The ability of a hospital's adverse drug reaction (ADR) database to identify common and repeated patterns of preventable adverse drug events (ADEs) was analyzed. ADR reports collected from 1994 through 2000 were extracted from a teaching hospital's ADR database. Reports were assessed concurrently in accordance with seven previously published explicit criteria for preventability. Only cases considered clinically significant were included in this analysis. Events that occurred in the ambulatory care setting were excluded. Preventable ADEs were categorized by drug or drug class, type of medication error, and the subsequent adverse outcome. Novel in this analysis was the linking of these three descriptors. Of the 2571 ADR reports assessed, 415 ADEs were deemed preventable. Of the preventable ADEs, 98 were not analyzed because they occurred in the ambulatory care setting, leaving 317 preventable ADEs in 275 inpatients (mean age +/- S.D., 48.5 +/- 23.9 years) for analysis. Although 93 drugs were associated with these ADEs, only 10 drugs accounted for more than 60% of the events. Analysis and categorization by type of error and outcome suggested that three high-priority preventable ADEs accounted for 50% of all reports: (1) overdoses of anticoagulants or insufficient monitoring and adjustments (according to laboratory test values) were associated with hemorrhagic events, (2) overdosing or failure to adjust for drug-drug interactions of opiate agonists was associated with somnolence and respiratory depression, and (3) inappropriate dosing or insufficient monitoring of insulins was associated with hypoglycemia. Analysis of a hospital ADR database identified prevalent and preventable clinically significant ADEs.