Measuring transactional sex in different contexts: How do tools to measure this practice perform in rural South Africa?

Background: Adolescent girls and young women across sub-Saharan Africa are at disproportionate risk of HIV infection compared to their male counterparts. Transactional sex has been identified as an important proximate risk for infection in this population. Definitions and measures of transactional sex vary, necessitating improved measures to better estimate prevalence across settings, over time, and to understand the mechanisms through which transactional sex increases HIV risk. This article describes the results of cognitive interviews in rural KwaZulu-Natal in South Africa to evaluate the performance of an improved measure of transactional sex. Methods: Data were collected between May and June 2017 with sexually active adolescent girls and young women (n = 10) and men (n = 10) drawn from a general population sample. Two questions were tested. Audio-recorded interviews were conducted in isiZulu using a structured tool. Matrices were used to summarise the data across participants which were then compared using constant comparative techniques. Results: Participants captured the instrumental nature of transactional sex relationships clearly and understood that the questions were about relationships that were primarily motivated by benefit. However, despite prior qualitative research in this setting describing transactional sex as widely practised, only one male participant answered either question in the affirmative in this face-to-face interview. This implies a judgement placed on relationships that are deemed as having been motivated mainly by exchange, perhaps compelling people to under-report such relationships. Conclusion: Participants' unwillingness to answer in the affirmative highlights the importance of understanding the research context and the possible social and historical influences which may influence how survey questions are answered. This has implications for measurement development, and highlights the need for measures that can be responsive to contextual differences. Further research is needed for refinements to measurement approaches in this and other settings.