Predicting Safety Risk of Working at Heights Using Bayesian Networks.

Although the construction industry has shown significant improvements in safety performance over the past 30 years, falls are still a leading cause of fatalities and serious injuries. Previous studies have focused on identifying factors affecting the risk of falls, but remained silent on investigating the evidential relationships among these factors to better prevent fall accidents. This research proposes a Bayesian network (BN) based approach to diagnose the accident risk of working at heights. The proposed approach consists of a conceptual and generic model with a protocol for assessing the risk of falls and a computational module. The generic BN model was developed on the basis of an extensive review and evaluation of causal factors leading to falls. The computational module was developed on the basis of Bayes' rule for inference to customize model input and job site characteristics. The results of the proposed approach provide probabilities associated with different states of safety risk. Additionally, sensitivity analysis allows practitioners to identify appropriate preventive actions and safety strategies to reduce risk of fall. The proposed approach was verified and tested with a construction operation in a condo-hotel project. This study contributes to the construction safety body of knowledge by providing an effective quantitative risk assessment tool to predict the safety risk of falls from heights. Researchers and practitioners may customize the model to assess and benchmark the fall risk for different operations in the construction industry. [ABSTRACT FROM AUTHOR]