The Gordon and Betty Moore Foundation is funding health services research projects to expand the base of rigorous evidence about pre-hospital diagnostic delays – i.e., those that occur before a patient reaches the care setting where their condition is diagnosed. The goal of this study is to improve Emergency Medical Services (EMS) care and patient outcomes through improvement programs informed by patient, disease, situational, and EMS factors associated with failure to recognize stroke in the pre-hospital setting. This study posits that there are two barriers hindering effective analysis of reasons for low detection of stroke amongst Emergency Medical Services providers: (1) fragmentation of documentation systems, and (2) prehospital data that does not allow one to determine potential reasons for misdiagnosis. This study seeks to understand whether narrative data in EMS notes may help address both gaps. The research team will analyze a statewide cohort of EMS transported strokes derived by linking the state stroke registry with the EMS information system. The study will use natural language processing (NLP) techniques to identify factors contributing to diagnostic error and develop a comprehensive structural equation model of EMS nonrecognition of stroke. This will enable researchers to quantify the contribution of these factors to delays in diagnosis and disparities in care. This information will be used to develop educational programs, feedback processes, and protocol adjustments for dissemination across a network of EMS partners. Deliverables will include a project work plan, narrative and financial reports, and a range of products to reach clinicians, policymakers, and other audiences for study findings.