Sample size and power

Although this is a three-arm study with the possibility of three between-group comparisons, our sample size estimation is based on comparison between the control group (community leader engagement only) and the group with all three interventions (community leader engagement, community day events, and peer groups). The other two comparisons will be exploratory to estimate the added individual effect of community day events or peer groups.

Based on country program data, we assumed that retention at 6–8 weeks post-delivery (as measured by HIV exposed infants returned for HIV diagnosis at 6–8 weeks) in the control arm will be about 50 %. In a simple randomized control trial, we need 126 HIV positive women per arm in order to detect at least a 20 percentage point increase in retention with 5 % significance level and 90 % power. We estimate that we can enroll at least 14 HIV positive women per health facility. For our cluster-randomized study, we assumed an intra-cluster correlation coefficient for the retention outcome of about 0.01 and an average cluster size of 14 HIV positive women, leading to a design effect of 1.13 and a sample size per arm of 143. With an average of 14 women per facility, this implies 11 facilities per arm. To allow for uncertainty of the intra-cluster correlation coefficient, we increased the number of facilities per arm to 15 (the maximum number of facilities we could afford to enroll in the study). This would allow us to detect the same effect of 20 % with 5 % significance and 90 % power at intra-cluster correlation coefficients as high as 0.05.

Across the trial 321 community leaders will be recruited and trained, an average of 7 leaders per cluster. Table 4 presents this information by country. This number of leaders will be selected based on the formative research into community structures and leadership and the feasibility and resources available to train and effectively support them. Efforts will be made to identify formal and informal leaders who have broad reach in terms of being able to mobilize other leaders.

Expected pregnancies and intervention sample size/recruitment targets

^aBased on national prevalence women 15–49 years

Across the countries a total of 70 community days will be held over the 24-month intervention period. The number of community days was estimated based on the desired/possible saturation levels, balanced by the feasibility, logistics, and available resources.

The sample size and recruitment targets for women in the MCH classes will be based on reaching 30 % of the projected number of pregnant women by country in order to effect community diffusion of information [29]. Across all the countries 4,163 pregnant and postpartum women will be recruited into the MCH classes, with Uganda contributing more than half of that number due to its larger cluster population sizes and higher fertility rate. The number of men in the male discussion groups was based on recruiting at least 30 % of the number of women in the MCH classes, for a total of 1,388 men. We based this number on the feasibility and practicality of recruiting these many men and on the resources available. As for the peer facilitators of these groups, we estimated that a peer facilitator could manage a group of up to 15 participants at a time and be able to conduct three groups during a year, so that we will recruit 93 peer facilitators for the women’s groups and 30 for the men’s group overall.