OBJECTIVE: To assess whether cross-checking of the physician ICU transfer report by ICU nurses may reduce transfer report errors. DESIGN: Prospective, observational study with random selection (according to patient registration code) of ICU transfer reports. SETTING: Eight-bed multidisciplinary intensive care unit of a teaching hospital. PATIENTS AND PARTICIPANTS: ICU transfer reports of 123 patients were randomly selected at discharge from the ICU between November 2006 and February 2007. INTERVENTIONS: Physician ICU transfer reports were cross-checked by nurses using defined review criteria. Inter-rater agreement (between nurses and the head of ICU) was assessed by kappa-values, and was excellent overall (0.9). All intercepted errors (100%) were consequently corrected by the interns. MEASUREMENTS AND RESULTS: Out of 123 transfer reports, 76 (62%) were affected by at least one error. Among 305 intercepted errors, 247 were prescription errors (26% of all prescriptions), 45 involved proposed procedures, and 13 were deficient in updating diagnoses. Most of the errors (248/305, 81%) were classified as simple, 43 (14%) as serious, or 14 (5%) as critical. Thirty-five (28%) transfer reports were considered potentially harmful (i.e., affected by at least one critical/serious error). In a multivariate model, only the number of medications included in the transfer report was associated with the occurrence of at least one critical/serious error. CONCLUSIONS: Errors in ICU transfer reports are frequent and may be potentially harmful. ICU nurses may help to effectively and accurately intercept those inaccuracies, and therefore reduce the exportation of errors from the ICU to the ward.