Improving Safety Performance of Construction Workers through Learning from Incidents

Abstract

Learning from incidents (LFI) is a process to seek, analyse, and disseminate the severity and causes of incidents, and take corrective measures to prevent the recurrence of similar events. However, the effects of LFI on the learner's safety performance remain unexplored. This study aimed to identify the effects of the major LFI factors on the safety performance of workers. A questionnaire survey was administered among 210 construction workers in China. A factor analysis was conducted to reveal the underlying LFI factors. A stepwise multiple linear regression was performed to analyse the relationship between the underlying LFI factors and safety performance. A Bayesian Network (BN) was further modelled to identify the probabilistic relational network between the underlying LFI factors and safety performance. The results of BN modelling showed that all the underlying factors were important to improve the safety performance of construction workers. Additionally, sensitivity analysis revealed that the two underlying factors-information sharing and utilization and management commitment-had the largest effects on improving workers' safety performance. The proposed BN also helped find out the most efficient strategy to improve workers' safety performance. This research may serve as a useful guide for better implementation of LFI practices in the construction sector.