A quasi-experimental prospective cohort study was conducted to examine differences in treatment outcomes between the intervention and comparison arms. The study entailed a baseline behavioral survey and a 6-month follow-up study visit (behavioral survey and viral load testing). A 12-month follow-up visit will be conducted 12 months after the baseline visit (this data is not yet collected at the writing of this paper).
The study was conducted in Njombe (intervention) and Mbeya (comparison) regions. Njombe was purposely selected because of its high HIV burden (32.9% in Iringa, of which Njombe was part until 2012) among FSWs , high estimated number of FSWs (3,871) , and was a priority region for the donor. A region where Sauti was similarly operating its CBHTC+ mobile and home-based platforms and with a sufficiently high HIV prevalence and high number of FSWs was need to serve as the comparison site. The selected region, Mbeya, has an HIV prevalence of 29.2% , and an estimated large FSW population size (10,152) . Both regions are part of the “Southern Highlands transportation corridor”, a major trucking route, and host many mobile seasonal workers, both of which make it ripe for a large sex work community. There are, however, some important differences. Mbeya region is four times larger than Njombe (approximately 2.7 million vs. 700,000) , has a higher proportion with no education among women (16.2% vs. 8.2%) , and is more urbanized (33% vs. 24%) . Further, according to the 2013 bio-behavioral survey, FSWs in Mbeya compared to those in Iringa are slightly younger, and had higher risk-taking behaviors and vulnerabilities to HIV (e.g., higher inconsistent condom use and average number of paid partners per day, and higher proportions never HIV tested). Lastly, Mbeya region has lower ART care and treatment clinic (CTC) coverage compared to Njombe as of 2014 (2.29 vs. 3.79 CTC per 1000 infected adults) .
Intervention Arm (Njombe)
The intervention was developed in close consultation with in-country partners, including Sauti, NACP, USAID and the Njombe Regional Health Medical Team. The intervention aligned closely with the National ART guidelines on the management of HIV and the national community-based HIV and AIDS services (CBHS) guidelines (in draft at the time of intervention development). Using an implementation science approach , this intervention was developed through a participatory and iterative process involving the aforementioned stakeholders to identify gaps and develop strategies that would be be acceptable to the target population, and supported by the NACP thus helping to ensure research uptake and sustainability of the intervention if found to be effective. The intervention was built upon Sauti’s existing Community-based HIV Testing and Counseling Plus (CBHTC+) intervention, which includes the following services for FSWs: HTC, STI screening and periodic presumptive treatment, escorted referrals of HIV positive clients to HIV treatment facilities, condom promotion and provision, family planning counselling and methods, referrals for cases of gender-based violence, TB screening, and alcohol and drug screening. Providers directed presumptive TB clients and those using drugs or alcohol to referral services by using an onsite available directory of the surrounding facilities. These services were offered to all FSWs in both the intervention and the comparison arm sites. Table 1 outlines the services offered in the intervention and comparison arms, where the latter refers to the standard of care. For the intervention arm, the community-based ART delivery was added. Specifically, clinical staff were recruited to form the community-based health services team who provided the ART services through the CBHTC mobile and home-based platform. As required by the government, clients must register for ART at a CTC site and must have a CTC number. For the purpose of this study, all CBHTC “sites” were considered “satellite” facilities and linked to a government-managed CTC in order to obtain a CTC number and ensure ARV supply. The team was trained on recruitment of eligible candidates to the study, assessing client’s readiness for ART initiation, ART delivery, ART adherence counselling, and protocol for when clients should be referred to facility-based ART services (e.g., complications, opportunistic infections or pregnancy). Each team comprised one clinician and two nurses both trained in HTC and ART services, and at least three peer educators.
Enrolled participants were referred to government-designated ART service CTC facilities for current standard-of-care ART services, which entailed test and start per the national guidelines. Patients return for monthly visits for the first 6 months for evaluation and drug refills as well as adherence counseling. The main difference between the two arms is that the comparison arm did not provide any community-based ART services.
Study Population and Sample Size
Eligible participants were females aged 18 years and above who sold sex for money or goods in the past 6 months, HIV positive and not currently on ART, and planning to reside in the region of recruitment for the next 12 months (or willing to return to recruitment district every 3 months for refill and check-up). Because community-based ART was delivered to stable clients, the World Health Organization (WHO) clinical stage was assessed at recruitment, and FSWs with stages 3 or 4 (with symptoms) were excluded and referred to government-designated ART facilities. Positive FSWs not on ART included: (i) newly diagnosed, (ii) previously diagnosed but not registered in care, (iii) in care but not on ART, and (iv) previously on ART but stopped ART for at least 3 months by the date of enrolment.
We estimated a sample size for the study to detect a difference of 20% points with 80% power in the key outcomes of ART initiation, ART retention, and viral suppression between the two arms at 12 months post-intervention at the p < 0.05 level and accounting for design effect of 1.5 and loss to follow-up (LTFU) of 20%. Based on Sauti’s routine programming data, we anticipated that 90% of the study participants in the intervention and 70% in the comparison arm would be initiated, and that 90% and 70% of those who initiated would be retained in care, and 80% and 60% of initiates would achieve viral suppression, respectively. The final sample size was determined as 300 per arm for the outcome of viral suppression, and 193 per arm for the outcome of initiation and retention.
Procedure and Measurement
Study participants were actively recruited in July to October 2017 through: (i) community-based HTC in hotspots, (ii) contacting FSWs who were previously diagnosed by Sauti HTC services but not yet on treatment; (iii) brochures and announcements at targeted health facilities, peer support groups, and HTC sessions through the Sauti prevention team. Previously diagnosed persons were tested again to confirm their HIV status prior to starting ART.
The survey elicited information on demographics, HIV-related risk behaviors, HIV testing history, health status, sexual abuse, self- and external stigma, and enrollment into ART (for those who knew their HIV status for at least a month prior to the survey). Participants were asked if they could be contacted for a 6-month and 12-month follow-up visits. The follow-up visits entail the same behavioral survey with additional questions regarding ART uptake, adherence, and experience with HIV treatment services; the follow-up visits also included viral load testing. Interviews were conducted face-to-face in Swahili using handheld devices in private venues such as the participant’s or a peer educator’s home or rented room in a guesthouse. At each study visit, each participant received 10,000 Tanzanian shillings (about 4.5 USD) to offset the costs of travel and time and to reduce loss to study follow-up.
At the follow-up visits, participants were considered linked to care if they self-reported having registered in HIV care at a Care and Treatment Center (CTC) or Sauti CBHTC+ ART program. ART initiation was based on self-report of having been prescribed ARVs by a provider and started taking them. Adherence was measured in two different ways: whether they missed any ARV dose in the past 7 days and whether they stopped taking ART for more than 30 days continuously. Internalized HIV-related stigma was measured using a validated six-item scale, which assessed participants’ feelings of shame and guilt because of living with HIV (e.g., “I sometimes feel worthless because I am HIV positive”) . Each item had two responses: “agree” and “disagree”. A composite index was computed and dichotomized at the median (low vs. high).
All analysis was performed using Stata analysis software (Version 14.1, College Station, Texas). Comparison of the baseline characteristics of the two arms, the comparison of the lost to follow-up participants to those retained in the study, and the comparison of treatment-related outcomes (e.g., ART initiation, currently on ART, missing dose in past 7 days, stopped taking ART for > 30 days continuously) in the two arms was analyzed using Chi square and Fisher’s exact tests for categorical variables and T-tests for continuous variables.
The main outcome for multivariate analysis was ART initiation. Multivariate analysis was performed to examine whether participation in the community-based ART intervention was associated with ART initiation at the 6-month follow-up. Because 100% of participants in the intervention arm started ART, this caused a complete separation in multiple logistic regression. Complete separation is defined as the outcome of each subject in the data set being perfectly predicted. We used an alternative to logistic regression called firth logistic. Firth logistic regression introduces a likelihood penalty that solves the separation problem [30, 31]. The selection of independent variables was initially determined through literature, theoretical concepts, and their levels of significance during bivariate analysis. We included key socio-demographic characteristics (age, education, marital status, and mobility) as covariates. As stigma is a critical barrier to accessing care, we also explored how it affects treatment initiation. Bivariate analysis was conducted to examine factors associated with ART initiation at the 6-month follow-up. Variables significant at p ≤ 0.15 in bivariate analysis were included in the final multivariate analysis [32, 33]. Unadjusted and adjusted odds ratios and 95% confidence intervals are reported.
The study was approved by the Population Council Institutional Review Board (USA), the National Institute for Medical Research, Medical Research Coordinating Committee (Tanzania), and the Mbeya Consultant Hospital, Mbeya Medical Research and Ethics Review Committee (Tanzania).