Driving under the influence (DUI) is illegal in the United States because a driver's mental and motor skills can be seriously impaired by alcohol or drugs. Consequently, DUI violators' involvement in severe crashes is high. Motivated by the spatial and temporal nature of traffic crashes, this study introduces an integrated spatio-temporal approach to analyzing highway safety data. Specifically, this study estimates Geographically and Temporally Weighted Regression (GTWR) models to understand the consequences of DUI in crashes. GTWR can theoretically outperform traditional regression methods by accounting for unobserved heterogeneity that may be related to the location and time of a crash. Using Southeast Michigan crash data, this study finds that DUI is associated with a 25% higher likelihood of injury in a crash. The association between injury severity and DUI varies significantly across space and time. From the spatial aspect, DUI crashes in rural or small-town areas are more likely to cause injuries than urban crashes. From the temporal aspect, different times are associated with varying relationships between injury severity and DUI. If focusing on DUI crashes in late nights and early mornings, on Fridays, the entire northeast part from Clinton Charter Township to Port Huron is associated with severer injuries than other regions including Detroit's urban area and its south. On Mondays, the DUI crashes in the northwest are also more likely to cause severe injuries. The methodology introduced in this study takes advantage of modern computational tools and localized crash/inventory data. This method offers researchers and practitioners an opportunity to understand highway safety outcomes in great spatial and temporal details and customize safety countermeasures for specific locations and times such as saturation patrols.
Keywords: Crash severity; Driving under the influence; Geographically and temporally weighted regression; Spatio-temporal variation.
Copyright © 2020 Elsevier Ltd. All rights reserved.