsible for this perception. Second, as per community members’ and service providers’ perceptions, sexual partner concurrency was fairly common in the ACB population. While there were no significant differences according to the chi-square tests comparing sexual partner concurrency in the different groups, it seems that women and people living at or below the LICO may be more likely to report being in a non-monogamous partnership. Third, abstinence was mentioned as a reason for low perception of risk by women, but Phase II showed that women were not significantly more likely than men to either have never had sex or be abstinent in the past year. In fact, never having sex was associated with poverty status, length of time in Canada and employment status. Furthermore, past year abstinence was significantly more likely to occur among: people living below the LICO, immigrants who had been in Canada for less time and people with less stable employment statuses. Fourth, Phase I participants said mixing sex with drugs or alcohol was a risk factor for HIV, and Phase II results show that women, immigrants who had been in Canada for less time, people with more stable immigration statuses, and people with less stable employment statuses were less likely to engage in this behaviour than other groups. Fifth, past year HIV testing was relatively low, and it was not associated with any of the markers of SSP in this paper. However, lifetime HIV testing might be higher than service providers perceived. Sixth, the prevalences of unprotected sex with regular and casual partners were high, thus confirming perceptions about unprotected sex being an issue within the ACB population. People living above the LICO and those who had regular full-time employment or were self-employed were more likely to engage in unprotected sex. Although not statistically significant, the results suggest that people who had been in Canada for more than 5 years and Canadian-born persons are more likely to engage in unprotected sex than new immigrants.
The data from Phase I show that there may be a disconnect between community members’ and service providers’ perceptions about HIV risk in the ACB population. Both groups agreed that unprotected sex, partner concurrency and low prevalence of HIV testing were important risk factors within the community. However, while community members spoke about abstinence as a protective factor, service providers did not seem to be aware that past-year and long-term abstinence were fairly common. Also, some of the barriers to protection that service providers cited (e.g. the need to be loved, cultural norms around disclosing information) were not mentioned by community members at all. These discrepancies may reflect service providers adopting a more analytical lens based on their overall observations versus community members sharing their individual experiences. Alternatively, these discrepancies may be due to service providers relying on research from the United States of America and other countries to inform their work. Consequently, they may not have contributed their own experiences and observations, but rather they could be repeating information from other service providers or researchers, or worse, they could be relying on stereotypes to inform their perceptions. Stereotypes and erroneous perceptions can be damaging to HIV prevention and care efforts, as they influence the types of actions that are taken to address HIV .
The data illustrate that poverty status, immigration experience and employment status are linked to the distribution of HIV risk and protective behaviours. The link between gender and HIV risk behaviours may be less apparent, because the effect of gender on HIV risk is likely dependent on its interaction with other markers of SSP, as Intersectionality Theory demonstrates . According to the data, those with higher SSP may be at greater risk of HIV exposure or transmission when compared to those with lower SSP. This is not surprising because, the combination of multiple marginalizations can create unique SSPs that simultaneously limit and enhance one’s agency [13, 47]. Hence, the combination of ACB identity and low SSP may protect an individual from engaging in particular HIV risk behaviours. Studies have also shown that early in an HIV epidemic, people with higher SSP are at greater risk for infection. However, as the epidemic matures and effective prevention interventions are designed, people with higher SSP are able to access and take advantage of the interventions. Hence people with lower SSP will begin to be at greater risk for infection, comparatively . Additionally, the “healthy immigrant effect” may be at play, which could explain why newer immigrants have lower risk than those who have been in Canada longer and Canadian-born persons . Furthermore, the data on immigration may reflect the effect of immigrants being exposed to HIV prevention messages in their home countries prior to immigration.
Given these findings, HIV prevention interventions should not be based on the assumption that low SSP automatically means high behavioural risk. Illustratively, fairly recent studies from Sub-Saharan Africa have shown that higher income [16, 19], higher educational attainment , being employed  and being male [24, 50] were associated with increased HIV risk, so these findings are not unusual. Paradoxically, British and North American studies show that HIV risk is associated with low income or poverty [51, 52], low educational attainment , female sex  and immigration experience . These contradictions are not surprising, however, as the impacts of SSP are context-specific and are influenced by governance, policies, cultures, and values . At minimum, prevention interventions for ACB people locally, and possibly in other parts of Canada, should include consideration of gender, poverty status, immigration experience and employment status.
Since the qualitative analyses were descriptive and more in-depth exploration is beyond the scope of this paper, deeper meanings of, and connections between perceptions were not explored in more detail. Furthermore, social desirability bias may be present when data about sensitive topics, such as HIV risk, are self-reported. This type of bias occurs when participants give inaccurate responses that others will view favourably, and it is more likely to occur when data are collected in less-anonymous ways, such as through in-person interviews . It could have accounted for some of the discrepancies between service providers’ and community member’s perceptions about HIV risk and protective behaviours within the local ACB population. The interview results should be interpreted cautiously.
As convenience sampling was used, Phase II of the study was subject to selection bias. However, non-response weights were applied to adjust for some of this bias. The sample’s size was smaller than the 384 participants required for a desired precision of 5%. Consequently, the confidence intervals for the prevalence estimates are wide, so the prevalence estimates should be interpreted in light of these wide ranges of plausible values. However, the study was adequately powered to detect statistically significant relationships, even with the wide confidence intervals. In all, 31% of the χ
2 tests (31/100) were significant at the p = 0.05 level. The percentage of significant χ
2 tests ranged from 15% for gender and immigration status to 50% for employment status. These percentages show that chance alone does not account for the results. Furthermore, the results demonstrate significant differences between groups and patterns in the distribution of risk behaviours by SSP.
Lastly, even though some aspects of Dillman’s “Tailored Design Method” were applied , only 32% of questionnaires were returned, which may be another source for selection bias. Other steps could have been taken to increase recruitment, such as: providing monetary incentives, further shortening the questionnaire, and having a web-based version of the questionnaire. Participants were given the option to request an interviewer to administer the questionnaire, but none requested one. Notably, the proportion of questionnaires returned is comparable to the proportion of participants who were successfully recruited into a similar study with East Africans in Toronto, Canada that offered monetary incentives and used interviewers .