Mental health problems among female sex workers in low- and middle-income countries: A systematic review and meta-analysis

Background The psychological health of female sex workers (FSWs) has emerged as a major public health concern in many low- and middle-income countries (LMICs). Key risk factors include poverty, low education, violence, alcohol and drug use, human immunodeficiency virus (HIV), and stigma and discrimination. This systematic review and meta-analysis aimed to quantify the prevalence of mental health problems among FSWs in LMICs, and to examine associations with common risk factors. Method and findings The review protocol was registered with PROSPERO, number CRD42016049179. We searched 6 electronic databases for peer-reviewed, quantitative studies from inception to 26 April 2020. Study quality was assessed with the Centre for Evidence-Based Management (CEBM) Critical Appraisal Tool. Pooled prevalence estimates were calculated for depression, anxiety, post-traumatic stress disorder (PTSD), and suicidal behaviour. Meta-analyses examined associations between these disorders and violence, alcohol/drug use, condom use, and HIV/sexually transmitted infection (STI). A total of 1,046 studies were identified, and 68 papers reporting on 56 unique studies were eligible for inclusion. These were geographically diverse (26 countries), representing all LMIC regions, and included 24,940 participants. All studies were cross-sectional and used a range of measurement tools; none reported a mental health intervention. Of the 56 studies, 14 scored as strong quality, 34 scored as moderate, and 8 scored as weak. The average age of participants was 28.9 years (age range: 11–64 years), with just under half (46%) having up to primary education or less. The pooled prevalence rates for mental disorders among FSWs in LMICs were as follows: depression 41.8% (95% CI 35.8%–48.0%), anxiety 21.0% (95% CI: 4.8%–58.4%), PTSD 19.7% (95% CI 3.2%–64.6%), psychological distress 40.8% (95% CI 20.7%–64.4%), recent suicide ideation 22.8% (95% CI 13.2%–36.5%), and recent suicide attempt 6.3% (95% CI 3.4%–11.4%). Meta-analyses found significant associations between violence experience and depression, violence experience and recent suicidal behaviour, alcohol use and recent suicidal behaviour, illicit drug use and depression, depression and inconsistent condom use with clients, and depression and HIV infection. Key study limitations include a paucity of longitudinal studies (necessary to assess causality), non-random sampling of participants by many studies, and the use of different measurement tools and cut-off scores to measure mental health problems and other common risk factors. Conclusions In this study, we found that mental health problems are highly prevalent among FSWs in LMICs and are strongly associated with common risk factors. Study findings support the concept of overlapping vulnerabilities and highlight the urgent need for interventions designed to improve the mental health and well-being of FSWs.


Introduction
Mental health problems are a significant cause of the global burden of disease. In 2010, mental, neurological, and substance use disorders were the leading cause of years lived with disability globally [1]. Worldwide, an estimated 300 million people are affected by depression, and 272 million people by anxiety, with women at higher risk compared with men [2,3]. The treatment gap for common conditions exceeds more than 90% in low-income countries [4]. Left untreated, mental disorders prevent people from reaching their full potential, impair human capital, and are associated with premature mortality from suicide and other illnesses [5]. Suicide is a health outcome strongly associated with mental, neurological, and substance use disorders. Nearly 800,000 people are estimated to die due to suicide each year, with 79% of global suicides occurring in low-and middle-income countries (LMICs) [6]. A range of social determinants affect the risk and outcome of mental disorders. These include demographic factors (such as age, gender, and ethnicity), socioeconomic factors (such as low income, unemployment, and low education), neighbourhood factors (such as inadequate housing and neighbourhood violence), and social change associated with changes in income and urbanisation [1].
Sex work-defined by The Joint United Nations Programme on HIV/AIDS (UNAIDS) as the receipt of money or goods in exchange for sexual services, either regularly or occasionally [7]-is criminalised in most regions of the world [8]. In addition to the social determinants described earlier, women who sell sex face a unique set of structural factors including police harassment and arrests, discrimination, marginalization, poverty, and gender inequality [8,9], as well as extreme occupational hazards such as violence, coercion, deception, alcohol and substance use, and human immunodeficiency virus (HIV)/sexually transmitted infection (STI) [10]. Together, these predispose female sex workers (FSWs) to increased psychological health vulnerabilities. Structural and occupational risks associated with sex work are highly dependent on sociocultural and economic contexts, which means that these hazards may differ for sex workers in LMICs and those in high-income countries. Evidence from high-income countries indicates a high prevalence of mental health morbidity among FSWs, especially post-traumatic stress disorder (PTSD), depression, anxiety, and psychological distress [11][12][13][14]. Three previous reviews have examined mental health in the context of STIs/HIV, alcohol use, and violence against sex workers [15][16][17]. However, no attempt has been made to date to synthesise the evidence or estimate the burden of mental health disorders for FSWs. This is vital to inform policy and programming at the global and country level. The aim of this systematic review is to estimate the prevalence of mental disorders among FSWs in LMICs, and to examine associations with factors that commonly affect their health and well-being (violence, alcohol and drug use, condom use, HIV/STI).

Search strategy and selection criteria
The review protocol has been registered with PROSPERO, number CRD42016049179 (https:// www.crd.york.ac.uk/prospero/). Ethics approval was not required for this study. This study follows the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) guidelines (Fig 1; S1 Prisma Checklist). We searched electronic peer-reviewed literature databases (Ovid, PubMed, Web of Science) from first record until 26 April 2020. Search terms included the following: "mental health" OR "mental well-being" OR "psycholog � health" OR "psycholog � distress" OR "mental illness � " OR "mental disorder � " OR "mental health problem � " OR "psychiatr � morbidit � " OR "anxiety" OR "depress � " OR "suicid � " OR "trauma" OR "post-traumatic stress disorder" OR "PTSD"; "sex work � " OR "female sex work � " OR "prostitut � " OR "female prostitut � " OR "sex trad � " OR "transact � sex" OR "FSW � " OR "commercial sex" OR "sex-trade worker � "; "low and middle income countr � " OR "LAMIC � " OR "LMIC � " OR "developing countr � " OR "names of all countries which fit the LMIC criteria." See S1 Text for full database list and search strategy.
Articles were included that measured the prevalence or incidence of mental health problems among FSWs, even if sex workers were not the main focus of the study. Only studies from countries defined as low or middle income, in accordance with the World Bank income groups 2019 [18], were included. Eligible studies had to be peer-reviewed, include females aged 16 or older who were actively engaged in sex work, and include the following study designs: cross-sectional survey, case-control study, cohort study, case series analysis, or experimental study with baseline measures for mental health. Studies were limited to English. We excluded studies that used qualitative methods only, were review papers, were conference abstracts, or were non-peer-reviewed publications. Studies exclusively focused on alcohol and substance use or victims of human trafficking sold into sex work were excluded from this review as reviews have recently been published in these areas [17,[19][20][21]. Studies focused on women engaged in transactional sex only were ineligible for review, as this practice-and its implications on health-is distinct from sex work [22].

Data extraction and quality assessment
All publications were screened by 2 independent reviewers (TB and BS). If either author classed a publication as relevant, the abstract was reviewed, with any disagreements discussed and a consensus reached. If eligibility could not be determined by screening of the title and abstract, the authors reviewed the full text. Three authors (BS, TB, and AM) assessed full texts using the eligibility criteria cited earlier, with any discrepancies resolved through discussion. Data were extracted by 3 authors (TB, BS, and AM) into a structured data extraction sheet.
Study quality was assessed by authors (TB, BS, and AM) using the Centre for Evidence-Based Management (CEBM) Critical Appraisal for Cross-Sectional Surveys Tool (S2 Text). One item on CEBM was removed (Item 12: "Can the results be applied to your organisation?") as it was not applicable to this review. To assess quality of studies, authors rated each article on 11 items, and an overall score was calculated, with higher scores indicating stronger quality. Studies scoring �7 out of 11 points were considered strong quality, between 5 and 7 were rated moderate quality, and �4 were scored as weak quality. Individual scores are presented in Table 1, and detailed scoring of each study is presented in S3 Text.

Data analysis
A narrative synthesis was conducted across all studies meeting inclusion criteria. Prevalence estimates were calculated from percentages or raw proportions, and we contacted authors of studies in which raw data were missing. If multiple publications reported results from a single  � Papers report findings on same study but explore different associations with outcome of interest. �� Papers report findings on same study but explore different associations with outcome of interest.
��� Papers report findings on same study but explore different associations with outcome of interest. † Papers report findings on same study but explore different associations with outcome of interest. § Papers report findings on same study but explore different associations with outcome of interest. ‡ Papers report findings on same study but explore different associations with outcome of interest. ≌ Papers report findings on same study but explore different associations with outcome of interest. study, we included all studies in Table 1 but only the original study in the narrative synthesis and prevalence analyses. Meta-analyses were conducted on studies that scored moderate to strong in the quality assessment and that used validated measures to assess mental health outcomes; we excluded studies from the meta-analyses that sampled participants based on characteristics that are known to be an independent risk factor for mental health problems (such as injecting drug use or HIV status) and could therefore bias the pooled mental health estimates. Analyses were completed using Comprehensive Meta-Analysis (CMA) software version 3 (Biostat, Englewood, NJ). Pooled estimates were calculated using a random effects model. Variation between studies was determined by heterogeneity tests with the Higgins' I 2 statistic. Relative weights were calculated using the formula 1/V + T 2 where V is the error variance and T 2 (Tau-squared) is the between-study variance. Subgroup analyses were completed to examine associations between mental health outcomes (e.g., depression) and the following covariates: violence/police arrest, alcohol/drug use, condom use, and HIV/STI. Due to variations between studies in the factors adjusted for in multivariate analyses, unadjusted odds ratios (ORs) were extracted or calculated from raw data. Pooled effect estimates were calculated using a randomeffects model.

Study characteristics
The initial electronic search yielded 1,035 results, with 11 more studies identified through reference list screening and online searches. After duplicate records were removed, the titles and abstracts of 630 publications were screened for eligibility. Of those, 208 were identified as potentially relevant publications and reviewed for inclusion. Sixty-eight papers reporting on 56 unique studies with 24,940 participants meeting the inclusion criteria (Fig 1). Eight of these studies did not provide prevalence data on mental health [23][24][25][26][27][28][29][30][31]; authors of these studies were contacted twice for further information, and 2 authors responded, providing prevalence data [23,27]. In total, 86 prevalence estimates from 48 studies were available (depression n = 37; anxiety n = 7; PTSD n = 8; suicide attempt n = 8; suicide ideation n = 17; psychological distress n = 7; mood disorders n = 2) ( Table 1). Studies were based in 26 LMICs: 13 countries in sub-Saharan Africa, 1 in the Middle East and north Africa region, 1 in Eastern Europe, 2 in South East Asia, 5 in the Western Pacific region, and 4 in Latin America and the Caribbean. Eleven studies reported findings from countries in the low-income group, 20 studies from the low-middle income group, and 26 studies from the upper-middle income group, as per the World Bank income classification. Twenty-nine studies used a purposive sample, 14 used respondent driven sampling techniques, 7 used a random sample, and 6 utilized random sampling techniques to select venues and purposive methodology to recruit FSWs within these venues (Table 1). Most studies recruited FSWs from a variety of venues, such as streets, bars, brothels, and entertainment establishments, with 3 studies selecting women from clinics or hospital settings [ [29,[42][43][44][45] HIV status (n = 5), and were excluded from the meta-analyses (regardless of CEBM score) to avoid biasing the pooled estimates. Analyses used a variety of validated scales and cut-off points to assess mental disorders (Table 1). None reported a mental health intervention.
The mean age of FSWs in the 42 studies that reported this was 28.9 years (age range: 11-64 years). Thirty-two studies reported sex work locations for their sample; among these studies, 66.3% of FSWs worked in brothels, lodges, bars, or other entertainment establishments; 51.7% worked in streets or public places; 24.7% worked at home; and 36.7% worked in other settings, e.g., via mobile phones (these categories were not mutually exclusive). Thirty-one studies reported education levels of their sample, and among these, nearly one-half of FSWs (45.7%) had an education level of primary school or less. Among the 40 studies that reported marital status for their sample, 48% of FSWs were never married; 32.9% were currently married or in a relationship; and 24.5% were divorced, separated, or widowed.

Associations between mental health and other factors
We conducted subgroup analyses to examine associations between mental health (e.g., depression) and factors commonly experienced by FSWs (violence/police arrest, alcohol/drug use, condom use, and HIV/STI) ( Table 2). Findings of the meta-analyses are summarised in Table 3 and displayed in forest plots in S1-S4 Figs.
Violence. Seventeen studies reported on associations between mental health problems and violence experience [30,34,36,37,39,43,44,46,49,52,53,57,59,75,79,80,82], usually by an intimate partner or a client ( , and 2 studies reported recent suicidal ideation and violence experience (ever or recent), with only one of these eligible for inclusion in a meta-analysis (unadjusted OR 2.5 [1.3-4.7], p = 0.004) [79]. Three studies reported recent suicide attempt      and violence experience (ever or recent), with 2 of these eligible for inclusion in a meta-analysis (S1 Fig). The pooled unadjusted OR of recent suicide attempt and violence experience (ever or recent) is 3.5 [2.2-5.5], p < 0.001. Three studies reported on police violence (harassment, arrest, or raids) and mental health problems ( Table 2). While no association was found between police harassment (ever) and current depression (unadjusted OR 0.9 [0.4-2.3], p = 0.9) [49], police arrest (ever) was associated with current depression in one study by Patel and colleagues (unadjusted OR 2.2 [1.7-2.8], p < 0.001) [53], and police raid in the past year was associated with a suicide attempt in the past 3 months (unadjusted OR 2.3 [1.2-4.3], p = 0.01) in a study by Shahmanesh and colleagues [80].
Alcohol and drug use. Associations between mental health problems and alcohol use were reported by 6 studies, but there was marked variation in how alcohol use was measured, with 2 studies asking about alcohol use in the past 30 days [53] or alcohol intoxication in the past 6 months [56] and 4 studies using Alcohol Use Disorders Identification Test (AUDIT) to  Table 2). The pooled unadjusted OR for depression and alcohol use is 1.6 (0. Three studies reported on mental health problems and illicit drug use, again with considerable variation in the way illicit drug use was measured (any illicit drug use ever [85,86] versus polydrug use past month [44]). Two studies were included in the meta-analysis (S3 Fig). The pooled unadjusted OR for depression and illicit drug use is 2.1 (1.4-3.1), p < 0.001.
Condom use. Nine studies reported on mental health problems and condom use with clients and regular partners [24,53,56,60,62,68,71,80,84]. Condom use measurement varied with studies either reporting frequency of condom use (always versus not always) or condom use at last sex (yes/no). The pooled unadjusted OR for depression and inconsistent condom use with clients is 1.6 (1.  [36,44,[47][48][49]60,69,73]. One study [48] was excluded from the meta-analyses because it did not use a validated tool to measure depression, and one study was excluded because it only sampled HIVnegative women [44]. The pooled unadjusted OR for depression and HIV is 1.

Discussion
In this systematic review and meta-analysis using data from 56 studies and 24,940 participants, we found that mental health problems are highly prevalent among FSWs in LMICs and are strongly associated with social and behavioural factors commonly experienced by FSWs. Of note, all studies were cross-sectional, and not a single intervention study designed to address mental disorders among FSWs was identified. The prevalence of mental disorders among FSWs in LMICs was much higher compared with the general population in LMICs. For example, data from 41 LMICs from the 2002-2004 World Health Survey found the prevalence of depression to range between 3.9% and 7.8%, with higher rates among women (7.0%-7.8%) compared with men (3.9%-4.9%) [87]. Additionally, the 12-month prevalence of suicidal behaviour among people in LMICs has been reported to be 2% for suicidal ideation and 0.4% for suicide attempts, with rates higher among women compared with men (ideation: 2.4% women versus 1.6% men; attempt: 0.5% women versus 0.4% men) [88]. FSWs face increased levels of key risk factors for mental disorders and suicidal behaviour, including financial stress, low education, inadequate housing, violence, alcohol and drug use, STIs including HIV, and stigma and discrimination [15,17,53,67], which may help explain the higher prevalence of mental health problems in comparison with the general population. Indeed, findings from our meta-analyses support this hypothesis. Understanding how these social determinants interact with mental disorders and which are modifiable within programmatic timeframes will be crucial to designing holistic interventions for FSWs. This review adhered to PRISMA guidelines and used a comprehensive search strategy, independent screening and quality appraisal of studies. This study had some limitations. By limiting the search to published studies only, and to literature written in English, we may have missed key studies. We used unadjusted ORs to examine associations between mental health problems and key risk factors to allow like-for-like comparisons between studies; not adjusting for potential confounders may have biased the findings although unadjusted and adjusted ORs were usually similar in individual studies. Where individual studies provided multiple estimates on co-linear outcomes (e.g., depression and violence; depression and police arrest), using unadjusted ORs to calculate the individual associations may have led to participants who had not experienced one outcome (e.g., police arrest) but who had experienced the other (e.g., violence) being included in the reference group and subsequent underestimation of the true association. The removal of studies that sampled participants based on characteristics that are known to be an independent risk factor for mental health problems (such as HIV status, harmful alcohol use) led to fewer studies being included and wider confidence intervals around prevalence estimates and pooled ORs. However, when we re-ran the analyses to include all qualifying studies, regardless of sampling criteria, we did indeed find that estimates were slightly higher, suggesting that inclusion of these studies would have led to an overestimation of the pooled estimates and associations. Several methodological issues across the studies were also observed. All studies were cross-sectional. Longitudinal studies are needed to ascertain direction of causality between mental health problems and other factors common to FSWs, although studies with the general population suggest that these relationships are likely to be bidirectional [89]. Most studies used nonprobability sampling across a wide variety of settings which may introduce selection bias and mean that the most vulnerable women will be missed from these surveys. This in turn may lead to underestimations of mental health estimates. A range of measurement tools was used to capture mental health outcomes, as well as violence, alcohol and drug use, and condom use. Even when studies used the same mental health outcome measures, different cut-off scores were applied. This limits the comparability and reliability of findings across studies and points to a need for establishing more rigorous guidelines on using validated tools with this study population.
To our knowledge, this systematic review is the first globally to estimate the prevalence of mental health problems among FSWs in LMICs and to examine associations between poor mental health and other risk factors common in sex workers' lives. Our findings and metaanalyses suggest that FSWs experience a high burden of depression, anxiety, PTSD, psychological distress, and suicidal behaviours and that poor mental health is strongly associated with violence experience, drug use, inconsistent condom use, and HIV/STI. Together, this supports the concept of overlapping vulnerabilities and has several important implications.
First, there are no existing studies that we are aware of that describe mental health interventions; low-cost, effective interventions for FSWs with mental health disorders are urgently needed. Among the general population attending primary care services in India and elsewhere, brief psychological interventions delivered by trained lay-counsellors have been shown to effectively treat depression [90,91]. Strategies to prevent suicide could include promoting mental health, limiting access to the means for suicide, reducing harmful alcohol use and violence experience, and training "gatekeepers" to support women at increased risk, such as those who have previously attempted suicide [6]. Such interventions should also be suitable for FSWs and could be adapted and embedded within existing HIV service provision. Second, the strong associations between mental health disorders and key occupational risk factors such as violence and harmful alcohol and drug use support the need for upstream structural interventions as part of holistic HIV prevention programming for FSWs. Again, violence interventions have been shown to be effective in reducing violence among women in LMICs [92,93] as well as among FSWs [94]. Low-cost, brief psychological interventions to treat harmful alcohol use could also be adapted to FSW settings [95]. Third, strong associations between poor mental health and reduced condom use with clients and with HIV infection suggest that treatment of mental health problems may also improve condom use with clients and the sexual and reproductive health of FSWs. In addition, women diagnosed with HIV may require on-going counselling and support, for example, by HIV testing and screening counsellors or FSW peer advocates, which goes beyond CD4 counts and treatment adherence, to also enquire about a woman's ongoing psychological well-being.