Predictive validity of safety leading indicators: Empirical assessment in the oil and gas sector

Improving safety performance is paramount to the success of oil and gas construction projects. Although considerable attention has been paid to developing and using new types of safety performance indicators, contractor safety has traditionally been measured and managed through lagging indicators. Alternatively, risk mitigation measures can be used to predict safety performance as leading indicators. Several research studies have been performed on safety leading indicators; however, no research has empirically validated the predictive validity of candidate indicators. To address this knowledge gap, empirical data were used to measure both potential safety leading and lagging indicators in an effort to test the hypothesis that variability in candidate safety leading indicators predicts variability in lagging indicators of safety performance. A total of 261 contractors were included in the study, with more than 60,000 data points. Using principal factor analysis, model building, and regression, factors with predictive power were identified. The predicted total recordable incident rate (TRIR) and severity rate (SR) provided strong correlation with the actual TRIR and SR, with coefficients of 0.7251 and 0.5338, respectively. The models of predictive TRIR and SR were validated and suggest a good fit when applied to new contractor safety data set. The models can be used as new safety leading indicators that provide warning signs for weakness in contractor safety performance. Consequently, clients can use the results to implement adaptive risk mitigation, reduce incidents, and continuously improve contractor safety performance. This research contributes to the existing body of knowledge by empirically identifying safety leading indicators, testing their efficacy, and validating the resulting models. This research set a foundation for establishing quantitative safety leading indicators in the construction industry in the future for other forms of project outcomes.