Development And Application Of Crash Severity Models For Highway Safety

The first edition of the Highway Safety Manual has provided methods and procedures for estimating total crashes, crashes by type, and crashes by severity at the site level, project level and corridor level. Crash prediction models are critical in the entire safety management system recommended by HSM, including network screening, economic analysis, project prioritization, and safety effectiveness evaluation. NCHRP Web-Only Document 351: Development and Application of Crash Severity Models for Highway Safety: Conduct of Research Report, from TRB's National Cooperative Highway Research Program, is supplemental to NCHRP Research Report 1047: Development and Application of Crash Severity Models for Highway Safety: User Guidelines. The document seeks to identify gaps and opportunities in the current severity prediction/estimation procedures within the HSM, to develop and validate new severity models to address the gaps and opportunities, and to develop a guidance document that includes protocols for the use and application of severity-based models in a format suitable for possible adoption in the HSM.

"This report presents guidelines on evaluating crash severity estimation models for use in different site conditions. The guidelines will be of interest to state departments of transportation (DOTs) seeking more informed model application, broader acceptance of model results, and, ultimately, improved safety decision making. The guidelines could also be applied to existing crash prediction models and serve to improve pertinent models and model elements in the Highway Safety Manual (HSM) and its associated tools." -- publisher's website

"The Highway Safety Manual (HSM) is a resource that provides safety knowledge and tools in a useful form to facilitate improved decision making based on safety performance. The focus of the HSM is to provide quantitative information for decision making. The HSM assembles currently available information and methodologies on measuring, estimating and evaluating roadways in terms of crash frequency (number of crashes per year) and crash severity (level of injuries due to crashes). The HSM presents tools and methodologies for consideration of 'safety' across the range of highway activities: planning, programming, project development, construction, operations, and maintenance. The purpose of this is to convey present knowledge regarding highway safety information for use by a broad array of transportation professionals"--p. xxiii, vol. 1.

Highway Safety Analytics and Modeling comprehensively covers the key elements needed to make effective transportation engineering and policy decisions based on highway safety data analysis in a single. reference. The book includes all aspects of the decision-making process, from collecting and assembling data to developing models and evaluating analysis results. It discusses the challenges of working with crash and naturalistic data, identifies problems and proposes well-researched methods to solve them. Finally, the book examines the nuances associated with safety data analysis and shows how to best use the information to develop countermeasures, policies, and programs to reduce the frequency and severity of traffic crashes. Complements the Highway Safety Manual by the American Association of State Highway and Transportation Officials Provides examples and case studies for most models and methods Includes learning aids such as online data, examples and solutions to problems

Modeling crash severity is an important component of reasoning about the issues that may affect highway safety. A better understanding of the factors underlying crash severity can be used to reduce the degree of crash severity injury, locate road hazardous sites, and adopt suitable countermeasures. In order to provide insights on the mechanism and behavior of the crash severity injury, a variety of statistical approaches have been utilized to model the relationship between crash severity and potential risk factors. Many of the traditional approaches for analyzing crash severity are limited in that they are based on the assumption that all observations are independent of each other. However, given the reality of vehicle movement in networked systems, the assumption of independence of crash incidence is not likely valid. For instance, spatial and temporal autocorrelations are important sources of dependency among observations that may bias estimates if not considered in the modeling process. Moreover, there are other aspects of vehicular travel that may influence crash severity that have not been explored in traditional analysis approaches. One such aspect is the roadway visibility that is available to a driver at a given time that can impact their ability to react to changing traffic conditions, a characteristics known as sight distance. Accounting for characteristics such as sight distance in crash severity modeling involve moving beyond statistical analysis and modeling the complex geospatial relationships between the driver and the surrounding landscape. To address these limitations of traditional approaches to crash severity modeling, this dissertation first details a framework for detecting temporal and spatial autocorrelation in crash data. An approach for evaluating the sight distance available to drivers along roadways is then proposed. Finally, a crash severity model is developed based upon a multinomial logistic regression approach that incorporates the available sight distance and spatial autocorrelation as potential risk factors, in addition to a wide range of other factors related to road geometry, traffic volume, driver's behavior, environment, and vehicles. To demonstrate the characteristics of the proposed model, an analysis of vehicular crashes (years 2013-2015) along the I-70 corridor in the state of Missouri (MO) and on roadways in Boone County MO is conducted. To assess existing stopping sight distance and decision sight distance on multilane highways, a geographic information system (GIS)-based viewshed analysis is developed to identify the locations that do not conform to AASHTO (2011) criteria regarding stopping and decision sight distances, which could then be used as potential risk factors in crash prediction. Moreover, this method provides a new technique for estimating passing sight distance along two-lane highways, and locating the passing zones and no-passing zones. In order to detect the existence of temporal autocorrelation and whether it's significant in crash data, this dissertation employs the Durbin-Watson (DW) test, the Breusch-Godfrey (LM) test, and the Ljung-Box Q (LBQ) test, and then describes the removal of any significant amount of temporal autocorrelation from crash data using the differencing procedure, and the Cochrane-Orcutt method. To assess whether vehicle crashes are spatially clustered, dispersed, or random, the Moran's I and Getis-Ord Gi* statistics are used as measures of spatial autocorrelation among vehicle incidents. To incorporate spatial autocorrelation in crash severity modeling, the use of the Gi* statistic as a potential risk factor is also explored. The results provide firm evidence on the importance of accounting for spatial and temporal autocorrelation, and sight distance in modeling traffic crash data.

In recent years, significant effort and money have been invested to enhance highway safety. As available funds decrease, the allocation of resources for safety improvement projects must yield the maximum possible return on investment. Identifying highway locations that have the highest potential for crash reduction with the implementation of effective safety countermeasures is therefore an important first step in achieving the maximum return on safety investment. This study was undertaken to develop safety performance functions (SPFs) for use in Virginia in conjunction with SafetyAnalyst, a computerized analytical tool that can be used for prioritizing safety projects. A safety performance function is a mathematical relationship (model) between frequency of crashes by severity and the most significant causal factors of crashes for a specific type of road. Although the SafetyAnalyst User's Manual recommends four SPFs for two-lane segments, these SPFs were developed using data from Ohio. Because the transferability of these SPFs to other states could not be guaranteed by the developers of the four recommended SPFs, it is necessary to calibrate or develop valid SPFs for each state using appropriate data from the state. In this study, annual average daily traffic (AADT) was used as the most significant causal factor for crashes, emulating the SPFs currently suggested by Safety Analyst. SPFs for two-lane roads in Virginia were developed for total crashes and combined fatal plus injury crashes through generalized linear modeling using a negative binomial distribution for the crashes. Models were developed for urban and rural areas separately, and in order to account for the different topographies in Virginia, SPFs were also separately developed for three regions in Virginia. A total of 139,635 sites were identified for use in this study. Each site is a segment of a rural or urban two-lane road without an intersection for which AADT data were available for the years 2003 through 2007 inclusive and no change in facility type had occurred over that period. A comparative analysis based on the Freeman-Tukey R2 coefficient was then conducted between the relevant Ohio SPFs suggested for use in the SafetyAnalyst User's Manual and those specifically developed in this study for Virginia to determine which set of models better fit the Virginia data. In general, the results indicated that the SPFs specifically developed for Virginia fit the Virginia data better. The final step in this methodology was to illustrate the value of SPFs developed through an analysis of sample sites and the need of the sites for safety improvement based on SPFs as compared to crash rates. The results indicated that prioritization using the empirical Bayes method that incorporates the SPFs resulted in a higher potential for reduction in crashes than did prioritization using crash rates. The effective use of SafetyAnalyst will facilitate the identification of sites with a high potential for safety improvement, which, in turn, with the implementation of appropriate safety improvements, will result in a considerable reduction in crashes and their severity.

Transportation Research Record contains the following papers: Incorporating crash risk in selecting congestion-mitigation strategies : Hampton Roads area (Virginia) case study (Garber, NJ and Subramanyan, S); Development of artificial neural network models to predict driver injury severity in traffic accidents at signalized intersections (Abdelwahab, HT and Abdel-Aty, MA); Transferability of models that estimate crashes as a function of access management (Miller, JS, Hoel, LA, Kim, S and Drummond, KP); Sensor-friendly vehicle and roadway cooperative safety systems : benefits estimation (Misener, JA, Thorpe, C, Ferlis, R, Hearne, R, Siegal, M and Perkowski, J); Interstate highway crash injuries during winter snow and nonsnow events (Khattak, AJ and Knapp, KK); Simulation of road crashes by use of systems dynamics (Mehmood, A, Saccamanno, F and Hellinga, B); Longitudinal analysis of fatal run-off-road crashes, 1975 to 1997 (McGinnis, RG, Davis, MJ and Hathaway, EA); Injury severity in multivehicle rear-end crashes (Khattack, AJ); Computing and interpreting accident rates for vehicle types driver groups (Hauer, E); Geographics information system-based accident data management for Mexican federal roads (Mendoza, A, Mayoral, EF, Vicente, JL and Quintero, FL); Bayesian identification of high-risk intersections for older drivers via gibbs sampling (Davis, GA and Yang, S); Automated accident detection system (Harlow, C and Wang, Y); Evaluation of inexpensive global positioning system units to improve crash location data (Graettinger, AJ, Rushing, TW and McFadden, J).

This study investigates the fatalities on the National Crash Severity Study (NCSS) of towaway, passenger car accidents. The analysis is in three stages. First, NCSS fatalities are compared to the fatally-injured occupants reported on the Fatal Accident Reporting System (FARS), as a tool for evaluating the representativeness of the NCSS data. Second, estimates of the probability of fatality for NCSS are computed for various conditions, such as the incidence of fire and the sex of the occupant. Third, in cases where two factors are highly correlated, such as is the case for rollover and ejection, modeling techniques are used to help quantify the effects of each variable. The results of this study suggest the following preliminary conclusions: (1) FARS and NCSS have similar distributions of many variables. These include urbanization, size of vehicles, type damage to vehicle, occupant seating location, sex, and restraint use. Differences resulting from the investigative methods and geographical areas of the two studies are identified and assessed. (2) On the NCSS file, many variables are associated with a much higher rate of fatality. These include (a) at the accident level: the number of vehicles involved, urbanization, and the incidence of fire or explosion; (b) at the vehicle level: the change of velocity at impact, the direction of the impacting force, and vehicle damage area; and (c) at the occupant level: seating position, age, sex, ejection, entrapment, and restraint use. (3) Rollover and ejection, which often occur together, are each independently associated with a higher rate of fatality. Of the two factors, ejection appears more related to a higher probability of fatality than does rollover alone. NCSS is the best currently-available source of accident data for analyzing injury-related factors. This report attempts to describe the accidents occurring in the NCSS sampling areas, and suggest ideas for further research.