An implementation science protocol of the Women’s Health CoOp in healthcare settings in Cape Town, South Africa: A stepped-wedge design

South Africa is the epicenter of the HIV epidemic; evidenced by the fact that 1% of the world’s population lives in South Africa, but about 17% of all persons living with HIV (5.6 million people) reside there [14]. Extremely vulnerable, impoverished women living in South African disadvantaged communities face myriad risks for HIV. For example, historically South Africa has had a patriarchal culture that fueled women’s social, legal, and economic disempowerment [15]; which, in turn, increases their risk of HIV infection. Research from over a decade in Pretoria has demonstrated that HIV prevalence remained stable among women who engage in sex work at about 68%, but increased among women who do not engage in sex work who use substances, including alcohol, from 34% in 2004–2007 to 47% in 2012–2014 [16]. Furthermore, recent research with women who engage in sex work and women who do not engage in sex work in Pretoria indicated that only 16% of women who were HIV-positive were on ART [17].

Achieving the 90–90-90 goal in South Africa will require reaching vulnerable populations, such as women who engage in sex work and/or substance use, and testing them for HIV, linking them to ART, and retaining them in care. Gender-focused, behavioral interventions that address multiple risk-taking behaviors can facilitate ART initiation and retention. However, contextual and structural factors—such as lack of transportation, food insecurity, myths and beliefs regarding HIV, and navigating the public health system—create barriers that interfere with ART initiation and adherence [18]. Additionally, many vulnerable South African women experience GBV or AOD use (or a combination of these factors), which can contribute to HIV risk [9‐12, 19]. Furthermore, multiple barriers exist regarding access to and the use of substance use rehab in South Africa, which is further complicated by the scarcity of affordable programs and a lack of knowledge regarding how to access the limited programs that are available [20].

Since 2001, our research team has developed and adapted the WHC, an evidence-based, woman-focused, behavioral HIV prevention intervention that reaches high-risk, hard-to-reach women in various contexts. The WHC addresses the intersecting risks of substance use, sex behaviors, and violence and victimization, with the primary goal of increasing skills and knowledge to reduce substance abuse and HIV risks. The original Women’s CoOp intervention was conducted in the United States and was named a Centers for Disease Control and Prevention “best-evidence” behavioral intervention for HIV prevention [21]. Since its inception, it has been adapted from previous iterations to new populations and settings in North Carolina, Rhode Island, Massachusetts, Russia, the Republic of Georgia, and several regions in South Africa [15, 22‐27]. To promote the use of EBIs in Africa, the U.S. Agency for International Development (USAID) included the first adaptation of the WHC intervention [22] in its Compendium of Gender Studies in Africa [28].

The WHC is grounded in empowerment and feminist theory to emphasize the inverse relationship between substance abuse and personal power and that reduced personal power leads to increased risk behaviors and victimization. Two intervention sessions are held approximately 1 week apart, with each session lasting about 1 h. Sessions can be conducted in groups, workshops, or one-on-one individual sessions. Key elements of the sessions educate participants about the risks of AOD use and how AOD use and sexual risk relate to HIV for women. The WHC also teaches risk-reduction methods such as proper condom use, sexual negotiation, and violence prevention strategies. Women also role-play and rehearse how to use male and female condoms correctly, as well as condom negotiation. At the end of the second session, participants create goals to reduce personal risk. Intervention staff also provide referrals to services and linkages to care.

Following numerous trials in which the WHC demonstrated efficacy, the program was packaged to facilitate widespread dissemination and implementation. The intervention package includes a brief marketing overview for decision-makers and materials for interventionists. It also includes a complete set of intervention materials, the curriculum, and an outreach manual with instructions related to reaching hard-to-reach vulnerable women. Based on the principles of translation frameworks, such as the RE-AIM model [29], the next logical step would be to evaluate the implementation of the packaged WHC in real-world clinical care settings. In this implementation science project, the WHC intervention will be implemented by clinic staff in healthcare clinics and substance use rehab centers as part of standard of care among women living with HIV who use substances. The WHC will address women’s multiple risk-taking behaviors and promote adherence to ART.

Prior research suggests the need to adapt and modify evidence-based programs to make them more flexible and amenable to scale-up in real-world settings [13]. The adaptation and modification steps need to be monitored and evaluated to ensure that they support continued fidelity to the overall model and lead to outcomes that are similar, but not necessarily identical, to efficacy trials. A framework for wider scale-up that leads to adoption of an innovation in real-world settings must consider multiple, nested levels—such as patient, provider, setting, organization, and environment—at which implementation and evaluation occur [30]. The approach posits that successful implementation begins with an evidence-based practice, in this case the WHC, to which there is initial training of clinical staff followed by preliminary and setting-specific modifications, then adoption and evaluation of multilevel outcomes. The process begins with formative work to identify any potential barriers to adoption and use of the WHC. Modification, implementation, and evaluation are iterative and interactive processes that form a feedback loop for monitoring service delivery, intervention refinements, implementation strategies, evaluation protocols, and quality of care in a staged progression prior to full implementation. As part of this project, these multiple steps will be implemented to refine and test the WHC program for wide-scale use in healthcare clinics and substance use rehab centers.

The overarching goal of this project is to assess the acceptability and feasibility of implementing the WHC in usual care settings and to assess the effectiveness of the intervention as it is implemented. The project has three specific aims. Aim 1 is to undertake qualitative data collection to develop, implement, and assess the appropriateness of a marketing plan or a recruitment strategy to facilitate entry into HTC/healthcare clinics and substance use rehab centers for the implementation of the WHC. Aim 2 is to test and evaluate a modified WHC program that has inputs from the first phase; evaluation will use a stepped-wedge design across the health clinics that provide HTC, ART, TB treatment, and antenatal care matched with substance use rehab programs to determine implementation outcomes (appropriateness, acceptability, adoption, cost, fidelity, feasibility, and sustainability) and service outcomes (comprehensiveness of services and timely service linkages). Aim 3 will assess the impact of the WHC program on patient outcomes at 6-month follow-up.

The WHC is being rolled-out sequentially in four HTC/healthcare clinics and four outpatient substance use rehab centers located in selected disadvantaged communities (townships) in Cape Town. These sites were identified and approved by the City of Cape Town Health Department. Each HTC/healthcare clinic was paired with a substance use rehab center based on geographic proximity, and each pair was randomized by computer into four succeeding 6-month implementation cycles where implementation of the WHC is taking place simultaneously at the paired sites.

The overall study uses a stepped-wedge design, as shown in Fig. 1. This design lends itself to implementation science research adaptation and mixed methods [31]. It also has several logistical and scientific advantages, as it involves incremental execution of the intervention; allows continued measurement of key process and outcome indicators; accommodates periods of feedback loops; allows for retrospective evaluation of sustainability; and accounts for temporal trends, such as changes in ART regimens (i.e., universal coverage) or funding for healthcare that may impact implementation. Our iterative approach will incorporate lessons learned at each step, making the integration process more efficient and conducive to identifying aspects of the clinic environment and the WHC that are acceptable, scalable, and sustainable.

As depicted in Fig. 1, this study has 4 cycles spanning several years. Each implementation cycle of the WHC program lasts 6 months. Based on the findings from each cycle, we undertake a brief formative “lessons learned” evaluation. These periods between cycles inform backward- and forward-implementation strategies through mixed methods of brief staff surveys and focus group discussions (FGDs), allowing for modification of the WHC while preparing for training in the next cycle. Pre-, mid-, and end-of-implementation FGDs are conducted to inform the modifications, including training materials, curriculum, and implementation data collection. Pre-implementation FGDs involve setting the stage for entry into the clinics, so that barriers and challenges can be addressed. Mid-implementation FGDs explore initial results and help determine if additional modifications to the intervention are needed. End-of-implementation FGDs with clinic staff and patients explore satisfaction with the intervention and concerns with the program’s sustainability. This format provides back-and-forth feedback and dialogue with clinic staff on the implementation processes and outcomes for further refinement. Consequently, the clinics randomized to Cycle 1 will have the longest post-intervention observation period, with implementation sustainability measured through fidelity checks and quantitative surveys. Conversely, based on lessons learned throughout the earlier cycles, sites in Cycle 4 benefit from an intervention that is more refined and in the most scalable form.

Prior to integrating the modified WHC program into the facilities, clinic staff are identified and presented with information on the WHC intervention. They then complete a survey to assess organizational readiness for change, including perceived barriers to change, acceptability, feasibility, and appropriateness of implementing the WHC in these facilities. This survey is then re-administered at the end of the implementation cycle. This process will be repeated at subsequent cycles. To assess long-term sustainability of the WHC, we will complete a brief checklist 6 months after the end of each cycle. These post-intervention sustainability checks will be repeated after each cycle such that Cycle 1 facilities have the longest post-intervention observation period on sustainability.

The patient-level study includes a baseline visit and a 6-month follow-up visit to evaluate intervention effectiveness. Eligibility criteria include (1) being between the ages of 18 and 45 years, (2) self-reporting use of at least one drug, including alcohol, at least weekly during the previous 3 months, (3) reporting unprotected sex with a male partner in the past 6 months, (4) having a positive verifiable HIV test result, (5) reporting the intention to remain in the area for at least the next 6 months, (6) providing contact information, and (7) being willing to participate in alcohol and other drug screening.

A field team hired by Kheth’Impilo, the local nongovernmental implementing partner, leads the outreach and data collection activities. Trained outreach workers operate as recruiters of women within clinics and surrounding communities and track eligible women for the study, while data capturers collect patient-level data from participants at the baseline and 6-month follow-up appointments. These are well-established methods from other studies conducted in the region [34, 35]. Once participants are enrolled, clinic staff trained and certified in the WHC offer the adapted intervention to all women enrolled and to other women in their clinics who may be interested in participating, so as to operate as a real-world setting.

At each data collection visit, a member of the research/implementing team administers a face-to-face interview using a secure tablet with skip patterns and quality-control checks programmed into the interview. The study instrument includes measures of socioeconomic status, sex risk behaviors, AOD use, substance abuse treatment readiness, ART use and barriers to use, and other self-reported clinical outcomes such as TB and symptoms for sexually transmitted infections. Biological data are obtained from a urine specimen tested for pregnancy and metabolites for recent drug use. Participants also undergo a breathalyzer test for recent alcohol use. The same measures are collected at 6-month follow-up.

HIV testing is not part of the research because it is performed by clinic staff as part of standard clinic operations. As proof of HIV-positive status, participants must provide a clinic-issued document such as an antiretroviral (ARV) card or ARV medication with identifiable data linking to the participant, or they must consent the research staff to access their hospital records for their results. Viral load is also collected at baseline and 6-month follow-up from the South African National Health Laboratory Service (NHLS) database in the clinics. Viral load testing is also part of the standard of care and the clinics routinely perform HIV viral load testing 6 months after a patient begins taking ART and then 12 months later; however, these intervals may vary by clinic. Viral load test information will be used to validate self-reported ART adherence.

In each cycle, the objective is to enroll 120 women (60 from a HTC/healthcare clinic and 60 from a substance use rehab center) for a sample total of 480, and to follow each enrolled participant at 6 months, with an expected follow-up rate of 90%. At the end of the 6-month follow-up period, we will assess effectiveness of the program using pre-post comparisons to determine whether women reduced their AOD use, are adhering to their ARV regimen, have reduced their sex risk behaviors (e.g., using a condom at every sex act; less transactional sex), and have experienced less to GBV. At the end of each cycle, we will conduct FGDs with a subsample of women who participated in both workshops of the intervention and completed their 6-month follow-up appointment. These study activities will enhance our understanding of women’s satisfaction with the WHC program and suggest modifications to the program for broader acceptability and implementation that will be incorporated into the next cycle.

To protect confidentiality, the study assigns each participant a unique study identification code (ID). This ID is the only link between the behavioral and biological data and the identifying information collected for locating participants for their follow-up interviews. Locator information is stored separately from other data in a double-locked file cabinet in a locked room with restricted access. Data collection for this study is conducted by highly trained, local implementing staff that develop a rapport with the study participants to engender trust and elicit the most accurate data possible. Data are transmitted each day from the field site to secure servers in the United States. The US-based data manager reviews additional automated quality control checks that the software generates each day. If any critical inconsistencies are noted, the data manager contacts the field project manager who works with the field staff members to resolve the inconsistencies. The principal investigator, other members of the research team, and the field staff also receive daily activity reports.

Adverse events (AEs) or severe adverse events (SAEs) related to data collection or any other research activities are reported to the principal investigator and co-investigators within 24 h, and the funding agency and the Institutional Review Board within 48 h, with appropriate action taken immediately. HTC/healthcare clinics and substance use rehab centers follow their standard procedures, including distressed respondent procedures, for any AEs related to the participants that they treat. The study does not interfere with the clinics’ standard operating procedures.

The team has made modifications to the protocol to strengthen the study design and facilitate implementation. Table 2 summarizes these changes, which were all approved by the appropriate ethical boards.

We used the Stata command, steppedwedge [36], which is based on the equation developed by Hussey and Hughes [37] to calculate the detectable difference between participants’ adherence before and after their exposure to the intervention, given a sample size that is feasible and realistic in these clinic settings (n = 480; or 120 participants per cluster). We conducted two analyses based on the following parameters: 120 participants per cluster, 240 observations per cluster, and power of .80. The first analysis was used to calculate the detectable difference between participants’ self-reported percentage of ART adherence during the past month before and after the intervention. No previous research has estimated the percentage of ART adherence among South African women who use substances since the government approved the provision of ART to all people living with HIV, regardless of CD4 count. Consequently, we based our analyses on a national estimate of the average percentage of ART adherence in South Africa since the ART rollout (87%; SD = 17.28). Based on this estimate, we will be able to detect an increase or decrease in adherence of 4% or greater. Next, we calculated the detectable difference between the proportion of adherent (i.e., report 95% or greater adherence or are virally suppressed) participants before and after their exposure to the intervention. We based this calculation on previous research that has reported on the proportion of people who engage in substance use that are adherent to ART (56%) [38] and national estimates indicating that 55% of all people living with HIV are virally suppressed [39]. In regards to self-report adherence, we estimate that we will be able to detect a difference of 13% or greater. Similarly, we also estimate that we will be able to detect a difference of 13% or greater to examine the change in the proportion of women who are virally suppressed before and after the intervention.

The analysis will treat count outcomes as Poisson distributed. A number of multilevel factors can impact the outcomes, including site/clinics, city, time, and the type of the site (HTC vs. substance use rehab). Because individuals are clustered within sites and sites are clustered within a city, we need to consider the variance inflation factor that results in this clustering. We considered a number of approaches: using random effects, fixed effects, or Generalized Estimating Equations (GEE) considering population-averaged methods. Although a random effects model for sites seems a natural approach, we needed to consider fixed effects of the city and of the site type, which would leave little power left to estimate the random effects distribution. Consequently, we will also use two other methods, GEE and the fixed effects model, to assess sensitivity by comparing the results across these models.

We will use Stata Release 14 (College Station, TX) and SAS version 9.2 (Cary, NC). Data will be examined to determine the extent to which they meet appropriate distributional assumptions. We will use methods described by Ecob and Der to identify outliers in longitudinal models and use methods of case downweighting when extreme cases are identified [40]. Missing data will be handled using multiple imputation methods [41]. The patient-level analyses will be organized into three groups to examine (a) successful reduction of risk behaviors targeted by the WHC (i.e., effectiveness), (b) patient satisfaction, and (c) increasing the number of women receiving ART and adhering to ART, and being admitted to substance use rehab.

To assess effectiveness, the analysis will compare changes from baseline to 6-month follow-up on the amount of alcohol consumption, biological and self-reported drug use, condom use at last sex, number of sex partners over the past 6 months, and values from scales that measure relationship power and communication. We will start with descriptive analyses, including measures of central tendency, and dispersion (e.g., means, medians, proportions, standard deviations). Bivariate analysis will include paired t-tests, Wilcoxon rank-sum, and tabular statistics and the chi-square test. For multivariable analysis, the xtlogit and xtgee function will be used to specify random effects or fixed effects, and will assume that the correlation matrix is exchangeable. The models will control for other covariates identified in bivariate analyses or hypothesized a priori (e.g., condom use is correlated with substance use). One of the most important covariates in the model will be measures of patient satisfaction. Model performance and fit will be assessed with incremental elimination of variables, likelihood ratio tests, and changes in the Akaike information criterion (AIC) and Bayesian information criterion (BIC) values.

Patient satisfaction analyses will be predominately descriptive to determine if satisfaction improves with each cycle, varies by type of clinic, and by participant characteristics (e.g., alcohol use, pregnancy status).

We assume that the prevalence of patients receiving or adhering to ART or receiving treatment for substance abuse are Poisson distributed and increase from baseline to follow-up. We will use a similar approach outlined in the effectiveness analysis for bivariate and multivariable analyses. Variables considered for multivariable analysis will include patient characteristics, measures of patient success with the WHC, patient satisfaction, and clinic-level characteristics (e.g., type of clinic, provider satisfaction).

The study is now in the field, with further modifications necessary. The first modification involved identifying women who use substances and are living with HIV, because many women are unwilling to admit that they are living with HIV or that they are using substances. Additionally, they may be unwilling to be tested because they have been treated poorly and are afraid to go to health services. Consequently, to increase access to healthcare services for study participants, the research team had to provide training to clinic staff and outreach workers to educate them about stigma surrounding HIV and substance use. Further, the electronic database systems to verify HIV status for those who had already tested were out of date (e.g., some names and dates of birth were not accurate based on the information given by the participants). Consequently, the research team established alternative protocols to verify HIV status.