A household survey was carried out in nine districts of Sri Lanka, one district randomly selected from each province. Each district constituted one study area, except the Colombo district which had two study areas: the Colombo Municipal Council area and other areas in the district. Data collection was done using an interviewer administered questionnaire and complemented with community interviews, key informant interviews, and market surveys. The estimated sample size per district was 617 households totaling to 6170 households in 10 study areas. A multistage cluster sampling method based on the probability proportional to size technique was used to identify 30 clusters per district, each cluster located in a Grama Niladhari division. A systematic random sampling technique was used within each cluster to identify 21 households.