Improved Prediction Models for Crash Types and Crash Severities

An important component of the AASHTO Highway Safety Manual (HSM) is the ability to estimate the safety performance of highways and the effects of proposed countermeasures. The HSM utilizes safety performance functions (SPFs) to estimate the number of crashes over a specific roadway over a specific time period. The move toward scientific approaches in safety analysis has driven the need for further development of SPFs so that they provide more detailed estimates by crash type and crash severity. These aspects are a natural and important step in the progression of crash analyses that is currently focused primarily on frequency reduction. It is envisioned that the SPFs and crash distributions will be refined to determine the expected severity and crash types of various road facilities. Planners and designers can then better target and select countermeasures to address these particular aspects resulting in improved project selection. Having a wide range of safety evaluation tools that facilitate a comparative analysis of crash severity and crash type will lead to potential systematic and system wide improvement scenarios. Currently, the models in HSM Part C differ by chapter. In Chapter 10 (rural two-lane highways), predictive models provide estimates for total crashes and then values are allocated to crash types and severity levels based on tabulated proportions. In Chapter 11 (rural multilane highways), separate predictive models are provided for total crashes, fatal-and-injury (KABC) crashes, and fatal-and-injury (KAB) crashes, with crash frequencies for property-damage-only (PDO) crashes determined by subtracting the KABC crash frequency from the total crash frequency. Crashes for any given crash severity level are then allocated to crash types based on tabulated proportions. In Chapter 12 (urban and suburban arterials), separate predictive models are provided for combinations of three crash severity levels (total, fatal-and-injury, and PDO) and five crash types (multiple-vehicle non-driveway crashes, single-vehicle crashes, multiple-vehicle driveway-related crashes, vehicle-pedestrian crashes, and vehicle-bicycle crashes). Within two of the crash types (multiple-vehicle non-driveway crashes and single-vehicle crashes), predicted crash frequencies can be broken down into even more specific crash frequencies by tabulated proportions.To provide more consistency for users, it is desirable that a HSM second edition provide a consistent approach to predictive modeling by crash severity and crash type. This would require that models from past research be refit using the databases that were used to develop those models or for new models to be developed from new databases. Certainly, users should expect all future HSM chapters to use a consistent set of crash severity and crash type. Candidate crash severity levels from which final choices should be made include total crashes (all severity levels combined) and K, A, B, C, PDO, KA, KAB, KABC, AB, and ABC crashes. There is a need to model individual crash types. User needs and sample size issues should be addressed in this modeling effort. In addition, there is a need to account for the variations of specific crash types by facility type. In other words, it may not be desirable to require modeling of a common set of crash types across all facility types, because some crash types are common (and, thus, easy to model) on some facility types and relatively uncommon (and thus difficult to model) on other facility types. A consistent approach, with reasonable variations by facility type, is needed for the HSM second edition. Once a consistent and practical set of categories for crash severity and crash type have been defined, these prediction models should be developed and implemented throughout HSM for future editions. The objectives of this research are to develop: (1) Crash severity and crash type SPFs or distributions or both that can be used in the estimation of the crash type and crash severity likely on the facility types contained or intended for use in the HSM; (2) Recommendations of how the research results can be incorporated into the HSM and associated tools, including the development of associated chapters or chapter content in AASHTO standard format for the HSM second edition and recommended procedures for consistent use of crash severity and crash type SPFs or distributions or both; and (3) A description of the statistical and practical advantages and disadvantages of the methodology developed in the research and potential barriers to implementation. This research should provide a consistent approach for developing and validating crash severity and crash type prediction models to improve the capabilities of the current HSM and associated tools to estimate the safety performance outcomes associated with modifications to the highway and road user environments.


  • English


  • Status: Active
  • Funding: $800000
  • Contract Numbers:

    Project 17-62

  • Sponsor Organizations:

    Federal Highway Administration

    1200 New Jersey Avenue, SE
    Washington, DC  United States  20590

    American Association of State Highway & Transportation Officials (AASHTO)

    444 North Capitol Street, NW, Suite 225
    Washington, DC  United States  20001

    National Cooperative Highway Research Program

    Transportation Research Board
    500 Fifth Street, NW
    Washington, DC  United States  20001
  • Project Managers:

    Bush, Mark

  • Performing Organizations:

    University of Connecticut, Storrs

    Storrs, CT  United States  06268-5202
  • Principal Investigators:

    Ivan, John

  • Start Date: 20130702
  • Expected Completion Date: 20171231
  • Actual Completion Date: 0
  • Source Data: RiP Project 37828

Subject/Index Terms

Filing Info

  • Accession Number: 01543916
  • Record Type: Research project
  • Source Agency: Transportation Research Board
  • Contract Numbers: Project 17-62
  • Files: TRB, RiP
  • Created Date: Nov 23 2014 1:01AM