Serum lactate is a potentially useful biomarker to risk-stratify patients with severe sepsis; however, it is plausible that elevated serum lactate is simply a manifestation of clinically apparent organ dysfunction and/or shock (i.e., refractory hypotension). To test whether the association between initial serum lactate level and mortality in patients presenting to the emergency department (ED) with severe sepsis is independent of organ dysfunction and shock. Single-center cohort study. The primary outcome was 28-day mortality and the risk factor variable was initial venous lactate (mmol/L), categorized as low (< 2), intermediate (2-3.9), or high (> or = 4). Potential covariates included age, sex, race, acute and chronic organ dysfunction, severity of illness, and initiation of early goal-directed therapy. Multivariable logistic regression analyses were stratified on the presence or absence of shock. The ED of an academic tertiary care center from 2005 to 2007. Eight hundred thirty adults admitted with severe sepsis in the ED. None. Mortality at 28 days was 22.9% and median serum lactate was 2.9 mmol/L. Intermediate (odds ratio [OR] = 2.05, p = 0.024) and high serum lactate levels (OR = 4.87, p < 0.001) were associated with mortality in the nonshock subgroup. In the shock subgroup, intermediate (OR = 3.27, p = 0.022) and high serum lactate levels (OR = 4.87, p = 0.001) were also associated with mortality. After adjusting for potential confounders, intermediate and high serum lactate levels remained significantly associated with mortality within shock and nonshock strata. Initial serum lactate was associated with mortality independent of clinically apparent organ dysfunction and shock in patients admitted to the ED with severe sepsis. Both intermediate and high serum lactate levels were independently associated with mortality.