Defining and Testing a Safety Cognition Framework Incorporating Safety Hazard Perception

AbstractThere has been insufficient research focusing on checking the reliability of construction employees’ hazard perceptions by comparing them with empirical safety data. There have also been li...

studies, more research could be performed in investigating the perceptions of workers and 48 other site employees' towards these commonly encountering hazards or accidents. 49 Specifically, more studiesare needed to investigate how the feature of the hazard or accident 50 (e.g., its frequency of occurrence) would affect site employees' safety hazard perception, as  100 further identified on construction sites' safety climate from employees with and without 101 management roles, as well as workers from various building trades (e.g., mechanical).

102
Safety climate can be measured by site employees' safety perceptions (Zohar, 1980;103 Brown and Homes, 1986). Employees' safety perceptions are formed by their own patterns of 104 safety cognition when exposed to hazards (Liu, 2018). Implicit memory influences and has a 105 significant role in unconscious cognition when making judgements (Jacoby and Witherspoon,  climate could be divided according to subgroup categories (Schein, 1996) This empirical study of construction employees' safety perceptions towards given site 129 hazard/scenarios started from the defined theoretical cognition framework illustrated in Fig.1.

130
The framework was set in the context of safety culture which was defined byMarquardt et al.   Fig.3 are discussed of their effects in workers' perceptions towards certain given hazards. The effect of employees' experience level on safety perceptions is also investigated leading to further discussions of how safety training could be implemented effectively to 169 employees by considering these internal influence factors (e.g., site experience). Fig.4 also 170 indicates that employees' experience is reflected through their cognition pattern.    Table 1. Following the questionnaire formatting procedure described in the study of Han et al. research students and academic staff. Coordinated by site managers, they were also given the 244 guide that they could either decline the survey request or accept to start the survey by 245 providing the answers to their best knowledge. They were also made aware that they could Since employees' perceptions towards each hazard were further categorized as occurrence, 295 severity, and controllability, the ranking deviation for each hazard between employees' 296 perception and the empirical ranking from Table 1   The detailed sample attributes are summarized in Table 2.  Table 3 were computed based on the five-point Likert-317 scale.

318
<Insert Table 3 Table 1. It was also found that fall (i.e., H2) ranked top among the 327 eight hazard scenes, with the lowest item-total correlation coefficient. The overall survey 328 population was then divided into two sub-samples according to their site experience level (i.e., 329 below or above five-year site experience). between less-experienced and more experienced employees in terms of the hazard occurrence.

333
It is seen in Table 4 that those with more than five years' site experience perceived all the 334 eight hazards with higher occurrence. Nevertheless, the two subsamples identified the same 335 three top-ranked hazards (i.e., fall, struck-by, and injuries caused by heavy equipment) with 336 highest occurrence. Among the top-ranked three hazards, two of them (i.e., fall and struck-by) 337 were consistent with the empirical ranking in Table 1. Fig.7 further compares the rankings in 338 terms of occurrence to each hazard scene among the empirical data, the overall sample, 339 subsample with those having more than five-year experience, and subsample with less than 340 five years' experience. Except fall and struck-by which received more consistent perceptions from subsamples, other hazards were perceived more differently, either between subsamples, or between the empirical data and employee perceptions.

Statistical analysis of site employees' perceptions towards hazards' severity 351
For the same eight defined hazards or accident scenes, the same statistical methods were 352 applied. The overall Cronbach's Alpha value at 0.8620 displayed in Table 5 indicated an 353 excellent internal consistency. 354 <Insert Table 5 here> 355 The individual Cronbach's Alpha value for H1 indicated that employees' perceptions 356 towards structural collapse tended to differ as they would perceive other hazards.

357
Electrocution was ranked as the most severe hazard, while hit by site vehicles were perceived 358 least severe. The subgroup difference is summarized in   The individual Cronbach's Alpha value for H1 higher than the overall one indicated that 383 employees held a differed view towards the controllability of structural collapse. Generally 384 consistent to the risk controllability ranking in Table 1, structural collapse was identified as 385 one of the least controllable hazard. Electrocution, although perceived with the highest degree 386 of severity by the overall sample, was also ranked as the most controllable hazard. The 387 subgroup analysis is summarized in Table 8.   401 The overall deviations bewteen subsamples, as well as between employee perceptions and 402 the empirical data are summarized in Fig.10.   According to Fig.2 and Fig.3, construction employees formed their hazard perceptions by 460 using their own safety knowledge, prior scenarios, or basic assumption. For more 461 experienced workers, they were more likely to apply their previous scenarios to match the 462 current hazard scene to form their perceptions leading to decision making. Less experienced 463 employees were more likely to perceive risks according to their basic assumption with 464 limited safety knowledge or prior scenarios to apply. As a result, hazard perceptions of less 465 experienced employees might be more biased and deviate more from the empirical data.

468
This study defined a theoretical safety cognition framework by addressing construction 469 employees' safety perceptions towards given site hazards within the context of safety culture.

470
The initially defined framework was then applied through site investigations in China's 471 construction industry. Firstly, the empirical safety accident data were collected and analyzed 472 in terms of the occurrence, severity, and controllability. Eight hazard or accident scenes (e.g.,