Evaluating Restricted Crossing U-Turn (RCUT) Intersections on Reducing Rear-End and Angle Crashes

Abstract: Traditional intersections pose significant safety risks due to multiple major conflict points, leading to frequent crashes. Per the Federal Highway Administration (FHWA), about 2.5 million crashes occur at intersections in the U.S., representing ~23% of fatal crashes and 50% of serious injury crashes. Similarly, more than 30% of crashes in Kentucky occur at intersections, representing ~18% of fatal crashes. In response, transportation agencies are exploring innovative solutions, such as the restricted crossing U-Turns (RCUTs), which have been implemented across the U.S. and Kentucky to reduce the severity of crashes, especially angle crash types. The safety effectiveness of RCUTs still remains underexplored, particularly using the empirical Bayes (EB) method, which accounts for regression-to-the-mean and temporal variation in the crash years.

This study takes the initiative and evaluates the safety effectiveness of RCUTs in reducing rear-end and angle crash types in the state of Kentucky while using the EB before-and-after approach. The focus is explicitly on these two crash types, not only because they were the most prevalent at all locations considered, but also due to their strong association with severe and injury-related crashes at intersections. In this regard, a total of 19 RCUT sites were selected for the safety evaluation. Crash data were retrieved from the Kentucky Transportation Cabinet (KYTC) in each of the before-and-after periods. The influence area considered for analyzing the before-period included crashes within a 250-foot (radius) circular buffer zone from the centroid of the intersection. To collect the after-period crashes for RCUT intersections, 250-foot (radius) circular buffer from the center of the main intersection and 50-foot (radius) circular buffer from the center of each median U-turn was considered. This was undertaken to account for all crashes due to the RCUT intersection in the after period. Furthermore, additional effort was made to collect the operational and geometric attributes (e.g., annual average daily traffic “AADT”, shoulder width, and median width) manually at each RCUT locations for developing crash prediction models in the before period (i.e., before conversions from traditional intersection to RCUTs).

Crash data exhibited over-dispersion; hence, the negative binomial (NB) model was employed to develop the safety performance functions (SPFs) for the two crash types (i.e., rear-end and angle crashes). The model results showed that angle crashes were significantly associated with AADT on minor approach and major approach median width. Consequently, rear-end crashes were significantly associated with AADT on minor approach, major approach posted speed limit, and major approach median width. To evaluate the safety effectiveness of RCUTs, the EB method was employed, which suggested that the conversion of traditional intersections into RCUTs showed a significant angle crash reduction of 33.8% at the 5% significance level, as well as a significant rear-end crash reduction of 75% at the 5% significance level.