Results of First Field Test of Telemetry Based Injury Severity Prediction

Identification of severely injured occupants is of utmost urgency following a crash event. Advanced automatic collision notification (AACN) has great potential to improve post-crash care if the risk of severe injury to a vehicle’s occupants can be accurately predicted. The National Expert Panel for Field Triage set a 20% risk of Injury Severity Score (ISS) 15+ injury [1] as the threshold for urgent transport to a trauma center. The objective of this study was to field test real world performance of the published injury severity prediction (ISP) algorithm in collisions involving recent model GM vehicles equipped with OnStar. This study was approved by the Institutional Review Board (IRB) of the Michigan Department of Community Health. There were 924 occupants in 836 crash events, involving vehicles equipped with AACN capabilities, in the state of Michigan which were identified from the OnStar records. The police crash report corresponding to the event was identified in the State of Michigan database and used to confirm data sent by telemetry from the vehicle. The injury status of all occupants in the case vehicles was determined. Occupants not transported for medical evaluation were assumed to have ISS<15. For occupants transported from the scene for evaluation and treatment, medical records and imaging data were obtained from the treating facility. Case reviews were conducted to jointly analyze crash, vehicle telemetry, and injury outcome data. The algorithm was used to calculate the predicted risk of injury based on transmitted telemetry data and this prediction was compared to the observed injury outcome for each vehicle as well as each occupant. In this field study, the ISP algorithm’s ability to predict whether a vehicle had a seriously injured (ISS>15) occupant was, in terms of sensitivity, at 63.64% compared to the model sensitivity of 39.6% and it also came very close to expectations of specificity at 96.12% compared to the model specificity of 98.3% with use of age and gender data. Without use of age and gender, for ISP calculation, the sensitivity performance was 45.45% while the specificity improved slightly to 97.58%. Detailed analysis of cases suggests that further performance gains could be obtained with more detailed definition of crash direction, seating position, and occupant age. There were 184 candidate crash occupants in 167 vehicles not included in the study analysis due to: A) missing Police accident reports, n=77 in 75 crashes; B) inability to retrieve medical records, n = 71 in 61 crashes; or C) rollover event, n=36 in 31 crashes. Analysis of these excluded cases did not reveal any bias in crash severity or injury that would confound the current study findings. This study confirms for the first time under real-world field conditions that occupant injury severity can be predicted using vehicle telemetry data. The ISP algorithm’s ability to predict a 20% or greater risk of severe (ISS15+) injury was better than anticipated and confirms ISP’s utility for the field triage of crash subjects. This analysis suggests that AACN technology can greatly facilitate the collection of field data with ISP also serving as a baseline for potential monitoring of the benefits resulting from vehicle safety design changes.

Language

  • English

Media Info

  • Media Type: Web
  • Features: Appendices; References; Tables;
  • Pagination: 12p
  • Monograph Title: 24th International Technical Conference on the Enhanced Safety of Vehicles (ESV): Traffic Safety Through Integrated Technologies

Subject/Index Terms

Filing Info

  • Accession Number: 01567536
  • Record Type: Publication
  • Report/Paper Numbers: 15-0388
  • Files: TRIS, ATRI, USDOT
  • Created Date: Jun 23 2015 5:49PM