Managers safety attitudes as organizational factors and pressure equipment risk predictor

2.1 Organizational factors as safety culture and process safety descriptor

Organizational factors are often used as safety culture descriptor, so, it is necessary to analyse the concept of safety culture [16–19], which came up all of a sudden, as a universal remedy. Other concepts stem from this one like safety climate and safety attitude [19, 20]. This concept is still difficult to define, despite being dealt with by many researchers. Safety culture is not an isolated structure, but derives from and is embraced by organizational culture. Safety climate is considered visible and measureable while culture is the essential content of human thought and action [21–23]. Just as better design and better material quality contributed to better process safety a decade ago, the focus today has shifted to human behavior and the impact of management. The critical role of wise leadership and good management nowadays has become very clear, because it is evident that if the top of the company does not take safety seriously and encourages employees to take care of safety, all workplace safety measures will not be effective [3]. Therefore, effective leadership is required at all organizational levels so that the climate of appropriate safety culture is created. Safety is something that must be managed, and represents the responsibility of line management. Namely, management is only a means by which joint human efforts ensure the achievement of desired goal. There is an extensive literature [24–28] regarding this topic, where survey methods are developed (questionnaires for safety climate assessment) with objective interpretation of their results, and they proved to be very useful, partly due to their verification by statistical methods. On the other hand, safety culture questionnaires describe employees’ attitudes, beliefs and attitudes about risk and safety and are closely related to personal safety, which is a significantly different concept from process safety, but can be considered one of many areas in the process safety.. For example, HSE tool [29] includes safety climate as one of 10 themes, however not among the areas marked as basic areas (defined as areas of crucial importance in all factories and at all places), but among the areas which are second in importance. Also, there is a general lack of connection of safety climate and safety culture either with the safety concept and risk management or with safety performances [25]. Accordingly, the questionnaires for safety climate assessment, despite being adeqately analyzed throughout the literature, validated statistically and checked, embrace only a small portion of a broader picture about organizational factors. Anyhow, studying these questionnaires may be useful in terms of defining part of questions to describe impact factor related to personal safety, however the coverage of organizational factors is not enviable.

2.2 Organizational factors as risk based inspection descriptor

Risk-based API inspection methodology contains organizational/managerial factors. It is presented in three parts [30]: a) Inspection Planning by implementing API RBI Technology, b) Defining the probability of failure by implementing API RBI Assessment, and c) Modelling consequences in API RBI. Probability of failure P _f (t) is defined as a product of generic failure frequency, gff, a damage factor, D _f (t) and a management system factor, F _MS [30]. Generic failure frequency is specified based on known data of the failure history. A damage factor is defined based on appropriate damage mechanisms (local and general corrosion, cracks, creep, etc.), which are associated with the materials that components are manufactured from, as well as with the nature of the process, physical condition of the components and testing techniques before which damages are quantified. A damage factor modifies generic failure frequency in industry and makes it closely related to a concrete component analyzed. A management system factor considers the impact of the management system on the equipment mechanical integrity, which means that certain, higher level organizational factors are well recognized. This factor also considers the probability of timely detection of carrying capacity loss due to damage accumulation and is directly proportional to the quality of the mechanical integrity program within the plant. As emphasized in API standard [30], the effectiveness of management system by process safety within a company may have a major impact on the mechanical integrity. The API RBI procedure includes management system assessment that most directly influence the probability of the component failure, whichconsists of a series of questions about the plant management, operations, inspections, trainings, etc. Each possible answer to each question has specific weight, depending on the adequacy of the answer and the significance of the question itself. Such system produces a quantitative result of management system assessment, which is easy to reproduce. Further, it is explained that the number of questions and the scope of the area they cover allow management system assessment to make the difference between PSM systems of various effectiveness level but without any generalized result that can be used as a basis to define agreement vs disagreement. A 1000-value is equivalent to the achievement of excellent results regarding the problems related to PSM that affect mechanical integrity. However, questions for Management System assessment within API procedure are not statistically processed. Also, there is no explanation for the manner of defining the questions. Quantification of the questions themselves and obtaining the overall factors in the end is not explained either. Further, as pointed out in the API document [30], in RBI the purpose of management system assessment is not to measure the overall agreement with API standard recommendations or with OHSA recommendations, but to stress the problems that are present in the mechanical integrity, the largest risk prone area, and those questions are related to areas closely connected to mechanical integrity. This leads to a slight negligence of very important areas that are affected by management system.

On the basis of American standard API 581, RIMAP project was initiated in Europe in 2001 [31–33]. The project resulted in 4 workbooks which were produced for specific industries: petrochemical industry, chemical industry, steel industry and power plants. The major RIMAP result was to make increased awareness of risk-based inspection, whose main role is to present the framework as the basis for assessment of more detailed risk approaches, i.e. as a general ‘good European practice’ [31, 35]. Risk-based inspection that RIMAP project is grounded on and defined as a method that uses risk for a basis to determine inspection priorities in the process plants components - tanks, pressure vessels, reactors, exchangers, pipelines... [31].

Seveso-Directive is a directive that deals exclusively with prevention and control of major industrial risks involving hazardous substances [36, 37]. The first Directive was enacted in 1982 (82/501/EC), followed by significant enhancements in 1996 (96/82/EC) and 2012 (2012/18/EU) [38]. These Directives require submission of a satisfactory report on safety prior to issuing a license, where operators show that ‘everything that is needed has been done to prevent severe accidents’.

However, both RIMAP project and Seveso directives are very similar to API 581 standard –focusing mainly on technical causes of accidents, costly and complex, and do not have developed any statistically valid tools for assessing the impact of human and organizational factors.

The OECD has issued guidelines through a senior management guide in the high-risk industry [39], whose goal is to establish a balance between risk and profit by drawing attention to the most responsible persons in high-risk industry [40]. In a booklet, recommendations are given for simple measures that should be reviewed by each director or president of the company, whose work involves severe risks, so they can check themselves in the safety field. The adoption of these guidelines, according to the guide authors, and their implementation in industry could demonstrate great devotion to the highest standards of process safety, which could result in a long-term sustainable development. As for the area of application, it is emphasized that despite the guidelines are intended for chemical and petrochemical industry, they will be useful for any industry or organization that is in touch with hazardous processes and substances due to the nature of its activity, which leads to the risk for a large number of humans or the environment. As regards the purpose of the guide, it clearly defines that it is intended for senior managers such as directors, presidents, members of boards (executive and non-executive) and other senior staff within the organization, that is, who have the opportunity to influence the direction in which the company safety culture will develop. It is considered that the guidelines are also useful for shareholders entering high-risk industries, as well as for regulators and stakeholders [14]. The guide consists of 5 areas, i.e. themes [39]: actions, risk awareness, information, competences, leadership and culture, the number of questions ranging from minimum 5 to maximum 8. For possible answers to each question, 3 options are offered: 1 = Yes, I can easily demonstrate this, 2 = I am not sure, I would have to find out, 3 = No, I do not think there is an omission here. Like in the case of the questionnaire from API standard these questions are not statistically processed.

2.3 Organizational factors as risk predictors of pressure equipment operation

Integration of risk assessment and organizational factors impact is scarcely present in the available literature, in general, although above mentioned references emphasize its high importance in accidents prevention. To this day, there is no universally accepted methodology for the assessment of organizational factors impact on risk in pressure equipment exploitation. All available techniques and methods, either complex or simple, more or less effective, have not been incorporated in the procedures for risk analysis, or associated with industrial accidents of pressure equipment with severe consequences. Considering all mentioned, there is an evident need to develop statistically valid tool for organizational factors assessment as a risk predictor of pressure equipment operation. So, further efforts aimed to prove hypothesis that it is possible to establish organizational factors by using statistical multivariate methods, as well as links between their dimensions, are necessary. This research aims at providing a valid and reliable measurement instrument to study the impact of these factors, which could be of great benefit for industries where pressure equipment operates.

3.1 Research operationalization and framework

So, in this paper, the research methodology comprises:

Creating a preliminary list outline of organizational factors’ constructs and their observed variables that are pressure equipment risk predictors, using current standards, available literature research, history of major accidents involving pressure equipment, conversations and consultations with experts in the related area.

All survey questions used were drawn from existing surveys [3, 41–44]. After the preliminary questionnaire was created, 30 experts in the area were interviewed in aim to prove its’ content validity, and thereafter the final questionnaire was formed to be distributed later on.

3.2 Survey construction

The first part of the questionnaire involves general (demographic) data such as industry type and company size and age and experience of the person which participates in survey, while the questions in the second part describe the following organizational factors as constructs - communication, potentially hazardous materials and equipment, safety and health at work, organizational changes, subcontractors from other companies, maintenance/inspection, human error, trainings and competences of employees for crisis situations and conducting research after accidents and their descriptors/observational variables which lean on a 5-level Likert scale (see Appendix). In the second part, in order that each question/observational variable describes given phenomenon viewed from all aspects and involves in the best possible way managers’ attitudes to a given topic, i.e. theme, two columns are defined, the first related to the current condition in the company and the second related to the significance of the item studied for potential pressure equipment failure. Responses from both columns were afterwards multiplied, because they can be perceived as the probability and consequence predictor, yielding a cumulative response from minimum value 1 to maximum value 25.

3.3 Sample and data collection

Questionnaires were distributed to the companies on the entire territory of Serbia, which cover electric power sector (thermal power plants, thermal plants –heating, and hydro power plants); oil, coal and natural gas sector; pharmaceutical industry, food industry and others, i.e. sectors where pressure equipment is in exploitattion. In total 321 questionnaire was distributed and they were provided with electronic surveys, by e-mail, printed surveys by mail, or taken in person, according to contacted persons’ choice. Participation was voluntary and anonymous. Lastly, 287 questionnaires were collected, of which 253 were valid (correctly filled in – exclusion criteria was an error in a control question). High response rate (79%) could be explained by five reminders during 3 months and different options of survey distribution to contacted persons which included even personal distribution on site. Two waves of responses have arrived during data collection period, according to inferential statistics results, with higher values of variables in the first wave.

Data on demographic facts on the sample surveyed are given as in Table 1:

After the described procedure of data collecting, it was necessary to check the sampling adequacy and the questionnaire reliability and validity [45]. Statistical Package SPSS (IBM Corp., SPSS Statistics Version 20, Armonk, NY, USA) was used for the needs of all analyses, together with additional module AMOS (AMOS 21). The questionnaire structure established by EFA analysis, after the reliability analysis, was further tested by confirmatory factor analysis, CFA, as one of the tools employed for structural equation modeling (SEM) [46]. The measurement model is given in a CFA diagram, represented by the links between certain measured values (items) and their related factors, including links between those factors. The ‘paths’ from latent factors to measured loadings are based on measurement theory. The only loadings that theoretically link measured items to their latent factors are calculated, whereas it is assumed for others that their value is zero [47]. The proposed model adequacy was defined by calculating the fit index, which is used to test the acceptance or rejection of the proposed model. There are a large number of such indices but none of them alone is sufficient to accurately assess fit into the CFA or SEM model. Recommendations for researchers are to apply multiple indices from different categories (cross validation) [48]. These indices give an idea on the measure of mutual correlation of the variables relative to the measure predicted by the model [49]. There are over 25 different model adequacy measures, i.e. fit indices, in the literature available, of which only a portion are commonly considered [50]. There are disagreements even on boundary values of these indices [49]. Therefore, all measured values should be interpreted within the context of a concrete sample size, assessment procedure, model complexity, essential assumptions on multivariate normality, etc. [42].

One of the most commonly applied fit indices is chi square χ², which is used to examine the approximation of the sample covariant matrix fit to the fitted covariant matrix [49]. A negligibly small value of χ² indicates that observed data does not differ significantly from assumed model. Since the formula for calculating the value of χ² is directly linked to the sample size, nearly all models are rated as wrong, with increasing sample size. For that reason, the ratio of χ² to the degrees of freedom (df) is often applied as an alternative fit index. If this value is about 2, the model is considered to have optimum fit [49]. Another commonly employed fit index is Root Mean Square Error of Approximation (RMSEA). The RMSEA is better than other fit indices, since RMSEA measures the lack of fit by the degree of freedom. The RMSEA values of 0.05 indicate good fit of the model, the values below 0.08 indicate the acceptable approximation error, while the values above 0.10 point to the model that does not fit well [42]. NFI index compares fit of two different models (assumed and zero model) with the same data set. Some authors [49, 50] recommend the adoption of NFI values of 0.90 and larger, as indicators of good fit. NNFI represents another fit index that considers model complexity as well, comparing the assumed model with the zero model. Although the interpretation of NNFI index is somewhat more complex, due to its range going beyond the 0 –1.00 limits, it is recommended to apply this index combined with at least three other types of indices [49].

The model adequacy measurement within this research was performed by applying the following fit indices: Model (χ²), χ² divided by value df (χ²/df), Comparative Fit Index (CFI), Root Mean Square Error of Approximation (RMSEA), Normed Fit Index (NFI), as well as Non-Normed Fit Index (NNFI), also referred to as Tucker-Lewis index (TLI).