OBJECTIVE: To assess the completeness of hospital discharge summaries and the efficiency of the discharge summary system in two urban teaching hospitals. DESIGN: Descriptive study, with follow-up telephone survey. SETTING: General internal medicine services at two urban tertiary care hospitals affiliated with the University of Ottawa. PATIENTS: A total of 135 patient charts, representing 10% of the patients discharged from the services between Aug. 1 and Dec. 31, 1993. Three charts were unavailable for review, and 26 were excluded because of patient death, early patient discharge (within 48 hours after admission) or lack of discharge summary; this left 106 summaries for analysis of completeness and 114 (including the charts without a summary) for analysis of efficiency. OUTCOME MEASURES: Completeness: proportion of summaries in which the following information was reported: admission diagnosis, drug allergies, physical examination, significant laboratory tests and results, discharge diagnosis, discharge medications and medical follow-up. Efficiency: time taken to generate the discharge summary and whether the patient's family physician received it. RESULTS: Of the 106 charts with a discharge summary, information was available from the dictation system database for all but one (99.1% complete). Information was missing on the admission diagnosis in 34.0% (36/106) of the summaries, the discharge diagnosis in 25.5% (27/106) and the discharge medications in 22.8% (23/101). Of the 268 significant laboratory tests and results noted in the charts 115 (42.9%) were not reported in the discharge summary. Of the 94 discharge summaries in charts with the patient's family physician listed on the facesheet, 38 (40.4%) were not received by the family physician. CONCLUSIONS: Considerable deficiencies in the completeness of the discharge summaries and the efficiency of the discharge summary system were found in the participating hospitals. Replication of this study in other settings is indicated, and strategies to improve the process should be pursued.