2.2 Sample size and sampling procedure
The sample size was determined with a software, the Open Source Epidemiologic Statistics for Public Health ( OpenEpi) 2.3 (Dean et al., 2009), with power was at 50% and 4% margin of error at 95% confidence level, while a design effect of one was used. A sample size of 600 beef farmers was obtained. A 10% contingency was added to take care of non-response, and 660 farmers were enrolled into the survey. For the laboratory based study, sample size was computed with the same software and power set at 50%, 5% absolute precision at 95% confidence level. A sample size of 384 cattle was determined.
A two-step cluster sampling procedure was adopted. It involved first, randomly selecting cattle fattening farms settlements in the northern area of the country, followed by stratification of farms based on type: intensive and semi-intensive (clusters). Secondly, systematic random sampling of the farms was conducted with a selection probability proportional to the farm size. Questionnaires were administered on the farm owners or their managers. A two-stage sampling method was also conducted for selection of animals. However, urine samples were collected from at least one cattle per farm. Voided urine was collected from 384 animals for antimicrobial residues screening.