Bee sting injuries can lead to serious medical complications such as anaphylaxis, respiratory distress, and cardiac arrest, particularly among vulnerable populations, including older adults and rural residents (1, 2). They also represent a significant public health concern. In Europe, annual mortality from hornet, wasp, and bee stings has been reported to range between 0.03 and 0.48 deaths per million people, amounting to hundreds of fatalities each year (3, 4). In South Korea, Health Insurance Review and Assessment Service (HIRAS) records from 2010 to 2014 documented 78,860 injuries and 49 deaths caused by Hymenoptera stings, corresponding to an annual average of 15,772 injuries and 9.8 fatalities. The total medical costs associated with wasp stings during this period were estimated at approximately 3.2 million USD (5).

It is important to investigate the environmental and social drivers of bee sting incidents to develop effective control strategies. Previous studies have shown that recognizing such factors can inform the creation of public health policies. For instance, analyses of climatic and ecological factors related to malaria transmission have supported the implementation of vector control programs, while identifying socioeconomic inequalities in injury occurrence has guided resource allocation for prevention strategies (6). In other words, clarifying the environmental and social factors associated with bee sting injuries is essential for designing targeted interventions and evidence-based policies to reduce their public health impact.

Several factors can be regarded as potential drivers of bee stings because of their plausible underlying mechanisms. First, temperature and sunlight levels may influence bee physiology, foraging behavior, and defensive responses, making them important candidates for epidemiological studies. Elevated temperatures are generally associated with increased bee activity and extended foraging periods, which may heighten the likelihood of human–bee encounters (7). Similarly, prolonged sunlight exposure throughout the year may extend diurnal foraging windows, thereby increasing the chances of spatial overlap between bees and human populations. Excessive heat or high solar radiation can also compromise colony health and reduce floral resources, which may indirectly increase human–bee interactions as bees expand their foraging ranges into anthropogenic environments (8–10). Second, precipitation and relative humidity may also play important roles. While increased precipitation can temporarily suppress bee activity by reducing flight opportunities, it may enhance floral abundance over time, indirectly supporting larger bee populations (11, 12). High humidity levels may further benefit nest maintenance by preserving adequate moisture within hive structures (13–15). Third, the proportion of protected areas, forest cover, and deforested land reflects habitat integrity and ecological disturbance. Extensive forest and protected regions provide stable nesting habitats, sustain floral and nesting resources, and may reduce colony displacement toward human settlements (16–18). By contrast, deforestation disrupts established nesting grounds, diminishes resource availability, and may encourage migratory behavior, thereby increasing the likelihood of human contact (19–23). Fourth, urban and agricultural land use patterns represent the intensity of human activity and may be associated with bee sting occurrence. Urban areas combine high human density, ornamental vegetation, and numerous structural nesting sites, all of which can increase the frequency of encounters between humans and bees (24–27). Agricultural regions, meanwhile, are characterized by heightened sting risk among outdoor laborers due to dense flowering crops and seasonal work patterns (12, 28, 29). However, agricultural activities may also negatively affect bee colonies through pesticide and insecticide use, as well as frequent human disturbance (30, 31).

In this study, we examined the epidemiological associations between potential environmental and social factors and bee sting injuries in South Korea. The outcome distribution showed excess zeros and overdispersion, and an ordinary zero-inflated negative binomial (ZINB) model was used to jointly model the count and zero components. Although the mechanisms underlying these factors have been discussed in previous research, epidemiological studies explicitly analyzing them remain limited. South Korea provides an especially suitable setting for this investigation because it offers systematic, region-specific data on bee sting injuries along with reliable, validated datasets for environmental and social variables from multiple sources. Bee sting injuries can be situated within a broader category of animal-related injuries associated with environmental change and land-use patterns. Climate variability and land-use modification have been reported to alter animal habitats and contact opportunities with human populations (32). In this context, bee sting injuries provide a specific and observable outcome for examining population-level associations with environmental and social factors.

In total, 2,000 observations (8 years × 250 regions) were included in the analysis. During the study period, 20,161 cases of bee sting injuries were reported in 2014, declining to 10,625 cases in 2020, with a subsequent increase to 12,762 cases in 2021. The observed annual number of bee sting injuries is presented in Supplementary Table S1. Annual counts of bee sting injuries ranged from fewer than 100 to more than 300 across districts. Group comparisons indicated that districts with bee sting injuries had a significantly higher mean annual temperature (13.2 °C vs. 13.0 °C, p = 0.0002) and a significantly lower proportion of protected area (13.3 vs. 15.9%, p = 0.0002) than districts without cases, while other variables showed no significant differences (Table 2). Spatial mapping of bee sting injury incidence proportion are shown in Figure 1A. Quartile maps of temperature, humidity, precipitation, sunlight, proportion of protected area, deforested region proportion, forest region proportion, agricultural land proportion, and urban region proportion are shown in Figures 1B–J.

In the count component of the ordinary ZINB model, a 1% increase in forest coverage was associated with a 2.4% increase in bee sting injuries (RR = 1.024, 95% CI: 1.017–1.032). A 10% increase in deforested area corresponded to a 9.2% increase in injuries (RR = 1.092, 95% CI: 1.028–1.160). A higher urban proportion was inversely associated with bee sting injuries (RR = 0.974, 95% CI: 0.969–0.980), while annual precipitation also showed a negative association (RR = 0.996, 95% CI: 0.992–0.999). Annual humidity, agricultural land proportion, proportion of protected area, and annual temperature were also significantly associated with injury counts, whereas annual sunlight duration was not. In the zero-inflation component, agricultural land proportion (OR = 1.015, 95% CI: 1.000–1.030) and proportion of protected area (OR = 1.025, 95% CI: 1.008–1.042) increased the probability of observing zero cases, while a higher urban proportion decreased this probability (OR = 0.714, 95% CI: 0.593–0.859). Estimates of RRs and ORs with 95% CIs are presented in Figure 2. The Moran's I statistic for residuals from the ordinary ZINB model was 0.034 (p = 0.183), indicating no significant residual spatial autocorrelation. A spatial ZINB model was not applied because Moran's I for the ordinary ZINB residuals was not statistically significant. Temporal correlation was not modeled as a separate correlation structure. Year was included in the ordinary ZINB model to account for temporal variation.

This study examined the environmental and social factors associated with bee sting injuries across 250 administrative districts in the Republic of Korea from 2014 to 2021. A gradual decline in incidence was observed over the study period. Among meteorological variables, mean temperature was negatively associated with bee sting injuries, whereas relative humidity was positively associated. Precipitation showed a negative association, and sunlight was not statistically significant. Regarding land-use factors, a higher proportion of deforested areas was positively associated with bee sting injuries, whereas urban region (%) showed a negative association in the count component of the ordinary ZINB model. In contrast, forest region (%) was positively associated with bee sting injuries. The proportion of agricultural land showed no significant association with injury occurrence. Overall, the findings suggest that environmental and social factors influence the spatial distribution of bee sting injuries, although some variables exhibited null associations.

Temperature was negatively associated with bee sting injuries in this study. A hospital-based analysis from Switzerland reported that the frequency and severity of Hymenoptera stings increased with higher daily temperatures, showing a marked rise when the mean temperature exceeded 24 °C (44). Similarly, population-level evidence from Alaska indicated that warmer summer periods were accompanied by a greater number of insect sting reactions requiring medical attention (45). These findings suggest that elevated temperatures are generally associated with an increased occurrence of sting-related injuries across diverse climatic regions.

Relative humidity showed a positive association with bee sting injuries in the present study. By contrast, hospital-based surveillance in Switzerland reported that lower humidity was associated with a higher frequency of bee sting injuries and more severe reactions (44). Meanwhile, reports from Northeastern Brazil documented seasonal clustering of bee sting injuries during warm and humid months, although humidity-specific estimates were not calculated (46). These findings indicate heterogeneity in the relationship across studies—an inverse pattern in a temperate European setting and higher occurrence in subtropical regions during humid periods. Direct epidemiologic evaluations of humidity in relation to bee sting injuries remain limited. Further investigations, including meta-analyses, may be warranted to clarify this association because such approaches allow the integration of heterogeneous findings and the assessment of consistency across studies.

Precipitation was negatively associated with bee sting injuries in this study, although epidemiological evidence on this relationship remains limited. In Switzerland, hospital-based surveillance of Hymenoptera stings reported that days with sting-related admissions had lower mean rainfall than non-sting days (p = 0.007) (44). Similarly, a Korean study of emergency calls for wasp stings identified an inverse association between daily rainfall and sting-related incidents, supporting the observation that increased precipitation suppresses sting occurrence (5). A study of European wasp ecology in metropolitan Adelaide also reported a weak negative correlation between total rainfall and nest destruction, though it did not examine human sting incident rates (47). Thus, consistency with the present findings is partial. Overall, the current results align with general seasonal patterns described in previous research, while differing in that precipitation was explicitly included as an explanatory factor in this analysis.

The proportion of protected areas was also associated with a higher occurrence of bee sting injuries in this study. In Sri Lanka's central hill country, sting injuries were reported within the Victoria–Randenigala–Rantembe Sanctuary, a designated forest reserve, with cases presenting to regional hospitals during periods of field activity in adjacent settlements (48). In the United States, surveillance of visitor injuries in Shenandoah National Park recorded cases of respiratory distress following insect stings or bites—including Hymenoptera stings—indicating that sting-related medical encounters can occur within large protected regions (49).

The proportion of deforested areas was positively associated with bee sting injuries in the present study. Previous epidemiological research has reported higher sting incidence in regions characterized by intensive human land use but has not directly assessed the proportion of deforested areas. In Ceará, Brazil, 1,307 cases were recorded between 2007 and 2013, with a predominance in urban regions (60.7 vs. 34.5%) (50). Similarly, in Campina Grande, Paraíba State, incidence was higher in urban than in rural zones, although deforested area was not evaluated as a specific variable (46). Thus, consistency with the present findings was partial—prior studies have noted elevated sting incidence in human-modified environments, but the proportion of deforested regions has rarely been explicitly examined.

Forest coverage was associated with a higher incidence of bee sting injuries in this study. In Sri Lanka's central hill country, a hospital-based analysis reported that most sting injuries occurred during outdoor activities in tea estates and settlements located within or near forest reserves, indicating frequent human exposure in forest-associated environments (48). Another study conducted in Anuradhapura District, Sri Lanka, found that 9.6% of all hospital admissions over 1 year were attributable to hornet stings, with most incidents occurring in paddy fields and chena cultivations situated near tropical dry evergreen forests (51). In Japan, an occupational epidemiological survey showed that more than 90% of forestry workers had experienced Hymenoptera stings, and systemic reactions were observed in 21% of them—a substantially higher proportion than among office-based controls (52). In Switzerland, a 5-year hospital-based analysis reported that 34% of patients presenting with Hymenoptera stings exhibited anaphylactic reactions, with a higher frequency of incidents in peri-urban and rural areas located near forested landscapes (44).

Bee sting injuries showed lower counts in districts with a higher proportion of urban areas according to the present findings. In Brazil, ecological studies reported that most bee sting injuries occurred in urban settings, suggesting greater exposure opportunities in densely populated regions (46). In Korea, severe systemic reactions following bee sting injuries were most often managed in metropolitan hospitals, reflecting the concentration of cases in large urban centers, despite the absence of spatially continuous urbanization indicators (53). Additionally, an analysis of emergency calls for wasp nest removal in Korea demonstrated a strong correlation with sting-related injuries in densely populated urban areas, further emphasizing the burden in metropolitan contexts (5). Although these studies differed in outcome definitions and exposure assessments, they consistently reported a high burden of bee sting injuries in urban environments. The consistency across different settings highlights the high burden of bee sting injuries in urban environments, although the present study showed a negative association between urban region (%) and injury counts in the count component of the ordinary ZINB model.

The proportion of agricultural land was not associated with bee sting injuries in this study. In Sri Lanka, hospital-based records from Anuradhapura District showed that a large proportion of sting injuries originated in paddy fields and chena cultivations within agricultural landscapes contiguous with forested regions (51). In Spain, a population survey in rural Mediterranean regions reported a higher prevalence of Hymenoptera sting reactions among agrarian populations compared with urban groups (54). Similarly, a population-based cohort study in Germany found increased Hymenoptera venom sensitization and systemic reactions among adults, with rural residence identified as a contributing factor (55). These findings contrast with the absence of a significant association in the present study and suggest heterogeneity across settings, likely reflecting differences in agricultural practices, exposure environments, and study designs.

These findings may be interpreted within the context of urbanization and land-use patterns characteristic of South Korea, where high-density residential development is closely interwoven with managed green spaces and peri-urban natural environments (56). Differences between highly urbanized districts with formalized environmental management and rural or forest-adjacent districts with comparatively limited protective infrastructure may be relevant to exposure conditions among districts reporting bee sting injuries (57). Furthermore, the concurrent associations observed for forest coverage and deforested areas may be considered in relation to landscape fragmentation in South Korea, where extensive forest cover coexists with localized land disturbance, rather than as opposing ecological processes (58).

This study has several limitations. First, the analysis was based on emergency department–derived surveillance data, which excluded cases treated in outpatient settings and likely led to an underestimation of bee sting injuries. Bee sting injury reporting during 2020 and 2021 may have been influenced by the COVID-19 pandemic and associated public health regulations, which could have affected healthcare-seeking behavior and outdoor activity patterns. As an ecological study, the present analysis is also subject to limitations in causal interpretation at the individual level, and residual confounding by unmeasured factors cannot be excluded (59, 60). Previous research has shown that incidence rates of psychiatric disorders were lower when outpatient data were omitted, suggesting a similar limitation for emergency department–restricted surveillance (61). Second, variations in access to emergency medical facilities—particularly in rural or remote regions—may have contributed to underreporting or misclassification. Evidence from healthcare access studies indicates that geographic barriers and limited regional connectivity are associated with reduced utilization of services in underserved areas (62). Third, unmeasured ecological and socioeconomic factors, such as floral resource availability, pesticide use, and microclimatic variability, were not incorporated. Methodological studies have demonstrated that residual and unmeasured confounding can account for variability in observed associations (63). Finally, the use of aggregated district-level data limited analyses to broad spatial scales and did not capture individual-level exposures, reducing the precision of the associations.

Despite these limitations, the results indicate that environmental, social, and land-use factors are associated with the occurrence of bee sting injuries and have important implications for prevention and planning. Regions characterized by higher urbanization or ongoing deforestation may require targeted awareness campaigns and strengthened emergency medical preparedness (64). The observed associations with temperature, humidity, and precipitation suggest that climate variability may influence future patterns of bee sting injuries (5). These findings underscore the importance of adaptive risk monitoring. Further research is warranted to incorporate outpatient and primary care data to improve case ascertainment, and to include additional variables such as floral abundance, pesticide exposure, and socioeconomic indicators, and to utilize spatially refined datasets for greater resolution (16, 65). Field-based investigations and multidisciplinary approaches—including behavioral and genetic studies of bee populations—are needed to provide a more comprehensive understanding of the ecological relationships underlying these patterns (66, 67). Such efforts are expected to support the development of region-specific strategies, enhance early warning systems, and inform policy measures addressing the public health burden of bee sting injuries under diverse environmental and social conditions.