Multi-dimensional model for determining the leading performance indicators of safety management systems

Abstract

BACKGROUND:

Health and safety performance measurements aimed to provide information on the progress and current situation of organizational strategies and activities.

OBJECTIVES:

We developed a model to determine and select safety key performance indicators in order to assess safety management systems.

METHODS:

This study has been designed in six steps aiming at defining a model of leading performance indicators (LPIs) and selecting key performance indicators (KPIs) using the AHP method.

RESULTS:

According to the results analysis, 116 structural and operational indicators were defined based on the components of the OHSAS 18001 management system. For this purpose, 19 structural, 27 operational and 33 active KPIs were selected by AHP and BN techniques.

CONCLUSION:

Development of LPIs is influenced by various organizational, managerial, and operational factors. LPIs extracted from the components of the OHS-MS deployed in an organization are often passive and cannot show the changes in the safety status of a workplace in a short period. The model presented in this study was designed with an emphasis on extraction of active and operational indicators, as they were capable of detecting performance changes in construction industries.

Keywords

Leading indicator safety performance Bayesian network AHP oil industries

1 Introduction

Performance evaluation indicators and criteria are important in all aspects of life [1]. These indicators create a proper reaction to incidents, so that we can take action to change and improve [2]. In the past, industrial safety performance was measured by indicators defined by the Occupational Health and Safety Administration (OSHA), including the Recordable Injury Rate (RIR), Lost Work Day (LWD) rate, Worker Compensation Rate (WCR) and so forth [3]. The LWD, RIR indicators, and disease rates measure the safety performance rate after loss and damage occurrence, which is defined as a lagging indicator [4]. Lagging indicators do not provide information necessary to prevent further accidents. In general, performance indicators are metrics showing the performance of one or more dimensions of an organization [5]. The main purpose of measuring the performance of occupational safety and health is to provide necessary information regarding the progress made and the current situation of the strategies and activities adopted by an organization to control health and safety risks [6]. Safety performance indicators fall into two main categories of leading and lagging indicators. Leading indicators show the activities of an organization regarding the prediction and prevention of accidents before they occur; while lagging indicators show the performance of the organization after an accident occurs to reduce impacts and consequences [7]. Safety research in the field of health and safety performance measurement indicators was started around 1998 [8]. Recent studies focus on the difference between leading and lagging indicators, with the research based on the need for safety or risk measurement. The term “index” is relatively new in the field of safety, but the measurement of safety was performed in 1980 and with expressions such as rate, index, and measurement [9]. Today, the terms “indicator” and “Key Performance Indicators (KPIs) ” are usually used [10]. The nuclear industry is known as a pioneer in the development of risk indicators, followed by the chemical process and petroleum industries [7, 11]. A few studies have been conducted on the development of leading safety performance indicators in the field of oil and gas. The Association of Oil and Gas Producers (OGP), the Center for Chemical Process Safety (CCSP), and the American Petroleum Institute (API) are active organizations in the development of leading safety performance indicators in the oil and gas industry [12 –15]. The reports on the safety performance of these centers are mainly at the managerial level, and focus on managerial elements such as change management, training, and management of risk rising from fatigue and maintenance procedures [16].

Cambon et al. divided the leading indicators into structural and operational ones [17]. Daniel, relying on the components of the ILO-OHS-2001 management system, defined the LPIs [18]. Jimmie Hinze et al. divided safety LPIs into two groups: active and passive [3]. Passive indicators are those that take a long time to measure, while the active ones can be measured within a shorter period [3]. The fatality rate in construction accidents can be 10–20 times more than the average fatality rate in other industries [19]. According to the reports of the safety performance indicators of the International Association of Oil and Gas Producers (OGP), 12.9% of total work days lost are related to construction phases in oil and gas industries [20]. Many studies indicate that the prevalence and incidence of occupational accidents and illnesses in construction activities are higher than those in other industries [21, 22]. Furthermore, it is difficult to use occupational health and safety costs as an indicator, since these costs cannot be easily calculated in the construction industry [23]. The main root causes of construction accidents include the failure to identify the unsafe conditions, work in unsafe conditions and the ignorance of workplace conditions, perhaps due to managerial mistakes, unsafe actions and the nature of construction activities [24]. Haslam et al. showed that the key factors responsible for accidents in the construction industry included human error, unsafe workplace, defective equipment and materials, lack of personal protective equipment and deficiencies in risk management [25]. The results show that construction operations and activities are of a high-risk nature. Therefore, it is essential that LPIs are detected timely along with safety warning characteristics [3, 26]. Hence, in the present study, we tried to develop a model to evaluate performance indicators and safety management systems, so that the determined LPIs could help to decide safety conditions at all levels of management during the construction phase in oil and gas refineries.

2 Methodology

This is a descriptive-analytical study conducted during the construction phase of oil and gas refineries. The methodology involves six steps as follows:

2.1 Step 1: Classification of safety performance indicators

In the first step, the LPIs are classified based on former studies [3 , 18] as shown in Fig. 1.

Fig. 1

The model of classification of LPIs.

According to Fig. 1, there are two LPIs: structural and operational. In fact, structural indicators show the status of safety management systems, including policies, objectives and plans, procedures, and guidelines [17]. In addition, operational performance indicators evaluate the effectiveness of the internal processes of safety and health management systems [18].

2.2 Step 2: Creation of a model to develop LPIs

As Fig. 2 depicts, the model to define safety LPIs was developed to determine and select KPIs. The model was based on structural and operational performance indicators [3, 18]. Two types divided both indicators: active and passive. The operational indicators are determined based on a project’s Work Breakdown Structure (WBS), hazard identification and an analysis of the construction phase accidents. In this model, the Bayesian network (BN) and an expert panel were used to analyze accidents.

Fig. 2

The process of defining and selecting key LPIs.

2.3 Step 3: Selection of an occupational health and safety management system

The current evidence indicates that the application of a systemic approach in different areas can improve efficiency and effectiveness and enhance performance output, and it may integrate all activities. The application of the major components of a management system is a key step in performance measurement to identify the strengths and weaknesses. Different studies demonstrated that among various systems of health and safety management, the OHSAS 18001 management system had the highest priority in different industries [5 , 27]. Here, based on the model, which is defined in the second step, the components of OHSAS 18001:2007 were chosen as the OHS management system guidelines in the construction phase of a refinery.

2.4 Step 4: Identification and classification of construction phase operations and hazards involved

In this step, to determine the active indicators as defined by Hinze [3], the construction operational processes were determined by the WBS plane. In fact, WBS is a network structure or tree graphic to illustrate the method of producing a product or service. It includes parts of hardware and software services, and some organizational tasks to provide some products or services. In the fourth step, the first main activities of the construction phase were determined by studying the project WBS. Afterward, the hazards of the unsafe conditions of the operation were analyzed using the PHA. The preliminary hazard analysis techniques were the first attempt to identify and classify the hazards of the operations or systems [28].

2.5 Step 5: Development of a casual network of accidents by the BN

In the fifth step of the study, the recorded accidents in five refinery construction projects were analyzed by the SPSS version 18 software, and major accidents were selected according to their “occurrence frequency” and “consequence severity”. The casual network of accidents and the assessment of deficiency probability are important factors for risk assessment, determining the safety performance indicator. The 14 experts working for that project determined the relationship between the causes of accident occurrence and their occurrence probability. Moreover, using the BN, the causative relationships and calculation of the probability of the major accidents in this study were accomplished. In brief, the BN method is the appropriate technique to the collect the scattered data, and it is an good framework in the field of subjective probabilities for decision-making under uncertainty [29]. In addition, it is a graphical probability model showing a set of random variables and conditional dependencies among them. The network shows a complete description of the different areas. Each variable of the joint distribution probability is calculated using the data from the network. A variable in the distribution can be set on a variable numbering like P (X₁ = x₁ … x_n = x_n). Regarding the data within the network, the amount of a variable can be calculated according to Equation 1: $P (x_{1} \dots x_{n}) = \prod_{i = 1}^{n} P (x_{i} | parents (X_{i})$ (1)

Figure 3 shows the causal network related to the E event that is a method to determine leading performance indicators using the BN. In addition, according to this figure, C₁ and C₃ are one of the causative factors for the occurrence of the E event. The calculation of the probability of the E event is based on relationships between nodes. Moreover, the probability of the event E in the specified path C₁ and C₃ concerning the chain rule in probability and calculated according to Equation 2.

Fig. 3

The method of extracting safety LPIs (LPI_S).

$P (C_{3} . C_{1} . E) = P (E | C_{1}, C_{3}) \times P (C_{1} | C_{3}) \times P (C_{3})$ (2)

The BN is exactly equal to the joint distribution, when Equation 3 is valid for any X_i variable in the network. $P (X_{i} | X_{i - 1} . \dots . X_{1}) = P (X_{1} | Parents (X_{i}))$ (3)

Assuming that parents (X1) ⊆ {X₁ . … X_i-1} the causal network has two structural and learning characteristics. By applying the network structural model, the safety performance indicators are determined through identification of the preventative measure. The extracted LPI demonstrates the status of the preventative measure in decreasing the probability of the occurrence of causes. The probability of occurrence of any cause is applied as one of the key criteria to select KPIs. In the present study, active safety performance indicators were developed using the concept of the BN.

2.6 Step 6: Selection of KPIs

It is essential to reduce the number of LPIs to some limited KPIs to create an effective method of measuring OHS MS performance. In other words, among the existing indicators, the best and most important indicators should be selected based on the considered criteria. Figure 4 shows the selection model of KPIs. In the scientific literature, different requirements may be found to choose a suitable indicator [30]. However, one of the most famous sets of criteria to select indicators in the field of management performance is SMART (specific, measurable, achievable, relevant, and time-bound) as mentioned previously by Kjellen, Rockwell, and Carlucci [31 –33].

Fig. 4

Model for the selection of key performance indicators.

In this step, KPIs were selected from a set of extracted indicators according to the components of OHSAS 18001 through the AHP method, and all indicators were compared to each other.

Furthermore, this paired comparison was accomplished based on the SMART criteria and experts’ opinions. Thus, a paired comparison table like a questionnaire was prepared. In addition, the comparison was calculated according to the following square matrix: $\tilde{A} = [\begin{matrix} 1 & {\tilde{a}}_{12} & \dots & {\tilde{a}}_{1 n} \\ {\tilde{a}}_{21} & 1 & \dots & {\tilde{a}}_{2 n} \\ ⋮ & ⋱ & ⋮ \\ {\tilde{a}}_{n 1} & {\tilde{a}}_{n 2} & \dots & 1 \end{matrix}] = [\begin{matrix} 1 & {\tilde{a}}_{12} & \dots & {\tilde{a}}_{1 n} \\ \frac{1}{{\tilde{a}}_{12}} & 1 & \dots & {\tilde{a}}_{2 n} \\ ⋮ & ⋮ & ⋱ & ⋮ \\ \frac{1}{{\tilde{a}}_{1 n}} & \frac{1}{{\tilde{a}}_{2 n}} & \dots & 1 \end{matrix}]$ (4)

In this step, the weighting of the indicators was based on the experience and expertise of the 14 selected safety experts who were employed in the construction of oil and gas industries. After obtaining the paired-wise comparison matrix from each expert, the results were merged using the geometric mean method, and the integrated matrix of paired comparisons was calculated by Equation 5.

$\begin{matrix} μ_{g} = {(\prod_{i = 1}^{n} a_{i})}^{1 / n} = \sqrt[n]{a_{1} . a_{2} \dots a_{n}} {\tilde{a}}_{ij} \\ = {({\tilde{a}}_{ij}^{1} \times {\tilde{a}}_{ij}^{2} \times \dots \times {\tilde{a}}_{ij}^{m})}^{\frac{1}{m}} \end{matrix}$ (5)

The weight was calculated by the geometric mean, and it was the final weight of the elements, but it was not the normal weight. When the geometric mean related to the paired comparison of each element was shown as π_i, the normal weight of the elements of each column was calculated as follows: $W_{i} = \frac{π_{i}}{\sum_{i = 1}^{n} π_{i}}$ (6)

Considering a relatively low complexity, software accessibility, and the possibility of applying this method to decision-making problems in economic, scientific, and technological sectors, the AHP is widely used in scientific contexts, as shown by Vaidya and Kumar, and Subramanian and Ramanathan [34, 35]. Therefore, using the AHP, the KPIs were selected based on the SMART criteria and the cause occurrence probability of accidents, which all of them were obtained from the BN. In addition, the GeNIe software and Expert Choice 11 software were used to generate the causal network of accidents and paired comparison of the LPIs, simultaneously.

3 Results

Table 1
Determination of the priorities of AHP in two-by-two comparison

Priorities Nominal value

Most priority 9

More important 7

Important 5

Somewhat important 3

Same 1

Priorities	Nominal value
Most priority	9
More important	7
Important	5
Somewhat important	3
Same	1

As per this methodology, 116 LPI_S were determined based on the OHSAS 18001:2007 components. Among them, 46 KPIs were selected according to the SMART criteria. Furthermore, Table 2 presents 19 structural performance indicators and 27 operational performance indicators.

Table 2

The final list of leading KPIs determined on the basis of the components of the OSH MS

Components	KPIs
1. –Policy
A. OHS Policy	Structural	•OHS policy approved and communicated by senior management
	Operational	•The percentage of employees who have a clear knowledge of the OHS policy
B. Personnel Participation	Structural	•Procedure to improve employee participation in safety are available
	Operational	•The number of consultations done by the managers and workers in the OSH field
		•The number of proposed measures for improving the OSH by the workers
2. – Planning
C. Planning for Hazard	Structural	•Procedures to identify, assess and control risks exist
Identification, Risk		•Identification and Risk Assessment Committee is formed
Assessment and Risk Control	Operational	•The percentage of risk assessment processes confirmed in terms of the accuracy and validity of the applied controlling measures
		•The percentage of operations was risk assessment had been done (in relation to the plan)
D. Legal and Other Requirements	Structural	•There are laws and requirements related to safety
	Operational	•The percentage of identified safety rules and requirements associated with the relevant operations
E. Objectives	Structural	•OHS goals of the Strategic Committee
	Operational	•The percentage of goals which have been totally realized
F. OH &S Management Programs	Structural	•There is an OHS action plan
	Operational	•The percentage of the budget spent on OHS improvement programs
3. –Implementation and Operation
G. Structure and Responsibility	Structural	•Percentage of workstations that the OHS scope of duties and responsibilities defined for it.
	Operational	•The percentage of workers who have been trained with regard to the duties and responsibilities of OHS
H. Training, Awareness, and Competence	Structural	•There are Training Needs Assessment documents
	Operational	•The percentage of workers who have passed basic safety training
		•The person hour rate of safety training
I. Consultation and Communication	Structural	•Procedures and processes of communication and consultation are available
	Operational	•The number of sessions held in the presence of managers about safety issues
		•The number of safety tours held in the presence of managers
J. Documentation	Structural	•The Documented procedure, and physical space archives available
	Operational	•The number of new guidelines developed in new working stations
K. Document and Data Control	Structural	•There is document procedure control
	Operational	•The percentage of guidelines that have been reviewed and upgraded periodically
L. Operational Control	Structural	•The ratio of the number of safety supervisors to the total number of personnel
	Operational	•The number of safety inspections conducted in relation to the program
M. Emergency Preparedness and Response	Structural	•The number of emergency response plans developed
	Operational	•The percentage of maneuvers held with regard to the program
		•The person-hour rate of training of personnel in ERP
4. –Checking and Corrective Action
N. Performance Measurement and Monitoring	Structural	•There is a health and safety management system performance monitoring procedure
	Operational	•The percentage of leading and lagging indicators, which are periodically reviewed and upgraded
		•The number of leading and lagging indicators developed for effective monitoring of safety improvement programs
O. Accident, Incident, Non-conformance, and Corrective and Preventive Action	Structural	•The process of implementing corrective
	Operational	•The percentage of controlling and preventive measures taken in relation to the defined measures obtained from risk assessment
		•The percentage of corrected anomalies
P. Records and Record Management	Structural	•There is a procedure to record
	Operational	•The number of accidents, diseases, and events analyzed in relation to the total recorded accidents
		•The number of databases related to the registration of the available data
Q. Audit	Structural	•Internal audit procedures
	Operational	•The number of person-days of the periodical internal audit performed
		•The percentage of nonconformity revealed by the results the of internal audit
5. –Management Review	Structural	•Identification of instructions and management review committee members.
	Operational	•The number of components and processes of OHS MS analyzed in OHS reviews

In addition, 2,744 accidents were statistically analyzed in five refinery construction projects and the results of them are illustrated in Fig. 5.

Fig. 5

The proportion of the frequency of recorded accidents and the mortality caused by the accidents.

Thus, Fig. 5 indicates that the accidents of falling (21%), objects collapse (18%), and falling objects (17%) account for the highest proportion of the recorded accidents. Furthermore, the accidents of falling (7), falling objects (3), and electrocution (2) have the highest frequency of recorded fatalities. According to the results of this stage and the conducted PHA, the accidents of fall from a height, falling objects, and electrocution were selected, and the causal network of the accidents related to these three accidents was developed.

Furthermore, active performance indicators were determined according to the results described in the previous paragraph.

As Table 3 shows, 32 active KPIs of construction and installation of the refinery were selected based on the SMART criteria and occurrence probability of the causes of accidents with the BN.

Table 3

The final list of determined KPIs based on construction operations

Operations	KPIs (major and proposed)
Civil operations	•The percentage of drivers having a special certificate
•Reinforcement	•The percentage of vehicles having technical inspection certificates
•Formatting	•The percentage of excavations that have a danger bar and hard barrier them
•Backfill	•The percentage of excavations with a safe access way toward the inside
•Concreting	•The percentage of drivers who have passed the defensive driving training course
•Soil transportation	•The percentage of mixers with guards
Installation of equipment and facilities	•The ratio of lifting with a lifting plan
	•There ratio of qualified riggers to the number of cranes
	•The number of trained riggers
	•The percentage of trained crane operators
	•The percentage of operators with special certificates
	•The percentage of cranes with technical certificates
	•The percentage of installation operations with work permits
Scaffolding	•The percentage of scaffolding operators who have received the necessary training
	•The percentage of operators with special cards for scaffolding
	•The percentage of safe scaffolds
	•The percentage of scaffolds with tag
Electrical	•The percentage of generators with grounding
	•The percentage of distribution panels with residual-current circuit breaker and grounding
	•The percentage of generators with the fire extinguisher
	•The percentage of generators with installed safety instructions
Insulation and color	•The ratio of MSDS to the existent chemicals
	•The percentage of safe grinding stone cutters, the percentage of safety support pipes
Fit up of pipes	•The percentage of cylinders with trolley during use
	•The percentage of cylinders with caps
	•The percentage of cylinders with technical inspection certificates
	•The percentage of cutting and welding blowpipes with flash-back arrestors
	•The percentage of fit-up and welding stations with safe support and stopper
General	•The ratio of the number of fire trucks to the required number, according to the HSE plan
	•The ratio of the number of ambulances to the required number, according to the HSE plan
	•The ratio of the number of safety personnel to the total required number, according to the confirmed chart
	•The ratio of the number of fire extinguishers to the standard number, according to the HSE plan of the project

4 Discussion

The aim of the present research was to provide a model to determine the safety LPIs in a safety management system in the construction of oil and gas refinery projects. For this purpose, the model was developed by studying previous researches and then prioritizing the leading safety performance indicators in construction projects. For this purpose, 116 safety performance indicators, both structural and operati-onal, were developed based on the OHSAS 18001:2007 management system. The main reason for using the OHSAS 18001 management system components was their use in the majority of companies active in the construction of oil and gas refineries. In previous studies like the one conducted by Podgorski, leading safety performance indicators were developed based on the ILO-OHS-2001 management system components, and KPIs were developed for use in all industries that only considered operational LPIs [18]. One of the main purposes of developing indicators to measure safety performance in construction projects was to rank and compare the safety performance of the project contractors. Thus, it is necessary to consider all aspects of the functional safety management system for an accurate assessment of the situation. However, the performance indicators presented by Podgorski do not cover all aspects, but the model presented in this project divides the indicators into two types: active and passive, which is according to the definition given by Hinze et al. [3]. In the first part, passive indicators in the structural and operational fields are defined. To develop active performance indicators, the construction phase activities of refineries and major accidents occurring in such projects should have been recognized. According to the statistical analysis of the events that occurred in five projects (Fig. 4), falling from a height was the most serious form of accident in terms of frequency and severity, being in consistent with the results of previous studies (such as [36, 37]).

Considering the statistics of incidents in construction industries and the nature of rapid changes in operational conditions in this field, LPIs should recognize rapid changes in the safety of activities. In this study, active indicators were developed based on of Hinze’s definitions as a part of LPIs (Table 3). Hinze et al. showed that active performance indicators were one of the operational performance indicators that are measurable within a short period. Abdelhamid TS et al. stated that unidentified unsafe conditions were one of the major causes of accidents in construction projects; therefore, active performance indicators should enable the measurement of safety condition in the short term. To overcome the limitations of active performance indicators defined in the study by Hinze, the indicators in the present study were designed to detect unsafe conditions. The strength of this study is that the BN was used to determine the direction of the causal pathway. The BN’s excellence in risk assessment and determination of the causal pathways lies in the fact that the variables can be considered together; thereby making it easier to determine relationships between the different levels and causes of accidents. Additionally, the calculated accuracy rate was higher for the probability of events. Correctly determining the accident causal network is the most important factor to identify the active safety performance indicators. The use of a proper method of accident analysis can improve the validity of performance indicators. Therefore, according to the learning capability and modeling of the structure of the causative relationships in the BN, the path of determining active LPIs is more authentic. In addition, studies by Khakzad et al. emphasize the validity and effectiveness of the BN compared to other methods [38, 39].

In the present investigation, the BN was used with two approaches. The first approach includes the extraction of performance indicators from the accident casual network using the structural and graphic capabilities of this method. Furthermore, the approach concerns the selection of active KPIs based on the rate of the calculated occurrence probability in different causal pathways of the main accident as one of the main criteria of indicator selection in the AHP. To select the active KPIs in addition to the SMART criteria, the indicator is highly important as a defined controlled action to prevent accidents. Herra defined active safety performance indicators as preventive control measurements [40]. According to this definition and as Fig. 3 shows, the active performance indicators extracted from accident pathways have varying importance based on the probability of the occurrence of accidents. Indicators derived from causal pathways with high probability of occurrence, according to the criteria and the AHP method, are more important and have greater weight. So, in addition to the SMART criteria, the probability of occurrence of the accident is considered to select active KPIs.

Cambon et al. described the difference between structural and operational indicators of the safety management system. Hinze et al. determined the characteristics of active and passive performance indicators in the construction industry, but the performance indicators were not comprehensively identified. In the study by Podgorski, the selected indicators were developed within the framework of the ILO-OHS-2001 system, but the active performance indicators to measure the safety level in high-risk operations were not specified. In this study, performance indicators were determined based on the challenges of previous studies. One of the main limitations of this study and former ones is the non-implementation of the validation of LPIs. It is recommended that in future studies, providing a proper model for validation of the performance indicators be considered.

5 Conclusion

The results show that determination of LPIs should be based on the type of operation and the identified hazards of the workplace in addition to the OHSMS components. LPIs extracted from the components of the safety management system deployed in an organization are often passive and cannot show the changes in the safety status of a workplace in a short period. One of the benefits of the present study was to solve the above-mentioned problem. The model presented in this study was designed with an emphasis on extraction of active and operational indicators, as they are capable of detecting performance changes in construction industries. Considering the high-risk nature of construction operations and rapid changes in safety status, it is necessary to select LPIs. One of the limitations of using this model in construction projects is the large number and variety of operations and associated accidents. In this study, weights of KPIs selection criteria and OHS-MS components were considered equal. It is recommended that to prioritize and select more accurately the KPIs, be weighted of these selection criteria and OHS components in the future study. It is also recommended that this model be validated in the construction industry and other manufacturing and service industries.

Conflict of interest

None to report.

Footnotes

Acknowledgments

The authors would like to thank all panel experts who cooperated in this study.

References

Sinelnikov

, Inouye

, Kerper

. Using leading indicators to measure occupational health and safety performance. Safety Science. 2015;72:240–8.

Franceschini

, Galetto

, Maisano

. Management by measurement: Designing key indicators and performance measurement systems: Springer Science & Business Media; 2007.

Hinze

, Thurman

, Wehle

. Leading indicators of construction safety performance. Safety Science. 2013;51(1):23–8.

Toellner

. Improving safety & health performance: identifying & measuring leading indicators. Professional Safety. 2001;46(9):42.

Falahati

, Karimi

, Mohammadfam

, Mazloumi

, Khanteymoori

, Yaseri

. Development of safety and health leading performance indicators in the phase of construction of a gas refinery plant using Bayesian network and AHP. International Journal Of Advanced Biotechnology And Research. 2017;8(2):1440–53.

Redinger

, Levine

. Development and evaluation of the Michigan Occupational Health and Safety Management System Assessment Instrument: a universal OHSMS performance measurement tool. American Industrial Hygiene Association Journal. 1998;59(8):572–81.

Øien

, Utne

, Herrera

. Building safety indicators: Part 1–theoretical foundation. Safety Science. 2011;49(2):148–61.

Øien

, Utne

, Herrera

. Building Safety indicators: Part 1 - Theoretical foundation. Safety Science. 2011;49(2):148–61.

Tarrants

. The measurement of safety performance: Uni-versity of Michigan-Dearborn; 1980.

10.

Kjellen

. Prevention of Accidents Through Experience Feedback: Taylor & Francis; 2002.

11.

Saqib

, Siddiqi

. Determination of weighting factors for safety performance indicators in NPPs. Safety science. 2016;84:245–9.

12.

Swuste

, Theunissen

, Schmitz

, Reniers

, Blokland

. Process safety indicators, a review of literature. Journal of Loss Prevention in the Process Industries. 2016;40:162–73.

13.

Wilkinson

. Progress on Process Safety Indicators–Necessary but Not Sufficient? : Discuss. Pap., US Chem. Saf. Hazard Investig. Board, Noetic Risk Sol., Washington, DC; 2012.

14.

Landucci

, Tugnoli

, Cozzani

. Inherent safety key performance indicators for hydrogen storage systems. Journal of Hazardous Materials. 2008;159(2):554–66.

15.

Frank

. Process safety culture in the CCPS risk based process safety model. Process Safety Progress. 2007;26(3):203–8.

16.

Gholami

, Nassiri

, Yarahmadi

, Hamidi

, Mirkazemi

. Assessment of Health Safety and Environment Management System function in contracting companies of one of the petro-chemistry industries in Iran, a case study. Safety Science. 2015;77:42–7.

17.

Cambon

, Guarnieri

, Groeneweg

. Towards a new tool for measuring Safety Management Systems performance. Learning from Diversity: Model-Based Evaluation of Opportunities for Process (Re)-Design and Increasing Company Resilience. 2006:53.

18.

Podgórski

. Measuring operational performance of OSH management system–A demonstration of AHP-based selection of leading key performance indicators. Safety Science. 2015;73:146–66.

19.

Hamid

ARA

, Majid

MZA

, Singh

. Causes of accidents at construction sites. Malaysian Journal of Civil Engineering. 2008;20(2):242–59.

20.

falahati

, karimi

, zokaei

, dehghani

. Development and ranking of safety performance indicators using Bayesian network and Analysis Hierarchical Process Case study: Work at height of the Oil and Gas refinery construction phase. Iran Occupational Health. 2018;15(3):172–85.

21.

Schwatka

, Rosecrance

. Safety climate and safety behaviors in the construction industry: The importance of co-workers commitment to safety. Work. 2016;54(2):401–13.

22.

Rodrigues

, Coutinho

, Cardoso

. Correlation of causal factors that influence construction safety performance: A model. Work. 2015;51(4):721–30.

23.

López-Alonso

, Ibarrondo-Dávila

, Rubio

. Safety cost management in construction companies: A proposal classification. Work. 2016;54(3):617–30.

24.

Abdelhamid

, Everett

. Identifying root causes of construction accidents. Journal of Construction Engineering and Management. 2000;126(1):52–60.

25.

Haslam

, Hide

, Gibb

, Gyi

, Pavitt

, Atkinson

, et al. Contributing factors in construction accidents. Applied Ergonomics. 2005;36(4):401–15.

26.

Versteeg

, Bigelow

, Dale

, Chaurasia

. Utilizing construction safety leading and lagging indicators to measure project safety performance: A case study. Safety Science. 2019;120:411–21.

27.

Jafari

, Eskandari

, Valipour

, Mehrabi

, Charkhand

, Mirghotbi

. Development and validation of a new safety climate scale for petrochemical industries. Work. 2017;58(3):309–17.

28.

Roland

, Moriarty

. Preliminary hazard analysis. System Safety Engineering and Management, Second Edition. 2009:206–12.

29.

Korb

, Nicholson

. Bayesian artificial intelligence: CRC press; 2010.

30.

Hale

. Why safety performance indicators? Safety Science. 2009;47(4):479–80.

31.

Kjellén

. The safety measurement problem revisited. Safety Science. 2009;47(4):486–9.

32.

Rockwell

. Safety performance measurement. Journal of Industrial Engineering. 1959;10(01).

33.

Carlucci

. Evaluating and selecting key performance indicators: an ANP-based model. Measuring Business Excellence. 2010;14(2):66–76.

34.

Vaidya

, Kumar

. Analytic hierarchy process: An overview of applications. European Journal of Operational Research. 2006;169(1):1–29.

35.

Taylan

, Zytoon

, Morfeq

, Al-Hmouz

, Herrera-Viedma

. Workplace assessment by fuzzy decision tree and TOPSIS methodologies to manage the occupational safety and health performance. Journal of Intelligent & Fuzzy Systems. 2017;33(2):1209–24.

36.

Leu

S-S

, Chang

C-M

. Bayesian-network-based safety risk assessment for steel construction projects. Accident Analysis & Prevention. 2013;54:122–33.

37.

Chen

T-T

, Leu

S-S

. Fall risk assessment of cantilever bridge projects using Bayesian network. Safety Science. 2014;70:161–71.

38.

Khakzad

, Khan

, Amyotte

. Safety analysis in process facilities: Comparison of fault tree and Bayesian network approaches. Reliability Engineering & System Safety. 2011;96(8):925–32.

39.