3.12.1 Sample size calculations
The sample size for the quantitative data required to obtain estimated proportions with 95% probability level was estimated using the CSURVEY design in Epi Info 6 version 3.22 (Centers for Disease Control and Prevention (CDC), 2004). The estimated total numbers of adults on ART in the study sites at the time of the study were 1425, 2400, 2055 and 1308 for Mahalapye, Serowe, Maun and Molepolole respectively. This was based on the assumption that 85% of the patients achieve optimal adherence (i.e. have adherence rates of > 95%). These estimates were arrived at using the reports from the health care providers and predicted rates for Botswana (Weiser et al., 2003). Using the CSURVEY design in Epi Info 6, version 3.22 (CDC, 2004) (with expected rates 85% and worst acceptable estimate 78%) the sample size was estimated to be 93, 96, 95 and 93 for Mahalapye, Serowe, Maun and Molepolole respectively.
3.12.2 Sampling and data collection
Some of the quantitative data were collected using the exit interview tool, while the bulk of these data were collected using an adherence tool. The exit interview was carried out at the end of the patient's visit after collecting the medication, while the adherence tool was administered at any point during the consultation process but before collecting the medication.
Research associates conducted interviews for health workers, national level policymakers and carried out the observations of health facilities. The research associates and the social workers conducted the 23 semi-structured interviews, 163 exit interviews and 514 adherence questionnaires with ARV users, while a research associate moderated 16 FGDs for community members and ARV users. Research associates collected the data and supervised the data collectors in a site where they are not resident. This was meant to increase the objectivity of the data collected and to address the issue of bias that might arise if research associates collected data in the facilities where they were employed.