Childhood injury is one of the leading causes of death and disability among children globally, constituting a major public health concern (1). Product-related injuries represent a significant component of childhood injury and are closely associated with children’s developmental stages, behavioral characteristics, and environmental exposures (2). The widespread implementation of non-pharmaceutical interventions (NPIs) such as school closures, stay-at-home orders, and restrictions on gatherings during the COVID-19 pandemic altered children’s activity spaces and patterns of product contact in a short period, which created a unique natural experiment to investigate the impact of macro-level social policies on the risk of childhood injury. Multiple studies have reported an overall decline in trauma admissions during the NPIs period (3, 4), along with a noticeable shift in the injury spectrum: road traffic injuries and sports-related trauma decreased substantially (5), while unintentional home injuries and violence-related injuries—particularly firearm injuries—increased (6, 7). However, existing research has largely been confined to describing changes in injury rates or proportional distributions, with limited in-depth assessment of changes in injury severity. More importantly, few studies have employed multivariable models to control for demographic and other confounding factors to examine whether the NPIs period independently influenced the risk of injury associated with specific products or locations, and whether effect modification exists. Therefore, utilizing long-term systematic injury surveillance data from Xiamen City, China, this study aims to investigate: (1) the structural changes in the epidemiology of product-related injuries among children—including but not limited to shifts in product types, injury locations, and population distributions—before, during, and after the implementation of NPIs; and (2) whether the NPIs period independently increased the risk of severe injury and modified the associations between product type, injury characteristics, and severe outcomes, after adjusting for demographic and injury-related confounders.

This study utilized large-scale injury surveillance data to elucidate the dynamic impact of COVID-19 NPIs on product-related injuries among children. Our findings reveal that NPIs not only altered the epidemiological distribution of injuries but also reshaped the risk landscape of injury severity, with the NPIs period acting as a powerful effect modifier. This shifting injurious potential suggests that macro-level social policies can fundamentally redistribute health risks across different environments and products.

A key finding is that while the NPIs period showed no significant independent association with injury severity in the main effects model, its interactions with injury location and product category were highly significant. This suggests that NPIs did not exert a uniform “blanket effect” on severity; instead, the lack of a significant main effect likely reflects competing countervailing trends: a substantial reduction in high-risk outdoor injuries (e.g., motor vehicle crashes) due to restricted mobility was statistically offset by a concurrent intensification of injury severity within the home environment.

The restructuring of activity spaces led to a marked concentration of injuries at home (peaking at 56.38%), aligning with the global “lockdown effect” (11, 12). However, our interaction analysis further delineates that the severity risk for injuries involving children’s toys (aOR = 3.47) and household appliances (aOR = 4.20) surged during NPIs. This intensification likely stems from dual mechanisms: “increased exposure dose” due to prolonged confinement and “caregiver fatigue” during remote work, which created supervision gaps and may have delayed intervention for minor incidents (13–17). Unlike pre-pandemic patterns, the home environment under NPI conditions became a site where domestic hazards possessed a higher potential to escalate into severe outcomes.

Conversely, the interaction analysis reveals a significant “Safety Resilience” in public institutions. Despite a post-pandemic rebound in the proportion of school-related injuries (18.07%), which aligns with other research (18), the risk of these injuries resulting in severe outcomes decreased by 53% (aOR = 0.47) compared to the pre-pandemic period. This seemingly paradoxical phenomenon of “high frequency but low severity” suggests that schools and public venues successfully institutionalized strengthened safety protocols and emergency response density during the pandemic (19). In contrast, NPIs triggered a “spillover effect” into informal spaces; when conventional recreational venues were restricted, children shifted toward “other/unclear” locations (e.g., agricultural sites, woodlands) where the severity risk rose significantly (aOR = 2.58) due to a lack of safety regulation and environmental adaptation (20).

Beyond the reshaped risk within built environments, the interaction analysis also captured the severe consequences of population displacement into natural or productive landscapes. The 15-fold surge in severity risk for agro-forestry-fishery products (aOR = 15.59) during NPIs highlights the impact of large-scale decentralized migration. Multiple international studies have confirmed a significant “decentralized” migration trend during COVID-19 lockdowns, with a massive outflow of people from high-density urban centers to suburban or rural areas (21, 22). That outflow exposed children to unfamiliar farming tools and more hazardous natural environments (23, 24). This “environmental mismatch” underscores that population shifts under restrictive measures require targeted safety education to bridge the gap between children’s prior experience and their new, more hazardous surroundings.

The robustness of these conclusions is supported by our multi-dimensional sensitivity analyses. The consistency of findings under a stricter outcome definition (Sensitivity Analysis 2) serves as a rigorous internal validation, confirming that the observed effect modifications are not artifacts of injury classification but reflect substantive changes across different clinical thresholds. Furthermore, the non-significant interaction between age group and period suggests that while NPIs altered activity scope, they did not modify the inherent developmental susceptibility to injury across the pediatric age spectrum.

This study has clear implications for public health practice. Prevention strategies should transition from generalized “home safety” to “precision intervention,” emphasizing adherence to toy safety standards, proactive safeguarding of household appliances, and environmental adaptation for families relocating during emergencies. Moreover, the experience of reduced injury severity in public places post-pandemic warrants institutionalization into sustainable safety standards.

This study also has several limitations. First, data were sourced from a single regional surveillance system, warranting caution regarding generalizability. Second, although we controlled for multiple confounders using multivariable models, unmeasured variables such as quality of home care, parental psychological stress, and socioeconomic status could not be included, potentially leading to residual confounding. Furthermore, as an observational study based on administrative data, we could not directly measure specific behavioral mediators of NPIs (e.g., actual time children spent at home, duration of online learning, or outdoor activity frequency). Therefore, although the observed changes in injury patterns (e.g., increased home injuries, decreased outdoor transport injuries, increased risk from home-related products) are highly consistent with the theoretical mechanism of “activity space shift to home” and align with global trends, alternative explanations cannot be entirely ruled out. This limitation primarily affects the in-depth interpretation of causal pathways but has a relatively limited impact on the main finding of a strong association between the NPIs period and injury patterns. Lastly, defining the entire 2020–2022 period as the “NPIs implementation period,” while based on the overarching presence of various NPIs, did not capture temporal variations in NPI intensity. China’s prevention policies followed a “dynamic zeroing” principle, encompassing forms ranging from comprehensive strict lockdowns to precise zonal control (25). This heterogeneous exposure definition may have led to a degree of non-differential misclassification, which typically biases effect estimates toward the null (i.e., dilution bias), potentially underestimating the true modifying effect of NPIs. However, this definition aligns with our research objective of examining the macro-policy “overall shift”. Future research could integrate multi-regional data and incorporate individual-level behavioral measurements, while adopting the ‘Stringency Index’ from the Oxford COVID-19 Government Response Tracker (OxCGRT) to further elucidate the underlying dynamic causal pathways (26).

In summary, COVID-19 NPIs dynamically reshaped the risk determinants of product-related injuries in children. By modifying the injurious potential of specific settings and products, NPIs led to a sharp increase in home-related injury severity while potentially catalyzing long-term improvements in public environmental safety. These findings call for the integration of targeted, context-specific injury prevention strategies within future public health emergency responses.