Analysis of first responder-involved traffic incidents by mining news reports

Abstract

Roadside service and incident response personnel face the risk of being killed or severely injured by passing vehicles when performing their duties on or along a road. This study investigated 5,113 responder-involved event news reports to understand the characteristics of first responder-involved incidents. Through text mining, this study examined and compared the characteristics of three types of responder-involved incidents: near-miss incidents, struck-by incidents, and line-of-duty-deaths (LODD). A higher proportion of struck-by and LODD incidents are associated with law enforcement agencies. In terms of the time of day, morning and night incidents are frequently reported in the news. Driving under the influence (DUI) or driving while intoxicated (DWI) is a major cause of LODD incidents. Compared to struck-by incidents, LODD incidents have a larger portion related to out-of-control vehicles. Further, this study built a logistic regression model to relate the incident characteristics to the odds of an incident being a LODD incident. The modeling result shows that tow truck drivers are associated with a greater likelihood of being involved in a news-reported LODD incident than other responders. LODD incidents are more likely to occur on early morning. Compare to entering/leaving/staying at the scene, responders are more likely to be involved in LODD event when assisting. The results offer insights into understanding the characteristics and possible reasons for first responder-involved incidents so that potential countermeasures could be developed to improve responder safety.

Keywords: Binary logistic regression; Line-of-duty-death; Roadside responder safety; Text-mining.