Crash Data Research Articles

Alternative unobserved heterogeneity models to analyze injury severity of expressway crashes in different tunnel types

This study examines the non-transferability of the determinants affecting injury severities of tunnels of three grades classified by different lengths. Based on tunnel crash data in Shanxi Province during (2012–2018), three possible crash injury severity categories (no injury, minor injury, and severe injury) are estimated. Two alternate models were used to account for unobserved heterogeneity: a random parameters multinomial logit (RPML) model with heterogeneity in means and variances and a latent-class multinomial logit (LCML) model with class probability functions. Non-transferability in the effects of explanatory variables is confirmed using two series of likelihood ratio tests based on the two alternate models. A wide variety of variables is observed to determine pedestrian-injury severities. The estimation results show significant non-transferability across the three grades in both RPML and LCML models. In addition, several explanatory variables produce relatively transferable effects, which can provide some valuable insights to implement effective countermeasures. Moreover, out-of-sample predictions are simulated to confirm the non-transferability, while the LC models produce higher differences for the non-transferability than RP models. According to estimation results, several practical measurements concerning education programs, traffic enforcement, emergency response systems, and traffic safety measurements are implemented to eliminate tunnel crash outcomes. Understanding and depth comparing the estimation results, likelihood ratio tests and out-of-sample predictions using alternate models is a promising direction for future research to explore how the observed and unobserved heterogeneity can be estimated.

A comprehensive study of child pedestrian crash outcomes in Ghana

Pedestrians are considered as one of the vulnerable road user groups. Among pedestrians of all ages, children are the most at risk. Previous studies have shown that children have inadequate knowledge of road safety and are unable to identify risks on road. Despite these limitations associated with children, society places the burden on them to protect themselves. However, to be able to adequately address child pedestrian safety issues there is the need to understand the factors that influence their crash involvement and severity of injury they sustain. To address this gap, this study performed a comprehensive analysis of historical crash data in Ghana to find holistic countermeasures for these crashes. The study used five years of child pedestrian (below 10 years) crash records obtained from the Building and Road Research Institute (BRRI) in Ghana. A temporal analysis of the data revealed that the highest number of the crashes coincide with when school-going children go to and close from school. A random parameter multinomial logit model was developed to identify crash variables that are significantly associated with child pedestrian crash outcomes. The estimation results revealed that children are likely to be killed in crashes when the driver is speeding and inattentive. Also, it was found that children walking along the road, crossing the road, and those in urban areas are more susceptible to incapacitating injury crashes. Male drivers accounted for 95.8% of child pedestrian crashes, and crashes involving male drivers are 7.8% more likely to be fatal.The findings from this study provide a deeper, data-driven understanding of child pedestrian crashes and how temporal characteristics, vehicle type, location of pedestrian, traffic operation, and environmental and human factors affect crash outcomes. These findings will help in developing countermeasures like providing conspicuous pedestrian crossings, footbridges on multi-lane high-speed roadways, and the use of school buses to convey students to help mitigate the number and severity of child pedestrian crashes in Ghana, and by extension other countries in the sub-region.

Evaluating Right-Turn Flashing Yellow Arrow for Vehicle–Pedestrian Interactions Using a Non-Probabilistic Regression Approach

Using video, interactions between right-turning vehicles and pedestrians were documented for sites with a permissive circular green or a flashing yellow arrow (FYA) permissive right-turn indication. Timestamps associated with key vehicle and pedestrian positions were documented and used to calculate three metrics: time for driver to complete right turn (with and without a pedestrian presence) and the time for the pedestrian to arrive at the theoretical conflict point of vehicle–pedestrian interactions. Using timestamp data, multiple non-probabilistic linear regression models were created to describe the relationship between the position of the pedestrian within the crosswalk and the time for a right-turning vehicle maneuver to be completed. Model results implicitly describe the attitude of drivers toward pedestrian presence on the crosswalk. Larger right-turning maneuver completion times are arguably an indicator of greater respect toward pedestrian presence. Given the nature of the model’s output, a pedestrian respect indicator (PRI) based on the value of model parameters is introduced as a potential indicator of the safety of vehicle–pedestrian interactions. PRI values can be computed for individual sites thus allowing its use as a safety indicator. The higher the PRI, the more “respect” toward pedestrians. Proof-of-concept modeling results based on available timestamp data suggest that interactions controlled by a right-turn FYA indication have a higher PRI than interactions controlled by a circular green. An argument is made that PRI values can be potentially used to proactively evaluate the effectiveness of traffic control devices and other safety countermeasures without the need for crash data.

Self-Paced Ensemble-SHAP Approach for the Classification and Interpretation of Crash Severity in Work Zone Areas

The identification of causative factors and implementation of measures to mitigate work zone crashes can significantly improve overall road safety. This study introduces a Self-Paced Ensemble (SPE) framework, which is utilized in conjunction with the Shapley additive explanations (SHAP) interpretation system, to predict and interpret the severity of work-zone-related crashes. The proposed methodology is an ensemble learning approach that aims to mitigate the issue of imbalanced classification in datasets of significant magnitude. The proposed solution provides an intuitive way to tackle issues related to imbalanced classes, demonstrating remarkable computational efficacy, praiseworthy accuracy, and extensive adaptability to various machine learning models. The study employed work zone crash data from the state of New Jersey spanning a period of two years (2017 and 2018) to train and evaluate the model. The study compared the prediction outcomes of the SPE model with various tree-based machine learning models, such as Light Gradient Boosting Machine, adaptive boosting, and classification and regression tree, along with binary logistic regression. The performance of the SPE model was superior to that of tree-based machine learning models and binary logistic regression. According to the SHAP interpretation, the variables that exhibited the highest degree of influence were crash type, road system, and road median type. According to the model, on highways with barrier-type medians, it is expected that crashes that happen in the same direction and those that happen at a right angle will be the most severe crashes. Additionally, this study found that severe injuries were more likely to result from work zone crashes that happened at night on state highways with localized street lighting.

Toolkit for the Assessment of Safety at Bus Stops

Safety at bus stops is a major concern of transit agencies. This research aimed to develop a data-driven method for estimating bus stop safety scores that will assist transit agencies in ranking and prioritizing bus stops for improvement considerations. This study presents a bus stop scoring methodology that uses a data-driven approach based on multiple data sources. The methodology was developed using six criteria: bus stop characteristics inventory, roadway characteristics inventory (RCI), land use and demographics, pedestrian crash frequency, annual average daily traffic, and annual ridership. Crash data were collected from SignalFour Analytics, and roadway characteristics data were retrieved from the Florida Department of Transportation RCI database. The Florida Geographic Data Library was used for retrieving census and land use data, while the Florida Transit Information System was employed to extract information on transit stops. A data-driven method was used to develop a priority list for improvements among the selected bus stops. The methodology was implemented in a case study including three transit agencies: Hillsborough Area Regional Transit Authority, Jacksonville Transportation Authority, and Miami-Dade Transit. A bus stop priority list was generated for each agency, with the highest-ranking bus stop location having the greatest need for safety improvements. Based on the analysis, safety measures that could improve safety at the highest-ranking bus stop locations include: installing a shelter, installing pedestrian safety features, conducting pedestrian safety outreach programs, and/or conducting further site investigation.

Collisions and cannabis: Measuring the effect of recreational marijuana legalization on traffic crashes in Washington State

Objective Washington State was among the first states in the US to legalize recreational consumption and retail sales of marijuana. Recreational use of cannabis was legalized December 6, 2012, following the passage of Initiative 502 30 days prior. Roughly 19 months later the first retail cannabis stores opened their doors for public sales (“commercialization”). I measure the impact of cannabis legalization and commercialization on traffic collisions in Washington State. Methods With county-level vehicle crash data from the Washington State Department of Transportation collected monthly, I utilize an interrupted time-series framework with Poisson estimation to compare traffic collisions with recreational retail cannabis sales revenue from 2011 (three years pre-commercialization) through 2017 (three years post-commercialization). First, I measure the shift in collisions brought about by Washington’s 2012 cannabis legalization. Then, I compare retail cannabis sales—a measure of commercialization—to traffic collisions based on severity of injury (fatal, severe injury, minor injury, non-injury, and all). Results After controlling for confounding factors, evidence suggests that recreational cannabis legalization led to fewer fatal and serious injury collisions. Retail cannabis sales generally correlate with more traffic collisions, particularly for less severe (minor injury) crashes. These findings are robust to the inclusion of additional control variables pertaining to county-level cannabis usage and driving behavior while intoxicated. Conclusions Cannabis legalization led to fewer fatal, serious, and minor injury collisions. Commercialization (cannabis sales) correlated with an increase in less severe crashes. Although cannabis use generally increased in Washington State following legalization/commercialization, survey data suggest that driving behavior while under the influence of cannabis did not change significantly over the post-commercialization period. Future research should focus on measuring the dose-dependent impact of cannabis consumption on traffic collisions. This should include recognition of the importance of cannabis dosing, timing, and route of consumption. Lastly, the dangers of poly-drug driving—particularly cannabis and alcohol—are well established and should be high priority for further research.

Traffic Order Analysis of Intersection Entrance Based on Aggressive Driving Behavior Data Using CatBoost and SHAP

Analyzing road risks and developing targeted countermeasures are essential for a safe and orderly traffic flow. However, previous intersection safety analyses were conducted based on crash data. Little research has been conducted on surrogate safety measures based on risky driving behavior. In this study, categorical boosting (CatBoost) and Shapley additive explanation (SHAP) were used to analyze the impact of features on traffic order using a set of multisource data that include roadway geometry, signal control, and land use. The traffic data for intersection entrances in Beijing were collected from navigation systems, field investigations, and application programming interfaces. The model results showed that CatBoost exhibits a prediction accuracy of 83.5%, a recall of 83.5%, and an F1 score of 81.1%. Moreover, the importance, total effects, main effects, and interaction effects of influence factors were analyzed by using SHAP. It was found that the congestion index (CI) has significant negative effects on traffic order. A larger number of lanes and more electronic traffic control were found to have a positive effect on traffic order. Intersection entrances with three-phase signals or off-peak intersection entrances helped increase traffic order. Moreover, a high green ratio for through vehicles can reduce the positive impact of CI on traffic order when the value of CI is 1.1–1.4, and the signal control scheme with a high left-turn green ratio would result in a safe and orderly traffic flow. The results from this study can be used for further studies on improving traffic safety at signalized intersections.

Influencing factors for right turn lane crash frequency based on geographically and temporally weighted regression models

Introduction: Right-turn lane (RTL) crashes are among the key contributors to intersection crashes in the US. Unfortunately, the lack of deep insights into understanding the effects of RTL geometric design factors on crash frequency impedes improving RTL safety performance. Method: Taking the crash data in ten counties in Indiana state from 2013 to 2016 as a case study, this study investigates the safety performance of RTL geometric configuration based on multi-sources. We introduce the geographically and temporally weighted negative binomial model (GTWNBR) to capture the space and time instability in crashes. Results: The results show that the impacts of RTL geometric design factors on crash frequency vary significantly among space and time. Several key insights can be obtained from the state-wide and multi-years crash analysis by associating the estimated parameters with road classes, localities, and counties. Conclusions: First, the RTL’s length, width, turning radius, and the installments of traffic roundabouts present higher spatiotemporal heterogeneity than other factors in modeling the crash frequency. Second, the effects of RTL’s geometric factors vary significantly across space and time. The presence of bicycle and pedestrian lanes is more likely to increase crashes in urban areas than in rural ones, especially at nighttime. Third, while exclusive RTLs decrease the crash frequency compared to the shared RTLs, the exclusive RTLs are more likely to increase the crashes for RTLs on the county road than on other road classes. Increasing RTL’s turning radius and decreasing RTL’s length is more likely to promote crashes for RTLs on county roads than on other road classes. Practical Applications: The insights provide vital guidance to improve the safety performance of geometric configuration for RTLs and intersections.

Assessing a Connected Environment’s Safety Impact During Mandatory Lane-Changing: A Block Maxima Approach

Previous studies have mostly hypothesised or provided preliminary evidence that crash risk during mandatory lane-changing manoeuvres in a connected environment would be reduced. However, the unavailability of crash data for a connected environment makes it challenging to confirm such a hypothesis. To this end, this study has adopted an Extreme Value Theory approach to estimate and compare crash risk using traffic conflicts during mandatory lane-changing manoeuvres in the traditional and connected environments. Using the CARRS-Q advanced driving simulator, seventy-eight participants performed mandatory lane-changing manoeuvres in three randomised driving conditions: baseline (without driving aids), connected environment with perfect communication (PC), and connected environment with communication delay (CD). Driving-related factors obtained from the driving simulator data such as speeds, spacings, lag gaps, and remaining distances were used as input to a block maximum (or generalised extreme value, GEV) model. The characteristics of the estimated GEV distributions and relative crash risk were employed as an indicator of safety. Results reveal that in the connected environment driving conditions, the mandatory lane-changing crash risk is significantly reduced compared to the baseline, with the highest reduction observed in the PC condition. The crash risk is found to be higher in the CD condition compared with the PC condition. Findings of this study confirm (i) that the connected environment has the potential to reduce the crash risk substantially by assisting drivers during mandatory lane-changing manoeuvres, and (ii) the efficacy of GEV models to quantify crash risk by using conflict data when crash records are unavailable.

Business Intelligence (BI) Approach for Traffic Accidents Analysis

‘FLCRASH’ is the data source used for this report that consists of crash data from the year 2008 to 2009. The purpose of this report is to find ways to reduce the number of fatal accidents on the road by identifying the main culprits on why these accidents happen so frequently. Business intelligence is used in the process of finding the relationships and trends in the major causes of the crashes. The technologies related to Business Intelligence that will help in aiding the organization to their goal are also discussed along with the supporting factors for the use of Business Intelligence for the organization. A dashboard has been created from various graphs and charts for easy viewing and understanding of the situation so logical decisions can be made based on it. This report seeks to help us better understand the nature of why traffic accidents happen in the first place and how to prevent them from happening in the future.

Spatial analysis of moving-vehicle crashes and fixed-object crashes based on multi-scale geographically weighted regression

Previous researches have demonstrated that traffic crashes in urban areas are geographical events and strongly linked to local characteristics such as road network and land attributes. However, with a significant emphasis on moving-vehicle crashes, the spatial pattern of fixed-object crashes is unclear so far. The difference between these two types of crashes, and whether existing spatial tools such as geographically weighted regression can interpretate the occurrence mode have not been investigated before. To fill this gap, this paper focuses on understanding the spatial features and occurrence of these two types of crash, i.e., moving-vehicle and fixed-object on the city level. Crash data from Dalian, China were aggregated into subdistricts and calibrated with multi-scale geographically weighted regression (MGWR) models. A noticeable but similar clustering pattern was revealed in both types, with spatial overlap of their accident-prone regions. The spatial influence of explanatory variables (road network, geographic, demographic, socio-economic, and land-use variables) was also found mostly similar in both types of crashes. However, fixed-object crash in downtown is more affected by node count, while POI entrance/exit count, especially those in areas with more industrial zones tend to significantly reduce crash risk. In both types of crashes, terrain slope rather than elevation is found to mitigate the crash risk, especially in the downtown area. Compared to traditional Geographically Weighted Regression (GWR) with a fixed bandwidth, the improvement in modeling performance using MGWR highlights the reasonability and benefits to consider the influence scale of each contributing factor in urban spatial analysis of traffic collisions. This study could help transportation authorities identify high-risk regions, understand their contributing factors and take precautions for improving the local traffic safety.

Comprehensive Investigation of Commercial Motor Vehicle Crashes at Signalized and Unsignalized Intersections in Kentucky

This study develops safety performance functions (SPFs) for “fatal and injury” (FI) commercial motor vehicle (CMV) crashes (here, a CMV is defined as a large truck or a bus) at both signalized and unsignalized intersections in Kentucky. Five count-response regression models—negative binomial, Conway–Maxwell–Poisson, heterogeneous Conway–Maxwell–Poisson, zero-inflated Conway–Maxwell–Poisson, and zero-inflated heterogeneous Conway–Maxwell–Poisson (ZI-HTCMP)—were developed and compared at each of the selected signalized and unsignalized intersections. FI CMV intersection-related crash data in Kentucky over a period of five years (2015 to 2019) were used. Information on road-specific characteristics, for example, shoulder width, median width and type, International Roughness Index (IRI) value, and traffic volume (including annual average daily traffic [AADT] and heavy vehicle percentage [HVP]) was collected. Additionally, information on intersection-specific characteristics, for example, number of through lanes and presence of exclusive left- and right-turn lanes on major and minor roads, was collected using Google Maps’ street view time slider back to the crash year. For both signalized and unsignalized intersections, the ZI-HTCMP model (with a varying dispersion parameter) outperformed the other comparative models. With regard to signalized intersections, a speed limit of greater than 45 mph on the major road, a speed limit greater than 45 mph on the minor road, and a median on the major road wider than 3 ft were significantly associated with increased FI CMV crashes. As far as unsignalized intersections were concerned, higher major road AADT, 4-legged configuration, a speed limit greater than 45 mph on the major road, and an IRI value of greater than 100 for the major road were significantly associated with increased FI CMV crashes. The SPF results (for each of the signalized and unsignalized intersections) were then used to identify the top 10 hazardous locations for each intersection type.

Statistical Model in Predicting Traffic Congestion Among Selected Routes in Metro Manila

Traffic congestion is a serious issue that contributes significantly to economic loss, increase in greenhouse gas emissions, and fuel wastage. Hence, an accurate congestion prediction model can help address these problems. This paper analyzes the status of the road transport infrastructure, public transportation system, volume of vehicles, road crash data, and government policies, rules, and regulations, as well as the quality of implementation. Moreover, traffic congestion prediction models were developed, using logistic regression, random forest, and neural networks. Seventeen months of daily traffic data were used in developing the models. Results showed that the Random Forest models have recorded the highest accuracy (77%), recall (77%) and F1-score (77%). On the other hand, the Neural Network model has better performance in predicting Free Flow traffic congestion at 81% F1-score, while the Random Forest model showed better results in predicting Moderate, Heavy, and Standstill Traffic.

Informing traffic enforcement leniency and discretion: Crash culpability and the effectiveness of written warnings versus citations

ObjectivesDeterrence of risky driving behavior is important for the prevention of crashes and injuries. Traffic law enforcement is a key strategy used to decrease risky driving, but there is little evidence on the deterrent effect of issuing warnings versus citations to drivers regarding the prevention of future crashes. The purpose of this study was to 1) investigate the difference between citations and written warnings in their association with future crash culpability and 2) investigate whether drivers who were issued written warnings or citations have different associations with future crash culpability likelihood than those without prior citations or written warnings. MethodsData for this study included Iowa Department of Transportation crash data for 2016 to 2019 linked to data from the Iowa Court Case Management System. A quasi-induced exposure method was used based on driver pairs involved in the same collision in which one driver was deemed culpable and one was non-culpable. Conditional logistic regression models were constructed to examine predictors of crash culpability. The main independent variable was traffic citation and warnings history categorized into moving warning, non-moving warning, moving citation, non-moving citation, or no citation or warning in the 30 days prior to the crash. ResultsThe study sample included a total of 152,986 drivers. Among drivers with moving violations, previously cited drivers were more likely to be crash culpable than previously warned drivers (OR = 1.64, 95% CI = 1.29–2.08). Drivers with prior non-moving citations were less likely to be the culpable party in a crash than a driver who had no recent warnings or citations (OR = 0.72, 95% CI = 0.58–0.89). Drivers with prior warnings (moving or non-moving) did not appreciably differ in crash culpability relative to drivers who had not received any citations or warnings in the previous 30 days. ConclusionsDrivers with prior moving citations were more likely to be culpable in a future crash than drivers with prior moving warnings, which may relate to overall driving riskiness as opposed to effectiveness of citations in deterring risky driving behaviors. Results from this study also suggest that officer discretion was being appropriately applied by citing the riskiest drivers, while giving lower risk drivers warnings. Results from this study may be useful to support strengthening of state driver improvement programming.

Effectiveness of bicycle helmets and injury prevention: a systematic review of meta-analyses

To mitigate the risk of injuries, many countries recommend bicycle helmets. The current paper seeks to examine the effectiveness of bicycle helmets by performing a systematic review focusing on meta-analyses. First, the current paper explores the findings of studies that employ meta-analyses using bicycle crash data. Second, the results are discussed considering the findings from research analyzing bicycle helmet effectiveness in a laboratory using simulation, and then are complemented with key methodological papers that address cycling and the overall factors contributing to the injury severity. The examined literature confirms that wearing a helmet while cycling is beneficial, regardless of age, crash severity, or crash type. The relative benefit is found to be higher in high-risk situations and when cycling on shared roads and particularly preventing severe head injuries. The results from the studies performed in laboratories also suggest that the shape and size of the head itself play a role in the protective effects of helmets. However, concerns regarding the equitability of the test conditions were found as all reviewed studies used a fifty-percentile male head and body forms. Lastly, the paper discusses the literature findings in a broader societal context.

Analysis of the injury severity of motor vehicle–pedestrian crashes at urban intersections using spatiotemporal logistic regression models

This paper conducted a comprehensive study on the injury severity of motor vehicle–pedestrian crashes at 489 urban intersections across a dense road network based on high-resolution accident data recorded by the police from 2010 to 2019 in Hong Kong. Given that accounting for the spatial and temporal correlations simultaneously among crash data can contribute to unbiased parameter estimations for exogenous variables and improved model performance, we developed spatiotemporal logistic regression models with various spatial formulations and temporal configurations. The results indicated that the model with the Leroux conditional autoregressive prior and random walk structure outperformed other alternatives in terms of goodness-of-fit and classification accuracy. According to the parameter estimates, pedestrian age, head injury, pedestrian location, pedestrian actions, driver maneuvers, vehicle type, first point of collision, and traffic congestion status significantly affected the severity of pedestrian injuries. On the basis of our analysis, a range of targeted countermeasures integrating safety education, traffic enforcement, road design, and intelligent traffic technologies were proposed to improve the safe mobility of pedestrians at urban intersections. The present study provides a rich and sound toolkit for safety analysts to deal with spatiotemporal correlations when modeling crashes aggregated at contiguous spatial units within multiple years.

Clustering and pedestrian crashes prediction modelling: Amman case

Pedestrian casualties are a severe domestic as well as international problem. This study analyses the spatial distribution of pedestrian casualties to define contributory factors and delineate the means for their prediction. Three years of crash data were collected along with other factors and analysed using kernel density estimation (KDE), spatial autocorrelation (Moran’s I), cluster K-Means, spatial regression, and general linear regressions (GLM). Kernel density estimate defines a cluster of pedestrian deaths within 1250 meters. According to Moran’s I, 17/22 attributes about casualties, road networks, demographics, and land use have positive values, indicating similar importance clustering. The spatial pattern of pedestrian casualties is random and insignificant and does not change with time. Casualties are negatively related to the surrounding attributes, indicating a tendency towards dispersion. A K-Means analysis of multiple variables revealed that when variables included in the clustering were higher, the variance explanation percentage was lower. In the multi-variable GLM assuming Poisson distribution, the road network length alone or with the house permits combined were the best predictors of casualties. Classic regressions were not significantly enhanced by spatial dimension, and none of the autoregressive coefficients were significant. The predictions from the Poisson-based GLM model are similar to the classic regressions.

GIS-based spatiotemporal analysis for road traffic crashes; in support of sustainable transportation Planning

Road traffic crashes pose a significant challenge worldwide, necessitating increased efforts to reduce them and promote sustainable transport systems. This study aimed to investigate spatiotemporal road traffic crashes and their causes in the State of Qatar by identifying hot spots of crashs and exploring whether they were primiarly attributed to behavioural practices and/or the geometrical design of roads and intersections. The study employed various methods, including Time-Space Cube analysis, Geographically Weighted Regression (GWR), Emerging Hot Spot analysis, and Spatial Autocorrelation analysis, with historical traffic crash data from 2015 and 2019. The findings indicated that crashes were mainly concentrated in the central-eastern region of Qatar and are related to driver behaviour. The analysis also revealed that crashes during the weekdays in 2019 were more strongly clustered than in 2015, suggesting a probable systematic cause of crashes. The results provide valuable information for policymakers to target high-incidence locations, prioritize interventions and develop more effective measures and policies to reduce crashs and promote a sustainable transportation system in Qatar. Overall, this study highlights the importance of continued research and policy development in this area and could potentially be applicable and transferable to similar regions.

Pedestrian- and bicyclist-involved crashes: Associations with spatial factors, pedestrian infrastructure, and equity impacts

Introduction: We analyze and compare the factors that influence the fatality of pedestrian and bicyclist involved crashes in New Jersey using available police-reported crash data between 2016 and 2020. Under three percent of crashes involve non-motorists statewide, but these account for about one third of all traffic fatalities in the state. Methods: Our analysis is broken down into five parts: we (1) analyze the relationship between minority and low-income communities and non-motorist involved crashes; (2) identify spatial differences between non-motorist involved crashes and non-motorist involved fatal crashes; (3) compare the factors affecting fatal pedestrian crashes in New Jersey and in four counties in southern New Jersey for which we have data on pedestrian infrastructure; (4) compare the factors affecting fatal pedestrian crashes and fatal cyclist crashes in New Jersey; and, (5) discuss priority areas for improving safety. Results: Crashes occur disproportionately more often in low-income communities. Moreover, we find that crashes are less likely to be geocoded if they take place in low-income and minority areas, a concerning finding considering that geocoded crashes are of paramount importance in identifying specific corridors for improvement. Light conditions, non-motorist age, posted speed, and vehicle type are significant factors influencing the fatality of non-motorist involved crashes. The proximity to a crosswalk or sidewalk is associated with decreased risk of a fatal crash for pedestrians. Cyclist crashes in low-income neighborhoods were more likely to be fatal – a finding that we attribute to lower access to bicycle facilities in low-income areas. Conclusions: We conclude with countermeasures, including a call for better data collection.

Modeling Pedestrian Injury Severity: A Case Study of Using Extreme Gradient Boosting Vs Random Forest in Feature Selection

Walking and bicycling are lauded for their negative net carbon impact and for their health benefits. However, national crash statistics suggest that pedestrians are disproportionately harmed in any vehicle–pedestrian conflict situation. Although automated transportation in the future is anticipated to increase overall safety, multiple incidents involving automated vehicles have been reported recently, indicating that the technology needs more training on real-world scenarios and conflicts. This research is motivated by the need for contextual data and related levels of harm in potential conflict scenarios in mixed traffic and we use a national police reported crash dataset, CRSS, to address this need. Our study uses a new gradient boosting algorithm, XGBoost, to identify important features among a host of seemingly significant variables. We compare the performance of XGBoost with the more frequently used random forest method and find that XGBoost is more reliable and robust for handling an unbalanced and sparse dataset like crash data, and the features extracted are more aligned to findings from previous research on the topic. We also compare feature importance between NASS-GES and CRSS—two national crash databases with different sampling strategies but the same objective—and find that sampling strategy influences selection of feature importance. We further use the features extracted using XGBoost in a multiclass logistic regression to quantify the effect of these features on different levels of pedestrian injury. Our findings indicate that speed limit, light conditions, pre-crash movements, and location of pedestrian are important contributors to crash severity, along with driver distraction and impairment.