Re-Examining the Association between Quality and Safety Performance in Construction: From Heterogeneous to Homogeneous Datasets

Abstract

Recent research undertaken revealed that a significant positive relationship exists between quality and safety performance. A major limitation of this research, however, was the nature of the sample; it was heterogeneous (i.e., a combination of U.S. and international projects) and the sample was restricted to 18 projects. Building upon initial research, this paper re-examines the association between quality and safety using a homogeneous sample of 569 projects, which were derived from an Australian construction company with an annual turnover in excess of $1 billion Australian dollars (AU$). A total of 19,314 nonconformances and 17,783 injuries were used to determine the validity and reliability of previous research. A weak association between quality and safety performance was found (p < 0.01). The p-values did not indicate any significant association between first aid and quality rates, except for the injury rate and rework frequency per million scope, which yielded an r-value of 0.307 and p-value 0.046 that is significant at 0.01 level. An association, however, between injuries and rework was identified (r2=0.70). The discrepancy between this research's findings and that of previous work led to an examination of the issues of using ratios in correlation analysis. Thus, the statistical and arithmetic issues associated with the use of ratios are discussed, and it is recommended that estimating the relationships between quality and safety should be examined using regression techniques or analysis of covariance. Linear regression, therefore, was performed with the injury data as the dependent variable, and rework frequency and personnel hours as the independent variables. The regression results demonstrated that there is a significant association between injuries, and rework and personnel hours; it was revealed that both predictors accounted for 68.2% of the explained variability in injury frequency. The replication of t