| Abstract |
HIV prevalence rates have recently been adjusted in resource-poor settings such as Sub-Saharan Africa due to the introduction of new estimation methods, in particular of population-based household surveys. However, a potentially important limitation of household surveys is bias due to data falsification by the interviewer. This issue is widely ignored in the international literature yet could have a major impact on HIV prevalence estimates as our study indicates. In a population-based household survey that we conducted in 2006 in Windhoek, Namibia according to standard DHS+ survey practices, a fraudulent nurse was discovered. The nurse had interviewed and taken oral fluid samples for HIV testing from 313 of the 2,452 individuals. Deleting the respondents interviewed by this particular nurse from the data resulted in a downward adjustment of the HIV rate in our sample from 12.7% to 9.6%. Considering that important policy decisions are made based on HIV prevalence estimates, it remains crucial to critically look at the reliability of these estimates. This also calls for more attention to the selection and supervision of interviewers as well as a retrospective analysis of databases that include interviewer ID-codes. |