Explanation and prediction of accidents using the path analysis approach in industrial units: The effect of safety performance and climate

Abstract

BACKGROUND:

The safety climate in an organization depends on people’s understanding of the safety policies and procedures, as well as the value, importance, and priority of safety in the workplace.

OBJECTIVE:

This study aimed to describe and predict accidents using the path analysis model (PAM) in industrial units though the analysis of the effect of safety performance and climate.

METHODS:

This cross-sectional study was conducted on 294 workers in industrial units in Hamadan, a province in the western part of Iran. The data on safety performance and climate was collected using a questionnaire. The first part of the questionnaire was a short version inventory (with 25 items on safety climate) that was used to assess five variables of management commitment, supportive environment, training, personal safety prioritization, and perceived work pressure. Moreover, the safety performance was measured using 10 items on safety rules and participation. The PAM was used to describe the effects of safety climate and performance on accidents.

RESULTS:

The results showed that the safety climate had the strongest negative impact on work pressure and safety compliance toward accident, followed by safety participation, and quality of training. Moreover, the negative influence of safety climate on accident was mainly mediated by two variables: work pressure and safety participation toward accident. The work pressure had the strongest indirect and total influence on accidents. However, none of the variables had a direct effect on accidents. Training was the most important direct cause of promoting personal safety priority. The safety compliance was more effective than safety participation in reducing accidents rates.

CONCLUSIONS:

Therefore, it seems that perceived work pressure has an indirect effect on accidents which is mediated by other variables, mainly personal safety priority and safety performance.

Keywords

Prediction of accidents risk assessment safety climate path analysis model

1 Introduction

With the development of technology, risk potential has been increased in industrial environments. Many studies have been conducted so far to identify the causes of accidents in various industries. They have shown that more than 85% of accidents are caused by unsafe acts [1]. Studies have also shown that 85% to 98% of occupational accidents are due to unsafe behavior resulting from workers’ attitude, behavior, and culture [2].

The safety climate in an organization depends on people’s understanding of the safety policies and procedures, as well as the value, importance, and priority of safety in the workplace [2]. In order to study the causes of accidents and improve the safety in an organization, it is necessary to determine the level of safety in that organization. To determine the level of safety in an organization, the most effective way, measuring the safety performance and climate [3]. The safety climate is also dependent on the organization’s safety culture, more precisely, a measurable form of the safety culture [4]. The safety culture is formed by the individual’s or group’s values, attitudes, perceptions, efforts, and behavioral patterns that determine the commitment, success, and effectiveness of the safety management system [4 –7]. The safety climate refers to people’s understanding of policies, guidelines, and safety procedures, as well as the values, importance, and priority of safety in an organization [8]. Therefore, the safety climate can be a benchmark for the assessment of safety programs. The safety climate assessment can be used to identify different areas of safety [9]. Moreover, the assessment of the safety climate can provide a picture of the status of organization’s safety at a specified time [3]. Therefore, in recent years both preventive indicators such as the safety climate along with reactive indicators such as the accident have been used to study the safety in work environments. Combining both preventive and reactive approaches helps to assess the effects of safety plans in organizations. In addition, the safety climate assessment can be used as a guide to develop a safety policy for an organization [10]. However, after 30 years of research on the safety climate, there are still controversies about different dimensions of the relationship between safety climate and safety performance and accidents [8, 9].

In general, limited studies have examined the relationship between the safety culture and climate, their dimensions and relationship with the safety behavior and the occurrence of accidents, and the main variables forming the relationship. The concept of safety climate has been highlighted by many studies as an important determinant of performance in workplaces. The safety climate has several dimensions that must be considered when measuring it. The dimensions of the safety climate include management values (prioritization of safety by the management), secure communications (suitable conditions for the exchange of safety information), safety training (availability, appropriate, and comprehensive training), and safety systems (effectiveness of safety procedures to prevent accidents). Many studies have reported a multi-level approach to the safety climate. The safety climate has been studied at organizational, group, and individual levels [11 –14]. Neal and Griffin provided a means for measuring the safety climate. They proved the multidimensional nature of the safety climate [15]. Furthermore, they concluded that the effect of the safety climate on the safety performance was partially mediated by the safety knowledge and motivation. However, the relationship between the safety climate and the safety performance and accidents is not quite clear. Management commitment, supportive environment, education, personal safety priority, and work pressure as the dimensions of the safety climate, and safety compliance and safety participation, as the dimensions of personal safety priority, are the factors reported in the literature to determine the safety climate and safety culture and assess the safety behavior of employees in workplaces. Moreover, according to the literature, the organizational climate, the safety climate, and the personal safety priority can influence individual safety behavior in the workplaces; on the other hand, the safety knowledge and safety motivation are strongly related to the safety performance. Hence, the aim of this study was to determine the effects of the safety climate variables on the personal safety priority and eventually on the accident. Moreover, using a path analysis model (PAM), this study aimed at providing a model for explaining the effects of safety climate and safety performance on the occurrence of accidents.

The initial model for explaining the effects of the safety climate on accidents was constructed on the basis of several hypotheses. According to a review study by Christian et al., the incidence of work-related accidents is more related to safety performance than to the safety participation and safety compliance [2]. Therefore, the model was constructed based on the following hypotheses:

Effects of the safety climate dimensions on accident are mediated by the safety performance dimensions.

Management commitment affects all other safety climate dimensions.

2 Methods

2.1 Data collection tools and preprocessing

This cross-sectional study was conducted on 294 workers employed in industrial units in Hamedan, a province in the western part of Iran. In this study the required data were collected using a questionnaire. In the first stage, the data on the safety climate of the units were collected using the Nordic safety climate questionnaire (with 25 items). The questionnaire assessed five dimensions of the safety climate including management commitment, supportive environment, training, personal safety priority, and perceived work pressure. Furthermore, 10 items were used to measure the safety performance; of all the items, five were related to the safety compliance and five to the safety participation. It is worth noting that all the items were designed as five-point Likert scale (strongly agree –agree –neutral –disagree –strongly disagree). Accidents were assessed using a questionnaire. In this study, the purpose of the events was the lost work days due to injury. In total, there were 120 lost work days due to injury. Confirmatory factor analysis model (CFAM) was used to assess the construct validity and convergent validity of the safety climate and safety performance questionnaire. CFAM is a statistical technique used to verify the factor structure of a set of observed variables. CFAM allows the researcher to test the hypothesis and investigate the presence of a relationship between the observed variables and their underlying latent constructs. The fitting of the model was evaluated using absolute and relative indices. The reliability of the questionnaire was evaluated through calculating Cronbach’s α.

2.2 Data analysis

Path analysis is a useful tool for assessing direct and indirect effects of some variables on a specific target variable. In the present study, the safety performance and climate was the target variable. The strength of a path is represented by a coefficient conceptually equal to standardized partial regression coefficients. A coefficient ranges from –1 to +1. The higher the coefficient, the greater is the effect of a variable. In order to assess the significance of a path in a path analysis model, the scholars use t value which is the ratio of the unstandardized estimate to standard error. If t > 1.96, the path will be significant at 0.05; if t > 2.56, the path will be significant at 0.01. In addition to each path, the goodness-of-fit of a path analysis model can also be determined using indices available for such evaluations. These indices can be categorized into two main groups: absolute fit indices and comparative fit indices. Absolute fit indices outline how well the hypothesized model fits the data. The model’s χ² value, root mean square error of approximation (RMSEA), goodness-of-fit index (GFI), and root mean square residual (RMR) are some indices categorized in this group [16].

The model’s χ² value is very sensitive to the sample size and normally its value increases as the sample size increases. To fix this problem, Wheaton et al. [17] suggested using the ratio of the χ² value to the df; accordingly, a ratio lower than two is indicative of a satisfactory model fit [16]. RMSEA is another absolute fit index that is popular because of its sensitivity and informative and easy-to-interpret nature. This index is calculated using the model χ² value, df, and the sample size (N) (Equation (1) [18]. A RMSEA value lower than 0.07 indicates a good fit, while values lower than 0.1 represent a moderate fit, and values higher than 0.1 represent unacceptable model fit [16]. $RMSEA = \sqrt{\frac{χ^{2} - df}{df (N - 1)}}$ where RMSEA = root mean square error of approximation; χ² = model χ² value; N = sample size.

In contrast, comparative fit indices explain how close the hypothesized model is to a baseline ideal model. Normed fit index (NFI) and comparative fit index (CFI) are two examples of such fit indices. Moreover, comparative fit indices with values higher than 0.95 indicate that a model is of good fit [16].

In this method, the convergence and divergence of the structure of the variables must first be tested. Convergent validity and divergent validity are ways to assess the construct validity of a measurement procedure. Convergent validity tests that constructs that are expected to be related are, in fact, related. Discriminant validity (or divergent validity) tests that constructs that should have no relationship do, in fact, not have any relationship. In this study, relationship between five dimensions (constructs that are related) of the safety climate including management commitment, supportive environment, training, personal safety priority, and perceived work pressure using convergent and divergent validity tests. It is a measure of the amount of variance that is captured by a construct in relation to the amount of variance due to measurement error and it has often been used to assess discriminant validity (or divergent validity).

3 Results

As shown in Table 1, the fit indices of CFAM was used for the safety climate and safety performance in the present study. As shown in this table, all the values of fit indices of the model were acceptable. Table 2 also presents the convergent and divergent validity of variables construct.

Table 1
Fit indices of the confirmatory factor analysis models for the safety climate and safety performance

Model fit index Measurement model of safety performance Measurement model of safety climate Acceptable level

χ²/df 52.1/31 2.130 <3.00

RMSEA 0.048 0.061 <0.08

CFI 0.989 0.937 <0.9

Model fit index	Measurement model of safety performance	Measurement model of safety climate	Acceptable level
χ²/df	52.1/31	2.130	<3.00
RMSEA	0.048	0.061	<0.08
CFI	0.989	0.937	<0.9

CFI = comparative fit index; RMSEA = root mean square error of approximation.

Table 2

Convergent and divergent validity of constructs

Variable	*AVE	**√AVE	Variable
			Mc	Se	Psp	Tr	Wp	Part	Com
Mc	0.750	0.870	1
Se	0.580	0.760	0.200	1
Psp	0.610	0.780	0.256	0.226	1
Tr	0.710	0.840	0.191	0.300	0.183	1
Wp	0.600	0.770	–0.365	–0.095	–0.243	–0.140	1
Part	0.610	0.780	0.062	0.112	0.217	0.061	–0.149	1
Com	0.680	0.820	0.177	0.278	0.289	–0.171	–0.148	0.274	1

*Average variance extracted, **The root of average variance extracted. (MC: management commitment, Se: supportive environment, Tr: training, Psp: personal safety priority, Wp: perceived work pressure, Com: safety compliance, Part: safety participation, and Accid: accidents.)

Based on the assumptions of the study, we constructed an initial model, but the fit indices of the model were not acceptable. In order to improve the model fit, we inserted some additional paths into the model. Figure 1 shows the paths depicted based on the assumptions of the study. The paths were directed from the five variables of the safety climate including management commitment, supportive environment, training, personal safety priority, and perceived work pressure to the safety participation and compliance and accident, resulting in a model with an acceptable fit. It was not possible to draw an arrow from management commitment or other variables to accident, because in the path analysis model an arch represents a causal effect, and accident normally is affected by other variables such management commitment, safety participation, compliance, etc.

Fig. 1

First model for explaining the impact of safety climate on accident. (MC: management commitment, Se: supportive environment, Tr: training, Psp: personal safety priority, Wp: perceived work pressure, Com: safety compliance, Part: safety participation, and Accid: accidents.)

With increasing significant paths in a model, the df increases and, as a result, the absolute fit indices such as χ²/df and RMSEA will improve. Accordingly, the paths from perceived work pressure and personal safety priority and the safety participation to the safety compliance were added to the model.

Hence, we selected the second model for further discussion. It should be noted that we tried all of the possible models which can be constructed by combining these five variables in different ways. The model presented in Fig. 2 was the best model with the most favorable fit indices. As shown in Fig. 2, in addition to the paths depicted, two additional paths were directed.

Fig. 2

Second model obtained by adding the paths linking work pressure to personal safety priority and the safety participation to the safety compliance. (MC: management commitment, Se: supportive environment, Tr: training, Psp: personal safety priority, Wp: perceived work pressure, Com: safety compliance, Part: safety participation, and Accid: accidents.)

Table 3 presents the fit indices of both models used in the present study. It can be observed that the safety climate (management commitment) had the strongest negative impact on work pressure and the safety compliance toward accident, followed by the safety participation, and training quality. Moreover, the negative influence of the safety climate on accident was mainly mediated by two variables: work pressure and the safety participation toward accident. In other words, negative safety climate can decrease safety participation toward safety and, consequently prevent employees to consider safety rules, regulations, norms, and procedures.

Table 3

Significance of each path in the second model

Path From	To	Unstandardized path coefficient	Standardized path coefficient	S.E.	T	P
Management commitment	Supportive environment	0.213	0.304	0.043	4.919	***
Management commitment	Training quality	0.207	0.268	0.050	4.133	***
Management commitment	Work pressure	–0.319	–0.322	0.061	–5.219	***
Management commitment	Personal safety priority	0.214	0.313	0.042	5.059	***
Training quality	Safety compliance	0.042	0.032	0.073	0.577	0.564
Training quality	Safety participation	–0.026	–0.024	0.061	–0.432	0.666
Supportive environment	Safety participation	0.079	0.066	0.072	1.101	0.271
Supportive environment	Safety compliance	0.274	0.191	0.089	3.087	0.002
Work pressure	Personal safety priority	–0.109	–0.158	0.047	–2.312	0.021
Work pressure	Safety participation	–0.132	–0.156	0.046	–2.866	0.004
Work pressure	Safety compliance	–0.076	–0.075	0.056	–1.371	0.170
Personal safety priority	Safety compliance	0.216	0.147	0.078	2.775	0.006
Personal safety priority	Safety participation	0.170	0.139	0.069	2.465	0.014
Safety participation	Safety compliance	0.246	0.204	0.069	3.536	***
Safety participation	Accident	–0.026	–0.043	0.035	–0.733	0.463
Safety compliance	Accident	–0.102	–0.204	0.029	–3.475	***

Furthermore, Table 4 presents the fit indices of the developed path analysis model.

Table 4

Fit indices of the developed path analysis model

Fit index	Value of the present model	Acceptable level
Chi-square/degree of freedom	2.688	<3.00
RMSEA	0.075	<0.08
GFI	0.978	>0.90

RMSEA: root mean square error of approximation, GFI: goodness-of-fit index.

Using a path analysis model, we are also able to quantify the direct and indirect effects of variables on each other. Tables 5 and 6 present the related data. As presented in Table 5, supportive environment and personal safety priority were the variables with the highest direct effect on the safety compliance. Management commitment had the highest indirect influence on the safety compliance. Moreover, supportive environment and management commitment, respectively, were the most influencing factors in terms of the total effect.

Table 5

Direct, indirect, and total effects of the safety climate variables on the safety performance

Variable	Direct effect		Indirect effect		Total effect
	Safety participation	Safety compliance	Safety participation	Safety compliance	Safety participation	Safety compliance
Management commitment	0.000	0.000	0.115	0.168	0.115	0.168
Work pressure	–0.156	–0.075	–0.022	–0.060	–0.178	–0.135
Training quality	–0.024	0.032	0.000	–0.005	–0.024	0.027
Supportive environment	0.066	0.191	0.000	0.014	0.066	0.205
Personal safety priority	0.139	0.147	0.000	0.028	0.139	0.175

Table 6

Direct, indirect, and total effects of the variables on accidents

Variable	Direct effect	Indirect effect	Total effect
Management commitment	0.000	–0.039	–0.039
Work pressure	0.000	0.035	0.035
Training quality	0.000	–0.005	–0.005
Supportive environment	0.000	–0.045	–0.045
Personal safety priority	0.000	–0.042	–0.042
Participation	–0.043	–0.042	–0.085
compliance	–0.204	0.000	–0.204

The work pressure had the highest indirect and total influence on accidents. However, none of the variables had a direct effect on accidents.

4 Discussion

Today, due to the high rate of accidents in the workplace, researchers are doing many efforts to study the workplace safety. However, research on the climate and culture of safety lacks theory, methodology, and optimal conceptualization of constructs. Employees have always been blamed for safety performance (safety non-compliance safety non-participation) and which has led to an accident [19, 20].

The results of the present study showed that the negative safety climate (high work pressure, non-commitment of management, lack of training) can negatively affect the safety performance and consequently lead to accidents. In this study, the effects of variables of the safety climate and safety performance on accidents were evaluated using path analysis model. The results of the present study showed that work pressure can negatively affect the safety participation and the safety compliance and result in accident. Moreover, it was demonstrated that work pressure had a strong negative effect on personal safety priority and safety compliance, while other dimensions of the safety climate (quality of training, supportive environment, and personal safety priority, excluding management commitment) had a positive effect on safety performance. Management commitment had a strong negative effect on work pressure. Consequently, perceived work pressure had a tremendous indirect negative effect on accidents. However, work pressure did not influence accidents directly, though it was the strongest factor influencing the safety performance. In the present study, it was also observed that the safety performance had a strong negative effect on accidents. The safety participation had a direct strong effect on the safety compliance. In other words, the safety participation increases the level of the safety compliance. Supportive environment and personal safety priority had a direct strong effect on the safety compliance. In addition, management commitment had an indirect strong effect on the safety compliance. Consequently, management commitment had a strong total effect on the safety participation and the safety compliance.

Overall, the model presented in this study provides useful information about how the effect of safety climate on accidents is mediated by different dimensions of the safety performance such as the safety participation and the safety compliance [21]. Therefore, it is necessary, the safety compliance improve to reduce accidents. According to the results of the path analysis model, a supportive environment should be increased to improve safety compliance. Because in the absence of a supportive environment, the safe acts of workers will be ridiculed. Finally, improving in a supportive environment requires training and increased awareness of management and supervisors.

Although path analysis is a strong method for evaluating direct and indirect effects, it has some limitations. Some of the limitations are discussed by Jeon [22]. For example, path analysis can only be used for explanation and not for prediction. For constructing a predictive model, other approaches such as artificial neural networks are recommended [22, 23]. The PAM is used to identify the root causes of accidents, however the prediction model using the artificial intelligence and neural network approach is more accurate compared with other applied methods such as PAM. In the other words, the PAM can be used to predict accidents but its accuracy is less than artificial intelligence. Cross-sectional studies do not provide a precision basis for establishing causality. Hence, the relationship between the variables of interest may be influenced by other variables. Another limitation is the assessing tools that were used. However, the answers of the workers in industrial units may not necessarily be accurate. Moreover, this study was conducted in workers in industrial units, which may not be representative workers in constructions.

5 Conclusion

According to the results, a supportive environment should be improved. Because it improves safety compliance. in another hand, perceived work pressure has an indirect effect on accidents which is mediated by other variables, mainly personal safety priority and safety performance. Training and increased awareness of management and supervisors can improve the supportive environment and reduce work pressure. In general, it can conclude that a supportive environment can improve safety compliance and safety performance. As a result, accidents can be reduced. Further research is required to generalize the results obtained in this study, hence, a replication of the study is suggested.

Conflict of interest

The authors have no conflict of interest to declare.

Funding

This study was financially supported by the Vice President for research at Hamadan University of Medical Sciences (project number: 9403191440).

Footnotes

Acknowledgment

The authors would like to thank the managers of factories and their workers for their cooperation. This study was approved by the Ethical Committee for Research in Hamadan University of Medical Sciences.

References

Jafari

, Eskandari

, Valipour

, Mehrabi

, Charkhand

, Mirghotbi

. Development and validation of a new safety climate scale for petrochemical industries. Work. 2017;58(3):309–317.

Christian

, Bradley

, Wallace

, et al. Workplace safety: a meta-analysis of the roles of person and situation factors. J Appl Psychol. 2009;94(5):1103. doi: 10.1037/a0016172.

Vinodkumar

, Bhasi

. Safety climate factors and its relationship with accidents and personal attributes in the chemical industry. Safety Science. 2009;47(5):659–667.

Choudhry

, Fang

, Mohamed

. The nature of safety culture: A survey of the state-of-the-art. Safety Science. 2007;45(10):993–1012.

Farrington-Darby

, Pickup

, Wilson

. Safety culture in railway maintenance. Safety Science. 2005;43(1):39–60.

Rao

. Safety culture and accident analysis—a socio-management approach based on organizational safety social capital. Journal of Hazardous Materials. 2007;142(3):730–740.

Antonsen

. The relationship between culture and safety on offshore supply vessels. Safety Science. 2009;47(8):118–28.

Mohandes

, Zhang

. Towards the development of a comprehensive hybrid fuzzy-based occupational risk assessment model for construction workers. Safety Science. 2019;115(2):294–309.

Bavafa

, Mahdiyar

, Marsono

. Identifying and assessing the critical factors for effective implementation of safety programs in construction projects. Safety Science. 2018;106(4):47–56.

10.

Kearney

, Balanay

, Mannarino

. Safety behavior and work safety climate among landscaping and groundskeeping workers in North Carolina: A pilot study. Work. 2017;56(1):45–53.

11.

Probst

, Estrada

. Accident under-reporting among employees: Testing the moderating influence of psychological safety climate and supervisor enforcement of safety practices. Accident Analysis and Prevention. 2010;42(5):1438–44.

12.

Choudhry

, Fang

, Lingard

. Measuring Safety Climate of a Construction Company. Journal of Construction Engineering and Management-Asce. 2009;135(9):890–899.

13.

Hope

, Overland

, Brun

, Matthiesen

. Associations between sleep, risk and safety climate: A study of offshore personnel on the Norwegian continental shelf. Safety Science. 2010;48(4):469–77.

14.

, Tsai

. The effects of safety climate on vessel accidents in the container shipping context. Accident; Analysis and Prevention. 2008;40(2):594–601.

15.

Barbaranelli

, Petitta

, Probst

. Does safety climate predict safety performance in Italy and the USA? Cross-cultural validation of a theoretical model of safety climate. Accident Analysis and Prevention. 2015;77:35–44.

16.

Hooper

, Coughlan

, Mullen

. Structural equation modelling: Guidelines for determining model fit. Electron J Bus Res Methods. 2008;6:53–60.

17.

Wheaton

, Muthen

, Alwin

, Summers

. Assessing reliability and stability in panel models. Sociological Methodology. 1977;8(1):84–136. doi: 10.2307/270754.

18.

Kline

, Santor

. Principles & practice of structural equation modelling. Canadian Psychology. 1999;40(4).

19.

Ghasemi

, Kalatpour

, Moghimbeigi

, Mohhamadfam

. A path analysis model for explaining unsafe behavior in workplaces: the effect of perceived work pressure. Int J Occup Saf Ergon. 2018;24(2):303–310.

20.

Natalie

, John

. Safety climate and safety behaviors in the construction industry: The importance of co-worker’s commitment to safety. Work. 2016;54(2)401–413.

21.

Jeon

. The strengths and limitations of the statistical modeling of complex social phenomenon: Focusing on SEM, path analysis, or multiple regression models. Int J Soc Behav Educ Econ Bus Ind Eng. 2015;9(5):1597–1605.

22.

Patel

, Jha

. Neural network model for the prediction of safe work behavior in construction projects. Journal of Construction Engineering and Management. 2014;141(1):401–406.

23.

Aliabadi

, Farhadian

, Darvishi

. Prediction of hearing loss among the noise-exposed workers in a steel factory using artificial intelligence approach. International Archives of Occupational and Environmental Health. 2015;88(6):779–787.