Objective Case death rates for severe childhood malnutrition remain stubbornly elevated in high HIV prevalence settings, despite the implementation of WHO guidelines. This study examined case death and other clinical outcomes in malnourished children with and without HIV infection. Methods A prospective, observational study was undertaken at three tertiary hospitals in Johannesburg, South Africa. All severely malnourished children had their HIV status established, and anthropometric, clinical and diagnostic findings and admission outcomes were analysed. Findings Just over half (51%) of the 113 severely malnourished children were HIV infected, but 31/58 (54%) of these children had their positive status diagnosed only after admission. Marasmic children were significantly more likely to be HIV infected (OR 9.7, 95% CI 3.5 to 29.1). Tuberculosis (TB) was strongly suspected and treated in 27 children (24%) although confirmed in only five (4%). The overall case death rate was 11.5%. HIV infection, pallor and shock were significant predictors of death. HIV-infected children were six times more likely to die compared with HIV-negative children (19% vs 3.6%, OR 6.2, 95% CI 1.2 to 59). HIV-‘affected’ children (HIV negative but exposed) and HIV-negative children had similar outcomes. Conclusion HIV infection significantly increases severe malnutrition case death. WHO guidelines for the management of severe malnutrition in high HIV prevalence settings need to be modified to include routine HIV and TB testing and offer guidance on the criteria and timing of TB treatment and highly active antiretroviral therapy initiation.