Although increases in severity of mortality from dengue infection have been observed in Brazil, their determinants are not fully known. A case-control study was conducted by using the National Notifiable Diseases Surveillance System, including patients with severe dengue during 2000-2005. Cases were defined as patients that died and controls were those who survived. Hierarchical multivariate logistic regression was performed. During the study period, there were 12,321 severe cases of dengue and 1,062 deaths. Factors independently associated with death included age ≥ 50 years (odds ratio [OR] = 2.29, 95% confidence interval [CI] = 1.59-3.29), < 4 years of schooling (OR = 1.83, 95% CI = 1.47-2.28), a rural area (OR =2.84, 95% CI = 2.19-3.69), hospitalization (OR = 1.42, 95% CI = 1.17-1.73), and a high hematocrit (OR = 2.46, 95% CI = 1.85-3.28). Factors associated with a lower chance of dying were female sex (OR = 0.76, 95% CI = 0.67-0.87), history of previous dengue (OR = 0.78, 95% CI = 0.62-0.99), positive tourniquet test result (OR = 0.47, 95% CI = 0.33-0.66), laboratory diagnosis of dengue (OR = 0.75, 95% CI = 0.61-0.92), and a platelet count of 50,000-100,000 cells/mm(3) (OR = 0.56, 95% CI = 0.36-0.87). The risk profile identified in this study should serve to direct public health interventions to minimize deaths.