OBJECTIVE: Methicillin-resistant Staphylococcus aureus (MRSA) is a major problem in intensive care units (ICU). International guidelines recommend screening patients for MRSA on admission, although consensus on sites required for optimum detection has not been reached. Our aim was to determine whether throat and rectal swabs identified a significant number of additional MRSA-colonised patients not captured by swabbing at keratinized skin carriage sites (anterior nares, perineum and axillae). DESIGN: Prospective cohort study. SETTING: 30-Bed medical and surgical ICU in a tertiary teaching hospital. PATIENTS: One thousand four hundred and eighty adult patients consecutively admitted over 15 months. MEASUREMENTS AND RESULTS: Swabs from carriage sites (anterior nares, perineum, axillae, throat and rectum), wounds and clinical samples taken within 48 h of ICU admission were analysed to identify patients admitted with MRSA. A complete set of carriage swabs were received from 1,470 patients. 105 (7%) patients were admitted with MRSA of which 63 (60%) were detected by a pooled keratinized skin swab (anterior nares, perineum, axillae). A further 36 (34%) patients were detected only by throat or rectal swabs. Indeed, throat and rectal swabs combined had a higher sensitivity than pooled keratinised skin swabs (76 vs. 60% P = 0.0247). Swabs from all carriage sites together detected 95% (100) of MRSA positive patients, with five patients being positive at wound sites only. CONCLUSIONS: The throat and rectum are important and potentially hidden sites of MRSA carriage in critically ill patients. These findings prompt the need for larger studies to determine the most cost-effective screening strategy for MRSA detection. DESCRIPTOR: Non-pulmonary nosocomial infections.