Summary of key findings
Overall, this review found evidence on the impact of selected community-based intervention packages on FSW continuation in HIV care across the HIV care cascade. The significant impact of the interventions was observed on three cascade stages namely; HIV testing, HIV diagnosis and ART use. However, for HIV testing and ART use, the improvement was short-lived in that the retention on ART and improved access to HIV testing was not sustained for the entire period of implementation. There were limited impactful interventions for HIV diagnosis with only one community service delivery model showing significance. Generally, impactful interventions were those that implemented targeted and comprehensive package of HIV services provided at one location within places where FSWs worked and lived, and with unique service delivery models for specific cascade stages. This review also found that community-based interventions led to the improvement of linkage to care and viral load suppression to undetectable levels, however, the improvement was not significant. In addition, the results showed that many of the projects were small-scale, research-based and have limited time-bound implementation periods. A few non-research based and large-scale HIV prevention efforts had limited systematic means of monitoring outcomes along the HIV care cascade and therefore, less methodical attention to constantly reviewing the effectiveness of interventions and altering delivery strategies accordingly.
Characteristics of included studies
Our literature search yielded 582 articles and after removing duplicates we remained with 565 studies. Following screening of the titles and abstracts, we retained 45 studies for full text review. After thoroughly reading the remaining 45 articles, 27 studies were excluded for the following reasons; six articles had reported outcomes that did not meet our eligibility criteria, six did not meet the study design criteria, eight studies had not reported community based interventions, two articles were not written in English, two had either not targeted FSWs or not reported disaggregated data for FSWs, two articles were abstracts from conferences and we failed to access the required information, and one had not been conducted in Sub-Saharan Africa.
The remaining 18 studies were found eligible for inclusion in the review (Table 1). The study designs were randomized controlled trials (n = 3), cross sectional studies (n = 10) and cohort studies (n = 4) and quasi experimental study (n = 1). These studies were conducted in seven sub-Saharan African countries (Table 1). Majority of the included studies were from Zimbabwe (n = 5) (24–28); Tanzania (n = 3) (29–31); Kenya (n = 2) (32, 33) and 5 studies, of which each was from Guinea (n = 1) (34); Zambia, (n = 1) (35); Burkina Faso (n = 1) (36); Uganda (n = 1) (37) and South Africa (n = 1) (38).
Table 1
Characteristics of selected studies
(#Ref) Author (Year)
|
Country
|
Study aim
|
Study design
|
sample size, and participant selection
|
(34) Aho et al. (2012
|
Guinea
|
To describe the acceptability and consequences
of VCT among a stigmatized and vulnerable group, female sex workers (FSWs), in Conakry, Guinea
|
Cross sectional study
|
Randomly selected 421 at FSW at baseline and 223 at end line. Recruited through
private or public centres providing adapted healthcare
services (AHS) for FSWs
|
(35) Chanda et al. (2017)
|
Zambia
|
To evaluating the effect of 2 different health system mechanisms (the active
approach of peer-based HIV self-test) for HIV self-test
delivery compared to referral to standard HIV testing
|
A 3-arm 1:1:1 cluster
randomized trial
|
Total randomized per arm (320,316 &329). peer educators recruitment of social network via direct contact, and referral
|
(26) Cowan et al. (2019)
|
Zimbabwe
|
To (i) present the current impact that engagement in the Sisters programme has on HIV incidence, prevalence and control in FSW. (ii) To describe the patterns and characteristics of sex work among FSW in Zimbabwe (iii) to assess the potential for wider population impact of sex worker programmes by modelling the impact on HIV incidence of eliminating transmission through FSW
|
Cross sectional study
|
Use of program data of 5083 FSW recruited through respondent driven sampling surveys through three studies conducted at 19 sites: in 2011 to 2015; 2013 to 2016 and 2017
|
(24) Cowan et al. (2017)
|
Zimbabwe
|
To describe the HIV diagnosis and care cascade among FSW in Zimbabwe.
|
Cross sectional study
|
Respondent driven sampling surveys of FSW in 14 sites. Recruited 2722 women, approximately 200 per site as the baseline for a cluster-randomized controlled trial investigating a combination HIV prevention and care package.
|
(25) Cowan et al. (2018)
|
Zimbabwe
|
To assess the efficacy of a targeted combination intervention for female sex workers in Zimbabwe.
|
Cluster-randomized trial from 2014 to 2016
|
Randomly assigned 14 clusters (1:1) to receive usual care cluster (n = 3612) and
an intervention cluster (n = 4619)
|
(36) Huet et al. (2011)
|
Burkina Faso
|
To describe the long-term virological, immunological and mortality outcomes of providing highly active antiretroviral therapy (HAART) with strong adherence support to HIV-infected female sex workers (FSWs) in Burkina Faso and contrast outcomes with those obtained in a cohort of regular HIV infected women.
|
A prospective observational study nested
within the Yerelon open cohort of high-risk women
|
47 FSWs and 48 non-FSWs recruited through a network of peer-educators and followed up at a dedicated clinic located within a public health facility.
|
(33) Kelvin et al. (2019)
|
Kenya
|
To assess whether informing female sex workers about the availability of HIV self-testing at clinics in Kenya using text
messages would increase HIV testing rates
|
Cohort study
|
A sample of 2196 female sex workers selected from electronic records.
|
(29) Kerrigan et al. (2019)
|
Tanzania
|
To determine the impact of a community empowerment
model of combination HIV prevention (Project Shikamana) among female sex workers (FSW) in Iringa, Tanzania.
|
A prospective
community-randomized trial conducted in 2 communities
matched on population size
|
Identified all active sex work venues (164 in total)
in the 2 study communities and enrolled 496 FSW through a time-location sampling
|
(40) (Lafort et al. (2018)
|
South Africa, Mozambique and Kenya
|
To enhance uptake of SRH services by FSWs through an implementation study
|
Cross sectional study (in the context of an implementation research project)
|
400 FSWs recruited by respondent-driven sampling
|
(39) Lillie et al. (2019)
|
Burundi, Cote d’Ivoire and DRC
|
To identify new HIV-positive cases among KPs so that newly diagnosed individuals could be linked to life-saving treatment for epidemic control.
|
Quasi experimental study
|
929 FSWs sampled. Selection was done through distribution of coupons by peer
|
(60) Lafort et al. (2016)
|
South Africa, Mozambique and Kenya
|
To identify gaps in the use of HIV prevention and care services and commodities for female sex workers with the aim of improving SRH services.
|
Cross-sectional survey (in the context of an implementation research project)
|
Used RDS to recruit 400 sex worker in Durban, 308 in Tete,
400 in Mombasa and 458 in Mysor
|
(32) Luchters et al. (2008)
|
Kenya
|
To evaluate the impact of five years of peer-mediated STI/HIV prevention interventions among
FSW in Mombasa, Kenya
|
Pre- and post- intervention
cross-sectional surveys
|
Initial respondents (seeds) were identified from FSW work places, with subsequent participants recruited using snowball sampling.
|
(27) Napierala et al. (2018)
|
Zimbabwe
|
1) To compare engagement in services and the HIV care cascade among FSWs aged 18–24 years compared with those aged 25 years and older. 2) To explore factors associated with young FSWs’ engagement in HIV services.
|
Cross-sectional survey
|
Sampled 2722 FSW through respondent-driven sampling from 14 communities
|
(28) Ndori-mharadze et al. (2018)
|
Zimbabwe
|
To compare key indicators related to FSW health seeking
behavior in 2011 and 2015 in three sites and explore
whether observed differences might be linked to the delivery
of intensified community mobilization.
|
Cross-sectional study
|
870 FSW sampled in 2011 and 915 in 2015. FSWs were selected
as seeds of the 2015 RDS Survey, and also reviewed programme data from the Sisters’ clinics between 2010 and 2015.
|
(37) Pande et al. (2019)
|
Uganda
|
To assess preference and uptake of the current community-based HIV testing services delivery models that are used to reach FSW and identify challenges faced during the implementation of the models.
|
Cross-sectional study design
|
Used cluster sampling
for hot spot selection and recruited 72 FSWs in each cluster
|
(38) Schwartz et al. (2017)
|
South Africa
|
To assess engagement
in the HIV care cascade and correlates of ART use among a
sample of South African FSWs.
|
Cross-sectional study
|
Selection was done through RDS by selecting seeds to represent FSWs across ages, race and locations
|
(31) Tun et al. (2019)
|
Tanzania
|
To examine differences in treatment outcomes between the intervention and comparison arms.
|
Quasi-experimental prospective cohort study
|
309 (intervention) and 308 (comparison) sampled at baseline. FSW selected randomly through; community-based HTC in hotspots, directly contacting former Sauti FSWs and use of brochures
|
(30) Vu et al. (2020)
|
Tanzania
|
To increase linkage to and retention in antiretroviral therapy (ART) care, by piloting a community based,
ART service delivery intervention for female sex workers
|
Quasi-experimental prospective cohort study
|
309 (intervention) and 308 (comparison) followed from baseline. FSW selected randomly through; community-based HTC in hotspots, directly contacting former Sauti FSWs and use of brochures
|
There were two multi-country studies conducted in Burundi, Cote d’Ivoire and DRC (n = 1) (39) and South Africa, Mozambique and Kenya (n = 1) (40). All studies in the review were conducted among FSWs, with two studies that included non FSWs and MSM. The first study compared uptake of HIV services among FSW and non-FSW (36) and the second one also included MSM (39) but reported disaggregated results for FSW.
The common implementation models of community interventions included: i) promotion of FSW participation with client-led approach, ii) augmented and intense community mobilization approaches through direct service delivery to places where FSW live and work, and iii) application of technological innovation (i.e. text messaging) as a tool for mobilization of increased uptake of HIV services. Other interventions also focused on provision of comprehensive package of services using a one-stop shop approach in the community, peer to peer implementation approaches and provision of static services in clinics based in hotspots of FSWs.
The vast majority of the studies focused on traditional FSW service delivery approaches that included peer focused approach of mobilization, condom promotion, HIV testing and promotion of regular STI screening (Table 2). Only six studies measured at least two or more HIV cascade outcomes downstream of HIV testing to viral suppression among FSWs at the end of the implementation period of the community-based interventions (26, 28, 29, 34, 35, 40).
Table 2
A summary table showing impact of community HIV intervention on continuation in HIV care across the treatment cascade, extracted from the included studies
Cascade Step
|
Combined Interventions that showed significant impact
|
Evidence
|
HIV testing
|
• Partnership with KP NGOs/CBOs based in the community/hotspots to deliver HTS services on behalf of national programmes
• Peer educator direct distribution of HIV self-test kits
• Repeated use of text messaging and communication on what’s up by peers informing FSW about the availability of testing services in the community
• Adapted health care: Creation of FSW safe spaces and integration of targeted FSW HIV services in the general health care (e.g STI screening and treatment, lubricants and condoms, direct escort by FSW peers within a public facility)
• Provision of testing through night Clinics (bars, brothels, DICs)
• Full time provision of testing at clinics based in hotspots
• Strengthening support networks FSW CSOs to encourage health-promoting behavior
• Venue-based peer education, free condom distribution, and HIV counseling and testing;
|
Chanda et al. (35)
Kelvin et al. (33)
Aho et al. (34)
Lafort et al. (60)
Pande et al. (37)
|
HIV Diagnosis
|
Enhanced peer outreach approach:
• Use of paid out reach peers that have not worked as peers before to find new FSWs from their network
• Use of short-term incentivized peer support to reach their hard to reach contacts-FSWs
|
Lillie et al. (39)
|
Linkage to care
|
None
|
|
ART Use:
|
• NGO initiated FSW-targeted mobile clinical services
• Provision of services at a community-led drop-in center
• Training of health workers in FSW-friendly approaches
• Provision of HIV services in the community clinic by a professional health provider
• Extending operating days at community based clinics with flex working hours
• Provision of broad package of HIV service offered in clinics based in hotspots
• Provision of on call services where FSW can consult anytime
• Police sensitivity trainings, violence prevention and campaigns for anti-stigma and discrimination
|
Kerrigan et al. (29)
Cowan et al. (25) Napierala et al. (27) Pande et al. (37)
|
Viral suppression
|
None
|
|
Interventions that showed a positive effect but with non-significant impact
|
Linkage to
|
• Enhancing referral mechanisms to the neighboring public health facilities by paying stipend for peers.
• Financial facilitation of FSW focal person’s based at public health facilities
• Establishment and incentivized peer referrals to the DICs
• Creation of a safe space at a public health facility in a community without a FSW DIC
• Conducting sensitivity trainings to all service providers including the non-professional staff within the clinics
• Peer referrals and linkages at the clinics based in hotspots
• Behavior change communication to educate and improve health seeking behaviors
• Extended hours of work to evenings, night and weekends
• Mobile HIV services to mitigate transport issues
|
Chanda et al. (35)
Kerrigan et al. (29) Pande et al. (37)
Lafort et al. (40)
|
Viral suppression:
|
• Usual HIV services augmented with additional community mobilization activities aimed at raising awareness of the benefits of ART.
• Building leadership skills among FSW groups
• Participation of FSW groups in selecting their fellow FSW adherence supporters
• Adherence training sessions for the FSW adherence supporters
• Mobile telephone messaging reminders for ART adherence
• SMS and follow-up phone to support clinic attendance.
• Empowering FSW to improve retention in care by targeting improved individual client oriented practices
|
Cowan et al. (25)
Kerrigan et al. (29)
Napierala et al. (27)
|
Table 3
A summary of reported outcomes across the care and treatment cascade for included studies
(#Ref) Author (Year)
|
HIV testing
|
HIV diagnosis
|
Linkage to care
|
ART Use
|
Viral Suppression
|
(34) Aho et al. (2012)
|
√
|
√
|
|
|
|
(35) Chanda et al. (2017)
|
|
√
|
√
|
√
|
|
(26) Cowan et al. (2019)
|
√
|
√
|
|
√
|
√
|
(25) Cowan et al. (2018)
|
|
√
|
|
√
|
√
|
(36) Huet et al. (2011)
|
|
√
|
|
√
|
√
|
(33) Kelvin et al. (2019)
|
√
|
|
|
|
√
|
(29) Kerrigan et al. (2019)
|
√
|
|
|
|
|
(40) Lafort et al. (2018)
|
|
|
√
|
√
|
√
|
(40) Lafort et al. (2018)
|
√
|
|
√
|
|
|
(39) Lillie et al. (2019)
|
√
|
√
|
|
|
|
(60) Lafort et al. (2016)
|
|
√
|
|
√
|
|
(32) Luchters et al. (2008)
|
|
√
|
|
|
|
(27) Napierala et al. (2018)
|
√
|
√
|
|
√
|
√
|
(61) Ndori-mharadze et al. (2018)
|
|
√
|
|
√
|
|
(37)Pande et al. (2019)
|
√
|
√
|
√
|
√
|
|
(38) Schwartz et al.
|
|
√
|
√
|
√
|
|
(31) Tun et al. (2019)
|
|
|
√
|
√
|
|
(30)Vu et al. (2020)
|
|
√
|
√
|
|
|
Risk of bias in included studies
The risk of bias in the included RCTs: Selection bias of allocation sequence generation was low in two studies and unclear in the remaining two. Allocation of concealment was unclear in three studies and low in one. In two RCTs, blinding of participants and personnel and blinding of outcome assessment was low and unclear in the remaining two studies. The risk of reporting incomplete data due to attrition and selective outcome reporting was low in three studies, high in one study and unclear in one study. The risk of bias in all RCTs was highest due to other forms of bias such that all had high risk due to reliance on self-report for outcomes, and short periods of intervention implementation with uncertain population effect. Other risk of bias were also related to low sample sizes including testing the intervention on limited number of FSW and FSW communities.
The risk of bias assessment in cohort studies: One cohort study had low risk of bias in relation to selection of representative samples and justification for case and control selection. However, for two cohort studies the risk of bias was unclear as information on justification for selection of control and cases was not provided. Never the less, both studies demonstrated that participants were not exposed to the intervention before the start of the study. All cohort studies had adequate measures of outcome assessment by using validated measurement scales, measurement of ART use by pill count rather than self-report and also had relatively long follow up periods of 6 to 12 months.
The risk of bias for cross-sectional studies: All the nine included cross-sectional studies had low risk of bias in regard to selection of sample size and its representativeness. However, the overall score between studies varied but the scoring grades weighed within the acceptable range of 4 to 6. In regard to risk of bias of assessing whether confounding factors were controlled, this information was not indicated in all the studies apart from one cross-sectional study which reported that a post-hoc pairwise comparison tests was conducted with RDS-adjusted weights while adjusting for the confounding effect. Lastly, one study did not describe outcome assessment, therefore its risk of bias was unclear. However, in the rest of eight cross- sectional studies, risk of bias for outcome assessment was low as there was adequate description of the validated measures used to control for risk of bias. These included review of medical records, and health assessment by qualified staff among others. All studies used statistical tests to control for bias in individual studies.
HIV testing services
HIV testing was done in two randomized control studies conducted in Kenya (33) and Zambia (35). The participants in the intervention arms for both studies were more likely to test for HIV than the standard of care arms. At one month’s follow up, Chanda (35) registered 94.9% of HIV testing, however, the testing rates dropped to 84.1% at four months of follow up in the intervention arm. Similarly, in an RCT (33), participants in the intervention arm were significantly (OR 1.9, p = 0.001) more likely to test for HIV, (81, 10.8%) compared to those in the enhanced standard of care (46, 6.1%) and usual standard of care (43, 6.2%) (OR 1.0 (p = 0.972).
Three cross sectional studies also reported HIV testing as an outcome (34, 37, 40). In a study conducted in Guinea (34) where adapted HIV care services were provided to FSW, there was 100% acceptance to HIV testing although there were reports of coercion by managers of FSW worksites affecting voluntary consent for an HIV test. Equally, Lafort (40) conducted context specific targeted community intervention in three cities; Durban in South Africa, Tete in Mozambique; and Mombasa in Kenya. In all the three cities, among all services provided, the greatest effect was on uptake of HIV testing, increasing from 40.9–83.2% in Durban, 56.0–76.6% in Tete, and 70.9–87.6% in Mombasa. Finally, a cross-sectional study (37) conducted in Uganda, compared three models of community interventions and assessed model preference determined by increased access and utilization of HIV testing services. This study showed that static clinics based in FSW hotspots were preferred (72% (279/390) compared to 25% (98/390) that used outreaches; and 3.3% (13/390) that used peer to peer mechanisms to have an HIV test. These models were implemented and assessed over a period of 12 months.
HIV diagnosis
HIV diagnosis was measured in six cross-sectional studies (27, 28, 32, 34, 37, 38), two randomized controlled trials (25, 35) and one quasi-experimental study (39). The quasi-experimental study (Fig. 2) was conducted in three different countries (Burundi, Cote d'Ivoire and DRC) and measured the proportions of HIV seropositivity during the implementation of an enhanced peer outreach approach (EPOA) for the three countries. Pooled analysis of data from the three countries showed statistically significant increase in proportion of participants who tested positive (OR 2.23; 95% CI 1.23–4.05; p < 0.001). However, pooled analysis of data from two RCTs (Fig. 3) that randomized participants to standard of care testing versus general peer support augmented with additional community mobilization, showed that there was no significant improvement in HIV diagnosis among FSWs (OR 0.99; 95% CI 0.79–1.24; p = 0.307). Similarly, data from the pooled analysis of three cross-sectional studies (Fig. 4) showed that there was a reduction in HIV diagnosis tending towards the negative impact, although this was not statistically significant (OR = 0.96; 95% CI 0.84–1.11; p = 0.554). The community-based interventions comprised of peer mediated service delivery to improve access of HIV services, intensified community mobilization and integrated adapted health services to suit specific needs of FSW.
Linkage to care
Two RCTs (29, 35) and two cross sectional studies (37, 38) reported linkage to care as an outcome. The data from pooled analysis of RCTs showed that there was improved linkage to care showing a tendency towards a positive impact of the intervention for the FSW who were provided with services in community-led drop-in centers, through venue-based peer education, introduction of social support and text messages to promote solidarity and engagement. Nonetheless, this was not statistically significant (OR 2.03; 95% CI 0.87-4.77; p = 0.085) (Figure 5). The data of pooled analysis from the cross-sectional studies showed that there was a 65% improvement in linkage to care (Overall estimate = 0.65; 95% CI 0.60 -0.69). This was observed at the end of the follow up period of implementing health worker-led HIV service delivery in clinics based in hotspots and outreaches in community settings as well as peer to peer mechanism (Figure 6).
Use of anti-retroviral therapy
ART use was measured in three RCTs (25, 29, 35) and three cross sectional studies (27, 37, 38). One of the RCTs had a two-time point measurement at one month and four months of ART initiation (35) with one year of follow up period, while the two RCTs (25, 29) had a two-year follow up period with two-time points measurement at 0 months and 18-24 months. The data from pooled analysis of the three RCTs showed that ART use was high with statistically significant results (OR 1.72; 95% CI 1.31- 2.25; p = 0.390) at the start of implementing the intervention (Figure 7), compared to the ART use reported data at the end of the follow up period (OR 2.21; 95% CI 1.38,-3.53), with no observed heterogeneity as evidenced by non-significant p values of heterogeneity (p = 0.663) (Figure 8). The data from pooled analysis of three cross sectional studies showed that there was an overall estimate of 0.60 use (95% CI 0.41-0.78; p < 0.001) which translates to 60% improvement with statistically significant rates of improved ART use (Overall estimate = 0.60 95% CI 0.41-0.78 p = 0.001) (Figure 9) and with evidence of significant heterogeneity between studies ((I^2 = 97.45%); we can therefore interpret these results with caution. The community intervention implemented for cross-sectional studies encompassed provision of targeted HIV services for FSW with augmented additional community mobilization activities and strengthened support networks.
Viral suppression
Viral suppression was reported in three studies; two RCTs (25, 29) and one cross sectional study (41). All the three studies reported data during implementation, while one RCT also reported effect at the end of follow up period (29). The RCT by Cowan (25) reported viral suppression rates for the participants that were assigned to clusters of usual care and those of the intervention clusters. At the end of the assessment period, 72% (588/828) in the intervention cluster, showed minimal difference in reduction of viral load to undetectable level less than 1000 copies per mL, compared with 68% (590/869) of the participants in the usual care clusters. Similarly, the RCT by Kerrigan (29), the viral suppression rates showed slight improvement in the intervention group 40.0% (n=36) to 50.6% (n=46) compared to the control group 35.9% (n=28) to 47.4% (n=36) at the end of 18 months follow up period. For the cross-sectional study (27), viral suppression was reported separately for young women aged (18-24 years) and older women aged (≥25). Among the older FSW, 79% showed viral load suppression to undetectable level compared to the 62% of younger FSW, after both groups received targeted HIV services (41).