In recent years, the number of gig workers, defined as individuals who provide labor through digital platforms, has increased significantly worldwide [1]. The rapid expansion of food delivery applications has markedly increased the number of motorcycle-based food delivery workers (MFDWs) [2]. Because MFDWs deliver orders using motorcycles, they are routinely exposed to road hazards and face a disproportionately high risk of traffic accidents when compared with other occupational groups [3]. Furthermore, algorithmic management and piece-rate payment systems often compel MFDWs to operate under intense time pressure, potentially shaping risky riding behaviors and elevating crash risks [4]. Moreover, the concept of rest in the gig economy is often ambiguous; periods labeled as breaks may conflate restorative rest with unpaid waiting times for algorithmically assigned tasks, potentially hindering effective recovery [5].

Although traffic accidents involving physical injury remain a primary concern [6], safety outcomes in this workforce encompass more than just accidents. Near-miss experiences and accident-related anxiety are critical yet understudied dimensions of traffic safety. Accident-related anxiety reflects not only a psychological response to danger but also a perceived vulnerability to crashes [7]. Accident-related anxiety, when combined with occupational fatigue, a state of exhaustion known to impair cognitive performance and elevate anxiety, may further compromise safety [8]. Importantly, from a preventative perspective, identifying early indicators before actual crash events is crucial for recognizing individuals at high risk of traffic accidents. Measures such as near-miss experiences and accident-related anxiety may provide valuable insight into preaccident risk and support the development of preventive interventions [9]. Therefore, understanding traffic safety requires a multidimensional approach that considers near-miss experiences and anxiety along with actual accidents among MFDWs.

Previous reviews have consistently reported sociodemographic characteristics and behavioral factors as key risk factors for adverse traffic safety outcomes among motorcycle riders. Sociodemographic factors, such as younger age and lower education level, along with behavioral characteristics, including alcohol consumption, smoking, nonhelmet use, and risky driving behaviors, have been consistently associated with traffic crashes and injuries [10,11]. Additionally, occupational status–related factors, such as long working hours and work schedules, are important correlates of traffic safety outcomes, and rider fatigue has been proposed as a significant risk factor [12]. Occupational fatigue has been identified as an important risk factor among motorcycle riders; however, it has rarely been assessed using validated instruments. Moreover, comprehensive analyses that simultaneously consider occupational fatigue, health-related behaviors, occupational status, and sociodemographic characteristics with adequate control for potential confounders remain limited.

Despite a growing literature on gig delivery workers, key gaps remain. Existing evidence has largely relied on crash- or injury-related outcomes, whereas near-miss or traffic-conflict events, widely discussed as proactive surrogate safety indicators, have been less commonly examined in this population [6,13]. Furthermore, the psychological aspects of safety, particularly traffic accident anxiety, have seldom been analyzed in relation to occupational fatigue and health behaviors. Additionally, existing studies lack comprehensive analyses that integrate occupational fatigue measured with validated instruments alongside health-related behaviors, occupational status, and sociodemographic characteristics, while adequately controlling for potential confounders. Addressing these gaps is vital to formulate prevention strategies that transcend regulatory enforcement. Therefore, this study aimed to bridge these limitations by investigating the work- and health-related determinants of traffic safety among MFDWs. Using a multidimensional framework, we assessed 3 distinct domains: Traffic Accident Risk Index (TARI), near-miss experiences, and accident-related anxiety. By integrating these experiential and psychological dimensions, this study aimed to provide comprehensive evidence to help occupational health policymakers improve the safety and well-being of platform delivery workers.

This study aimed to (1) assess the levels of traffic accident risk, including near-miss experiences, other-rated accident anxiety, self-rated accident anxiety, and occupational fatigue; (2) describe the demographic characteristics, occupational status, and health behaviors of MFDWs; (3) investigate the relationship between TARI and occupational fatigue; and (4) identify factors associated with TARI and its 3 subdomains among MFDWs.

This study contributes to the growing body of literature on MFDWs by adopting a multidimensional view of traffic safety beyond crash-based outcomes. We examined TARI, near-miss experiences, and 2 forms of accident-related anxiety (self-rated and other-rated anxiety). This distinction is meaningful because behavioral precursors to accidents (near-miss experiences) and the psychological dimensions of safety (accident anxiety) may not share identical correlates. By modeling these outcomes separately, our findings offer a more granular understanding of how occupational fatigue and health behaviors are related to safety precursors and perceived vulnerability in platform-based delivery work.

Across all the models, occupational fatigue emerged as the most consistent correlate of adverse safety outcomes. Higher fatigue was associated with a higher overall traffic accident risk, more near-miss experiences, and greater accident-related anxiety. This pattern is broadly consistent with prior evidence indicating that fatigue is linked to degraded cognitive performance and safety-critical decision-making [20,21]. Additionally, fatigue has been associated with reduced situational awareness and slower reaction times, which have been linked to near-miss events and judgment errors in complex traffic environments [22,23]. The consistency of the association across multiple safety domains suggests that fatigue is a salient marker of increased safety risks and psychological strain in MFDWs. Although causal inference was not warranted in this cross-sectional study, the consistency of fatigue associations across all safety domains suggests that fatigue management warrants prioritization in this workforce. At the platform level, immediate steps could include in-app advisory alerts recommending rest after extended delivery periods, as well as the integration of fatigue-related considerations into the delivery system design, followed by rider-facing educational interventions that raise awareness of fatigue-inducing factors and promote monitoring. Looking further ahead, the development of cumulative fatigue monitoring systems integrated into delivery platforms, as well as policy-level requirements for platforms to disclose algorithmic workload indicators to occupational health authorities, should be pursued.

Several modifiable health behaviors were independently associated with specific safety domains, even after adjusting for fatigue and other covariates. The inconsistent use of helmets was associated with a higher overall accident risk and greater levels of self-rated and other-rated accident anxiety, consistent with the evidence that helmet use remains a key protective factor for motorcyclists [10,24]. Insufficient physical activity was associated with higher overall risk and self-rated accident anxiety. This finding aligns with broader evidence linking physical activity to lower anxiety symptoms and improved mental well-being [25] and may plausibly relate to self-regulation or coping capacity relevant to safety, although the underlying mechanisms in this occupational context remain unclear and warrant further investigation. Smoking was specifically associated with higher near-miss experiences, which may reflect distraction-related or broader risk-prone patterns [11,26-28]. Together, these findings provide preliminary evidence supporting a multicomponent view of safety promotion, suggesting that fatigue management may be most effective when considered alongside targeted health-behavior strategies.

The multivariable models explained a significant proportion of the variance in the outcomes (adjusted R²=29.9% for TARI, 21.6% for near-miss experiences, 29.2% for other-rated accident anxiety, and 23.5% for self-rated accident anxiety), suggesting that fatigue and the included behavioral and occupational covariates capture important correlates of multidimensional safety. At the same time, the remaining unexplained variance indicates that additional work- and context-related determinants may also contribute to safety outcomes and should be incorporated in future studies.

We also observed a counterintuitive inverse association between longer self-reported break duration and safety outcomes, with riders reporting breaks of 2 hours or more showing higher accident risk and near-miss experiences. This counterintuitive finding should not be interpreted as evidence that shorter breaks are protective or that reducing rest improves safety among MFDWs. This finding should be regarded as exploratory and hypothesis generating, as alternative explanations cannot be excluded, including reverse causation and potential sensitivity to how break duration was categorized. One plausible hypothesis raised by this study is that the observed pattern may reflect a structural measurement limitation inherent in algorithmically managed delivery. In platform-based food delivery, periods that MFDWs label as “breaks” may frequently coincide with unpaid on-duty waiting time, during which riders remain logged into the app and available for dispatch [5]. If so, those reporting the longest “breaks” may not represent a well-rested MFDW, but rather one with prolonged on-duty availability that sustains cognitive load rather than enabling genuine recovery. Recovery research suggests that the restorative value of rest depends not only on its duration but also on psychological detachment from work demands [29], and such detachment may be structurally difficult to achieve within algorithmically managed gig work. Previous studies further suggested that prolonged idle periods can induce compensatory time pressure once orders resume, which may be associated with riskier riding behavior [4,30]. However, these explanations remain speculative, and this hypothesis requires direct empirical testing in future studies. Future studies should employ more precise measures that distinguish restorative rest from on-duty waiting, ideally complemented by objective indicators such as platform activity logs or ecological momentary assessments.

This study had some limitations. First, its cross-sectional design precluded causal inferences. However, given that MFDWs represent a particularly vulnerable occupational group in which traffic incidents frequently result in serious injuries [3], identifying strong correlates such as occupational fatigue provides critical and immediate evidence for prioritizing interventions, even in the absence of longitudinal confirmation. Second, self-reported measures may have introduced recall or reporting bias. In particular, break times may not clearly distinguish actual rest from idle waiting times. Nevertheless, it is important to note that psychological constructs, such as “accident anxiety” and “near-miss experiences,” are inherently subjective. These internal vulnerability states cannot be captured through objective accident records alone, making self-reporting an appropriate and valid assessment method. As the 2 primary psychological constructs, occupational fatigue and TARI, were assessed using a single self-report survey, the possibility of common method bias cannot be fully excluded. To examine this issue, we conducted Harman single-factor test, which showed that the first factor accounted for 45.78% of the total variance, below the commonly used 50% threshold. Third, several important unmeasured variables may limit the interpretation of our findings. Objective exposure indicators such as riding distance, number of deliveries per shift, shift timing, route type, urban traffic exposure, weather conditions, incentive pressure, and platform algorithm demands were not captured. Given that these factors may vary considerably across MFDWs and could independently influence safety outcomes, the findings regarding TARI and its subdomains should be interpreted with caution. Furthermore, residual confounding cannot be ruled out, as health behaviors such as smoking and physical inactivity may partly reflect unmeasured risk-prone tendencies, and unmeasured work-context factors may have contributed to the observed associations. Finally, because convenience sampling was employed, the findings should be generalized with caution. In particular, South Korea’s food delivery market is characterized by a predominance of algorithmic dispatch systems, peak-hour demand concentration, and highly developed road infrastructure, which may differ substantially from delivery contexts in other countries [31,32]. Therefore, findings should be interpreted with caution when applied to other national settings. Consequently, future studies should use longitudinal designs and more precise measurements of rest and workload, along with objective exposure indicators, to validate these findings. Real-time approaches, such as ecological momentary assessments and complementary objective indicators, may help clarify temporal relationships and support more actionable guidance on fatigue management and recovery for MFDWs working under algorithmically managed piece-rate conditions.

This study contributes to the literature on MFDWs by conceptualizing traffic safety as a multidimensional construct encompassing accident risk, near-miss experiences, and accident-related anxiety. Occupational fatigue was consistently associated with all the outcomes, suggesting its utility as an important indicator of elevated safety risks and accident-related anxiety. Modifiable health behaviors, including inconsistent helmet use, insufficient physical activity, and smoking, were also associated with specific safety domains, indicating the potential value of considering fatigue management alongside health-behavior approaches. The inverse association observed between longer self-reported break duration and safety outcomes warrants cautious interpretation, as break time may partly reflect on-duty waiting rather than restorative rest. Future studies using longitudinal designs and more precise measures of rest and exposure are required to clarify temporality and inform context-sensitive strategies for supporting recovery during platform delivery.