حوادث ایمنی منفی به عنوان عوامل تنش ایمنی کار

We investigate the extent to which three types of self-reported negative safety events (i.e., being injured at work, working in an unsafe way, witnessing others working in an unsafe way) correlate with work-safety tension. Work-safety tension is bi-dimensional, defined here as the perceived conflict between production and following safety rules (“barriers to safety compliance”) and production and proactive ways of working more safely (“barriers to safety participation”). Directly experiencing or witnessing negative safety events may send signals to employees about the extent to which their organization prioritizes production over safety. We tested a model of negative safety events as predictors of both barriers to safety compliance and barriers to safety participation using survey data from 316 front-line supervisors (97% male, mean age = 44 years) working for a UK rail maintenance company. The number of injuries directly experienced had a positive relationship with perceived barriers to safety compliance, whereas the number of times respondents witnessed others work in an unsafe way had a positive relationship with perceived barriers to safety participation.

Employees in many organizations experience a tension between completing work tasks (production pressure) and completing work tasks safely (e.g., Brown et al., 2000, Gouldner, 1954 and Janssens et al., 1995). This work-safety tension can translate into less attention and less energy devoted to ensuring work tasks are conducted safely (Probst, 2002), with growing evidence that such trade-offs encourage unsafe work behavior (e.g., Clarke, 2006, McLain and Jarrell, 2007, Morrow et al., 2010 and Seo, 2005) such as short-cuts or workarounds (Halbesleben, 2010), and may result ultimately in more workplace injuries (Humphrey et al., 2004). Much research has focused on work-safety tension and other dimensions of safety climate as determinants of negative safety events like work injuries and unsafe work behavior (McGonagle and Kath, 2010, McLain and Jarrell, 2007 and Morrow et al., 2010). However, cross-sectional research in this vein makes two assumptions about work-safety tension. First, it assumes that work-safety tension is unidimensional, despite related research that acknowledges the multi-dimensional nature of safety climate – a shared sense of how the organization values safety (Guldenmund, 2007 and Zohar, 1980) – and employee safety performance (Christian et al., 2009). Second, it assumes work-safety tension precedes negative safety events – such that work-safety tension predicts negative safety events. In a recent meta-analysis comparing safety climate → injuries and injuries → safety climate relationships, Beus et al. (2010) showed that injuries → safety climate had a stronger overall relationship than safety climate → injuries, suggesting that negative safety events may not only be caused by but also contribute to perceptions of safety climate. The current study makes two contributions. First, it measures work-safety tension as two dimensions, consistent with bi-dimensional models of employee safety performance. Second, it investigates how constructs traditionally considered as outcomes – workplace injuries, unsafe behaviors, witnessing others work unsafely – predict work-safety tension. This model invokes social learning and reinforcement theories to justify why negative safety events may contribute to employee perceptions of work-safety tension.

5.1. Measurement model To establish a measurement model for the seven items developed to measure work-safety tension, we conducted an exploratory factor analysis on a randomly-selected half of the sample. As recommended by Conway and Huffcutt (2003), principal axis factoring and a scree plot were used in the extraction process, with an oblique rotation employed given the likely non-zero correlation of the underlying constructs. The best solution, explaining 69% of the variance, comprised 2 factors of work-safety tension: one measured by four items, interpretable as ‘barriers to safety compliance’; the other, by three items measuring ‘barriers to safety participation’. A confirmatory factor analysis was performed on the other half of the sample (to sidestep the upward bias on fit caused by building and testing the model on the same respondents); this verified the adequacy of this 2-factor model (chi-square = 28.73 on 13 df, comparative fit index [CFI] = .95, root mean squared error of approximation [RMSEA] = .09, standardized root mean square residual [SRMR] = .05) using the indices and benchmarks suggested by Hu and Bentler, 1998 and Hu and Bentler, 1999 as a guideline for assessing fit. There was a strong positive correlation between the factors (r = .62) though not approaching levels that would suggest they were not distinct ( Kline, 1998); their divergent validity was supported by a 1-factor solution indicating a substantially weaker fit (chi-square = 44.51 on 14 df, CFI = .91, RMSEA = .12, SRMR = .06). A multi-group confirmatory factor analysis indicated that the structural components of the measurement model (item-factor loadings and factor inter-correlations) were invariant across the two largest occupational groups in this sample; that is, items comprising the two factors were commonly understood, grouped together in the same way for different occupations, and were thus valid to measure differences between them. Both sets of items achieved satisfactory levels of internal consistency reliability (barriers to safety compliance: α = .70; barriers to safety participation: α = .83) (see Fig. 1).The incidence of negative safety events, correlations between these scales, and potentially confounding variables are given in Table 1 alongside sample statistics. Employees reported an average of .68 reportable injuries in the last year (SD = 2.02, range: 0–15), .88 incidences of witnessing unsafe behavior enacted by others (SD = 2.56, range: 0–25), and .11 incidences of personal unsafe behavior at work (SD = .53, range: 0–5). At the zero-order level, greater incidence of workplace injuries is correlated with greater perceptions of barriers to safety compliance (r = 0.169, p < .05), while greater incidence of witnessing unsafe behavior is correlated with greater perceptions of barriers to safety participation (r = 0.174, p < .05).5.2. Structural model We then formulated and tested a structural equation model using the whole sample. We extended the confirmatory factor analysis model by regressing the factors for barriers to safety compliance and barriers to safety participation on the log of workplace injuries, the log of unsafe behavior, and the log of witnessed unsafe behavior, while controlling for age, occupational group, and geographical territory (the latter two variables were dummy-coded). Again, the model fit appeared adequate, χ2(68) = 135.66, CFI = 0.92, RMSEA = 0.06, SRMR = 0.03, and with fit indices meeting standards of robustness given model complexity and sample size ( Hu and Bentler, 1999). The path coefficients are presented in Table 2. The model predicts 13.9% of the variance in barriers to safety compliance and 13.7% of the variance in barriers to safety participation. Having accounted for background variables of age (linear and curvilinear), occupational group, and geographic territory, barriers to safety compliance are predicted by workplace injuries experienced (in partial support of H3), and barriers to safety participation are predicted by witnessing others behave in an unsafe way (in partial support of H1)