A data-driven approach to child pedestrian crash analysis using dimension reduction, clustering, and explainable AI.

Little crash equity research has controlled for active transportation rates, examined the Southeastern United States, or used composite sociodemographic metrics. This study analyzed disparities in pedestrian and cyclist crashes and crash severity according to level of social vulnerability (SV) across South Carolina. Data about SV and its 4 dimensions (socioeconomic status, household composition and disability, minority status and language, and housing type and transportation) were compiled for all census tracts (N = 1103) within South Carolina. Data for all crashes involving a pedestrian (n = 10,688) and/or cyclist (n = 4802) from 2011 to 2021 were obtained from the South Carolina Department of Transportation and geocoded to the respective census tract. Total average pedestrian and cyclist crash severity (Equivalent Property Damage Only) were also calculated for each tract. Crash frequency and severity scores were adjusted using the average number of walking and cycling trips in the census tract per year using StreetLight Data. Mixed-model linear regression analyzed the relationships between overall SV and the 4 SV dimensions and 4 crash measures-pedestrian and cyclist frequency and severity. Stratified analyses were conducted for urban and rural tracts. Overall SV was positively and significantly associated with all 4 crash outcomes in urban areas: pedestrian crashes per trip (B = 0.048, SE = 0.012, P < .001), pedestrian crash severity per trip (B = 9.018, SE = 2.516, P < .001), cyclist crashes per trip (B = 0.093, SE = 0.029, P < .01), and cyclist crash severity per trip (B = 16.370, SE = 5.482, P < .01). In urban areas, greater SV was associated with more severe pedestrian and cyclist crash outcomes. Targeted policy and programmatic and infrastructure interventions are needed to improve active transportation safety and public health.

A novel approach for analyzing severe crash patterns on multilane highways

Driving at night is riskier in terms of crash involvement than it is during the day. Fortunately, it is clearly established that illumination on roadways can reduce the number and severity of nighttime crashes. However, state and municipal departments of transportation (DOTs) lack the available illumination data. Therefore, the objective of this research is threefold, as follows: (i) to develop machine learning models that use readily available roadway characteristic data to predict the severity of nighttime crashes; (ii) determine the effect that illumination has on crash severity; and (iii) develop a tool to assist DOT decision makers in collecting illumination data. To accomplish this objective, we have extracted data from the Texas Department of Transportation (TxDOT) Crash Record Information System (CRIS) database, which was then further split into a training and a test dataset. Then, seven machine learning techniques, namely binary logistic regression, k-nearest neighbors, naïve Bayes, random forest, artificial neural network, Extreme Gradient Boosting (XGBoost), and a Long Short-Term Memory (LSTM) model, were all applied to the unseen test data. The random forest model produced the most promising results by predicting severe crashes with 97.6% accuracy. In addition, we conducted a pilot study to test the collection of illumination data using a light meter. In the future, we aim to complete the development of a smartphone application, which can be used in conjunction with the random forest model presented in this paper, to collect crowdsourced illumination data and predict nighttime crash hotspots. This may assist DOT decision makers to prioritize funding for illumination at the hot spots.

Investigating the relationship between crash severity, traffic barrier type, and vehicle type in crashes involving traffic barrier

BackgroundDespite substantial progress, motor vehicle crashes remain a leading killer of US children. Previously, we documented significant positive impacts of Safe Routes to School interventions on school-age pedestrian and pedalcyclist...

Crash severity, as a major concern in the routing and scheduling of hazardous material shipments, has caused great loss of lives and property damage every year. Although abundant studies have been conducted to identify the relationship between different factors on crash severity, the analysis of the severity of hazard material transportation (HMT) crashes is very limited. Factors including road, vehicle, driver, and environment are not well considered in previous studies. This article analyzed the influence of various factors on HMT crash severity using Highway Safety Information System data. The random forest combined with the ordered logistic model is used for factor analysis. The results showed that annual average daily traffic, fatigues/asleep, number of lanes, speeding, adverse weather, and light are the six most important factors affecting HMT crash severity. Different from the non-HMT crashes, driver factor (e.g., driver age, gender, and drug/alcohol influence) was found to be not significantly related to crash severity. Speeding should be strictly forbidden for HMT drivers, considering the potential increased crash severity. Increasing the level of lighting can help reduce the number of severe crashes. The corresponding recommendations were provided based on the regression results.

Urban crash-related child pedestrian injury incidence and characteristics associated with injury severity

Objective: The objective in this study, using data from crashed cars fitted with on-board crash pulse recorders, was to present differences in average crash severity, distribution of crash severity, and injury outcomes, based on an independent safety rating of roads, also taking road type and speed limit into consideration. Furthermore, the objective was to evaluate differences in injury risk, based on the distribution of crash severity. Methods: The investigation included both frontal two-vehicle crashes and single-vehicle crashes with known injury outcome. In total, 209 real-world crashes involving cars fitted with crash pulse recorders were included. For all crashes, average mean acceleration and change of velocity of the vehicle acceleration pulse were measured and calculated. All crash spots were classified according to an independent road safety rating program (European Road Assessment Programme Road Protection Score), where the safety quality of roads is rated in relation to posted speed limits. The crash severity and injury outcome in crashes that occurred on roads with good safety ratings were compared with crashes on roads with poor safety ratings. The data were also divided into subcategories according to posted speed limit and road type, to evaluate whether there was a difference in crash severity and injury outcome within the categories. Results and Conclusions: In total, crash severity was statistically significantly lower in crashes occurring on roads with good safety ratings than in crashes occurring on roads with poor safety ratings. It was found that crash severity and injury risk were lower on roads with good safety ratings with a speed limit of above 90 km/h compared with roads with poor safety ratings, irrespective of speed limit. On the other hand, crash severity was higher on roads with good safety ratings with speed limit of 70 km/h than on roads with poor safety ratings with the same speed limit. Though it was found that a higher speed limit resulted in higher crash severity on roads with poor safety ratings, the opposite was found on roads with good safety ratings. The main reason for this was that lanes for traffic traveling in opposite directions were more often separated at higher speeds on roads with good safety ratings. On divided roads with good safety ratings, there were no crashes resulting in crash severity above the level corresponding to a 10 percent risk of sustaining serious or fatal injury. This indicates that one of the most important safety measures is divided roads.

&lt;div&gt;As pedestrians are among the most critical road users, this research analyzes their vulnerability characteristics and predicts the injury severity of pedestrian crashes through decision tree techniques, rather than using statistical regression models that have particular predefined causal relationships between dependent and independent variables. Five years have been studied in Nablus Governorate/Province (2012–2016), one of 16 governorates in Palestine, as a case study based on reported crash frequencies for developing countries. Tree techniques (CART [Classification and Regression Tree] and CHAID [Chi-Square Automatic Interaction Detector]) were applied to extract the main impacting factors on injury severity for pedestrian crashes. The main contributions considered a small regional context in developing countries and found differences between the results of various methods in injury severity. Fourteen independent variables have been analyzed. A CART model with Gini splitting has produced the most accurate model. The most important variables were the victim’s gender, followed by area classification as rural, and the age categories of pedestrians older than 65 and younger than 15 years. The least important variables were found to be the driver’s gender, land use, and pavement conditions. Results also showed that the proximity of crashes to schools is relatively high; therefore, some policies were suggested regarding children’s awareness, school zones, and driver behavior. It was found that the majority of factors influencing pedestrian crashes are related to human characteristics such as age, gender, or attitude whereas, in developed countries, they were related to vehicles and infrastructure. Based on the results of the study, tree techniques were considered effective in the analysis of injury severity of pedestrians in the context of developing countries to identify the main factors of vulnerability.&lt;/div&gt;

This study investigated the main factors affecting the severity of injury to pedestrians in taxi–pedestrian crashes on urban arterial roads. Video data recorded by an in-car black box were used. Because the video data provided direct crash observation, they were more reliable than the crash data, and video images and speed profiles retrieved from the black box were advantageous for safety studies. For analysis of the black box data, this study defined new explanatory variables that affected injury severity; these variables could not have been identified by the conventional method, which was based on crash reports. A multiple-indicator and multiple-cause model was used to investigate the relationship between the explanatory variables and injury severity. A total of 484 taxi–pedestrian crash scenes over 2 years was used for the multivariate analysis in the city of Incheon, South Korea. The crash characteristics most strongly associated with increased crash severity were failure by the pedestrian to watch for approaching vehicles, jaywalking by the pedestrian, the pedestrian being elderly, excessive vehicle speed, failure by the driver to immediately stop, limited driver vision, and nighttime. This study emphasized the potential of individualized black box video recording data for crash severity analysis and investigation of the causal factors of crashes.

This study investigates the relationship between crash severity and risk factors, addressing the limitations of conventional methods in capturing spatial and temporal heterogeneity. We developed and compared five binary logit models: a basic model, temporal heterogeneity model, spatial heterogeneity model, spatiotemporal heterogeneity, and a partially constrained spatiotemporal heterogeneity model. The latter leverages the posterior distribution properties to ensure similar adjacent time and space regions share distribution. These models were calibrated and validated using 3,222 pedestrian crashes occurring at mid-blocks in Hong Kong (2010-2019). The results showed that both the unconstrained and partially constrained spatiotemporal models outperformed others in fit and predictive accuracy. Temporal heterogeneity analysis revealed evolving risk factors, while spatial heterogeneity analysis identified hot zones for crash severity. Key risk factors included pedestrian age, head injury, pedestrian actions, driver operations, vehicle type, and the first collision point. These insights support targeted, region- and time-specific pedestrian strategies.

This study examined the spatial, temporal, and predictive dimensions of traffic crashes in Somerville, Massachusetts, to contribute to the evidence base for Vision Zero initiatives. Using crash data, spatial autocorrelation analysis, temporal distribution assessments, and machine learning models, the research identified patterns and predictors of crash severity. Spatial analysis revealed significant clustering of crashes, particularly around intersections and arterial corridors, underscoring the structural vulnerabilities of the urban road network. Temporal analysis demonstrated that crashes peaked during midday and afternoon hours, with seasonal spikes in January, September, and December, reflecting the combined influence of commuting cycles, weather conditions, and increased travel activity. Predictive modeling using logistic regression, random forest, and gradient boosting highlighted the challenges of forecasting severe crashes in imbalanced datasets. While gradient boosting achieved high accuracy (0.885) and precision (0.959), its ROC‑AUC score of 0.50 indicated poor discriminatory power, revealing a bias toward the majority class. The confusion matrix further confirmed that severe crashes were frequently misclassified as non‑severe, limiting the model’s utility for proactive safety interventions. Nonetheless, feature importance analysis identified time of day, intersection type, and pedestrian involvement as key predictors of crash severity, reinforcing the systemic interplay of temporal, infrastructural, and behavioural factors. The implications extend to Vision Zero frameworks, suggesting that traffic safety outcomes are embedded within discernible patterns that can be studied and anticipated. This study underscores the value of spatiotemporal and predictive analyses in advancing the understanding of urban traffic safety and informing long‑term strategies to eliminate severe crashes.

Electric vehicles (EVs) are increasing in popularity and are among the fastest-growing segments in new vehicle sales. Battery-electric vehicles (BEVs) are the highest-selling class of EV. BEVs use a unique frame construction (i.e., a “skateboard” chassis) with a heavy battery pack located below the occupant’s floorpan, and are heavier and have lower centers-of-gravity than internal combustion engine vehicles (ICEVs)—factors that could affect their crashworthiness with roadside features. A crash study was performed to compare the crash conditions for BEVs to ICEVs and hybrid-electric vehicles (HEVs) covering the numbers of vehicles, posted speed limits, crash locations, time of day, weather conditions, crash severities, and roadside fixed objects struck. Eight state departments of transportation provided crash data spanning 5 years that included vehicle identification numbers or make, model, and year combinations, which were used to isolate BEVs from ICEVs. Crash results were reviewed and EV crash conditions were found to be similar to those of ICEV and HEV crash conditions. Moreover, the frequency of run-off-road crashes was similar for BEVs and ICEVs. Because no substantial differences were observed between the two datasets, it is recommended that impact conditions described in the Manual for Assessing Safety Hardware be utilized during preliminary evaluation of roadside hardware using electric vehicles.

Discovering injury severity risk factors in automobile crashes: A hybrid explainable AI framework for decision support

The rapid advancement of autonomous vehicle systems (AVS) has introduced complex challenges to road safety. While some studies have investigated the contribution of factors influencing AV-involved crashes, few have focused on the impact of vehicle-specific factors within AVS on crash outcomes, a focus that gains importance due to the absence of a human driver. To address this gap, the advanced machine learning algorithm, LightGBM (v4.4.0), is employed to quantify the potential effects of vehicle factors on crash severity and collision types based on the Autonomous Vehicle Operation Incident Dataset (AVOID). The joint effects of different vehicle factors and the interactive effects of vehicle factors and environmental factors are studied. Compared with other frequently utilized machine learning techniques, LightGBM demonstrates superior performance. Furthermore, the SHapley Additive exPlanation (SHAP) approach is employed to interpret the results of LightGBM. The analysis of crash severity revealed the importance of investigating the vehicle characteristics of AVs. Operator type is the most predictive factor. For road types, highways and streets show a positive association with the model’s prediction of serious crashes. Crashes involving vulnerable road users can be attributed to different factors. The road type is the most significant factor, followed by precrash speed and mileage. This study identifies key predictive associations for the development of safer AV systems and provides data-driven insights to support regulatory strategies for autonomous driving technologies.