Background: Sepsis from bloodstream infection (BSI) is an important cause of morbidity and mortality among surgical patients. Our hypothesis was that fever and leukocytosis during BSI would be associated with gram-negative pathogens and worse outcomes among hospitalized surgical patients. Study design: A prospectively collected dataset of all infections diagnosed on the adult general and trauma surgery services between December 1996 and December 2005 at the University of Virginia Hospital was reviewed. Fever was considered a temperature of ≥ 38.5°C, and leukocytosis was defined as a white blood cell count ≥ 15,000/μL within 24 hours of treatment for infection. Logistic regression was used to identify predictors of fever and mortality. Results: Over 9 years, 823 BSIs were analyzed. One hundred forty-eight BSIs resulted in death (18.0%), and 541 (65.7%) patients were febrile at diagnosis; mortality for these two groups were 12.9% and 27.7%, respectively (p < 0.0001). Febrile patients had a trend toward fewer gram-negative infections (27.0% versus 31.9%, p = 0.13), 403 had a leukocytosis at diagnosis and 420 did not; mortality for the two groups was 19.1% and 16.9%, respectively (p = NS). Higher maximum temperature was protective against mortality in the logistic regression analysis (odds ratio = 0.60 per C°, p < 0.0001). Conclusions: Among surgical patients with sepsis, fever during BSI was not associated with a gram-negative cause and correlated with survival, although increasing WBC had little effect. Mortality after BSI appears associated more with an initially blunted physiologic response than with a robust, proinflammatory response. In addition, a threshold for blood culture other than temperature ≥ 38.5°C should be considered.