Sampling design
A two-stage cluster design will be used to generate a probability sample of households within the study area for each of the survey domains (control versus intervention groups). Detailed sampling frames consisting of a count of all households within primary sampling unit (PSU - village) are currently available from the NMCP. PSUs will consist of villages within the study area of Gash Barka, divided into intervention and control villages. To maximize sampling efficiency, 58 PSUs will be selected for each survey domain at the first stage, with 15 households selected from each PSU at the second stage, thereby obtaining a total sample size of 1,740 houses (after adding non-response, the total number of houses is equal to 1,914). The list of eligible PSUs for the study was provided by the NMCP. All households within each selected PSU will be enumerated by the Orotta School of Medicine to create the second stage sampling frames for households. Enumeration of households within each PSU will also allow data to be weighted during analysis if needed. For the second stage, 17 households (ultimate sampling units) will be systematically (systematic random sampling) selected within each enumerated PSU. Data will then be collected at the household level. Survey fieldwork will be conducted just following/during the rainy season in September, 2009.
Sample Size and Power Calculations
A sample size of 870 households in the treatment group (58 villages, 15 households per village) and the same number in the control group will allow the detection (at a 5% level of significance) of a reduction of 10.47% points in malaria parasite prevalence from a baseline of about 15-20% with 80% power. This assumes that the intra-cluster correlation is 0.2. The total sample size is 1,740 households. Adding 10% to account for non-response yields a total sample size of 1,914 houses. Given that there are approximately 4.5 individuals per household, the total sample of individuals will be 8,613.
