OBJECTIVE: To assess whether three proposed quality-of-care indicators (unplanned readmissions, hospital-acquired bacteraemia, and postoperative wound infection) can be accurately identified from State health department databases. DESIGN: Algorithms were applied to State health department databases to maximise the identification of individuals potentially positive for each indicator. Records of these patients were then examined to determine the percentage of cases that met the precise indicator definitions. SETTING: 10 public, acute-care hospitals from Victoria, South Australia and New South Wales. Data from the 1994-95 and 1995-96 financial years were collected. PARTICIPANTS: Individuals 18 years of age or older who were identified from State health department administrative databases as potentially meeting the indicator criteria. MAIN OUTCOME MEASURES: The proportion of screened cases that met the precise indicator definitions, and the elements of the indicator definitions which could not be extracted from the administrative databases. RESULTS: The proportions of cases confirmed by medical record review to be positive for the indicator events were 76.3% for unplanned readmissions within 28 days, 20% for hospital-acquired bacteraemia, 43.5% for wound infections after clean surgery, and 34.8% for wound infections after contaminated surgery. The clinical elements of each indicator definition were not easily extracted from the administrative databases. CONCLUSIONS: The three proposed clinical indicators could not be extracted from current State health department databases without an extensive process of secondary medical record review. If administrative databases are to be used for assessing quality of care, more systematic recording of data is needed.