Advanced monitoring techniques which require a long duration of insertion of catheters into patients have caused the increase of nosocomial bloodstream infections, especially in the intensive care setting. To provide information to prevent this catheter-related infection, this prospective follow-up study goal is to estimate the occurrence of the primary nosocomial bloodstream infection in severely ill patients, and to explore the possible relation with the use of invasive devices. From Jan. 1 to Dec. 31 1996, all 858 patients admitted into the ICU of a medical center were enrolled into this study. The severity of diseases, type and duration of each in vasive apparatus inserted into each patient were recorded. Cases of primary nosocomial bloodstream infection were identified, and the infection rate for each invasive apparatus was determined as the number of infected cases divided by the person-day experience of patients received the same apparatus. The rate of primary nosocomial bloostream infection in the ICU was 7.35/1000 (65 cases/8842 person-day). The rate increases with the duration of stay in the ICU (p<0.005). Patients who were more sick were prone to get the in fection (p=0.025). The risk of nosocomial infection increases as the duration of invasive catheters in situ increases. By a logistic regression analysis, the severity of diseases (APACHE II score), the insertion of urinary catheter, central venous line, or arterial line were all associated with an elevated risk. The findings of this study are similar to other reports. Health care providers should pay more attention to following the guidelines to control the nosocomial infection. Future study can focus on the adverse outcomes of the infected patients in order to estimate the cost of these nosocomial infections.