Traffic Inj Prev [JOURNAL]
Sun
200 papers
Examination of risk factors in bus accidents: merging the AcciMap framework with natural language processing methods.
Wang R, Niu X, Zhao D … +1 more , Wang J
Traffic Inj Prev · 2025 Dec · PMID 41401396 · Publisher ↗
OBJECTIVE: This study develops and validates an automated, text-driven pipeline that integrates Natural Language Processing (NLP) with the AcciMap systemic-accident framework to identify, cluster, and quantify multi-leve...
OBJECTIVE: This study develops and validates an automated, text-driven pipeline that integrates Natural Language Processing (NLP) with the AcciMap systemic-accident framework to identify, cluster, and quantify multi-level risk factors in Chinese bus-accident reports. The aims were to (1) automatically extract causal factor statements from unstructured accident narratives, (2) map extracted factors onto the six socio-technical AcciMap levels, and (3) construct a quantitative, network-based representation of cross-level causal linkages to inform targeted safety interventions. METHODS: We analyzed 127 Chinese bus accident reports (2019-2024) from the Safehoo platform. A domain-adapted NLP pipeline was implemented: (1) Named Entity Recognition (NER) using BiLSTM-CRF with lexicon-enhanced tokenization to extract factor instances across six AcciMap levels; (2) Generating semantic embeddings using CoSENT-fine-tuned BERT (Bidirectional Encoder Representations from Transformers); (3) Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN), optimized Optuna Bayesian search, identified 18 distinct risk factor categories; (4) AcciMap network models quantified relationships using Total Co-occurrence Indices (TCI) and centrality metrics (degree, betweenness, closeness, eigenvector). Core nodes were identified using composite centrality scores (80th percentile threshold). RESULTS: The pipeline produced 18 interpretable factor clusters spanning government, regulator, enterprise, management, employee, and work environment levels. Network analysis of the TCI matrix revealed regulatory deficits and employee behaviors as primary hubs: The three highest composite centrality scores corresponded to driver operational errors (emp2), weak driver safety awareness (emp1), and formalistic regulatory supervision (reg1). Emp2 and emp1 had the largest degree and closeness centralities (emp2 degree = 25.8; emp1 degree = 24.5), and emp2 showed the highest betweenness (0.36), indicating its role in bridging disparate factor groups. Government-level factors were relatively peripheral but contributed indirectly by weakening regulatory and enterprise controls. Environment factors (e.g., aging infrastructure, reduced visibility) acted as triggering conditions with moderate centrality. Strong propagation paths highlighted reg1-mgmt1-emp2 and reg1-ent2 as frequent, high-weight associations. CONCLUSIONS: Combining CoSENT-enhanced sentence embeddings, unsupervised clustering, and AcciMap-based network quantification enables scalable extraction and systemic interpretation of causal factors from unstructured bus-accident narratives. Results indicate that strengthening regulatory enforcement, improving enterprise safety management and managerial accountability, and targeting frontline driver behaviors should be coordinated to break multi-level risk propagation chains. The pipeline offers a reproducible method for automated, policy-relevant causal analysis and can guide prioritized interventions and monitoring in urban passenger transport safety.
Joint assessment of casualty scale and injury severity in single-vehicle crashes with passenger injuries: A heterogeneity approach considering comprehensive severity.
Meng Y, Wang Y, Chen H … +1 more , Wang L
Traffic Inj Prev · 2025 Dec · PMID 41397286 · Publisher ↗
OBJECTIVES: Focusing on the phenomenon of single-vehicle crashes resulting in severe injuries and numerous casualties, this study investigates the influencing factors of occupant casualties in single-vehicle crashes by e...
OBJECTIVES: Focusing on the phenomenon of single-vehicle crashes resulting in severe injuries and numerous casualties, this study investigates the influencing factors of occupant casualties in single-vehicle crashes by employing an integrated analysis framework incorporating injury severity and casualty scale. METHODS: A joint assessment of both injury severity and casualty scale in single-vehicle crashes involving passenger injuries is conducted using a k-Prototypes-random parameter logit model with heterogeneity in means and variances (K-P-RPLHMV). RESULTS: The statistical model identified 14 crash-significant variables with fixed parameters. It is noteworthy that driver home area exhibit significant heterogeneity characteristics. The sources of heterogeneity include vehicle age exceeds 10 years and driver have intermediate socioeconomic. CONCLUSIONS: This study contributes to the existing literature not only by providing a comprehensive characterization of factors influencing both injury severity and casualty scale, but also by proposing targeted countermeasures for traffic management authorities and drivers.
Spatial variation characteristics of drivers' visual search behavior at urban expressway exits.
He Z, Fan D
Traffic Inj Prev · 2025 Dec · PMID 41397283 · Publisher ↗
OBJECTIVES: This study aims to reveal the spatial variation characteristics of drivers' visual search behavior in urban expressway exit areas by analyzing the evolution patterns of their fixation characteristics and fixa...
OBJECTIVES: This study aims to reveal the spatial variation characteristics of drivers' visual search behavior in urban expressway exit areas by analyzing the evolution patterns of their fixation characteristics and fixation transition patterns along the driving path. The ultimate goal is to establish a human factors foundation for enhancing safety and optimizing the design of traffic facilities in these areas. METHODS: Two types of exit ramps-short and long-were selected as experimental scenarios. A naturalistic driving experiment was conducted, collecting eye-movement data from 25 participants. Their eye-movement data were analyzed by dividing the visual field into three primary Areas of Interest (AOIs): Guiding, In-car, and Look-ahead fixations, and by segmenting the exit routes spatially. Key metrics, including the proportion of fixation duration, fixation entropy, fixation transition probability, and fixation transition entropy, were calculated for each segment to quantify visual search behavior. RESULTS: As vehicles approached the exit, visual attention progressively concentrated on Look-ahead fixations, peaking at 91% within 50-100 meters before the exit. Drivers exhibited high fixation entropy and broader search dispersion in two critical sections: from ramp entry to 50 m past the gore point, and 50 m before the ramp terminal. However, attention to Look-ahead fixations increased again toward the end of the ramp, indicating a stabilization of visual scanning patterns. For short ramps, drivers primarily focused on Guiding fixations, with fixation duration exceeding 50% across all segments from the gore point to the ramp terminal, and the transition probabilities from both Look-ahead fixations and In-car fixations to Guiding fixations exceeding 56%. In contrast, drivers navigating long ramps exhibited higher fixation entropy and fixation transition entropy, reaching local peaks before entering curved sections. CONCLUSIONS: This study reveals that drivers dynamically adjust their visual search strategies based on their spatial position at urban expressway exits. As the vehicle approaches the exit, visual attention prioritizes information acquisition for decision-making, before shifting back to vehicle control near the exit point. In short ramps, drivers exhibited a heightened demand for lane control, whereas the complex geometry of long ramps increased task complexity, necessitating simultaneous information search and vehicle control and leading to more sophisticated visual strategies. These findings elucidate the intrinsic link between drivers' visual search behavior and their spatial positioning, providing a human factors foundation for optimizing the spatial compatibility of traffic facilities.
Driving under the influence of alcohol and cannabis: Associations with substance use and behavioral health characteristics.
Choi NG, Marti CN
Traffic Inj Prev · 2025 Dec · PMID 41397281 · Publisher ↗
OBJECTIVES: Driving under the influence of alcohol (DUIA) and cannabis (DUIC) continues to increase. Given the increased number of states with cannabis legalization in recent years, updates on driving under the influence...
OBJECTIVES: Driving under the influence of alcohol (DUIA) and cannabis (DUIC) continues to increase. Given the increased number of states with cannabis legalization in recent years, updates on driving under the influence (DUI) prevalence and risk factors are needed to inform targeted prevention and intervention efforts aimed at reducing substance-impaired driving. METHODS: We used data from the 2021-2023 U.S. National Survey on Drug Use and Health ( = 139,524 individuals age 18 years and older) for self-reported DUI. After presenting the prevalence of alcohol, cannabis, and other substance use and DUI, we used binary logistic regression models to examine sociodemographic and clinical correlates of DUIA and DUIC, and a multinomial logistic regression model to examine DUIC but no DUIA (DUIC-NA) and both DUIA and DUIC (DUIA&C), compared to DUIA but no DUIC (DUIA-NC). RESULTS: Among those who used alcohol in the past year, 8.6% had DUIA; among those who used cannabis, 20.6% had DUIC. Among those who reported DUIA and/or DUIC, 47.9% had DUIA-NC, 34.0% had DUIC-NA, and 18.1% had DUIA&C. Any severity of alcohol and cannabis use disorder, initiation of substance use during adolescence, mental health problems, risk propensity, self-recognition of substance use problem, and criminal justice involvement were associated with DUIA and DUIC. Age 65+, relative to age 18-25, was also a risk factor for DUIC among those who used cannabis. Compared to DUIA-NC, DUIC-NA was higher among residents of medical cannabis legal states. CONCLUSIONS: Substance use treatment, including mental health screening and treatment, is the most important DUI prevention approach. As cannabis use rises among older adults, prevention strategies must increasingly include this group. States with legalized medical cannabis suggest a need for complementary public safety measures.
Functional brain network-based identification of special weather conditions on grassland road.
Gao M, Liang Y, Li H
Traffic Inj Prev · 2025 Dec · PMID 41397265 · Publisher ↗
OBJECTIVE: The grassland roads in Inner Mongolia have the characteristics of low traffic density, simple road geometry, and few environmental disturbances, making them an ideal scenario for the research of autonomous dri...
OBJECTIVE: The grassland roads in Inner Mongolia have the characteristics of low traffic density, simple road geometry, and few environmental disturbances, making them an ideal scenario for the research of autonomous driving technology. However, the frequent crossing of livestock and the high incidence of sandstorms and rain and snow pose significant challenges to the autonomous driving system. Currently, the system is still in the conditional autonomous driving stage, and it still requires the driver to take over in time. The takeover process involves complex cognitive activities involving multiple brain regions. Analyzing the brain functional network characteristics can help reveal the takeover mechanism. Further, combining brain network features with machine learning algorithms can improve the ability to identify takeover states under different weather conditions, providing more adaptive takeover decision support for the system. METHODS: A driving simulation experiment was conducted, with three weather conditions: sunny, sandstorm, and rain and snow. The typical takeover event was livestock crossing the road, and the participants' takeover response time and electroencephalogram (EEG) signals were recorded. The phase lag index was used to construct a phase synchronization network, and two-factor repeated measures variance analysis was performed on the topological characteristics of the brain network in the θ, α, and β frequency bands. Finally, a K-nearest neighbor algorithm was used to establish a weather classification model. RESULTS: The takeover response time was the longest in the sandstorm condition, indicating that visual interference inhibited cognitive processing and executive control; severe weather weakened the connections between brain regions, especially in the β frequency band, affecting the stability of takeover. In the rain and snow, the integration of the θ and α frequency band networks was enhanced, which was conducive to information processing; in the sunny, the connections between brain regions remained stable, the cognitive load was lower, and thus the takeover performance was better. The accuracy of the KNN model was 96%, with good generalization ability, and it can be used for special weather recognition in autonomous driving. CONCLUSIONS: The analysis based on brain functional networks not only helps to reveal the takeover mechanism in complex weather environments but also provides neurophysiological support for the optimization of autonomous driving strategies under special weather conditions.
Segment-based truck crash risk analysis on highways: clustering and logistic regression for safety enhancement.
Kim K, Hong J
Traffic Inj Prev · 2025 Dec · PMID 41397262 · Publisher ↗
OBJECTIVE: Truck-involved crashes on highways represent a significant threat to traffic safety because of their disproportionate severity and frequency. Existing crash analyses frequently overlook the spatial heterogenei...
OBJECTIVE: Truck-involved crashes on highways represent a significant threat to traffic safety because of their disproportionate severity and frequency. Existing crash analyses frequently overlook the spatial heterogeneity of highways by treating them as uniform networks, thereby limiting their effectiveness in identifying risk-prone areas. This study proposes a segment-level risk assessment framework that integrates machine-learning-based clustering with logistic regression modeling to evaluate truck crash patterns on South Korean expressways from 2020 to 2022. METHODS: Four clustering techniques-K-means, DBSCAN, Gaussian mixture model, and hierarchical clustering-were applied based on the crash frequency, severity metrics, and traffic exposure indicators. Among these, K-means clustering offered the most interpretable segmentation, identifying six distinct highway segment groups. Logistic regression was then used to examine the influence of various risk factors, such as driver fatigue, overloading, geometric features, and nighttime driving, on crash occurrences within each cluster. RESULTS: The analysis revealed that crash risks varied considerably across segments: fatigue-related crashes were concentrated in freight-dense corridors; brake and overloading failures were more prevalent on steep or curved segments; and nighttime crashes were notably higher on port-access routes. CONCLUSIONS: These findings support the implementation of targeted safety strategies tailored to local roadway characteristics, including rest-area expansion, automated fatigue detection, and lighting improvement. The proposed methodology offers a practical data-driven approach for transportation agencies seeking to prioritize high-risk highway segments and implement more effective crash prevention policies.
Analysis of head injury risk in vehicle to pedestrian collisions using a large-scale simulation dataset.
He W, Yong H, Dongmei G … +1 more , Changchun Z
Traffic Inj Prev · 2025 Dec · PMID 41369161 · Publisher ↗
OBJECTIVE: Head injury is the main cause of pedestrian death in traffic accidents, and rapid prediction of injury severity can win more time for accurate treatment and vehicle decisions. This paper aims to establish an a...
OBJECTIVE: Head injury is the main cause of pedestrian death in traffic accidents, and rapid prediction of injury severity can win more time for accurate treatment and vehicle decisions. This paper aims to establish an applicable large-scale numerical database for the development of head injury prediction models. METHODS: A refined multi-body vehicle model was first constructed to consider the influence of vehicle front-end stiffness properties. Then a Head Weighted Injury Study (HWIS) Database containing 60,750 simulations was established by using a full factorial design of experiments with car speed, car type, collision position, collision angle, pedestrian speed, pedestrian size, and pedestrian gait as input variables, and peak linear acceleration, HIC, peak angular velocity and acceleration as outputs. The distribution of head impact locations and injury severity in the numerical database was counted, and the effects of different factors on the risk of head injuries were further qualitatively analyzed. RESULTS: The results showed that the head impact locations were distributed in almost all regions of the vehicle's front-end, and the proportion of each head injury level exceeded 10%, which demonstrated the balance and applicability in the database. In vehicle to pedestrian collisions, pedestrian size significantly influences head linear and rotational kinematic metrics. Considering the scaling, the proportion of severe head injuries in the 6YOC and AF05 was 1.37 times and 1.39 times higher than in the AM50, respectively. CONCLUSIONS: This study established a reliable large-scale database for the development of head injury risk prediction models. Through analyzing the effects of various factors on head injury risk indicates that in headform testing, for regions with WAD values between 1,000 and 1,500 millimeters, the performance criteria for each grid point should be reevaluated to accommodate the HIC tolerance limits for children and small-size pedestrians.
Beyond helmet use: enhancing fit and safety in a youth bicycling program.
Ghanbari A, Jansson S, Pae G … +6 more , Cavanaugh J, Dusil R, Spears S, Plumert JM, Peek-Asa C, Hamann C
Traffic Inj Prev · 2025 Dec · PMID 41369140 · Publisher ↗
OBJECTIVE: Bicycling is a popular form of transportation and recreation among youth, yet head injuries remain one of the most common and serious outcomes in bicycle-related crashes. While helmet use can substantially red...
OBJECTIVE: Bicycling is a popular form of transportation and recreation among youth, yet head injuries remain one of the most common and serious outcomes in bicycle-related crashes. While helmet use can substantially reduce head injury risk, the protective benefit is compromised when helmets are worn improperly. Educational interventions aimed at both helmet use and correct fit remain limited, particularly those using objective, naturalistic methods for evaluation. This study aimed to evaluate the effectiveness of a school-based bicycle safety program, with and without a parental component, in improving helmet use and proper helmet fit among adolescents. METHODS: This randomized controlled trial evaluated a bicycle safety program among adolescents aged 9-12 years. Participants were assigned to one of two intervention groups, Bike Club and Bike Club Plus, or a control group. Both intervention groups received a 12-h program combining safety education and practical bicycling skills; Bike Club Plus included an additional parent session. Participants' bicycle rides were video recorded for one week before and one week after the intervention. For each ride, helmet use fit (position, side/chin straps, buckle) was coded. Generalized linear models fit using weighted generalized estimating equations were used to evaluate changes in helmet fit and use across the three study groups. RESULTS: Helmet use was high at baseline across all groups (≥90%) and remained stable post-intervention. Proper helmet positioning improved in both intervention groups relative to their baseline (Bike Club: odds ratio [OR] = 1.52, 95% confidence interval [CI] 1.02-2.26; Plus: OR = 1.60, 95% CI: 1.04-2.44), although differences between study arms were not statistically significant. Side/chin strap fit remained low overall, though improvements were greatest in the Plus group. Stratified analyses revealed that boys in the Plus group benefited most, suggesting sex-specific responses to different components. CONCLUSIONS: This study provides evidence that a school-based intervention, particularly when enhanced with a parental component, may improve aspects of helmet fit among youth cyclists. Future interventions should consider tailoring strategies by sex and target specific aspects of helmet fit, not just usage.
Analyzing crash severity through impairment and protection: A hybrid XGBoost-Bayesian network approach.
Dumka A, Kandiboina R, Joshi A … +2 more , Hawkins N, Sharma A
Traffic Inj Prev · 2025 Dec · PMID 41363838 · Publisher ↗
OBJECTIVE: This study analyzes how impairment, protection status, vehicle type, age group, and road characteristics collectively influence crash severity. While previous research has examined these factors in isolation,...
OBJECTIVE: This study analyzes how impairment, protection status, vehicle type, age group, and road characteristics collectively influence crash severity. While previous research has examined these factors in isolation, this study adopts a hybrid framework combining XGBoost to identify key predictors with Bayesian networks for modeling conditional dependencies and estimating risk under various scenarios. Additionally, the study introduces a temporal dimension by comparing crash severity patterns across pre-COVID and post-COVID periods using relative risk scores. This integrated approach supports data-driven, scenario-specific, and time-sensitive safety interventions to mitigate crash severity and improve road safety. METHODS: Historical crash data from Iowa (26,111 records for fatal, major, and minor crashes) were analyzed using 14 variables covering infrastructure, environment, driver, and vehicle characteristics. An XGBoost model was applied for feature selection, with SHAP values used to interpret the most influential predictors. A Bayesian network, built using the tree-augmented naive Bayes method, was then used for probabilistic inference. The Bayesian network's conditional probability tables were leveraged to compute relative risk scores under various evidence settings. A temporal analysis was conducted by segmenting the data into pre-COVID and post-COVID periods, enabling a comparative assessment of evolving crash risk patterns across user categories. RESULTS: The study identifies impairment and lack of protection as key drivers of crash severity, with relative risk scores differentiating high-risk groups. The temporal analysis comparing pre- and post-COVID periods reveals a consistent rise in relative risk scores post pandemic, underscoring shifts in crash risk patterns and reinforcing the need for adaptive, data-driven safety interventions over time. CONCLUSIONS: This study examines the interplay among protection status, impairment, driver demographics, vehicle type, and road characteristics, providing a deeper understanding of how these factors collectively influence crash severity. In addition to analyzing these complex relationships, the study introduces a probabilistic and robust framework that moves beyond traditional regression models. By integrating XGBoost for feature selection and Bayesian networks for conditional probability modeling, this approach captures conditional dependencies among key factors, enhancing interpretability. In addition to scenario-based relative risk estimation, the study introduces a temporal component by comparing pre-COVID and post-COVID crash patterns. The findings support data-driven, targeted interventions and policies, offering a flexible and interpretable tool for monitoring crash severity trends and guiding effective safety strategies over time.
Evaluating submarining in female human body models: Predicting occurrence and timing through biomechanics and restraint metrics.
Kalmar Gonzalo AJ, Von Kleeck W, Robinson A … +2 more , Koya B, Gayzik FS
Traffic Inj Prev · 2025 Dec · PMID 41363837 · Publisher ↗
OBJECTIVE: Submarining occurs during vehicle crash when the lap belt slips over the pelvis and loads the abdomen. Countering this motion is an interdisciplinary challenge, requiring expertise in biomechanics and restrain...
OBJECTIVE: Submarining occurs during vehicle crash when the lap belt slips over the pelvis and loads the abdomen. Countering this motion is an interdisciplinary challenge, requiring expertise in biomechanics and restraint design to develop countermeasures and quantitative metrics to predict it. Most prior research into submarining has focused on the 50 percentile male. This study compares small female models to reference sled tests and evaluates two methods of identifying submarining occurrence and timing in Human Body Models (HBMs). METHODS: A 5th percentile female model and age-adjusted 70-year-old 5th percentile female model (GHBMC) were simulated in sled environments based on Trosseille et al. Three seating configurations were run, two of which were designed to induce submarining. Submarining was assessed in the HBMs in two ways; analyzing the divergence between pelvis strain and belt force timing CORA's phase score, and analyzing an internally derived metric based on abdominal organ strain energy density (SED) time history, which is independent of restraint system-based measures. RESULTS: HBMs compared well to experimental corridors and time-history data (average CORA score of 0.73). Like the PMHS, neither model submarined in the low-speed case, while both submarined in the high-speed cases. The pelvic strain and belt force phase correlation scores were on average 0.998 (out of 1) in the non-submarining cases and 0.217 in the submarining cases. The peak SED values were on average 12 times greater in the submarining cases than non-submarining cases. The models tended to predict submarining earlier than the experiments. CONCLUSIONS: This study illustrates that both restraint based and internally derived HBM metrics can be used to predict the occurrence and timing of submarining. It further addresses a gap in the literature to relate to female HBM validation in frontal impact conditions. The findings will enable engineers to more effectively utilize HBMs to contribute to the development of safer vehicles, linking biomechanics to restraint design.
Belted driver fatalities in frontal crashes in 2017-2023 CISS.
Viano DC
Traffic Inj Prev · 2025 Dec · PMID 41359003 · Publisher ↗
OBJECTIVE: The study analyzed lap-shoulder belted driver fatalities in frontal crashes over 7 years (2017-2023) and evaluated the usefulness of NHTSA's 90 km/h oblique OMDB (offset moving deformable barrier) crash test t...
OBJECTIVE: The study analyzed lap-shoulder belted driver fatalities in frontal crashes over 7 years (2017-2023) and evaluated the usefulness of NHTSA's 90 km/h oblique OMDB (offset moving deformable barrier) crash test to bring about meaningful reductions in fatalities. METHODS: NHTSA online database of CISS (Crash Investigation Sampling System) crashes was searched for driver fatalities in frontal crashes (0 deg PDOF) with delta ≥ 48 km/h (30 mph). Each case was downloaded and summarized for: 1) vehicle make, model and year and crash damage, 2) the road type, weather and lighting conditions, 3) the pre-crash maneuvers and type of frontal damage, 4) driver demographics, alcohol-drug use and number of occupants in the vehicle, 5) restraint performance, steering system damage and intrusion and 6) the injury severity, body region, type and source. RESULTS: There were 16 driver fatalities with known lap-shoulder belt use in 2017-2023 CISS. The seatbelt wearing rate was 30.8%. The vehicles were 12.2 ± 7.3 years old with 68.8% ≥10 years old. The delta V was 72.2 ± 15.6 km/h (44.8 ± 9.7 mph) with 101.4 ± 30.5 cm of maximum frontal crush. Most (68.8%) of the crashes occurred on 2-lane roads. The crashes clustered into: 1) near-center impacts with a tree or pole (31.3%), 2) small overlap impacts with another vehicle (31.3%), 3) full width frontal impacts (25%) and 4) oblique or offset impacts with frame engagement (12.5%). The drivers were 61.5 ± 15.5 years old. Many were obese (61.5%). Most (68.8%) were alone in the vehicle. Four used alcohol (25%), two were also positive for drug use. Most drivers (57.1%) had severe chest injuries MAIS 4-6 with multiple rib fractures, some with heart and aorta lacerations. The most common injury source (61.5%) was the shoulder belt and steering wheel. In many cases (62.5%), the steering wheel was deformed and/or the column was displaced. The upper body kinetic energy was 8,698 ± 4,178 J in the fatal crashes. CONCLUSIONS: The drivers were typically alone, obese and older. Most vehicles were older; they typically moved left (75%) into an oncoming vehicle or off road hitting a tree or pole. The combination of collision severity and driver BMI pulled webbing from the load-limiting retractor (when equipped), compressed the airbag and deformed the steering wheel resulting in severe chest injuries (85.7%). The restraints allowed too much excursion into the steering system to protect the drivers. NHTSA's 90 km/h oblique OMDB test involves 4,516 ± 535 J of upper body energy and 57.6 ± 6.8 cm crush. It is not severe enough to bring about meaningful changes in vehicle structures or restraints. A 113 km/h OMDB impact has upper body energy consistent with the fatal crashes, but further testing is needed to determine if a higher speed OMDB test is useful. A 72 km/h (45 mph) sled test with the 95 or 82 km/h (51 mph) with the 50 dummy would be useful because it has upper body kinetic energy to bring about meaningful design changes to reduce driver fatalities.
Examining the association between clinical tests and simulated driving performance in long-haul truck drivers.
McKeown ML, Lemke MK, Crizzle AM
Traffic Inj Prev · 2025 Dec · PMID 41343223 · Publisher ↗
OBJECTIVES: The objectives of this study are to profile LHTD health, clinical test scores, and simulated driving performance; examine associations between clinical test scores and simulated driving errors in LHTD; and ex...
OBJECTIVES: The objectives of this study are to profile LHTD health, clinical test scores, and simulated driving performance; examine associations between clinical test scores and simulated driving errors in LHTD; and examine differences between LHTD simulated driving performance in different weather, lighting, and traffic conditions. METHODS: LHTD were recruited from various provincial and federal trucking associations and trucking companies across Canada. A sample of 36 LHTD completed a demographic questionnaire, objective health measures, a battery of cognitive, visual, and motor tests, and two simulated drives with different environmental conditions. RESULTS: The mean age of the sample was 47.9 ± 12.3 years (range 22-69); 94.4% were men. When operating in nighttime, rural, and winter conditions, LHTD made significantly more lane maintenance and speed regulation errors compared to the daytime, urban, and summer drive. In contrast, LHTD made significantly more signaling errors during the daytime urban drive compared to the nighttime, winter drive. Together these findings show that LHTD face significant challenges in a variety of driving environments. When combining both simulated drives, our findings show that poorer TMTA scores were significantly associated with more speed regulation (i.e., over-speeding and hard braking) and total driving errors. Additionally, poorer TMTB and UFOV-2 scores were significantly associated with more adjustment to stimuli, speed regulation, and total driving errors. CONCLUSIONS: Our study highlights the critical role of visual search, processing speed, and divided attention on driving performance, and the significant impact of environmental factors (e.g., lighting; weather; traffic) on the occurrence of specific driving errors and crashes. The integration of cognitive assessments (e.g., UFOV; TMTB) should be considered for inclusion as part of the mandatory medical examinations to ensure LHTD can safely operate their commercial motor vehicle.
Spatial clustering and route-based analysis of DUI incidents for enhancing traffic enforcement deployment.
Lee J, Choe CW, Park HC
Traffic Inj Prev · 2025 Dec · PMID 41343202 · Publisher ↗
OBJECTIVE: Driving under the influence (DUI) remains a persistent threat to public safety, especially in dense urban areas where alcohol use and vehicle traffic often intersect. This study develops one of the first integ...
OBJECTIVE: Driving under the influence (DUI) remains a persistent threat to public safety, especially in dense urban areas where alcohol use and vehicle traffic often intersect. This study develops one of the first integrated frameworks that link DUI origins, travel patterns, crash locations, and enforcement points to inform adaptive, route-based enforcement strategies. METHODS: Using 17,075 DUI incidents in Seoul from 2021-2022, the study combines network-based shortest-path estimation and logistic regression to model offender movement, assuming the Place of Last Drink (POLD) as the trip origin. A Monte Carlo sensitivity test confirms the robustness of high-risk corridor identification under spatial uncertainty. The study employs multiscale spatial analysis and hotspot typology, including Global Moran's I, Getis-Ord Gi*, hotspot classification, performance assessment, and enforcement strategies. RESULTS: DUI travel was highly localized, with nearly half of the trips happening within 5 km of the starting point. High-risk corridors stayed consistent under different origin assumptions, supporting the use of POLD as a reliable proxy. Spatial analysis revealed strong clustering of origins and enforcement close to each other, while crash sites were more spread out. A hotspot classification identified about 3.7% of nodes as high-risk areas with limited enforcement, emphasizing the need for proactive and targeted DUI prevention strategies. CONCLUSIONS: The findings emphasize route-based DUI enforcement along urban networks, combining static checkpoints and dynamic patrols to improve efficiency in high-risk areas. Despite potential biases in self-reported origin data, the framework remains robust and offers practical insights for DUI prevention and policy development.
Rotating convertible child restraint systems and their potential effect on correct use in rear-facing mode.
Maltese MR, Haverstick S, Dahle E
Traffic Inj Prev · 2026 · PMID 41343189 · Publisher ↗
OBJECTIVES: Field obervation of CRS use shows that a large portion (65% to 94%) of CRS are misused. In this study, we compared the field use of three different 5-point harness CRSs in rear-facing convertible mode: the fi...
OBJECTIVES: Field obervation of CRS use shows that a large portion (65% to 94%) of CRS are misused. In this study, we compared the field use of three different 5-point harness CRSs in rear-facing convertible mode: the first was a "rotating" car seat (RCRS), where the seat portion of the RCRS can be completely detached from the base and can be rotated to face the parent/caregiver during child ingress and egress. The other two CRSs were conventional convertibles (CRS1 and CRS2). We hypothesize that the RCRS has higher correct use rates on initial inspection compared to two other CRS designs in rear-facing mode. METHODS: The National Digital Car Seat Check Form (NDCF) is a data collection instrument used by certified Child Passenger Safety Technicians to record data on child restraint use and interventions during child passenger safety inspections throughout the United States. The NDCF database was searched for children using an RCRS, CRS1 or CRS2. Outcome variables were binary (1 = proper; 0 = improper) measures of common misuse parameters - correct harness use, correct recline angle, and correct lower attachment. Child age in whole years, state law, body mass, and vehicle type and age were considered as explanatory variables. RESULTS: 752 records were included in the dataset from inspections in 48 U.S. states. Compared to conventional car seats, children seated in an RCRS had 1.4 [95% CI: 1.0, 2.0] times greater adjusted odds of correct harness use when controlling for state law and vehicle type. RCRS children had 2.6 [95% CI: 1.8, 3.8] times greater odds of having the correct recline angle with a modest effect of vehicle age (aOR = 0.97 [95% CI: 0.944, 0.999]). RCRS children had 1.9 [95% CI: 1.3, 2.8] times greater odds of having a correct lower attachment to the vehicle. Outboard seating and LATCH were significant factors in proper CRS-to-vehicle lower attachment. CONCLUSIONS: Rear facing rotating-type convertible car seats increased overall proper use in a population that sought child passenger safety instruction from qualified technicians. More research is necessary to determine if these effects persist in the general population, in forward facing modes of use, and with other rotating car seat models.
Trajectory prediction for multiple types of traffic participants at a signalized intersection based on Heterogeneous Spatio-Temporal Multi-Scale Attention Network.
Gao L, Pang M, Liang H
Traffic Inj Prev · 2025 Dec · PMID 41343183 · Publisher ↗
OBJECTIVES: The work is to investigate the trajectory prediction for multiple types of traffic participants in signalized intersection scenarios within intelligent connected environments based on Heterogeneous Spatio-Tem...
OBJECTIVES: The work is to investigate the trajectory prediction for multiple types of traffic participants in signalized intersection scenarios within intelligent connected environments based on Heterogeneous Spatio-Temporal Multi-Scale Attention Network (HST-MSAN), where participants include Connected and Automated Vehicles (CAVs), Human Vehicles (HVs), cyclists, and pedestrians. METHODS: A novel method of trajectory prediction that integrates spatio-temporal interaction features and multi-scale map features is proposed based on HST-MSAN. The interaction model is established based on Spatio-Temporal Graph Attention Network (STGAN). The trajectory prediction model is constructed based on STGAN and Multi-Scale Squeeze-and-Excitation Network (MS-SENet). First, an STGAN is developed to differentially encode the historical trajectory of each participant, model the complex interactions, and quantify the interaction intensity among participants. Second, an MS-SENet that integrates Multi-Scale Convolutional (MSC) and a Squeeze-and-Excitation (SE) module is proposed, where multiple parallel convolutional kernels are employed to extract both local and global map features. RESULTS: The proposed model is validated through the INTERACTION dataset. The results of three-second trajectory prediction show that the average displacement error (ADE) and final displacement error (FDE) can reach to 0.17 and 0.68 m, respectively. ADE is reduced by 26.1%, 22.7%, 10.5%, and 29.2%, respectively, and FDE is reduced by 10.5%, 12.8%, 8.1%, and 5.6%, respectively, compared with prediction methods of multiple participants of Heterogeneous Edge-enhanced graph attention network (HEAT), Heterogeneous Driving Graph Transformer (HDGT), Hybrid transformer trajectory network (HTTNet), and Flock-inspired network (FN). The ablation experiments show that ADE is reduced by 22.2% and 19.0%, respectively, and FDE is reduced by 10.0% and 5.6%, respectively, compared with the models without STGAN and without MS-SENet. CONCLUSIONS: This model of trajectory prediction jointly models the temporal interaction features of the participants, the spatial interaction features with the surrounding participants, and the multi-scale map features that are most suitable for the current state of the participants. By accurately predicting the future movement trajectories of the surrounding participants, CAVs can identify potential conflict points in advance, optimize their trajectory planning, and reduce the risk of traffic accidents.
SHAP-based interpretation of holistic traffic risk causation patterns using high-resolution trajectory data and risk field theory.
Li B, Hua J, Li P … +2 more , Li Z, Zhang W
Traffic Inj Prev · 2025 Nov · PMID 41289156 · Publisher ↗
OBJECTIVE: Establishing the relationship between traffic states and vehicle conflicts using high-resolution trajectory data is an effective approach for assessing road traffic safety. However, predicting the likelihood o...
OBJECTIVE: Establishing the relationship between traffic states and vehicle conflicts using high-resolution trajectory data is an effective approach for assessing road traffic safety. However, predicting the likelihood of a single type of conflict cannot fully capture the holistic risk level of a road segment. This study aims to develop a method for evaluating road segment-level traffic risk irrespective of conflict types and to analyze the key factors influencing such risks, thereby exploring their underlying mechanisms. METHODS: We propose an integrated surrogate safety measure based on risk field theory, enabling comprehensive identification of vehicle conflicts regardless of relative motion states. Road segment-level holistic risk assessment models were then developed using binary logistic regression and five machine learning methods, with traffic state variables as inputs and conflict occurrence as the output. Resampling techniques were applied to mitigate the effect of imbalanced sample categories. RESULTS: The random forest model trained on the 60-s oversampled dataset achieved the best performance, with an overall prediction accuracy of 80.6%, a precision of 89.0%, and a recall of 86.8% for identifying conflict cases. CONCLUSION: Using SHAP analysis, we interpreted the contribution of individual traffic state variables and their interactions to traffic risk. The vehicle type distribution in traffic flow was identified as a key factor influencing conflict occurrence. This study not only provides accurate risk prediction but also enhances the interpretability of machine learning-based traffic risk models, bridging prediction and causal insight.
Analysis of factors influencing the severity of crashes between new-energy vehicles and pedestrians based on theoretical and data-driven models.
Zhang D, Yang J, Liao W … +2 more , Dong X, Zhao C
Traffic Inj Prev · 2025 Nov · PMID 41264671 · Publisher ↗
OBJECTIVE: With the increasing market share of new energy vehicles (NEVs), the incidence of NEV-related crashes has risen sharply, making the study of influencing factors of NEV crashes an urgent research priority. This...
OBJECTIVE: With the increasing market share of new energy vehicles (NEVs), the incidence of NEV-related crashes has risen sharply, making the study of influencing factors of NEV crashes an urgent research priority. This study aims to identify and assess the key determinants affecting injury severity in NEV-pedestrian collisions to inform safety interventions. METHODS: This study employs a hybrid method combining the SHAP approach, enhanced by information gain ratio, with random parameter logit model to explore the factors influencing the severity of NEV-pedestrian crashes in the UK from 2018 to 2022. RESULTS: The results show that the model's predictive performance improved to 87.00% and 87.76% after random oversampling and hyperparameter optimization, respectively. The SHAP method, improved by the information gain weight, effectively distinguished the contribution rates of different influencing factors when multiple factors interact, significantly enhancing the differentiation of each factor's importance. Pedestrian age was found to be the most significant characteristic factor, increasing the probability of severe injury by 7.82%. Turning (or preparing to turn) and lane changing or overtaking increased the likelihood of severe injury by 2.17% and 1.19%, respectively. Heterogeneous factors included female drivers, drivers aged 56 and older, pedestrians aged 46-65, pedestrians crossing without pedestrian facilities, and crashes occurring on weekends. CONCLUSIONS: By employing optimized data-driven models, theoretical models, and an improved SHAP method, a combined predictive model was constructed to predict and analyze the severity of NEV-pedestrian crashes. The findings of this study provide a foundation for formulating NEV collision safety strategies to reduce the severity of future collisions.
Prediction of driver alertness levels on mountain roads using machine learning models: A naturalistic driving study in China.
Liu T, Xie D, Liu T … +2 more , Shan J, Tang C
Traffic Inj Prev · 2025 Nov · PMID 41264665 · Publisher ↗
OBJECTIVE: A method for evaluating driver alertness on mountain roads was developed to enhance dynamic safety monitoring in high-risk sections. An indicator system integrating human and environmental factors was establis...
OBJECTIVE: A method for evaluating driver alertness on mountain roads was developed to enhance dynamic safety monitoring in high-risk sections. An indicator system integrating human and environmental factors was established, with 13 variables used for alertness classification and 17 initial variables applied for quantification. METHODS: Field tests were conducted in Guizhou, China, where data on drivers' heart rates, eye movements, and demographics were collected. Kernel principal component analysis (KPCA) was employed to extract four representative factors from 13 driving-related indicators. K-means clustering was employed to categorize drivers into high- and low-alertness groups. Logistic regression scoring, XGBoost, and Tabular Prior-Data Fitted Network (TabPFN) models were developed to assess driver alertness probabilities. RESULTS: Superior performance was demonstrated by the XGBoost model, achieving an area under the receiver operating characteristic curve (AUC) value of 0.97 compared with 0.80 for logistic regression scoring model, indicating improved classification accuracy and robustness. TabPFN further enhanced performance, yielding the highest results (AUC: 0.98; KS: 0.85; F1: 0.90), confirming its robustness and adaptability in small-sample, high-dimensional data settings. CONCLUSION: Driver alertness levels were effectively quantified by integrating human and environmental factors. TabPFN model was found to be more effective than logistic regression scoring and XGBoost, making it the preferred approach for monitoring driver alertness on mountain roads.
Developing and evaluating the feasibility and usability of desktop simulator-based driver training for international college students seeking a US driver's license: a pilot study.
Zhang F, Kerwin T, Rundus CRM … +3 more , Wrabel C, Kopelcheck A, Yang J
Traffic Inj Prev · 2025 Nov · PMID 41264664 · Publisher ↗
OBJECTIVE: International college students seeking their first US driver's license represent a potentially risky and yet understudied road user group. They are mostly not regulated by graduated driver licensing (GDL) laws...
OBJECTIVE: International college students seeking their first US driver's license represent a potentially risky and yet understudied road user group. They are mostly not regulated by graduated driver licensing (GDL) laws, making them miss out on the safety benefits associated with the laws. Available resources for them to gain driver knowledge and accumulate driving experience before licensure also remain limited. Simulator-based training offers the potential to train young drivers and improve their safe driving. The objectives of this study were to 1) develop a desktop driving simulator-based driver training program prototype tailored to international older novice drivers seeking to obtain their first US driver's license; and 2) assess the feasibility and usability of the training. METHODS: Following human-centered design principles, our research team composed of experts in human factors, driving simulator research, and injury prevention, first developed a driver training program prototype on a desktop simulator. After that we enrolled 8 international college students to trial the program prototype, surveyed and interviewed them about their experience and attitudes toward the training. We analyzed both qualitative and quantitative data collected from their simulated driving session, survey responses, and interviews to extract participants' experience with and subjective view of the feasibility and usability of desktop simulator-based training, as well as challenges and difficulties driving in the United States. RESULTS: Most participants found the desktop simulator-based training easy to use and helpful in training novice drivers. They also offered practical guidance on how it may be improved. Those who had not driven prior to the study appeared to have extra difficulties following the instructions and completing the training on the simulator as planned. Language barriers, unfamiliarity with US traffic rules, and difficulties finding licensed drivers to accompany them for the road test were identified as the top challenges in obtaining a driver's license in the United States. CONCLUSIONS: Desktop simulator-based driver training was perceived feasible, acceptable, and helpful to train international college students seeking their first US driver's license. Few participants struggled with simulator use due to language barriers and limited prior driving experience. Participants also highlighted key challenges faced in obtaining a US driver's license and adapting to US driving. This research is among the first to explore the use of desktop simulator-based novice driver training for this population, and offers a foundation for future development of relevant driver training programs. Our next steps will involve finalizing the development of the training program and empirically testing its effectiveness in improving the success in passing the road test and reducing post-licensure real-world risky driving behaviors.
Plugged in, tuned out: How earbuds on roads are becoming a silent menace.
Chanda N, Gupta I, Kanjilal MK
Traffic Inj Prev · 2025 Nov · PMID 41264662 · Publisher ↗
Abstract loading — click title to view on PubMed.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe