Strategies for construction hazard recognition

68 Crews 1 and 2 demonstrated a level change improvement immediately after receiving the intervention of 27% and 31%, respectively. Since Model II was ideal for both cases, it can be inferred that both crews did not reveal a slope change in performance. Crew 3, however, revealed a level change improvement of 27% and also revealed a significant slope change coefficient of
3.309 (p<0.05). This implies that crew 3 had a statistically significant improvement of 3% (3.309
– 0.314) in each successive work period following the intervention. The weighted overall level change improvement, calculated using Equation 3, indicated that the crews improved by 28%
(p<0.05).
Similarly, the corroborative two-sample t-test for independent measures using photographs of construction scenarios revealed statistically significant improvements among all crews. Crew 1,
2 and 3 exhibited an increase in hazard recognition of 41% (p<0.05), 34% (p<0.05), and 44%
(p<0.05), respectively.

Potential limitations and methods used to improve validity and reliability
A primary issue inmost observational studies is the observer effect where subjects modify their behavior due to the presence of an external researcher. Although literature vastly suggests concealing the observer by using hidden cameras or other remote equipment (Desai 2002), it was impossible to implement such methods due to disclosure and confidentiality requirements imposed by the contractors. Instead, we adapted other methods that have been vastly supported in literature to minimize observer effects. First, we included in-house safety managers who were familiar to the workers and frequently conducted such safety observations as required of their role (Pellegrini and Bjorklund 1998). Second, the longitudinal nature of the study allowed us to

69 introduce the safety intervention after the researchers frequently visited the site and became familiar with the workers. Third, the selected sites were accustomed to external visitors for various assessment and training purposes, and the workers were unaware that data was being collected for this research effort. It is important to note that a significant improvement in the HR index was measured for each crew although the external researcher was present for several work periods before the intervention. That is, the observer effects may have existed in both the baseline and the intervention phase. Yet, a significant improvement in hazard recognition was caused by the intervention Also, similar improvements in the HR index were observed in the validation study using construction images. Finally, although research participants tend to work harder to attain higher performance levels due to the hawthorne effect (Roethlisberger and Dickson 1939), skill- sets such as hazard recognition is not likely to be altered especially given that their performance was significantly lower in the baseline phase. Similarly, hazard recognition skills cannot improve simple because the workers understand the purpose of this research effort. One significant limitation of this study is the method used for the computation of the HR index. We used a site-based panel with significant safety experience to identify and catalogue the hazards encountered by the workers, however, there is a possibility that some hazards were not identified. This maybe the case because safety professionals, like workers, may not recognize every hazard in dynamic environments. Nevertheless, we followed a replicable and rigorous protocol using the same panel throughout the study to make reliable inferences. Also, it is

70 expected that the two safety managers, the researchers, and the workers may have recognized all identifiable hazards when all data were combined. Despite the longitudinal nature, another limitation associated with this study is that the long-term effects of the intervention. While the study clearly indicates a statistically significant improvement in the hazard recognition capability of workers, the sustenance of this impact maybe questionable. Unfortunately followup studies, in this case, were not feasible because several crews that received the intervention were dispersed after the study. Also, dropout rates among some crews were very high. Although interviews with the safety managers appear to indicate that the intervention effects will sustain, constant reinforcement and site leadership commitment is indispensable. We endeavor to conduct followup studies to ascertain the impacts of such intervention over longer periods.

Download 2.75 Mb.

Share with your friends: