The advancement of intelligent road systems in developing countries poses unique challenges in identifying risk factors and implementing safety strategies. The variability of factors affecting crash injury severity leads to different risks across levels of roadway smartness, especially in hazardous terrains, complicating the adaptation of smart technologies. Therefore, this study investigates the temporal instability of factors affecting injury severities in crashes across various terrains, with a focus on the evolution of road smartness. Crash data from selected complex terrain regions in Shaanxi Province during smart road adaptation were used, and categorized into periods before, during, and after smart road implementations. A series of mixed logit models were employed to account for unobserved heterogeneity in mean and variance, and likelihood ratio tests were conducted to assess the spatio-temporal instability of model parameters across different topographic settings and smart processes. Moreover, a comparison between partially constrained and unconstrained temporal modeling approaches was made. The findings reveal significant differences in injury severity determinants across terrain conditions as roadway intelligence progressed. On the other hand, certain factors like pavement damage, truck and pedestrian involvement were identified that had relatively stable effects on crash injury severities. Out-of-sample predictions further emphasize the need for modeling across terrain and roadway development stages. These insights are crucial for developing tailored safety measures for smart road retrofitting in different terrain conditions, thereby supporting the transition towards smarter road systems in developing regions.

This study investigates the impacts of various factors on e-bike riders\' injury severity in crashes with motor vehicles, based on the in-vehicle recording video crash data in China. Variables from human factors, vehicle characteristics, road conditions, and environmental attributes are extracted from the video, especially for drivers and riders\' illegal and avoidance behaviour before the crash, and sun shade canopy use. Results of mixed logit models reveal that drivers\' speeding, running red lights, slow-down and swerve behaviour, light trucks, heavy trucks, and buses have significantly varied impacts on riders\' injury. Moreover, both drivers and riders\' illegal behaviour leads to an increased injury, while their avoidance behaviour before crashes can protect riders. In addition, types of visual obstacles, accidents occurring at night, large vehicles\' involvement, and the application of sunshade canopies by riders increased the probability of severe injury, while helmet use can protect riders in accidents with motor vehicles.

Understanding injury severity is essential to inform injury prevention practice. The aims of this scoping review were to investigate how running-related injury (RRI) severity is measured, compare how it differs across studies, and examine whether it influences study outcomes (i.e., injury rates and risk factor identification). This scoping review was prospectively registered with Open Science Framework. A systematic electronic search was conducted using PubMed, Scopus, SPORTDiscuss, MEDLINE, and Web of Science databases. Included studies were published in English between January 1980 and December 2023, investigated RRIs in adult running populations, and included a measure of injury severity. Results were extracted and collated. Sixty-six studies were included. Two predominant primary criteria are used to define injury severity: the extent of the effect on running and/or the extent of the physical description. When secondary definition criteria are considered, 13 variations of injury severity measurement are used. Two approaches are used to grade injury severity: a categorization approach or a continuous numerical scale. Overall, the measurement of RRI severity is relatively inconsistent across studies. Less than half of studies report incidence rates per level of injury severity, while none report specific risk factors across levels, making it difficult to determine if the approach to measuring injury severity influences these study outcomes. This lack of information is possibly contributing to inconsistent rates of RRIs reported, and the lack of clarity on risk factors.

Rear-end (RE) crashes are notably prevalent and pose a substantial risk on freeways. This paper explores the correlation between speed difference among the following and leading vehicles (Δν) and RE crash risk. Three joint models, comprising both uncorrelated and correlated joint random-parameters bivariate probit (RPBP) approaches (statistical methods) and a cross-stitch multilayer perceptron (CS-MLP) network (a data-driven method), were estimated and compared against three separate models: Support Vector Machines (SVM), eXtreme Gradient Boosting (XGBoost), and MLP networks (all data-driven methods). Data on 15,980 two-vehicle RE crashes were collected over a two-year period, from January 1, 2021, to December 31, 2022, considering two possible levels of injury severity: no injury and injury/fatality for both drivers of following and leading vehicles. The comparative performance analysis demonstrates the superior predictive capability of the CS-MLP network over the uncorrelated/correlated joint RPBP model, SVM, XGBoost, and MLP networks in terms of recall, F-1 Score, and AUC. Significantly, numerous shared variables influence the injury severity outcomes for the following and leading vehicles across both statistical and data-driven approaches. Among these factors, the following vehicle (a truck) and the leading vehicle (a passenger car) demonstrate contrasting effects on the injury severity outcomes for both vehicles. Furthermore, the SHapley Additive exPlanations (SHAP) values from the CS-MLP network visually show the relationship between Δν and injury severity, revealing non-linear trends unlike the average effects shown by statistical methods. They indicate that the least injury outcomes for both following and leading vehicles occurs at a Δν of 0 to 10 mph, matching observed patterns in RE crash data. Additionally, a marked variation in the trend of SHAP values for the two vehicles is noted as the speed difference increases. Therefore, the findings affirm the superior performance of joint model development and substantiate the non-linear impacts of speed difference on injury outcomes. The adoption of dynamic speed control measures is recommended to mitigate the injury outcomes involved in two-vehicle RE crashes.

Speeding, a risky act of driving a vehicle at a speed exceeding the posted limit, has consistently emerged as a leading contributor to traffic fatalities. Identifying the risk factors associated with injury severity in speeding-related crashes is essential for implementing countermeasures aimed at preventing severe injury incidents and achieving Vision Zero goals. With the wealth of traffic crash data collected by various agencies, researchers have a valuable opportunity to conduct data-driven studies and employ various modeling methods to gain insights into the correlated factors affecting injury severity in traffic crashes. Machine learning models, owing to their superior predictive power compared to statistical models, are increasingly being adopted by researchers. These models, in conjunction with interpretation techniques, can reveal potential relationships between crash injury severity and contributing factors. Traffic crashes are inherently tied to geographic locations, distributed across road networks influenced by diverse socioeconomic and geographical factors. Recognizing spatial heterogeneity in traffic safety is crucial for tailored safety measures to address speeding-related crashes, as a one-size-fits-all approach may not work effectively everywhere. However, most existing machine learning models are unable to incorporate the spatial dependency among observations, such as traffic crashes, which hinders their ability to uncover spatial heterogeneity in traffic safety. To address this gap, this study introduces the Geographically Weighted Neural Network (GWNN) model, a spatial machine-learning model that integrates neural network (NN) and geographically weighted modeling approaches to investigate spatial heterogeneity in speeding-related crashes. Unlike the traditional NN model, which trains a single set of model parameters for all observations, the GWNN trains a local NN model for each crash location using a spatially weighted subsample of nearby crashes, allowing for the quantification of corresponding local effects of features through calculating local marginal effects. To understand the spatial heterogeneity in speeding-related crashes, this study extracted two years (2020 and 2021) of speeding-related crash data from Alabama for the development of the GWNN local models. The modeling results show significant spatial variability among several factors contributing to injury severity in speeding-related crashes. These factors include driver condition, vehicle type, crash type, speed limit, weather, crash time and location, roadway alignment, and traffic volume. Based on the GWNN modeling results, this study identified three types of spatial variations in relationships between contributing factors and crash injury severity: consistent positive associations, consistent negative associations, and inverse associations (i.e., marginal effects can vary between positive and negative depending on the location). This study contributes by integrating advanced machine learning and spatial modeling approaches to uncover intricate spatial patterns and factors influencing injury severity in speeding-related crashes, thereby facilitating the development of targeted policy implementations and safety interventions.

Although seatbelt use is known to reduce motor vehicle occupant crash injury and death, rear-seated adult occupants are less likely to use restraints. This study examines risk and protective factors associated with injury severity in front- and rear-seated adults involved in a motor vehicle crash in New York State. The Crash Outcome Data Evaluation System (CODES) (2016-2017) was used to examine injury severity in front- and rear-seated occupants aged 18 years or older (N = 958,704) involved in a motor vehicle crash. CODES uses probabilistic linkage of New York State hospitalization, emergency department, and police and motorist crash reports. Multivariable logistic regression models with MI analyze employed SAS 9.4. Odds ratios are reported as OR with 95% CI. The mortality rate was approximately 1.5 times higher for rear-seated than front-seated occupants (136.60 vs. 92.45 per 100,000), with rear-seated occupants more frequently unrestrained than front-seated occupants (15.28% vs. 1.70%, p < 0.0001). In adjusted analyses that did not include restraint status, serious injury/death was higher in rear-seated compared to front-seated occupants (OR:1.272, 1.146-1.412), but lower once restraint use was added (OR: 0.851, 0.771-0.939). Unrestrained rear-seated occupants exhibited higher serious injury/death than restrained front-seated occupants. Unrestrained teens aged 18-19 years old exhibit mortality per 100,000 occupants that is more similar to that of the oldest two age groups than to other young and middle-aged adults. Speeding, a drinking driver, and older vehicles were among the independent predictors of serious injury/death. Unrestrained rear-seated adult occupants exhibit higher severe injury/death than restrained front-seated occupants. When restrained, rear-seated occupants are less likely to be seriously injured than restrained front-seated occupants.

This study examines the variability in the impacts of factors influencing injury severity outcomes of elderly pedestrians (age >64) involved in vehicular crashes at intersections and non-intersections before, during, and after the COVID-19 pandemic. To account for unobserved heterogeneity in the crash data, a random parameters logit model with heterogeneity in the means approach is utilized to analyze vehicle-elderly pedestrian crash data from Seoul, South Korea, occurring between 2018 and 2022. Preliminary transferability tests revealed instability in factor impacts on injury severity outcomes, highlighting the need to estimate individual models across various road segments and time periods. Thus, the dataset was segregated by crash location (intersection/non-intersection) and period (before, during, and after COVID-19), with individual models estimated for each group. Results obtained from the analyses revealed that back injuries positively influenced fatalities at non-intersections after the pandemic and was negatively associated with fatalities at intersections before the pandemic. Additionally, several indicators demonstrated significant instability in their impact magnitudes across different road segments and crash years. During the pandemic, head injuries increased the probability of fatalities higher at non-intersections. After the pandemic, crosswalk locations decreased the possibility of fatalities more at intersections. Compared to intersection segments, the female indicator reduced the likelihood of fatal injuries at non-intersections more before, during, and after the pandemic. Before the pandemic, much older pedestrians experienced a greater decline in fatalities at intersections than non-intersections. This instability could be attributed to altered mobility patterns stemming from the COVID-19 pandemic. Overall, the study findings highlight the variability of determinants of fatal/severe injury outcomes among elderly pedestrians across various road segments and years, with the underlying cause of this fluctuation remaining unclear. Furthermore, the findings revealed that accounting for heterogeneity in the means of random parameters enhances model fit and provides valuable insights for safety professionals. The factor impact variability in the estimated models carries significant implications for elderly pedestrian safety, especially in scenarios where precise projections of the effects of alternative safety measures are essential. Road safety experts can leverage these findings to refine or update current policies to enhance elderly pedestrian safety at intersections and non-intersections.

OBJECTIVE: Traumatic injury surveillance can be enhanced by describing injury severity trends. This study reports trends in work-related injury severity for males and females over the period 2004-2017 in Ontario, Canada.
METHODS: A weighted measure of workers\' compensation benefit expenditures was used to define injury severity, obtained from the linkage of workers\' compensation claims to emergency department (ED) records where the main injury or illness was attributed to work. Denominator counts were obtained from Statistics Canada\'s Labor Force Survey. Trends in the annual incidence of injury, classified as low, moderate, or high severity, were examined using regression modeling, stratified by age and sex.
RESULTS: Over a 14-year observation period, there were 1,636,866 ED records included in the analyses. Overall, 57.6% of occupational injury records were classified as low severity, 29.5% as moderate severity, and 12.8% as high severity conditions. There was an increase in the incidence of high severity injuries among females (annual percent change (APC): 1.52%; 95% CI: 0.77, 2.28), while the incidence of low and moderate severity injuries generally declined for males and females. Among females, injuries attributed to animate mechanical forces and assault increased as causes of low, moderate, and high severity injuries. The incidence of concussion increased for both males (APC: 10.51%; 95% CI: 8.18, 12.88) and females (APC: 16.37%; 95% CI: 13.37, 19.45).
CONCLUSIONS: The incidence of severe work-related injuries increased among females in Ontario between 2004 and 2017. The methods applied in this surveillance study of traumatic injury severity are plausibly generalizable to applications in other jurisdictions.

This study simultaneously modelled the injury severity of motorcycle riders and their pillion passengers and determine the associated risk factors. The analysis is based on motorcycle crashes data in Ashanti region of Ghana spanning from 2017 to 2019. The study implemented bivariate ordered probit model to identify the possible risk factors under the premise that the injury severity of pillion passenger is endogenously related to that of the rider in the event of crash. The model provides more efficient estimates by considered the common unobserved factors shared between rider and pillion passenger. The result shows a significant positive relationship between the two injury severities with a correlation coefficient of 0.63. Thus, the unobservable factors that increase the probability of the rider to sustain more severe injury in the event of crash also increase that of their corresponding pillion passenger. The rider and their pillion passenger injury severities have different propensity to some of the risk factors including passengers\' gender, day of week, road width and light condition. In addition, the study found that time of day, weather condition, collision type, and number of vehicles involved in the crash jointly influence the injury severity of both rider and pillion passenger significantly.

UNASSIGNED: A lower helmet-wearing rate and overspeeding in Pakistan are critical risk behaviors of motorcyclists, causing severe injuries. To explore the differences in the determinants affecting the injury severities among helmeted and non-helmeted motorcyclists in motorcycle crashes caused by overspeeding behavior, single-vehicle motorcycle crash data in Rawalpindi city for 2017-2019 is collected. Considering three possible crash injury severity outcomes of motorcyclists: fatal injury, severe injury and minor injury, the rider, roadway, environmental, and temporal characteristics are estimated.
UNASSIGNED: To provide a mathematically simpler framework, the current study introduces parsimonious pooled random parameters logit models. Then, the standard pooled random parameters logit models without considering temporal effects are also simulated for comparison. By comparing the goodness of fit measure and estimation results, the parsimonious pooled random parameters logit model is suitable for capturing the temporal instability. Then, the non-transferability among helmeted and non-helmeted overspeeding motorcycle crashes is illustrated by likelihood ratio tests and out-of-sample prediction, and two types of models provide robust results. The marginal effects are also calculated.
UNASSIGNED: Several variables, such as age, cloudy and weekday indicators illustrate temporal instability. Moreover, several variables are observed to only show significance in non-helmeted models, showing non-transferability across helmeted and non-helmeted models.
UNASSIGNED: More educational campaigns, regulation and enforcement, and management countermeasures should be organized for non-helmeted motorcyclists and overspeeding behavior. Such findings also provide research reference for the risk-compensating behavior and self-selected group issues under overspeeding riding considering the usage of helmets.