Randomisation of sites
Following the 2008 baseline survey, in 2009 we randomised the 77 clusters included in the survey: 25 each in Botswana and Swaziland and 27 in Namibia. At the time of the baseline survey, we sought and obtained informed consent from community leaders for their community to participate in the trial, including being randomised to receive one or more of trial interventions or to be initially a control site for the trial.
An epidemiologist not involved in the fieldwork randomised the clusters within each country separately, using a random number sequence generator. Within each country, we first stratified the clusters
- into urban and rural clusters
- by HIV prevalence in the baseline survey (above or below the average for the stratum) within urban and rural strata
In a factorial design, we assigned clusters to receive no intervention, or combinations of one or more of the three interventions. The resulting pattern of a control and three intervention arms allows each intervention to be tested on its own and in combination with others.
Randomisation of clusters to combinations of the trial interventions
Return to the Implementation page.
Young women in Swaziland set up a choice disability campaign in their community.
Service providers in Botswana get together to organise a drama group competition in their village.
A group of young women in Namibia organise a community feast to celebrate their enterprise. 
