HIV prevalence is relevant for policy makers for planning control pro-grams and interventions. HIV prevalence estimated from population-based surveys is aected by the presence of non ignorable missing data. We use a bounding approach in order to asses the uncertainty in the estimation of the HIV prevalence due to the presence of missing data in the three waves of the Malawi Diusion and Ideational Change Project (MDICP). We propose dynamic bounds which exploit the absorbing nature of HIV to narrow the worst-case bounds without any additional restrictions. We then use instrumental and monotone instrumental variables restrictions to further narrow the bounds.