Study design
The study design was a three-arm cluster-randomized controlled trial conducted at six HIV clinics operated by The AIDS Support Organization (TASO). A cluster design was used to limit risks of contamination biases across treatment conditions. Randomization and masking: Using a blind manual drawing, clinics were randomly assigned to one of three conditions by the project director: one of two implementation approaches for integrating SCC into FP services (more intensive=SCC1, less intensive=SCC2) or usual care (existing FP services). Allocation was not concealed to providers, nor individual participants (although clients did not know if their clinic was SCC1 or SCC2). The assigned condition was applied clinic-wide for all clients.
Participants completed assessments at baseline, month 6 and month 12. If the client experienced a pregnancy by month 12, the final study assessment was conducted approximately one month after pregnancy completion. Participants were followed as long as they were still in a relationship with the partner they enrolled with at baseline; if they separated from this partner mid-study, the next assessment would be their final, unless they had an ongoing pregnancy which would lead to an assessment at the completion of that pregnancy. The protocol was approved by Institutional Review Boards at TASO and RAND, and described in a prior publication [14]. The research conformed to the principles embodied in the Declaration of Helsinki. The trial was registered with the NIH clinical trial registry (clinicaltrials.gov) and assigned the number NCT03167879.
Patient and Public Involvement. HIV clients provided input into the intervention development and outcome measures through participation in prior research that involved piloting the intervention and formative research including focus groups and in-depth interviews. Furthermore, HIV clients who volunteer as “expert clients” at the clinic sites were involved in aspects of implementing the intervention including community outreach and childbearing screening. These same clients will also be involved in dissemination of the study findings.
Study setting and participants
TASO is the oldest indigenous non-governmental organization in Uganda providing comprehensive HIV care. Each site provides care to 6,000-8,000 clients and has a staff of 15-20 medical providers. Clients were eligible if they met the following criteria: 1) in a serodiscordant relationship (partner’s HIV-negative status confirmed by rapid HIV test prior to enrollment), 2) of reproductive age (men age 15-60 years; women age 15-45), 3) considering childbearing with their partner (determined via triage screening item), 4) not currently pregnant (determined by a pregnancy test prior to enrollment), and 5) reports having disclosed HIV status to partner. Recruitment was stratified by sex to ensure a 50/50 balance and took place between July 2017 and January 2019. Clients who were potentially eligible were informed of the study by clinic staff and referred to the study coordinator for consent procedures. All enrolled participants provided written informed consent.
Intervention conditions
Our Choice. Informed by our earlier research [10, 15] and guided by an ecological adaptation of the Information, Motivation and Behavioral skills (eIMB) model of behavior change [16], we developed a multi-component, structured intervention that engages HIV clients and their partners with fertility desires in SCC. The goal of the Our Choice intervention is for providers to facilitate an informed childbearing decision making process and support each couple’s decision with training on the use of contraception or SCM in accordance with their reproductive goal. The intervention components are: 1) Client outreach to increase awareness and uptake of services, 2) Routine screening of childbearing desires at triage, and 3) Provision of SCC, starting with an initial consult conducted by an HIV counselor to facilitate an informed decision from the couple to pursue or delay pregnancy, followed by subsequent referral to FP nurses for either provision of contraception or monthly SCC sessions (see Supplement Figure 1). The client’s partner was encouraged to attend sessions, if possible. SCC was implemented by trained FP nurses, using a structured protocol and manual, the content of which is summarized in Supplement Table 1.
Low vs. high intensity approaches to implementation. Our Choice was implemented using two approaches, SCC1 and SCC2; both used the same content, manual and tools, but differed on the method, duration and frequency of training and supervision of FP nurses and HIV counselors (see Supplement Table 2). SCC1 was the more intensive approach, with the study team providing initial training of providers over two days and supervision starting twice-a-month before transitioning to monthly contact at month 6. SCC2 was less intensive and followed the standard model used by the Uganda Ministry of Health (MoH) to integrate new services: initial one day training and quarterly supervision sessions (although in reality supervision occurred every 6-9 months) provided by MoH supervisors who had been trained by the study team.
Usual care control. Control sites received no training and provided FP services as usual, with no use of routine screening of childbearing desires or SCC.
Measures
Assessments included measures of the primary outcome (see below), secondary outcomes (any use of SCM/contraception, pregnancy, partner seroconversion), and sociodemographic, HIV disease, and relationship/partner characteristics, as well as reproductive history and behaviors (see Table 1). Self-report measures underwent a translation, back-translation and group consensus review process in Luganda and Runyakitara (local languages used in the study settings) [17], and were interviewer-administered using computer-assisted software. All measures were assessed at each assessment time point.
Primary Outcome
The primary outcome was the accurate use of SCM [timed condomless intercourse (TCI), manual self-insemination (MSI), sperm washing] or modern contraception consistent with the client’s reproductive goal. This goal was determined at follow-up assessments by asking whether they had tried to conceive a child with their partner at any time in the past 6 months. Clients with missing responses were classified as trying to conceive if they endorsed “currently trying to conceive” at their most recent prior assessment. To assess accurate use of SCM, clients were asked in an open-ended format to describe exactly how they implemented their chosen SCM. Interviewers listened, probed for specific criterion not spontaneously mentioned, and rated the presence or absence of pre-defined criterion for each SCM (see Supplement Table 3). All criteria needed to be present for clients to be classified as accurately using the method. Accurate use of contraception was defined as using modern contraception as defined by FP practice in Uganda [18]. Clients were asked if they or their partner were currently using modern contraception methods (i.e., birth control pills, medroxyprogesterone acetate injection, intrauterine device, implant, tubal ligation/vasectomy and condoms), condoms or abstinence. Male participants were asked to consent to the interviewer calling their partner during the interview to assess use of contraception (calls were made in private and responses were not shared with the male partner). We also assessed an alternative outcome that combined use of modern contraception, consistent condom use or abstinence to allow for comparison to other studies.
Data Analysis
With a planned sample size of 400, enrolled evenly across 6 sites, and using an intra-class correlation (ICC) of 0.01 and assumed attrition of 10% attrition at month 12, our power analysis determined we would have > 80% power (2-tailed test) to detect a 4.5 percentage point difference (small effect size) for our comparison of the usual care arm to the combined SCC1 & SCC2 arms on the primary outcome, and a 7-8 percentage point difference between the SCC1 and SCC2 intervention arms.
We conducted initial bivariate analyses to compare the study arms on baseline measures of sociodemographic characteristics, HIV disease characteristics, partner/relationship characteristics, and reproductive health history/behavior for participants. We assessed differences using chi-square or Fisher’s exact test (FET) for categorical variables, and two-tailed independent t-tests for continuous variables. We repeated this analysis to compare those who completed the study with those who dropped out.
The primary analyses followed an intent-to-treat (ITT) approach (i.e., outcomes with missing data are designated as not engaging in the desired behavior or achieving the desired pregnancy status) and had two main objectives. First, to examine efficacy of Our Choice, we examined whether the primary outcome was more prevalent in the intervention group (SCC1 and SCC2 combined) compared to the usual care group. Second, we examined the relative efficacy of the two implementation approaches to administering Our Choice by comparing the effects of the higher intensity SCC1 versus the lower intensity SCC2. In both analyses, we first conducted bivariate chi-square or Fisher’s exact tests to assess associations between the outcome and treatment condition. We further tested for a significant association using covariate-adjusted Firth logistic regressions [19, 20]. Each logistic regression was adjusted for these covariates: age, sex, any secondary education, time since HIV diagnosis, marital status, length of relationship with partner, and whether the participant had a child with their partner. The model included fixed effects for site to account for clustering and allow for time-invariant differences across sites. (A mixed model approach was not possible for outcomes with quasi-separation or all zero values as was observed in the usual care arm). We relied on the effect sizes (estimated by the differences across study arms) as demonstration of practically significant results. Effect sizes were estimated using Cohen’s d. This same approach was used to examine effects on the secondary outcomes. We replicated all analyses with study completers only. The results of this analysis (not presented) were very similar to that of the ITT analysis.
Cost-effectiveness Analysis. We conducted a cost-effectiveness analysis based on program costs for personnel, supervision, training, and materials across both active study arms. Personnel costs included those for counselors, nurses and expert clients, who over the course of the study were asked to recall how much time they spent weekly delivering intervention activities. Supervision costs were estimated by multiplying the duration of supervision sessions by the respective supervisor’s hourly salary and number of supervisory sessions conducted and including any transportation costs. Training costs included those for the initial trainings for providers (including transport and per diem), as well as any refresher trainings conducted, and costs incurred for training supervisors. Material costs included those for brochures, videos, phone top-ups for nurses, and transport for participants.
To examine the cost-effectiveness of the two different implementation approaches used in SCC1 and SCC2, we compared total intervention costs across the two active study arms for the primary outcome. Because this is one of the first analysis of SCC in this setting, we also assessed the costs for accurate use of SCM on its own among those trying to conceive. The cost-effectiveness of contraception provision in Uganda has already been established [21]. Because supervisors in SCC1 were research staff with significantly higher salaries than the MoH supervisors in SCC2, we conducted an additional more realistic “scale-up scenario” showing the costs of SCC1 if its supervisors’ salaries were the same as the MoH supervisors’ salaries. The cost-effectiveness ratio was calculated as the cost per participant divided by the relative effect size in that group compared to the usual care control. Costs in local currency were converted at 3600 Ugandan Shillings=US $1 and reported in 2020 prices adjusted for inflation.