OBJECTIVES:: To study the prognostic value of fever in the emergency department in septic patients subsequently admitted to the ICU. DESIGN:: Observational cohort study from the Swedish national quality register for sepsis. SETTING:: Thirty ICU’s in Sweden. PATIENTS:: Two thousand two hundred twenty-five adults who were admitted to an ICU within 24 hours of hospital arrival with a diagnosis of severe sepsis or septic shock were included. INTERVENTIONS:: None. MEASUREMENTS AND MAIN RESULTS:: Body temperature was measured and classified according to four categories (< 37°C, 37–38.29°C, 38.3–39.5°C, ≥ 39.5°C). The main outcome was in-hospital mortality. Odds ratios for mortality according to body temperature were estimated using multivariable logistic regression. Subgroup analyses were conducted according to age, sex, underlying comorbidity, and time to given antibiotics. Overall mortality was 25%. More than half of patients had a body temperature below 38.3°C. Mortality was inversely correlated with temperature and decreased, on average, more than 5% points per °C increase, from 50% in those with the lowest temperatures to 9% in those with the highest. Increased body temperature in survivors was also associated with shorter hospital stays. Patients with fever received better quality of care, but the inverse association between body temperature and mortality was robust and remained consistent after adjustment for quality of care measures and other factors that could have confounded the association. Among vital signs, body temperature was best at predicting mortality. CONCLUSIONS:: Contrary to common perceptions and current guidelines for care of critically ill septic patients, increased body temperature in the emergency department was strongly associated with lower mortality and shorter hospital stays in patients with severe sepsis or septic shock subsequently admitted to the ICU.