Identifying Roadway Sections With \"Critical\" Large Truck Accident Rates

A considerable amount of transportation-related data exists at the national and state levels which may be of assistance In the evaluation and modification of routing and risk assessment models currently utilized by the Department of Energy's Transportation Technology Center at Sandia National Laboratories. This Information, Including the Federally-supported Fatal Accident Reporting System· FAAS (U.S. Department of Transportation 1984), National Accident Sampling System· NASS (U.S. Department of Transportation 1987), and Highway Performance Monitoring System· HPMS (Kittell1984), can be used to evaluate both existing and proposed model appications as well as to formulate strategies for future model and associated data base Improvements. Current efforts at the national level appear to be focusing on attempts to create a National Safety Information System. This system would merge existing information relating to highway accident, highway inventory (such as geometric data), and traffic volume data which is currently maintained in separate computer files. Any model applications involving accident rates would thus have to access this combined file since such calculations require roadway section length and volume Information as well as data on accident occurrences. State traffic record systems also appear to be moving toward this consolidated approach, although routine use of such special-purpose files appears to be a few years away (Wolff 1989). Existing truck accident analyses at the state-wide level, then, must rely on highway accident and roadway inventory data bases residing in separate, often incompatible, computer files. The purpose of this paper is to present an overview of efforts currently underway to analyze roadway geometric data and accident rate information from a variety of sources to develop a relationship between these data and accident probability indicators for specific roadway segments, and to develop a computerized methodology to determine parameters of interest. This methodology will then be included on the TRANSNET system of models and data bases where the accident probability value may be used to perform analyses in codes such as StateGEN/StateNET or RAOTRAN. A technique for merging accident record and roadway Inventory files and for utilizing the combined data set to Identify sections of a rural Interstate highway system that have unusually high accident rates involving large truck vehicles is presented here as a prototype of the accident probability methodology. This statistical technique, known as the rate/quality control method, Involves calculating a •critical rate• for each roadway section. This critical rate is a function of the traffic volume on the section, the overall systemwide accident rate, and a desired level of statistical significance. Observed accident rates on each section are then compared to the section's critical rate and sections are ranked according to their \"criticality,\" that is, the numerical difference between their observed and \"critical\" rates. With ongoing data collection efforts and concomitant improvements in data quality, particularly involving large truck accident rates. the technique may ultimately have application to the formulation of routing decisions regarding the transport of radioactive or other hazardous materials and in the evaluation of routing alternatives. Subsequent sections of the paper describe in detail the rate/quality control technique and its application to a state-wide, rural Interstate highway network, develop the computer code for merging the accident and inventory files, and describe the selection of critical roadway sections. Conclusions and recommendations for further work follow.