A "base accident rate", also known as "expected value", allows safety engineers or planners to objectively determine whether an accident pattern at a study location is significantly higher than the same accident pattern at other locations with similar geometric, traffic, and environmental factors. This study was conducted to develop base accident rates for rural highways at non-intersections in Ohio using all the available data from the Ohio Department of Transportation's database. Using a random sampling technique, 30% of the data for each of the 12 districts was extracted and a comprehensive database was created for each district. Then, the highway sections were generally divided into uniform segments of length 0.25 mile. For each highway segment, population density data within one-mile radius was generated. Additional data, namely, number of residential and business driveways, number of passing zones, horizontal and vertical curves, and guardrail length were manually recorded using photolog discs. A master database was created for each district using the above-mentioned data. The base accident rates were developed for ten accident types, namely, (i) Total accidents, (ii) Injury accidents, (iii) Property Damage Only accidents, (iv) Fixed-Object accidents, (v) Sideswipe accidents, (vi) Rear-End accidents, (vii) Left-Turn accidents, (viii) Right-Angle accidents, (ix) Wet Pavement accidents, and (x) Night accidents. Fourteen independent variables were used in the development of the base rates. Automatic Interaction Detection (AID) technique was used to evaluate the complex interaction between the dependent and independent variables. The AID technique assisted in identifying a set of independent variables that have significant relationship with each dependent variable. Based on the results of this study, it is recommended that accident density (and not accident rate) be considered as the primary unit for accident analysis for non-intersections. It is recommended that the means and standard deviations for each accident type as provided in the report be used as a method for calculating base accident densities/rates for the individual districts as well as for the entire State. It is also recommended that the regression equations developed for the individual districts and the whole State be used as an additional tool for estimating base accident densities/rates for each accident type on rural non-intersections in Ohio. An additional work completed in this project was the new and upgraded version of the knowledge-based expert system called RITA (Rural Highway Safety Advisor) that codifies within its knowledge base the procedures, techniques and rules of thumb (heuristic) involved in the process of conducting highway safety analysis for non-intersections, signalized intersections and non-signalized intersections that commonly exist in the rural highways in Ohio. The results of the base accident rate study and RITA are amenable for immediate implementation. (Author/publisher)
Abstract