Washington State Post-Crash EMS Data Science Initiative - ABSTRACT Motor vehicle injuries (MVI) have been a leading cause of death in the United States since the 1930s. Fatalities had trended downwards from the 1970s until 2019; however, these trends appear to have reversed recently. Forty percent of crash decedents were alive when emergency medical services (EMS) providers arrived at the crash scene. Improved post-crash care may prevent MVI fatalities. The National EMS Information System (NEMSIS) contains EMS information from over 14,000 EMS agencies nationally. NEMSIS includes case definitions for EMS scenarios including crashes. However, NEMSIS lacks the level of detail available in EMS data systems and does not include information from other data sources such as crash records. Linking EMS to crash records and health outcomes can provide better alignment of MVI classification across all data sources. Washington State has linked the Washington EMS Information System (WEMSIS), which includes 95% of state EMS calls between 2020 and 2024, to crash reports records as part of the Traffic Records Integration Program (TRIP) program. In this project, we will complete TRIP’s integration of Washington’s Rapid Health Integration NetwOrk (RHINO), a syndromic surveillance system containing ED, urgent care and primary care medical encounter data, and Washington Comprehensive Hospital Abstract Reporting System (CHARS), which collects hospital discharge data. We will then develop a technical report explaining the processes of building TRIP, including both a quantification of linkage failure rate and potential for selection biases and a qualitative guide to data governance discussions. We will also explore the governance steps necessary to submit a de-identified limited data set of the TRIP repository to an NIH data repository. Next, we will conduct several data science-driven analyses. We will use the TRIP to identify crashes, then compute sensitivity and positive predictive value of the current NEMSIS case definition. We will use data science, including machine learning variable importance metrics, to identify elements that can improve NEMSIS crash case definitions. We will also use the same linked data and data science approaches to explore predictors of 30-day mortality to improve NEMSIS’s injury severity scoring algorithm. Finally, we will leverage outcome information from linked ED and vital statistics data to explore associations between time to care, transport time, care facility, and crash victim mortality. Specifically, carefully considering Washington’s trauma triage protocol, we will estimate the impact of EMS response time (scene response time, scene time, and transport time) and ED facility trauma designation status (trauma level I through V or non-trauma center), on 30-day post-crash mortality.
