GLOBAL EPIDEMIOLOGY OF HIV AMONG FEMALE SEX WORKERS: INFLUENCE OF STRUCTURAL DETERMINANTS

Systematic review

To consider the centrality of structural determinants in HIV epidemics among FSWs, we mapped present epidemiological reports on a structural determinants framework for HIV and sex work (figure 1).¹⁹ Of the 3214 relevant publications retrieved, 149 (73%) of 204 unique studies reported at least one structural determinant (table, appendix), but only 87 (43%) were designed a priori to examine one or more structural determinants of HIV, HIV and sexually transmitted infection (STI), or condom use. Of these, 81 (93%) were from low-income and middle-income countries, with most from Asia (35 from India and 13 from China), followed by Latin America and the Caribbean (14 studies). Few reports originated from heavy or increasing HIV burden settings of sub-Saharan Africa, Russia, or eastern Europe. Most studies documented associations between structural factors and condom use, rather than HIV (or HIV and STI) infection.

Macrostructural factors

Macrostructural factors increasingly play a central part in HIV epidemic structures among FSWs and operate in iterative pathways with other structural determinants and behavioural and biological factors. An increasing number of reports show how punitive laws and policies governing sex work, including criminalisation of some or all aspects of sex work, incarceration, demolition of red-light districts (due to punitive policy changes) and legal restrictions on where sex workers operate, elevate HIV acquisition and transmission risks.^{23, 24} Criminalisation and punitive policies on sex work have been shown to enact stigma,^{25, 26} increase food and economic insecurity,^{27, 28} and residential instability (due to evictions)²⁹ among FSWs, and have been associated with inconsistent condom use. Where FSWs have constrained rights and access to resources due to their status as sex workers and women, gender-based violence^{26, 30} and sex trafficking (forced sexual labour)^{28, 30, 31} have been consistently linked to increased odds of HIV infection and condom non-use. By contrast, gender empowerment²⁹ and higher education and literacy^{29, 32} continue to mitigate HIV risk among FSWs.

Migration and mobility have particularly complex and non-linear effects on HIV risk pathways among FSWs, both mitigating and conferring HIV risk. Internal domestic and circular migration and mobility (eg, intraurban or intradistrict mobility, and short-term travel to sex-work hotspots)^{33, 34, 35, 36} have been associated with enhanced HIV vulnerability, whereas long durations of mobility and international migration from non-endemic settings have been linked to high rates of condom use³⁷ and low HIV prevalences.³⁸ The complexity of migration and HIV is characterised by other geographical features and epidemic structures (eg, immigration or emigration to higher-prevalence settings and rural vs urban migration) that might confer or mitigate HIV risk among FSWs.

Community organisation

The role of community organisation in the reduction of HIV risk through increased condom use and lower HIV prevalence among FSWs at the population level is increasingly important. The available data for sex worker organisation is largely restricted to India,^{25, 39, 40, 41, 42, 43} where substantial resource investments have been made in community empowerment efforts among sex workers as part of broader HIV prevention efforts. Aside from a few reports from China,³² Kenya,⁴⁴ and Latin America,^{45, 46} gaps in data for community empowerment up to now can be attributed in large part to scarcity of resources to assess grassroots sex worker efforts and macrostructural constraints of criminalisation and stigma that restrict the ability of FSWs to organise. Community organisation has been considered both as a broad process or intervention of community empowerment, as exemplified by Songachi and Avahan models in India,^{25, 39, 47} or as a specific component of community organisation (eg, sex work drop-in spaces or social cohesion). For example, substantial increases in condom use (and decreased STI prevalence) were noted after the Ashodaya sex work collective intervention in Mysore, India, which combined community mobilisation, sex worker-led outreach, sex worker advocacy to police and local government, and improved sexual health services tailored to sex workers.²⁵

Work environment

The work environment consists of intersecting social, physical, policy, and economic features of places within which FSWs work (eg, venues, streets and public spaces, online, and other off-street self-advertising spaces).

The social features of work environments assessed in more than 100 publications are measured as both downstream products of and interactions with macrostructural factors (eg, laws and policies). Within criminalised environments, physical and sexual violence in the workplace, whether by clients, police, managers, pimps, or predators posing as clients, are among the most ubiquitous and influential determinants of HIV acquisition and transmission risk among FSWs, linked to inconsistent condom use, client condom refusal, condom use failure and breakage,^{22, 24, 26, 28, 33, 43, 46, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57} and HIV infection.^{5, 33, 58, 59, 60, 61, 62} In addition to police abuse as a human rights violation, law enforcement strategies and local policing of sex work, including arrests and incarceration,^{26, 38, 61, 63, 64, 65} raids,⁶⁴ displacement,²⁴ and confiscation of condoms or syringes,^{64, 66} are key barriers to HIV prevention efforts among FSWs worldwide, which reduces or eliminates the ability to negotiate male condom use^{24, 51, 61, 63, 64, 66} and increases HIV prevalence and incidence.^{38, 65, 66}

Physical features of the work environment (eg, venue characteristics or typology) are very context-specific and heterogeneous in structuring of HIV risk patterning in FSWs.^{23, 24, 34, 37, 41, 55, 57, 63, 67, 68, 69, 70, 71, 72, 73, 74, 75} Few investigators have examined the nuanced and intersecting influence of policy, social, and physical features of the work environment in the mitigation or conferment of HIV risk among FSWs. The authors of these studies examined how supportive venue-based policies^{30, 32, 76, 77, 78, 79, 80, 81} or managerial practices (eg, client sign-in, safety mechanisms, or removal of violent clients)^{48, 69, 76, 77, 80, 81, 82, 83, 84, 85} and physical features (eg, types or layout of venues) of sex work establishments are associated with increased condom use, often through synergistic effects with other social features of increased peer or sex worker support.^{32, 52, 53, 69, 72, 76, 79, 81, 83, 85, 86, 87, 88, 89, 90} Work environments for FSWs are shaped by economic features (eg, economic pressures, client financial incentives for non-condom use, refusal of payment, and bribes or fines by state agents [eg, police] to avoid arrest) resulting from macrostructural forces of poverty, laws, and access to resources and are associated with non-condom use^{27, 28, 29, 30, 40, 41, 48, 64, 70, 83, 84, 87, 91, 92, 93} and HIV infection among FSWs.^{36, 65, 75, 94} Conversely, higher income and absence of economic dependence among FSWs mitigate HIV risks, including increased condom use^{40, 48, 80, 86, 87} and lower HIV prevalence.^{43, 71, 73, 74, 81, 94}

Suboptimum access to safe and appropriate condoms, sexual health care, HIV testing, and ART remain major shortfalls in the fight against the worldwide HIV epidemic among FSWs. Where FSWs report adequate access to condoms, sexual health care (eg, STI testing and contraceptives) and HIV care (eg, HIV testing, ART, and sex worker-tailored clinics), increases are noted in condom use and reduced condom breakage,^{39, 40, 32, 44, 47, 67, 79, 95, 96, 97, 98, 99} and lower HIV prevalence.^{100, 101} Condom coverage must include condom access (eg, free or subsidised condoms at the workplace, the ability to carry condoms while working, and contact with peer condom distribution), availability and affordability, linked to reduced HIV acquisition and transmission among FSWs.^{25, 30, 39, 40, 44, 47, 68, 70, 79, 82, 84, 87, 93, 95, 97, 98, 99, 102, 103, 104, 105, 106, 107}

Case 1: Vancouver, Canada

Vancouver provides a key example of a concentrated, high-prevalence epidemic among key populations. As elsewhere in North America, the epidemic in Vancouver first emerged in men who have sex with men in the 1980s.¹¹⁰ By the mid-1990s, an explosive HIV outbreak occurred in people who inject drugs, with HIV incidence peaking at around 18 cases per 100 person-years in 1996 and 1997,¹¹¹ declining to fewer than three per 100 person-years in 2007 because of substantially improved coverage of syringe distribution, other harm reduction, and ART.¹¹² The epidemic among street FSWs emerged alongside the epidemic among people who inject drugs, partly due to overlap between drug-using FSWs, clients, and non-paying partners.^{113, 114} Because HIV incidence due to syringe sharing declined substantially, a temporal shift has occurred from illicit drug injection-acquired to sexually transmitted HIV infections in people who inject drugs and FSWs in the past 10 years.^{115, 116} Alongside the HIV epidemic and decline of HIV within people who inject drugs, STI epidemics have continued to escalate, suggesting that sexual risks precipitate HIV transmission in key populations.¹¹⁶ Data for non-injecting FSWs previous to 2010 are scarce. Nevertheless, modelling suggests that overall HIV prevalence in FSWs might have peaked at around 18% in 2000 and declined to 12% in 2011 (figure 3), and that HIV transmission after 2002 seems to be due predominantly to sexual transmission and, to a lesser extent, unsafe injection practices. Many infections have been averted with increased access to the HIV care continuum (eg, HIV testing and ART) through government-sponsored treatment as prevention efforts, although suboptimal access still remains among women and other subpopulations of indigenous and migrant FSWs (Deering et al, unpublished).¹¹⁷

Work environments in Canada

Patterns of HIV prevalence in FSWs are very heterogeneous by work environment exposures. In Vancouver, estimates put HIV prevalence at 20% in FSWs working in informal indoor venues (eg, bars or hotels), 12% on the street, and 3% in formal sex work establishments (eg, in-call venues).¹¹⁸ Because work environments represent a dynamic interplay of changing structural conditions (eg, legal changes and police crackdowns), historical exposure to street-based sex work¹¹⁴ could remain important to understand the epidemic structure among FSWs in concentrated epidemics.

Despite the criminalised nature of sex work in Canada, in-call venues have existed for a long time in the form of licensed brothels (ie, massage parlours and health enhancement centres) and microbrothels. As described by FSWs (panel 2), efforts have been made to implement new models of safer indoor work spaces within women-only, low-income housing (eg, with supportive management policies and safety measures onsite).¹²³ Longitudinal data suggests that working in formal or in-call sex work establishments reduces rates of non-condom use¹¹⁸ and STIs¹²⁴ compared with street and informal indoor venues. Modelling predicts that interventions to promote access to safer work environments for FSWs could avert 37% (95% UI 16–61) of HIV infections among FSWs and their clients in Vancouver in the next decade through the combined effect on reduced violence, police harassment, and non-condom use (figure 3).

Criminalisation, policing, and violence in Canada

Within a criminalised environment, workplace sexual violence is very prevalent and seems to have a sustained negative effect on condom use (Figure 2 and Figure 3). 30% of FSWs have experienced police harassment or workplace violence in formal indoor establishments compared with 70% in informal indoor or outdoor venues.¹¹⁸ The elimination of future exposure to sexual violence might only have a small effect (2% [95% UI 1–4] in 10 years) in view of the sustained negative effects on condom use negotiation. However, if support and resources were in place to address both immediate and sustained effects of historical sexual violence among FSWs (eg, peer support, counselling, and housing) on non-condom use, the elimination of client sexual violence alone could avert 20% (95% UI 3–39) of HIV infections among FSWs and their clients over the next decade in Vancouver (figure 3).

As described by sex workers (panel 2), police harassment (without arrest) can directly influence HIV acquisition risk by forcing FSWs to rush transactions with their clients, forgo condoms, or engage in risky sexual practices, or by displacement of FSWs to isolated or hidden venues, where they have less ability to control transactions (eg, client selection, types of sexual acts, or condom use).¹²⁰ Police harassment without arrest (eg, raids, forced detainment, or coercion) is very prevalent in Vancouver—62% of FSWs report ever experiencing police harassment in their lifetimes and 40% do so within the past 6 months. Recent exposure to police harassment is estimated to increase the rate of client violence by 1·0–5·2 times¹²⁵ due to reduced ability to screen clients and displacement to more isolated and hidden venues, thereby increasing FSWs' risk of HIV acquisition (figure 2). Police harassment and physical violence alone have an immediate and transient effect on the reduction of condom use after a recent exposure to the violence, whereas sexual violence seems to negatively affect condom use in the short and long term (figure 2). Because of the dynamic and iterative effects of violence and policing, the independent addressing of police harassment or physical violence alone would have a negligible effect. However, elimination of both police harassment and sexual and physical violence coupled with support to address long-term effects of violence could avert 24% (95% UI 8–45) of HIV infections among FSWs and their clients in Vancouver in the next decade (figure 3) through its direct and sustained effect on condom use.

In Canada, a Supreme Court decision to strike down criminal laws targeting sex work has allowed the government to potentially decriminalise sex work by the end of 2014, providing a unique opportunity to model the potential effect of law reforms on the HIV epidemic. Our modelling suggests that nearly 39% (95% UI 16–63) of infections could be averted among FSWs and their clients in the next decade by decriminalisation of sex work (figure 3) through immediate and sustained effects on violence, police harassment, and safe work environments, and associated condom use. These predictions represent the maximum effect that interventions reducing violence or police harassment or legislation can have because the complete elimination of violence or stigma could be challenging.

Case 2: Mombasa, Kenya

Kenya has a generalised epidemic of HIV, with prevalence remaining about 6–7% since 2007.¹²⁶ FSWs have the highest HIV burden in Kenya, with an explosive HIV epidemic emerging in the 1980s among FSWs, peaking at 80% in 1983.¹²⁷ HIV incidence declined substantially after the scale-up of HIV prevention and treatment for FSWs in Nairobi from an estimated 18 cases per 100 person-years in 1985 to fewer than five cases per 100 person-years in 2005.¹²⁷ Estimates suggest that 29·3–47·0% of all Kenyan FSWs are living with HIV.^{44, 127, 128} Continued high HIV burden among FSWs and absence of sustained coverage of HIV prevention and care services tailored to FSWs is largely attributed to a dearth of strategies to mitigate structural determinants of HIV (eg, stigma, discrimination, violence, and criminalisation).¹²⁸

Mombasa is an economic centre of Kenya's Coast Province, with important tourism, port, rail, and industrial enterprises serving as a client base, and a large FSW population (appendix). An estimated 6% of the female population in Mombasa engaged in sex work in the past 12 months.⁴⁴ Modelling results show that with increased condom use, HIV prevalence (and incidence) among Mombasa FSWs declined from roughly 50% (18 cases per 100 person-years) in the mid-1990s to less than 30% (fewer than five cases per 100 person-years) in 2010–15 (figure 4).^{44, 129}

Since 2006, ART has been provided to individuals with HIV for free in Kenya. Despite high coverage in the general Kenyan population (estimated at 83% of ART-eligible people in 2011; ie, roughly 40% of all HIV-positive people),¹²⁶ about 20–30% of all HIV-positive FSWs were reportedly receiving ART based on a CD4 cell count of less than 200 cells per μL.130 Once on ART, much the same rates of retention and virological suppression were reported among FSWs compared with the general population of women, highlighting existing gaps in HIV testing and access to and initiation of ART. As described by sex workers (panel 2), denial of health care access, including ART refusal, is a major structural barrier to HIV prevention and care among FSWs in Kenya, as elsewhere in East Africa, and is associated with heavy criminalisation, stigma, and violence.¹² Scale-up of ART access to meet WHO guidelines of a CD4 cell count of less than 500 cells per μL for FSWs would have substantial effect (appendix), with the largest gains on the epidemic seen in ART scale-up for both FSWs and their clients, averting 34% (95% UI 25–42) of HIV infections over the next decade (figure 4), although evidence suggests that removal of structural barriers, including law reform and stigma, remain crucial.^{1, 12} In addition to the critical need for ART access for FSWs, the ART prevention capacity, particularly in generalised epidemics, of infection avoidance among FSWs by enhanced ART coverage for clients remains pivotal.

Criminalisation and client, police, and stranger sexual violence in Kenya

Sex work is criminalised by both national laws and municipal bylaws in Kenya, with sex worker arrests for loitering for the purpose of selling sex, importuning, and indecent exposure, rather than for clients. Kenyan sex workers describe similar experiences to Vancouver—of how criminalisation compromises HIV prevention by rushed transactions due to fear of arrest, bribes, extortion, sexual coercion, and the forgoing of condoms, and the deterrence of sex workers from reporting violence to authorities (panel 2).¹²²

Sexual violence against FSWs by police, clients, or strangers is endemic in Kenya, with estimates in Mombasa ranging from 29% to 32% in 2007.¹²⁹ Sexual violence is precipitated by macrostructural determinants, including stigma, criminalisation, and gendered cultural norms (panel 2).^{44, 129} Modelling suggests that elimination of sexual violence could avert 17% (95% UI 1–31) of HIV infections among FSWs and their clients over the next 10 years, coupled with support to address long-term effects of historical sexual violence, due to its immediate and sustained effect on non-condom use (figure 4, appendix).

Work environments in Kenya

Sex work in Mombasa is most common within entertainment establishments (72% in bars or nightclubs), followed by homes (21%) and the street (7%).⁴⁴ Sparse data exists for features of the work environment, although binge alcohol use is widespread in bars and associated with a reduced ability among FSWs to negotiate client condom use and condom breakage, heightened sexual violence, and HIV infection.¹²⁹ Substance use in the coastal province is common, with 33% of FSWs estimated to binge drink and most (80%) reporting exchanging sex while intoxicated, with increasing use of khat (77%; amphetamine derivatives), heroin, cocaine, and glue (6%).^{129, 131} No data for client substance use patterns exist, although estimates among the general male population in Mombasa are similar to that of FSWs.^{129, 131} Nightclub-based FSWs (who report the most binge drinking) are younger, earn more, and use condoms the most compared with bar or lodge-based FSWs, who charge less and use condoms less frequently, showing the influence of social and physical features of bar venues.¹²⁹ Modelling suggests that 21% (95% UI 5–35) of HIV infections of FSWs and their clients could be averted through safer work environments over 10 years (figure 4) through combined elimination of increased sexual violence and non-condom use risks associated with binge drinking. Interventions that moderate the work environment (eg, supportive management and venue-based policies, and peer harm reduction) remain pivotal.

Peer and sex worker-led interventions in Kenya

Mapping estimates suggest that large gaps exist in condom access for FSWs, with an estimated 25% of bars with condom dispensers and 29% accessed by peer outreach.^{105, 132} Sex worker-led outreach can promote consistent condom use directly through enhanced condom coverage in venues and indirectly through peer education around condom negotiation with clients and shifting sex industry norms on condom use.^{32, 69, 72, 76, 79, 81, 83, 85, 86, 87, 88, 89, 90} Sex worker-led interventions are a key component of community empowerment, as shown in several low-income and middle-income countries. In Mombasa, peer or sex worker-led outreach implemented between 2000 and 2005 was associated with a more than three-fold increase in consistent condom use,⁴⁴ although coverage was less than 30%. Human rights violations and harassment of outreach workers have been reported in Mombasa, which could deter scale up in the absence of structural change (panel 2).¹¹⁹ Scaling up of sex worker-led outreach to even modest coverage could prevent 20% (95% UI 8–36) of HIV infections in the next 10 years (figure 4). Combined structural changes to decriminalise sex work could avert 33% (95% UI 13–53) of HIV infections among FSWs and their clients over the next 10 years through a combined effect on sexual violence, safer work environments, and increased condom use (figure 4).

Case 3: Bellary, India

South India is experiencing a predominately heterosexual HIV epidemic among FSWs.133 Although national adult HIV prevalence is less than 1%, India has the largest burden of HIV after South Africa and Nigeria, with an estimated 2·5 million people living with HIV. In the past decade, India has seen substantial progress in the curbing of HIV prevalence among FSWs, although substantial heterogeneity exists. FSWs remain 54-times more likely to be infected with HIV over their lifetime compared with the general population of women of reproductive age.²

In response to a growing HIV epidemic among FSWs in 2003 in Karnataka and other high-prevalence states, the Bill & Melinda Gates's Avahan, Indian AIDS initiative rapidly scaled up HIV prevention efforts. By 2008, more than 75% of the estimated target population (217 000 FSWs) in the 69 Avahan districts across four southern states were being contacted monthly.¹³⁴

Karnataka (population ~54 million) ranks in the top four Indian states for HIV epidemic severity. Bellary, a high-prevalence district (overall HIV prevalence of 1·2% in 2011) in central Karnataka has a predominately urban epidemic (population of Bellary ~2 million), concentrated among FSWs and men who have sex with men.⁹⁶ Work environments vary substantially between districts, with home or independent-based and street-based sex work, and to a lesser extent, brothel-based work, with an increased use of phone or text communications and mobility.⁴⁹ As in Mombasa, industries (mining and chalking) in urban Bellary provide a large client base. Avahan was initiated in Bellary in 2004 and substantial progress has been made (a reduction in HIV prevalence from 15·7% in 2005 to 6·4% in 2011; figure 5).71

Alongside Avahan, a slower rollout of ART across India began in 2004, with eligibility criteria shifting from a CD4 cell count of up to 250–350 cells per μL in 2012 to meet international guidelines.¹³⁵ By 2012, 19% of the 2·3 million adults with HIV in India were estimated to be on ART, with much variability in coverage.¹³⁶ However, because referral for semi-annual HIV screening among FSWs is a key component of India's targeted intervention programme,¹³⁷ HIV screening among FSWs is higher than it is for the overall population, although linkage to and retention of ART could be lower.¹³⁷

Criminalisation, violence, and policing in India

Although buying or selling of sex is not illegal in India, the Indian Immoral Traffic Prevention Act, enacted in 1956, criminalises most aspects of sex work, including solicitation and living off the avails by third parties. Violence is common, although in Karnataka, structural and sex worker-led interventions (policy level, police or stakeholder engagement, and training) have reduced violence against FSWs.^{22, 25} Prevalence of recent client physical and sexual violence in Bellary reduced from 35% to 9% between 2006 and 2008.^{22, 33} Elimination of sexual and physical violence could avert a further 5% (95% UI 2–14) of HIV infections among FSWs and their clients over the next 10 years through its immediate effect on condom use (figure 5). Fear of condom confiscation due to recent arrest or intimidation is common and associated with a transient increased risk of non-condom use.^{7, 119} Based on the large-scale reductions in violence already reported because of structural and sex worker-led efforts, elimination of both police condom confiscation and physical and sexual violence through their immediate and sustained effect on condom use could avert 6% (95% UI 2–15) of HIV infections among FSWs and their clients over the next decade (figure 5). However, decriminalisation of sex work would have the largest effect: it would avert 46% (95% UI 33–56) of HIV infections among FSWs and their clients over the next decade (figure 5) through its immediate and sustained effect on violence, policing, and safer work environments.

Community empowerment in India

After the Sonagachi model in Kolkata,¹³⁸ the Avahan intervention in Karnataka and other states included structural and community empowerment components to address both macrostructural (eg, stigma) and work environment (eg, violence) determinants of HIV. Ashodaya, the first of the Avahan interventions, is now regarded as best practice.^{25, 121} Avahan assessments suggest that the combined multicomponent structural and community empowerment intervention could have averted 33–62% of all infections in FSWs and 39–66% of all infections in clients within 5 years in Bellary⁹⁷ through membership in sex work collectives, access to STI clinics and peer-led outreach, 24 h crisis lines and drop-in centres, and sex worker advocacy efforts directed towards local government and police (table).²¹

Systematic review and modelling

The role that structural determinants have in HIV epidemics and access to care for female sex workers (FSWs) is poorly understood. We aimed to review available published data for HIV prevalence and incidence, condom use, and structural determinants among this group through a systematic search of published sources. We extracted details regarding study information (year, setting, study dates, and participants); study design (sample size, geographical location, type of study [longitudinal or cross-sectional], and whether pathways were explored through interaction or stratified analyses); specific outcomes of interest (HIV infection, HIV or sexually transmitted disease infection, or condom use); individual behavioural or biological partner or dyad factors, or structural factors (risks or protective). We recorded whether or not each study had aimed, a priori, to measure and analyse structural determinants of HIV. We then modelled key structural drivers from the systematic review to assess the population-level effect of structural changes using deterministic transmission dynamic models to simulate the course of HIV epidemics and potential HIV infections averted in regions with concentrated and generalised epidemics, and high HIV prevalence among FSWs (Canada, Kenya, and India). Full details of the search and modelling are presented in the appendix.

Structural determinants

To identify structural determinants, we mapped present epidemiological reports on a structural determinants framework for HIV and sex work¹⁹ (figure 1, table), and drew on earlier work by Diez Roux and Aiello¹⁷, Blanchard and Aral¹⁶, Rhodes and colleagues¹⁸, Overs²⁰, Strathdee¹³, and others. Structural determinants (risk or protective factors) operate at macrostructural (eg, social, economic, and health-related policies governing sex work; mobility and migration of sex workers and their clients; geography and sociopolitical transitions; and stigma and cultural norms), community organisation (eg, community empowerment, sex work collectivisation, or leadership), and work environment (eg, physical, social, policy, and economic features, such as venue-based policies, violence, policing, and managerial practices) levels. These various structural determinants can act iteratively and dynamically with interpersonal (eg, partner or dyad factors, such as condom use, types of sexual exchanges, and sexual networks), individual behavioural (eg, drug use or duration in sex work), and biological (eg, sexually transmitted disease co-infection) factors to confer or mitigate HIV acquisition and transmission risk among FSWs.

Inclusion and exclusion criteria

We included studies that examine multivariate risk or mitigating factors for HIV infection (or HIV or sexually transmitted infections) or (male) condom use among FSWs. We excluded studies that focused solely on adolescent (<18 years), transgender, or male sex workers, or transactional sex (exchange of sex for non-monetary goods). We excluded qualitative studies, non-primary research (eg, reviews, modelling studies, and commentaries), studies in which HIV infection, HIV or sexually transmitted infection, or condom use were not analysed as individual outcomes, studies that did not report multivariate analyses of the outcome, and those that did not report stratified results among FSWs.

Epidemiological data gaps, biases, and contexts

Our systematic review of available epidemiological literature (appendix) does not account for the vast empirical and theoretical social science and grey literature on factors that mitigate or potentiate epidemics among FSWs. Structural determinants that have not been well operationalised, such as stigma, support the increasing need for dialogue between epidemiologists, social scientists, and the community to ensure the diversity of the sex industry and FSWs' lived experiences are shown. Despite efforts to review non-English language reports in Russia, China, Kenya, and India, other countries are not represented here.

Both the available evidence and our modelling of key structural factors identified elucidate the heterogeneous nature of epidemics among FSWs, including the dynamic interplay of structural, behavioural, and biomedical factors that are context specific and modified by variations in epidemic structure (eg, concentrated or generalised; Figure 1 and Figure 2). As such, scarce or unavailable data (including longitudinal, qualitative, and grey literature) on key structural determinants parameters (eg, policing exposures in Kenya and India) and their evolution over time hindered our ability to consider some key contextual elements in the modelling of epidemic trajectories, which led to a potential underestimation of their effect on HIV risk structuring among FSWs. Of particular importance is the concerning dearth of epidemiological, qualitative, and grey literature on structural determinants of HIV among FSWs in high-burden countries of sub-Saharan Africa, given that this region has a substantial portion of the epidemic among FSWs worldwide. In addition, most of the available data are drawn from cross-sectional studies. In our modelling analysis, we show many different influences of structural determinants through recurrent dynamic stages and we highlight the gaps in epidemiological data on non-linear, dynamic, and iterative HIV transmission pathways, and the need for more complex, multilevel, and mixed methods studies. For example, data for Vancouver show that to tackle only one form of violence is probably insufficient to curb HIV acquisition and transmission risk, supporting the need for combined structural HIV interventions to both support FSWs' rights and ensure access to biomedical interventions for prevention, treatment, and care. Of note, our findings might overestimate the effect of structural interventions because we provided estimates of HIV infections averted if the longer-term excess risk associated with non-condom use was wholly eliminated immediately after implementation. In reality, a reduction in structural determinants (eg, violence or police harassment) and an increased ability of FSWs to negotiate condom use, for example, might occur slowly. In some instances, community empowerment might lead to a transient period of increased violence by clients or regular partners,^{21, 22} which would reduce short-term intervention effect. Access, not just coverage, remains crucial for prevention and treatment; however, our modelling of antiretroviral therapy coverage does not take into account potential structural barriers to access to treatment, such as stigma.

Direct and indirect effects of policing

Violence by clients, police, strangers, intimate partners, and third parties

Stigma and denial of health services and antiretroviral therapy

Safer work environments (with supportive management and venue policies, and sex worker and peer support)

Search strategy and selection criteria

We searched PubMed, EMBASE, Science Citation Index, BIOSIS Previews, PsycINFO, CINAHL, Social Sciences Citation Index, Sociological Abstracts, and CAB Direct (CAB Abstracts & Global Health) for peer reviewed reports published in any language from Jan 1, 2008, to Dec 31, 2013, assessing determinants of HIV infection or incidence, or condom use among female sex workers. The search terms used were “Sex work*” OR “sex worker” OR prostitute* OR prostitut* OR “prostitution” OR “commercial sex worker*” OR “sex workers” for sex work, “risk factor” OR “risk factors” OR correlate* OR determinant* OR predictor* OR risk* for risk or protective correlates, “Condom use” OR “non-condom use” OR “condom non-use” OR “unprotected sex” OR “condom refusal” OR condom* OR “condom negotiation” OR “condoms utilization” OR “condoms utilisation” OR “condoms/utilization” OR “condoms/utilization” for condom use, HIV* OR “human immunodeficiency virus” OR “HIV infections” OR AIDS* OR “acquired immunodeficiency syndrome” OR “acquired immune deficiency syndrome” OR HIV/AIDS* for HIV. SMG and PD did initial screening and eight reviewers (SMG, PD, PM, SR-P, MR, JL, SAS, and KS) extracted data from relevant reports.