A randomized group antenatal care pilot showed increased partner communication and partner HIV testing during pregnancy in Malawi and Tanzania

Background: HIV testing at antenatal care (ANC) is critical to achieving zero new infections in sub-Saharan Africa. Although most women are tested at ANC, they remain at risk for HIV exposure and transmission to their infant when their partners are not tested. This study evaluates how an HIV-enhanced and Centering-based group ANC model-Group ANC+ that uses interactive learning to practice partner communication is associated with improvements in partner HIV testing during pregnancy. Methods: A randomized pilot study conducted in Malawi and Tanzania found multiple positive outcomes for pregnant women (n = 218) assigned to Group ANC+ versus individual ANC. This analysis adds previously unpublished results for two late pregnancy outcomes: communication with partner about three reproductive health topics (safer sex, HIV testing, and family planning) and partner HIV testing since the first antenatal care visit. Multivariate logistic regression models were used to assess the effect of type of ANC on partner communication and partner testing. We also conducted a mediation analysis to assess whether partner communication mediated the effect of type of care on partner HIV testing. Results: Nearly 70% of women in Group ANC+ reported communicating about reproductive health with their partner, compared to 45% of women in individual ANC. After controlling for significant covariates, women in group ANC were twice as likely as those in individual ANC to report that their partner got an HIV test (OR 1.99; 95% CI: 1.08, 3.66). The positive effect of the Group ANC + model on partner HIV testing was fully mediated by increased partner communication. Conclusions: HIV prevention was included in group ANC health promotion without compromising services and coverage of standard ANC topics, demonstrating that local high-priority health promotion needs can be integrated into ANC using a Group ANC+. These findings provide evidence that greater partner communication can promote healthy reproductive behaviors, including HIV prevention. Additional research is needed to understand the processes by which group ANC allowed women to discuss sensitive topics with partners and how these communications led to partner HIV testing. We use data from a 2-arm randomized pilot study conducted in Malawi and Tanzania that compared outcomes for pregnant women randomly assigned to individual or Group ANC+. Prior to enrollment, computer assigned random assignment slips representing each arm of the study were placed in identical envelops and manually shuffled to randomized order. After completing the baseline survey, the woman selected the first envelope in the batch which revealed the assignment to the woman and study team. Previously published work describes details about the randomization process, retention rates, methods, and primary outcomes [32].Before data collection, we received necessary approvals from three institutional review boards, the College of Medicine Research and Ethics Committee in Malawi, the National Institute for Medical Research in Tanzania, and the University of Illinois Chicago. We also received approval from the Ministries of Health and administrators at participating clinics. Malawi and Tanzania are low-income sub-Saharan African countries with high rates of maternal and infant morbidities and mortality. This pilot was launched in 2014 in two rural clinics in central Malawi and one urban clinic in Dar es Salaam, Tanzania where ANC followed focused antenatal care guidelines for four visits [35]. Women over the age of 15, with a gestational age between 20 and 24 weeks were recruited for participation. After completing the informed consent process, pregnant women completed the baseline survey and then were randomly assigned to one of two study conditions [32]. Participants were compensated with the equivalent of US$5 for taking the surveys. The compensation was not linked to their level of care engagement. A consort diagram with detailed recruitment and retention was previously published [32]. Services are provided on a first-come, first-served basis. At all visits, women met with a midwife individually for a brief physical assessment. Laboratory tests (including HIV testing) were undertaken at their first visit. Although not required, often women are present for a health lecture that provides a rapid overview of important topics. Women were expected to complete four visits and return to the clinic for two postnatal visits at one and 6 weeks after delivery. Attendance is recorded, but there is no reminder system in place. Women had the same number of scheduled visits as women in the individual arm. However, after an individual first (intake) visit, the other ANC visits and their 6-week postnatal check-ups occurred with the same consistent group of women with an approximately similar expected delivery date. Each scheduled 2-h appointment included women’s self-measurement and recording of their blood pressure and weight, followed by a one-on-one physical assessment in a group space with the midwife. The group then gathered in a circle, and a trained midwife and assistant facilitated interactive educational health promotion activities and discussions that focused on partner communication and HIV testing. Opportunities for community building occur throughout the session. The randomization indicator, Type of ANC (individual ANC or Group ANC+), was the primary variable of interest. Partner communication was measured by asking women whether they discussed three sexual health topics with their partner since coming to ANC: safer sex, HIV testing, and family planning. These three items were combined to produce the total number of items discussed, possible range, 0-3. Partner HIV testing was measured by asking women whether their partner had an HIV test since the woman started coming to ANC. Covariates included age at baseline (< 20, 20-34, 35+), gravidity (1 or > 1), education category (less than primary school, completed primary school, or more than primary school), relationship status (whether married or living with a partner, coded yes or no), parity (0 or ≥ 1), religion (Christian or Muslim), and access to an independent source of income (yes or no). We also included the country (Malawi or Tanzania) as a covariate because it encapsulates many economic and sociodemographic differences. Women completed a survey in late pregnancy (third trimester) that included the partner communication and HIV testing questions. Of the 218 women enrolled at baseline, 88% completed the late pregnancy survey. Women assigned to individual ANC in Malawi had the lowest retention rate (40/58 [69.0%]). Bivariate relationships were examined between the type of ANC and partner communication and partner HIV testing using Chi-squared tests. We then used the mediation analysis process using the procedures described by Baron and Kenny (1986) [36], which requires establishing that: (1) the causal variable (the type of ANC) significantly affected the outcome (partner HIV testing); (2) the causal variable also significantly affected the mediator (partner communication); and (3) when the mediator is added to the regression model, the relationship between the causal variable and outcome is no longer significant. We used multivariate logistic regression and cumulative ordinal regression models to examine the impact of the type of ANC on partner HIV testing and communication, respectively. Stepwise model selection method was employed in these regression models so that the estimates of the type of ANC effects were adjusted for significant covariates. We then introduced partner communication as a predictor of partner HIV testing to examine whether this variable mediated the effect of type of ANC on the relationship. Effect sizes from all logistic regression models were reported using Odds Ratios (OR) with 95% confidence interval estimates. The indirect effect of type of ANC on partner testing outcome through partner communication was calculated and tested using the Sobel method (1982) [37] given in the formula as follows, where Sa and Sb refered to the standard errors of the effects a and b in the mediation process (Fig. 1). All statistical tests were two-sided, controlled for Type I error probability of 0.05. Sobel test for mediation effect [37]