Objective: To assess the incidence and to evaluate the feasibility of inter-unit continuous surveillance of intensive care unit (ICU)-acquired infections. Design: Prospective multicentre, longitudinal, incidence survey. Setting: Five ICUs in university hospitals in western France. Patients: All patients admitted to the ICU during two 3-month periods (1994–1995). Measurements and results: The main clinical characteristics of the patients, ICU-acquired infections, length of exposure to invasive devices and the micro-organisms isolated were analysed. The study included 1589 patients (16 970 patient-days) and the infection rate was 21.6 % (13.1 % of patients). The ventilator-associated pneumonia rate was 9.6 %, sinusitis 1.5 %, central venous catheter-associated infection 3.5 %, central venous catheter-associated bacteraemia 4.8 %, catheter-associated urinary tract infection 7.8 % and bacteraemia 4.5 %. The incidence density rate of ICU-acquired infections was 20.3 ‰ patient-days. Ventilator-associated pneumonia and sinusitis rates were 9.4 and 1.5 ‰ ventilation-days, respectively. Central venous catheter-associated infection and central venous catheter-associated bacteraemia rates were 2.8 and 3.8 ‰ catheter-days, respectively. The catheter-associated urinary tract infection rate was 8.5 ‰ urinary catheter-days and the bacteraemia rate 4.2 ‰ patient-days. Six independent risk factors for ICU-acquired infection were found by stepwise logistic regression analysis: absence of infection on admission, age > 60 years, length of stay, mechanical ventilation, central venous catheter and admission to one particular unit. A total of 410 strains of micro-organisms were isolated, 16.8 % of which were Staphylococcus aureus (58.0 % methicillin-resistant). Conclusion: This prospective study using standardised collection of data on the ICU-acquired infection rate in five ICUs identified six risk factors. It also emphasized the difficulty of achieving truly standardised definitions and methods of diagnosis of such infections.