A national recruitment strategy for HIV-serodiscordant partners living in Canada for the Positive Plus One study: a mixed-methods study

As of 2018, approximately 62,050 individuals in Canada were estimated to be living with HIV, with an incidence of 5.6 per 100,000 individuals per year [1]. Approximately 85% of diagnosed individuals were receiving treatment, of whom 94% reported viral suppression [2]. Since the introduction of combination antiretroviral therapy (ART) and adoption of Undetectable = Untransmittable (U=U), HIV has become a chronic and manageable disease in resource-rich settings where individuals reporting durable viral suppression had comparable life expectancy to individuals in matched controls [3‐6]. A Swiss study found that the life expectancy of individuals living with HIV at the age of 20 increased from 11.8 years during the monotherapy era (i.e., 1988–1991) to 54.9 years in the early stages of the ART era (i.e., 2006–2013) [3].

While the incidence of HIV has stabilized in Canada and elsewhere following the introduction of ART, the number of individuals living with HIV in Canada is expected to grow, and with it, the number of primary HIV-serodiscordant couples [7, 8]. Data from Sub-Saharan Africa suggest that approximately 50 to 75% of HIV-positive individuals are involved in HIV-serodiscordant relationships [9, 10]. Previous studies conducted in Zambia and in the USA have shown that 60–94 and 68% of incident HIV cases among heterosexual individuals and men who have sex with men (MSM), respectively, were attributable to primary sex partners [10, 11]. In recent years, the focus has shifted to addressing HIV risk within relationships and couple-based interventions to promote safer sex and prevent HIV transmission [12‐14]. However, there are gaps in knowledge regarding experiences of HIV-serodiscordant couples and their management of HIV transmission risk in the ART era, including the quality and extent of supportive services from the perspective of both HIV-positive and HIV-negative partners in a relationship and, given partners’ differential experiences and perceptions, of the couple as a unit [15‐22]. In a recent scoping review, our team identified gaps in the evidence available in Canada, particularly among those involved in stable, long-term, or primary serodiscordant relationships [15]. Many studies recruited participants from HIV clinics and other clinical settings, missing HIV-negative partners who did not require direct HIV care services and potentially missing HIV-positive individuals not linked to or retained in care [15, 23‐25]. Data were lacking on HIV-serodiscordant couples within key populations including transgender, Indigenous, and immigrant communities, including those from HIV-endemic areas, who may experience significant barriers to HIV education, care, and supportive resources [15].

Dyadic studies (i.e., studies involving both partners in a relationship) can provide a better understanding of the interplay of individual and dyadic experiences within the serodiscordant relationship [15]. For example, the HIV-negative partner’s estimate of their HIV-positive partner’s adherence to ART was shown to be a better predictor of viral suppression than the HIV-positive partner’s self-reported adherence [26, 27].

Several challenges exist in the recruitment of representative cohorts of HIV-serodiscordant dyads, introducing gaps and biases in our understanding of HIV-serodiscordant relationships. Given the lack of a means to systematically identify people in HIV-serodiscordant relationships, relationships that are undisclosed to health care providers are hidden [28]. Previous studies have reported challenges in recruiting both partners of dyads with the following characteristics: long relationships with older partners, relationships involving at least one bisexual man, relationships with higher satisfaction [29‐31]. A USA study found that the successful recruitment of both partners in a dyad varied significantly by race and ethnicity, geographical region, education, and relationship type [32]. Previous studies have underscored the need to develop new and creative methods of recruiting and enrolling dyads to obtain a large, diverse sample and increase the independent participation of both partners in the dyad to limit coercion [30, 32].

To address the previously identified challenges to dyad recruitment and to design a study that was relevant, feasible, and addressed needs articulated by serodiscordant couples in the Canadian setting, the Positive Plus One (PP1) study team conducted a feasibility study among staff at AIDS Services Organizations (ASOs) and individuals in serodiscordant relationships [33]. Findings indicated the need to use multiple approaches for recruitment and survey delivery (in both English and French) to involve a national sample representing regional differences with a diverse range of sociodemographic backgrounds and sexual identities [33].

In this paper, we evaluate methods used to recruit and collect data on dyads and individuals involved in current or past HIV-serodiscordant relationships in Canada between 2016 and 2018, discuss the strengths and limitations of our approach, and make recommendations to inform the interpretation of study results and the design of future studies.

The PP1 investigative team comprised 31 academics, clinicians, HIV/AIDS service providers, and people living with HIV from across Canada. The project aimed to understand sociodemographic characteristics, relationship satisfaction, HIV transmission risk, perceived needs and access to supportive services, and subjective experiences of individuals living within an HIV-serodiscordant relationship in Canada via an online/telephone survey followed by an in-depth telephone interview for more complex, open-ended questions. The survey was designed to take approximately 30 minutes to decrease the risk of respondent fatigue [34]. To reduce the risk of participation bias identified in previous dyad studies, PP1 used a multi-pronged recruitment strategy to survey one or both partners in a current or recently concluded serodiscordant relationship in Canada from 2016 to 2018. While our team planned for recruitment to take one year, it ultimately took two years to recruit our sample. In this paper, we describe participant enrollment from various recruitment sources. We also compare the sociodemographic characteristics of HIV-positive participants in PP1 to HIV-positive individuals included in Canada’s public health surveillance data (detailed later in Methods) and participants’ relationship characteristics by whether their partner was also recruited to the study [35]. Furthermore, we examine the proportion of missing data across the survey questions and correlates of participation for individuals recruited by partners in their relationships.

We recruited 613 participants (355 HIV-positive; 258 HIV-negative) over two years, including 540 participants in a current HIV-serodiscordant relationship at the time of the study and 73 participants from past relationships that ended within two years prior to survey completion. At the dyad level, 306 participants were recruited from 153 relationships that included both partners (i.e., complete dyads); 307 individuals participated without their current partner. In total 460 relationships were represented.

Led by a large diverse group of investigators, PP1 provided key insight on the use of a mix of in-person and online strategies to recruit a diverse sample of individual HIV-positive and HIV-negative partners and dyads involved in a current or past HIV-serodiscordant relationship in Canada from 2016 to 2018. We found that main sources of recruitment varied between HIV-positive and HIV-negative partners. Additionally, the majority of index partners were willing to recruit their primary partner to the study and 40% were successful. Given differences in characteristics between recruited dyads and relationships where only one partner was enrolled in the study and between individuals involved in a current or past HIV-serodiscordant relationship, our findings support the need for the inclusion of both dyads and individual partners involved in a HIV-serodiscordant relationship and those involved in both current and past HIV-serodiscordant relationships to gain a full understanding of the experiences of living in a HIV-serodiscordant relationship.

Due to the hidden nature of HIV-serodiscordant relationships, our study used active recruitment by ASO staff and clinicians when possible and passive recruitment when staffing resources were not available. While we recruited the majority of our participants through staff or physicians at HIV clinics and/or ASOs, a large portion of our participants learned of the study through poster, pamphlet, or cards displayed at these and other recruitment sites and through their partner and/or friends. Among those involved in a current HIV-serodiscordant relationship, HIV-positive participants were more likely to learn of our study through HIV clinics and/or ASOs via physicians, staff, and/or newsletters whereas HIV-negative participants were more likely to learn of our study through their partner. These findings suggest that our multi-pronged approach to participant recruitment was needed to recruit HIV-positive and HIV-negative individuals involved in different types of HIV-serodiscordant relationships. Since the majority of HIV-negative individuals heard about our study through their partner, the PP1 recruitment strategy expanded as the study progressed to include the placement of posters and pamphlets at pharmacies dispensing PrEP, anonymous HIV testing sites, and methadone clinics. Ultimately these were effective at reaching HIV-negative individuals.

Similar to a study conducted by Starks and colleagues, we found that a high proportion of index participants (78%) were willing to invite their primary HIV-serodiscordant partner to participate, and 40% of these index participants successfully recruited their primary partner to enrol [30]. Following the lessons learned from previous studies, each partner was asked to complete the survey independently to limit coercion and partner bias and increase the participation of both partners in the dyad. We found that dyads provide more complete information compared individual partners on some variables of interest. For example, 13% of individuals who were the sole representative of their relationship reported the viral suppression status of the HIV-positive partner in the relationship as “unknown” or “missing” compared to 7% of dyads. However, viral suppression reported by dyads may result from healthier relationships and therefore may also be biased towards healthier outcomes. Additionally, studies examining HIV-serodiscordant relationships that only examine dyads may be biased towards those with higher relationship and/or sexual satisfaction (i.e., “happy couple” bias). Participants with higher relationship satisfaction and longer relationship duration were more likely to successfully engage their partner to take part in the study. Previous dyadic studies conducted among gay and heterosexual couples have noted that their samples included disproportionately more satisfied couples and couples with more confidence in their relationship [29‐31]. More creative strategies are needed to help index partners recruit their partner in relationships with lower satisfaction and shorter duration as these populations may be under-represented in the current study and may be in need of different types of support.

Our study was not designed to capture detailed information on polyamorous HIV-serodiscordant relationships. A previous study among individuals involved in a hierarchical polyamorous relationship reported lower relationship satisfaction with their secondary and tertiary partners compared to non-hierarchical and primary partners [46]. While polyamorous partners were eligible to participate in the study, we did not ask whether their relationship was polyamorous, and could only link one serodiscordant partner to an index partner. Additional serodiscordant partners completed the survey as unlinked individuals. Future studies may wish to expand their inclusion criteria to individuals involved in casual relationships and expand analytic approaches to incorporate polyamorous relationships. Longitudinal studies are needed to observe the impact of changes in relationship satisfaction and sexual behaviour over time and correlates of relationship termination, an outcome of interest for HIV-serodiscordant couples. While PP1 tried to capture these outcomes retrospectively, the cross-sectional design limited our ability to draw firm conclusions.

PP1 has provided insights into several challenges of a multi-pronged approach for recruiting individuals involved in current or past HIV-serodiscordant relationships. Although we used a multi-pronged recruitment approach, the majority of our sample was recruited from ASOs and/or clinics, which may have led to potential biases in our sample. A previous study conducted in Ontario, Canada revealed that individuals living with HIV who used ASOs were less healthy, had lower quality of life, and lower income compared to those who did not receive services from ASOs [47]. Hence, our study may have over-represented individuals involved in HIV-serodiscordant relationships with poorer health and lower quality of life and income. Additionally, recruitment via physicians and clinics may have introduced bias towards those who were more connected to HIV care, and therefore, ART initiation and viral suppression [48]. Despite our best efforts to reach as many HIV-serodiscordant couples as possible, a comparison with the PHAC national HIV surveillance database indicated that our sample under-represented HIV-positive individuals who were ≥ 30 years of age at the time of diagnosis, men, those self-identifying as Black or Indigenous, and those who resided in British Columbia, Alberta, Quebec, and the three territories. The under-representation of Black or Indigenous HIV-positive individuals in our study may indicate that we were unable to adequately reach these populations despite our efforts to engage recruitment sites that work with them. These populations are generally known to be underrepresented in research studies [32]. Previous studies have shown that internet-based recruitment strategies including Facebook advertising can be a cost-effective method of recruiting a diverse sample of participants who are at risk of acquiring HIV infection, particularly those identifying as racial minorities [36, 37]. PP1 only recruited a small proportion of participants through online ads (e.g., Facebook), which could explain under-recruitment. Additional discussions with key stakeholders may be necessary to improve participation among these populations [32]. It may also be possible that HIV-positive individuals who were underrepresented in PP1 relative to national surveillance data are less likely to be involved in HIV-serodiscordant relationships compared to other HIV-positive individuals. However, since HIV-serodiscordant relationship status is frequently unknown to clinicians and ASO staff and there is a lack of a routine database of individuals involved in HIV-serodiscordant relationships in Canada, these findings are difficult to confirm. It should also be noted that PHAC data may not be representative of the current population living with HIV as it includes individuals who were diagnosed since the beginning of the HIV epidemic, a large proportion of whom identified as White and MSM and may no longer be alive [42]. As such, PHAC data may have over-represented the proportion of men living with HIV, partially explaining the lower proportion of men living with HIV enrolled in PP1 compared to PHAC surveillance data.

Our study did not have a budget to reimburse ASOs, NGOs, and clinic staff for their time spent actively recruiting participants. When organizations were initially approached about involvement in recruitment, many asked about cost recovery for recruitment activities. As a result of governmental ASO funding cutbacks during the course of study recruitment, active recruitment likely decreased as the study progressed [49]. Without a dedicated budget for recruitment efforts, accessing hard-to-reach populations likely worsened over time and overall recruitment took longer than expected.

Finally, PP1 took place in a Canadian setting, and sought to recruit individuals involved in a primary HIV-serodiscordant relationship. As such, findings may be generalizable to resource-rich settings.

Our findings provide important insights that support the use of a multi-pronged approach to recruit a diverse sample of individuals involved in current or past HIV-serodiscordant relationships in Canada that include a mix of complete dyads and lone participants in current HIV-serodiscordant relationships. Our findings suggest several considerations to facilitate the recruitment, enrollment, and engagement of a diverse sample of individuals involved in HIV-serodiscordant relationships for future studies. In particular, the recruitment of both dyads, lone participants, and individuals involved in past HIV-serodiscordant relationships can help researchers obtain a more diverse sample in terms of relationship duration and satisfaction. Additionally, future studies should employ creative strategies to recruit individuals involved in past, casual, and/or polyamorous HIV-serodiscordant relationships and relationships with lower satisfaction in order to further minimize the risk of “happy couple” bias. Future studies that focus on hard-to-reach areas and populations should involve more discussion with stakeholders, expand the use of social media platforms, and approach more sites utilized by HIV-negative individuals (e.g., pharmacies prescribing PrEP, anonymous HIV-test sites, methadone clinics) to increase enrollment from underrepresented populations to ensure the recruitment of a more representative sample. Budgeting for active recruitment in ASOs and clinics where staffing may be limited is essential to increase enrollment.