This study investigated the effect of education and audit on the identification and management of medical inpatients with excessive alcohol consumptions by use of a cross-sectional survey of medical and nursing notes, compared with historical controls. We report the proportion of medical and nursing notes containing qualitative and quantitative drinking histories, the use of the CAGE questionnaire, and notification of alcohol consumption in discharge summaries. A total of 1,979 out of 2,680 (74%) notes of eligible patients were examined. The number of medical notes containing quantitative alcohol histories rose from 175/792 (22%) among historical controls, to 242/690 (35%) after training of junior medical and nursing staff, and to 241/497 (48%) after feedback of results in a clinical audit meeting. The number of medical notes with no recorded alcohol history changed from 309/792 (39%) to 287/690 (42%) to 152/497 (31%), respectively (chi 2 for trend = 97.2, p < 0.0001). The number of nursing notes containing quantified alcohol histories rose from 31/792 (4%) among historical controls, to 123/690 (18%) after training, and to 237/497 (48%) after feedback of results in a clinical audit meeting. The number of nursing notes with no recorded alcohol history fell from 309/792 (90%), to 512/690 (74%), to 205/497 (41%) respectively (chi 2 for trend = 353.7, p < 0.0001). The proportion of patients found to be drinking excessively rose from 40/792 (5.1%) to 57/690 (8.3%) to 45/497 (9.1%) (chi 2 for trend = 8.25, p = 0.004). There was little worthwhile improvement in use of the CAGE questionnaire and in discharge summaries. Education and audit are effective ways of improving the identification of excessive drinkers by both junior doctors and nurses alike. Other aspects of management are less responsive.