Contextual background

The study was conducted in three cities in Africa and one in India where sex work is common, criminalised and FSWs are marginalised. Durban, South Africa, has a population of 3.4 million and an estimated FSW population of 6300 [15]. The Tete-Moatize area in Mozambique has approximately 250,000 inhabitants and an enumeration exercise counted 4415 FSWs in 2008 [16]. Mombasa, Kenya, with a population of about 1.2 million has approximately 9000 FSWs [17]. Mysore, in Karnataka State India, has approximately 900,000 inhabitants and about 2000 FSWs. Female sex worker-targeted services are provided by non-governmental organisations in each site. Services in Durban are delivered primarily through outreach, while in the three other cities service provision occurs within stand-alone clinics. In Tete-Moatize this consists of a small clinic on the outskirts of Moatize [16], in Mombasa there are three drop-in clinics in different divisions of the city [18] and in Mysore there is a clinic operated by the FSWs organisation Ashodaya Samithi [19].

The Diagonal Interventions to Fast-Forward Enhanced Reproductive Health (DIFFER) study is a 5-year implementation research project that aims at improving access to HIV/SRH services for FSWs by a better linkage between interventions targeted at FSWs and the general health services [20]. It applies a methodological framework for health systems research, starting with a detailed situation and policy analysis, followed by the development of context-specific packages of interventions to improve access to SRH services in each city. These packages will combine a strengthening of services specifically targeted at FSWs with improving access to the general health services, and developing linkage systems between both. The performance of the packages is evaluation after at least 18 months of implementation [21]. The baseline situational analysis used a mixed-methods approach combining quantitative surveys among FSWs and SRH service clients, with qualitative focus group discussions and key informant interviews. To assess the effect of the intervention, quantifiable indicators of SRH commodities and services utilization are measured at baseline and end-line through a cross-sectional survey among a representative sample of FSWs in each city. This paper reports on the baseline cross-sectional survey, presenting a comparison of the indicators of use of contraception and services for cervical cancer screening, unwanted pregnancies and sexual violence, by FSWs across the four cities. The findings on the use of HIV prevention and care services and of where FSWs seek care have been published elsewhere [22, 23]. The findings on the policy analysis and the other - qualitative - components of the situational analysis are also available [24–26].

Cross-sectional surveys

We recruited FSWs using Respondent-Driven-Sampling (RDS), which is similar to snowballing, but corrects for the bias towards FSWs with large social networks through statistical adjustments [27]. Respondent-driven sampling begins with the selection of “seeds” i.e. known members of the population of interest. Each seed is asked to identify a number of other FSWs from their social network for the survey, who subsequently identify other FSWs, and so on, until the desired sample size is reached.

In Durban, Tete, Mombasa and Mysore, 11, 13, 16 and 8 seeds were recruited, respectively. Seeds were identified by experienced peer educators who work within long-standing peer outreach programmes in each city. To ensure branching into different FSW networks, seeds were categorised according to locally relevant sub-populations. In Durban, they were categorised according to age, location of soliciting (indoors/outdoors) and migration status; in Tete, according to nationality (Mozambican/Zimbabwean), place of residence (Tete city/Moatize city) and type of FSW (full-time/occasional); and in Mombasa, according to location of soliciting sex (bar/club based, street/truck based, brothel/home based, and beach based). In Durban, Tete and Mombasa each participant recruited, using coupons, up to 3 new participants and in Mysore up to 5. Handling of coupons was monitored with Electronic RDS Coupon Manager Version 3.0 in Durban, Tete and Mombasa, and manually through a coupon log notebook in Mysore.

To detect substantial changes in project indicators between the baseline survey and the end-of-project survey with a significance level of 0.05 and a power of 0.80, the minimum estimated sample size was 400 per study site. In Tete, the refusal rate was high, in particular among FSWs of Mozambican nationality, allegedly because false rumours had spread about the intentions of the survey, and recruitment was stopped at 308 FSWs because of time constraints.

Potential respondents were screened for eligibility and invited to answer a standardised, interviewer-administered questionnaire. The questionnaire collected information on socio-demographic and sex work characteristics, main contraceptive method currently used, ever having used emergency contraception, unwanted pregnancies in the last 5 years and action taken for an unwanted pregnancy, ever having tested for cervical cancer, having experienced forced sex in the past 12 months and if medical care was sought subsequently. Respondents were considered eligible to participate in the study if they were females 18 years or older and a practicing FSW (having received money or gifts for sex at least three times in the last 6 months). Eligible and consenting respondents were interviewed using either a paper-based questionnaire (Durban, Mombasa and Mysore) or Computer-Assisted Personal Interview (QDS™) software (Tete). Confidentiality was safeguarded through the use of non‐identifying survey codes and keeping all collected information locked or protected.

In Durban, Mombasa and Mysore, the questionnaires were entered in an MS-Access database and in Tete uploaded in a QDS data warehouse. The survey data were merged with the coupon data, and imported into STATA version 12. In the first stage of the analysis, we calculated population point estimates with their 95% confidence intervals (CI) by non-parametric bootstrapping, adjusted for social network size and homophily within networks, using the STATA RDS analysis package and the Volz-Heckathorn estimator [28]. Then, for the analysis comparing outcomes between cities, we performed post-hoc pairwise comparison tests after fitting a logistic regression model with RDS-adjusted weights, using jack-knife resampling and Dunn–Šidák correction for multiple comparisons [29]. The analysis focused on the comparison of the proportion using contraceptives, action taken for unwanted pregnancies, cervical cancer screening and SGBV services across the four cities. Sociodemographic sex worker characteristics that were associated with both the outcome and the city were stepwise introduced in the regression model and retained in the model if they changed the odds ratio (OR) by at least 10%. To assess overall use of SRH commodities and services we established a composite index. In the denominator we added ‘1’ for each service the FSW was in need of: a non-barrier contraceptive method if she did not want to become pregnant; ever having been screened for cervical cancer if she was older than 30 years; and seeking medical care if she had been forced to sex. In the numerator we added ‘1’ for each of these services that the FSW had used. The index thus represents the proportion of needed services that a FSW used.

The applicable ethical review boards in each country (the University of Witwatersrand’s Human Research Ethics Committee in South Africa, the National Committee of Bioethics for Health in Mozambique, the KNH/UoN Ethics and Research Committee in Kenya, and the Asha Kirana Institutional Ethics Committee in India) and the ethical board of the coordinating agency in Belgium (Commission for Medical Ethics of the University Hospital Ghent) approved the study protocols.

Socio-demographic characteristics (Table 1)

FSWs were on average older in Mysore than in the African cities. Educational levels were higher in the African cities than in Mysore. More than half of FSWs in Durban and Mombasa reported never have being married or cohabiting with a partner, while in Mysore this proportion was very small (3.5%). The proportion currently living with a partner, be it married or not, was particularly small in Mombasa (1.2%). In Tete most FSWs had previously been in a steady relationship, but were currently single (62%).

The number of commercial sex acts reported in the past month was substantially lower in Mysore than in the African cities. The proportion of FSWs that reported a regular non-commercial partner was much higher in Mysore (97%) and substantially more FSWs reported other non-commercial partners in Mysore and Tete (60% and 49%, respectively) than in Durban and Mombasa (20% and 24%, respectively). Fewer FSWs had another source of income in Durban (11%) than in the other cities. The most common sources of income were activities such as hairdressing, selling at the market, or waitressing, across cities.

Occurrence of SRH risks and use of SRH commodities and services (Table 2 and 3)

The majority of FSWs with a need for contraception (not wanting to become pregnant, not pregnant and not infertile) were using a contraceptive method in all cities, though the proportion was significantly higher in Mombasa than in Durban and Tete (p < 0.05). In all cities, but particularly in Durban and to a lesser extent Mombasa, a large proportion used condoms alone as contraceptive method. The proportion of FSWs who used a non-barrier contraceptive method was substantially lower in Durban (33%) than in Mombasa (66%), Tete (70%) and Mysore (85%) (p < 0.01). Of the non-barrier methods, injectable hormonal contraceptives were the most often used method in Durban and female sterilisation in Mysore. In Mombasa the implant was often used. Oral hormonal contraceptives were only commonly used in Tete and the IUD was rarely used across cities. Emergency contraception was less commonly used in Mysore (2.4% ever used it) than in the African cities (ranging from 13% in Tete to 38% in Mombasa) (p < 0.01).

In Durban and Mombasa about a third of FSWs reported to have experienced a pregnancy in the last 5 years, when they didn’t want to get pregnant at that time. This proportion was substantially lower in Tete and Mysore (7.6% and 8.0%, respectively, p < 0.01). In Durban, Tete and Mombasa, most FSWs kept the pregnancy, despite being unwanted, while in Mysore only a small proportion (6.3%) did so.

Cervical cancer screening was more commonly accessed in FSWs in Durban (29% reported to have ever been screened) than in the other cities (p < 0.01). If we only consider FSWs of at least 30 years old, the proportion of FSWs ever screened was higher, but still low in Mombasa and Mysore. In Tete, none of the interviewed FSWs had ever been screened for cervical cancer.

The proportion of FSWs who reported that in the past 12 months they were forced to have sex was highest in Durban (36%) and the lowest in Mysore (7.1%). Condom use during forced sex was reported by about half of the victims in Durban, but fewer in Tete and Mombasa. In all cities, a large proportion of the victims did not seek medical care following sexual violence.

Using the composite index, a similar proportion of FSWs (ranging between 25.8% and 32.7%) accessed all the services they needed in Durban, Tete and Mysore, but in Mombasa the proportion was significantly higher (40.9%) (p = <0.001, 0.058 and 0.006 compared to Durban, Tete and Mysore, respectively).