Black or African American (hereafter referred to as Black) and Hispanic/Latino gay, bisexual, and other men who have sex with men (BMSM and HLMSM) in the United States (US) are disproportionately affected by HIV and jointly account for over two-thirds of new HIV diagnoses¹. Unfortunately, annual HIV testing rates are suboptimal, with screening rates lower than desired, particularly for older patients and patients in primary care settings², limiting linkage to care. Given barriers to accessing health facilities, including stigma and medical mistrust due to historic mistreatment faced by many US MSM^3–5, especially MSM of color^6–10, HIV self-testing may be uniquely well-suited to reach this high-need population.

HIV self-testing is a feasible and acceptable strategy to increase HIV testing and serostatus awareness among US MSM^11–14 without risk compensation¹². TakeMeHome, a public-private partnership to distribute HIV self-tests (HIVSTs) to US MSM, successfully engaged MSM, many of whom had not previously tested (36%) or had not tested in the past year (56%)¹⁵. This suggests that HIV self-testing may be a good mechanism to reach people who do not regularly engage with facility-based healthcare. More recently, the US Centers for Disease Control and Prevention (CDC) launched Together TakeMeHome, a program that aims to distribute up to 1 million free HIVSTs between 2022–2027 to US populations nationwide¹⁶. During the first twelve months of distributing HIVSTs, nearly half of HIVST orders were placed by persons who had never tested for HIV (24%) or who had not tested in the last year (25%)¹⁷.

Online recruitment using culturally tailored marketing has been proposed to increase the distribution of HIVSTs to MSM^18,19. We previously reported that an intervention (Evaluation of HIV Self-Testing Among MSM Project [eSTAMP]) providing HIVSTs to online-recruited MSM increased both HIV testing and diagnoses compared to MSM who did not receive HIVSTs, with no reported risk compensation in those who received HIVSTs¹². However, BMSM and HLMSM were under-enrolled relative to their population proportions in this study, limiting generalizability to these groups.

Subsequently, the Implementation of Rapid HIV Self-Testing Among MSM Project (iSTAMP) was developed to determine best practices, effectiveness, costs, and cost-effectiveness of online recruitment to distribute HIVSTs to BMSM and HLMSM in 11 US states²⁰. We previously reported that iSTAMP was feasible, acceptable, and safe²¹. Here, we describe the cost and cost-effectiveness of iSTAMP in HIV self-testing among BMSM and HLMSM recruited through online dating and general interest sites.

To our knowledge, iSTAMP is the first report on the costs and cost-effectiveness of HIV self-testing via online recruitment of US BMSM and HLMSM. Recruiting BMSM and HLMSM via online sites is an effective and cost-effective approach to identifying infections in undiagnosed persons.

iSTAMP identified new diagnoses at relatively low cost ($2,337-$6,004/new diagnosis). These findings suggest that HIV self-testing via online recruitment may be highly cost-effective. Prior cost-effectiveness studies have estimated that a cost per new diagnosis could be as high as $35,732 (in 2022 US dollars) for a testing program to be cost-saving and $98,490 to be cost-effective in the US³⁰. Additionally, the cost per new diagnosis in iSTAMP is generally favorable compared to other US testing programs. For example, the eSTAMP HIV self-testing program, which primarily recruited white MSM, found a cost per new HIVST diagnosis of $17,455²⁵. The TakeMeHome HIV self-testing program reported a cost per confirmed new HIVST diagnosis of $10,288³¹. The CDC reports a range of cost per confirmed HIV diagnosis in other non-clinical settings (e.g., jails, offices, shelters, bathhouses, mobile vans, and community-based organizations) from $4,949-$29,753^32–35. Meanwhile, in clinical settings, costs per new confirmed diagnosis, ranging from $2,465-$12,975^36–41, are generally lower than for non-clinical settings. Of course, testing in non-clinical settings remains critical due to barriers to accessing clinical care, in particular those who may be at highest risk of HIV⁴².

The cost per HIVST used in iSTAMP ($274-$329/HIVST used) is higher compared with other HIV self-testing programs, but within the range of other US testing programs in non-clinical settings. For example, eSTAMP found a cost per person tested of $172²⁵. TakeMeHome reported a cost per person tested of $71³¹. In other non-clinical settings like mobile vans and community-based organizations, cost per person tested ranges from $197-$899^34,35. Differences may be driven by increasing costs of advertising over time as well as higher costs for the HIVST kits used in iSTAMP relative to other programs. Meanwhile, in clinical settings, where advertising and HIVST kits are not needed, costs per HIV test are generally lower, ranging from $24-$91^41,43–45. However, many patients with HIV risk factors face barriers to accessing services in clinical settings^46–48, as indicated by the fact that almost a quarter of our participants had never been HIV tested.

Online ad placement was one of the largest costs (33%–38%) and potentially the most uncertain cost of the study. The selected sites had differing ad display, targeting, and pricing options. Given that these options change often and the specific sites we employed may not be popular or even exist during future studies and programs, what follows are general implementation and cost considerations.

Different sites had different options to display ads and different cost structures. For example, some charged per screen takeover/pop-up (e.g., Jack’d, Grindr). Grindr also offered a more expensive inbox messaging option. Meanwhile, some sites had banner ad options and charged by cost per click or cost per impression (e.g., Snapchat, Grindr, Facebook, Reddit) or charged a fixed cost per posting (e.g., Adam4Adam). Some sites required a minimum spend (e.g., Adam4Adam). Facebook additionally had a bidding system to possibly have ads displayed at prime times. No site would cycle ads automatically, so each payment was for one ad copy.

Different sites also offered different options for ad targeting. Some allowed for no targeting at all, whereas others allowed targeting by gender (e.g., Facebook, Snapchat), location (e.g., Grindr), or proxy indicators of the target population (e.g., attendance at Historically Black Colleges and Universities; stated media, influencer, shopping interests) generally at no extra cost. Importantly, the landscape for ad targeting and pricing has and will likely continue to change. The European Union ruled in October 2023 that Meta must seek user consent before using their personal information to sell targeted ads on Facebook and Instagram⁴⁹, and in the US, Facebook no longer targets ads by sexual orientation.

Implementers of HIVST programs may benefit from lessons learned during our program. Given this evolving landscape, the implementers of online recruitment may need to be familiar with which sites are popular with their target population and what ad copy would work well for them. In our study, a community advisory board (CAB) was valuable to ensure ads were appropriate and appealing. To reach a diverse target population, implementers may need to diversify ads and platforms – for example, if we had only pursued recruitment via Jack’d because it was performing well among BMSM, we could have missed recruitment of HLMSM from Snapchat. Thus, cost is not the only consideration if trying to engage individuals from multiple target populations. Programs may consider regularly monitor site and ad copy performance to identify when they are not performing well. For example, Lesbian, Gay, Bisexual and Transgender interest sites not geared toward dating (Pride Media, Pink News) were initially used for recruitment but were stopped due to low yield. Monitoring demand is also important as too many ads can lead to too much demand, which could overload staff and reduce quality.

Test kits were the other large driver of program costs (32%–34%). We apply the current market cost of HIVST kits, but this could change or potentially be negotiated with bulk purchases. Additionally, program costs may vary by prior resource availability (e.g., images) or across different states. For example, we applied the estimated hourly wage for staff with an MPH in Georgia; the current equivalent salary in California is roughly double. The hours per activity shown in Appendix 2 may help implementers project costs for their context. Importantly, we used a time-motion study to more accurately estimate these hours by activity using a real-time tracker (Clockify) relative to studies which often use surveys, self-report, or expenditure records³².

The proportion of positive HIVSTs in iSTAMP was higher (6% general interest sites and 12% dating sites) than what has been seen in the TakeMeHome HIVST program (1.4%⁵⁰). TakeMeHome and Together TakeMeHome are large programs accessible to the general population, and the percentage of reactive tests is likely an underestimate due to low reporting of results (the report of results was not incentivized). Additionally, iSTAMP was restricted to two US populations most likely to have undiagnosed HIV infection (BMSM and HLMSM¹), and studies have shown that MSM using dating sites have a greater frequency of condomless sex, more STIs, and more sex partners^51–55. Finally, participants eligible for iSTAMP had to have engaged in anal sex with a man in the past year and not be on PrEP, two criteria which place them at higher HIV risk than the general population.

Our findings should be interpreted considering limitations. Because HIVST use is self-reported, it undercounts people who used a test but did not complete the survey (meaning ACER and ICER estimates are likely conservative, as indicated by sensitivity analyses). Some data could be misreported, especially SNA HIVST use and results (most participants reported their own HIVST result correctly based on the images provided²¹). Again, we expect that any information bias related to reporting would underreport outcomes, leading to more conservative estimates. Costs and cost-effectiveness estimates are based on online recruitment of BMSM and HLMSM living 11 states, limiting generalizability of the findings. Additionally, our analysis of the costs or cost-effectiveness estimates do not consider longer-term effects (e.g., HIV infections averted due to linkage to ART or PrEP) and health consequences (e.g., quality-adjusted life-years) – incorporation of longer time horizons may make the strategies cost-saving. The time and motion study was conducted after iSTAMP was running efficiently, and the average number of participants enrolled during this time period was similar to that of the overall program, both of which lends support that the time and motion data are generally representative of the overall program. However, these data may not exactly quantity staff time for the entire program, which is not expected to have a substantive impact on the results given that staff time was a relatively small cost of the program. We acknowledge that our findings are less directly comparable to studies that include costs of confirmatory testing, though we note that confirmatory testing costs only apply to those with a positive HIVST result and therefore have minimal influence on overall program costs. Some HIVST test results may be incorrect which could affect downstream costs related to linkage to care and confirmatory testing. Finally, our analyses were conducted from the healthcare provider’s perspective, and therefore participant costs are not included. However, it is unlikely that participant time, primarily time spent taking and returning the HIVST, would differ across online recruitment venues, and thus we would not expect any impact on the ICER estimates. Participant costs and time spent on HIV self-testing is likely lower than participant time spent seeking HIV testing in a clinical setting.

HIV self-testing through dating and general interest site recruitment may be an effective and cost-effective approach to enhance HIV testing and increase new HIV diagnoses. Sites geared toward dating may reach people with HIV risk factors and could comprise a useful component of an online HIV self-testing recruitment strategy, though multiple platforms may be needed to reach all members of the target population. Our findings inform efficient implementation of online recruitment strategies to engage US BMSM/HLMSM in HIV self-testing. Future studies of linkage to care after HIVST are needed.