Heat Exposure and Fall Accidents in Construction: Informing Occupational Safety and Health Policy in Korea

As the frequency and intensity of heat waves increase due to climate change, concerns regarding construction worker safety are growing. While research on heat-related illnesses is extensive, studies examining the impact of heat exposure on construction fall accidents—the leading cause of occupational fatalities—remain limited. This study investigated the relationship between daily maximum temperature and fall accidents. We analyzed 42,220 fall accidents reported to the Korea Occupational Safety and Health Agency (KOSHA) between 2015 and 2019. Daily maximum temperature and relative humidity data were linked to accident dates. The association between temperature and fall accidents was examined using negative binomial regression with a segmented term at the identified threshold. Exploratory quadratic and segmented regression models were used to identify the temperature threshold, and stratified analyses were conducted by age group and workplace size. A nonlinear relationship was observed, with fall accidents peaking at approximately 29°C, followed by a plateau. The segmented regression model incorporating a knot at 29°C demonstrated the highest explanatory power (R 2 =0.894). Negative binomial regression confirmed significant overdispersion (α=0.123) and indicated that each 1°C increase in temperature above 29°C was associated with a 16.7% increase in fall accident counts (IRR=1.167, 95% CI: 1.145–1.188). This relationship was most pronounced among workers aged 50–59 years (R 2 =0.573) and ≥60 years (R 2 =0.593), as well as in workplaces with fewer than 50 employees (R 2 =0.589–0.595). Construction fall accidents were concentrated around a daily maximum temperature of 29°C, which is lower than the current heatwave advisory threshold (apparent temperature: 33°C). The stronger associations observed among older workers and in small-scale workplaces indicate that prioritized protection for these groups is necessary.