Point 'Oh!' Five


Posts in Publications
Stigma, the Media, and Pre-Exposure Prophylaxis for HIV Prevention: Observations for Enhancing Knowledge Translation and Resisting Stigma in the Canadian Context

Drafted Version; Final version available at AIDS and Behaviour.

ABSTRACT (140/150)                                                                                                                     

Pre-Exposure Prophylaxis (PrEP) is an effective, though sometimes stigmatized, strategy for HIV prevention. With the goal of examining how PrEP stigma can be addressed, this study examined the media’s handling of stigma related to PrEP by searching the Canadian Newsstream and Daily Xtra news databases for key terms related to PrEP. Overall, 101 media articles were thematically coded in triplicate; 36.3% of which included mentions of PrEP stigma. LGBT media sources were more likely than mainstream sources to have included content coded as relating to PrEP stigma (p = 0.02). In these articles, uncertainty regarding PrEP, and neo-liberal attitudes towards sexual responsibility were major factors associated with media discussion of PrEP stigma. We discuss the role that heuristics play in shaping lay readers perceptions and interpretation of PrEP media coverage and discuss methods for overcoming stigma using evidence-based communication strategies.


Pre-exposure prophylaxis (PrEP) with Tenofovir and Emtricitabine is an effective HIV prevention strategy [1–5] supported by a number of clinical trials which together demonstrate a strong dose-response relationship between PrEP adherence and reduction in HIV transmission [6–8]. At a population-level, network models suggest that PrEP can reduce HIV incidence even in face of rising risk compensation and declining condom use [9,10]. However, access to PrEP is largely dependent on prescription drug coverage; and while some Canadian provinces have added PrEP to provincial drug coverage plans, in most settings PrEP remains uncovered by either provincial or private drug insurers [11,12].

Among several potential barriers to expanding PrEP access, stigma towards PrEP use has been regularly reported [13–15]. Conceptually, stigma can be described as a form of social control in which particular attitudes, behaviors, or characteristics are devalued, treated with contempt by others, or used as a form of social distinction. Theoretical discussions of stigma distinguish between felt stigma (i.e., perceived or anticipated stigma), enacted stigma (i.e., expressed discrimination), and internalized stigma (i.e., incorporation of stigma into one’s own beliefs about oneself) [16]. When these stigmas are pervasive in a society and begin to shape social policy, negative social control can also be described as “structural stigma.” In any case, stigma gives rise to normal and non-normal patterns of behavior or identity-formation [17], and, in turn, these patterns have the potential to negate the effectiveness of otherwise promising prevention strategies such as PrEP [18,19]. Each of these delineations highlight the ways that stigma can be manifest to the detriment of stigmatized populations, even when actual experiences of enacted stigma are rare. With respect to PrEP, the pervasiveness of social stigma has hindered the expansion of PrEP coverage by supporting evidence-neutral health policies (i.e., enacted and structural stigma) and by discouraging the widespread uptake of PrEP (i.e., internalized and felt stigma) [13].

Articulating the processes by which stigma is perpetuated, multiple related theories of stigma and risk perception, highlight the role of the media (both the news media and social media) in initiating, perpetuating, and maintaining social stigma [20,21]. These media-conscious approaches highlight how even minor risks have the potential to become embroiled in long standing ideological controversies [22,23] and play a role in communicating stigma within and between social networks and communities [24]. More broadly, the media has been shown to bear significant influence over what their audience is aware of, what their opinions are, and how they behave [25]. For example, successful mass media campaigns have resulted in changes to smoking behavior in jurisdictions across North America [26]. Other examples of the media’s impact on health have been previously described with respect to its role in shaping the public’s perceptions of HIV [27], mental illness [28,29], vaccines [30], genetically modified foods [31], and climate change [32,33]. Indeed, in each of these cases, the media has contributed to significant confusion, misunderstanding, and stigmatization even in the face of near-scientific consensus on these issues [34–36]. Furthermore, the media is understood to exert considerable political and ideological control by framing issues and setting policy agendas [37] – thus underscoring its importance to promoting new policy-based interventions, such as those regarding PrEP.

Ironically, the media also plays an important role in shaping social discourse about – not only scientific innovations themselves – but also the stigma surrounding these innovations. This is particularly important when considering the media’s role in perpetuating felt and internalized stigma such as with respect to PrEP. Indeed, regarding HIV-stigma, research from as early as the 1990’s shows that individuals are prone to overestimate the degree to which HIV is stigmatized [38]. Considering this, we hypothesize that media portrayals of stigma may have the unintended effect of reinforcing felt and internalized stigma. Therefore, we sought to identify news articles related to PrEP, with a focus in the present article on PrEP stigma.


Data Collection

To systematically sample Canadian media coverage of PrEP, two national news databases – Canadian Newsstream and Daily Xtra – were searched in January 2017 for key terms related to PrEP (i.e., PrEP, Pre-Exposure Prophylaxsis, Truvada, Tenofovir disoproxil, HIV Medication, HIV Drug, HIV Treatment). Articles included in our search were published between 2008 and 2016, as this sampling frame was inclusive of the earliest mentions of PrEP in the media until the time the study was conducted. The selected sources were chosen to capture a systematic subset of both mainstream and gay news across Canada. From the keyword search, a total of 3,020 search results were reviewed by three trained reviewers with the aim of identifying PrEP-related news articles. Most articles identified by our search strategy were not related to PrEP.

Thematic Coding

Aiming to identify themes covered in relevant articles, analyses were conducted using an inductive thematic approach (ITA) [39] wherein a code-book was collaboratively generated and validated against an iterative review of articles by paired coders. Inductive thematic analysis was selected as we expected that new coverage would cover a variety of themes not related to any specific pre-established theory. As such, ITA is similar to so called “grounded theory approaches” with the exception that the latter is applied in the context of theory development, while in the present study was pre-occupied with identifying what role, if any, the media plays in originating, facilitating, and perpetuating stigma towards PrEP use. Consistent with ITA protocol [40], codes and coding practices were refined until consensus was reached between all three coders using a test-sample of 10 articles. When finalized, the code-book contained nine codified themes focusing on PrEP portrayal, regulation, efficacy, awareness, side-effects, accessibility, adherence, gender-issues, and stigma. Final coding for each article was jointly reviewed and adjudicated to ensure consistency with the codified definitions for each theme. In the second stage of theme development, we reviewed articles with coded material (i.e., expressions) related to stigma. A subset of inductively defined themes was then developed examining (i) sources of stigma (i.e., friends and partners), (ii) the underlying rationale for stigma (e.g., personal responsibility), and (iii) rhetorical strategies to react to PrEP stigma (e.g. appeal to authority or data).

Quantitative Analysis

All quantitative analyses were conducted in R [41]. Descriptive statistics were stratified by the type of media (i.e., LGBT or mainstream) the article was published in and linear and exponential regression models examined trends in frequency of themes over time. Regression coefficients were compared to assess differences in trends between LGBT and mainstream media sources [42]; and student’s t-tests were used to examine whether some themes were more likely in LGBT media than in mainstream media. Linear regression models tested whether the relative proportion of articles addressing each theme changed over time. Phi correlation coefficients were calculated to examine the intercorrelations between the coded themes.


Descriptive Results

Out of an initial 3,020 search results, a total of 101 media articles were coded. Most articles provided a description of PrEP (78.2%) and discussed regulatory implications (53.5%). Other important topics included efficacy (48.5%), accessibility (46.5%), and stigma (35.6%). Meanwhile, a relatively low proportion of articles discussed awareness (20.8%), side effects (25.7%), and adherence (28.7%). LGBT media articles were more likely than mainstream media articles to have included content coded as relating to awareness (p < 0.01), accessibility (p = 0.02), and stigma (p = 0.02). There were no other statistically significant differences between LGBT and mainstream media. Within LGBT media sources, themes for accessibility and regulation were correlated (p < 0.01), as were themes for adherence and side-effects (p < 0.01), and accessibility and awareness (p = 0.01). Within mainstream media articles, the description and awareness themes (p = 0.03), regulations and accessibility themes (p = 0.03), and side effects and stigma themes (p = 0.04) were correlated. Summarizing the number of stigma-related articles across time, the share of these articles that included stigma coding, and the average word count dedicated to stigma-related themes across time. , we note that stigma has increasingly become a dominant topic in PrEP journalism and that the average number of words addressing PrEP stigma significantly increased in the wake of early media activism by PrEP skeptics [43]. Notably in 2014, there was a dramatic increase in the average word count related to stigma themes potentially coinciding with the endorsement of PrEP by the U.S. Center for Disease Control and Prevention [44]. Overall, there was an exponential increase in the number of articles published over time (p < 0.001). However, comparing LGBT and mainstream media sources, there was no difference in trends (p = 0.78). Overall, the relative proportion of articles addressing each theme was stable. Indeed, only the description theme decreased in frequency over time (p = 0.012).

Thematic Focus: PrEP Stigma

Sources of PrEP Stigma. Overall, stigma was primarily discussed in generic terms (n = 6), though specific sources of stigma included health care providers (n = 4), friends and partners (n = 4), and the media itself (n = 1). For instance, health care providers were characterized as being “disappointed” in their patients who had “given up on condoms,” opting for PrEP instead [45–47]. Regarding PrEP access one advocate commented that

"the process of getting a prescription for Truvada as PrEP can seem stigmatizing because it's only made available to men who identify themselves as people who don't use condoms systematically and have more than one sexual partner."  [45]

Outside the medical establishment, people using PrEP were also characterized as facing stigma from their friends and sexual partners [45,48–50]. For instance, one gay man was quoted as saying that his friends called him a “sex addict” and “a whore” when they found out he was taking PrEP [45] and another PrEP advocate stated that he believed “negative media coverage of the drug” contributed to low PrEP uptake [51]. Providing evidence for PrEP stigma, four articles made note of the term “Truvada whore” which was initially used to critique PrEP, but quickly became a badge of honor: “#TruvadaWhore” [46,52–54]. Six articles also cited instances of PrEP being referred to as a “party drug” [44,55–60] – a reference to mixing Truvada with traditional sex drugs – “a combo known as ‘MTV’” [61].

Writers also represented PrEP skepticism as being primarily concerned with worries that PrEP would encourage "risky and irresponsible behaviour" such as condomless anal sex [56]. Indeed, 26 of the studies related to PrEP stigma referenced, either directly (n = 4) or indirectly, the phenomena of risk compensation. Of primary concern was the impact that PrEP would have on condom abandonment (n = 22) and promiscuity (n = 7).  In context of an overwhelming focus on behavior, only 5 articles linked stigma specifically to worries that PrEP would contribute to increased HIV or STI infections. Other PrEP skeptics went so far as to say that “Truvada is for cowards” — emphasizing a priori judgments about the moral character of people on PrEP [54]. Indeed, questions like “Why can’t people behave themselves?” (emphasis added; 47), provide an example of the implicit and explicit assumptions made about promiscuity and behaviors which prioritize personal pleasure.

Impact of PrEP Stigma. The most commonly reported impact of PrEP stigma was its role as a barrier to PrEP uptake. This was, despite a common acknowledgement that PrEP stigma was an implicit sub-category of HIV-stigma or sex-negativity. Illuminating the negative impact of stigma, one health care provider questioned why he would proscribe PrEP when his patients already have "highly effective tools" available to them [58]. Another man on PrEP noted that stigma, not only meant his partners were more willing to engage in condomless sex, but that sometimes there was an “expectation” for it – leading to “some awkward situations” [49]. Finally, PrEP was also framed as a linchpin in dividing the gay community, forcing individuals to choose “for PrEP or against,” as one advocate put it [48].

Responses to PrEP Stigma. Seeking to address PrEP stigma, writers and PrEP advocates alike sought to undermine PrEP skepticism through the media. For instance, one PrEP skeptic was quoted as saying that he felt PrEP was "irresponsible" but could not explain why beyond "it's just what he feels" [63]. This framing of PrEP skepticism as naïve or values-driven was widespread. For example, one PrEP advocate was quoted as likening the battle for PrEP to the battle for birth control: "There was a value judgement attached,” he concluded [64]. PrEP advocates, on the other hand, were often depicted as separating "feelings from the actual facts" and were poised to call out “false arguments” against PrEP [50,65]. For instance, in the following excerpt a writer backs up a PrEP advocate’s urging for evidence-based PrEP policy:

“‘It is critical that PrEP access be governed by science and not by personal values,’ Calabrese rightfully claims. This is particularly true when the goal should be to end the epidemic.” (Emphasis added, 30).

However, despite an overwhelming majority of the included articles being supportive of PrEP and despite the general acknowledgement that “we cannot be judgmental [about PrEP] " [66], many of the arguments supporting PrEP skepticism were left unanswered or were even subtly reinforced. For instance, one writer commented,

“Most of the Canadians I spoke to for this story — on and off the record, inside and outside the AIDS establishment — are to some degree hesitant" [67].

More specifically, when writers represented the positions of academic and clinical professionals, experts were often portrayed as cautious regarding the potential impact, limitations, and side-effects of PrEP. This was particularly true when considering the role that PrEP might play in risk compensation. For instance, health care providers were said to be concerned that prescribing PrEP would contribute to "unsafe sex" or that it would give "patients a false sense of confidence" [47]. Further, writers and experts alike, sought to provide fair and balanced coverage, often leading them to rebut their own rebuttals to PrEP skepticism. For instance, in the following example a writer notes that one group of researchers did not find evidence for risk compensation, but in the next sentence they use a direct quote from these researchers to undermine their own argument:

“Goicochea says some critics have expressed concern that people taking the drug will have sex more often and freely and engage in unsafe sex leading to higher instances of other sexually transmitted infections. But, he says, condom use actually increased during the study. ‘But of course, this is under the conditions of the clinical trial with monthly visits. So participants were consulted and given condom supplies on a monthly basis,’ he says” [68].

However, there were also several examples where writers did make use of forceful quotes, such as the one below, in an attempt to resolve ambiguity and doubt:

“You can worry all you want, but once the evidence is in, you cannot deny it. This is based on science, and the science says that Truvada taken as prophylaxis is effective, it’s safe, and complications are extremely rare.” [47].


Primary Findings

In the present study, we reviewed a systematically sampled subset of news media articles related to stigma throughout the early emergence of PrEP in Canada and inductively developed three themes which identified sources of stigma, the underlying rationale for stigma, and the rhetorical strategies imbedded within the media’s portrayal of stigma. In doing so, we note that our findings are easily contextualized within a growing body of literature that highlights the media’s role in shaping not only their audience’s awareness and knowledge of given health topics, but also their attitudes towards the subject matter covered [69–71]. Within this literature, it has become apparent that the media has, at times, become, even if unwittingly, instrumental to the spread of stigma and prejudice [72]. Examples of this include the media’s role in the development of stigma towards mental illness [29], body weight [73], sexuality [74,75], and HIV [76].

Consistent with these findings, we found that even though our news articles framed stigma as problematic and identified common sources of PrEP-related stigma, these articles did not necessarily work to undermine PrEP stigma. This is particularly worrisome given that negative portrayal can contribute to stigma and stigma can, in turn, negate the efficacy of PrEP by tying it to rejected stereotypes, behaviors, or identities [17,24]. In the present study, the most commonly identified rationale for stigma was scientific uncertainty regarding the effectiveness of PrEP and the potential for risk compensation. This is consistent with previous research which has shown that scientific uncertainty was a common theme in U.S. News coverage, especially prior to the endorsement of PrEP by the U.S. Center for Disease Control and Prevention [77]. So, while writers provided persuasive rebuttals to many commonly reported worries of PrEP skeptics, the out-sized focus on dissenting voices gave considerable weight to these arguments, even if unintentionally. By doing so, news media coverage of PrEP provides subtle reinforcement of arguments that do not necessarily represent informed scientific consensus. Indeed, this problem has been regularly documented with respect to other scientific topics such as vaccines, genetically modified foods, and climate change [30]. Given previous misreporting in the media regarding PrEP [78,79] and risk perception theories that emphasize the difficulty of disseminating politicized findings [21,22], our assessment of historical and current media coverage supports the need for greater cooperation between academic researchers and news writers to correct common misconceptions and negative assumptions about PrEP. In practice, this means that researchers should (a) invest time in working with journalists to communicate findings and participate in public conversations, (b) make themselves available to work with journalists by building relationships with the respective writers and editors covering their research area, (c) gain sufficient training to communicate effectively with lay audiences, and (d) understand and support the journalists duty to ensure that content is both relevant to readers and newsworthy [80].

More broadly, our study highlights how the emergence of new prevention strategies presents academic, policy, and journalistic institutions with a serious challenge – especially in the face of already inflicted damage. Primarily, our review raises the question of how to promote potentially controversial interventions while hoping at the same to communicate caution with respect to emergent scientific and biomedical innovations. This challenge of communicating uncertainty and risk while at the same time not destabilizing public confidence in scientific consensus is an increasingly important area of research – one that requires the support of academics, clinicians, policy leaders, and journalists [81].

Underscoring this challenge, policy makers are increasingly subjected to democratized decision-making and must, therefore, navigate public policy decisions with careful attention to the mediating influence of news coverage and commentary [37]. This is particularly true given the media’s documented influence over agenda-setting, framing, and priming — leading some political scientists to describe media outlets as bona fide policy actors [82,83]. Thus, media coverage plays a central role in determining the policy implications for stigmatized and politicized public health interventions, such as PrEP [84].

One strategy to help lay audiences make accurate judgements about PrEP is to address the heuristic processes that govern the way information is perceived and processed. One of the common heuristic devices used in media coverage is to appeal to expert opinion [35]. However, communications research shows that such appeals can be executed on a sliding scale of effectiveness, especially with respect to already controversial topic areas. This is because individuals distinguish between so-called expert opinion and their own personal views when forming judgements [85]. With that said, a growing number of studies suggest that the weight-of-evidence (i.e., the certainty of consensus in the scientific community) does in fact persuade individuals to agree with expert opinion [35]. This suggests that, whenever possible, expert opinion should be presented within the broader context of existing evidence – and that lay readers should not be asked to rely on only the viewpoint of a single practitioner (a common practice in person-driven pieces).

A second strategy to improve lay judgement of scientific coverage is to minimize false equivalencies. While there is an obvious and well-meaning desire for journalists to give equal weight to each perspective, available research suggests that this style of reporting effectively undermines scientific consensus and distorts lay people’s ability to accurately understand what is being conveyed [86]. Similarly, a number of studies have shown that contrasting-view narratives increase uncertainty in lay-readers – highlighting single-view narratives as more effective in conveying important health information [87]. With that said, writers should not attempt to resolve scientific issues before scientific consensus is reached. Communications research has shown that providing evidentiary balance (i.e., acknowledging limitations of current research, avoiding personal predictions, and explaining next steps to fill in knowledge gaps) increases lay trust in the scientific process [88,89] – thus strengthening the weight of consensus once it is achieved.

While not common practice for all media outlets, linking to scientific articles and providing measures of uncertainty (e.g., margins of error) can also improve trust and increase the perceived credibility of news articles [81]. As noted earlier, researchers should also seek to develop working relationships with policy makers and journalists who regularly discuss and report on their research areas. This allows for smoother dissemination and facilitates better reporting practices, relieving journalists and policy makers of the burden for developing appropriate error estimates and providing accessible content such as lay summaries or infographics [90,91]. However, journalists should be cautious when conveying news on behalf of researchers in areas where broader scientific consensus has not been achieved; and policy makers should take care to minimize the political function of emerging scientific research until sufficiently broad evidence is available to support policy action. Doing so will reduce the political utility of interventions by resigning scientific uncertainty to the appropriate academic and scientific channels where it can best be adjudicated [92,93]. Conversely, when scientific consensus has been reached, it is important to identify such consensus as news-worthy, particularly if previous reports conveyed doubt. Applied beyond PrEP, such action might be taken with respect to preventing stigma against people living with HIV by accurately portraying the scientific consensus surrounding viral load suppression and undetectability – namely, that people living with HIV cannot pass on the virus if their viral load is suppressed or undetectable [94,95].


Regarding the limitations of the present study, readers should be aware that some relevant media articles may not have been captured in our sampling process. Indeed, while news databases provide an expansive and powerful source for examining new coverage, indexing limits make it difficult for any single database to accurately and completely catalog relevant news media. Further, as social media has come to play an increasingly important role in media dissemination [96–98], research is urgently needed to assess how traditional media and scientific research is disseminated via these media and how public health leaders can influence these newly democratized outlets for knowledge translation.


In conclusion, the present study demonstrates the relationship between scientific uncertainty in emergent prevention strategies (e.g., PrEP) and the stigmatization that occurs as the public attempts to incorporate scientific innovation into existing ideological frameworks (e.g., neoliberalism). Considering this with respect to communications research, we argue that reporting standards developed for knowledge translation must account for the heuristic processes of readers – especially with respect to the presentation of expert opinion and scientific consensus.  This is the responsibility of both academics, who provide source quotes and scientific guidance, and journalists, who convey these messages to the public. We, therefore, conclude that enhancing cooperation between these two actors is paramount to reducing stigma and misinformation in scientific reporting.


1. Baeten JM, Donnell D, Ndase P, Mugo NR, Campbell JD, Wangisi J, et al. Antiretroviral Prophylaxis for HIV-1 Prevention among Heterosexual Men and Women. N Engl J Med. 2012;367:399–410.

2. Choopanya K, Martin M, Suntharasamai P, Sangkum U, Mock PA, Leethochawalit M, et al. Antiretroviral prophylaxis for HIV infection in injecting drug users in Bangkok, Thailand (the Bangkok Tenofovir Study): a randomised, double-blind, placebo-controlled phase 3 trial. Lancet Lond Engl. 2013;381:2083–90.

3. Grant RM, Lama JR, Anderson PL, McMahan V, Liu AY, Vargas L, et al. Preexposure Chemoprophylaxis for HIV Prevention in Men Who Have Sex with Men. N Engl J Med. 2010;363:2587–99.

4. Grant RM, Anderson PL, McMahan V, Liu A, Amico KR, Mehrotra M, et al. Uptake of pre-exposure prophylaxis, sexual practices, and HIV incidence in men and transgender women who have sex with men: a cohort study. Lancet Infect Dis. 2014;14:820–9.

5. McCormack S, Dunn DT, Desai M, Dolling DI, Gafos M, Gilson R, et al. Pre-exposure prophylaxis to prevent the acquisition of HIV-1 infection (PROUD): effectiveness results from the pilot phase of a pragmatic open-label randomised trial. Lancet Lond Engl. 2016;387:53–60.

6. Fonner VA, Dalglish SL, Kennedy CE, Baggaley R, O’Reilly KR, Koechlin FM, et al. Effectiveness and safety of oral HIV preexposure prophylaxis for all populations. AIDS Lond Engl. 2016;30:1973–83.

7. Hendrix CW. Exploring concentration response in HIV pre-exposure prophylaxis to optimize clinical care and trial design. Cell. 2013;155:515–8.

8. Okwundu CI, Uthman OA, Okoromah CA. Antiretroviral pre-exposure prophylaxis (PrEP) for preventing HIV in high-risk individuals. Cochrane Database Syst Rev. 2012;CD007189.

9. Jenness SM, Sharma A, Goodreau SM, Rosenberg ES, Weiss KM, Hoover KW, et al. Individual HIV Risk versus Population Impact of Risk Compensation after HIV Preexposure Prophylaxis Initiation among Men Who Have Sex with Men. PloS One. 2017;12:e0169484.

10. Jenness SM, Weiss KM, Goodreau SM, Gift T, Chesson H, Hoover KW, et al. Incidence of Gonorrhea and Chlamydia Following HIV Preexposure Prophylaxis among Men Who Have Sex with Men: A Modeling Study. Clin Infect Dis Off Publ Infect Dis Soc Am. 2017;

11. CBC News. Ontario to cover HIV prevention pill under public health plan [Internet]. CBC News. 2017 [cited 2017 Oct 5]. Available from: http://www.cbc.ca/news/health/hiv-prep-coverage-1.4302184

12. Bell N. Want insurance to cover your PrEP? Good luck. Xtra [Internet]. 2016 Feb 18 [cited 2017 Oct 5]; Available from: https://www.dailyxtra.com/want-insurance-to-cover-your-prep-good-luck-70225

13. Calabrese SK, Underhill K. How Stigma Surrounding the Use of HIV Preexposure Prophylaxis Undermines Prevention and Pleasure: A Call to Destigmatize “Truvada Whores.” Am J Public Health. 2015;105:1960–4.

14. Freeborn K, Portillo CJ. Does Pre-exposure prophylaxis (PrEP) for HIV prevention in men who have sex with men (MSM) change risk behavior? A systematic review. J Clin Nurs. 2017;

15. Hubach RD, Currin JM, Sanders CA, Durham AR, Kavanaugh KE, Wheeler DL, et al. Barriers to Access and Adoption of Pre-Exposure Prophylaxis for the Prevention of HIV Among Men Who Have Sex With Men (MSM) in a Relatively Rural State. AIDS Educ Prev Off Publ Int Soc AIDS Educ. 2017;29:315–29.

16. Herek GM, Gillis JR, Cogan JC. Internalized stigma among sexual minority adults: Insights from a social psychological perspective. J Couns Psychol. 2009;56:32–43.

17. Goffman E. Stigma: Notes on the Management of Spoiled Identity. Reissue edition. New York: Touchstone; 1986.

18. Morgan E, Ryan DT, Newcomb ME, Mustanski B. High Rate of Discontinuation May Diminish PrEP Coverage Among Young Men Who Have Sex with Men. AIDS Behav. 2018;

19. Van der Elst EM, Mbogua J, Operario D, Mutua G, Kuo C, Mugo P, et al. High Acceptability of HIV Pre-exposure Prophylaxis but Challenges in Adherence and Use: Qualitative Insights from a Phase I Trial of Intermittent and Daily PrEP in At-Risk Populations in Kenya. AIDS Behav. 2013;17:2162–72.

20. Douglas M. Risk Acceptability According to the Social Sciences. Psychology Press; 2003.

21. Kasperson RE, Renn O, Slovic P, Brown HS, Emel J, Goble R, et al. The Social Amplification of Risk: A Conceptual Framework. Risk Anal. 1988;8:177–87.

22. Douglas M, Calvez M. The self as risk taker: a cultural theory of contagion in relation to AIDS. Sociol Rev. 1990;38:445–64.

23. Masuda JR, Garvin T. Place, culture, and the social amplification of risk. Risk Anal Off Publ Soc Risk Anal. 2006;26:437–54.

24. Smith R. Language of the Lost: An Explication of Stigma Communication. Commun Theory. 2007;17:462–85.

25. Wakefield MA, Loken B, Hornik RC. Use of mass media campaigns to change health behaviour. Lancet. 2010;376:1261–71.

26. Bala MM, Strzeszynski L, Topor-Madry R. Mass media interventions for smoking cessation in adults. Cochrane Database Syst Rev. 2017;11:CD004704.

27. Mykhalovskiy E, Hastings C, Sanders C, Hayman M, Bisaillon L. Callous, cold, and deliberately duplicitous: racialization, immigration and the representation of HIV criminalization in Canadian Mainstream Newspapers. Toronto; 2016 Nov.

28. Klin A, Lemish D. Mental disorders stigma in the media: review of studies on production, content, and influences. J Health Commun. 2008;13:434–49.

29. Benbow A. Mental illness, stigma, and the media. J Clin Psychiatry. 2007;68 Suppl 2:31–5.

30. Dixon GN, Clarke CE. Heightening Uncertainty Around Certain Science: Media Coverage, False Balance, and the Autism-Vaccine Controversy. Sci Commun. 2013;35:358–82.

31. Frewer LJ, Miles S, Marsh R. The Media and Genetically Modified Foods: Evidence in Support of Social Amplification of Risk. Risk Anal. 2002;22:701–11.

32. Anderson A. Media, Politics and Climate Change: Towards a New Research Agenda. Sociol Compass. 2009;3:166–82.

33. Renn O. The social amplification/attenuation of risk framework: application to climate change: Social amplification/attenuation of risk framework. Wiley Interdiscip Rev Clim Change. 2011;2:154–69.

34. Corbett JB, Durfee JL. Testing Public (Un)Certainty of Science: Media Representations of Global Warming. Sci Commun. 2004;26:129–51.

35. Dunwoody S, Kohl PA. Using Weight-of-Experts Messaging to Communicate Accurately About Contested Science. Sci Commun. 2017;39:338–57.

36. Parkinson S, Jackson T, Berg K, editors. Risk, Media and Stigma: Understanding Public Challenges to Modern Science and Technology. London ; Sterling, VA: Earthscan; 2001.

37. Bennett WL, Entman RM, editors. Mediated Politics: Communication in the Future of Democracy. 1 edition. Cambridge, UK ; New York: Cambridge University Press; 2000.

38. Green G. Attitudes towards people with HIV: are they as stigmatizing as people with HIV perceive them to be? Soc Sci Med 1982. 1995;41:557–68.

39. Bernard H. Analyzing Qualitative Data: Systematic Approaches. Ryan G, editor. Los Angeles Calif.: Sage Publications; 2009.

40. Braun V, Clarke V. Using thematic analysis in psychology. Qual Res Psychol. 2006;3:77–101.

41. R Core Team. R: A language and environment for statistical computing. [Internet]. 2017. Available from: http://www.R-project.org/

42. Clogg CC, Petkova E, Haritou A. Statistical Methods for Comparing Regression Coefficients Between Models. Am J Sociol. 1995;100:1261–93.

43. Associated Press. Divide over HIV prevention drug Truvada persists. USA TODAY [Internet]. 2014 Apr 6 [cited 2017 Sep 14]; Available from: https://www.usatoday.com/story/news/nation/2014/04/06/gay-men-divided-over-use-of-hiv-prevention-drug/7390879/

44. Barsotti N. US: CDC recommends expanded use of PrEP. Xtra [Internet]. 2014 May 15 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/us-cdc-recommends-expanded-use-of-prep-60661

45. Christopher N. Coming out of the PrEP closet. Xtra [Internet]. 2015 Dec 1 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/coming-out-of-the-prep-closet-69540

46. Miksche M. Can we end the North American HIV epidemic? Xtra [Internet]. 2016 Nov 7 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/can-we-end-the-north-american-hiv-epidemic-72327

47. Bell N. PrEP is now approved in Canada. What happens now? Xtra [Internet]. Vancouver, Canada; 2016 Mar 2 [cited 2017 Sep 13]; Available from: https://www.dailyxtra.com/prep-is-now-approved-in-canada-what-happens-now-70344

48. Miksche M. Waiting in purgatory (Part 1). Xtra [Internet]. 2015 Mar 11 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/waiting-in-purgatory-part-1-66665

49. Miksche M. Why I had to stop taking PrEP. Xtra [Internet]. 2016 Mar 14 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/why-i-had-to-stop-taking-prep-70437

50. Miksche M. What the rise of the alt-right means for HIV prevention. Xtra [Internet]. 2016 Nov 21 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/what-the-rise-of-the-alt-right-means-for-hiv-prevention-72462

51. Cruikshank J. An ounce of prevention. Xtra [Internet]. Vancouver, Canada; 2014 Sep 25 [cited 2017 Sep 13]; Available from: https://www.dailyxtra.com/an-ounce-of-prevention-63917

52. Miksche M. PrEP and the birth control pill. Xtra [Internet]. 2016 Oct 11 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/prep-and-the-birth-control-pill-72142

53. Glenwright D. For the love of fucking [Internet]. Xtra. 2014 [cited 2017 Sep 14]. Available from: https://www.dailyxtra.com/for-the-love-of-fucking-57930

54. Gilbert S. What they don’t want you to know about Truvada [Internet]. Xtra. 2015 [cited 2017 Sep 14]. Available from: https://www.dailyxtra.com/what-they-dont-want-you-to-know-about-truvada-67669

55. Houston A. Concerns raised about HIV pill study. Xtra [Internet]. 2010 Nov 22 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/concerns-raised-about-hiv-pill-study-8646

56. Miksche M. How bareback culture has evolved since the AIDS epidemic. Xtra [Internet]. 2016 Dec 12 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/how-bareback-culture-has-evolved-since-the-aids-epidemic-72629

57. Miksche M. A safer way to party and play? [Internet]. Xtra. 2016 [cited 2017 Sep 14]. Available from: https://www.dailyxtra.com/a-safer-way-to-party-and-play-71310

58. Prendergast F. If PrEP is a party drug, then cue the DJ. Xtra [Internet]. 2014 Apr 29 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/if-prep-is-a-party-drug-then-cue-the-dj-60116

59. Miksche M. Why the AIDS vigil should remind communities to work harder. Xtra [Internet]. 2016 Jul 1 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/why-the-aids-vigil-should-remind-communities-to-work-harder-71420

60. Hunter P. Party pill or drug of hope?: Increasingly, the drug Truvada is being used to prevent HIV transmission. But some fear it will encourage unsafe sex. Tor Star Tor Ont. Toronto, Ont., Canada; 2014 Nov 23;IN.1.

61. Ren P. Can a pill prevent HIV? Xtra [Internet]. 2009 Jan 28 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/can-a-pill-prevent-hiv-13617

62. LGBTQ RIGHTS; Community fights stigma; Event to promote open dialogue on race, sex, HIV status, experience of young people. Chron - Her Halifax NS. Halifax, N.S., Canada; 2016 Mar 26;A6.

63. Miksche M. Getting the facts right about PrEP, serosorting and your sexual health. Xtra [Internet]. 2016 Oct 24 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/getting-the-facts-right-about-prep-serosorting-and-your-sexual-health-72243

64. Ellis E. Insurer stops covering HIV prevention drug. Vanc Sun [Internet]. 2016 May 6 [cited 2017 Sep 14]; Available from: http://vancouversun.com/news/local-news/insurer-stops-covering-hiv-prevention-drug

65. Syms S. Got PrEP? Xtra [Internet]. 2008 Jan 30 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/got-prep-38667

66. Barsotti N. Almost half of gay men bareback: study. Xtra [Internet]. 2013 Jan 14 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/almost-half-of-gay-men-bareback-study-1568

67. McCann M. Canadians already on PrEP while drug sits in regulatory limbo. Xtra [Internet]. 2014 Feb 7 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/canadians-already-on-prep-while-drug-sits-in-regulatory-limbo-57946

68. Christopher N. US approves drug to prevent HIV infection. Xtra [Internet]. 2012 Jul 19 [cited 2017 Sep 14]; Available from: https://www.dailyxtra.com/us-approves-drug-to-prevent-hiv-infection-3110

69. Lecheler S, de Vreese CH. News Framing and Public Opinion: A Mediation Analysis of Framing Effects on Political Attitudes. Journal Mass Commun Q. 2012;89:185–204.

70. Gunther AC, Christen CT. Effects of News Slant and Base Rate Information on Perceived Public Opinion. Journal Mass Commun Q. 1999;76:277–92.

71. Hanitzsch T. Deconstructing Journalism Culture: Toward a Universal Theory. Commun Theory. 2007;17:367–85.

72. Betton V, Borschmann R, Docherty M, Coleman S, Brown M, Henderson C. The role of social media in reducing stigma and discrimination. Br J Psychiatry J Ment Sci. 2015;206:443–4.

73. Flint SW, Nobles J, Gately P, Sahota P. Weight stigma and discrimination: a call to the media. Lancet Diabetes Endocrinol. 2018;6:169–70.

74. Birch P, Ozanne R, Ireland J. Examining the portrayal of homophobic and non-homophobic aggression in print media through an integrated grounded behavioural linguistic inquiry (IGBLI) approach. J Forensic Pract. 2017;19:239–44.

75. Venzo P, Hess K. “Honk Against Homophobia”: Rethinking Relations Between Media and Sexual Minorities. J Homosex. 2013;60:1539–56.

76. Hutchinson P l., Mahlalela X, Yukich J. Mass Media, Stigma, and Disclosure of HIV Test Results: Multilevel Analysis in the Eastern Cape, South Africa. AIDS Educ Prev. 2007;19:489–510.

77. Schwartz J, Grimm J. Uncertainty in Online U.S. News Coverage of Truvada. Health Commun. 2016;31:1250–7.

78. Duran D. Truvada Whores? [Internet]. Huffington Post. 2012 [cited 2017 Sep 14]. Available from: http://www.huffingtonpost.com/david-duran/truvada-whores_b_2113588.html

79. Cadelago C, White JB. Spending big on sex and drug initiatives, AIDS activist Michael Weinstein says he ‘can’t lose.’ Sacram Bee [Internet]. 2016 Oct 10 [cited 2017 Sep 14]; Available from: http://www.sacbee.com/news/politics-government/capitol-alert/article107156677.html

80. Waddell C, Lomas J, Lavis JN, Abelson J, Shepherd CA, Bird-Gayson T. Joining the Conversation: Newspaper Journalists’ Views on Working with Researchers. Healthc Policy. 2005;1:123–39.

81. Guenther L, Bischoff J, Löwe A, Marzinkowski H, Voigt M. Scientific Evidence and Science Journalism. Journal Stud. 2017;0:1–20.

82. Soroka S, Lawlor A, Farnsworth S, Young L. Mass media and policymaking. Routledge Handb Public Policy Process Lond Routledge. 2013;204–14.

83. Scheufele DA. Framing Effects. Int Encycl Commun [Internet]. John Wiley & Sons, Ltd; 2008 [cited 2017 Oct 5]. Available from: http://onlinelibrary.wiley.com/doi/10.1002/9781405186407.wbiecf039.pub2/abstract

84. Smith K. Beyond Evidence Based Policy in Public Health: The Interplay of Ideas. Springer; 2013.

85. Thomm E, Bromme R. “It should at least seem scientific!” Textual features of “scientificness” and their impact on lay assessments of online information. Sci Educ. 2012;96:187–211.

86. Koehler DJ. Can journalistic “false balance” distort public perception of consensus in expert opinion? J Exp Psychol Appl. 2016;22:24–38.

87. Clarke CE, Weberling McKeever B, Holton A, Dixon GN. The Influence of Weight-of-Evidence Messages on (Vaccine) Attitudes: A Sequential Mediation Model. J Health Commun. 2015;20:1302–9.

88. Clarke CE, Dixon GN, Holton A, McKeever BW. Including “evidentiary balance” in news media coverage of vaccine risk. Health Commun. 2015;30:461–72.

89. Jensen JD. Scientific Uncertainty in News Coverage of Cancer Research: Effects of Hedging on Scientists’ and Journalists’ Credibility. Hum Commun Res. 2008;34:347–69.

90. Dixon GN, McKeever BW, Holton AE, Clarke C, Eosco G. The Power of a Picture: Overcoming Scientific Misinformation by Communicating Weight-of-Evidence Information with Visual Exemplars. J Commun. 2015;65:639–59.

91. Peters HP. Gap between science and media revisited: Scientists as public communicators. Proc Natl Acad Sci U S A. 2013;110:14102–9.

92. Salas S, Beca I. [Mass media communication of biomedical advances]. Rev Med Chil. 2008;136:1348–52.

93. Guidotti TL. Evaluation of scientific evidence in law, adjudication and policy: when occupational health takes the witness chair. Med Lav. 2006;97:167–74.

94. Cohen MS, Chen YQ, McCauley M, Gamble T, Hosseinipour MC, Kumarasamy N, et al. Prevention of HIV-1 Infection with Early Antiretroviral Therapy. N Engl J Med. 2011;365:493–505.

95. Rodger AJ, Cambiano V, Bruun T, Vernazza P, Collins S, Lunzen J van, et al. Sexual Activity Without Condoms and Risk of HIV Transmission in Serodifferent Couples When the HIV-Positive Partner Is Using Suppressive Antiretroviral Therapy. JAMA. 2016;316:171–81.

96. Gil de Zúñiga H. Social Media Use for News and Individuals’ Social Capital, Civic Engagement and Political Participation. J Comp-Med Commun. 2012;17:319–336.

97. Kwak H, Lee C, Park H, Moon S. What is Twitter, a Social Network or a News Media? Proc 19th Int Conf World Wide Web [Internet]. New York, NY, USA: ACM; 2010 [cited 2017 Sep 14]. p. 591–600. Available from: http://doi.acm.org/10.1145/1772690.1772751

98. Lee CS, Ma L. News sharing in social media: The effect of gratifications and prior experience. Comput Hum Behav. 2012;28:331–9.




Using Geosocial Networking Apps to Understand the Spatial Distribution of Gay and Bisexual Men: Pilot Study


Background: While services tailored for gay, bisexual, and other men who have sex with men (gbMSM) may provide support for this vulnerable population, planning access to these services can be difficult due to the unknown spatial distribution of gbMSM outside of gay-centered neighborhoods. This is particularly true since the emergence of geosocial networking apps, which have become a widely used venue for meeting sexual partners.

Objective: The goal of our research was to estimate the spatial density of app users across Metro Vancouver and identify the independent and adjusted neighborhood-level factors that predict app user density.

Methods: This pilot study used a popular geosocial networking app to estimate the spatial density of app users across rural and urban Metro Vancouver. Multiple Poisson regression models were then constructed to model the relationship between app user density and areal population-weighted neighbourhood-level factors from the 2016 Canadian Census and National Household Survey.

Results: A total of 2021 app user profiles were counted within 1 mile of 263 sampling locations. In a multivariate model controlling for time of day, app user density was associated with several dissemination area–level characteristics, including population density (per 100; incidence rate ratio [IRR] 1.03, 95% CI 1.02-1.04), average household size (IRR 0.26, 95% CI 0.11-0.62), average age of males (IRR 0.93, 95% CI 0.88-0.98), median income of males (IRR 0.96, 95% CI 0.92-0.99), proportion of males who were not married (IRR 1.08, 95% CI 1.02-1.13), proportion of males with a postsecondary education (IRR 1.06, 95% CI 1.03-1.10), proportion of males who are immigrants (IRR 1.04, 95% CI 1.004-1.07), and proportion of males living below the low-income cutoff level (IRR 0.93, 95% CI 0.89-0.98).

Conclusions: This pilot study demonstrates how the combination of geosocial networking apps and administrative datasets might help care providers, planners, and community leaders target online and offline interventions for gbMSM who use apps.


In British Columbia, Canada, HIV and other sexually transmitted infections continue to disproportionately impact gay, bisexual, and other men who have sex with men (gbMSM) [1,2]. Yet, because the spatial geography of gbMSM may not correlate with that of the broader population, it remains difficult to ensure that sexual health and other services are optimally tailored for these individuals [3]. Previous research examining the social geography of gbMSM has shown that their spatial distribution is nonrandom [4] within the general population. For example, research suggests that the marginalization of sexual minorities along with other forces has given rise to gay neighborhoods—areas that often have a higher than expected concentration of gay men, gay-centered amenities, and homonormative cultural artifacts [5]. However, changing attitudes toward gbMSM in Western society have supposedly reshaped these communities, leading to changes in where these men live, work, and socialize [6]. Additionally, current literature indicates that the introduction of geosocial networking apps, which allow gbMSM to use smart devices to connect with other gbMSM within their geographic proximity, has reduced the need for traditional gay enclaves to facilitate connection [7,8]. These changes challenge the assumption that sexual health services tailored for gbMSM are only needed (or appropriate) within these historically gay neighborhoods [9]. Further compounding the difficulty of targeting app users, their spatial geography may not correlate with that of the wider gbMSM population. For example, previous research has found that only 10% of rural gbMSM sought sex online, compared with 56% in medium sized cities, 50% in suburban areas, and 48% in urban centers [10]. However, dating and online hookup apps largely appeared on the scene in 2009, after this research was conducted; therefore, it is unclear whether these patterns hold true today. These realities make it difficult to identify where and how sexual health services can best meet the needs of app users who are at elevated risk for HIV and other sexually transmitted infections.

Methods in examining app user density have not been widely explored. This study is the first of its kind in Canada and is only preceded by the work of Delaney et al [3], who used similar methods in estimating app user density in Atlanta, Georgia. In their pilot, the authors used a geosocial networking app designed for gbMSM to manually sample 2666 app users across 79 sampling locations. Sampling locations were selected by starting at the home of 1 of the researchers and driving along road networks to create 2-mile sampling intervals throughout the city. In areas where app user density was greater than 50 users per 2-mile radius, they recorded the maximum distance to the 50th closest user and traveled to the next sampling point outside of that buffer. This sampling strategy resulted in 79 data collection points across the city, many of which overlapped. The data were then smoothed using ArcGIS’s kernel density tool (Esri) [11] to create a density map of app users. While Delaney’s objectives were primarily descriptive, our study seeks to modify and leverage their sampling methods to estimate the spatial density of app users across Metro Vancouver and identify the independent and adjusted neighborhood-level factors that predict app user density. The latter of these 2 objectives has not yet been explored despite studies in other research contexts suggesting that neighborhood-level factors are related to the health and behavior of gbMSM [12,13].



Study Setting

This pilot study took place in Metro Vancouver, a regional district of British Columbia, Canada. Metro Vancouver is a favorable location for examining the delivery of sexual health services as it offers a highly supportive environment for sexual minorities and for people living with HIV [14-16]. Since the late 1990s, the province has provided HIV medications and testing services free of charge, with much of the HIV treatment services being administered centrally by the British Columbia Centre for Excellence in HIV/AIDS [16]. Further, the province has led the way in several global initiatives, including the Joint United Nations Programme on HIV/AIDS 90-90-90 worldwide strategy for HIV prevention [17]. Further, Metro Vancouver is an ideal location to consider app use and the spatial variation in gender and sexual minority populations, as it has an active lesbian, gay, bisexual, and transgender (LGBT) community, evidenced by its hosting of an annual gay pride parade, several community-based organizations for lesbian, gay, bisexual, transgender, and queer people, gay bathhouses and bars, and other attractive amenities. Many of these attractions are in the downtown West End (Vancouver’s historically gay neighborhood), however smaller municipalities such as New Westminster are also home to gay bathhouses and gay-owned businesses.

Data Collection

App User Density

Like Delaney et al [3], we used a popular geosocial networking app designed for gbMSM and primarily used by people looking for casual sexual partners, dates, or relationships [7]. While several similar apps exist—targeting a wide range of gbMSM subgroups—the app selected for our study was chosen because it is among the most popular apps for gbMSM [18]. When creating or editing their profile, users of this app can elect to provide a picture and headline for their profile, which is displayed in a grid alongside other users, organized by increasing Euclidian distance [19]. Only active or recently active (ie, within 1 hour) profiles are displayed. Tapping on each photo reveals volunteered information, composing a user’s profile. Further, and of greatest relevance to this study, users are also asked whether they would like to grant access to their location data, which in turn is displayed to other users as real-time Euclidian distance [19]. We should note that the app used in this pilot study is not necessarily representative of all apps used by gbMSM, and we expect that future analyses will explore and compare the results from available platforms. Nevertheless, using this platform, we modified Delaney’s data collection method by systematically sampling app users across a grid of predetermined data collection points throughout Metro Vancouver. The first collection point was selected randomly from a location in Metro Vancouver, and the grid was created by calculating the coordinates for points at 2-mile intervals. Rather than physically traversing the city, as in Delaney et al [3], this approach allowed us to estimate app user density by putting the coordinates of each sampling location into our phone and then counting the number of profiles within a 1-mile radius of each sampling location. This distance was chosen because the app allows users to see the distance (in feet) of other app users up to a 1-mile radius, beyond which the distance of other users is measured with less precision (in miles). As we were only counting the number of users within each sampling radii, no data were collected from user profiles. Collection of other profile data was avoided as an extra precaution beyond traditional ethics guidelines due to the need for further ethical guidance on the use of internet-embedded, publicly available geotagged data for public health and research purposes [20].

As some users did not display their location on their profile, we did not count users who withheld their location and were listed on our screen such that it was unclear whether they were within 1 mile of our virtual sampling location (although we did count users without location information when their inclusion was unambiguous). Recognizing that the desire for greater privacy might vary spatially, this limitation has the potential to underestimate the number of users at some sampling locations (eg, where discreet users worry that they might be identified based on their location). In evaluating the extent to which this limitation impacted our results, we sampled 500 profiles across 5 spatially diverse sampling locations and found that 25.4% (127/500, range 19 to 32) of users did not provide location information. Of these, 5.5% (7/127, range 0 to 3) were listed such that their privacy settings made their inclusion ambiguous (ie, less or greater than 1 mile). The remaining 120 participants did not provide location information but were listed such that dichotomizing their location (eg, 1 mile or more, less than 1 mile) was not difficult (ie, they appeared earlier in the distance-ordered list of users than the farthest participant within 1 mile, thus indicating they resided within 1 mile).

As previous research has shown that app use is higher in the evening and on weekdays [21], data were collected between 5:45 pm and 11:00 pm, Monday through Wednesday, in the last week of November 2016. Dates were selected to represent a normal weekday (eg, no holidays or LGBT events). To further control for variance in use across time (ie, peak hours), we used a random number generator to randomize the order in which geographic locations were sampled. As users can access apps from anywhere (eg, work, home, bars, bathhouse), it is likely that some users access the app from multiple locations throughout their day or week; therefore, individuals were blocked so that they were not counted multiple times. When accessing the app platform, we used a blank profile and did not respond to private messages.

Neighborhood Factors

Recognizing that social and demographic factors have previously been associated with app use [22-25], risky sexual behavior [4,26-29], and neighborhood residence among gay and bisexual men [28,30-32], selected sociodemographic variables for each dissemination area were derived from the 2016 Canadian Census using the Census Analyzer developed by Computing in the Humanities and Social Sciences at the University of Toronto. Additional information on this data source is available elsewhere [33]. Brief definitions for each variable included in our study are provided in Textbox 1. Selection of included variables was made based on their ubiquity in administrative datasets and scientific surveys, thus improving the reproducibility of our study [34]. Furthermore, measuring urbanity, gender, age, ethnicity, socioeconomic status, family situation, and immigration status, the selected variables represented a variety of factors which have regularly been associated with health-related outcomes [35-40].

Statistical Analysis

Spatial data were generated in ArcMap version 10.5 (Esri), and statistical modeling was conducted in R version 3.4.4 (The R Foundation). Bivariate and multivariable Poisson regression models were used to identify neighborhood-level factors associated with greater app user density. The spatial unit of analysis for this regression was the 1-mile sampling radius around each virtual sampling point. For each unit, app user density, rounded to the nearest integer, was calculated by dividing the number of app users observed at each sampling location by the land area within the 1-mile sampling radius. As explanatory variables were on the dissemination area level, we created a combined area and population-weighted average for each factor, which took into account the population size of each dissemination area as well as the proportion of the dissemination area within each sampling radius [41]. Final multivariable models were constructed by initially including all candidate variables of interest and then optimizing the Akaike information criterion (AIC) by backwards elimination. As our sampling method may have biased the app user density of location, we forced inclusion of an interaction term that controlled for time of day (ie, before 8 pm, 8 pm or later) and day of week (ie, Monday, Tuesday, or Wednesday). As a widely used variable selection method [42], particularly for exploratory analyses such as those conducted in our study, this backwards elimination procedure allowed us to identify the relatively best fitting statistical model achievable from our candidate variables, thus simultaneously improving the reproducibility of our study procedures and ensuring the optimal inclusion of candidate variables under conditions where closely related measures (eg, income and education) might limit model accuracy or performance. Comparing the final multivariable model to 1 including only population density and our time-day interaction term, we used a likelihood ratio test [43] and a Bonferroni outlier test [44], the latter of which allowed us to assess the relative performance of the models and detect geographic areas of interest with statistically unexpected app user densities.

Model fit was assessed using the McFadden likelihood-based pseudo r2 and by reviewing other postmodel evaluation criteria (such as the distributions of residuals). The Office of Research Ethics at Simon Fraser University waived ethics approval, as we collected only publicly accessible data (ie, counted the number of profiles near each sampling location) and did not engage users.



A total of 2021 app user profiles were counted within 1 mile of 263 sampling locations. In our simplified model examining the association between app user density and population density (controlling for time and day of sampling), each 100-person increase in population density was associated with a 6.2% increase in app user density (incidence rate ratio [IRR] 1.06, 95% CI 1.06-1.07). As suggested by an increase in model fit (pseudo r2 .650 to .760), the results of a likelihood ratio test (P<.001), and a 4-fold reduction in the number of outliers identified by a Bonferroni model outlier test (ie, 4 to 1), an AIC optimized model including all dissemination area characteristics of interest had superior performance relative to this population density–only model.

This expanded model showed that app user density was positively associated with population density, average age of male residents, proportion of male residents who were not married, proportion of males with a postsecondary education, proportion of male residents who were immigrants, proportion of males living below the low income cutoff (LICO) level, and average household size of residents.


Principal Findings

Using a popular geosocial networking app designed for gbMSM, we sampled over 2000 profiles that were within 1 mile of 263 randomly selected sampling sites in Metro Vancouver, Canada. While our methodology extends those originally piloted by Delaney et al [3], this study is novel in its use of this approach to evaluate the relationship between app user density and other neighborhood-level factors. In doing so, this pilot study supports the use of geographic information systems in aiding public health specialists to understand the spatial distribution of app users. With that said, we acknowledge that the associations identified in our study may be the result of ecological fallacy. Addressing this possibility, we also recognize that several of the factors associated with app user density in this pilot study have also been shown to predict app use among gbMSM at the person level.

Beginning with the social geography of app use, we note that each 100-person increase in population density was associated with a 6% increase in app user density in unadjusted models and a 3% increase when accounting for other factors. Furthermore, we see

that app user density is dramatically higher in downtown Vancouver, particularly in the historically gay neighborhood of Davie Village. This, along with increased app user density in New Westminster (the location of several LGBT-friendly amenities including a gay bathhouse), shows that app user density tracks the distribution of other gay-centric amenities quite well, perhaps indicating that the social geography of online sex seeking has changed from the patterns observed earlier in the internet’s history, when online sex seekers were more likely to identify as bisexual, be closeted, live outside major urban centers, and be disconnected from the gay community [45]. If true, these patterns agree with recent community-based research among gbMSM in Metro Vancouver that suggests that online sex-seeking gbMSM actually spend more time with other gbMSM and are equally as likely to participate in the gay community compared with those who do not seek sex online [46]. With that said, these findings should not be interpreted to mean that rural gbMSM do not use online venues. To do so would be to conflate app use with app user density, the latter of which being a composite measure that includes both the spatial distribution of gbMSM and the prevalence of app use among these men. As such, we note that previous studies have shown that rural men rely on internet-enabled technologies to connect with one another, particularly in rural localities where gbMSM are stigmatized [47]. Interpreted with respect to this, it is possible that app user density is higher in urban areas due to both a preference among gbMSM to live in these areas [48] and the increased motivation for app use proffered by greater opportunities to meet nearby partners [49-51]. Regarding the first hypotheses, we should comment that a growing body of literature has come to question unidirectional migration patterns (ie, from rural to urban) of LGBT people [6,52,53], and research regarding the latter highlights how different motivations for technology use (eg, to meet nearby partners for casual sex) may motivate urban MSM to specifically use apps. With these varied perspectives in mind, we acknowledge that the relationship between online sex seeking, identity, disclosure, and community connectedness remain important areas of study for the health and social sciences [54].

More squarely within the focus of our pilot study, we found that each 1% increase in the proportion of males who were not married and each 1-person increase in average household size were associated with a respective 8% increase and 74% decrease in app user density. The opposing effects here are consistent on face value: with increasing household size being negatively associated with an increasing proportion of residents who are married. Likewise, given that previous research has shown that the technographics of online dating are heavily biased toward single and nonmonogamous users [22], an increasing proportion of single residents in a given neighborhood is expectedly associated with increasing app user density.

As with measures assessing marital status and household size, the observation that each 1-year increase in the average age of the male population was associated with a respective 7% decrease in app user density is unsurprising. Again, the technographics of app use tend to skew toward young gbMSM [46,55]. Thus, neighborhoods with a greater proportion of young men (and a lower average age) would be expected to have more app users. However,

 we can see that the outliers identified by our pilot study included the sampling area in which the University of British Columbia is located. Underscoring this spatial observation, we also documented a 6% increase in app user density for each 1% increase in the proportion of males who had a postsecondary education. This finding too is supported by recent person-level research in Metro Vancouver that has shown an association between greater educational attainment and online sex seeking [22]. Likewise, studies have documented higher educational attainment among adult sexual minorities [56]. Together, these disparate findings are suggestive of nuanced interrelationships between residential location, app use, educational attainment, and age. However, these cannot be fully explained by our findings here and require additional research regarding the life course of gay and bisexual men.

Moving to other closely related sociodemographic measures, our study found that each 1% increase in the proportion of males who were living below the LICO level and each Can $1000 (US $1300) increase in the median income of males were associated with a 7% and 4% decrease in app user density, respectively. As these associations present seemingly contradictory findings, we should first point out that median income and the proportion of residents living below the LICO threshold represent considerably different neighborhood and household conditions despite both serving as measures of socioeconomic status [57]. Median incomes are the median total income residents receive throughout a year. LICO thresholds are the income levels in each dissemination area below which a household would devote at least 20% more than the average family would on basic necessities (ie, food, clothing, and shelter) [58]. An increasing proportion of people living below LICO thresholds can indicate an increasing proportion of impoverished residents as well as an increasing cost of living in a given neighborhood. Therefore, the negative associations between app user density and these 2 measures may indicate that app user density is lower in both cash-strapped neighborhoods (regardless of overall income levels) and those where incomes are depressed. In either case, these trends may be associated with greater constraints placed on the time of residents or attributable to differing lifestyles of residents in these neighborhoods. Supporting this interpretation, previous research examining the association between individual income and app use found that app use on weekdays (during which this study was conducted) is associated with having lower income [21]. As such, caution should be taken when interpreting these findings, as patterns of app user density on weekends might eliminate or reverse this association. In any case, further qualitative research may be needed to understand how app use, neighborhood residence, and socioeconomic status relate to one another.

The same is likely true regarding the final measure included in our multivariable model. Indeed, as is often the case with research addressing multiple intersecting identities [59], to our knowledge little attention has been specifically devoted to the diverse phenomenon of app use among immigrant gbMSM or those living in semisegregated immigrant neighborhoods [60], yet in our study we found that each 1% increase in the proportion of males who were immigrants was associated with a 4% increase in app user density. It is possible that immigrants rely on apps as ways to connect with other gay men, perhaps due to the lack of LGBT venues available to them in ethnically segregated neighborhoods [61] or, alternatively, due to their desire to explore their sexuality discreetly [60]. In either case, this association highlights the importance of diversifying sexual health services and ensuring that they are accessible to those living outside traditional gay villages that often have the reputation of being for wealthy, white, gay men and their straight allies [62,63].


Given the findings outlined, future studies are needed to assess the generalizability of these piloted methods and determine the generalizability of these results outside Metro Vancouver. Laying groundwork for such a validation, our pilot study provides a proof of concept for methods that might be used by public health leaders to optimize the delivery and focus of HIV prevention services by targeting populations at elevated risk for HIV transmission using administrative and geotagged data. While we are not aware of any studies that have leveraged this type of data to improve the delivery of HIV services (ie, location of new services, mobile testing vans) to high-risk neighborhoods, some work has shown that administrative data can be used to identify neighborhoods at risk for other adverse health outcomes [26]. Combining spatial data from various sources (such as dating apps) with administrative data may, therefore, provide an important opportunity for knowledge translation in the context of sexual health, allowing providers to deliver health care services to at-risk neighborhoods. This is especially true for jurisdictions that have invested in mobile testing services [64], online-initiated testing services [65], or other flexible health promotion programs. Further, by planning HIV care using a neighborhood-level perspective [66], public health and community leaders can better justify support for targeted interventions that can address the varied context-specific needs and concerns of local communities [4].


That said, the findings discussed are limited by several potential biases. First, and perhaps most importantly, readers should be aware that sociodemographic census-level factors may not reflect the characteristics of the app users sampled here. Second, because our explanatory variables are averaged across several dissemination areas, the accuracy of our estimates may be limited. However, because dissemination areas are administrative boundaries that are not necessarily reflective of the natural gradation of the characteristics, it is unclear to what extent these units might have biased our results. Future studies should employ a more purposeful sampling design that might better capture app user density within natural communities. Third, our data do not describe from where sampled users are accessing apps (eg, from bars or their home). Therefore, the data generated for this study do not necessarily reflect the residential location of gbMSM but rather where they use the apps on a typical weekday evening. Importantly, while the time and days selected for sampling were purposeful, the effects of sampling error may introduce bias into our study design. To account for this, we randomly assigned the order in which location points were sampled. However, it is still possible that temporal patterns of app use vary by some nonrandom factor (eg, daily routines). Indeed, it is not entirely clear how patterns of app use might vary across the day or week. Future analyses should explore these temporal patterns to determine why and to what degree app use varies across time and under what conditions gbMSM use apps. Fourth, this study was conducted using only a single app. While the app we selected is among the most popular apps for gbMSM [18], few studies have examined differences between apps that are targeted to and as a result taken up by specific subcultures or subgroups within the gay community. It is therefore possible that the spatial density of app users is reflective of only a subset of gbMSM who use apps to find sexual partners. Future work should investigate whether our results are reproducible with other apps such as those targeting older men, ethnic minority men, or men interested in “kink.” That said, previous research has shown that there is a large amount of overlap in the apps used by gbMSM. For instance, 1 study reported a median number of apps per user as 3.11 [21]. Fifth, as our multivariable model had a pseudo r2 of .76, omitted variables not accounted for in this study may also affect app user density. These likely include factors that are difficult to measure using administrative data or are at least rarely measured in these data sources, such as sexual orientation, prevalence of HIV, the social climate toward sexual minorities in a given neighborhood, or a person’s ability to meet sexual partners via other venues. Similarly, our models have yet to be validated for other settings and given that they were developed as exploratory, proof-of-concept models, further research is needed before these or similar models are used authoritatively to inform the deployment of health resources. Therefore, future studies should seek out other datasets and data sources from which models might be derived, thus providing a more complete and empirically valid picture of the ecological factors associated with app user density (eg, male population density vs general population density, same-sex households).


Findings from this pilot study highlight the potential utility of using geographic information systems to better understand the spatial density of gbMSM, particularly among those who use geosocial networking apps and live in urban settings. While additional analyses are needed to validate the modeling techniques explored here and understand the impact of various sampling decisions (eg, time of day, choice of app provider), our findings suggest that these methods may be useful for public health and community leaders hoping to better understand the communities of gbMSM they serve.

  1. British Columbia Provincial Health Officer. 2014. HIV, stigma, and society: tackling a complex epidemic and renewing HIV prevention for gay and bisexual men in British Columbia. Provincial Health Officer's 2010 Annual Report   URL: https:/ /www2. gov.bc.ca/ assets/ gov/ health/ about-bc-s-health-care-system/ office-of-the-provincial-health-officer/ reports-publications/ annual-reports/ hiv-stigma-and-society. pdf [accessed 2018-07-30] [WebCite Cache]
  2. Public Health Agency of Canada. 2014. Population-specific HIV/AIDS status report: gay, bisexual, two-spirit and other men who have sex with men   URL: https:/ /www. canada.ca/ en/ public-health/ services/ hiv-aids/ publications/ population-specific-hiv-aids-status-reports/ bisexual-two-spirit-other-men-who-have-sex-men. html [accessed 2018-07-30] [WebCite Cache]
  3. Delaney KP, Kramer MR, Waller LA, Flanders WD, Sullivan PS. Using a geolocation social networking application to calculate the population density of sex-seeking gay men for research and prevention services. J Med Internet Res 2014;16(11):e249 [FREE Full text] [CrossRef] [Medline]
  4. Latkin CA, German D, Vlahov D, Galea S. Neighborhoods and HIV: a social ecological approach to prevention and care. Am Psychol 2013;68(4):210-224 [FREE Full text] [CrossRef] [Medline]
  5. Black D, Gates G, Sanders S, Taylor L. Why do gay men live in San Francisco? J Urban Econ 2002 Jan;51(1):54-76. [CrossRef]
  6. Ghaziani A. There Goes the Gayborhood?. Princeton: Princeton University Press; 2014.
  7. Gudelunas D. There's an app for that: the uses and gratifications of online social networks for gay men. Sex Cult 2012 Jan 14;16(4):347-365. [CrossRef]
  8. Tudor M. Cyberqueer techno-practices: digital space-making and networking among Swedish gay men [doctoral dissertation]. Stockholm: Stockholm University   URL: http://www.diva-portal.org/smash/get/diva2:532984/FULLTEXT01.pdf [accessed 2018-07-30] [WebCite Cache]
  9. Simon Rosser BR, West W, Weinmeyer R. Are gay communities dying or just in transition? Results from an international consultation examining possible structural change in gay communities. AIDS Care 2008 May;20(5):588-595 [FREE Full text] [CrossRef] [Medline]
  10. Horvath KJ, Simon Rosser BR, Remafedi G. Sexual risk taking among young internet-using men who have sex with men. Am J Public Health 2008 Jun;98(6):1059-1067. [CrossRef] [Medline]
  11. Silverman B. Density Estimation for Statistics and Data Analysis. Boca Raton: Chapman and Hall; 1986.
  12. Gueler A, Schoeni-Affolter F, Moser A, Bertisch B, Bucher HC, Calmy A, Swiss HIV Cohort Study‚ Swiss National Cohort. Neighbourhood socio-economic position, late presentation and outcomes in people living with HIV in Switzerland. AIDS 2015 Jan 14;29(2):231-238. [CrossRef] [Medline]
  13. Burke-Miller JK, Weber K, Cohn SE, Hershow RC, Sha BE, French AL, et al. Neighborhood community characteristics associated with HIV disease outcomes in a cohort of urban women living with HIV. AIDS Care 2016 Oct;28(10):1274-1279 [FREE Full text] [CrossRef] [Medline]
  14. Hull MW, Montaner JSG. HIV treatment as prevention: the key to an AIDS-free generation. J Food Drug Anal 2013 Dec;21(4):S95-S101 [FREE Full text] [CrossRef] [Medline]
  15. Lourenço L, Colley G, Nosyk B, Shopin D, Montaner JSG, Lima VD, STOP HIV/AIDS Study Group. High levels of heterogeneity in the HIV cascade of care across different population subgroups in British Columbia, Canada. PLoS One 2014;9(12):e115277 [FREE Full text] [CrossRef] [Medline]
  16. Montaner JSG, Lima VD, Harrigan PR, Lourenço L, Yip B, Nosyk B, et al. Expansion of HAART coverage is associated with sustained decreases in HIV/AIDS morbidity, mortality and HIV transmission: the “HIV Treatment as Prevention” experience in a Canadian setting. PLoS One 2014 Feb;9(2):e87872 [FREE Full text] [CrossRef] [Medline]
  17. 90-90-90—an ambitious treatment target to help end the AIDS epidemic.: UNAIDS; 2014 Oct.   URL: http://www.unaids.org/sites/default/files/media_asset/90-90-90_en_0.pdf [accessed 2018-07-30] [WebCite Cache]
  18. Badal HJ, Stryker JE, DeLuca N, Purcell DW. Swipe right: dating website and app use among men who have sex with men. AIDS Behav 2018 Apr;22(4):1265-1272. [CrossRef] [Medline]
  19. Roth Y. Zero feet away: the digital geography of gay social media. J Homosex 2016;63(3):437-442. [CrossRef] [Medline]
  20. Zwitter A. Big Data ethics. Big Data Soc 2014 Nov 20;1(2):205395171455925. [CrossRef]
  21. Goedel WC, Duncan DT. Geosocial networking app usage patterns of gay, bisexual, and other men who have sex with men: survey among users of Grindr, a mobile dating app. JMIR Public Health Surveill 2015;1(1):e4 [FREE Full text] [CrossRef] [Medline]
  22. Card KG, Lachowsky NJ, Cui Z, Shurgold S, Gislason M, Forrest JI, et al. Exploring the role of sex-seeking apps and websites in the social and sexual lives of gay, bisexual and other men who have sex with men: a cross-sectional study. Sex Health 2017 Jun;14(3):229-237. [CrossRef] [Medline]
  23. Burrell ER, Pines HA, Robbie E, Coleman L, Murphy RD, Hess KL, et al. Use of the location-based social networking application GRINDR as a recruitment tool in rectal microbicide development research. AIDS Behav 2012 Oct;16(7):1816-1820 [FREE Full text] [CrossRef] [Medline]
  24. Kakietek J, Sullivan PS, Heffelfinger JD. You've got male: internet use, rural residence, and risky sex in men who have sex with men recruited in 12 U.S. cities. AIDS Educ Prev 2011 Apr;23(2):118-127 [FREE Full text] [CrossRef] [Medline]
  25. Downing MJ, Schrimshaw EW. Self-presentation, desired partner characteristics, and sexual behavior preferences in online personal advertisements of men seeking non-gay-identified men. Psychol Sex Orientat Gend Divers 2014 Mar 14;1(1):30-39 [FREE Full text] [CrossRef] [Medline]
  26. Gabert R, Thomson B, Gakidou E, Roth G. Identifying high-risk neighborhoods using electronic medical records: a population-based approach for targeting diabetes prevention and treatment interventions. PLoS One 2016 Jul;11(7):e0159227 [FREE Full text] [CrossRef] [Medline]
  27. Ramjee G, Wand H. Geographical clustering of high risk sexual behaviors in “hot-spots” for HIV and sexually transmitted infections in Kwazulu-Natal, South Africa. AIDS Behav 2014 Feb;18(2):317-322 [FREE Full text] [CrossRef] [Medline]
  28. Kelly BC, Carpiano RM, Easterbrook A, Parsons JT. Sex and the community: the implications of neighbourhoods and social networks for sexual risk behaviours among urban gay men. Sociol Health Illn 2012 Sep;34(7):1085-1102 [FREE Full text] [CrossRef] [Medline]
  29. Kerr JC, Valois RF, Siddiqi A, Vanable P, Carey MP. Neighborhood condition and geographic locale in assessing HIV/STI risk among African American adolescents. AIDS Behav 2014 Aug 10;19(6):1005-1013. [CrossRef] [Medline]
  30. Buttram ME, Kurtz SP. Risk and protective factors associated with gay neighborhood residence. Am J Mens Health 2013 Mar;7(2):110-118 [FREE Full text] [CrossRef] [Medline]
  31. Nash CJ, Gorman-Murray A. LGBT neighbourhoods and “new mobilities”: towards understanding transformations in sexual and gendered urban landscapes. Int J Urban Reg Res 2014 Jan 06;38(3):756-772. [CrossRef]
  32. Vaughan AS, Kramer MR, Cooper HLF, Rosenberg ES, Sullivan PS. Activity spaces of men who have sex with men: an initial exploration of geographic variation in locations of routine, potential sexual risk, and prevention behaviors. Soc Sci Med 2017 Dec;175:1-10 [FREE Full text] [CrossRef] [Medline]
  33. Statistics Canada. 2015. 2011 Census data quality and confidentiality   URL: http://www12.statcan.gc.ca/census-recensement/2011/ref/quality-qualite-eng.cfm [accessed 2018-07-30] [WebCite Cache]
  34. Tarrant C, Wobi F, Angell E. Tackling health inequalities: socio-demographic data could play a bigger role. Fam Pract 2013 Dec;30(6):613-614. [CrossRef] [Medline]
  35. Chen J, Vargas-Bustamante A, Mortensen K, Ortega AN. Racial and ethnic disparities in health care access and utilization under the Affordable Care Act. Med Care 2016 Feb;54(2):140-146 [FREE Full text] [CrossRef] [Medline]
  36. Wong A, Wouterse B, Slobbe LCJ, Boshuizen HC, Polder JJ. Medical innovation and age-specific trends in health care utilization: findings and implications. Soc Sci Med 2012 Jan;74(2):263-272. [CrossRef] [Medline]
  37. Filc D, Davidovich N, Novack L, Balicer RD. Is socioeconomic status associated with utilization of health care services in a single-payer universal health care system? Int J Equity Health 2014 Nov 28;13:115 [FREE Full text] [CrossRef] [Medline]
  38. Derose KP, Escarce JJ, Lurie N. Immigrants and health care: sources of vulnerability. Health Aff (Millwood) 2007 Sep;26(5):1258-1268. [CrossRef] [Medline]
  39. Tiagi R. Access to and utilization of health care services among Canada's immigrants. Intl J Migration Health Soc Care 2016 Jun 13;12(2):146-156. [CrossRef]
  40. Umberson D, Montez JK. Social relationships and health: a flashpoint for health policy. J Health Soc Behav 2010;51 Suppl:S54-S66 [FREE Full text] [CrossRef] [Medline]
  41. Hallisey E, Tai E, Berens A, Wilt G, Peipins L, Lewis B, et al. Transforming geographic scale: a comparison of combined population and areal weighting to other interpolation methods. Int J Health Geogr 2017 Aug 07;16(1):29 [FREE Full text] [CrossRef] [Medline]
  42. Dohoo IR, Martin SW, Stryhn H. Methods in Epidemiologic Research. Charlottetown: Ver Inc; 2012.
  43. Kraemer W, Sonnberger H. The Linear Regression Model Under Test. Heidelberg: Physica-Verlag; 2012.
  44. Cook R, Weisberg S. Residuals and Influence in Regression. New York: Chapman and Hall; 1982.
  45. Tikkanen R, Ross MW. Technological tearoom trade: characteristics of Swedish men visiting gay Internet chat rooms. AIDS Educ Prev 2003 Apr;15(2):122-132. [Medline]
  46. Card K, Lachowsky N, Cui Z, Shurgold S, Gislason M, Forrest J, et al. Exploring the role of sex-seeking apps and websites in the social and sexual lives of gay, bisexual and other men who have sex with men: a cross-sectional study. Sex Health 2016:16. [CrossRef]
  47. Williams ML, Bowen AM, Horvath KJ. The social/sexual environment of gay men residing in a rural frontier state: implications for the development of HIV prevention programs. J Rural Health 2005;21(1):48-55 [FREE Full text] [Medline]
  48. Wimark T. Beyond Bright City Lights: The Migration Patterns of Gay Men and Lesbians. Stockholm: Stockholm University; 2014.
  49. Van de Wiele C, Tong S. Breaking boundaries: the uses & gratifications of Grindr. 2014 Presented at: Proc ACM Int Jt Conf Pervasive Ubiquitous Comput; 2014; New York p. 619-630.
  50. Crooks R. The Rainbow Flag and the Green Carnation: Grindr in The Gay Village. 2013 Nov 26.   URL: http://firstmonday.org/ojs/index.php/fm/article/view/4958 [accessed 2018-07-30] [WebCite Cache]
  51. Rice E, Holloway I, Winetrobe H, Rhoades H, Barman-Adhikari A, Gibbs J, et al. Sex risk among young men who have sex with men who use Grindr, a smartphone geosocial networking application. J AIDS Clin Res 2012 Jul.
  52. Annes A, Redlin M. Coming out and coming back: rural gay migration and the city. J Rural Stud 2012 Jan;28(1):56-68. [CrossRef]
  53. Gorman-Murray A. Rethinking queer migration through the body. Soc Cult Geogr 2007 Feb;8(1):105-121. [CrossRef]
  54. Grov C, Breslow AS, Newcomb ME, Rosenberger JG, Bauermeister JA. Gay and bisexual men's use of the Internet: research from the 1990s through 2013. J Sex Res 2014;51(4):390-409 [FREE Full text] [CrossRef] [Medline]
  55. Allman D, Meyers T, Xu K, Steele S. The social technographics of gay men and other men who have sex with men (MSM) in Canada: implications for HIV research, outreach and prevention. Digit Cult Educ 2012 Jan;1(4):126-144.
  56. Pearson J, Wilkinson L. Same-sex sexuality and educational attainment: the pathway to college. J Homosex 2017;64(4):538-576. [CrossRef] [Medline]
  57. Zhang X. Low income measurement in Canada: what do different lines and indexes tell us?. Ottawa: Statistics Canada; 2015.   URL: https://www150.statcan.gc.ca/n1/pub/75f0002m/75f0002m2010003-eng.htm [accessed 2018-07-30] [WebCite Cache]
  58. Giles P. Low income measurement in Canada. Ottawa: Statistics Canada; 2004.   URL: https://www150.statcan.gc.ca/n1/en/pub/75f0002m/75f0002m2004011-eng.pdf?st=Kac5rG5-[accessed 2018-07-30] [WebCite Cache]
  59. Bauer GR. Incorporating intersectionality theory into population health research methodology: challenges and the potential to advance health equity. Soc Sci Med 2014 Jun;110:10-17 [FREE Full text] [CrossRef] [Medline]
  60. Dhoest A, Szulc L. Navigating online selves: social, cultural, and material contexts of social media use by diasporic gay men. Soc Media Soc 2016 Oct 07;2(4). [CrossRef]
  61. O'Neill H, Kia B. Cent Excell Res Immigr Divers. 2012. Settlement experiences of lesbian, gay, and bisexual newcomers in BC   URL: http://mbc.metropolis.net/assets/uploads/files/wp/2012/WP12-15.pdf [accessed 2018-07-30] [WebCite Cache]
  62. Dalgleish J, Porter E. Inclusivity and othering in Montréal's gay village. Sackville: Mount Allison University; 2016.   URL: http:/ /ocpm. qc.ca/ sites/ ocpm.qc.ca/ files/ pdf/ P87/ 8. 3. 1_mount_allison_final_research_project_paper_dalgleish_porter. pdf [accessed 2018-07-30] [WebCite Cache]
  63. Barrett DC, Pollack LM. Whose gay community? Social class, sexual self-expression, and gay community involvement. Sociol Q 2016 Dec 02;46(3):437-456. [CrossRef]
  64. Lipsitz MC, Segura ER, Castro JL, Smith E, Medrano C, Clark JL, et al. Bringing testing to the people—benefits of mobile unit HIV/syphilis testing in Lima, Peru, 2007-2009. Int J STD AIDS 2014 Apr;25(5):325-331 [FREE Full text] [CrossRef] [Medline]
  65. Gilbert M, Salway T, Haag D, Fairley CK, Wong J, Grennan T, et al. Use of GetCheckedOnline, a comprehensive Web-based testing service for sexually transmitted and blood-borne infections. J Med Internet Res 2017 Mar 20;19(3):e81 [FREE Full text] [CrossRef] [Medline]
  66. Fitzpatrick K, LaGory M. Unhealthy Places: The Ecology of Risk in the Urban Landscape. London: Routledge; 2002.


PublicationsKiffer Card
A Systematic Review of the Geospatial Barriers to Antiretroviral Initiation, Adherence, and Viral Suppression among People Living with HIV

Drafted Version; Final version available at Sexual Health.


Background: Antiretroviral therapy (ART) has become a cornerstone of not only HIV clinical care, but also of HIV prevention—a strategy referred to as Treatment as Prevention (TasP). However, despite the efficacy of treatment-based programs and policies, structural barriers to ART initiation, adherence, and viral suppression have the potential to reduce TasP effectiveness in key populations. Providing a framework for examining these barriers at a population-level, Geographic Information Systems (GIS) have been used widely to study a variety of HIV-related outcomes. While previous reviews have examined the GIS literature with respect to HIV-testing, – an essential antecedent to clinical care – to date no reviews have summarized the research with respect to other ART-related outcomes. Methods: Therefore, the present review leveraged the PubMed database to identify studies that leveraged GIS to examine the barriers to ART initiation, adherence, and viral suppression with the overall goal of understanding how GIS can be used to improve TasP programs. Joanna Briggs Institute criteria were used for the critical appraisal of included studies. Results:  In total, 33 relevant studies were identified, excluding those not utilizing explicit GIS methodology or not examining TasP-related outcomes. Conclusions:  Findings highlight geospatial variation in ART success and inequitable distribution of HIV care in racially segregated, economically disadvantaged, and, by some accounts, increasingly rural areas – particularly in North America. Furthermore, this review highlights the utility and current limitations of using GIS to monitor health outcomes related to ART and the need for careful planning of resources with respect to the geospatial movement and location of people living with HIV (PLWH).


The use of antiretroviral therapies (ART) to prevent HIV/AIDS-related morbidity, mortality, and transmission is referred to as Treatment as Prevention (TasP)1–8 and conceptualized as a care continuum in which people living with HIV (PLWH) are (i) diagnosed soon after infection, (ii) initiate ART soon after HIV diagnosis, (iii) remain adherent to ART across their lives, and (iv) achieve viral suppression and/or undetectability.9 Despite the proven efficacy of TasP programs in reducing HIV/AIDS-related morbidity, mortality, and transmission,10–13 HIV continues to be of concern in key demographic and behaviourally-defined populations.14 To understand the drivers underlying these population-specific epidemics, researchers have used a variety of epidemiological approaches, including Geographic Information Systems (GIS).15–17 These methods have shown that TasP-related outcomes (e.g., community viral load) are salient predictors of neighborhood- and area-level HIV incidence.18–20 However, despite the effective use of GIS to evaluate broadly-targeted epidemiological and prevention strategies, such as HIV testing and HIV incidence,26–30 outcomes related to post-diagnosis HIV care represent a uniquely difficult challenge for GIS because they must target specific individuals within communities that are defined by broader and less subtle forces. Therefore, while other reviews have previously examined structural barriers to HIV-testing,26,27 the present literature review aims to systematically sample and summarize available geospatial research examining ART initiation, adherence, and viral suppression among PLWH.


In October 2016, we systematically sampled articles related to geospatial research (i.e., spatial OR geograph* OR geospatial OR neighborhood), HIV (i.e., HIV OR AIDS), and ART (i.e., antiretroviral OR treatment OR care OR viral load OR suppression OR initiation OR adherence). Sampling was conducted using a Boolean keyword search of the PubMed database. Non-English language articles were omitted, and relevant articles were identified from the titles and abstracts of each search result. Each study was then screened twice for inclusion, and articles were excluded if they did not (i) assess treatment-related measures (e.g. ART initiation, adherence, viral suppression, or some proxy measure for these variables), (ii) did not include geospatial methods, or (iii) did not present original research. The sensitivity and specificity of our search strategy was tested a posteriori by searching for “HIV” in four prominent health geography journals: Applied Geography, Health & Place, International Journal for Health Geographics, International Journal of Geographic Information Science. While our search protocol was not indexed, our review was conducted to comply with PRISMA guidelines for systematic reviews and each article was evaluated by our study team based on the Johanna Briggs Institute’s Critical Appraisal Checklist for Observational Studies 28,29


Study Selection

Our PubMed keyword search returned 4,427 studies, 4,209 of which were available in English. Of these, 184 were considered for review based on the relevance of their titles and abstracts. Based on a more thorough review 23 did not present original research findings; 79 did not assess ART initiation, adherence, or viral suppression; 50 did not utilize geospatial approaches. This search strategy resulted in the inclusion of 33 studies – or approximately 0.7% (n = 33/4427) of returned articles (Figure 1). Sensitivity analyses of our search strategy returned 336 search results, 0.6% (n = 2 / 336) of which met our inclusion criteria and 0.3% (n = 1 / 336) of which was not included in our original search results (and was therefore added to our final inclusion list).

Critical Appraisal of Studies

As all studies relied on some level of aggregation or clustering, there was broad risk for ecological bias. Risk for ecological bias was particularly high for risk analysis studies, where ecological-level data was used to represent or modeled alongside person-level data. With that said, there is a fundamental benefit of using ecological data alongside person-level data, as it can characterize environmental factors not necessarily represented by person-level factors. However, due to both day-to-day movements and permanent migration, assigning individuals to specific geographic areas can be problematic. This is because one’s place of residence or the location where they access care may not be representative of where they live and work.30 Similarly, as few studies conducted sensitivity analyses for the level of aggregation (e.g., by country, province, city, zip code, census block) for explanatory factors and outcomes it was difficult to assess the extent to which modifiable units might have biased study results.31

Despite these limitations, most studies were of acceptable quality: Each provided a clear description of their study samples and data sources; Many (but not all) provided sufficient contextual information regarding study settings; Limitations (exempting those noted above) were cautiously outlined; and sources of bias were often well-defined. Statistical analyses, while wide-ranging, were conventional and well-reasoned (though not always described in sufficient detail for lay readers). Outcomes and explanatory factors were generally appropriate and interpreted correctly. Finally, studies protected participant confidentiality by mapping and reporting aggregated and censored data where small counts or detailed geographic features might have raised privacy concerns.

Overview of Findings

Spatial Heterogeneity of TasP Outcomes. As a first step to examining the relationship between geography and TasP outcomes, a number of studies attempted to map and describe the spatial variation of TasP outcomes: Althoff et al. (2016) found that spatial patterns of ART use (range: 8 – 90%), viral suppression (range: 69 – 95%), and retention (range: 21 – 91) varied independently with significant inter- and intra-regional heterogeneity in North, Central, and South America. Likewise, in North America, Hanna et al. (2013) identified significant inter- and intra-regional (i.e., Northeast U.S.A., Western U.S.A., Midwest U.S.A., Southern U.S.A., and Canada) heterogeneity in ART initiation (range: 35 – 94%; p < .001) and viral suppression (range: 45 – 78%; p < .001). Raboud et al. (2010) found that viral load gaps among Canadians living with HIV were more likely among those residing in Quebec (aOR = 1.72, p < 0.0001) and Ontario (aOR = 1.78, p < 0.0001) compared to those living in British Columbia. On a smaller scale, Gordon et al. (2015) found that residence in New York City (compared to elsewhere in New York State) was associated with poorer linkage to care (OR = 0.73, p < 0.01). Sayles et al. (2012) identified heterogeneity in viral suppression within Los Angeles, California (range: 21.9 – 32.3%). Eberhart et al. (2013) found that patients living in “hotspots” associated with attrition, were more likely to not link to care (OR = 1.76, p < 0.05), link to care late (OR = 1.49, p < 0.05), not be retained in care (OR = 1.84, p < 0.05), and not achieve viral suppression (OR = 3.23, p < 0.05). However, the authors noted that the areas associated with each outcome did not overlap geographically. Laraque et al. (2013) found that community viral load undetectability varied by borough of residence (range: 49.8 - 59.2%, p < 0.0001) and public health area (range: 49– 64%, p < 0.0001). Rebeiro et al. (2016) found that retention varied by region in the United States (i.e., Midwest, Northeast, West, and South; range: 7 – 45%).

Rural and Urban Differences in TasP Outcomes. Generalizing these results, several authors compared TasP outcomes by urban-rural residence. Results from these studies have been mixed, but most studies suggest that rural residence is associated with greater attrition from care and poorer viral suppression. For example, Lourenço et al. (2014) found that compared to the urban health authorities, where attrition from the cascade was the lowest (range: 13 - 44%), residents in the rural health authorities were less likely to be linked to care (aOR = 0.54, p < 0.05), retained in care (aOR = 0.55, p < 0.05), on ART (aOR = 0.55, p < 0.05), or virally suppressed (aOR = 0.45, p < 0.05). Similarly, Ohl et al. (2010) found that CD4 count at time of ART initiation was negatively associated with rural residence (68.7 vs. 74.9%, p < 0.01); and Ohl et al. (2011) found that urban residence was associated with faster ART uptake within 180 days (aOR = 1.72, p < 0.05) and 360 days (aOR = 1.63, p < 0.05), but not 720 days (aOR = 1.26, p < 0.05). Likewise, King et al. (2008) found that both urban residence (vs. rural; OR = 1.38, p < 0.001) and state of residence (range: 46.7 – 71.8%, p < 0.01) were associated with receiving highly active antiretroviral therapy. Meanwhile, other studies show that the effect of rural-urban residence varied based on the outcome of interest. For example, Wilson et al. (2011) found that while neither CD4 counts, virologic suppression, nor ART usage differed based on geographic location, rural patients were less likely than urban patients to report >4 annual outpatient visits compared to urban patients (OR = 0.60, p = 0.003). Furthermore, other studies showed that collinearity with demographic variables may neutralize or reverse the observed relationships associated with rural-urban residence. For example. Ohl et al. (2012) found that while residence in micropolitan (OR = 0.75, p < 0.0001) and rural (OR = 0.79, p < 0.0001) areas was associated with lower adherence on the univariate level, this association reversed after adjusting for age, race/ethnicity, substance use, Hep. C. coinfection, and AIDS-status (OR = 1.24, p < 0.05). Similarly, Chakraborty et al. (2015) also found that while rural residence was associated with smaller declines in viral load (Difference: -23,474 copies/mL; p < 0.01), this was not so for CD4 counts (Difference: 5 cells/mm3; p = 0.59). Furthermore, the effect on viral load became non-significant (β = 0.003; SE = 0.014; p = 0.85) after adjusting for age, gender, race, and exposure type (e.g., MSM, IDU). Cooke et al. (2010) found that rural residence was associated with lower ART uptake (OR = 0.65, p = 0.02), but not after accounting for socioeconomic status, education, age, sex, and distance to care (aOR = 0.94, p = 0.77). Finally, several studies simply found non-significant relationships between TasP outcomes and rural-urban residence. For example, Joy et al (2008) found that rural residence was not associated with CD4 counts (OR = 0.89, n.s.). Likewise, Cherutich et al. (2016) found that neither programmatic region (Range: 45 – 69%; p = 0.349) nor urban-rural residence (range: 59 - 63%; p = 0.158) were associated with viral-load.

Area-level Sociodemographic Factors Associated with TasP Outcomes. Irrespective of rurality, authors also sought to identify other neighborhood-level factors that might underlie patterns of spatial heterogeneity. These studies generally show that area-level factors (e.g., racial segregation, economic inequality, neighborhood disorder, crime) were negatively associated with positive TasP outcomes. For example, Arnold et al. (2009) found that racial disparities in mortality disappeared after accounting for neighborhood socio-economic factors, which explained between 19 and 22% of these disparities (p > 0.05). Furthermore, the effect of race and neighborhood-level socioeconomic factors primarily impacted AIDS survival by impacting ART initiation. Similarly, Burke-Miller et al. (2016) found that racial segregation (aOR = 2.45, p = 0.04) and poor-quality of the built environment (aOR = 2.61, p = 0.03) were associated with having a low CD4 count, but not having an unsuppressed viral load (p = 0.87, p = 0.91, respectively). Castel et al., (2012) found that the areas with the highest community viral loads also had the highest poverty rates and the lowest rates of educational achievement. Goswami et al. (2016) found that linkage to care was associated with higher neighborhood-level education (r = 0.12, p = 0.24), lower income inequality (r = -0.19, p – 0.04), lower vacancy (r = -0.20, p < 0.05), higher proportion of telephone serviced homes (r = 0.05, p < 0.001), and fewer HIV/AIDS service providers (r = -0.15, p < 0.01). Furthermore, viral suppression was associated with lower neighborhood-level income inequality (r = 0.13, p = 0.03), higher proportions of telephone serviced homes (r = 0.08, p = 0.006), and more HIV/AIDS service providers (r = 0.02, p = 0.006). However, there was also a moderating effect of income on transportation with car ownership associated with greater linkage to care (p = 0.017) and viral suppression (p = 0.013) in areas with low poverty; and the number of bus stops was negatively associated with viral suppression in areas with high poverty (p = 0.009). Gueler et al. (2015) found that neighborhood socioeconomic status was negatively associated with late presentation (OR = 0.79, p < 0.05), presenting with advanced HIV (OR =0.71, p < 0.05), presenting with AIDS (OR = 0.59, p < 0.05), delayed ART initiation (OR = 0.62), and loss to follow-up (OR = 0.76, p < 0.05); and positively association with achieving viral suppression (OR 1.52, p < 0.05). Hernández-Romieu et al. (2016) found that 42.2% of those entering care in states with low socio-economic marginalization had CD4 counts <200 cells/mm3 compared to 52.8% in states with higher marginalization. Joy et al. (2008) found that higher census-level unemployment predicted delayed ART access (aOR = 1.41, p < 0.05). Rebeiro et al. (2016) found that greater retention in care was associated with older median age (OR = 1.10, p < 0.05) and a lower proportion of African American residents (OR = 0.37, p < 0.05). Shacham et al. (2013) found that higher poverty was associated with higher CD4 cell counts (aOR = 1.56, p < 0.05) and greater unemployment was associated with greater odds or receiving an ART prescription (aOR = 1.47, p < 0.05). Wood et al. (2000) found that ART use was higher in areas with a higher population density (β = 0.02, SE = 0.002, p < 0.01) and more indigenous residents (β = 0.09, SE = 0.12, p < 0.01); and lower in areas with a higher proportion of female residents (β = -0.11, SE = 0.02, p < 0.01). Surratt et al. (2015) found that neighborhood disorder was distally associated with non-adherence among HIV-positive substance users who sold or traded their medications.

Several studies also demonstrated mixed results – representing significant nuance in the relationship between neighborhood level factors and TasP outcomes. For example, Eberhart et al. (2015) found that lower economic deprivation (aOR = 0.92, p < 0.05), increased public transportation (aOR = 1.04, p < 0.05), longer average distance to pharmacies (aOR = 2.41, p < 0.05), and decreasing distance to HIV care (aOR = 0.85, p < 0.05) were associated with residing in poor retention hotspots. Meanwhile, greater economic deprivation (aOR = 1.09, p < 0.05) and shorter average distance to pharmacies (aOR = 0.12, p < 0.05) were, as expected, associated with residence in a poor viral suppression hotspot. Similarly, Kahana et al. (2016) found that socioeconomic disadvantage was negatively associated with ART use (aOR = 0.85, p < 0.05), but was positively associated with ART adherence for at least six months (OR = 1.32, p < 0.05). Ransome et al. (2017) found that higher neighborhood social participation was associated with higher prevalence of late diagnosis (β = 1.37, SE = 0.32, p < 0.001), linkage to care (β = 1.13, SE = 0.20, p < 0.001), and lower prevalence of engagement in care (β = -1.16, SE = 0.30, p < 0.001); higher collective engagement was associated with lower linkage to care (β = -0.62, SE = 0.32, p < 0.05); the percent of residents with more than a 9th grade education was negatively associated with late diagnosis (β = -1.07, SE = 0.22, p < 0.001) and positively associated with engagement in care (β = 0.94, SE = 0.21, p < 0.001); higher unemployment was negatively associated with linkage to care (β = -0.48, SE = 0.17, p < 0.01); and higher assault rates were negatively associated with engagement in care (β = -0.08, SE = 0.02, p < 0.01).

The Impact of Access to Care on TasP Outcomes. Several studies also incorporated spatial measures into their analyses to determine how distance to care impacted TasP outcomes. These studies generally demonstrated that greater distance to care had a negative impact on TasP outcomes. For example, Siedner et al. (2013) found that GPS-tracked distance (β = 0.03, p < 0.001) and Euclidean distance (β = 0.02, p < 0.001) of transportation were associated with delayed HIV care. Likewise, Eberhart et al. (2013) found that greater distance to care was associated with reduced odds of achieving viral suppression (OR = 0.63, p < 0.05). Cooke et al. (2010) found that greater Euclidean distance was associated with lower ART uptake (OR = 0.73, p < 0.01). Relatedly, Wood et al. (2000) found that ART use was associated with residing close to public transit (β = 0.56, SE = 0.93, RR = 1.75, p < 0.01).

However, not all studies have linked distance to care to poorer TasP outcomes. For example, Kloos et al. (2007) found that ART use was association with population size (r = 0.99, p < 0.01), urbanity (r = 0.98, p < 0.01), number of hospitals and health centers nearby (r = 0.95, p < 0.01), but not the actual distance from the nearest health center to an ART hospital (r = 0.22, p = 0.15). Similarly, Johnson et al. (2013) found that distance to clinic was associated with timely ART initiation at one clinic (OR = 0.97, p < 0.01), but not for patients attend a second clinic (OR = 0.99, p > 0.05). Furthermore, several of the studies assessing the impact of access to care show that access can have different effects on different outcomes. For example, Ransome et al. (2017) found that greater distance to care was associated with both late HIV diagnosis (β = 0.96, SE = 0.18, p < 0.01) and greater engagement in care (β = 0.62, SE = 0.27, p < 0.05) – seemingly contradictory trends.


The studies included in our review demonstrate the use of GIS to (i) survey spatial heterogeneity in ART initiation, adherence, and viral suppression;32–45 (ii) link area-level risk factors to health outcomes;46–58 and (iii) describe the impact of healthcare access on ART outcomes.59–64 Overall, our findings suggests that GIS is a useful and necessary tool to monitor the implementation of ART programs and policies worldwide. This is particularly important given the significant challenges posed to urban and rural regional planners, public health officials, and other stakeholders within the context of an epidemic that is increasingly a concern in ethnically segregated neighborhoods, economically depressed regions, and, by some accounts, increasingly rural areas.65 However, while some local findings from these studies are generalizable to other settings and populations, the articles in our review highlighted the use of GIS for examining local epidemics — making the suite of GIS tools especially powerful when used by those with the ability to effect change, inform policy, and plan healthcare delivery.

            Considering the increasingly complex and geographic demands placed on HIV prevention systems, future GIS applications might include the development of geographically flexible healthcare delivery and prevention services that can quickly respond to changing spatial patterns of HIV. The utility of such approaches have long been identified within the context of mobile HIV testing,66,67 and therefore the application of these principles to HIV treatment is likely to offer similar benefits for those who experience difficulty initiating or adhering to HIV care.68,69 Indeed, in an epidemic that has in many cases transitioned outside traditional gay communities, traditional brick-and-mortar community-based care must, in some settings, be retooled to meet the changing dynamic of the HIV epidemic—this includes expanding current programs to bring about greater geographic coverage.70 With that said, implementation research on the expansion and delivery of care services remains scant,71 making it difficult to identify which specific strategies result in optimal coverage. Meanwhile, HIV surveillance, monitoring, and clinical services should be centralized — administratively, not geographically72 — to allow for coordinated planning and administration.73 Though limited, investigations of population-based ART management within centralized healthcare settings have shown that central ART programming is increasingly important in an era of micro-epidemics and disease clustering.74–76 As such, centralization is likely to ensure coverage in areas of increasing demand and inadequate supply of ART services.77

            With respect to future research, our review highlights the utility of using GIS to engage and monitor ART programs, services, and policies. Further, considering that most studies in our review focused on socioeconomic status (as measured using census data), spatial distance to care (using Euclidian and Network distance), and other social determinants of health (i.e., poverty, race, etc.), we feel that additional analyses are needed to understand how more proximal social factors – such as stigmatization, economic forces, and political geography – might also interrupt TasP implementations in some areas. Indeed, other methodologies suggest that the determinants of TasP success are diverse, making it important to take these factors into consideration when planning geographically-specific interventions aimed at addressing these social determinants of local health.78 Secondly, for the most part, the analyses included in our review were focused primarily on general populations and did not sufficiently delineate across key strata such as GBM, people who inject drugs, and racial/ethnic minorities. Future research should therefore assess the unique spatial geographies of these populations and the syndemic factors which contribute to poorer health outcomes among them.79 Third, our review highlights the importance of combining person-level research with ecological research, to ensure that results derived from clustered data also reflect the real-world phenomenon experienced by individuals. Such combinations might also utilize GPS technology to better understand and articulate the spatial behavior of individuals within the geographies they live. As discussed, geocoding individuals to specific spatial units may not provide sufficient nuance for understanding their movement and migratory patterns – particularly in settings where individuals must traverse between geographic units with some regularity.

Given these research needs, there is a pressing need for open-access and easy-access data across all levels of private and public administration, regardless of industry or focus. Indeed, the success of GIS and other informatics-based studies rests largely on the availability of novel, high quality, and robust data. Therefore, and when possible, research and administrative data should be made available in formats that, while protecting any potential privacy concerns, can be used by informatics researchers. This can be achieved by providing data files that have been pre-aggregated by useful spatial dimensions (e.g., mailing codes, census areas) and by making documentation about license-only data sources more accessible to the public.

            As with other literature reviews, the present study has several important limitations. First and foremost, our search terms and database selection may have caused us to miss important studies relevant to this review. As with all literature reviews there is a tradeoff between sensitivity and specificity. With that said, based on our sensitivity analysis and considering the significant overlap in study themes, we believe that our search strategy resulted in a systematic and generally representative sample of GIS studies examining post-diagnosis related outcomes. Second, as we excluded non-English language papers, most papers dealt with North American contexts, particularly focusing on studies from the US and Canada. Third, because non-significant findings are rarely discussed in the abstracts of published papers it may be that some articles with non-significant findings were inappropriately excluded. Fourth, regarding those studies which used geocoding, but not necessarily geostatistical methodology, the selection of articles was somewhat subjective. While geocoding alone is not necessarily considered a geospatial method, inclusion of studies with a significant focus on geography were still judged to be relevant, and therefore included. With respect to these articles, there is a significant risk that the framing, rather than particular features of the analytic approach, merited their inclusion. However, as this was not a meta-analysis, our primary concern was to provide a review of evidence and therefore we felt this was appropriate. Finally, as this review was focused on intermediate TasP-related outcomes, our results are not inclusive of studies assessing mortality or the efficacy of TasP (i.e., using measures such as community viral load to assess the impact of TasP on HIV incidence). While these outcomes are surely of interest, distinct characteristics of these studies have been reviewed elsewhere.20,80 Despite these limitations, we feel that our general findings—especially those regarding the necessity and utility of conducting geospatial analysis to monitor the development of ART programs—likely represent the most salient needs of future HIV research and prevention programming.


In summary, the findings of this review highlight the feasibility and utility of GIS to monitor health outcomes related to ART. Further, in light of the documented shift of the epidemic into socially and geographically segregated and economically disadvantaged communities, our review also highlights the necessity of using GIS studies to track the changing epidemic as it shifts into new populations and conditions. Indeed, understanding the geospatial variations in ART outcomes can help ensure that resources are equitably distributed and accessible to individuals living in areas and neighborhoods where they are most needed. Central planning of these resources, especially when leveraging GIS, can therefore help optimize care for PLWH as well as HIV prevention.




The authors would like to thank Kirk J. Hepburn for editing this manuscript prior to publication and Meghan Winters for here guidance and expertise in assisting with the development of this manuscript. KGC is supported by the Momentum Health Study as part of his doctoral training. Momentum is funded through the National Institute on Drug Abuse (R01DA031055-01A1) and the Canadian Institutes for Health Research, through both project grant and foundation grant awards (MOP-107544, 143342, PJT-153139). JM is supported with grants paid to his institution by the British Columbia Ministry of Health and by the U.S. National Institutes of Health (R01DA036307).  He has also received limited unrestricted funding, paid to his institution, from Abbvie, Bristol-Myers Squibb, Gilead Sciences, Janssen, Merck, and ViiV Healthcare. These sponsors had no role in the design and conduct of the study; in the collection, management, analysis and interpretation of the data; or in the preparation, review or approval of the manuscript. We have no conflicts of interest to declare.




1.            Attia S, Egger M, Müller M, Zwahlen M, Low N. Sexual transmission of HIV according to viral load and antiretroviral therapy: systematic review and meta-analysis. AIDS Lond Engl 2009; 23(11): 1397-1404. doi:10.1097/QAD.0b013e32832b7dca.

2.            Cohen MS, Chen YQ, McCauley M, et al. Prevention of HIV-1 Infection with Early Antiretroviral Therapy. N Engl J Med 2011; 365(6): 493-505. doi:10.1056/NEJMoa1105243.

3.            Donnell D, Baeten JM, Kiarie J, et al. Heterosexual HIV-1 transmission after initiation of antiretroviral therapy: a prospective cohort analysis. The Lancet 2010; 375(9731): 2092-2098. doi:10.1016/S0140-6736(10)60705-2.

4.            Rodger A, Cambiano V, Brunn T, et al. Sexual activity without condoms and risk of hiv transmission in serodifferent couples when the hiv-positive partner is using suppressive antiretroviral therapy. JAMA 2016; 316(2): 171-181. doi:10.1001/jama.2016.5148.

5.            Antiretroviral Therapy Cohort. Life expectancy of individuals on combination antiretroviral therapy in high-income countries: a collaborative analysis of 14 cohort studies. Lancet Lond Engl 2008; 372(9635): 293-299. doi:10.1016/S0140-6736(08)61113-7.

6.            Montaner JSG, Lima VD, Harrigan PR, et al. Expansion of HAART Coverage Is Associated with Sustained Decreases in HIV/AIDS Morbidity, Mortality and HIV Transmission: The “HIV Treatment as Prevention” Experience in a Canadian Setting. PLoS ONE 2014; 9(2): e87872. doi:10.1371/journal.pone.0087872.

7.            Montaner JSG, Hogg R, Wood E, et al. The case for expanding access to highly active antiretroviral therapy to curb the growth of the HIV epidemic. Lancet Lond Engl 2006; 368(9534): 531-536. doi:10.1016/S0140-6736(06)69162-9.

8.            Montaner JSG, Lima VD, Barrios R, et al. Association of highly active antiretroviral therapy coverage, population viral load, and yearly new HIV diagnoses in British Columbia, Canada: a population-based study. Lancet Lond Engl 2010; 376(9740): 532-539. doi:10.1016/S0140-6736(10)60936-1.

9.            UNAIDS. 90–90–90 - An Ambitious Treatment Target to Help End the AIDS Epidemic.; 2014. Available at: http://www.unaids.org/sites/default/files/media_asset/90-90-90_en_0.pdf [verified September 2016].

10.          Lima VD, Lourenço L, Yip B, Hogg RS, Phillips P, Montaner JSG. Trends in AIDS incidence and AIDS-related mortality in British Columbia between 1981 and 2013. Lancet HIV 2015; 2(3): e92-e97. doi:10.1016/S2352-3018(15)00017-X.

11.          Lima VD, Eyawo O, Ma H, et al. The impact of scaling-up combination antiretroviral therapy on patterns of mortality among HIV-positive persons in British Columbia, Canada. J Int AIDS Soc 2015; 18: 20261.

12.          Hull M, Lange J, Montaner JSG. Treatment as Prevention–Where Next? Curr HIV/AIDS Rep 2014; 11(4): 496-504. doi:10.1007/s11904-014-0237-5.

13.          Nosyk B, Min J, Lima V, Hogg RS, Montaner J. Modelling the cost-effectiveness of population-level HAART expansion in British Columbia. Lancet HIV 2015; 2(9): e393-e400. doi:10.1016/S2352-3018(15)00127-7.

14.          Cohen MS, Smith MK, Muessig KE, Hallett TB, Powers KA, Kashuba AD. Antiretroviral treatment of HIV-1 prevents transmission of HIV-1: where do we go from here? Lancet 2013; 382(9903). doi:10.1016/S0140-6736(13)61998-4.

15.          Hixson BA, Omer SB, del Rio C, Frew PM. Spatial Clustering of HIV Prevalence in Atlanta, Georgia and Population Characteristics Associated with Case Concentrations. J Urban Health Bull N Y Acad Med 2011; 88(1): 129-141. doi:10.1007/s11524-010-9510-0.

16.          Lindberg LD, Orr M. Neighborhood-Level Influences on Young Men’s Sexual and Reproductive Health Behaviors. Am J Public Health 2011; 101(2): 271-274. doi:10.2105/AJPH.2009.185769.

17.          Ramjee G, Wand H. Geographical clustering of high risk sexual behaviors in “hot-spots” for HIV and sexually transmitted infections in Kwazulu-Natal, South Africa. AIDS Behav 2014; 18(2): 317-322. doi:10.1007/s10461-013-0578-x.

18.          Das M, Chu PL, Santos G-M, et al. Decreases in community viral load are accompanied by reductions in new HIV infections in San Francisco. PloS One 2010; 5(6): e11068. doi:10.1371/journal.pone.0011068.

19.          Solomon SS, Mehta SH, McFall AM, et al. Community viral load, antiretroviral therapy coverage, and HIV incidence in India: a cross-sectional, comparative study. Lancet HIV 2016; 3(4): e183-190. doi:10.1016/S2352-3018(16)00019-9.

20.          Smith MK, Powers KA, Muessig KE, Miller WC, Cohen MS. HIV Treatment as Prevention: The Utility and Limitations of Ecological Observation. PLOS Med 2012; 9(7): e1001260. doi:10.1371/journal.pmed.1001260.

21.          Geanuracos CG, Cunningham SD, Weiss G, Forte D, Henry Reid LM, Ellen JM. Use of Geographic Information Systems for Planning HIV Prevention Interventions for High-Risk Youths. Am J Public Health 2007; 97(11): 1974-1981. doi:10.2105/AJPH.2005.076851.

22.          Gilliam G, Hanchette C, Fogarty K, Gibbs D. A Geospatial Analysis of CDC-funded HIV Prevention Programs for African Americans in the United States. J Health Disparities Res Pract 2012; 2(2). Available at: http://digitalscholarship.unlv.edu/jhdrp/vol2/iss2/3.

23.          Koblin BA, Egan JE, Rundle A, et al. Methods to measure the impact of home, social, and sexual neighborhoods of urban gay, bisexual, and other men who have sex with men. PloS One 2013; 8(10): e75878. doi:10.1371/journal.pone.0075878.

24.          Latkin CA, German D, Vlahov D. Neighborhoods and HIV: A Social Ecological Approach to Prevention and Care. Am Psychol 2013; 68(4): 210-224. doi:10.1037/a0032704.

25.          Leibowitz AA, Taylor SL. Distance to public test sites and HIV testing. Med Care Res Rev MCRR 2007; 64(5): 568-584. doi:10.1177/1077558707304634.

26.          Institute of Medicine (US) Committee on HIV Screening and Access to Care. HIV Screening and Access to Care: Health Care System Capacity for Increased HIV Testing and Provision of Care. Washington (DC): National Academies Press (US); 2011. Available at: http://www.ncbi.nlm.nih.gov/books/NBK209480/ [verified September 2017].

27.          Coovadia HM. Access to voluntary counseling and testing for HIV in developing countries. Ann N Y Acad Sci 2000; 918: 57-63.

28.          Moher D, Shamseer L, Clarke M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Syst Rev 2015; 4: 1. doi:10.1186/2046-4053-4-1.

29.          Joanna Briggs Institute. Critical Appraisal Checklist for Analytical Cross Sectional Studies.; 2017. Available at: http://joannabriggs.org/assets/docs/critical-appraisal-tools/JBI_Critical_Appraisal-Checklist_for_Systematic_Reviews2017.pdf.

30.          Vaughan AS, Kramer MR, Cooper HLF, Rosenberg ES, Sullivan PS. Activity spaces of men who have sex with men: An initial exploration of geographic variation in locations of routine, potential sexual risk, and prevention behaviors. Soc Sci Med 1982 2017; 175: 1-10. doi:10.1016/j.socscimed.2016.12.034.

31.          Gehlke CE, Biehl K. Certain Effects of Grouping Upon the Size of the Correlation Coefficient in Census Tract Material. J Am Stat Assoc 1934; 29(185): 169-170. doi:10.2307/2277827.

32.          Althoff KN, Rebeiro PF, Hanna DB, et al. A picture is worth a thousand words: maps of HIV indicators to inform research, programs, and policy from NA-ACCORD and CCASAnet clinical cohorts. J Int AIDS Soc 2016; 19(1): 20707.

33.          Cherutich P, Kim AA, Kellogg TA, et al. Detectable HIV Viral Load in Kenya: Data from a Population-Based Survey. PLOS ONE 2016; 11(5): e0154318. doi:10.1371/journal.pone.0154318.

34.          Hanna DB, Buchacz K, Gebo KA, et al. Trends and Disparities in Antiretroviral Therapy Initiation and Virologic Suppression Among Newly Treatment-Eligible HIV-Infected Individuals in North America, 2001–2009. Clin Infect Dis 2013; 56(8): 1174-1182. doi:10.1093/cid/cit003.

35.          Lourenço L, Colley G, Nosyk B, et al. High levels of heterogeneity in the HIV cascade of care across different population subgroups in British Columbia, Canada. PloS One 2014; 9(12): e115277. doi:10.1371/journal.pone.0115277.

36.          Ohl ME, Lund B, Belperio PS, et al. Rural Residence and Adoption of a Novel HIV Therapy in a National, Equal-Access Healthcare System. AIDS Behav 2011; 17(1): 250-259. doi:10.1007/s10461-011-0107-8.

37.          Ohl ME, Perencevich E, McInnes DK, et al. Antiretroviral Adherence Among Rural Compared to Urban Veterans with HIV Infection in the United States. AIDS Behav 2012; 17(1): 174-180. doi:10.1007/s10461-012-0325-8.

38.          Raboud JM, Loutfy MR, Su D, et al. Regional differences in rates of HIV-1 viral load monitoring in Canada: Insights and implications for antiretroviral care in high income countries. BMC Infect Dis 2010; 10: 40. doi:10.1186/1471-2334-10-40.

39.          Sayles JN, Rurangirwa J, Kim M, Kinsler J, Oruga R, Janson M. Operationalizing treatment as prevention in Los Angeles County: antiretroviral therapy use and factors associated with unsuppressed viral load in the Ryan White system of care. AIDS Patient Care STDs 2012; 26(8): 463-470. doi:10.1089/apc.2012.0097.

40.          Wilson LE, Korthuis T, Fleishman JA, et al. HIV-related medical service use by rural/urban residents: a multistate perspective. AIDS Care 2011; 23(8): 971-979. doi:10.1080/09540121.2010.543878.

41.          King W, Minor P, Ramirez Kitchen C, et al. Racial, gender and geographic disparities of antiretroviral treatment among US Medicaid enrolees in 1998. J Epidemiol Community Health 2008; 62(9): 798-803. doi:10.1136/jech.2005.045567.

42.          Chakraborty H, Iyer M, Duffus WA, Samantapudi AV, Albrecht H, Weissman S. Disparities in viral load and CD4 count trends among HIV-infected adults in South Carolina. AIDS Patient Care STDs 2015; 29(1): 26-32. doi:10.1089/apc.2014.0158.

43.          Gordon DE, Bian F, Anderson BJ, Smith LC. Timing of Entry to Care by Newly Diagnosed HIV Cases Before and After the 2010 New York State HIV Testing Law: JAIDS J Acquir Immune Defic Syndr 2015; 68: S54-S58. doi:10.1097/QAI.0000000000000394.

44.          Laraque F, Mavronicolas HA, Robertson MM, Gortakowski HW, Terzian AS. Disparities in community viral load among HIV-infected persons in New York city: AIDS 2013; 27(13): 2129-2139. doi:10.1097/QAD.0b013e328360f619.

45.          Ohl ME, Tate J, Duggal M, et al. Rural residence is associated with delayed care entry and increased mortality among veterans with human immunodeficiency virus infection. Med Care 2010; 48(12): 1064-1070. doi:10.1097/MLR.0b013e3181ef60c2.

46.          Eberhart MG, Yehia BR, Hillier A, et al. Individual and community factors associated with geographic clusters of poor HIV care retention and poor viral suppression. J Acquir Immune Defic Syndr 1999 2015; 69 Suppl 1: S37-43. doi:10.1097/QAI.0000000000000587.

47.          Arnold M, Hsu L, Pipkin S, McFarland W, Rutherford GW. Race, place and AIDS: The role of socioeconomic context on racial disparities in treatment and survival in San Francisco. Soc Sci Med 1982 2009; 69(1): 121-128. doi:10.1016/j.socscimed.2009.04.019.

48.          Burke-Miller JK, Weber K, Cohn SE, et al. Neighborhood community characteristics associated with HIV disease outcomes in a cohort of urban women living with HIV. AIDS Care 2016; 28(10): 1274-1279. doi:10.1080/09540121.2016.1173642.

49.          Castel AD, Befus M, Willis S, et al. Use of the community viral load as a population-based biomarker of HIV burden. AIDS Lond Engl 2012; 26(3): 345-353. doi:10.1097/QAD.0b013e32834de5fe.

50.          Goswami ND, Schmitz MM, Sanchez T, et al. Understanding Local Spatial Variation Along the Care Continuum: The Potential Impact of Transportation Vulnerability on HIV Linkage to Care and Viral Suppression in High-Poverty Areas, Atlanta, Georgia. J Acquir Immune Defic Syndr 1999 2016; 72(1): 65-72. doi:10.1097/QAI.0000000000000914.

51.          Gueler A, Schoeni-Affolter F, Moser A, et al. Neighbourhood socio-economic position, late presentation and outcomes in people living with HIV in Switzerland: AIDS 2015; 29(2): 231-238. doi:10.1097/QAD.0000000000000524.

52.          Hernández-Romieu AC, Rio C del, Hernández-Ávila JE, et al. CD4 Counts at Entry to HIV Care in Mexico for Patients under the “Universal Antiretroviral Treatment Program for the Uninsured Population,” 2007–2014. PLOS ONE 2016; 11(3): e0152444. doi:10.1371/journal.pone.0152444.

53.          Joy R, Druyts EF, Brandson EK, et al. Impact of neighborhood-level socioeconomic status on HIV disease progression in a universal health care setting. J Acquir Immune Defic Syndr 1999 2008; 47(4): 500-505. doi:10.1097/QAI.0b013e3181648dfd.

54.          Kahana SY, Jenkins RA, Bruce D, et al. Structural Determinants of Antiretroviral Therapy Use, HIV Care Attendance, and Viral Suppression among Adolescents and Young Adults Living with HIV. PloS One 2016; 11(4): e0151106. doi:10.1371/journal.pone.0151106.

55.          Rebeiro PF, Gange SJ, Horberg MA, et al. Geographic Variations in Retention in Care among HIV-Infected Adults in the United States. PLOS ONE 2016; 11(1): e0146119. doi:10.1371/journal.pone.0146119.

56.          Shacham E, Lian M, Önen N, Donovan M, Overton E. Are neighborhood conditions associated with HIV management? HIV Med 2013; 14(10): 624-632. doi:10.1111/hiv.12067.

57.          Surratt HL, Kurtz SP, Levi-Minzi MA, Minxing Chen. Environmental Influences on HIV Medication Adherence: The Role of Neighborhood Disorder. Am J Public Health 2015; 105(8): 1660-1666. doi:10.2105/AJPH.2015.302612.

58.          Ransome Y, Kawachi I, Dean LT. Neighborhood Social Capital in Relation to Late HIV Diagnosis, Linkage to HIV Care, and HIV Care Engagement. AIDS Behav 2017; 21(3): 891-904. doi:10.1007/s10461-016-1581-9.

59.          Cooke GS, Tanser FC, Bärnighausen TW, Newell M-L. Population uptake of antiretroviral treatment through primary care in rural South Africa. BMC Public Health 2010; 10: 585. doi:10.1186/1471-2458-10-585.

60.          Johnson DC, Feldacker C, Tweya H, Phiri S, Hosseinipour MC. Factors associated with timely initiation of antiretroviral therapy in two HIV clinics in Lilongwe, Malawi. Int J STD AIDS 2013; 24(1): 42-49. doi:10.1177/0956462412472312.

61.          Eberhart MG, Yehia BR, Hillier A, et al. Behind the Cascade: Analyzing Spatial Patterns Along the HIV Care Continuum. JAIDS J Acquir Immune Defic Syndr 2013; 64: S42-S51. doi:10.1097/QAI.0b013e3182a90112.

62.          Siedner MJ, Lankowski A, Tsai AC, et al. GPS-measured distance to clinic, but not self-reported transportation factors, are associated with missed HIV clinic visits in rural Uganda. AIDS Lond Engl 2013; 27(9): 1503-1508. doi:10.1097/QAD.0b013e32835fd873.

63.          Wood E, Chan K, Montaner JS, et al. The end of the line: has rapid transit contributed to the spatial diffusion of HIV in one of Canada’s largest metropolitan areas? Soc Sci Med 1982 2000; 51(5): 741-748.

64.          Kloos H, Assefa Y, Adugna A, Mulatu MS, Mariam DH. Utilization of antiretroviral treatment in Ethiopia between February and December 2006: spatial, temporal, and demographic patterns. Int J Health Geogr 2007; 6: 45. doi:10.1186/1476-072X-6-45.

65.          Schafer KR, Albrecht H, Dillingham R, et al. The Continuum of HIV Care in Rural Communities in the United States and Canada: What Is Known and Future Research Directions. J Acquir Immune Defic Syndr 1999 2017; 75(1): 35-44. doi:10.1097/QAI.0000000000001329.

66.          Bassett IV, Regan S, Luthuli P, et al. Linkage to care following community-based mobile HIV testing compared with clinic-based testing in Umlazi Township, Durban, South Africa. HIV Med 2014; 15(6): 367-372. doi:10.1111/hiv.12115.

67.          Liang TS, Erbelding E, Jacob CA, et al. Rapid HIV testing of clients of a mobile STD/HIV clinic. AIDS Patient Care STDs 2005; 19(4): 253-257. doi:10.1089/apc.2005.19.253.

68.          Gilman B, Hidalgo J, Thomas C, Au M, Hargreaves M. Linkages to care for newly diagnosed individuals who test HIV positive in nonprimary care settings. AIDS Patient Care STDs 2012; 26(3): 132-140. doi:10.1089/apc.2011.0305.

69.          Mdege ND, Chindove S. Bringing antiretroviral therapy (ART) closer to the end-user through mobile clinics and home-based ART: systematic review shows more evidence on the effectiveness and cost effectiveness is needed. Int J Health Plann Manage 2014; 29(1): e31-47. doi:10.1002/hpm.2185.

70.          Houben RM, Van Boeckel TP, Mwinuka V, et al. Monitoring the impact of decentralised chronic care services on patient travel time in rural Africa - methods and results in Northern Malawi. Int J Health Geogr 2012; 11: 49. doi:10.1186/1476-072X-11-49.

71.          Lazarus JV, Safreed-Harmon K, Nicholson J, Jaffar S. Health service delivery models for the provision of antiretroviral therapy in sub-Saharan Africa: a systematic review. Trop Med Int Health TM IH 2014; 19(10): 1198-1215. doi:10.1111/tmi.12366.

72.          Koole O, Tsui S, Wabwire-Mangen F, et al. Retention and risk factors for attrition among adults in antiretroviral treatment programmes in Tanzania, Uganda and Zambia. Trop Med Int Health TM IH 2014; 19(12): 1397-1410. doi:10.1111/tmi.12386.

73.          Gerberry DJ, Wagner BG, Garcia-Lerma JG, Heneine W, Blower S. Using geospatial modeling to optimize the rollout of antiretroviral-based pre-exposure HIV interventions in Sub-Saharan Africa. Nat Commun 2014; 5: 5454. doi:10.1038/ncomms6454.

74.          Davila JA, Miertschin N, Sansgiry S, Schwarzwald H, Henley C, Giordano TP. Centralization of HIV services in HIV-positive African-American and Hispanic youth improves retention in care. AIDS Care 2013; 25(2): 202-206. doi:10.1080/09540121.2012.689811.

75.          Morrow C, Munro A, Wilke M, Stark R, Wood R. Remote sensing of HIV care programmes using centrally collected laboratory results: can we monitor ART programme effectiveness? South Afr Med J Suid-Afr Tydskr Vir Geneeskd 2012; 102(6): 501-505.

76.          Tanner AE, Philbin MM, Ott MA, et al. Linking HIV+ adolescents into care: The effects of relationships between local health departments and adolescent medicine clinics. J HIVAIDS Soc Serv 2013; 12(3-4). doi:10.1080/15381501.2013.817280.

77.          Mallitt K-A, Jansson J, Crooks L, McGuigan D, Wand H, Wilson DP. Demand for HIV clinical services is increasing in Australia but supply is decreasing. Sex Health 2013; 10(1): 43-46. doi:10.1071/SH12051.

78.          Pulerwitz J, Michaelis A, Weiss E, Brown L, Mahendra V. Reducing HIV-Related Stigma: Lessons Learned from Horizons Research and Programs. Public Health Rep 2010; 125(2): 272-281.

79.          Singer M, Bulled N, Ostrach B, Mendenhall E. Syndemics and the biosocial conception of health. The Lancet 2017; 389(10072): 941-950. doi:10.1016/S0140-6736(17)30003-X.

80.          Castel AD, Befus M, Willis S, et al. Use of the community viral load as a population-based biomarker of HIV burden. AIDS Lond Engl 2012; 26(3): 345-353. doi:10.1097/QAD.0b013e32834de5fe.

Assessing the Longitudinal Stability of Latent Classes of Substance Use among Gay, Bisexual, and Other Men Who Have Sex with Men

Submitted Draft; Final Version in Drug and Alcohol Dependence


Background: Association between substance use and HIV-risk among gay and bisexual men (GBM) is well documented. However, their substance use patterns are diverse and it is unknown whether self-reported use patterns are stable over time.

Methods: Sexually-active GBM, aged >16 years, were recruited in Metro Vancouver using respondent-driven sampling and followed across 5 study visits at six-month intervals (n=449). To identify distinct patterns of substance use and their longitudinal stability, Latent Transition Analysis (LTA) was conducted for drugs reported by at least 30 participants. Intraclass correlation coefficients (ICC) quantified the stability of class assignments.

Results: Six classes characterizing ‘limited drug use’ (i.e., low use of all drugs, except alcohol), ‘conventional drug use’ (i.e., use of alcohol, marijuana, and tobacco), ‘club drug use' (i.e., use of alcohol, cocaine, and psychedelics), 'sex drug use’ (i.e., use of alcohol, crystal meth, GHB, poppers, and erectile dysfunction drugs), 'street drug use' (i.e., use of alcohol and street opioids) and ‘assorted drug use’ (i.e., use of most drugs) were identified. Across five visits (2.5 years), 26.3% (n=118/449) of GBM transitioned between classes. The prevalence of limited use trended upwards (Baseline:24.5%, Visit 5:28.3%, p<0.0001) and assorted use trended downwards (13.4% to 9.6%, p=0.001). All classes had strong longitudinal stability (ICC>0.97).

Conclusion: Stability of latent substance use patterns highlight the reliability of these measures in identifying patterns of substance use among people who use drugs – potentially allowing for better assessment of these groups and interventions related to their health.


1.1. Background

Gay, bisexual, and other men who have sex with men (GBM) represent a priority population for public health intervention (Institute of Medicine, 2011) as they are known to be at elevated risk for a variety of deleterious health outcomes (Cochran and Mays, 2007; Coker et al., 2010). Among these, GBM are especially vulnerable to poorer mental and emotional health and have greater risk for sexually transmitted infections (STIs), such as HIV (CDC, 2015; Plöderl and Tremblay, 2015). Concomitant with these priority health concerns, substance use has been repeatedly identified as a syndemic factor associated with adverse health – making it a priority concern for GBM communities (Halkitis et al., 2015; Lachowsky et al., 2017).

Recent examinations of GBM’s substance use have attempted to characterize heterogenous patterns of use within their communities with the goal of targeting those patterns that are most strongly associated with adverse health outcomes, such as HIV transmission, problematic substance use, and poor mental health. In doing so, latent class analysis (LCA; Lazarsfeld & Henry, 1968) has emerged as an increasingly popular method to characterize GBM’s polysubstance use behavior (Lim et al., 2015a; McCarty-Caplan et al., 2014; Newcomb et al., 2014a; Tobin et al., 2015; Yu et al., 2015). These studies have shown that patterns of substance use are highly context and population dependent. Indeed, among various samples of GBM, three-class (Lim et al., 2015b; Newcomb et al., 2014b; Tobin et al., 2015), four-class (McCarty-Caplan, Jantz, and Swartz 2014), and six-class (Yu et al., 2015) latent models of substance use have been previously identified. These studies highlight not only a gradient in the number of substances used, but also distinct categories of substance use, such as sex drug use (McCarty-Caplan, Jantz, and Swartz 2014) and recreational drug use (Yu et al., 2015). However, while these studies can empower researchers to understand the complex substance use patterns of GBM, none have yet assessed the longitudinal stability of LCA classes among GBM.

Addressing this limitation, a closely related procedure called Latent Transition Analysis (LTA) has been developed to assess the stability of longitudinal class membership by examining whether individuals transition between classes over time (Collins and Lanza, 2013). While LTA is difficult to implement due to the relative scarcity of longitudinal data, previous studies have leveraged LTA to examine trends in GBM’s sexual behavior (Wilkinson et al., 2017), smoking habits (Gamarel et al., 2017), and sexual orientation (Fish and Pasley, 2015). Among other populations, LTA studies have found that while substance use classes are relatively stable, transition does in fact occur even across relatively short study periods. For example, one study found that over the course of 18 months, 10% of German vocational students transitioned from “alcohol use” to “polysubstance use” (Tomczyk et al., 2016). Similarly, another study among women at risk for HIV reported that 10% of those initially classified as “smokers” transitioned to “crack, cocaine, and heroin use” after six months (Lanza and Bray, 2010). Other epidemiological studies have likewise shown that among a subset of people who use drugs, individuals progress from relatively less harmful and more socially acceptable substances (e.g., alcohol, tobacco, and marijuana) to those which are less widely available and potentially more harmful (Cougle et al., 2016; Flórez-Salamanca et al., 2013; Kirby and Barry, 2012; Nkansah-Amankra and Minelli, 2016; Otten et al., 2017; Secades-Villa et al., 2015; Weinberger et al., 2016). Considering these findings, it is likely that some GBM also transition between latent substance use classes over time.

1.2. Theoretical Framework

While substance use classes are believed to be largely stable due to personal and cultural attitudes and preferences towards substance use (e.g., drug of choice, social norms; Fast et al., 2009; Golub et al., 2005), two primary mechanisms have been proposed to explain transitions from less severe to more severe substance use. The first posits that commonly available drugs, such as alcohol, tobacco, and marijuana, act as neurobiological primers that predispose individuals to subsequent use of other drugs (Kandel, 2002; Kandel and Kandel, 2015; Kandel and Yamaguchi, 1993; Kirby and Barry, 2012; Secades-Villa et al., 2015; Weinberger et al., 2016). For example, Kandel and Kandel report that, in mice, nicotine exerts a non-reciprocal priming effect on cocaine-induced neurobiological addiction. Similar studies have shown priming effects of alcohol in both human and animal studies (Kirby and Barry, 2012). While by no means universal, these findings suggest that neurological primers can contribute to a gateway-like effect where exposure to less severe drugs (e.g., nicotine) predisposes one to increased risk for more severe drug use (e.g., cocaine). The second mechanism, known as the common liability hypothesis, disregards the temporal sequencing of using different substances and posits that shared risk-factors (referred to as liabilities) predispose individuals to substance use (Vanyukov et al., 2017). Proponents of this hypothesis argue that these common liabilities better explain the observed co-occurrence and temporal patterns of substance use behavior (Vanyukov et al., 2012; Vanyukov and Ridenour, 2012). These liabilities include genetic and biological propensities, as well as factors related to the social environment of individuals (e.g., access to healthcare, mental well-being, community connectedness, and social support). Of course, as is often the case with competing hypotheses, empirical investigations comparing these mechanisms show that both primer and liability effects likely contribute to substance use transition (Mayet et al., 2016).

Conversely, access to social and economic capital, access to care and social services, and better mental well-being might contribute to at least temporary transitions towards less severe substance use (Savic et al., 2017). Indeed, previous analyses of Momentum data have shown that GBM’s substance use is strongly associated with socioeconomic and mental health conditions (Card et al., 2017; Lachowsky et al., 2017). Therefore, increased access to health care and social services has the potential to reduce substance use by addressing these contributing factors. Further, several behavior change models highlight the role that communities and health care providers play in helping individuals recognize the potentially harmful effects of their substance use – thus providing motivations for these individuals to reduce their substance use (Chang et al., 2014; Prochaska and Velicer, 1997).

1.2. Objective

Recognizing (i) the limited research on substance use transitions among GBM, (ii) the strong theoretical and empirical support for shifting patterns of substance use, and (iii) the evidence that transition occurs between latent substance use classes in other populations, the present study examined the longitudinal stability of substance use classes among GBM to assess the utility of latent substance use analyses. We hypothesized that while the overall latent class structure would remain stable, a sizeable proportion of individuals would transition towards more frequent and severe substance use.


2.1. Study Protocol

Data for this study were collected as part of the Momentum Health Study, a longitudinal cohort of sexually active GBM, aged >16 years, in Metro Vancouver, British Columbia. Additional information about this cohort has been previously reported (Forrest et al., 2014, 2016; Lachowsky et al., 2016; Moore et al., 2016). In short, participants were recruited using respondent-driven sampling (RDS; Heckathorn, 1997). Eligible GBM presenting an RDS-voucher were screened for enrollment, provided informed consent, completed a 45-minute computer-administered questionnaire, and underwent STI screenings administered by a study nurse. Participants completed follow-up visits every six months. At the completion of each visit, participants were provided a $50 honorarium and received an additional $10 for each eligible referee recruited into the study. Inclusion criteria for this analysis further restricted responses to those which were not lost to follow-up before the 5th study visit and who provided responses for outcome factors. Ethical approval was granted by the research ethics boards at the University of British Columbia, the University of Victoria, and Simon Fraser University.

2.2. Variables

2.2.1. Substance Use

Participants reported their use (any vs. none) of alcohol, cannabis, tobacco, crystal meth, crack, cocaine, speed, heroin, poppers, erectile dysfunction drugs, gamma-Hydroxybutyric acid (GHB), ecstasy, ketamine, mushrooms, Lysergic acid diethylamide (LSD), benzodiazepines, codeine, oxycodone, and prescription steroids over the past six months (P6M). For each substance, reported frequency of use (more than weekly vs. weekly or less) was also assessed. Tobacco use (daily vs. less frequently) was assessed over the past six months; and cannabis use (more than weekly vs. weekly or less) was also assessed over the past three months.

2.2.2. Descriptive Characteristics

Descriptive characteristics were collected to assess the representativeness of sociodemographic and community connectedness variables in the study sample. Sociodemographic variables included age, race/ethnicity, sexual orientation, annual income, employment status, other income sources (i.e., welfare, disability, sex work, drug sales), current housing situation, level of educational attainment, relationship status, and HIV status. Community connectedness variables included attendance over the past six months at gay bars or clubs, group sex events, and the most recent annual pride parade. Participants also reported whether they read gay newspapers, used gay apps and websites to find sexual partners, and how much of their social time they spent with other GBM.

2.3. Statistical Analysis

All statistical analyses were conducted in SAS (SAS, n.d.). Class membership, item response, and transition probabilities were calculated using the PROC LTA procedure (PROC LCA & PROC LTA, 2015). Indicator variables included all substances reported at any frequency by >30 participants. As few missing observations were observed (n = 115/2245), missing indicators due to a missing study visit were carried over from the previous visit to allow LCA models to include individuals who skipped only 1 study visit. This was necessary in order to ensure that our analysis was sufficiently powered to identify the correct number of latent classes. To assess whether this procedure impacted our results, we estimated the number of expected transitions based on the prevalence of observed transitions and number of missing events. Measurement invariance over time was confirmed by comparing class structure and item response probabilities at two separate visits. Final models were built using data from participant’s 1st (February 2012-February 2015) through 5th (March 2014 – February 2017) visits. The number of latent classes was identified based on model parsimony, class distinguishability, theoretical interpretability, and optimization of the Bayesian Information Criterion (BIC; (Dziak and Donna, 2012; Nylund et al., 2007). Supplemental Figure S1 provides the fit criterion used in selecting the number of classes. For each substance, intraclass correlation coefficients (ICC; Koo and Li, 2016) were also calculated to test the longitudinal stability of regular use (i.e., more than weekly use) with scores greater than 0.9 indicating excellent longitudinal stability. To assess statistical significance of trends, regression models were constructed with visit number as an explanatory factor.


Among 774 GBM recruited, 698 enrolled in the longitudinal cohort. Of these, 519 completed second visits, 485 completed third visits, 452 completed fourth visits, and 451 completed fifth visits. In total, 449 participants provided all the data necessary to be included in the LTA (i.e., provided a response for each indicator variable). From these, 2130 observed visits were provided (with 115 visits missing, for which previous observations were carried forward). At enrollment, the median age for this restricted sample was 35 years (Q1, Q3: 27, 48), 88.6% identified as gay (vs. 11.4% as bisexual/other), 76.8% were white, 40.8% had a current regular partner, 28.7% were HIV-positive, 84.4% had some post-secondary education, 66.6% were employed, 92.1% were stably housed, and 42.8% had incomes above $30,000 CAD. Other income sources included welfare (22.1%), disability (6.5%), sex work (4.2%), and drug sales (2.2%). Most participants read gay newspapers (84.2%), attended gay bars/clubs (80.2%), attended or participated in the most recent annual gay pride event (65.5%), and spent >50% of their social time with other GBM (53.5 %). Further, 55.2% sought partners on gay apps and 25.4% attended at least one group sex event in the past six months.

The BIC value, which has been shown to be one of the best performing fit statistics for LCA models, was minimized at a six-class solution (Dziak and Donna, 2012; Nylund et al., 2007). Further, as the six-class model provided the best interpretability and distinguishability compared to other models, the six-class solution was selected. Membership in Class 1 was characterized by limited use of all substances, except alcohol (65.9% reported use in past six months), compared to other classes. Membership in Class 2 was characterized by past six month use of alcohol, tobacco, and marijuana. Membership in Class 3 was characterized by use of alcohol (100%), tobacco (63.6%), marijuana (85.3%), cocaine (63.5%), ecstasy (86.2%), mushrooms (34.3%), and LSD (15.7%). Membership in Class 4 was characterized by past six month use of erectile drugs (88.0%), poppers (59.5%), ecstasy (23.5%), crystal methamphetamine (18.7%), and steroids (13.5%). Membership in Class 5 was characterized by high levels of tobacco use (76.8%) and by elevated use relative to other classes of crack (20.7%), crystal methamphetamine (29.4%), speed (3.5%), heroin (4.1%), codeine (10.9%), and oxycodone (10.8%). Membership in Class 6 was characterized by elevated overall and relative use of most drugs evaluated. Based on deductive interpretations of these classes, we described Class 1 as ‘limited drug use,’ Class 2 as ‘conventional drug use,’ Class 3 as ‘club drug use,’ Class 4 as ‘sex drug use,’ Class 5 as ‘street drug use,’ and Class 6 as ‘assorted drug use.’

Figure 1 shows the distribution of individuals across each of the six latent classes at each study visit. Overall, classes were relatively stable in the proportion of men assigned to each class. Following 449 men across all 5 visits revealed that 99.1% of GBM in Cass 1 (i.e., limited drug use) stayed in their originally assigned class, as did 93.4% of GBM in Class 2 (i.e., conventional drug use), 85.0% of GBM in Class 3 (i.e., club drug use), 93.4% of GBM in Class 4 (i.e., sex drug use), 84.3% of GBM in Class 5 (i.e., street drug use), and 82.1% of GBM in Class 6 (i.e., assorted drug use). In terms of trends, the prevalence of limited drug use increased over time from 24.5 to 28.3% (p < 0.0001), while assorted drug use declined (p = 0.001). Meanwhile, conventional drug use (p = 0.749), club drug use (p = 0.393), sex drug use (p = 0.550), and street drug use (p = 0.216) remained stable.

Summarizing observed transitions , 26.3% (n = 118) of GBM ever transitioned across classes, with most individuals transitioning either once (n = 96) or twice (n = 20). Notable transition pathways included reciprocal relationships between ‘conventional drug use’ (Class 2) and ‘club drug use’ (Class 3), ‘street drug use’ (Class 5) and ‘assorted drug use’ (Class 6); as well as unidirectional transitions from ‘street drug use’ (Class 5) to ‘limited drug use’ (Class 1), ‘assorted drug use’ (Class 6) to ‘conventional drug use’ (Class 1), ‘club drug use’ (Class 3) to ‘assorted drug use’ (Class 6), and ‘sex drug use’ (Class 4) to ‘conventional drug use’ (Class 2). No transition groups were large enough to power further analysis on predictors of class transition.

Generally, significantly fewer participants reported > weekly use of substances than reported any use of substances. Based on point estimates, > weekly use of erectile dysfunction drugs and GHB had “excellent” longitudinal stability (e.g., were internally consistent within individuals across time); tobacco, marijuana, prescription steroids, and benzodiazpenes had “good” longitudinal stability; cocaine, alcohol, and crystal methamphetamine had “moderate” longitudinal stability; and codeine, poppers, crack, heroin, and oxycodone had “poor” longitudinal stability. Only a small number of participants reported > weekly use of speed, ecstasy, ketamine, mushrooms, or LSD. By class, limited drug use (ICC = 0.99, 95% CI: 0.99 – 1.00), conventional drug use (ICC = 0.98, 95% CI: 0.97 – 0.99), club drug use (ICC = 0.98, 95% CI: 0.95 – 0.98), sex drug use (ICC = 0.99, 95% CI: 0.97 – 0.99), street drug use (ICC = 0.97, 95% CI: 0.91 – 0.98), and assorted drug use (ICC = 0.97, 95% CI: 0.91 – 0.98) all had excellent longitudinal stability.


4.1. Primary Findings

The present study provides evidence supporting the longitudinal stability of latent substance use classes – a previously noted limitation of LCA studies among GBM (Lim et al., 2015a; McCarty-Caplan et al., 2014). While studies among other populations (e.g., heterosexual men and women) have previously shown that latent substance use classes are relatively stable (Lanza and Bray, 2010; Tomczyk et al., 2016), none, to our knowledge, have examined the stability of latent classes beyond a single follow-up visit (e.g., six to twelve months) or among GBM. However, our study supports these previous findings and suggests that latent classes in this population are relatively stable – at least over 2.5 years. This finding highlights the potential utility of using latent classes as explanatory factors in longitudinal studies, particularly for hierarchical models using repeated measures. However, while the latent classes themselves persisted at a population-level and most individuals within each class retained class membership, at the individual level approximately 25% of GBM underwent class transitions. This may be due to uncertainty in class assignment or actual changes in substance use behavior. As such, caution should be taken when interpreting cross-sectional relationships between person-level characteristics and latent class membership.

            Regarding specific patterns of transition, our study shows that assorted drug use declined over time and limited drug use increased over time – potentially suggesting an overall decline in severity of substance use patterns within this cohort. This finding is likely reflective of long standing evidence suggesting that substance use varies with period (i.e., changes in substance use patterns in society), age (i.e., changes in substance use patterns across the life course), and cohort (i.e., changes in age cohorts) effects (O’Malley et al., 1984). Indeed, the increasing proportion of participants who reported limited drug use and attrition from the assorted drug use class is likely indicative of a curvilinear relationship between substance use patterns and the life course – with younger participants increasing their substance use and older participants limiting their use as they age (Hser et al., 2007, 2009). Furthermore, previous research has found that severity of substance use is a strong predictor of individuals obtaining substance use treatment (Evans-Polce et al., 2014). Therefore, assorted drug use class members may be more likely to access care, seek to intentionally scale back their use, or substitute less harmful drugs for more harmful ones (Grella and Lovinger, 2011). However, the lack of specific analyses examining the underlying factors predicting class transitions makes it difficult to speak directly to these phenomena. As such, larger scale quantitative studies and carefully targeted qualitative studies are needed to assess specific patterns of substance use transition.

With that said, several transitions observed in the present study merit attention. First, the largest transition pathway (n = 24) was among individuals transitioning from the conventional drug use class to the club drug use class – with a smaller number (n = 10) transitioning from club drug use to conventional drug use. While both classes were characterized by prevalent alcohol use, the club drug use class was characterized by increased use of cocaine and a number of other party drugs (Lea et al., 2016; Noor et al., 2017). This provides some empirical support to biological studies demonstrating a gateway effect of alcohol on cocaine priming (Griffin et al., 2017; Kecojevic et al., 2017), as well as behavioral economic studies showing alcohol substitution for cocaine (Petry, 2001). Furthermore, the link between conventional drug use and club drug use is likely reflective of common liabilities as a previous cross-sectional analysis from our study has shown that membership in these classes is associated with patronage of gay bars and clubs (Card et al., 2017). Together, the transitions mentioned above underscore the influence of biological (e.g., related biological pathways) and socio-ecological factors (e.g., social environments) on both typology and transition patterns of substance use (Terry-McElrath et al., 2009).

Reinforcing the saliency of these factors, the next largest transition pathway (n = 20) was among those transitioning from the assorted drug use class to the street drug use class – with an additional 7 transitioning in the opposite direction. Transition from assorted drug use to a narrower subset of addictive substance (i.e., street drug use) may be reflective of the financial and market-related barriers to assorted drug use. Indeed, previous studies have discussed the profound implications of socioeconomic status and resource availability on peoples substance use patterns (Bourgois, 2003; Carpenter et al., 2017; Chalmers et al., 2010; Dwyer and Moore, 2010; Fast et al., 2009; Petry, 2001). Additionally, there is likely a naturally tendency for heroin users to maintain membership in these classes (where heroin use is the highest) – as heroin has previously been shown to be highly addictive, even compared to other drugs (Hser et al., 2008; Nutt et al., 2007). This is despite our observation that frequent heroin use (i.e., weekly or more) had poor reliability – which is supportive of the erratic nature of substance use. Indeed, previous studies have demonstrated that even addictive substances are characterized by transitions between periods of treatment, abstinence, non-daily, and daily use (Nosyk et al., 2014). These erratic usage patterns may also underlie the transition from the street drug use class to the limited (n = 11) and conventional drug use classes (n = 8). While these transitions may very well reflect intentional, health-motivated, or legally-mandated abstention from drug use (Klingemann, 1991; Klingemann et al., 2010), they may also reflect the natural volatility associated with specific substance use patterns. This suggests that latent class analysis may be relatively less reliable for some typologies of substance use than for others. Therefore, additional assessment of the individual and circumstantial factors that contribute to periods of sobriety among GBM are needed. Among several factors meriting future interest, these assessments should examine the impact of drug treatment and support group participation. Such studies may also provide learning opportunities for researchers and public health leaders hoping to instigate transitions from harmful use to abstinence or managed use.

4.2. Limitations

With consideration to the findings outlined above, readers should be aware of several important limitations. First, generalizability of these findings may be limited due to loss to follow up. However, the relationships we observed are likely still robust (i.e., internally valid). Second, to increase the flexibility of our analytic design we included participants who had missed 1 study visit, carrying forward responses from their previous visit into the missed visit. This may partially have underestimated the number of latent transitions observed (Lachin, 2016). Indeed, if transition probabilities were stable across all periods, we might expect that as many as 9 additional transitions would have been observed. However, given this small number, it seems unlikely that our conclusions would have changed. Yet, because the distribution of carried over observations was higher among those classified as engaging in conventional drug use, readers should be alert to the potential that transitions might have been higher than we report here. Third, due to the length of our follow-up periods (i.e., six-month intervals) and the use of period prevalence measures of substance use, it is possible that our study design does not provide sufficient nuance to understand how GBM transition between latent classes (Cooper, 2010). It is possible for instance that individuals transition ‘back-and-fourth’ between classes multiple times within a single six-month period. Though given the infrequency of transitions, this seems unlikely. Fourth, use of any one substance in the past six months is not necessarily indicative of frequency of use – even for addictive maintenance drugs. Fifth, readers should be aware that the naming and interpretation of latent classes is subjective. Similarly, readers should pay careful attention to the underlying latent variable being assessed by this analysis. Indeed, the indicators used in constructing our latent model represent not only a diverse sampling of psychoactive drugs (e.g., erectile dysfunction drugs, steroids, amphetamines, opioids) but also a broad spectrum of social ambiguities (e.g., legality, social acceptability). Given this, readers should be careful in comparing the results of the present study to those which have been conducted using a more restrictive or targeted selection of indicators.


In conclusion, the present study supports the longitudinal stability of latent substance use classes and highlights several notable transition pathways worth exploring in future research of GBM’s substance use. Overall, transitions do not represent a progression from less severe to more severe substance use as we initially hypothesized. Rather, transitions reflect the biological and socio-ecological propensities and vulnerabilities that underlie specific substance use patterns. Future qualitative studies are therefore needed to better describe the biological and social motivators that instigate transition between classes across the life course.


Bourgois, P., 2003. In Search of Respect: Selling Crack in El Barrio. Cambridge University Press.

Card, K.G., Armstrong, H., Cui, Z., Zhu, J., Lachowsky, N.J., Moore, D.M., Rother, E.A., 2017. A Latent Class Analysis of Substance Use and Culture among Gay, Bisexual, and Other Men Who Have Sex with Men. Cult. Health Sex.

Carpenter, C.S., McClellan, C.B., Rees, D.I., 2017. Economic conditions, illicit drug use, and substance use disorders in the United States. J. Health Econ. 52, 63–73. https://doi.org/10.1016/j.jhealeco.2016.12.009

Centers for Disease Control and Prevention, 2015. CDC Fact Sheet: HIV Among Gay and Bisexual Men.

Chalmers, J., Bradford, D., Jones, C., 2010. The effect of methamphetamine and heroin price on polydrug use: A behavioural economics analysis in Sydney, Australia. Int. J. Drug Policy 21, 381–389. https://doi.org/10.1016/j.drugpo.2010.06.002

Chang, S.J., Choi, S., Kim, S.-A., Song, M., 2014. Intervention Strategies Based on Information-Motivation-Behavioral Skills Model for Health Behavior Change: A Systematic Review. Asian Nurs. Res. 8, 172–181. https://doi.org/10.1016/j.anr.2014.08.002

Cochran, S.D., Mays, V.M., 2007. Physical Health Complaints Among Lesbians, Gay Men, and Bisexual and Homosexually Experienced Heterosexual Individuals: Results From the California Quality of Life Survey. Am. J. Public Health 97, 2048–2055. https://doi.org/10.2105/AJPH.2006.087254

Coker, T.R., Austin, S.B., Schuster, M.A., 2010. The health and health care of lesbian, gay, and bisexual adolescents. Annu. Rev. Public Health 31, 457–477. https://doi.org/10.1146/annurev.publhealth.012809.103636

Collins, L.M., Lanza, S.T., 2013. Latent Class and Latent Transition Analysis: With Applications in the Social, Behavioral, and Health Sciences. John Wiley & Sons.

Cooper, M.L., 2010. Toward a person x situation model of sexual risk-taking behaviors: illuminating the conditional effects of traits across sexual situations and relationship contexts. J. Pers. Soc. Psychol. 98, 319–341. https://doi.org/10.1037/a0017785

Cougle, J.R., Hakes, J.K., Macatee, R.J., Zvolensky, M.J., Chavarria, J., 2016. Probability and correlates of dependence among regular users of alcohol, nicotine, cannabis, and cocaine: concurrent and prospective analyses of the National Epidemiologic Survey on Alcohol and Related Conditions. J. Clin. Psychiatry 77, e444-450. https://doi.org/10.4088/JCP.14m09469

Dwyer, R., Moore, D., 2010. Beyond neoclassical economics: Social process, agency and the maintenance of order in an Australian illicit drug marketplace. Int. J. Drug Policy 21, 390–398. https://doi.org/10.1016/j.drugpo.2010.03.001

Dziak, J., Donna, L., 2012. Sensitivity and specificity of information criteria (Technical Report No. #12-119), Technical Report Series. The Methodology Center, The Pennsylvania State University, Pennsylvania.

Evans-Polce, R.J., Doherty, E.E., Ensminger, M.E., 2014. Taking a life course approach to studying substance use treatment among a community cohort of African American substance users. Drug Alcohol Depend. 142, 216–223. https://doi.org/10.1016/j.drugalcdep.2014.06.025

Fast, D., Small, W., Wood, E., Kerr, T., 2009. Coming “down here”: young people’s reflections on becoming entrenched in a local drug scene. Soc. Sci. Med. 1982 69, 1204–1210. https://doi.org/10.1016/j.socscimed.2009.07.024

Fish, J.N., Pasley, K., 2015. Sexual (Minority) Trajectories, Mental Health, and Alcohol Use: A Longitudinal Study of Youth as They Transition to Adulthood. J. Youth Adolesc. 44, 1508–1527. https://doi.org/10.1007/s10964-015-0280-6

Flórez-Salamanca, L., Secades-Villa, R., Hasin, D.S., Cottler, L., Wang, S., Grant, B.F., Blanco, C., 2013. Probability and predictors of transition from abuse to dependence on alcohol, cannabis, and cocaine: results from the National Epidemiologic Survey on Alcohol and Related Conditions. Am. J. Drug Alcohol Abuse 39, 168–179. https://doi.org/10.3109/00952990.2013.772618

Forrest, J.I., Lachowsky, N.J., Lal, A., Cui, Z., Sereda, P., Raymond, H.F., Ogilvie, G., Roth, E.A., Moore, D., Hogg, R.S., 2016. Factors Associated with Productive Recruiting in a Respondent-Driven Sample of Men who Have Sex with Men in Vancouver, Canada. J. Urban Health Bull. N. Y. Acad. Med. 93, 379–387. https://doi.org/10.1007/s11524-016-0032-2

Forrest, J.I., Stevenson, B., Rich, A., Michelow, W., Pai, J., Jollimore, J., Raymond, H.F., Moore, D., Hogg, R.S., Roth, E.A., 2014. Community mapping and respondent-driven sampling of gay and bisexual men’s communities in Vancouver, Canada. Cult. Health Sex. https://doi.org/10.1080/13691058.2014.881551

Gamarel, K.E., Neilands, T.B., Conroy, A.A., Dilworth, S.E., Lisha, N., Taylor, J.M., Darbes, L.A., Johnson, M.O., 2017. A longitudinal study of persistent smoking among HIV-positive gay and bisexual men in primary relationships. Addict. Behav. 66, 118–124. https://doi.org/10.1016/j.addbeh.2016.11.019

Golub, A., Johnson, B.D., Dunlap, E., 2005. Subcultural evolution and illicit drug use. Addict. Res. Theory 13, 217–229. https://doi.org/10.1080/16066350500053497

Grella, C.E., Lovinger, K., 2011. 30-year trajectories of heroin and other drug use among men and women sampled from methadone treatment in California. Drug Alcohol Depend. 118, 251–258. https://doi.org/10.1016/j.drugalcdep.2011.04.004

Griffin, E.A., Melas, P.A., Zhou, R., Li, Y., Mercado, P., Kempadoo, K.A., Stephenson, S., Colnaghi, L., Taylor, K., Hu, M.-C., Kandel, E.R., Kandel, D.B., 2017. Prior alcohol use enhances vulnerability to compulsive cocaine self-administration by promoting degradation of HDAC4 and HDAC5. Sci. Adv. 3, e1701682. https://doi.org/10.1126/sciadv.1701682

Halkitis, P.N., Kapadia, F., Bub, K.L., Barton, S., Moreira, A.D., Stults, C.B., 2015. A Longitudinal Investigation of Syndemic Conditions Among Young Gay, Bisexual, and Other MSM: The P18 Cohort Study. AIDS Behav. 19, 970–980. https://doi.org/10.1007/s10461-014-0892-y

Heckathorn, D., 1997. Respondent-Driven Sampling: A New Approach to the Study of Hidden Populations*. Soc. Study Soc. Probl. 44.

Hser, Y.-I., Evans, E., Huang, D., Brecht, M.-L., Li, L., 2008. Comparing the dynamic course of heroin, cocaine, and methamphetamine use over 10 years. Addict. Behav. 33, 1581–1589. https://doi.org/10.1016/j.addbeh.2008.07.024

Hser, Y.-I., Hamilton, A., Niv, N., 2009. Understanding Drug Use Over the Life Course: Past, Present, and Future. J. Drug Issues 31, 231–236.

Hser, Y.-I., Longshore, D., Anglin, M.D., 2007. The life course perspective on drug use: a conceptual framework for understanding drug use trajectories. Eval. Rev. 31, 515–547. https://doi.org/10.1177/0193841X07307316

Institute of Medicine (US) Committee on Lesbian, Gay, Bisexual, and Transgender Health Issues and Research Gaps and Opportunities, 2011. The Health of Lesbian, Gay, Bisexual, and Transgender People: Building a Foundation for Better Understanding, The National Academies Collection: Reports funded by National Institutes of Health. National Academies Press (US), Washington (DC).

Kandel, D. (Ed.), 2002. Stages and Pathways of Drug Involvement: Examining the Gateway Hypothesis, 1 edition. ed. Cambridge University Press, Cambridge, UK ; New York.

Kandel, D., Kandel, E., 2015. The Gateway Hypothesis of substance abuse: developmental, biological and societal perspectives. Acta Paediatr. Oslo Nor. 1992 104, 130–137. https://doi.org/10.1111/apa.12851

Kandel, D., Yamaguchi, K., 1993. From beer to crack: developmental patterns of drug involvement. Am. J. Public Health 83, 851–855.

Kecojevic, A., Jun, H.-J., Reisner, S.L., Corliss, H.L., 2017. Concurrent polysubstance use in a longitudinal study of US youth: associations with sexual orientation. Addiction 112, 614–624. https://doi.org/10.1111/add.13681

Kirby, T., Barry, A.E., 2012. Alcohol as a Gateway Drug: A Study of US 12th Graders. J. Sch. Health 82, 371–379. https://doi.org/10.1111/j.1746-1561.2012.00712.x

Klingemann, H., 1991. The motivation for change from alcohol and heroin use. Br. J. Addict. 86, 727–44. https://doi.org/10.1111/j.1360-0443.1991.tb03099.x

Klingemann, H., Sobell, M.B., Sobell, L.C., 2010. Continuities and changes in self-change research. Addict. Abingdon Engl. 105, 1510–1518. https://doi.org/10.1111/j.1360-0443.2009.02770.x

Koo, T.K., Li, M.Y., 2016. A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research. J. Chiropr. Med. 15, 155–163. https://doi.org/10.1016/j.jcm.2016.02.012

Lachin, J.M., 2016. Fallacies of last observation carried forward analyses. Clin. Trials Lond. Engl. 13, 161–168. https://doi.org/10.1177/1740774515602688

Lachowsky, N.J., Dulai, J.J.S., Cui, Z., Sereda, P., Rich, A., Patterson, T.L., Corneil, T.T., Montaner, J.S.G., Roth, E.A., Hogg, R.S., Moore, D.M., 2017. Lifetime Doctor-Diagnosed Mental Health Conditions and Current Substance Use Among Gay and Bisexual Men Living in Vancouver, Canada. Subst. Use Misuse 52, 785–797. https://doi.org/10.1080/10826084.2016.1264965

Lachowsky, N.J., Lal, A., Forrest, J.I., Card, K.G., Cui, Z., Sereda, P., Rich, A., Raymond, H.F., Roth, E.A., Moore, D.M., Hogg, R.S., 2016. Including Online-Recruited Seeds: A Respondent-Driven Sample of Men Who Have Sex With Men. J. Med. Internet Res. 18, e51. https://doi.org/10.2196/jmir.5258

Lanza, S.T., Bray, B.C., 2010. Transitions in drug use among high-risk women: an application of latent class and latent transition analysis. Adv. Appl. Stat. Sci. 3, 203–235.

Lazarsfeld, P.F., Henry, N.W., 1968. Latent structure analysis. Houghton, Mifflin.

Lea, T., Mao, L., Hopwood, M., Prestage, G., Zablotska, I., de Wit, J., Holt, M., 2016. Methamphetamine use among gay and bisexual men in Australia: Trends in recent and regular use from the Gay Community Periodic Surveys. Int. J. Drug Policy 29, 66–72. https://doi.org/10.1016/j.drugpo.2016.01.003

Lim, S.H., Cheung, D.H., Guadamuz, T.E., Wei, C., Koe, S., Altice, F.L., 2015a. Latent class analysis of substance use among men who have sex with men in Malaysia: Findings from the Asian Internet MSM Sex Survey. Drug Alcohol Depend. 151, 31–37. https://doi.org/10.1016/j.drugalcdep.2015.02.040

Lim, S.H., Cheung, D.H., Guadamuz, T.E., Wei, C., Koe, S., Altice, F.L., 2015b. Latent class analysis of substance use among men who have sex with men in Malaysia: Findings from the Asian Internet MSM Sex Survey. Drug Alcohol Depend. 151, 31–37. https://doi.org/10.1016/j.drugalcdep.2015.02.040

Mayet, A., Legleye, S., Beck, F., Falissard, B., Chau, N., 2016. The Gateway Hypothesis, Common Liability to Addictions or the Route of Administration Model? A Modelling Process Linking the Three Theories. Eur. Addict. Res. 22, 107–117. https://doi.org/10.1159/000439564

McCarty-Caplan, D., Jantz, I., Swartz, J., 2014. MSM and drug use: A latent class analysis of drug use and related sexual risk behaviors. AIDS Behav. 18, 1339–1351. https://doi.org/10.1007/s10461-013-0622-x

Moore, D.M., Cui, Z., Lachowsky, N.J., Raymond, H.F., Roth, E., Rich, A., Sereda, P., Howard, T., McFarland, W., Lal, A., Montaner, J., Corneil, T., Hogg, R.S., 2016. HIV Community Viral Load and Factors Associated With Elevated Viremia Among a Community-Based Sample of Men Who Have Sex With Men in Vancouver, Canada: JAIDS J. Acquir. Immune Defic. Syndr. 72, 87–95. https://doi.org/10.1097/QAI.0000000000000934

Newcomb, M.E., Ryan, D.T., Greene, G.J., Garofalo, R., Mustanski, B., 2014a. Prevalence and patterns of smoking, alcohol use, and illicit drug use in young men who have sex with men. Drug Alcohol Depend. 141, 65–71. https://doi.org/10.1016/j.drugalcdep.2014.05.005

Newcomb, M.E., Ryan, D.T., Greene, G.J., Garofalo, R., Mustanski, B., 2014b. Prevalence and patterns of smoking, alcohol use, and illicit drug use in young men who have sex with men. Drug Alcohol Depend. 141, 65–71. https://doi.org/10.1016/j.drugalcdep.2014.05.005

Nkansah-Amankra, S., Minelli, M., 2016. “Gateway hypothesis” and early drug use: Additional findings from tracking a population-based sample of adolescents to adulthood. Prev. Med. Rep. 4, 134–141. https://doi.org/10.1016/j.pmedr.2016.05.003

Noor, S.W., Adam, B.D., Brennan, D.J., Moskowitz, D.A., Gardner, S., Hart, T.A., 2017. Scenes as Micro-Cultures: Examining Heterogeneity of HIV Risk Behavior Among Gay, Bisexual, and Other Men Who Have Sex with Men in Toronto, Canada. Arch. Sex. Behav. https://doi.org/10.1007/s10508-017-0948-y

Nosyk, B., Li, L., Evans, E., Huang, D., Min, J., Kerr, T., Brecht, M., Hser, Y., 2014. Characterizing longitudinal health state transitions among heroin, cocaine, and methamphetamine users. Drug Alcohol Depend. 140, 69–77. https://doi.org/10.1016/j.drugalcdep.2014.03.029

Nutt, D., King, L.A., Saulsbury, W., Blakemore, C., 2007. Development of a rational scale to assess the harm of drugs of potential misuse. The Lancet 369, 1047–1053. https://doi.org/10.1016/S0140-6736(07)60464-4

Nylund, K.L., Asparouhov, T., Muthén, B.O., 2007. Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation Study. Struct. Equ. Model. Multidiscip. J. 14, 535–569. https://doi.org/10.1080/10705510701575396

O’Malley, P.M., Bachman, J.G., Johnston, L.D., 1984. Period, age, and cohort effects on substance use among American youth, 1976-82. Am. J. Public Health 74, 682–688.

Otten, R., Mun, C.J., Dishion, T.J., 2017. The social exigencies of the gateway progression to the use of illicit drugs from adolescence into adulthood. Addict. Behav. 73, 144–150. https://doi.org/10.1016/j.addbeh.2017.05.011

Petry, N.M., 2001. A behavioral economic analysis of polydrug abuse in alcoholics: asymmetrical substitution of alcohol and cocaine. Drug Alcohol Depend. 62, 31–39.

Plöderl, M., Tremblay, P., 2015. Mental health of sexual minorities. A systematic review. Int. Rev. Psychiatry Abingdon Engl. 1–19. https://doi.org/10.3109/09540261.2015.1083949

PROC LCA & PROC LTA, 2015. . The Methodology Center, Penn State., University Park.

Prochaska, J.O., Velicer, W.F., 1997. The transtheoretical model of health behavior change. Am. J. Health Promot. AJHP 12, 38–48.

SAS, n.d. . SAS Institute Inc., Cary, NC, USA.

Savic, M., Best, D., Manning, V., Lubman, D.I., 2017. Strategies to facilitate integrated care for people with alcohol and other drug problems: a systematic review. Subst. Abuse Treat. Prev. Policy 12, 19. https://doi.org/10.1186/s13011-017-0104-7

Secades-Villa, R., Garcia-Rodríguez, O., Jin, C.J., Wang, S., Blanco, C., 2015. Probability and predictors of the cannabis gateway effect: a national study. Int. J. Drug Policy 26, 135–142. https://doi.org/10.1016/j.drugpo.2014.07.011

Terry-McElrath, Y.M., O’Malley, P.M., Johnston, L.D., 2009. Reasons for Drug Use among American Youth by Consumption Level, Gender, and Race/Ethnicity: 1976–2005. J. Drug Issues 39, 677–714.

Tobin, K.E., Yang, C., King, K., Latkin, C.A., Curriero, F.C., 2015. Associations Between Drug and Alcohol Use Patterns and Sexual Risk in a Sample of African American Men Who Have Sex with Men. AIDS Behav. https://doi.org/10.1007/s10461-015-1214-8

Tomczyk, S., Pedersen, A., Hanewinkel, R., Isensee, B., Morgenstern, M., 2016. Polysubstance use patterns and trajectories in vocational students--a latent transition analysis. Addict. Behav. 58, 136–141. https://doi.org/10.1016/j.addbeh.2016.02.027

Vanyukov, M., Cornelius, M., Genna, N.D., Reynolds, M., Kirillova, G., Maher, B., Kirisci, L., 2017. Measurement of Liability to Addiction: Dimensional Approaches. Int. J. Pers. Centered Med. 6. https://doi.org/10.5750/ijpcm.v6i4.612

Vanyukov, M., Ridenour, T.A., 2012. Common liability to drug addictions: Theory, research, practice. Drug Alcohol Depend. 123, S1–S2. https://doi.org/10.1016/j.drugalcdep.2012.01.005

Vanyukov, M., Tarter, R.E., Kirillova, G.P., Kirisci, L., Reynolds, M.D., Kreek, M.J., Conway, K.P., Maher, B.S., Iacono, W.G., Bierut, L., Neale, M.C., Clark, D.B., Ridenour, T.A., 2012. Common liability to addiction and “gateway hypothesis”: Theoretical, empirical and evolutionary perspective. Drug Alcohol Depend. 123, S3–S17. https://doi.org/10.1016/j.drugalcdep.2011.12.018

Weinberger, A.H., Platt, J., Goodwin, R.D., 2016. Is cannabis use associated with an increased risk of onset and persistence of alcohol use disorders? A three-year prospective study among adults in the United States. Drug Alcohol Depend. 161, 363–367. https://doi.org/10.1016/j.drugalcdep.2016.01.014

Wilkinson, A.L., El-Hayek, C., Fairley, C.K., Roth, N., Tee, B.K., McBryde, E., Hellard, M., Stoové, M., 2017. Measuring Transitions in Sexual Risk Among Men Who Have Sex With Men: The Novel Use of Latent Class and Latent Transition Analysis in HIV Sentinel Surveillance. Am. J. Epidemiol. 185, 627–635. https://doi.org/10.1093/aje/kww239

Yu, G., Wall, M.M., Chiasson, M.A., Hirshfield, S., 2015. Complex drug use patterns and associated HIV transmission risk behaviors in an Internet sample of U.S. men who have sex with men. Arch. Sex. Behav. 44, 421–428. https://doi.org/10.1007/s10508-014-0337-8


PublicationsKiffer Card
Predictors of Facebook User Engagement with Health-Related Content for Gay, Bisexual, and Other Men Who Have Sex with Men

Submitted Draft; Final Version in JMIR - Public Health


Background: Social Media is used by community-based organizations (CBOs) to promote the well-being of gay and bisexual men (GBM). However, few studies have quantified which factors facilitate the diffusion of health content tailored for sexual minorities.

Objectives: To identify post characteristics that can be leveraged to optimize the health promotion efforts of CBOs on Facebook.

Methods: The Facebook application programming interface (API) was used to collect 5 years of posts shared across 10 Facebook pages administered by Vancouver-based CBOs promoting GBM health. Network analysis assessed basic indicators of network structure. Content analyses were conducted using informatics-based approaches. Hierarchical negative binomial regression of post engagement data was used to identify meaningful covariates of engagement.

Results: In total, 14,071 posts were shared and 21,537 users engaged with these posts. Most users (n=13,315) engaged only once. There was moderate correlation between the number of posts and the number of CBOs users engaged with (r=0.53, p < 0.0001). Higher user engagement was positively associated with positive sentiment, sharing multimedia, and posting about PrEP, stigma, and mental health. Engagement was negatively associated with asking questions, posting about dating, and sharing posts during or after work (vs. before).

Conclusions: Results highlight the existence of a core group of Facebook users who facilitate diffusion. Factors associated with greater user engagement present CBOs with a number of strategies for improving the diffusion of health content.


Gay, bisexual, and other men who have sex with men (GBM) are at elevated risk for a number of adverse health outcomes [1,2]. Stall et al. [3], argues that gay communities experience a syndemic of co-occurring sexual, substance use, and psychosocial conditions which, according to Singer [4], work synergistically under “deleterious social and physical conditions” (p.15) to adversely affect the health of this population [5]. In response, public health and community leaders have advanced holistic approaches to gay men’s health which address not only individual and biological factors, but also the broader psychosocial and structural factors that affect their health and well-being [6].

In implementing such programs, social media is widely used by community-based organizations (CBOs) to disseminate health information and engage with GBM [7–9]. Indeed, social media has come to play a significant and diverse role in a variety of health contexts. Articulating this role, Kietzmann et al. [10] highlight seven personal and interpersonal needs that social media has come to fulfill. Broadly, we summarize these needs by three activities: identity management, communication, and social bonding. In the context of GBM health, sexual minorities have always needed spaces where they can engage in these activities, and social media has come to provide such spaces [11,12].

While the internet provides a platform whereby CBOs can reach GBM, the success of these interventions is far from guaranteed [13]. Rogers’ Diffusion of Innovations Theory describes the challenges to CBO’s in terms of diffusion, reach, and uptake [14,15]. In brief, Rogers posits that key characteristics of individuals (whom he describes as “adopters”) and the network ties that connect them to others in a social network are fundamental to the spread of information, behavior, and products. While a number of factors have been identified as impacting adoption and diffusion (e.g., age, social network structure, personality types), media richness theory describes how specific media (i.e., routes of content delivery) detract or promote to diffusion [16] and argues that more “life like” interactions better promote uptake of new ideas.

In the age of social media, specific engagement indicators (i.e., reactions, comments, and shares) on Facebook provide rudimentary markers for diffusion – and in fact, are used by Facebook’s edgerank algorithm to govern which messages are shown to other users [17]. Barriers to diffusion are particularly relevant to efforts targeting GBM, who represent a diverse and uniquely organized group of individuals [18]. For example, Cassidy [19] notes that campaigns to amass likes, comments, and shares can often be at odds with an individual’s need to manage their public identity. After all, not all sexual minorities openly acknowledge their sexuality online—especially in spaces where multiple social circles collide [20]. Yet, if social media strategies among GBM are to be successful, CBOs must find ways to encourage users to engage with their content. This is because many social media platforms rely on engagement-based algorithms to determine if social media content is viewed by other users. For example, according to Facebook,

“The stories that show in your News Feed are influenced by your connections and activity on Facebook. This helps you to see more stories that interest you from friends you interact with the most. The number of comments and likes a post receives and what kind of story it is (ex: photo, video, status update) can also make it more likely to appear in your News Feed.” [17]

Consistent with this, increasing user engagement (defined by Facebook as the composite of reactions, comments, and shares on a post) has become a primary objective of social media campaigns, and a handful of studies have sought to identify predictors of user engagement. For example, Veale et al. [21] identified 10 Twitter and Facebook profiles with high user engagement and found that these organizations gained prominence by posting regularly, engaging with individual users, encouraging interaction and conversation by posing questions, sharing multimedia, and highlighting celebrity involvement. In a similar study, Kite et al. [22] found that higher post engagement among 20 Facebook health profiles was associated with positive sentiment, providing factual information, inclusion of videos, and celebrity marketing. Likewise, Rus & Cameron [23] explored 10 Diabetes-related health pages and found that imagery was a strong predictor of engagement. Further they identified other characteristics, such as sentiment, crowdsourcing, and providing factual information that were associated with some, but not all, forms of engagement. However, as campaigns addressing sensitive subjects and those targeting sexual minorities might be uniquely constrained by users’ willingness to publicly endorse or share CBO-generated content, context-specific evaluations of user engagement are needed. As such, the primary objective of this study was to identify strategies to enhance user engagement.

Additionally, it is unclear whether Facebook is even an effective platform for CBOs to reach sexual minority populations [24]. Indeed, while social media campaigns might gain the attention of local network members, they may miss those who are not directly associated with CBOs. Despite widely-held assumptions of Facebook’s communication potential [9], little research has been conducted on the Facebook network structure of sexual minorities. Optimistically, that which has, suggests that the Facebook network structure of sexual minorities is scale-free [25], meaning that some individuals are more embedded in the social network than others, and that these individuals act like “hubs” diffusing information into their local networks. However, while scale-free networks are said to effectively transmit information [26], their efficiency relies on the ways these networks are organized [14]. For example, scale-free networks with high modularity (i.e., the appearance of distinct clusters or communities within a network) promote strong bonds between network members and thus saturation of local networks, while those with low modularity promote weak ties between individuals, but broad global diffusion [27,28]. Both modular and non-modular network structures offer benefits and limitations, for example experimental research by Bakshy et al. [27] shows that strong ties increase the likelihood that individuals will share content shared by other network members, while weak ties facilitate the diffusion of information between network clusters. Therefore, as a secondary objective, the present study aimed to complement our understanding of the diffusion of information through the Facebook networks of CBOs in Vancouver, British Columbia.


Consistent with these objectives, the present study leveraged data collected from 10 Facebook pages (i.e., all pages identified as being administered by selected organizations) belonging to 8 community-based organizations (CBOs) in Vancouver, British Columbia (BC). Pages were purposively selected (i.e., all identified organizations were included) that were (1) well known to our study team (i.e., community-based partners or those otherwise highly visible), (2) inclusive of or targeted towards sexual minorities (i.e., page content relevant, at least in part, to sexuality, sexual health, or community social issues), and (3) dealt primarily with health promotion (i.e., health promotion was main goal of the organization). To ensure user privacy and compliance with Facebook’s end-user agreement, data were downloaded using Facebook’s public application programming interface (API), accessed through the Netvizz Facebook application [29]. Data collected between January 1, 2010 and August 31, 2016 via Netvizz were hierarchically organized by page and post. The first year, 2010, was selected based on the completion of the iPrEx trial examining the efficacy of Pre-Exposure prophylaxis – one of the key topics assessed in this analysis [30]. At the page-level, we identified the number of followers for each page. On the post-level, we identified the number of likes, comments, and shares on each post. Netvizz also assigned unique identifiers to each user, allowing us to examine user engagement across multiple posts and multiple pages. As such, we used Spearman’s Rank Correlation to determine whether there was an association between frequency of participation and participation across multiple pages. Further, a network diagram showing the ways individuals interacted with posts from the 10 CBOs was constructed in Gephi 0.9.1. using the ForceAtlas2 layout algorithm [31]e. Modularity clusters were also identified using Gephi’s modularity tool with the resolution set to 1 in order to maximize the modularity [32]. As this study leveraged publicly-available data, the research ethics board at Simon Fraser University deemed the study exempt from review. As an extra precaution on behalf of the users whose data were included in the present analysis, the names of the Facebook pages included in our study have been omitted. 

Network diagram illustrating user engagement with each post. Colors represent modularity clusters. Numbered symbols represent each Facebook page with the location indicating the modularity class in which most posts were located.

Network diagram illustrating user engagement with each post. Colors represent modularity clusters. Numbered symbols represent each Facebook page with the location indicating the modularity class in which most posts were located.

The content of each post was then analyzed using informatics-based methodology [33–35]: First, using researcher-generated search taxonomies, we identified posts relating to 8 topics: Pre-Exposure Prophylaxis (i.e., PreP, preexposure, pre-exposure, prophylaxis), treatment (i.e., treatment, undetect*, viral load, viral-load), condoms (i.e., condom*), mental health (i.e., mental, emotion*, depress*, anxiety), stigma (i.e., stigma, discriminat*), testing (i.e., test*, screening, checked online), dating (i.e., dating, relationship), and research (i.e., research*, study). Posts that utilized questions to engage users were also recorded by identifying posts with a question mark (i.e., “?”). Similarly, posts which directly encouraged user engagement were identified by searching for key terms inviting participation (i.e., like, comment, share, take, visit). Further, each sentence of each post was scored using the Bing Liu Sentiment Lexicon [34]. The Bing Liu Sentiment Lexicon, which is widely used in sentiment analysis and opinion mining, was selected because it provides a freely accessible word database which assigns positive and negative values to key words, including commonly misspelled words. After each word within each sentence was scored, an average sentiment score was assigned to each post indicating whether the post had an overall negative or positive affect.

We then used multivariable hierarchical negative binomial regression to identify the post characteristics associated with greater user engagement. In this analysis Facebook’s engagement score was used, since this is presumably an important variable used in their News Feed algorithm. According to Facebook’s API, the number is calculated as the combined total number of reactions, shares, and comments on each post. Hierarchical negative binomial regression modelling was selected as the statistical approach for this study as the Facebook engagement count data were over-dispersed, highly skewed towards 0 and 1, and came from 10 separate Facebook pages—each with a varying number of Facebook “fans” and with differing rates of activity.  All coding and statistical analysis were conducted in r-studio.


During the study period between January 1, 2010 and August 31, 2016, 14,071 posts were shared. In total, 21,537 unique users were identified as having engaged with at least one post. Most users engaged only once (n = 13,315), two to five times (n = 4,872), or six to nine times (n = 1,197). Approximately 10% (n = 2,153) of users engaged more than 10 times. Similarly, most users engaged with content from only one (n = 18,837) or two (n = 1,978) groups. Only a small minority of users (n = 722) interacted with more than three groups. Despite low overall engagement (low number of users who ‘engaged’ with content more than once), high modularity (Q = 0.62) was observed in the ways individuals interacted with shared content. Indeed, eight modularity clusters accounted for 74.5% of posts (n = 10,481 / 14,071) and 93.3% of users (n = 20,097 / 21,537). There was moderate correlation between the number of posts and the number of CBOs users engaged with (r=0.53, p < 0.0001).

Higher user engagement was positively associated with positive sentiment (IRR = 1.68), sharing photos (IRR = 3.00), videos (IRR = 2.32), and links (IRR = 1.66), and posting about PrEP (IRR = 3.64), stigma (IRR = 1.60), and mental health (IRR = 1.52). Figure 2 shows the frequency of health messaging over time for the key terms assessed in the present analysis. Engagement was negatively associated with asking a question (IRR = 0.90), posting about dating (IRR = 0.72), sharing posts during (IRR = 0.76) or after work (IRR = 0.79) compared with before work, and with sharing events (IRR = 0.70).

Loess smoothed mention of health messages overtime (2010-2016), stratified by keyword

Loess smoothed mention of health messages overtime (2010-2016), stratified by keyword


The present study collected post data from 10 Facebook pages promoting health or health-related events to GBM in Vancouver, British Columbia. Together, these 10 pages had approximately 24,000 followers, shared approximately 14,000 posts, and amassed over 25,000 engagements (i.e., likes, comments, shares) during the seven years of data analyzed. While our data do not speak empirically to the true network structure of Facebook’s gay communities in Vancouver, we can make several important inferences regarding the network structure that underlies the present analysis. First, based on the correlation between the number of groups and the number of engagements, our results point to the existence of a core group of users who may promote the diffusion of health content. Indeed, only a minority (38.2%) of users engaged more than once over the seven-year period we studied. These observations suggest that the true Facebook network structure of Vancouver’s gay community is indeed scale-free, as shown by Silenzio et al. [25]. Second, as most users only engaged once over the extended timeframe of this analysis, our findings also suggest that shared content is broadly diffusing into distal regions of the network among individuals who may not be directly linked to the Facebook pages included in this analysis [28]. Third, as we observed modularity in user-post engagement, our findings also suggest that together, the 10 Facebook pages included in the present analysis are serving multiple, distinct, though linked clusters. Indeed, while some Facebook pages overlap in their outreach, our findings suggest that the combined effort of these organizations reaches into distinct user-communities. This suggests that both strong and weak ties make the Facebook platform an ideal location for the diffusion of health content [27].

With that said, our analysis also identified several factors that may enhance the diffusion of health content by increasing user engagement. These findings may be of help to CBOs, as unlike social network factors, they are amenable to intervention and change. For instance, we found that posts shared in the morning diffused better than those shared during working hours or after work. These results are consistent with previous studies which show that posts can be strategically timed to take advantage of when users are active e. Similarly, the richness of posts was also shown to be an important covariate of user engagement with higher engagement associated with photos, videos, and links, and lower engagement associated with sharing events. This is consistent with previous research [21] and with Media Richness Theory [37], which suggests that “richer” media (i.e., those with greater ability to efficiently convey messages, social cues, personalization, and feedback) better engages target audiences.

However, contradicting this theory, we also found that specific strategies to engage users, such as asking questions, were associated with lower user engagement. This supports other research which shows that inviting engagement, ironically, may be a less effective way to promote engagement [23]. Other research has shown more generally that traditional marketing elements discourage user engagement on Facebook [22]. This may reflect a distrust for traditional marketing and a desire for more authentic communication [38]. Indeed, Fromm, Butler, and Dickey [39] recommend that marketers approach younger audiences, not as target populations, but as partners in the advertising process. Consistent with this approach, social media strategies should identify ways to authentically promote health with, not to, GBM [21]. With that said, posts with positive affect did elicited higher engagement – perhaps reflecting the well-documented heuristic bias towards positive messaging [23,40].

Closely related to the form of posts, the content of posts was also seen to have a significant effect on user engagement. Posts about PrEP, stigma, and mental health exhibited greater engagement, while posts about dating had lower engagement. While it is difficult to assess why some subjects engaged users better in the current research, these finding may reflect the health priorities, or perhaps current controversies, in gay communities. Therefore, higher user engagement is expected when pages are posting content that might be trending and amenable to gay communities—highlighting the importance of community-conscious agendas for health promotion. Indeed, during the time of this study, community-driven campaigns around PrEP (www.getpreped.ca) and stigma (www.resiststigma.com) may have served as driving forces behind user engagement with posts regarding PrEP, stigma, and mental health. Conversely, posts relating content regarding HIV-related behaviours (e.g., testing and condoms) seemed to attract fewer engagements – potentially highlighting the difficulty of using social media to promote well-established prevention strategies. This may be particularly true for those with which audiences have become fatigued – such as has long been reported among GBM in San Francisco [41]. Based on our results, future analyses should investigate whether integrating better diffusing content – such as PrEP and stigma – into posts promoting more traditional prevention strategies has the potential to improve the diffusion of this content.

Regular assessment of how users are engaging with posts relevant to specific key themes may therefore provide public health and community leaders with insight into the diffusion of social discourse surrounding important topics of concern. To this point, we note significant temporal variation in the frequency at which key themes were included in CBO posts. As mentioned before, PrEP and Stigma increased throughout the observation period likely due to specific prevention campaigns in Metro Vancouver. Similarly, the frequency at which research and testing were discussed increased dramatically during the first half of the observation period, with research-related posts peaking in early 2013 and declining thereafter; and testing-related posts leveling off at the same time As this study was primarily focused on engagement and not CBO’s rationale for content selection, future studies might improve our understanding of what factors contribute to the ebb and flow of specific key themes.

Further, future research should examine individual-level data, particularly that of core users, whom our findings suggest may play an important role in the diffusion of post content. Such examinations might be conducted by each CBO, as they may have greater access and interest in these specific analyses. More generally, our findings also highlight the importance of the user experience in shaping the diffusion of health content. Therefore, ongoing cooperation with users is needed to identify the features that should be leveraged in health promotion—especially as users, not social media specialists, are the ultimate arbiters of whether content is shared with their networks. Consistent with this, CBOs may benefit from examining the network dynamics of their followers and leverage the approaches used in the present study to identify specific users who might be willing to partner with CBOs to promote their content.

These findings should be interpreted with consideration to the limitations of the present study. First, as community-based organizations were not selected using a randomized approach, it is difficult to say whether our findings are generalizable to all Facebook-based health promotion efforts. However, we included most of the major pages associated with organizations in Metro Vancouver. Therefore, our results best represent the health priorities of Vancouver’s gay community, though they may not be the same as those in other communities. Second, as we used relatively simple informatics-based analytic approaches to identify and code posts, our analysis is subject to measurement error. In particular, the selection of key terms may limit the accurate classification of posts relevant to the post features and health messages we explored. However, based on the consistency of our findings with studies conducted regarding other health areas, it seems that our approach produced similar results to studies which included manual-coding techniques [22,23]. Nevertheless, validation of the results of the present study is needed, both in other geographic settings and with other sexual or gender minority communities. Third, as the engagement factors for Facebook reactions, comments, and shares may differ [21,23], further analysis is needed on how to elicit the type of participation that will best promote health awareness. This is especially important given that the predictors of likes, comments, and shares may not be the same. Indeed, as we summed across these three types of user engagement, we may be obscuring important differences or patterns. For example, posts which elicit comments may elicit fewer shares—thus misestimating user engagement with shared posts. Furthermore, Facebook’s edgerank algorithm, which determines whether content is diffused and shown on people’s Facebook pages, is constantly updated and the relative weighting of various types of interaction may change – making it important to understand the unique determinants of various types of engagement (i.e., reactions, comments, shares). Future analyses should thus expand our findings by evaluating the factors associated with specific engagement indicators. Lastly, other important factors, which we have not considered, may also shape user engagement. These include individual-level factors, which require a different analytic and sampling approach to understand how specific user characteristics may shape user engagement. Indeed, while engagement at the individual level is difficult to study, integrating Facebook plug-ins into study questionnaires might allow researchers to match social media participation to survey responses. Other important considerations may also include specific factors which might persuade different individuals to engage with post content — underscoring the need for further examination of gay and bisexual men’s social media engagement. Likewise, exploration of additional themes which were not examined in the present analysis is needed. Indeed, only a minority of posts were relevant to the themes we selected and examined. CBO’s undoubtedly have interest in sharing and promoting content which may not necessarily be directly related to health outcomes studied by public health researchers. Despite these limitations, the present study supports the use of Facebook for health promotion among sexual minorities and highlights multiple factors that can be leveraged to optimize user engagement, thus enhancing the diffusion of health information and the reach of community-based organizations.


1.       Isacco A, Yallum NK, Chromik LC. A Review of Gay Men’s Health: Challenges, Strengths, and Interventions. Am J Lifestyle Med 2012 Jan 1;6(1):45–62. [doi: 10.1177/1559827611402580]

2.       Lee R. Health care problems of lesbian, gay, bisexual, and transgender patients. West J Med 2000 Jun;172(6):403–408. PMID:10854396

3.       Stall R, Mills TC, Williamson J, Hart T, Greenwood G, Paul J, Pollack L, Binson D, Osmond D, Catania JA. Association of Co-Occurring Psychosocial Health Problems and Increased Vulnerability to HIV/AIDS Among Urban Men Who Have Sex With Men. Am J Public Health 2003 Jun;93(6):939–942. PMID:12773359

4.       Singer M. Pathogen-pathogen interaction: a syndemic model of complex biosocial processes in disease. Virulence 2010 Feb;1(1):10–18. PMID:21178409

5.       Singer M, Bulled N, Ostrach B, Mendenhall E. Syndemics and the biosocial conception of health. The Lancet 2017 Mar 4;389(10072):941–950. PMID:28271845, 28271845

6.       Halkitis PN, Wolitski RJ, Millett GA. A holistic approach to addressing HIV infection disparities in gay, bisexual, and other men who have sex with men. Am Psychol 2013 Jun;68(4):261–273. PMID:23688093

7.       Noar SM, Willoughby JF. eHealth interventions for HIV prevention. AIDS Care 2012;24(8):945–952. PMID:22519523

8.       Schnall R, Travers J, Rojas M, Carballo-Diéguez A. eHealth interventions for HIV prevention in high-risk men who have sex with men: a systematic review. J Med Internet Res 2014;16(5):e134. PMID:24862459

9.       Taggart T, Grewe ME, Conserve DF, Gliwa C, Roman Isler M. Social Media and HIV: A Systematic Review of Uses of Social Media in HIV Communication. J Med Internet Res [Internet] 2015 Nov 2 [cited 2016 Jan 30];17(11). PMID:26525289

10.    Kietzmann JH, Hermkens K, McCarthy IP, Silvestre BS. Social media? Get serious! Understanding the functional building blocks of social media. Bus Horiz 2011 May;54(3):241–251. [doi: 10.1016/j.bushor.2011.01.005]

11.    Grov C, Breslow AS, Newcomb ME, Rosenberger JG, Bauermeister JA. Gay and bisexual men’s use of the Internet: research from the 1990s through 2013. J Sex Res 2014;51(4):390–409. PMID:24754360

12.    Harper GW, Serrano PA, Bruce D, Bauermeister JA. The Internet’s Multiple Roles in Facilitating the Sexual Orientation Identity Development of Gay and Bisexual Male Adolescents. Am J Mens Health 2015 Jan 13; PMID:25585861

13.    Korda H, Itani Z. Harnessing Social Media for Health Promotion and Behavior Change. Health Promot Pract 2013 Jan 1;14(1):15–23. [doi: 10.1177/1524839911405850]

14.    Rogers M, Chapman C, Giotsas V. Measuring the diffusion of marketing messages across a social network. J Direct Data Digit Mark Pract 2012 Nov 1;14(2):97–130. [doi: 10.1057/dddmp.2012.25]

15.    Rogers EM. Diffusion of Innovations, 5th Edition. Simon and Schuster; 2003. ISBN:978-0-7432-5823-4

16.    Dennis AR, Kinney ST. Testing Media Richness Theory in the New Media: The Effects of Cues, Feedback, and Task Equivocality. Inf Syst Res 1998 Sep 1;9(3):256–274. [doi: 10.1287/isre.9.3.256]

17.    Facebook. Help Centre [Internet]. [cited 2017 Jan 31]. Available from: https://www.facebook.com/help/166738576721085

18.    Hallett J, Brown G, Maycock B, Langdon P. Changing communities, changing spaces: the challenges of health promotion outreach in cyberspace. Promot Educ 2007;14(3):150–154. PMID:18154224

19.    Cassidy EM. Gay men, social media and self-presentation : managing identities in Gaydar, Facebook and beyond [Internet] [Thesis]. Queensland University of Technology; 2013 [cited 2017 Jan 20]. Available from: http://eprints.qut.edu.au/61773/

20.    Rubin JD, McClelland SI. “Even though it’s a small checkbox, it’s a big deal”: stresses and strains of managing sexual identity(s) on Facebook. Cult Health Sex 2015;17(4):512–526. PMID:25587894

21.    Veale HJ, Sacks-Davis R, Weaver ER, Pedrana AE, Stoové MA, Hellard ME. The use of social networking platforms for sexual health promotion: identifying key strategies for successful user engagement. BMC Public Health 2015;15:85. [doi: 10.1186/s12889-015-1396-z]

22.    Kite J, Foley BC, Grunseit AC, Freeman B. Please Like Me: Facebook and Public Health Communication. PLOS ONE 2016 Sep 15;11(9):e0162765. [doi: 10.1371/journal.pone.0162765]

23.    Rus HM, Cameron LD. Health Communication in Social Media: Message Features Predicting User Engagement on Diabetes-Related Facebook Pages. Ann Behav Med Publ Soc Behav Med 2016 Oct;50(5):678–689. PMID:27059761

24.    Capurro D, Cole K, Echavarría MI, Joe J, Neogi T, Turner AM. The use of social networking sites for public health practice and research: a systematic review. J Med Internet Res 2014 Mar 14;16(3):e79. PMID:24642014

25.    Silenzio VMB, Duberstein PR, Tang W, Lu N, Tu X, Homan CM. Connecting the invisible dots: Reaching lesbian, gay, and bisexual adolescents and young adults at risk for suicide through online social networks. Soc Sci Med 2009 Aug;69(3):469–474. [doi: 10.1016/j.socscimed.2009.05.029]

26.    Caldarelli G. Scale-Free Networks: Complex Webs in Nature and Technology. OUP Oxford; 2007. ISBN:978-0-19-921151-7

27.    Bakshy E, Rosenn I, Marlow C, Adamic L. The Role of Social Networks in Information Diffusion. Proc 21st Int Conf World Wide Web [Internet] New York, NY, USA: ACM; 2012 [cited 2017 Mar 18]. p. 519–528. [doi: 10.1145/2187836.2187907]

28.    Nematzadeh A, Ferrara E, Flammini A, Ahn Y-Y. Optimal network modularity for information diffusion. Phys Rev Lett 2014 Aug 22;113(8):088701. PMID:25192129

29.    Rieder B. Studying Facebook via Data Extraction: The Netvizz Application. Proc 5th Annu ACM Web Sci Conf [Internet] New York, NY, USA: ACM; 2013 [cited 2017 Jan 20]. p. 346–355. [doi: 10.1145/2464464.2464475]

30.    Grant RM, Lama JR, Anderson PL, McMahan V, Liu AY, Vargas L, Goicochea P, Casapía M, Guanira-Carranza JV, Ramirez-Cardich ME, Montoya-Herrera O, Fernández T, Veloso VG, Buchbinder SP, Chariyalertsak S, Schechter M, Bekker L-G, Mayer KH, Kallás EG, Amico KR, Mulligan K, Bushman LR, Hance RJ, Ganoza C, Defechereux P, Postle B, Wang F, McConnell JJ, Zheng J-H, Lee J, Rooney JF, Jaffe HS, Martinez AI, Burns DN, Glidden DV. Preexposure Chemoprophylaxis for HIV Prevention in Men Who Have Sex with Men. N Engl J Med 2010 Dec 30;363(27):2587–2599. PMID:21091279

31.    Jacomy M, Venturini T, Heymann S, Bastian M. ForceAtlas2, a Continuous Graph Layout Algorithm for Handy Network Visualization Designed for the Gephi Software. PLOS ONE 2014 Jun 10;9(6):e98679. [doi: 10.1371/journal.pone.0098679]

32.    Blondel VD, Guillaume J-L, Lambiotte R, Lefebvre E. Fast unfolding of communities in large networks. J Stat Mech Theory Exp 2008;2008(10):P10008. [doi: 10.1088/1742-5468/2008/10/P10008]

33.    Bellazzi R, Diomidous M, Sarkar IN, Takabayashi K, Ziegler A, McCray AT. Data analysis and data mining: current issues in biomedical informatics. Methods Inf Med 2011;50(6):536–544. PMID:22146916

34.    Liu B. Sentiment Analysis and Opinion Mining. San Rafael, Calif.: Morgan & Claypool Publishers; 2012. ISBN:978-1-60845-884-4

35.    Nadkarni PM, Ohno-Machado L, Chapman WW. Natural language processing: an introduction. J Am Med Inform Assoc 2011 Sep 1;18(5):544–551. [doi: 10.1136/amiajnl-2011-000464]

36.    Cvijikj IP, Michahelles F. Online engagement factors on Facebook brand pages. Soc Netw Anal Min 2013 Dec 1;3(4):843–861. [doi: 10.1007/s13278-013-0098-8]

37.    Daft RL, Lengel RH. Organizational Information Requirements, Media Richness and Structural Design. Manag Sci 1986 May 1;32(5):554–571. [doi: 10.1287/mnsc.32.5.554]

38.    Brown S. Marketing to Millennials: Improving Relationships with Millennial Consumers Through Online Advertising and Social Media Networking. Marketing [Internet] 2016 Mar 1; Available from: http://digitalcommons.calpoly.edu/marktsp/1

39.    Fromm J, Butler C, Dickey C. How to engage Millennials: Re-imagining the consumer as a partner, not a target audience, to increase engagement. J Brand Strategy 2015 Mar 1;4(1):27–36.

40.    Dodds PS, Clark EM, Desu S, Frank MR, Reagan AJ, Williams JR, Mitchell L, Harris KD, Kloumann IM, Bagrow JP, Megerdoomian K, McMahon MT, Tivnan BF, Danforth CM. Human language reveals a universal positivity bias. Proc Natl Acad Sci 2015 Feb 24;112(8):2389–2394. PMID:25675475

41.    Stockman JK, Schwarcz SK, Butler LM, de Jong B, Chen SY, Delgado V, McFarland W. HIV Prevention Fatigue Among High-Risk Populations in San Francisco. JAIDS J Acquir Immune Defic Syndr 2004 Apr 1;35(4):432.

42.    Lachowsky NJ, Lal A, Forrest JI, Card KG, Cui Z, Sereda P, Rich A, Raymond HF, Roth EA, Moore DM, Hogg RS. Including Online-Recruited Seeds: A Respondent-Driven Sample of Men Who Have Sex With Men. J Med Internet Res 2016;18(3):e51. PMID:26980147