55 Development of a Novel Taxonomy for Adverse Events in the Emergency Department: Critical Review and Testing

Abstract

To 1) perform a critical review of existing taxonomies that describe harm; 2) develop a usable event taxonomy for the emergency department; 3) perform a pilot-test of this taxonomy. Safety science argues that improvement efforts should focus on preventing and mitigating harm rather than on error, which is commonplace but infrequently leads to adverse events (AEs). Describing and categorizing AEs is important for quality and safety assessment and identifying areas for improvement. However, most taxonomies fail to describe harm experienced by patients (eg, hypoxia, hemorrhage), focusing instead on errors or using categorizations that are too broad to be useful (eg, “medication error”). A patient-centered, ED-specific common language and framework for describing AEs and near misses is needed to advance quality and safety in the acute care setting. We followed the PRISMA guidelines for critical reviews. Our search strategy included taxonomies from major patient safety organizations and ontology/taxonomy databases including: Linked Open Vocabularies, Basel Register of Thesauri, Ontologies & Classifications, Taxonomy Warehouse, BioPortal and SCOPUS 1823, using the concepts “patient safety,” “adverse event,” “adverse reaction,” and “taxonomy,” followed by a title and abstract review of the 50 most-cited manuscripts. Taxonomy Development and Testing: The taxonomy was empirically developed on ∼600 AEs identified in a companion project that used the NCC MERP Index to describe severity. Our iterative process balanced parsimony with specificity describing both harm and non-harm events. We collapsed rare events into higher level sub-categorizations and added modifiers to provide further description. Safety and quality experts at 5 medical centers trained on use of the taxonomy on a conference call. To estimate reliability, collaborators independently reviewed 25 AE scenarios. Inter-rater reliability was assessed using the Fleiss kappa statistic for multiple raters. The 50 most-cited articles (out of 165) and the taxonomies from 10 major patient safety organizations yielded 2 that included descriptions of harm: The Joint Commission’s Patient Safety Event Taxonomy and the taxonomy from Adventist Health System. We used the latter as starting point, defining 7 categories and 51 subcategories. We added up to 3 “modifiers” per event (from a total of 46 modifiers available) to better describe events. Twenty-one distinct category-subcategory dyads were represented in the 25 test scenarios. There was complete agreement between reviewers in 20/25 scenarios. Inter-rater reliability was excellent for both overall-category scoring (κ = 81%) as well as for dyad scoring (κ = 79%). We developed a taxonomy of adverse events and non-harm events for the ED modified from an existing framework. Use of the tool with minimal training yielded excellent performance with high interrater reliability. We will perform additional reliability testing to cover all possible dyads and once validated, plan to use this taxonomy in further work related to adverse event detection in the ED.