Introduction: Road traffic crashes pose a significant public health challenge, with traditional crash data often relying on broad classifications that obscure critical details. This study addresses the knowledge gap created by ambiguous categories, “Other Improper Action” and “Not Discernible,” within the Ohio crash dataset. Method: Employing a multi-faceted analytical framework that combines descriptive statistics, N-gram analysis, and Latent Dirichlet Allocation (LDA) topic modeling on over 67,000 free-text crash narratives from 2020 to 2024, the study uncovers the latent contributing circumstances previously masked by these labels. Results: The analysis reveals that “Other Improper Action” incidents are disproportionately linked to adverse environmental conditions and a lack of formal traffic control. Text mining further extracted hidden behavioral and environmental factors, predominantly severe spatial awareness deficits, striking legally parked vehicles in urban environments and environmentally induced loss of vehicle control resulting in infrastructure strikes. In contrast, “Not Discernible” crashes are more prevalent in daylight and at signalized intersections. Rather than a lack of physical information, the NLP models revealed that investigative ambiguity primarily stems from conflicting driver accounts, specifically mutual lane change encroachment and right-of-way disputes. Conclusions: These findings not only validate the presence of recognized crash mechanisms, but transform vague data classifications into actionable intelligence. Practical Applications: This methodology provides transportation safety professionals with the granular evidence needed to refine statewide data collection protocols and develop targeted, engineering-backed safety interventions.
Introduction: Dementia is a progressive neurodegenerative condition associated with declines in cognitive functions essential for safe driving. While on-road and simulator studies consistently demonstrate impaired driving performance, evidence regarding motor vehicle crash (MVC) risk remains mixed. This review synthesised contemporary evidence on the impact of dementia on real-world driving behaviour and crash involvement. Methods: A systematic search of CENTRAL, MEDLINE, Embase, Transport, PsycINFO, and Scopus was conducted in accordance with PRISMA guidelines. Quantitative studies published from January 2015 were included if they examined licensed adult drivers with clinically diagnosed dementia and reported outcomes related to MVC involvement, driving performance, or unsafe behaviour compared with cognitively healthy drivers. Methodological quality was assessed using the Joanna Briggs Institute tools. The review was registered with PROSPERO (CRD420251107288). Results: Thirteen studies met the inclusion criteria, including cross-sectional, cohort, and case–control designs. Findings for MVC involvement were heterogeneous: population-based studies generally reported no increased crash risk, whereas studies of medically referred drivers indicated elevated risk in selected subgroups. In contrast, studies examining driving errors and unsafe behaviours consistently demonstrated poorer performance among drivers with dementia, particularly in cognitively demanding situations such as hazard response and intersection negotiation. Naturalistic studies further identified persistent unsafe events during everyday driving, even among individuals adopting compensatory strategies in early disease. Conclusions: Dementia is associated with clinically meaningful impairments in driving behaviour and performance, evident even in the early stages of the disease. Differences in exposure, self-regulation, and regulatory processes largely explain variability in MVC findings. A multidimensional approach integrating MVC, performance-based, and naturalistic data is required to inform fitness-to-drive decisions. Practical applications: MVC involvement alone is an insensitive safety indicator. Performance-based and naturalistic measures provide more sensitive markers of impairment, supporting earlier risk identification, targeted intervention, and the development of in-vehicle monitoring and assistive technologies to enhance safety and prolong safe mobility.
Introduction: High-quality crash data are essential for traffic safety analysis, yet police-reported crash databases often suffer from underreporting and miscoding, particularly for secondary crashes. This study evaluates advanced natural language processing (NLP) techniques to enhance crash data quality by mining crash narratives, using secondary crash identification in Kentucky as a case study. Method: Drawing from 16,656 manually reviewed narratives from 2015 to 2022, with 3803 confirmed secondary crashes, we systematically compared 11 models across four paradigms: zero-shot open-source large language models (LLMs), fine-tuned transformers, deep learning model with word embeddings, and logistic regression. Statistical significance was assessed using pairwise McNemar’s tests, and 95% bootstrap confidence intervals were computed for all metrics. Results: Fine-tuned transformers achieved statistically superior performance, forming a top-performing cluster that was indistinguishable internally. RoBERTa yielded the highest F1 (0.90) and accuracy (95.4%) while requiring only seconds of inference on the test set. Among zero-shot LLMs, Llama3:70B reached the best F1 (0.86) but required 139 min of inference. The BiLSTM baseline (F1: 0.79) was statistically indistinguishable from Qwen3:32B and Gemma3:27B, while logistic baseline lagged well behind (F1: 0.66). Qualitative error analysis revealed that RoBERTa and Llama3:70B exhibit complementary failure patterns, supporting ensemble deployment strategy. Conclusions: For agencies with labeled training data, fine-tuned RoBERTa is the recommended deployment choice, which offers the strongest accuracy at negligible computational cost. For agencies lacking labeled data, zero-shot LLMs such as Llama3:70B provide a viable alternative that can be deployed readily and simultaneously accumulate a labeled dataset for eventual transition to fine-tuned models. Practical applications: These findings allow transportation agencies to automate labor-intensive narrative reviews, addressing chronic data quality issues like secondary crash miscoding. Practical deployment considerations are discussed, which emphasize privacy-preserving local deployment, ensemble approaches for improved accuracy, and incremental processing for scalability, providing a replicable scheme for enhancing crash-data quality with advanced NLP.
Introduction and method: Employing an explanatory sequential mixed-methods design, data were collected from 340 survey respondents and 35 interviewees across logistics facilities. Quantitative analysis using moderated multiple regression revealed that social norms (β = 0.19, p = 0.002) and workplace behavior (β = 0.48, p < 0.001) significantly predict safety practices. However, workplace behavior did not moderate this relationship (p = 0.329). Results: Qualitative findings elucidated this null effect, highlighting dominant operational logics where productivity pressures and conformity override individual behavioral consistency. Three interlocking themes emerged: Productivity over safety, where institutional output targets structurally decouple individual routine from safety protocol; Conformity as survival, wherein peer-enforced norms function as coercive mandates that neutralize personal discipline; and Precarious Employment, which institutionalizes a culture of safety silence. Together, these themes demonstrate that macro-level structural constraints systematically override individual agency. Conclusion and Practical Applications: The study concludes that effective injury reduction requires interventions that address institutional coercive pressures and employment precarity, rather than relying solely on individual behavioral modification in high-tempo warehouse environments.
Introduction: Subjective safety is a key determinant of cycling behavior and traffic risk perception, yet it is rarely assessed systematically in infrastructure evaluation. Because perceived safety does not necessarily correspond to objective indicators such as crash statistics or design standards, reliable methods are needed to capture experiential safety dimensions relevant to research and practice. Immersive technologies such as head-mounted displays (HMDs) may provide a controlled, scalable, effort- and resource-efficient addition or even alternative to rarely conducted on-site assessments. Method: This study investigated whether static stereoscopic 360° images presented via an HMD can approximate real-world evaluations of subjective cycling safety. Following two preparatory studies on representativeness (N1 = 92) and usability (N2 = 20), a comparative study was conducted with two independent cohorts (N3 = 129). One group assessed a bicycle street in real traffic conditions (n = 64), while the other evaluated the same locations using twelve 360° panoramas in an HMD (n = 65). Subjective safety was measured using five items adapted from a national cycling safety survey. Bayesian independent-samples t tests were conducted to examine convergence or divergence between assessment modes. Results: Findings indicated convergence for overall subjective safety (BF10 = 0.67) and perceived conflicts with motor vehicles (BF10 = 0.81), suggesting that HMD-based assessments can reproduce general safety impressions. However, participants in the HMD condition reported higher perceived conflict frequency with pedestrians (BF10 = 735.95), more obstacles (BF10 = 7.14), and lower perceived suitability for children and older adults (BF10 = 2.83). Usability ratings were highly positive. Conclusion and Practical Applications: Static HMD-based assessments seem to validly capture key aspects of perceived cycling safety while offering a resource-efficient and controlled alternative to field surveys. The approach enables systematic integration of user perception into infrastructure planning and supports the development of safer cycling environments.
Introduction: Despite years of research challenging their validity, injury rate metrics continue to significantly influence decision-making in many organizations. In this paper, we argue that both the source of this problem and a solution rest in understanding the narrative power of these metrics. We explore this idea by developing and applying a set of alternative injury measurement tools specifically designed to provide more comprehensive and compelling narratives of injury performance. Comparing alternative measures with traditional stand-alone injury rates demonstrates how the overall narrative of injury performance changes through metric design. Our overall goal is not to fix the statistical validity of the metrics, but to improve safety decision-making through more complex discourse. Method: The analysis in this paper rests on three large organization case studies. In each case, study data from the organization was presented back to senior executives using: the injury rate alone; the average injury severity; and a weighted combination of rate and severity known as SAIF. The resultant conversations were then analyzed, supplemented by individual interviews with conversation participants. Results: This research highlights the importance of presenting injury rate and severity measures together and demonstrating their interplay, rather than as stand-alone metrics. Four guidelines are proposed for improving narratives and conversations based on injury metrics. The four guidelines are: pairing measures, continuous variability, a simple communication mechanism, and highlighting differences. Conclusions and Practical Applications: Together, these guidelines enable users of injury measures to generate more dynamic discussions from their metrics.
Introduction: Rapid digitalization has evolved from a technical facilitator into a significant psychosocial demand. This study analyzes the transformation of psychosocial risk management in Spain and Denmark, testing whether digital intensity acts as a universal stressor or interacts with national contexts. Method: Drawing on the Job Demands-Resources (JD-R) model, the study employs a repeated cross-sectional design based on the European Survey of Enterprises on New and Emerging Risks (ESENER) in 2019 (N = 3,385) and 2024 (N = 3,462). A moderated mediation model was tested to connect Digital Intensity to Preventive Measures via Organizational Reactivity, examining Worker Participation as a moderator. Results: The analysis reveals a structural stratification in 2024: traditional industrial sectors are now significantly disadvantaged regarding safety management compared to the service sector. While Digital Intensity universally triggers Organizational Reactivity, the role of Worker Participation shifted from a dormant factor in 2019 to an active participatory shield in 2024, significantly buffering the negative link between reactivity and prevention. Conclusions: The study challenges the assumption of digital homogenization. In the post-pandemic era, high digital intensity creates a reactive environment that requires active worker voice—not merely as a right, but as an operational necessity—to decouple safety management from the logic of failure demand. Practical Applications: Safety managers should prioritize worker participation mechanisms in highly digitalized environments to counteract organizational reactivity. Targeted interventions are urgently needed in industrial sectors, which are currently falling behind in the digital safety transition.
Introduction: Serious injuries and fatalities (SIFs) in construction have remained stubbornly high despite substantial safety efforts. Prior research suggests that these events frequently arise in contexts where controls are absent, ineffective, poorly defined, or inconsistently implemented. Prevention appears to depend less on adding more controls and more on designing controls that are explicitly intended for SIF prevention, usable in the field, and precise enough to support consistent decisions at the point of risk. Within the energy-based safety framework, the concept of Direct Control has been introduced as a scientifically defined type of control for high-energy hazards with SIF potential. However, Direct Controls are not always feasible, leaving a recurring gap in practice: what constitutes “enough” control when high-energy hazards must be managed without a Direct Control? Method: This study addresses that gap through a two-phase research process. In Phase 1, an expert panel of 25 construction safety professionals participated in structured brainstorming, focus groups, and a modified Delphi method to develop a definition and criteria for Alternative Controls. The resulting framework specifies three categories and rules to ensure that Alternative Controls are timely, tangible, targeted, and reliably implemented at the point of risk. In Phase 2, a noise experiment with a separate group of 20 practitioners tested the framework’s usability. Variability in adequacy decisions decreased by 47%, inter-rater agreement improved from fair to substantial (Fleiss’ κ from 0.21–0.32 to 0.61–0.62), and practitioner confidence increased significantly (paired t-test, p < 0.001). Results: Findings indicate that structured criteria for Alternative Controls can reduce decision noise and support more consistent, SIF-focused control decisions when Direct Controls are not feasible.
Introduction: The construction industry remains one of the most hazardous work environments despite advances in safety technologies. Construction safety technology (CST) research has increasingly focused on hazard detection, monitoring, and prediction; however, the conceptualization of injury within this literature remains unclear. Method: This study examines how injuries are represented in CST research through a theory-informed analytical review of 148 studies. The analysis is structured around three domains: safety risk context, safety management function, and injury outcome and burden. Results: CST research was strongly oriented toward visible hazards and detection-based functions, with general or non-specific safety framing appearing in 93% of studies and prevention/detection functions represented in 90% and 86%, respectively. Injury representation was dominated by physical injury outcomes (95%), while burden dimensions such as severity, frequency, long-term impact, cost, and time loss were minimally represented (∼0–5%). Exploratory regression showed limited but improved explanatory power after adjustment for publication year and study type (R2 = 0.150 to 0.251), suggesting a structural misalignment between risk framing, technological function, and injury representation. Practical Applications: Hazard detection metrics alone are insufficient for evaluating construction safety technologies. Burden-informed indicators should be incorporated as preliminary evaluation dimensions, calibrated to specific trades, tasks, and site conditions, and linked to response, recurrence, recovery, fatigue, and corrective-action data. Conclusions: The results highlight the need for a shift toward outcome-oriented and human-centered approaches in CST research to better align technological development with multidimensional injury burden.
Introduction: Occupational risk management has traditionally focused on the prevention of risks that seem more obvious than psychosocial risks. In recent years, the importance of psychosocial risks in the workplace has become apparent. The European Agency for Safety and Health at Work developed guidelines and information campaigns focusing on these risks. Psychosocial risks are multi-causal and difficult to identify due to the particularities of the factors that characterize them. Research on psychosocial risks has provided different classifications of the factors and effects of this type of risk, where we can highlight the one developed by the European Union in the Psychosocial Risk Management-Excellence Framework (PRIMA-EF). The aim of this research is to determine the influence of gender, company size, or productive sector in relation to psychosocial risks in the workplace. Method: A multivariate analysis of the II Andalusian Working Conditions Survey, using the PRIMA-EF theoretical model and by means of the elaboration of a synthetic indicator, allowed us to identify and confirm eight psychosocial risk factors and to define the Andalusian Composite Indicator of Psychosocial Risks (ACIPR). This indicator is used to measure differential exposure. Results: The main findings show a higher psychosocial risk in men, where the construction sector is the one with the highest psychosocial risk. It also highlights that medium-sized companies have the lowest psychosocial risk.
Introduction: Older cyclists in the Netherlands continue to cycle at a higher age. While this is generally considered as positive, older cyclists are more likely to be involved in crashes leading to severe injuries compared to younger cyclists. However, little is known about how cycling behavior changes as people age. Method: This longitudinal study explored potential behavioral changes by measuring cycling behavior in 112 older cyclists (mean age = 72 years, SD = 5.7) over a three-year period, with measurements in spring and autumn to assess seasonal effects. Cycling speed and swerving were recorded during short bicycle rides and mounting and dismounting behavior were observed on a closed track. Additionally, cyclists reported on their cycling experiences over this period. Linear mixed-effects models were used to assess the effects of time, age, gender and location of the measurements. Results: Participants were generally highly active cyclists. Linear mixed-effects models showed that cycling speed and swerving behavior remained stable over time, with no meaningful effects for age, gender, or season. Participants greatly increased their helmet use over the study period. Substantial gender differences were observed for mounting and dismounting behavior, which are related to differences in saddle height. Conclusions: Active older cyclists maintained stable cycling performance over a period of three years. Differences in saddle height between male and female cyclists result in different mounting and dismounting strategies. Practical applications: Saddle height differences between older male and female cyclists highlight opportunities for improving bicycle design aimed at facilitating older cyclists.
Introduction: Scholarly evidence suggests that the effectiveness of Occupational Health and Safety Committees (OHSCs) may rely on various factors. However, it remains unclear which specific factors influence OHSC functionality and how they may be interconnected. This scoping review aimed to identify, map, and analyze empirical studies from four major bibliographic databases (Scopus, EBSCOhost, PubMed, and Web of Science) to explore which factors, attributes, or characteristics contribute to the effective operation of OHSCs. Method: Nine studies from 2014 to 2024 were identified across seven different industries, with the healthcare sector being the most represented. The findings revealed 63 factors grouped into eight themes: OHSC composition, OHSC governance, OHSC resources, OHSC roles and responsibilities, OHSC external communication, OHSC influence, OHSC members’ attitudes, management, and workers’ support for OHSC. The discussion of these themes in relation to other literature led to the development of a preliminary conceptual framework for the effective operation of OHSCs, illustrating the interconnectedness and interdependence of the themes and their associated factors. Conclusions and practical applications: This framework, along with the identified factors, can play a vital role in guiding improvements in industry practices and standards. It can also inform future research by contextualizing its themes and factors within local parameters, such as organizational complexity, cultural elements, resources available, and legislative expectations.
Introduction This study evaluates the safety effectiveness of newly implemented Midblock Pedestrian Signals (MPS) at 14 locations across Florida. Data collection and processing: A total of 2,645 pedestrian–vehicle interactions were extracted from CCTV footage and processed using advanced computer vision techniques. Pedestrian–vehicle conflicts were categorized as serious, moderate, and non-conflicts based on the Relative Time to Collision (RTTC) measure. Methodology To estimate the safety impact, a two-stage joint modeling framework was developed, addressing two key methodological challenges: potential selection bias due to non-random MPS assignment and temporal and baseline differences across sites. The selection model estimated the probability of MPS treatment using a Probit model, while the outcome model predicted conflict severity through a penalized multinomial logistic regression with integrated Difference-in-Differences (DiD)-style variables. Joint likelihood estimation corrected for selection bias by linking treatment assignment to outcome patterns within a unified likelihood structure. Results The Average Treatment Effect (ATE) results demonstrated that MPS installations significantly improved pedestrian safety outcomes by reducing both moderate and serious conflicts across all control group comparisons, including locations with Pedestrian Hybrid Beacons (PHBs), Rectangular Rapid Flashing Beacons (RRFBs), and Flashing Beacons. The DiD analysis confirmed that these improvements were not merely driven by general time trends but were directly attributable to the MPS treatment itself. Practical applications The findings provide strong empirical support for transportation safety policies that prioritize MPS deployment at midblock crossings, suggesting that MPS can serve as an effective and practical alternative to traditional pedestrian crossing treatments under appropriate conditions. The proposed framework also offers a methodological foundation for evaluating non-randomized interventions and can inform future safety research, policy development, and data-driven signal implementation strategies.
Introduction: Road traffic fatalities and injuries remain a leading cause of death worldwide, with risks varying significantly across age groups. While existing research primarily focuses on aggregate fatality trends, limited attention has been given to how socioeconomic development reshapes the distribution of fatality risk between children and older adults. This study examines this relationship across countries over time. Method: Using panel data from 160 countries over a 30-year period, a fixed-effects analytical approach was employed to assess the relationship between development and age-specific road mortality. The child-to-older adult death ratio (CTODR) was used as the key outcome measure to capture the relative distribution of fatalities between children (aged 0–14 years) and older adults (aged 60 years and above). Results: Higher levels of socioeconomic development were associated with improvements in road infrastructure, traffic regulation, and post-crash care, which disproportionately reduced exposure and fatality risk among children relative to older adults. Population aging further reinforced this pattern by increasing the relative vulnerability of older road users. Urbanization and income inequality were also linked to lower CTODR values, indicating comparatively higher risks for older adults in more urbanized and unequal societies. Additionally, a negative time trend suggested that road safety improvements over time have benefited children more than older adults. Conclusions: The findings provide evidence of a “transitional risk shift,” whereby the burden of road mortality shifts from children to older adults as societies develop. This indicates that road safety improvements are not distributionally neutral and may inadvertently increase relative risks for older populations. Practical Applications: Policymakers should consider age-specific interventions in road safety strategies. As countries develop, targeted measures to protect older road users—such as safer urban design, improved pedestrian infrastructure, and tailored post-crash care—are essential to address the shifting burden of road mortality.
Introduction: Visible police presence is an effective approach to improve compliance with posted speed limits in work zones. While police presence improves speed limit compliance and driver attentiveness, it can also contribute to sudden speed reduction and harsh braking, which may lead to traffic queues and increase the risk of end-of-queue (EOQ) collisions. Despite many studies examining the effects of police presence on driver speeds, the effects on EOQ crash risk are yet to be comprehensively examined. This paper aims to evaluate the effects of police presence on vehicle speeds, collision risk, and driver perceptions in work zones. Method: Speed data, collected using automatic traffic counters at six locations in a roadwork site, were analyzed using a Tobit regression model to quantify the effect on driver speeding behavior. Near-misses were identified from vehicle-trajectory analysis of video images recorded using roadside cameras and analyzed using the Extreme Value Theory approach. Driver perceptions regarding the effectiveness of police presence were gathered through a roadside survey. Results: The results showed that police presence was associated with a 35% reduction in the probability of speeding near where police were located and 11–20% reduction closer to the worksite. However, the risk of serious EOQ conflicts increased by around 10% when police were present. The survey results revealed mixed driver views about police presence at roadworks. Conclusions: Visible police presence reduced speeding cases but increased the probability of serious EOQ conflicts. Practical applications: These findings suggest that while improved driver behavior can be achieved using visible police presence, attention also needs to be given to using other safety measures to reduce the risk of EOQ crashes, when police are present on site.
Background: The huge number of road traffic crashes (RTCs) resulting in death, disability or injury renders road safety a crucial issue worldwide. Nevertheless, most of the available literature focus on the pre-RTC factors rather than the post-crash period. As a result, little is known about the physical and mental burden following a RTC. The aim of this scoping review was to summarize the evidence on the factors associated with the long-term consequences on physical and mental health of individuals involved in RTCs. Method: Long-term consequences, physical and mental health, quality of life, road traffic crash were some of the keywords used to search the PubMed, Scopus, and Web of Science electronic databases. Eligible studies were observational, published between 2010 and 2024, investigating the physical and mental long-term consequences of RTCs on any type of road users. Results: Starting from 419 papers and following the PRISMA 2020 guidelines, 50 studies were recorded as relevant. After reading the full text and retrieving two further articles through citation searching, 35 studies were included in this review. Findings showed that variables associated with the long-term consequences on physical and mental health can be categorized into demographic, clinical, psychological, socioeconomic, and crash-related factors. When distinguishing among road user types, cyclists showed better health outcomes compared to other groups. Conclusions: This scoping review focuses on the long-term physical and psychological impact that RTCs have on individuals, emphasizing the need to regard recovery as an extended and multidimensional process. Methodological heterogeneity across studies limits comparability, thus underscoring the need for standardized measures and consistent follow-up periods in future research. Practical applications: These findings carry important implications for clinical care and public health and highlight the importance of identifying individuals at risk of unfavorable long-term consequences and implementing coordinated approaches that encompass both physical and psychological aspects.
Objectives: Despite procedural justice in police-citizen interactions being widely studied, there is limited research regarding police perceptions of these interactions. This study addresses this by exploring police perceptions of two different types of random breath test (RBT) – standard and procedurally just – in an experimental trial. Officers were asked to report on whether they believe that their RBT interaction with drivers and riders changed behaviour, developed more positive perceptions of police and resulted in slower travel. Additionally, they were asked if they thought that the RBT interaction took too long. Methods: This study was a six month randomised control trial with two operational conditions: a control condition with a standard (n = 18 days) and an experimental condition with a procedurally just (n = 17 days) RBT. Police officers completed post-operation surveys (n = 80). Results: Police officers incorporating the principles of procedural justice into RBT had limited impact on officer perceptions. However, those officers who were more open to innovation believed that the interaction would result in more positive perceptions of police. Police officers who believed that their RBTs with motorcycle riders would make a difference perceived that the riders would both change their behaviour and travel more slowly after the interaction. There were no differences in perceived interaction length by police officers across both conditions. Conclusions: This study has shown that police officer openness to innovation has an important role to play in police officers having more positive perceptions regarding the outcomes of RBT interactions. As such police organisations should undertake activities, such as training, to improve innovation. The lack of perceived differences in RBT length between standard and procedurally just interactions suggests that these enhanced interactions can be implemented without concerns that police officers believe that they ‘take too long’.
