BACKGROUND: Malaria risk may be heterogeneous in urban areas of Africa. Identifying those at highest risk for malaria may lead to more targeted approaches to malaria control. METHODS: A representative sample of 558 children aged 1-10 years were recruited from a census population in a single parish of Kampala and followed up for 2 years. Malaria was diagnosed when a child presented with a new episode of fever and a thick blood smear positive for parasites. Multivariate analysis was used to identify independent predictors of malaria incidence. RESULTS: A total of 695 episodes of uncomplicated malaria were diagnosed after 901 person years of follow-up. Sickle cell trait (relative risk [RR], 0.68 [95% confidence interval {CI}, 0.52-0.90]), glucose-6-phosphate dehydrogenase deficiency in female children (RR, 0.48 [95% CI, 0.31-0.75]), and use of an insecticide-treated bed net (RR, 0.52 [95% CI, 0.32-0.83]) were associated with a lower risk of malaria. The distance of the subject's residence from a swamp bordering the parish showed a strong "dose-response" relationship; living in the swamp was the strongest predictor of malaria risk (RR, 3.94 [95% CI, 2.61-5.97]). CONCLUSION: Malaria incidence was highly heterogeneous in this urban cohort of children. Malaria control interventions in urban areas should target populations living in pockets of high malaria risk.