Citation

Owen S, Wang S, Zhang P. 27th International Technical Conference on the Enhanced Safety of Vehicles (ESV); April 3-6, 2023; Abstract #: 23-0123, pp. 6p. Washington, DC USA: US National Highway Traffic Safety Administration, 2023 open access.

Copyright

(Copyright © 2023 open access, US National Highway Traffic Safety Administration)

Abstract

27th International Technical Conference on the Enhanced Safety of Vehicles (ESV): Enhanced and Equitable Vehicle Safety for All: Toward the Next 50 Years

https://www-esv.nhtsa.dot.gov/Proceedings/27/27ESV-000123.pdf

Injury Severity Prediction (ISP) models provide emergency responders with a rapid assessment of potential serious injury for occupants in vehicle accidents around the globe. ISP models predict the need for high level trauma care so that appropriate Emergency Medical Services (EMS) can be dispatched as quickly as possible to improve patient outcomes. In 2020, OnStar implemented its first occupant-based ISP models which predict outcomes for specific seating locations.[1] Models were developed and validated with NHTSA NASS CDS [2] and CISS [3] data. This paper seeks to assess model performance in the field using vehicle-based crash data and real-world occupant outcomes. This study leverages data from a sample of over 1,500 Michigan Advanced Automatic Crash Notification (AACN) events involving over 1,700 front row occupants to assess model performance. Vehicles include model years 2013 to 2020 and span several segments, including passenger cars, SUVs, and light trucks. AACN telemetry data and ISP-predicted outcomes are compared to actual Injury Severity Scores (ISS) for transported occupants. For non-transport cases, police reported injury severities (KABCO scores) are also examined. Measures of sensitivity, specificity, and likelihood ratio are calculated. False negative cases are used to understand model limitations. A range of threshold values used to assess "high" injury risk are also explored to highlight potential tradeoffs. Statistical analyses show that front row occupant models predict ISS 15+ injuries with high levels of accuracy. Metrics compare occupant-based model performance to prior vehicle-based ISP formulations. This study demonstrates that models based on government-sampled data sets are producing reliable results in the field.

Language: en