OBJECTIVE: To evaluate the current definitions for sepsis and clarify and quantify the risk for intensive care unit (ICU) patients with sepsis. DESIGN: A prospective cohort analysis of 519 patients with a primary clinical diagnosis of sepsis treated in the ICUs of 40 US hospitals drawn from a nationally representative sample of 17,440 admissions. MEASUREMENTS: Patient's age, treatment location prior to ICU admission, comorbidities, origin of sepsis, daily physiologic measurements, therapeutic intensity, and subsequent hospital mortality rate. INTERVENTION: Patients were categorized into subgroups by important risk factors and into current clinical definitions of sepsis. Patients also were provided an individual risk of hospital mortality based on their individual predicted risk by using the first ICU day APACHE III score, treatment location prior to ICU admission, and etiology of sepsis. RESULTS: Patients with a designated urinary source of sepsis had a significantly lower baseline risk of death (30 percent) than patients with other causes (54 percent, p less than 0.01). Patients admitted to the ICU from the emergency department also had significantly lower mortality (37 percent) than patients admitted from hospital wards, other units within the hospital, or transferred from other hospitals (55 percent, p less than 0.01). Recognized definitions such as "sepsis syndrome" and "septic shock" identified groups of patients with significantly different mortality rates, 40 percent and 64 percent, respectively (p less than 0.01), but the range of individual patient risks within these groups were indistinguishable from the 211 patients (41 percent) that did not meet these definitions during the initial seven days of ICU treatment. Multivariate analysis using initial APACHE III score, etiology (urosepsis or other), and treatment location prior to ICU admission provided the greatest degree of discrimination (ROC = 0.82) of patients by risk of hospital death. CONCLUSIONS: Sepsis is a complex clinical entity and could be viewed as a continuum with substantial variation in initial severity and risk of hospital death. One accurate description of sepsis is the continuous measure of hospital mortality risk estimated primarily from physiologic abnormalities.