Prevalence of suicidality, depression, post-traumatic stress disorder and anxiety among female sex workers: a systematic review and meta-analysis

Purpose: To assess prevalence of suicidality, depression, post-traumatic stress disorder (PTSD) and anxiety among female sex workers (FSW). Methods: A systematic review and meta-analysis was performed. Search strategy was performed in MEDLINE, Scopus, Web of Science, EMBASE, Ovid and Cochrane Central Database from inception until March 2020. Considered for inclusion were cross-sectional studies performed on FSW that assessed prevalence of any of the following: suicide attempt or suicidal ideation, depression, PTSD, or anxiety. Five reviewers independently and in duplicate to select all eligible articles in an abstract and full-text screening phase and to moreover extract information from each study. A binomial-normal generalized linear mixed model was employed to estimate prevalence of the conditions. Results: From 8,035 studies yielded in the search strategy, 55 were included for analysis. The overall prevalence of suicidal ideation and attempt was 27% (95% C.I. 18-39%) and 20% (95% C.I. 13-28%), respectively. Furthermore, overall prevalence of depression and PTSD was 44% (95% C.I. 35-54%) and 29% (95% C.I. 18-44%), respectively. Eleven studies were classied as high quality. Conclusion: Findings indicate there is an overall high prevalence of suicidality, depression, and PTSD among FSW. Development of accessible large-scale interventions that assess mental health among this population remains critical.

reviewers was passed on to the full-text phase where disagreements were resolved by consensus or intervention of another reviewer. Any non-enligsh published article was translated by study authors.

Data collection process
All ve reviewers independently and in duplicate extracted data from included articles to a previously standardized web-based form. Extracted data included only summary estimates. The following variables were extracted: rst author, year of publication, country, and sex work legal status where study was performed, number of participants, age, sex work setting and percentage of HIV positive participants, above mean drug use or exposure to physical/sexual violence compared to general population (United Nations, 2020;WHO, 2020). For scales evaluating depression, anxiety or PTSD, data on whether it was selfreported or rater-administered was obtained as well as cut-off points to determine diagnosis when it applied. For scales that evaluated symptom severity rather than diagnosis, moderate and/or high levels of symptoms were considered as a positive screening for the evaluated outcome. Suicidality was evaluated as direct questions regarding suicidal ideation or attempt during any speci c time span. Regarding outcomes of interest, absolute number and proportions were directly extracted or calculated with a simple cross multiplication when either of those was missing. Discrepancies at this stage were resolved by consensus or through intervention of another reviewer.

Quality Assessment
Quality of individual studies was rated by ve reviewers (JMMA, PCLC, FCN, HLG, AFGM) independently and in duplicate using the Appraisal Tool for Cross Sectional Studies (AXIS), a 20-item critical appraisal instrument developed to assess standard of quality of reporting by cross-sectional studies (Downes, Brennan, Williams, & Dean, 2016). AXIS evaluates quality design (clarity of objectives, design, and rationale for sample size), reporting (consistency between methodology, reported results and conclusions, incorporated measures to account for non-responders) and declaration of funding sources, con icts of interest and statement of informed consent attainment. Discrepancies were resolved by consensus or arbitration of another reviewer. Since no evidence regarding score classi cation has been described, a score of 17-20 was classi ed as high quality, 14-17 as moderate quality and ≤13 as low quality.

Data Synthesis
Although we consider that our analysis is meant to estimate the pooled proportion rather than the pooled prevalence, we will use the latter term throughout the manuscript for practical purposes. Numeric data was reported as central tendency and dispersion measures, categorical data as frequencies and percentage.
Data integration is comprised of a qualitative synthesis and a quantitative synthesis (meta-analysis) where pooled prevalence for each outcome is reported.
To estimate prevalence, we used a binomial-normal model for meta-analysis of prevalence via a Generalized Linear Mixed Model (GLMM) (Stijnen, Hamza, & Ozdemir, 2010). Prevalence estimates were reported as binomial proportions with 95% con dence intervals. Heterogeneity was assessed with the I2 statistic and Kendall rank correlation coe cient. Levels between 25-50% were considered as low, 50-75% as moderate and >75% as substantial heterogeneity.
Statistical signi cance was classi ed as a p value <.05. Meta-analysis was performed in R statistical software version 4.0.2.

Subgroup and sensitivity analysis
Subgroup analyses were performed by stratifying studies according to sex-work setting (studies performed on streets only, studies performed on combined settings) and sex-work legal status on country where study was performed (legal, partially legal, illegal) (Studies IoD; ProCon, 2020). Moreover, sensitivity analyses were performed by stratifying studies according to their quality (high, moderate, low), sample size (higher than 220, lower or equal than 220) and sampling method (adequate sampling method, inadequate sampling method). We selected a cut-off point of 220 participants for sample size sub-analysis since it was estimated as the median number of samples. Sampling was considered as appropriate if study reported enough information to generate a conclusion, recruited participants in numerous sites and used any of the following sampling methods: quota, respondent-driven, snowball, peer-referral or probabilistic.

Results
A total of 8,035 references were screened by title and abstract from which 155 were passed to full-text screening phase. Ultimately, a total of 55 studies, which were performed across 28 countries and 5 continents met the inclusion criteria ( Figure 1) (See Table 1 for studies that met inclusion criteria).
Mean/median age of participants ranged between 20.2 and 38.6 years. A total of 8 (14.5%) studies recruited participants only from the streets, 2 (3.6%) only from brothels and the rest of them (81.8%) recruited them on either a mixed setting or did not provide enough information. Thirteen studies (23.6%) provided information on proportion of HIV positive participants where it ranged between 2% and 100%. Above mean drug use and exposure to physical or sexual violence when compared to general population parameters was observed in most studies. Full details regarding study characteristics and outcome measures are available on Table 1.
Eleven studies (20%) were classi ed as possessing high quality. Moreover, 33 (60%) and 11 (20%) studies possessed moderate and low quality, respectively. Some consistently de cient domains across studies were sample size justi cation, categorization and description of non-responders (See Table 1 for the overall quality; see Online Resource 2- Table 1 for full assessment).
Twelve studies (21.8%) evaluated outcomes encompassing anxiety related diagnoses. Due to the observation of signi cant heterogeneity between anxiety diagnoses (generalized anxiety disorder, obsessive compulsive disorder, phobias, etc.), pooled analysis of this outcome was not performed, and results are reported qualitatively on Table 2.
Strati cation of studies according to sex work site and legal status of sex work on the country where study was performed revealed a higher prevalence of overall PTSD among street-only sex workers when compared to non-street sex workers or samples with combined settings (

Discussion
FSW are an extremely vulnerable group susceptible to intense psychological stress derived from multiple health, social, economic, and legal barriers where the exposure to different types of violence during their lifetime also plays a key role in the psychopathology of mental health disorders (Romans et al., 2001;Rössler et al., 2010;H. L. Surratt & Kurtz, 2012). It comes as no surprise that prevalence for psychiatric symptoms marginally surpasses that of the general population. In addition to the previously mentioned, when comparing them to other similarly vulnerable populations, especially considering trauma-exposed populations, proportions of depression and PTSD are still higher among FSW (Morina, Stam, Pollet, & Priebe, 2018).
To the best of our knowledge, this is the largest meta-analysis to examine the prevalence of depressive symptoms in FSW, and the only one regarding anxiety, PTSD, and suicidal behavior. Furthermore, our analysis encompasses results from studies performed on more than 25 countries worldwide, which provides a global perspective on this important topic. Our results highlight the signi cantly elevated prevalence of psychiatric morbidity in FSW. Depressive symptoms were the most amply studied with a pooled prevalence of 44% from 40 studies and 17,581 subjects. This is staggeringly higher compared to the general population (12.9%) (Lim et al., 2018). A previous review obtained an estimate of 62%, although this result was based on only 4 studies (Yuen et al., 2016). Regardless, the occurrence of depressive symptoms is still alarmingly elevated, irrespectively of self-reported or rater-administrated instrument use. Similarly, the prevalence of other outcomes was higher in FSW when compared to general population: PTSD (29.3% vs 11.1%), suicidal ideation (26.9% vs 9.2%) and attempts (19.7% vs 2.7) (Nock et al., 2008;Spottswood, Davydow, & Huang, 2017). Although a quantitative analysis was not possible for anxiety due to diversity of anxiety-related diagnoses, results from the included studies ranged from 8% to 61.4% (Devóglio, Corrente, Borgato, & Godoy, 2017;Iaisuklang & Ali, 2017).
As part of our subgroup analyses, we observed higher proportions of PTSD among street-based sex workers when compared to other setting (bars, venues, hotels. etc). This phenomenon could be explained due to lack of access to health services, work-related stigma, illegal status, and more importantly exposure to higher rates of workplace violence among FSW working in these settings; however, several other factors could also be associated into this observation. Another interesting nding in our subgroup analysis was also related to PTSD, where we observed higher rates of it among studies that were performed in countries where sex work is partially legal compared to completely legal (61.9% vs 12.9%, p < •001). FSW who work in countries where sex work is not legal may be more prone to being exposed to different types of violence, social and cultural stigma, as well as lack of access to health services (Romans et al., 2001;Rössler et al., 2010;H. L. Surratt & Kurtz, 2012).
The proportion of violence in every study included in this review was above the general proportion of violence against women established by the WHO (World Health Organization, 2020). Violence endured by FSW is constant and prolonged as it occurs in an everyday setting (home, work, public spaces, etc). Perpetrators are usually family members, friends, strangers, intimate partners, clients, peers and even health care workers. Although not generalizable for the entire FSW population, this is a major distinction from other traumatic environments which could elevate rates of PTSD among them (Evens et al., 2019).
Among the included studies, there was a highly consistent presence of perpetuating interpersonal trauma among FSW, including sexual abuse and physical assault. Most studies reported a history of childhood physical, emotional, psychological, and sexual violence, where there also was a signi cant correlation between these variables and a current diagnosis of PTSD. The proportion of affected individuals was reported to be as high as 82% in some cohorts. Hence, in addition to current work and domestic violence, history of complex childhood trauma as frequently observed among the studies in this systematic review may be another heavy contributing factor to the pooled prevalence for PTSD and other stress-related symptoms in our meta-analysis, in which interpersonal trauma plays a more critical role compared to non-interpersonal and context trauma. The previous hypothesis being supported from ndings in several studies involving participants with a history of exposure to childhood violence, neglect, and sexual abuse (Roxburgh. Degenhardt & Copeland, 2006;Coetzee et al. 2018;Dorahy et al. 2009).
The predicament of FSW in these different countries is not the same; most are victims of gender-based violence, but possibilities of protection vary considerably between countries. The same holds for stigma and family support.
In the interpretation of our results, it is important to consider that mental health issues for FSW are present across cultures. Sex work is a profession that is practiced worldwide and as evidenced in our review, the included references come from more than 25 countries and 5 continents. The manifestation of mental health in general among this population could vary depending on their country of origin as well as their country of residence and work for the following considerations: legal status regarding the profession, cultural perceptions of sex work and mental illness or psychological symptoms, access to physical and mental health services as well as to legal services, country economic status, among others. Adverse circumstances regarding some of the previously mentioned could negatively affect the working environment surrounding sex work and directly elevate rates of stigmatization, gender based and workplace violence among them. Also, noting that cultural factors could in uence and affect the reporting of mental health problems should also be considered in the design of much needed implementation strategies to promote mental health treatment among this vulnerable population. These strategies should be individualized based on each country and culture due to some of the previously mentioned aspects.
Another signi cant issue is the high HIV prevalence among FSW (Paz-Bailey, Noble, Salo, & Tregear, 2016). Evidence suggests a relation between poor mental health and HIV-risk behaviors such as unprotected vaginal or oral sex, IV drug use, sex under the in uence of alcohol or drugs, and syringe exchange, etc (Grosso et al., 2019;Lau et al., 2010;Rogers, Lemstra, & Moraros, 2015;Sagtani et al., 2013;Shen & Zou, 2016;Ulibarri et al., 2013). Depression, for example, may increase HIV-risk behaviors due to several factors such as impairment in cognitive function, decreased impulse control, reduced motivation, and increased fatalism. Moreover, following Beck's theory of depression, negative thoughts and feelings surround how depressed people perceive themselves and could ultimately in uence their behavior and prevent them from making rational decisions, which will often be based on emotions (Khan et al., 2009;Seth et al., 2011). Regarding PTSD, two main mechanisms have been described to be involved in the process of impulsive and risky behaviors. The rst one is an impairment in emotion regulation, a multi-component construct composed by lack of acceptance, di culties in engaging in goal-directed behaviors when distressed, di culty in impulse control, limited access to effective emotion regulation strategies, lack of emotional awareness and clarity. The second one is impulsivity, as it is a well-established correlate of risky behaviors such as the addressed in our review (Strom et al., 2012;Tull et al., 2014).
Understanding the complex relation between the abovementioned is substantial for promoting health policies and effective interventions for HIV-prevention and mental health treatment on this population. Behavioral, social, and community-empowerment interventions have shown promising results on reducing rates of risk behaviors and even the incidence of HIV and other sexually transmitted diseases. These interventions should be considered in the process of developing future health policies for vulnerable populations such as FSW (Chow et al., 2015;Kerrigan, Fonner, Stromdahl, & Kennedy, 2013).
Strengths from this review include a systematic approach to identify and synthesize all available evidence from the most frequent and debilitating mental disorders worldwide, with over 8,000 screened references (Gu et al., 2010;Hilary L Surratt et al., 2005;Wang et al., 2007). Moreover, we only included studies whose outcomes were evaluated with a validated scale which provides more validity to our results. Although we presented the general pooled prevalence for each outcome, subgroup analyses were also provided for self-reported and rater-administered instruments, sex work legal status based by country, work setting, sample size and methods, study quality and time presentation for suicidal behavior. This strati cation allows for a more comprehensive analysis on the implications for each outcome and for nding sources of heterogeneity. We also decided to implement a Generalized Linear Mixed Method for metaanalysis due to evidence indicating misleading results from the back-transformation to the original prevalence when using the inverse of the Freeman-Tukey double arcsine transformation (Schwarzer & Chemaitelly, 2019).
The present review is subject to some limitations. The most important one is that when evaluating study quality, we identi ed that studies consistently did not justify sample size or report non-response rate and the characteristics of non-responders. Moreover, nearly a third of the included studies employed inadequate sampling methods and study quality could be a source of bias in our results. As FSW comprise a heterogeneous and hard-to-reach population, this potentially leads to selection bias and non-representativeness, consequently affecting the interpretation of the results. Additionally, we acknowledge that the results may not fully re ect the real prevalence and rather represent pooled proportions from observational studies. Another important consideration is that the real prevalence may be overestimated since most measures were obtained through screening instruments. Also, although we carefully evaluated each article for this issue, we acknowledge that participant repetition could be present as some studies might have been performed using the same population cohorts; however, by summing up participants only by outcome instead of overall, the probability of this scenario being present was diminished. We consider that these matters do not hinder the observed results as they highlight the alarmingly high occurrence of psychiatric morbidity in FSW, most probably related to violence exposure, comorbidities, and the multidimensional barriers they face.
An important topic to address is the consistently high heterogeneity throughout the analysis. Although our subgroup and sensitivity analysis did not provide su cient explanations, the unexplained heterogeneity could be the result of three main factors. Firstly, FSW have a great diversity of social, working, economic, political, and more importantly, cultural circumstances. Moreover, stigma associated with sex work itself also varies across settings and countries, as has been previously reported (Sawicki et al., 2019). As a result, some FSW could be at greater risk of mental health problems compared to others. Secondly, clinimetrics employed to measure outcomes varied considerably across studies. Although evidence suggests strong positive correlations between many of them, we still consider this a source of heterogeneity (Smarr & Keefer, 2011;Titov et al., 2011). Finally, the observational nature of the studies is another potential source. However, it is highly expected to observe high heterogeneity in prevalence meta-analyses, and this topic should be addressed in future research on meta-analytical methods.
As a conclusion, high prevalence of depression, PTSD, suicidal ideation, and attempt was observed among FSW. Although results could be overestimated due to usage of screening scales, prevalence is still higher when compared to other literature. Understanding the critical impact of poor mental health on this population, its triggers as well as its relationship with HIV-risk behaviors is crucial in the design of new strategies that promote access to physical and mental health services for them. There is an urgent need for FSW to be empowered and demand adequate health access as well as to enforce their human rights and be provided with not only the former but also with decent living conditions and important factors such as workplace protection and access to legal help to prevent the onset of the studied conditions. In the process diminishing risk factors for HIV-risk behaviors while promoting human and decent life and work conditions for this population. Government agencies and institutions should prioritize this as a global health problem.

Declarations Funding
The present study did not receive any grants or funding from an independent organization or sector, neither public, commercial, nor non-pro t.

Con icts of Interest
The authors do not have any con icts of interest to state.

Ethics Approval
Not applicable Consent to participate Not applicable Consent for publication All authors have reviewed the nal version of this manuscript and authorize its submission & publication.

Availability of Data and Material
Data used in the present study will be made available upon request to the corresponding author.
Code Availability Not applicable.  Pooled prevalence of suicide ideation, suicide attempt, depression and post-traumatic stress disorder (PTSD) among included studies