BACKGROUND: Severe malnutrition in childhood associated with HIV infection presents a serious humanitarian and public health challenge in Southern Africa. The aim of this study was to collect country wide data on HIV infection patterns in severely malnourished children to guide the development of integrated care in a resource limited setting. METHODS: A cross sectional survey was conducted in 12 representative rural and urban Nutrition Rehabilitation Units (NRUs), from each of Malawi's 3 regions.All children and their caretakers admitted to each NRU over a two week period were offered HIV counselling and testing. Testing was carried out using two different rapid antibody tests, with PCR testing for discordant results. Children under 15 months were excluded, to avoid difficulties with interpretation of false positive rapid test results. The survey was conducted once in the dry/post-harvest season, and repeated in the rainy/hungry season. RESULTS: 570 children were eligible for study inclusion. Acceptability and uptake of HIV testing was high: 523(91.7%) of carers consented for their children to take part; 368(70.6%) themselves accepted testing. Overall HIV prevalence amongst children tested was 21.6%(95% confidence intervals, 18.2-25.5%). There was wide variation between individual NRUs: 2.0-50.0%. Geographical prevalence variations were significant between the three regions (p < 0.01) with the highest prevalence being in the south: Northern Region 23.1%(95%CI 14.3-34.0%), Central Region 10.9%(95%CI 7.5-15.3%), and Southern Region 36.9%(95%CI 14.3-34.0%).HIV prevalence was significantly higher in urban areas, 32.9%(95%CI 26.8-39.4%) than in rural 13.2%(95%CI 9.5-17.6%)(p < 0.01).NRU HIV prevalence rates were lower in the rainy/hungry season 18.4%(95%CI 14.7-22.7%) than in the dry/post-harvest season 30.9%(95%CI 23.2-39.4%) (p < 0.001%). CONCLUSION: There is a high prevalence of HIV infection in severely malnourished Malawian children attending NRUs with children in urban areas most likely to be infected. Testing for HIV is accepted by their carers in both urban and rural areas. NRUs could act as entry points to HIV treatment and support programmes for affected children and families. Recognition of wide geographical variations in childhood HIV prevalence will ensure that limited resources are initially targeted to areas of highest need. These findings may have implications for the other countries with similar patterns of childhood illness and food insecurity.