Risk assessment is used to determine the need for isolation in single rooms. Limited availability of isolation rooms and/or operational needs may compromise this process. This article reports the results of a 12-month prospective observational study of every infection control request for isolation in a 1100-bed teaching hospital. In addition, four point-prevalence surveys of the usage of single rooms were carried out. Data were collected on the incidence of new clinical meticillin-resistant Staphylococcus aureus (MRSA) isolates per ward and these were correlated with rates of isolation failures for MRSA cases. There were 845 requirements for patient isolation, of which 185 (22%) could not be met (isolation failures). Three-quarters of the requirements for isolation were due to MRSA or Clostridium difficile. The proportion of isolation failures was consistent for most organisms and conditions but varied markedly between clinical specialities (0-57%). Reasons for failure to isolate included no single rooms available, all single/isolation rooms occupied (for both isolation and non-infection-control reasons), limitations on the use of single rooms in mixed-sex wards and patient-specific reasons. Only a minority of the available single rooms were occupied for infection control reasons (12-19%). There was a statistically significant correlation between isolation failures and MRSA incidence (Spearman's rho 0.596, P<0.001). In only one case where a ward had >or=30% of its beds provided in single rooms was there an instance of failure to isolate. In conclusion, insufficient capacity to isolate patients with potentially transmissible pathogens is common and may compromise infection control requirements. Either isolation capacity must be increased or evidence-based risk assessment must be applied to situations where demand for isolation exceeds availability. Further information is needed on the consequences of isolation failure.