Multivariate analyses for infrastructure-based crash-prediction models for rural highways

Crash-prediction models can be used to assess road safety during highway planning and design. The main objective of this study is to develop an Infrastructure coefficient that reflects the overall safety level of a highway and can be used as an independent variable in a crash-prediction model. Infrastructure is defined as the highway and its geometric features. It includes the road alignment, roadside environment, sight distance along the highway, presence of guardrails, number of access points, roadway consistency alignment, lane and shoulder width and percentage of access points with a speed-change lane. These geometric features measure the overall quality of the highway. Two different infrastructure coefficients are developed and calibrated by two different statistical methods. The infrastructure coefficient developed by using the principal component analysis method consists of 11 infrastructure characteristics, and that developed by using the analytic hierarchy process method consists of 5 infrastructure characteristics. For each highway section, a value reflecting its infrastructure quality was calculated according to each of the infrastructure coefficients developed. The results showed a significant correlation between highway infrastructure quality and crash rates. Based on the infrastructure coefficients and crash records, two crash-prediction models are developed. It is suggested that these models can be used to evaluate the safety level of existing or planned highways.