In this prospective cohort of 71 elderly patients undergoing cardiac surgery, each subject was interviewed before and after surgery to detect incident delirium using the Confusion Assessment Method (CAM), the Mini-Mental State Examination (MMSE), the Clock Test, and a health record review. The first 41 were assessed by a physician and the remaining 30 by two study nurses. Delirium was then diagnosed by a physician using DSM-III-R criteria. Delirium was present in 23 subjects (32.4%). The sensitivity of the CAM differed significantly when administered by physicians compared to nurses (1.00 vs. .13). When standard cutoffs were used, neither the MMSE nor the Clock Test were found to be sensitive markers for delirium (.30 and .09, respectively). Recognition of delirium by charting was superior in nurses compared to physicians (.83 vs. .30). We conclude that the sensitivity of markers for delirium, such as the CAM and health record documentation, is dependent on the training background of the operator.