Since the early 1960's, highway engineers have become increasingly aware of fixed object hazards that border U.S. Highways, and more than 1500 roadside safety improvement projects have been undertaken. Although beneficial, such projects hardly represent a solution to the roadside hazard problem. The problem that remains is manifestly too large for treatment by inventory and analysis of individual obstacles. Roadside improvement programs are needed covering extensive segments of highways with focus on those areas most likely to experience encroachment by errant vehicles.

In view of the enormity of the task of removing all unsafe highway conditions, highway engineers, faced with limited funds, should select those sections for safety improvements which will yield the greatest benefits. This implies a need for a sound, rational basis for the selection of the most beneficial roadway safety projects. In recognition of this need, a study was undertaken at Georgia Institute of Technology in 1974 with the goal of developing priorities for the improvement of roadway geometric features and roadsides for safety. Results of that study were reported at the 20th Conference of AAAM. In that study, roadside surveys and inventories were made at 300 sites in Georgia of fatal fixed object crashes and at 300 comparison sites 1 mile away along the road the vehicle had travelled. Since the 1974 study was confined to fatal crashes, it left open the question of whether the priority strategy recommended would ameliorate a larger set of crashes, i.e., those of all levels of severity.

The research reported herein involved roadside analyses for 300 additional locations where fixed object crashes occurred. The crash samples were for all levels of severity and were randomly sampled from police files in three Georgia counties.

This latter study generally validated the findings of the fatal crash study and confirmed that roadside modification projects should be focused first on locations with curvature greater than six degrees and negative (downhill) gradient of two percent