Since its inception, the National Highway Traffic Safety Administration (NHTSA) has been concerned with providing the most complete and technologically feasible crash data collection. The collaboration with the Federal Highway Administration (FHWA) also dates back to the inception of the data sets. Funding issues and interest of primary users have limited coded infrastructure variables and attributes. In 2005, NHTSA embarked upon the congressionally-mandated National Motor Vehicle Crash Causation Study (NMVCCS) data collection. With on-scene reporting, nearly crashtime graphic data became available to end-users. In 2008, the Fatality Analysis Reporting System (FARS) published the first geographical coordinates for its cases. This eventually resulted in the rerelease of data from 2001 through 2007. Although not temporally compatible, those interested in infrastructure and relevant elements would be able to complement the coded variables and attributes. The improved graphic reporting was noted in the National Automotive Sampling System (NASS) Crashworthiness Data System (CDS) pursuant to 2007 and potentially drawing from the NMVCCS model.

This paper offers an approach to mine, previously unconsulted NMVCCS data, rooted in precedent and established using FARS and NASS CDS. Using new data sources, the safety community might yield additional insights about crashes and the influence of various factors. In the narrowest sense, the findings might support the knowledge derived from crash testing and the limited extent of in-service evaluations of roadside safety elements that have been undertaken to date. As a natural by-product, this paper suggests that aggregated knowledge might populate an infrastructure dataset to aid those involved in roadway design, especially those addressing roadway departure issues, as supported by the overwhelming FARS incidence. During the feasibility study to identify the roadway elements and the value of image review, the digital image information has been enlightening. Tangentially, unlike NHTSA, FHWA may reference the unweighted data sets, as this furthers understanding of crash causation rather than underpinning rulemaking activities, thereby maximizing the use of unweighted NMVCCS data, predating the sampling plan. In the past, FWHA has consulted state-reported roadway features inventories and their resulting crashes when possible, aspiring to a macro view of roadside element description. As inconsistencies exist in the way that data has been collected, stored, and eventually processed at the state-level, this study seeks to review untapped digital images from national crash reporting, filling a void present in roadway design using a micro approach of roadside element description based upon crash scene locations. The present study seeks to address highway safety data needs by leveraging new data resources and tools.