This paper describes the construction and validation of a prognostic model for predicting post-bloodstream infection survival up to Day 21. A Weibull multiple regression model was adopted in a prospective cohort study of all patients diagnosed with true bacteremia or fungemia in a teaching hospital between 1991 and 1994 (training set, 1,577 patients). The final model included six variables easily detected in any institution: source of infection, underlying neoplasm, septic shock, community-acquired, age over 65, and polymicrobial bacteremia. Using this model, it is possible to obtain a graphic representation of survival probability for any combination of these risk factors. The model was tested on a second set of patients diagnosed in the same hospital between 1996 and 1997 (validation set, 952 patients), confirming its reliability in predicting survival. In conclusion, the Weibull function, together with variables easily identified at bedside, enables a precise prediction of the short-term, post-bloodstream infection mortality of a given patient.