We tested the performance of a previously developed index to diagnose occult bacterial infection and bacteremia in febrile patients. A total of 113 patients consecutively hospitalized because of an acute febrile disease, without a recognizable source of fever, were divided into four groups, with increasing probability of bacterial infection and bacteremia. None of the patients in the first group had bacterial infection or bacteremia. The incidence of bacterial infection and bacteremia was 27% and 11% in the second group, 32% and 17% in the third group, and 53% and 35% in the fourth group. No patient in the first group died, as opposed to 29% of patients in the fourth group. The use of the index at admission would have probably changed the treatment of 11% of patients. Thus, the index could be used to identify patients to be discharged from the emergency department (first group) or hospitalized and treated empirically (fourth group).