According to the World Health Organization, 98% of fatal dengue cases can be prevented; however, endemic countries such as Colombia have recorded higher case fatality rates during recent epidemics. We aimed to identify the predictors of mortality that allow risk stratification and timely intervention in patients with dengue. We conducted a hospital-based, case-control (1:2) study in two endemic areas of Colombia (2009-2015). Fatal cases were defined as having either 1) positive serological test (IgM or NS1), 2) positive virological test (RT-PCR or viral isolation), or 3) autopsy findings compatible with death from dengue. Controls (matched by state and year) were hospitalized nonfatal patients and had a positive serological or virological dengue test. Exposure data were extracted from medical records by trained staff. We used conditional logistic regression (adjusting for age, gender, disease's duration, and health-care provider) in the context of multiple imputation to estimate exposure to case-control associations. We evaluated 110 cases and 217 controls (mean age: 35.0 versus 18.9; disease's duration pre-admission: 4.9 versus 5.0 days). In multivariable analysis, retro-ocular pain (odds ratios [OR] = 0.23), nausea (OR = 0.29), and diarrhea (OR = 0.19) were less prevalent among fatal than nonfatal cases, whereas increased age (OR = 2.46 per 10 years), respiratory distress (OR = 16.3), impaired consciousness (OR = 15.9), jaundice (OR = 32.2), and increased heart rate (OR = 2.01 per 10 beats per minute) increased the likelihood of death (AUC: 0.97, 95% confidence interval: 0.96, 0.99). These results provide evidence that features of severe dengue are associated with higher mortality, which strengthens the recommendations related to triaging patients in dengue-endemic areas.