Over 35,000 mi (56,315 km) of arterial highways in the United States are multi-lane, non-Interstate roads in rural areas. Fatality rates on rural federal-aid primary highways have been significantly higher compared to the fatality rates for urban and rural Interstate highways and urban primary highways. Unfortunately, very little is known concerning the effects of geometric design elements on the safety for rural, multi-lane, non-freeway highways since little past research has concentrated on these roads. This paper presents a study of the effects of the various cross-section related design elements on the frequency of accidents for rural, multi-lane, non-freeway roads. Data extracted from the Highway Safety Information System (HSIS) for four states were utilized for data exploration and descriptive analysis. Minnesota data were used for a statistical modeling due to the availability of accident, traffic, roadway inventory and supplemental inventory data for the selected data elements. Supplemental roadway variables which were needed included roadside condition and intersection/driveway access points. To collect those supplemental data elements, an advanced Photolog Laser Videodisc (PLV) data recording system was developed and applied for the study. These data were integrated into the HSIS database for the modeling analysis. The objective of the statistical modeling analysis was to identify cross-section related variables that were statistically associated with the occurrence of accidents on selected roadway segments and to estimate model parameters. A Poisson regression model was used to model the relationship between expected accident frequency and various roadway and traffic variables.

The study results establish a quantitative relationship between accident frequency and various cross-section related roadway design elements on rural, multi-lane, non-freeway highways.