Existing methods for surveillance of patient harm in the ED setting are inadequate, without any meaningful change in decades. Trigger tools, popularized by the Institute for Healthcare Improvement’s Global Trigger Tool, have been developed for multiple clinical areas and are used across the world, outperform traditional approaches for surveillance of adverse events. These tools use a two-tiered review process where a nurse screens records for triggers (predefined findings that make the presence of an AE more likely) and reviews records with triggers for AEs, discarding those without triggers. We developed a consensus-based ED trigger tool (EDTT) using a multicenter, transdisciplinary modified Delphi approach, subsequently pilot testing this in a multicenter fashion with encouraging results. This was followed by a recently completed, AHRQ-funded single center study to automate, refine and validate this tool. This study demonstrated that the EDTT is a high-yield and efficient instrument for identifying adverse events in the ED. The present study will evaluate the refined, automated EDTT), in a multicenter study. We will evaluate the EDTT’s generalizability and robustness at three sites with large emergency departments, with a planned in-depth review of 9,000 ED admissions. We will use natural language processing of electronic medical record narratives and machine learning to improve the EDTT efficiency in trigger detection and AE discovery. We will establish the basis for a wider use and prepare for scalability and usability of the tool, creating standardized, streamlined and free online training materials, and by evaluating the tool in a real-world manner consistent with intended use.