Sources and Impact of Misclassification Errors in HIV Rapid Test Algorithms in a Hyper-Endemic HIV Setting in sub-Saharan Africa 

Staff: Simon Gregson, Louisa Moorhouse, Tawanda Dadirai, Haynes Sheppard, Justin Mayini, Nadine Beckmann, Morten Skovdal, Janet Dzangare, Brian Moyo, Rufurwokuda Maswera, Benjamin A Pinsky, Sungano Mharakurwa, Ian Francis, Owen Mugurungi,Constance Nyamukapa 

There is growing evidence that HIV rapid tests are providing incorrect results to many infected and uninfected individuals (misclassification errors) when used in routine health programmes in sub-Saharan Africa – despite having high sensitivity and specificity under controlled laboratory conditions. This study is an in-depth investigation of the causes, impacts and acceptability thresholds of misclassification error in HIV test algorithms through a collaborative project with the Zimbabwe Ministry of Health and Child Care (ZMOHCC) on the latest round of national antenatal clinic-based HIV surveillance. The over-riding purpose of the study therefore is to inform efforts to reduce AIDS mortality, improve the health of people living with HIV, and prevent new infections in sub-Saharan Africa, by reducing errors in the HIV rapid test results given to people tested in routine health programmes. 

We measured the performance of HIV rapid test kits under local field conditions, investigated the use of incorrect or malfunctioning test kits, errors in application and interpretation of HIV rapid test kit results, and recording errors through a combination of quantitative and qualitative methodologies. Using data from this investigation, we assess the epidemiological and health service impacts of misclassification errors using a mathematical model. Findings from these investigations will be discussed with key stakeholders to establish acceptability thresholds and technological and quality assurance solutions considered necessary to reduce misclassification errors. The data provides robust scientific evidence to inform decisions on the need to reduce misclassification errors and on what needs to be done to achieve this.