# Sampling Strategy

Participant sampling entailed a two-stage sampling scheme. In stage one, a representative sample of 575 enumeration areas (EA) were selected using a probability proportional to size sampling plan. In stage two, all households in the 575 EAs were enumerated using Global Positioning Systems (GPS) receivers to record the geographic coordinates of each household. A random selection of 26 households was identified from each EA. A census of household residents was conducted in each of the randomly selected households. Household members meeting study inclusion criteria were recruited for participation in the pre-cohort survey at T1.

### SHIMS Sample: 575 Enumeration Areas (EAs)

Sample size calculations were developed to achieve 80% power to detect a 45% reduction in male HIV incidence from before and after *Soka Uncobe*. Assuming a male HIV incidence rate of 2.0 per 100 person-years prior to *Soka Uncobe* [UNAIDS 2010], the targeted Cohort 1 male sample size was 5832.

To identify and enroll 5832 men, the following assumptions were made to estimate the number of households needed in the sample:

- A refusal and non-contact rate of 18% of sampled households,
- One male participant recruited into the study from every 2 households,
- 0.5 years of follow-up time per participant in the longitudinal cohort,
- A 10-15% loss-to-follow-up rate, and
- A design effect of 1.25. These assumptions resulted in a sample of 14,884 households. Among a sample of 14,884 households, it was estimated that 7106 women would be recruited for Cohort 1.