IHSDM: Crash Prediction Module

The crash prediction module estimates the frequency of crashes expected on a roadway based on its geometric design and traffic characteristics. The crash prediction algorithm considers the effect of a number of roadway variables: lane width, shoulder width and type, horizontal curve length and radius, presence of spiral transition, superelevation, grade, driveway density, passing lanes and short four-lane sections, two-way left-turn lanes, and roadside hazard rating. Intersection variables considered include skew angle, traffic control, presence of left- and right-turn lanes, and sight distance.

The algorithm for estimating crash frequency combines statistical base models and accident modification factors. FHWA derived the base models from negative binomial regression analyses of crash data from four States. Base models are available for roadway segments and for three types of intersections: three-legged intersections with stop control on the minor-road approach, four-legged intersections with stop control on the minor-road approaches, and four-legged signalized intersections.

The accident modification factors adjust the base model estimates for individual geometric design element dimensions and for traffic control features. The factors are the product of an expert panel review of related research findings and consensus on the best available estimates of quantitative safety effects of each design and traffic control feature. The algorithm can be calibrated by State or local agencies to reflect roadway, topographic, environmental, and crash-reporting conditions. The algorithm also provides an Empirical Bayes procedure for weighted averaging of the algorithm estimate with project-specific crash history data.

The crash prediction module can provide input for scoping improvement projects on existing roadways, comparing the relative safety performance of design alternatives, and assessing the safety cost-effectiveness of design decisions.

See IHSDM Library for References.