An adverse event identification strategy based on physician self-referral uncovers as many adverse events as does a record review and is less costly. In addition, physician-identified events are more likely to be preventable and, thus, are targets for quality improvement.

No abstract

No abstract

Study Objective: To evaluate the potential ability of computerized information systems (KS) to identify and prevent adverse events in medical patients.Design: Clinical descriptions of all 133 adverse events identified through chart review for a cohort of 3,138 medical patients were evaluated by two reviewers. Measurements:For each adverse event, three hierarchical levels of IS sophistication were considered: Level l-demographics, results for all diagnostic tests, and current medications would be available on-line; Level 2-all orders would be entered on-line by physicians; and Level 3-additional clinical data, such as automated problem lists, would be available on-line. Potential for event identificationand potential for event prevention were scored by each reviewer according to two distinct sets of event monitors.Results: Of all the adverse events, 53% were judged identifiable using Level 1 information, 58% were judged identifiable using Level 2 information, and 89% were judged identifiable using Level 3 information.The highest-yield event monitors for identifying adverse events were "panic" laboratory results, unexpected transfer to an intensive care unit, and hospital-incurred trauma. With information from Levels 1, 2, and 3, 5%, 13%, and 23% of the adverse events, respectively, were judged preventable. For preventing these adverse events, guided-dose algorithms, drug-laboratory checks, and drug-patient characteristic checks held the most potential.