General Estimates System Research Articles

Examination of Recent Pedestrian Safety Patterns at Intersections through Crash Data Analysis

Pedestrians are the most vulnerable road users and are at risk of severe consequences when involved in traffic accidents. The purpose of this research is to determine the factors that have significant impacts on the increasing likelihood of pedestrians being seriously injured or killed when involved in a collision with a single vehicle at an intersection over a recent 6-year period. Both 2013–2015 General Estimates System (GES) and 2016–2018 Crash Report Sampling System (CRSS) crash data were used in the analysis. Logistic regression models for the two crash datasets showed that there were four common significant variables affecting pedestrians’ injury levels. The following pairwise comparisons of these common significant factors using the Wald chi-square statistic test showed similar log-odds with few exceptions, suggesting that these affecting factors share similar effects from 2013 through 2018. In both datasets, results showed that a high likelihood of pedestrians’ severe injuries was associated with pedestrians older than 25, dark lighting conditions, light trucks and buses, and vehicles’ straight maneuver. Furthermore, the GES data distinguished further factors imposing higher threats on pedestrians as being drivers’ 19–25 age group, speeding, pedestrians’ roadway crossings maneuvers, and rain conditions. Crashes that occurred at intersections with more than two lanes or during summertime had significantly higher odds of resulting in severe injuries for pedestrians than crashes at two-lane intersections or during wintertime, respectively, in the CRSS dataset. Results of this study contribute to a better understanding of the recent changes in pedestrian safety at intersections and potential countermeasure design suggestions.

U.S. vehicle occupancy trends relevant to future automated vehicles and mobility services

Objective Identifying current occupant travel patterns can inform decision making regarding the design, regulation, and occupant protection systems helpful for automated vehicle systems and mobility services. Methods Two travel data sets were analyzed to quantify travel patterns: the 2017 National Household Travel Survey (NHTS), which provides data on household trips logged for a 24-h period, and the 2011–2015 National Automotive Sampling System–General Estimates System (NASS-GES), which contains data sampled from police-reported crashes. Analysis identified trends with driver age and gender, occupant age and gender, time of day, day of week, trip purpose, trip duration, vehicle type, as well as occupant role as solo driver, driver of others, single passenger, or multiple passengers. Results In NHTS, the median trip duration is 15 min; only 10% of trips last longer than 40 min. Trip duration does not vary with occupant role or vehicle type. Variations with trip time of day and day of week show a unimodal pattern for weekends, as well as weekday trips for those aged 55 years and older and non-solo occupants aged 18 to 29 years. Other occupant groups have a bimodal weekday travel pattern with peak trips corresponding to morning and evening rush hours. In GES, approximately half of occupants are solo drivers. Female drivers aged 55 and older travel alone 60% of the time, and drivers under age 18 and female drivers aged 30 to 54 drive alone on less than 45% of trips. Approximately 13% of occupants are single passengers, and 16% travel with a driver and at least 1 other passenger. About 16% of occupants are front seat passengers. Conclusions This analysis of vehicle occupancy provides insights on what ridership of future automated vehicles and expanded ride-hailing services may look like. Because half of occupants are solo drivers, only 16% are multiple passengers, and median trip length is just 15 min, proposed alternative seating arrangements intended to promote comfort and passenger interaction may not represent the typical future vehicle use case in the United States. Knowledge of current occupancy patterns can help automated vehicle designers and regulators develop safe seating scenarios that meet customer needs.

Potential Benefits of Animal-Detecting Automatic Emergency Braking Systems Based on U.S. Driving Data

In 2015, there were 319,195 police reported vehicle-animal crashes, resulting in 275 vehicle occupant fatalities. Animal-detecting automatic emergency braking (AEB) systems are a promising active safety measure which could potentially avoid or mitigate many of these crashes by warning the driver, utilizing automatic braking, or both. The purpose of this study was to develop and characterize a target population of vehicle-animal crashes applicable to AEB systems and to analyze the potential benefits of an animal-detecting AEB system. The study was based on two nationally representative databases, Fatality Analysis Reporting System and the National Automotive Sampling System’s General Estimates System, and a naturalistic driving study, SHRP 2. The target population was restricted to vehicle-animal crashes that were forward impacts or road departures and involved cars and light trucks, with no loss of control. Crash characteristics which may influence the performance of AEB such as lighting, weather, pre-crash movement, relation to junction, and first and worst harmful events, were analyzed. The study found that the major influences on the effectiveness of animal AEB systems were: weather, lighting, pre-crash movements, and the crash location. Six potential target populations were used to analyze the potential effectiveness of an animal AEB system, with effectiveness ranging between 21.6% and 97% of police reported crashes and between 4.1% and 50.8% of fatal vehicle-animal crashes. An AEB system’s ability to function in low light and poor weather conditions may enable it to avoid a substantially higher proportion of crashes.

A cluster-based taxonomy of bus crashes in the United States

Accident taxonomy or classification can be used to direct the attention of policymakers to specific concerns in traffic safety, and can subsequently bring about effective regulatory change. Despite the widespread usage of accident taxonomy for general motor vehicle crashes, its use for analyzing bus crashes is limited. We apply a two-stage clustering-based approach based on self-organizing maps followed by neural gas clustering to construct a data-driven taxonomy of bus crashes. Using the 2005–2015 data from general estimates system, we identify four clusters and expose the qualitative traits that characterize four distinct types of bus crash. Our analysis suggests that cluster characteristics are largely stable over time. Consequently, we make targeted policy recommendations for each of the four subtypes of bus crash.

Predictive modeling of maximum injury severity and potential economic cost in a car accident based on the General Estimates System data

In this paper, we aim to identify the significant variables that contribute to the injury severity level of the person in the car when an accident happens and build a statistical model for predicting the maximum injury severity level as well as estimating the potential economic cost in a car accident based on those variables. The General Estimates System data, which is a representative sample of police-reported motor vehicle crashes of all types collected by the National Highway Transportation Safety Administration, from the years 2012 to 2013 is the main data source. Some other data sources such as the car safety rating from the United State Department of Transformation and the state-specific cost of crash deaths fact sheets are also used in the predictive model building process. An interactive system programmed in HyperText Markup Language, Cascading Style Sheets and JavaScript is developed based on the results of predictive modeling. The system is hosted on a website at http://gessmu.azurewebsites.net for public access. The system allows users to input variables that are significant contributors in car accidents and obtain the predicted maximum injury severity level and potential economic cost of a car accident.

The 2016 Data Challenge of the American Statistical Association

The Sections on Statistical Graphics and Statistical Computing of the American Statistical Association have a long history of issuing Data Challenges with the first one starting in 1982/1983. The challenge is now an annual event where most of them use data collected and disseminated by the U.S. government. The data set for the 2016 Data Challenge was the Department of Transportation’s General Estimates System. The GES is collected by the National Highway Transportation Safety Administration and is a representative sample of police-reported motor vehicle crashes. This editorial introduces the five papers submitted by contestants in the data challenge.

A hierarchical Bayes approach to adjust for selection bias in before–after analyses of vision zero policies

American cities devote significant resources to the implementation of traffic safety countermeasures that prevent pedestrian fatalities. However, the before–after comparisons typically used to evaluate the success of these countermeasures often suffer from selection bias. This paper motivates the tendency for selection bias to overestimate the benefits of traffic safety policy, using New York City’s Vision Zero strategy as an example. The NASS General Estimates System, Fatality Analysis Reporting System and other databases are combined into a Bayesian hierarchical model to calculate a more realistic before–after comparison. The results confirm the before–after analysis of New York City’s Vision Zero policy did in fact overestimate the effect of the policy, and a more realistic estimate is roughly two-thirds the size.

Analysis of drowsy driving: exploring subpopulation risk with weighted contingency table tools

The incidence of driving accidents due to human error, and drowsy driving in particular, is an important topic in the field of public health research, and might be considered preventable. Our study investigates the 606 police-reported drowsy driving accidents in 2013 as recorded in the NASS General Estimates System from the NHTSA. This study seeks to examine how the prevalence of drowsy driving in accidents differs between subpopulations and how this prevalence changes depending on the time of day. We explore these interactions using recent developments in survey-weighted ROC analysis. By doing so, we hope to offer employers and government agencies insight into what can be done to reduce the rate of injuries and fatalities related to drowsy driving.

An analysis of crash-safety ratings and the true assessment of injuries by vehicle

Each year the National Highway Traffic Safety Administration and Insurance Institute for Highway Safety release safety ratings for popular makes and models of vehicles produced. We link these safety ratings with the crash data provided in the National Automotive Sampling System General Estimates System to study the efficacy of safety ratings as a predictor of the likelihood a passenger will be injured in a crash. We also consider the case of predicting a severe injury or death as a function of safety rating. A web-based dashboard was developed to graphically explore the relationship among these datasets. By considering the proportion of occupants (severely) injured per crash, we find that vehicle safety ratings generally do not have as much of an effect on likelihood of injury as one might expect, although vehicles with higher ratings can result in significant relative improvement. Our web application allows a user to explore different aspects of crashes (e.g., use of alcohol, speeding, etc.) and compare injuries and safety rating performance based on these conditions. Lastly, our dashboard allows a user to see differences between the NHTSA and IIHS ratings systems and how they correspond to the crash records.

Red light camera interventions for reducing traffic violations and traffic crashes: A systematic review.

Red light camera interventions for reducing traffic violations and traffic crashes: A systematic review.

Factors Affecting Child Injury Risk in Motor-Vehicle Crashes.

Current recommendations for restraining child occupants are based on biomechanical testing and data from national and international field studies primarily conducted prior to 2011. We hypothesized that analysis to identify factors associated with pediatric injury in motor-vehicle crashes using a national database of more recent police-reported crashes in the United States involving children under age 13 where type of child restraint system (CRS) is recorded would support previous recommendations. Weighted data were extracted from the National Automotive Sampling System General Estimates System (NASS-GES) for crash years 2010 to 2015. Injury outcomes were grouped as CO (possible and no injury) or KAB (killed, incapacitating injury, nonincapacitating injury). Restraint was characterized as optimal, suboptimal, or unrestrained based on current best practice recommendations. Analysis used survey methods to identify factors associated with injury. Factors with significant effect on pediatric injury risk include restraint type, child age, driver injury, driver alcohol use, seating position, and crash direction. Compared to children using optimal restraint, unrestrained children have 4.9 (13-year-old) to 5.6 (< 1-year-old) times higher odds of injury, while suboptimally restrained children have 1.1 (13-year-old) to 1.9 (< 1-year-old) times higher odds of injury. As indicated by the differences in odds ratios, effects of restraint type attenuate with age. Results support current best practice recommendations to use each stage of child restraint (rear-facing CRS, forward-facing harnessed CRS, belt-positioning booster seat, lap and shoulder belt) as long as possible before switching to the next step.

Urban freight and road safety in the era of e-commerce

Objective: E-commerce has increased freight volumes and shifted freight vehicles to local streets and arterials. These changes have caused increasing concern over traffic safety in urban areas, but there has been limited investigation of road safety impacts. To advance understanding of this issue, we examined recent trends in urban freight-related safety in the United States and characteristics of these crashes.Methods: We estimated annual rates of fatal and nonfatal urban freight-related injuries from 2005 through 2015 using data from the NHTSA’s Fatal Analysis Reporting System and National Automotive Sampling System General Estimates System. We compared trends in fatal and nonfatal urban freight-related injuries, stratified by road type (interstate vs. not), to overall crash injury trends. We used piecewise linear regression models to test trend changes. We examined and compared specific crash and injury characteristics associated with urban freight-involved crashes, compared to other types of crashes, for the most recent year of data analyzed.Results: Though freight and overall crash fatality rates per vehicle mile traveled declined between 2005 and 2009, increases were observed between 2009 and 2015. Fatality rates increased 3% for all vehicle crashes, 17% for urban freight-involved crashes, and 15% for urban freight-involved crashes not occurring on interstates. Nonfatal injury rates for urban freight-involved crashes rose even more rapidly between 2009 and 2015. Nonfatal injury rates for urban freight-involved crashes increased 45%, rates for urban freight-involved crashes not occurring on interstates increased 40%, and overall nonfatal injury rates for all vehicle crashes increased 3% during this time. Of all urban freight-related crashes resulting in nonfatal injury, the proportion occurring on non-interstate roads increased from 17% to 25% between 2005 and 2015, and these crashes were more likely to occur on weekdays between 6 a.m. and 4 p.m., when freight demand is high, than weekends or evenings.Conclusions: Freight-involved injury and fatality rates are rising more rapidly than overall road traffic–related rates, both in all areas and in urban areas. These crashes are also increasingly occurring on local roads and arterials as opposed to interstates. These findings can help policymakers better understand the changing patterns of freight-related safety issues. As freight volumes increase in commercial and residential areas, planners must increasingly consider freight needs and ensure that space is allocated to this function. Additionally, changes to surveillance systems are necessary to better track burden and risk factors associated with these crashes and inform crash prevention efforts.

Intersection characteristics that influence collision severity and cost

Introduction: Traffic engineers require robust tools to assist with their day-to-day decision making, and there is no better example of this than traffic signal warrants. North American traffic signal warrant systems are lacking in how they incorporate motor-vehicle collisions from both a severity and prediction perspective. The objective of this study was to produce reliable collision costs for the development of improved traffic signal warrants that accounted for the variations in severity that practitioners should expect based on the characteristics of the intersection being studied. Method: The primary data used for this analysis were from the National Automotive Sampling System (NASS) Crashworthiness Data System, with adjustments from the NASS General Estimates System and Fatality Accident Reporting System. Generalized ordered logit models were used to identify the most significant intersection characteristics, which were then used to segregate the data to determine expected the collision severity profiles and average costs of both casualty and total collisions at intersections. Results: The average collision at a signalized intersection was found have a lower severity than the average collision at a stop-controlled intersection. A combination of posted speed limit, urban/rural, and divided/undivided were identified as the most significant intersection characteristics in most cases and were used to delineate the data for developing collision cost estimates. Conclusions: Posted speed limit, rural/urban land use, and the presence of divided approaches are intersection characteristics that traffic engineers can readily determine and/or control for that have significant effects on intersection collision severity. Practical applications: The collision costs produced through this process give traffic engineers a reliable estimate that can provide a more substantial foundation for justifying a proposed change in intersection traffic control.

Evaluating temporal variability of exogenous variable impacts over 25 years: An application of scaled generalized ordered logit model for driver injury severity

The current study undertakes a unique research effort to quantify the impact of various exogenous factors on crash severity over time. Specifically, we examine if over time, the impact of exogenous variables has changed and if so what is the magnitude of the change. The research contributes to driver injury severity analysis both methodologically and empirically by proposing a framework that addresses the challenges associated with pooled (or pseudo-panel) data. For our analysis, we draw data from the General Estimates System (GES) over a span of twenty-five years. The data is compiled for driver injury severity in single or two vehicle crashes from 1989 through 2014 in 5-year increments (1989, 1994, 1999, 2004, 2009 and 2014). The alternative econometric frameworks considered for the analysis include ordered logit, generalized ordered logit, scaled generalized ordered logit and mixed generalized ordered logit models. A host of comparison metrics are computed to evaluate the performance of these alternative models in examining the pooled data. The model development exercise is conducted with a host of exogenous variables including driver characteristics, vehicle characteristics, roadway attributes, environmental factors, crash characteristics and temporal attributes. The model estimation results are further augmented by performing a detailed policy scenario analysis, probability profile representations and elasticity effects for different driving and situational conditions across different years.

An examination of the increases in pedestrian motor-vehicle crash fatalities during 2009–2016

IntroductionPedestrian fatalities increased 46% in the United States during 2009–2016. This study identified circumstances under which the largest increases in deaths occurred during this period. MethodAnnual counts of U.S. pedestrian fatalities and crash involvements were extracted from the Fatality Analysis Reporting System and General Estimates System. Poisson regression examined if pedestrian fatalities by various roadway, environmental, personal, and vehicle factors changed significantly during 2009–2016. Linear regression examined changes over the study period in pedestrian deaths per 100 crash involvements and in horsepower per 1000 pounds of weight among passenger vehicles involved in fatal single-vehicle pedestrian crashes ResultsPedestrian deaths per 100 crash involvements increased 29% from 2010, when they reached their lowest point, to 2015, the most recent year for which crash involvement data were available. The largest increases in pedestrian deaths during 2009–2016 occurred in urban areas (54% increase from 2009 to 2016), on arterials (67% increase), at nonintersections (50% increase), and in dark conditions (56% increase). The rise in the number of SUVs involved in fatal single-vehicle pedestrian crashes (82% increase) was larger than the increases in the number of cars, vans, pickups, or medium/heavy trucks involved in these crashes. The power of passenger vehicles involved in fatal single-vehicle pedestrian crashes increased over the study period, with larger increases in vehicle power among more powerful vehicles. ConclusionsEfforts to turn back the recent increase in pedestrian fatalities should focus on the conditions where the rise has been the greatest. Practical applicationsTransportation agencies can improve urban arterials by investing in proven countermeasures, such as road diets, median crossing islands, pedestrian hybrid beacons, and automated speed enforcement. Better road lighting and vehicle headlights could improve pedestrian visibility at night.

Forward collision warning system impact

Objective: The objective of this research is to use historical crash data to evaluate the potential benefits of both high- and low-speed automatic emergency braking (AEB) with forward collision warning (FCW) systems.Methods: Crash data from the NHTSA’s NASS–General Estimates System (GES) and Fatality Analysis Reporting System (FARS) databases were categorized to classify crashes by the speed environment, as well as to identify cases where FCW systems would be applicable.Results: Though only about 19% of reported crashes occur in environments with speeds greater than 45 mph, approximately 32% of all serious or fatal crashes occur in environments with speeds greater than 45 mph. The percentage of crashes where FCW systems would be relevant has remained remarkably constant, varying between about 21 and 26% from 2002 to 2015. In 2-vehicle fatal crashes where one rear-ends the other, the fatality rates are actually higher in the struck vehicle (33%) than the striking vehicle (26%). The disparity is even greater when considering size–class differences, such as when a light truck rear-ends a passenger car (15 vs. 42% fatality rates, respectively).Conclusions: NHTSA and the Insurance Institute for Highway Safety (IIHS) proposed the Automatic Emergency Braking Initiative in 2015, which is intended to make AEB (also called crash-imminent braking) with FCW systems standard on nearly all new cars by September 2022. Twenty automakers representing 99% of the U.S. auto market voluntarily committed to the initiative. Though the commitment to safety is laudable, the AEB component of the agreement only covers low-speed AEB systems, with the test requirements set to 24 mph or optionally as low as 12 mph. The test requirements for the FCW component of the agreement include 2 tests that begin at 45 mph. Only 21% of relevant serious injury or fatal accidents occur in environments at speeds under 24 mph, whereas about 22% of serious or fatal crashes occur in environments with speeds greater than 45 mph. This means that the AEB with FCW systems as agreed upon will cover only 21% of serious or fatal crashes and will not cover 22% of serious or fatal crashes. Because these systems are protective not only for the occupants of the vehicle where they are installed but also other vehicles on the roads, the data indicate that these systems should be a standard feature on all cars for high-speed as well as low-speed environments for the greatest social benefit.

Modified Mixed Generalized Ordered Response Model to Handle Misclassification in Injury Severity Data

Traditional crash databases that record police-reported injury severity data are prone to misclassification errors. Ignoring these errors in discrete ordered response models used for analyzing injury severity can lead to biased and inconsistent parameter estimates. In this study, a mixed generalized ordered response (MGOR) model that quantifies misclassification rates in the injury severity variable and adjusts the bias in parameter estimates associated with misclassification was developed. The proposed model does this by considering the observed injury severity outcome as a realization from a discrete random variable that depends on true latent injury severity that is unobservable to the analyst. The model was used to analyze misclassification rates in police-reported injury severity in the 2014 General Estimates System (GES) data. The model found that only 68.23% and 62.75% of possible and non-incapacitating injuries were correctly recorded in the GES data. Moreover, comparative analysis with the MGOR model that ignores misclassification not only has lower data fit but also considerable bias in both the parameter and elasticity estimates. The model developed in this study can be used to analyze misclassification errors in ordinal response variables in other empirical contexts.

Development of test scenarios and bicyclist surrogate for the evaluation of bicyclist automatic emergency braking systems

Purpose To support the standardized evaluation of bicyclist automatic emergency braking (AEB) systems, test scenarios, test procedures and test system hardware and software tools have been investigated and developed by the Transportation Active Safety Institute (TASI) at Indiana University-Purdue University Indianapolis. This paper aims to focus on the development of test scenarios and bicyclist surrogate for evaluating vehicle–bicyclist AEB systems. Design/methodology/approach The harmonized general estimates system (GES)/FARS 2010-2011 crash data and TASI 110-car naturalistic driving data (NDD) are used to determine the crash geometries and environmental factors of crash scenarios including lighting conditions, vehicle speeds, bicyclist speeds, etc. A surrogate bicyclist including a bicycle rider and a bicycle surrogate is designed to match the visual and radar characteristics of bicyclists in the USA. A bicycle target is designed with both leg pedaling and wheel rotation to produce proper micro-Doppler features and generate realistic motion for camera-based AEB systems. Findings Based on the analysis of the harmonized GES/FARS crash data, five crash scenarios are recommended for performance testing of bicyclist AEB systems. Combined with TASI 110-car naturalistic driving data, the crash environmental factors including lighting conditions, obscuring objects, vehicle speed and bicyclist speed are determined. The surrogate bicyclist was designed to represent the visual and radar characteristics of the real bicyclists in the USA. The height of the bicycle rider mannequin is 173 cm, representing the weighted height of 50th percentile US male and female adults. The size and shape of the surrogate bicycle were determined as 26-inch wheel and mountain/road bicycle frame, respectively. Both leg pedaling motion and wheel rotation are suggested to produce proper micro-Doppler features and support the camera-based AEB systems. Originality/value The results have demonstrated that the developed scenarios, test procedures and bicyclist surrogate will provide effective objective methods and necessary hardware and software tools for the evaluation and validation of bicyclist AEB systems. This is crucial for the development of advanced driver assistance systems.

Epidemiology of Pedestrian–Motor Vehicle Fatalities and Injuries, 2006–2015

Pedestrian road safety remains a public health priority. The objective of this study is to describe trends in fatalities and injuries after pedestrian-motor vehicle collisions in the U.S. and identify associated risk factors for pedestrian fatalities. This is a cross-sectional study of U.S. pedestrian-motor vehicle collisions from 2006 to 2015 (performed in 2017). Pedestrian fatality and injury data were obtained from the National Highway Traffic Safety Administration's Fatality Analysis Reporting System and National Automotive Sampling System General Estimates System. Frequencies of fatalities, injuries, and associated characteristics were calculated. Multivariable logistic regression was performed for risk of fatality, controlling for demographic and crash-related factors. There were 47,789 pedestrian fatalities and 674,414 injuries during the 10-year study period. Fatality rates were highest among the elderly aged 85 years and older (2.95/100,000 population), whereas injury rates were highest for those aged 15-19 years (35.23/100,000 population). Predictors associated with increased risk for death include the following: male sex (AOR=1.36, 95% CI=1.15, 1.62), age ≥65 years (AOR=3.44, 95% CI=2.62, 4.50), alcohol involvement (AOR=2.63, 95% CI=1.88, 3.67), collisions after midnight (AOR=5.21, 95% CI=3.20, 8.49), at non-intersections (AOR=2.76, 95% CI=2.21, 3.45), and involving trucks (AOR=2.15, 95% CI=1.16, 3.97) and buses (AOR=5.82, 95% CI=3.67, 9.21). Potentially modifiable factors are associated with increased risk of death after pedestrian-motor vehicle collisions. Interventions including elder-friendly intersections and increasing visibility of pedestrians may aid in decreasing pedestrian injuries and deaths.

Analyzing the Injury Severity Sustained by Non-Motorists at Mid-Blocks considering Non-Motorists’ Pre-Crash Behavior

Non-motorized travel is considered as one of the most beneficial transportation modes. Compared with other road users, non-motorists as a whole account for about 13% of all fatal transportation-related accidents, and from 2002 to 2009 nearly 30% of those fatalities occur at mid-blocks. In addition, there are few reported studies that investigated the impact of non-motorists’ pre-crash behavior on injury severities. To examine the risk factors of non-motorist injury severity at mid-blocks, 8-year crash-related data from the General Estimates System were explored, based on the mixed logit model. The data contain various information including time characteristics, crash features, environmental conditions, roadway attributes, non-motorists’ characteristics, and their pre-crash behaviors. The results show that five factors tend to have mixed effects on injury severities, including the speed limit between 30 and 55 mph, night time indicator, right-side collision, and hit-and-run action on the incapacitating injury, as well as no action of motorists on the non-incapacitating injury. Moreover, heavy and light truck, dark not lighted indicator, and age over 65 are found to increase the likelihood of fatal injury, while age below 25 decreases the likelihood of fatality. Other indicators including roadway alignment, number of lanes, and so forth also affected injury severity. After controlling for these factors, non-motorists’ pre-crash behaviors such as darting or running into the road, activities in the roadway, and improper passing are found to have a significant impact on severity outcomes.