OBJECTIVES: To determine the incidence and identify risk factors of critical incidents in an ICU. DESIGN: Prospective observational study of consecutive patients admitted over 1 year to an ICU. Critical incidents were recorded using predefined criteria. Their causes and consequences were analysed. The causes were classified as technical failure, patient's underlying disease, or human errors (subclassified as planning, execution, or surveillance). The consequences were classified as lethal, leading to sequelae, prolonging the ICU stay, minor, or without consequences. The correlation between critical incidents and specific factors including patient's diagnosis and severity score, use of monitoring and therapeutic modalities was analysed by uni- and multivariate analysis. SETTING: An 11-bed multidisciplinary ICU in a non-university teaching hospital. PATIENTS: 1,024 consecutive patients admitted to the ICU. INTERVENTION: None. MEASUREMENTS AND MAIN RESULTS: The median length of ICU stay by the 1,024 patients was 1.9 days. Of the 777 critical incidents reported 2% were due to technical failure and 67 % to secondary to underlying disease. There were 241 human errors (31%) in 161 patients, evenly distributed among planning (n = 75), execution (n = 88), and surveillance (n = 78). One error was lethal, two led to sequelae, 26 % prolonged ICU stay, and 57 % were minor and 16 % without consequence. Errors with significant consequences were related mainly to planning. Human errors prolonged ICU stay by 425 patient-days, amounting to 15 % of ICU time. Readmitted patients had more frequent and more severe critical incidents than primarily admitted patients. CONCLUSIONS: Critical incidents add morbidity, workload, and financial burden. A substantial proportion of them are related to human factors with dire consequences. Efforts must focus on timely, appropriate care to avoid planning and execution mishaps at the beginning of the ICU stay; surveillance intensity must be maintained, specially after the fourth day.