Spiraling Risk: Visualizing the multilevel factors that socially pattern HIV risk among gay, bisexual & other men who have sex with men using Complex Systems Theory

Purpose of review Global disparities in HIV infection, particularly among gay, bisexual, and other men who have sex with men (GBMSM), indicate the importance of exploring the multi-level processes that shape HIV’s spread. We used Complex Systems Theory and the PRISMA guidelines to conduct a systematic review of 63 global reviews to understand how HIV is socially patterned among GBMSM. The purpose was to conduct a thematic analysis of the reviews to (1) synthesize the multi-level risk factors of HIV risk, (2) categorize risk across the socioecological model, and (3) develop a conceptual model that visualizes the interrelated factors that shape GBMSMS’s HIV “risk.” Recent Findings We included 49 studies of high and moderate quality studies. Results indicated that GBMSM’s HIV risk stems from the individual, interpersonal, and structural levels of the socioecological model. We identified a few themes that shape GBMSM’s risk of HIV infection related to biomedical prevention methods; sexual and sex-seeking behaviors; behavioral prevention methods; individual-level characteristics and syndemic infections; lived experiences and interpersonal relationships; country-level income; country-level HIV prevalence; and structural stigma. The multi-level factors, in tandem, serve to perpetuate GBMSM’s risk of HIV infection globally. Summary The amalgamation of our thematic analyses from our systematic reviews of reviews suggests that the risk of HIV infection operates in an emergent, dynamic, and complex nature across multiple levels of the socioecological model. Applying complex systems theory indicates how multilevel factors create a dynamic and reinforcing system of HIV risk among GBMSM. Supplementary Information The online version contains supplementary material available at 10.1007/s11904-023-00664-y.


Introduction
Gay, bisexual, and other men who have sex with men (GBMSM) around the world experience inequitable disparities in their risk for acquiring HIV. In Europe, 50% of new HIV cases in its western region are among GBMSM [1]. In the Eastern European region, GBMSM HIV transmissions increased eight-fold from 2008-2017 [1,2]. In 2019, in the United States, 69% of new HIV cases were among GBMSM [3] but Black GBMSM in the US account for 37% of new HIV diagnoses among GBMSM [4]. In South America, GBMSM are one of three key population groups, including transgender women and female sex workers, that account for nearly half of HIV infections [5]. In Africa, a meta-analysis indicates that 61% of GBMSM ever tested for HIV, 46% tested within the last 12 months, and only 15% of African GBMSM were engaged in care [6]. In Asia, HIV prevalence among GBMSM increased from 2007 to 2015 and ranged from 11-32% depending on the country [7]. An exploration of the factors that shape GBMSM inequities is paramount.
The literature on health behaviors and HIV among GBMSM show that current per-act probabilities of acquiring HIV, without incorporating any preventive efforts, show the highest risk stems from receptive anal intercourse at 138 infections per 10,000 exposures and, third highest from insertive anal intercourse at 11 infections per 10,000 exposures [8••]. There is upwards of a 90% reduction in HIV risk if condoms are used consistently and correctly [9]. These probabilities can be altered by pre-exposure prophylaxis (PrEP), which indicates nearly 100% protection if the regimens are adhered to appropriately [10••, 11••]. Additionally, Undetectable = Untransmittable indicates that persons living with HIV whose viral load is suppressed through adherence to treatment regimens do not transmit HIV to sexual partners [12••]. The science of HIV risk showcases numerous interrelated factors that can shape risk.
Theoretical models aim to incorporate structural and social determinants of health, which is vital for HIV science. The Health Stigma and Discrimination Framework emphasizes the social, cultural, political, and economic structures that shape stigma itself, and not merely the stigma enacted interpersonally [13]. While the framework is useful theoretically, the framework does not offer specificity in elucidating the various relationships between the social, cultural, and political levels. Rhodes et al.'s " [HIV] risk environment" shifts the focus toward the social and structural processes rather than on endogenous factors of the individual. However, the "risk environment" framework was initially conceptualized in spaces of rapid social, political, and economic transition and again didn't elucidate the specificity of relationships [14]. The socioecological model, often presented as concentric circles, is a model that recognizes relationships between individual, interpersonal, community, organizational, and societal level factors that can all influence health, but the relationships usually are not specific [15]. Numerous studies have examined factors such as stigma, access and quality of services, and policies that can influence HIV risk. Given that multifaceted factors shape HIV risk, studying their relationships, including reinforcing and structural processes, and how the totality of factors shape HIV risk is critical. Theoretically, we can End the HIV Epidemic, yet the science of HIV is not serving all communities equitably.
Complex Systems Theory is arguably a valuable application to study HIV risk. Emergence is a key concept, which refers to the formation of the collective properties (HIV risk) arising from the systems' numerous interactions. Complex systems, as applied to HIV, recognizes that the interacting, interrelated, and dynamic processes are what create one's HIV risk [16]. Complex systems are adaptive and dynamic in nature and have reinforcing processes that interact with one another and create dependencies and feedback loops to different parts of the system, which all, in tandem, serve to create and elevate GBMSM'S HIV risk. The emergent collective properties (HIV risk) cannot be understood by looking at system parts in isolation. Aristotle stated, "the totality is not, as it were, a mere heap, but the whole is something besides the parts…" [17]. The whole is the focus of complex systems, including how the whole gets created and emerges into existence. Complex Systems Theory has been used to visualize and specify the system of factors that shape health outcomes such as obesity, which portrays a high level of complexity of different interrelated factors across levels of the socioecological model [18]. Complex systems have not yet been applied to the body of HIV literature to identify and visualize the multi-level processes that (re)produce and socially pattern GBMSM's HIV "risk" as a product of the system [18].
While GBMSM experience numerous risk factors for HIV that have been well studied in the literature, no single study can examine all the risk factors simultaneously. However, studying the entirety of the HIV literature will allow us to identify and visualize the multi-level factors that (re)produce HIV risk. Therefore, we used the Complex Systems Theory approach to conduct a systematic review of the HIV literature to (1) synthesize the multi-level risk factors of HIV risk, (2) thematically categorize risk across the socioecological model, and (3) develop a conceptual model that visualizes the interrelated factors that shape GBMSMS's HIV "risk."

Materials and Methods
We conducted a systematic review of systematic reviews that aligned with the PRISMA criteria [19]. We worked with a librarian and research analyst to conduct the review. We used the following search terms HIV/AIDS, human immunodeficiency virus, acquired immunodeficiency syndrome; gay, bisexual, men who have sex with men, male-to-male sexual intercourse; systematic review, scoping review, metaanalysis. The full MESH search is provided in a supplementary file. We searched in CINAHL (n = 226), Global Health (n = 267), Psycinfo (n = 128), PubMed (n = 473), Scopus (n = 678), and Web of Science (n = 330) databases. We used the Covidence evidence synthesis software, which has systematized the systematic review process [20]. We identified 2,102 studies and removed 1,249 duplicates, leaving 853 systematic reviews and meta-analyses for screening (Table 1).
To be included, publications were required to satisfy the following criteria: (1) had HIV infection or seropositivity as the primary outcome; (2) studied factors associated with HIV infection, whether quantitatively or qualitatively; (3) had GBMSM as a focus population; (4) were peer-reviewed systematic review or meta-analytic papers; and (6) written in English or Spanish. There was no limit on the timeframe or geographical area of the studies. Given the study is a metasynthesis, we focused on systematic reviews & meta-analyses. Additionally, we used reviews because they interrogate discrepant information that arises in our scientific literature, ensuring high-quality data for visualization of HIV risk as a complex system. Exclusion criteria included: not having HIV infection or serostatus as the main outcome, not global in nature, no incorporation of GBMSM in study samples, non-review articles, and no examination of factors associated with HIV infection or positive serostatus. Two independent reviewers conducted the title and abstract screening (KS and SO). First, we examined titles and abstracts independently and in duplicate for inclusion and exclusion criteria. Any study that did not meet inclusion criteria after the title and abstract review was excluded (n = 633), leaving 220 studies for full-text review. Most studies excluded in this initial screening were either because the outcome of the studies was not HIV infection or positive serostatus or because the studies did not include GBMSM in the sample. During the full-text review, 157 papers were excluded, mainly because they did not explicitly examine factors associated with HIV infection or positive serostatus. This resulted in 63 systematic review studies for data extraction, which totaled 2,448 individual articles. Figure 1 captures the flow of title and abstract screening (reasons for exclusion are not provided at this stage in line with evidence synthesis methods) and full-text review (reasons for exclusion must be provided at this stage) from Covidence.

Analysis and conceptualizing HIV risk
Two reviewers (KS and KG) conducted data extraction and qualitative thematic analysis of the research literature. During the full paper review, we separately extracted the following information: (1) authors; (2) year of publication; (3) geographic area of study; (4) populations under study & their frequency, and (5) identified risk factors associated with the HIV infection outcome within each review. We extracted the information and organized them into tables. We conducted quality assessments of each review using the AMSTAR-2 guidelines for systematic reviews [21]. Supplementary tables one and two present each systematic review's strengths, weaknesses, quality assessment, and the data and results extracted from each systematic review. If multiple populations were present, we only used extracted findings specific to GBMSM. After the primary results were extracted, KS and KG categorized them into higher-order themes and respective subthemes under each level of the socioecological model (Table 2) using Bronfenbrenner's social-ecological model. For example, condom use, the number of sexual partners, and sexual position were coded as sexual and sex-seeking behaviors (theme) under the individual level of the socioecological model. We only extracted, synthesized, and visualized data from articles with moderate or high-quality assessments (n = 49). KS used the systems mapping software, Kumu.io, to create a conceptual model [22]. The output is a conceptual model that visually represents the HIV risk environment for GBMSM, which comprises the risk factors and their potential interrelationships identified through the systematic review of systematic reviews. The risk factors identified in the review were thematically analyzed to identify the relationships between the various risk factors and HIV infection. For each risk factor (e.g., sexual behaviors), we identified relationships from the literature that conceptualized specific HIV risk pathways (e.g., no condom use < -> increased HIV risk). We used text to represent the relationships between the risk factors (e.g., decreased condom use < -> increased HIV risk). Lines portray which extracted factors are directionally associated with each other. If the studies were quantitative associational/cross-sectional, we used nondirectional arrows. We utilized unidirectional arrows when the quantitative research examined a potential causal pathway (e.g., cohort studies, intervention studies). Quantitative studies that addressed confounding were of particular interest if observational. For qualitative studies, we used major thematic outputs (i.e., family/main theme), especially for studies that had methods to achieve data saturation. In the conceptual model, we aimed to ensure the directionality of arrows was theoretically relevant and aligned with the study design (e.g., limited PrEP policies and financing would influence access and use of PrEP, rather than vice versa). The conceptual model is a visual representation of the thematic analyses that portray the complex and diverse nature that systematically pattern HIV risk among GBMSM's. The complex systems model is a dynamic system model and thus is not meant to serve as a static visualization despite its representation within the text. If studies found negative relationships, theoretically, in a complex system, that pathway would "turn off." The dynamic model, which respects the principles of complex systems, is available in an onlin e suppl ement ary file.

Results
We included 49 systematic reviews and meta-analyses of high and moderate quality, which included a total of 1,721 studies. Among the included reviews, 35 were global (two or more continents), five only in the United States, six only in China, one in Brazil, one in Africa, and one in Europe. Thirty-two reviews explicitly focused on GBMSM, seven extended the study population beyond GBMSM to include high-risk populations such as injection drug users, sex workers, transgender people, incarcerated people, and LGBT populations more broadly. The remaining ten reviews were broader but included GBMSM in the inclusion criteria for the review. The risk of bias assessments indicated that 35 reviews were of high quality and 14 were of moderate quality. Four reviews were low quality, and nine were critically low quality (of which their data are not presented here). The

Identified themes
Individual-level of the socioecological model Theme 1A: Biomedical prevention methods shape HIV infection There were four reviews, one in Europe and three globally that provided high-quality information that the use of pre-exposure prophylaxis (PrEP) reduces HIV infection, emphasizing that greater adherence leads to more protection [22][23][24]25••]. Wang et al. reviewed 74 studies reporting that post-exposure prophylaxis (PEP) reduces HIV risk, reporting 2.6% seroconversions were observed among    between age and HIV infection were mixed. Two highquality reviews indicated that younger age was associated with more significant HIV infection [32, 34••]. In contrast, one moderate quality review showed older aged MSM had higher rates of HIV [30••]. The association between circumcision status and HIV risk has been studied extensively with mixed results. Six systematic reviews examined circumcisions and HIV, of which four reviews of three high quality and one moderate quality found no significant relationship with reduced HIV infection [25••, 40-42]. Two other high-quality reviews found some protection against HIV with male circumcision but mostly in low-middle-income countries and among insertive partners [44,45]. The two reviews suggest that mixed results may be due to variations in study design, geography, or the country's economic status.  [47]. However, the risk among incarcerated GBMSM varied by geography such that HIV prevalence among incarcerated GBMSM is 10 times higher in Latin America and 20 times higher in Western Europe [47]. Another high-quality review indicated that the continuity of HIV services, such as testing and testing pre-and-post release could reduce HIV cases among Black GBMSM who experience incarceration and reenter their communities post-release [48]. Three systematic reviews (two moderate and one high quality) examined the relationship between HIV infection and experiences of abuse (childhood and intimate partner violence), all indicating that GBMSM who experienced abuse had elevated HIV infections [30••, 48, 49]. Lastly, a review examined motivational interviewing as an intervention to support the prevention of HIV among MSM. However, the results concluded that evidence of motivational interviewing's effectiveness was lacking [51]. The various interpersonal-level factors that shape HIV infection serve to further complicate conceptualizations of HIV "risk."  [53••]. Countries with protective language for men who sell sex had a 7% lower prevalence of HIV, as compared to those without such protections [54 ••]. A systematic review and meta-analysis of 75 papers found that GBMSM were more likely to be aware of their HIV status in countries with less repressive legislation (22% vs. 6.7%) and less severe penalties for same-sex relations [55]. These findings indicate that structural factors could influence HIV infection and complicate our understanding of HIV risk further.

Dynamic and emergent nature of HIV risk
In totality, the amalgamation of our thematic analyses and findings from our systematic reviews of reviews suggests that the risk of HIV infection operates in an emergent, dynamic, and complex nature across multiple levels of the socioecological model that interact with one another to elevate GBMSMS's risk of HIV further. Structural factors such as stigmatizing policies, macroeconomic factors, and population-level epidemiology can shape environments in ways that produce HIV infection-moreover, interpersonal factors such as experiences of abuse and homophobia further shape HIV infection among GBMSM. Lastly, numerous individuallevel factors such as sexual behaviors, use of biomedical prevention methods, and sexual relationships, amongst others, further perpetuate the risk of HIV infection among GBMSM. This systematic review of reviews and meta-analyses and the developed visualization indicated that HIV risk among GBMSM worldwide is socially patterned by numerous interrelated factors, such that the ecosystem itself is the driver of the disproportionate risk of HIV that burdens GBMSM (Fig. 2).

Discussion
The results of our systematic review portray that, globally, HIV infections among GBMSM arise from a complex interplay of structural, interpersonal, and individual-level factors. The visualization positions that structural, interpersonal, and individual-level forces operate in concert to structure GBMSM's risk of HIV infection. The totality of these relationships influences the emergence of HIV infection in ways that disproportionately burden GBMSM throughout the globe. The analyses and conceptual model suggest that focusing on the system that shapes HIV infection is needed to address the interactional and collective nature of HIV risk that GBMSM experience.
The growing application of complex systems models improves our understanding of how several intersecting processes shape population-level HIV infection and prevalence. A recent agent-based model (ABM) of a complex system explored U.S. racial inequities in PrEP use and showed that PrEP uptake will reduce overall HIV infections, it will do nothing to address the HIV disparities currently existing [56]. In the model, when Black and White agents use PrEP at equal rates, the HIV disparity ratio between the two groups continues to increase because the underlying disparity was never addressed [56]. Goedel's findings support the need to simultaneously address the diversity of risk factors such as accessibility, acceptability, and quality of healthcare service, which align with the results of our study, portraying the system that shapes HIV infection. This aligns with a metaanalysis of 194 studies in Canada, the United States, and the United Kingdom, which showed that Black GBMSM were more likely to engage in individual-level prevention behaviors (e.g., fewer sexual partners, more condom use, less likely to use substances), as compared to White GBMSM, yet had greater odds of testing positive for HIV and six-fold greater odds of having undiagnosed HIV [57]. However, as our systematic review indicates, the scientific literature still predominately focuses on individual-level risk factors for HIV, as evidenced by the 68% of articles from 2000-2021 that focused on individual-level factors. Our findings add further credence to how HIV infection arises from a complex system of factors, not simply individual-level behaviors, and indicate that enhanced conceptualizations of HIV "risk" are needed.
This is the first study to identify and visualize the interacting processes that socially pattern the emergence of HIV risk and infection among GBMSM. Global responses, such as the Joint United Nations Programme on HIV/AIDS (UNAIDS), must adequately address the system of factors together, instead of in isolation, to achieve zero new HIV infections. Scholarship implores us to think of interventions not merely as a "package" of activities but rather, 1 3 alternatively, to focus on the dynamic processes of the environment and context where interventions are introduced [58]. The focus on the dynamic processes is significant for HIV, which, as our results show, arises from multi-faceted routes across the socioecological model. Studies have also shown how country-level policies shape the ecosystem of HIV risk in ways that elevate GBMSM's risk of HIV [59]. Research also identifies how combining multiple HIV intervention strategies may be necessary given the complexity in which HIV risk is socially patterned [60]. Further research needs to examine how risk factors are situated within the social and structural contexts to fully grapple with the multiple levels of influence that can weaken or support the capacity of GBMSM to reduce their risk of HIV infection.
Limitations in our synthesis include aspects of the approach used and generalizability. While we used a systematic review of review studies, the system of processes outlined in Fig. 2 is only partial. This is especially true as our primary outcome was HIV infection, and we didn't include studies that might have explored risk factors for the risk factors themselves. The dynamic processes will change over time as scientific research and interventions occur. Many of the pathways developed were identified as associations, given the cross-sectional nature of much of our HIV research literature. To enhance the rigor, we removed the 22% of reviews that were low or critically low quality. Lastly, given that our data focused on reviews or meta-analyses rather than individual studies, we may be missing risk domains if no review study exists for that domain. However, focusing on reviews and meta-analyses allowed for a meta-review of our scientific literature on HIV risk to begin to grapple with the complexity of "HIV risk." Future research should further explore the system of HIV risk by examining what may shape the HIV risk factors (risks of risks). Automating systematic review processes using natural language processing (NLP) and machine learning could enhance the synthesis of the research literature. NLP is a collection of algorithms that can be used to identify, extract, parse, and analyze textual data, such as written text in journal articles, which is growing with tools such as GPT, ElicitAI, amongst others [61]. The application of Complex Systems Theory and our conceptual model can be leveraged to better hypothesize and study the mechanisms that shape population-level HIV inequities among priority populations across different environments, geographies, and interventions.

Conclusion
The thematic analysis and visualization indicate that HIV infection among GBMSM, globally, is socially patterned by a diversity of numerous interacting risk factors. Applying complex systems theory indicates how multilevel factors create a dynamic and reinforcing system of HIV risk among GBMSM. Having this more complete understanding of what shapes HIV risk is critical for developing more effective HIV prevention efforts for highly burdened populations and for strengthening the global commitment to achieve zero new HIV infections. Author contributions KS, EK, and AG (equal) conceptualized the study. KS led the writing of the original draft. KS, SO, and KG (equal) curated the data and conducted screening and review. KS and KG (equal) conducted the data extraction and analysis. KG supported writing the results. EK, AG, and KT (equal) reviewed drafts, and supported with writing and editing.
Funding This work was supported by the Partners for Advancing Health Equity grant (78477) through the Robert Wood Johnson Foundation.
Data Availability Data are available as tables in supplementary files.

Conflict of interest
The authors have no conflicts.

Human and Animal Rights and Informed Consent
This article does not contain any studies with human or animal subjects performed by any of the authors.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.