Assessing safety culture in offshore environments - Industrial Centre

112 S.J. Cox, A.J.T. Cheyne / Safety Science 34 (2000) 111±129In the wake of such changes, a Cross Industry Safety Leadership Forum has beenestablished to facilitate the sharing of safe working practices. The members of thisforum recently highlighted a number of indicators of improved safety performance(Cross Industry Safety Leadership Forum, 1997), including an overall decrease inreported injury frequency rates and an estimated $5 billion in investments in healthand safety-related improvements. However, despite these overall e€orts, there issome evidence to suggest that the rate of improvement, as measured by accident/injury frequency, has slowed down over the past 2 years (HSE, 1997; Cox andCheyne, 1999). Furthermore the Cross Industry Safety Leadership Forum (1997)have also con®rmed that much of the existing e€orts in support of these improvementshave been focused upon technology and management systems rather thanhuman factors. They also suggest that potential for future improvements may bestbe realised through enhanced e€orts in the areas of human factors and through theassociated developments in health and safety culture. In light of this initiative thispaper considers how o€shore installation managers can gauge their safety culture.Discussion here focuses on the development of a generic safety culture assessmentmethodology which was designed to support improvements in safety performanceacross the sector.This paper describes a series of studies that are focused on the practical needs ofo€shore managers and workers for the monitoring of safety culture through thedevelopment of a self-assessment tool. It aims to give a broad overview of thedevelopment of this methodology (Cox and Cheyne, 1999) and draws heavily uponrecent work carried out in the UK sector of the North Sea. It also explores thepractical utility of the concept in o€shore environments. The methods describedhere are concerned with both the development and piloting of this tool. Whereas themethodology is based on sound safety science, the overall aim of the assessmentprocess is to chart a path through, what is increasingly becoming, a conceptualmine®eld (Cox and Flin, 1998), and one in which researchers have been challengedby industry to address practical issues (Cox and Lacey, 1998). Although much hasbeen done in other sectors (e.g. Hale and Hovden, 1998; Hofmann and Stetzer,1998) in these studies the cultural context for investigation is set within the o€shoreindustry.1.1. Previous studiesA number of previous studies in o€shore environments have, either directly orindirectly, considered employee perceptions and elements of health and safety culture.In an early study, Marek et al. (1985) examined risk perceptions, stress andaccidents among di€erent work groups on an o€shore installation. They identi®edthe main in¯uences on safety as the implementation of a safety policy, `correct'safety management and safety promoting activities, and incorporation of employees'views in designing safety programmes. Similarly Rundmo (1993) implemented aquestionnaire survey of personnel on eight o€shore installations in the Norwegiansector of the North Sea; ®rstly to determine the personnel evaluations of riskresources and, secondly, to identify di€erences in risk perceptions among di€erent

S.J. Cox, A.J.T. Cheyne / Safety Science 34 (2000) 111±129 115data collection varies. Multiple and independent measures, if they provide supportiveevidence, can thus provide a more certain picture (or pro®le) of the issue underreview (Denzin, 1978; Cox and Cheyne, 1999).In the same vein James and Jones (1974) describe three di€erent approaches to theassessment of organisational culture (and climate) and, in doing so, o€er commentaryon its di€erent (conceptual) loci. First, is the `multiple measurement-organisationalattribute approach' which regards organisational climate exclusively as a setof organisational attributes (or main e€ects), measurable by a variety of methods;e.g. organisational structure or organisational systems measured by propriety auditsystems. Second, there is the `perceptual-organisational attribute approach', whichviews organisational climate as a set of perceptual variables which are still seen asorganisational e€ects, e.g. views of the organisation's commitment, etc. Finally,there is the `perceptual measurement-individual attribute approach' which capturesorganisational climate through perceptions of individual attributes, e.g. individuals'feelings and attitudes towards organisational issues, their related behaviour, etc.Given the importance of both organisational and individual factors in in¯uencingsafety culture (James and Jones, 1974; Denison, 1996) it is suggested that in anattempt to measure culture these di€erent approaches should be treated as complementaryand not as alternatives. Thus a multiple perspective was taken whichalso combined di€erent approaches to assessment, as recommended by Jick (1979).This is illustrated in the framework model shown in Fig. 1.In such a model, the representation of organisational safety culture is consistentwith one sponsor's preferred (HSE, 1997) de®nition. For example, the AdvisoryFig. 1. Multiple perspective model of safety culture assessment.

employed here for two reasons: (1) to elicit constructs based on individuals' notionsof `safety culture'; and (2) to provide initial indications of any di€erences in overallperceptions between di€erent work groups in the study organisations. The focusgroup studies were deemed to be particularly appropriate at this stage of the investigationssince the data collected would be utilised as the basis for the toolkit developmentand, in particular, the development of the questionnaire tool. Focus groupsas a method are well accepted (Krueger, 1994) but they can be biased by a numberof factors, e.g. domination of the group by one member or too much direction bythe facilitator. These biases were controlled for by the use of a small number ofgroup facilitators who followed a structured discussion format.2.1. MethodA series of focus group discussions (n=40) were conducted using both o€shoreand onshore personnel in three separate organisations, at a total of six di€erentlocations. These 40 discussion groups involved 375 employees in groups rangingbetween three and 12 individuals. One hundred and eighty-one participants workedin onshore locations and 194 were based on o€shore installations. Eighty-threemanagers and supervisors took part in the study. Wherever possible the groups werehomogenous and comprised either: (1) managers and supervisors; or (2) members ofthe workforce. This promoted an atmosphere of open enquiry and, wherever possible,avoided inhibiting the group members (Carroll, 1998).Each of the discussions involved four questions being put to the group participants,namely what they understood by the term safety, what they understood bythe term company culture, how safety ®ts into their picture of company culture andwhat they understood by the term safety culture. The ®rst three questions weredesigned to put participants at their ease and help focus their discussion of theconcept of safety culture. All members of the group were encouraged by the facilitatorto make a response to the ®nal question.Each of the focus group discussions were recorded on audiotape and then transcribedverbatim. These transcriptions were then subjected to a preliminary contentanalysis (Holsti, 1969; Dane, 1990) involving two independent raters, and theresulting keywords (based on the most commonly occurring constructs in the discussionsof the concept of safety culture) were tabulated. There was 96% agreementbetween raters on the classi®cation of concepts. Where the raters did not agree onthe categorisation of a concept (in 13 cases), a third rater arbitrated. In addition tothe extraction of individual key constructs, general group feelings and perceptions ofsafety were recorded as appropriate.2.2. ResultsS.J. Cox, A.J.T. Cheyne / Safety Science 34 (2000) 111±129 117All groups actively participated with a good degree of co-operation. The numberof occasions each concept was mentioned during the focus discussion groups isshown in Table 1, together with a breakdown by organisation. The numbers shownin brackets indicate the number of times each issue was raised in a negative sense.

118 S.J. Cox, A.J.T. Cheyne / Safety Science 34 (2000) 111±129Table 1Focus group keywordsKeyword Organisation 1 Organisation 2 Organisation 3 TotalInvolvement in safety issues 6(3) 67(20) ± 73Physical work environment 2(2) 52(34) 2(1) 56Priority of safety over other issues 11(10) 44(37) ± 55Management actions for safety 3(2) 20(15) 27(21) 50Management commitment 9(3) 21(6) 7(2) 37Apportioning of blame 5(4) 29(20) ± 34Safe systems/procedures ± ± 23(17) 23Accidents ± ± 11(10) 11Attitudes to personal safety ± ± 10(2) 10Safety training and instruction 2(1) 5(1) ± 7Safety legislation ± ± 6(4) 6Total number of discussion groups 7 19 15 40For example, in the case of involvement, the number of time individuals felt thatthey were not involved in safety issues.Many of the non-managerial participants, especially within Organisation 2,recounted experiences of not feeling that they were adding value to safety-relatedprocesses and perceived that they were not involved in any safety decision-makingprocesses a€ecting anything other than routine day-to-day operations. Responsesfrom o€shore workers suggest that they have a greater appreciation of risk thanonshore workers which, in turn, a€ects their perception of safety issues and may gosome way to explaining the many emotive comments made about the importance ofsafety o€shore. The responses from the onshore groups represented a more sanitisedviewpoint, given with much less emotion than those received from the o€shoregroups. Several groups expressed the view that, although attitudes to safety hadimproved, they perceived that the principles of risk management were providing themain impetus for improvements in all three organisations. Combined responsessuggest that management were quite committed to safety although improvementscould still be made. Although some non-managerial participants, particularly ino€shore environments, related feelings that a blame culture existed.2.3. DiscussionOf the constructs identi®ed, management commitment, management actions forsafety and the physical work environment were the more common across all threeorganisations, although involvement in safety issues and the priority given to safetyissues were also highlighted as important. These results give some indication of howsafety culture is perceived by those working in the o€shore environment and areconsistent with earlier ®ndings both on and o€shore (Zohar, 1980; Flin et al., 1996).The constructs uncovered provided the basis, together with the review of currentliterature on safety culture assessment (Cox and Flin, 1998) and climate identi®cation,for the construction of a set of climate assessment tools. It should be noted

120 S.J. Cox, A.J.T. Cheyne / Safety Science 34 (2000) 111±129safety attitude measures have also been developed and utilised in other industrialsectors (e.g. Zohar, 1980; Brown and Holmes, 1986; Cox, 1988; Dedobbeleer andBe land, 1991; Cooper and Philips, 1994; Mearns et al., 1998). An initial questionbank, derived from the initial in-house studies, was re®ned in the light of otherpublished studies. Consideration was given to developments within: (1) the nuclearindustry (Lee, 1995); (2) the energy supply industry (see Donald, 1995, for details);(3) a variety of manufacturing settings (Cox and Cox, 1991; Cheyne and Cox, 1994;Cox et al., 1998); and (4) the HSE attitudinal indicator of safety climate (Byrom andCorbridge, 1997). Extra dimensions, not included in the common threads from previousstudies in participating organisations, were included if common to two ormore of the external studies and if highlighted by the focus group discussions (Study1). The full set of dimensions and their origin is shown in Table 2.In summary, the initial question bank for the questionnaire instrument was,therefore, based on:1. an initial review of o€shore surveys within the study organisations and theestablishment of common themes;2. comparisons between these and instruments in other industrial sectors (in particularthe nuclear and chemical/process industries); and3. a review of the constructs identi®ed in the focus discussion groups (Study 1 above).These reviews and comparisons allowed a pilot questionnaire to be developed toinclude 47 items covering the areas of Management Commitment, Communication,Priority of Safety, Safety Rules and Procedures, Supportive Environment, Involvement,Personal Priorities and Need for Safety, Personal Appreciation of Risk andWork Environment. This attitude questionnaire was piloted using a population ofo€shore personnel in two locations. These personnel were asked not only to completethe questionnaire but also to comment on the general content and any speci®citems they felt to be unclear. Sixty completed questionnaires were returned. Themain objective of this pilot study was to test the face validity of the items in thequestionnaire with an appropriate group. Comments were restricted to individualitems and resulted in two items being reworded and four items being deleted. Thenew, 43-item questionnaire was tested on a larger pilot population.3.2. AnalysisThe questionnaire was distributed to 350 employees on three o€shore installations.Two-hundred and twenty-one questionnaires were returned, representing a63% response rate. There was no evidence that this sample was unrepresentative ofthe total population. These data were subjected to a series of statistical tests includingcon®rmatory factor analysis (CFA), internal-scale consistency and alternateforms reliability tests.3.2.1. Factor analysisThe data were subjected to CFA using version 5.1 of the EQS (structural equation)program (Bentler and Wu, 1995). In CFA the researcher postulates a model (a

S.J. Cox, A.J.T. Cheyne / Safety Science 34 (2000) 111±129 123Table 5Alternate-forms reliability coecientsItemInter-item correlationManagement acts decisively when a safety concern is raised 0.622*Managers and supervisors express concern if safety procedures are not adhered to 0.353*Management considers safety to be equally as important as production 0.583*There is good communication here about safety issues which a€ect me 0.036Safety information is always brought to my attention by my line0.089manager/supervisorSome health and safety rules and procedures are not really practical 0.589*Employees are not encouraged to raise safety concerns 0.561*When people ignore safety procedures here, I feel it is none of my business 0.361*I can in¯uence health and safety performance here 0.555*I am never involved in the ongoing review of safety 0.292Safety is the number one priority in my mind when completing a job 0.564*I understand the safety rules for my job 1*It is important that there is a continuing emphasis on safety 1*I am rarely worried about being injured on the job 0.664*I cannot always get the equipment I need to do the job safely 0.361*Operational targets often con¯ict with safety measures 0.363*Sometimes I am not given enough time to get the job done safely 0.728**Signi®cance at the 0.05 level.the factors in Table 3 are presented in Table 4. It can be seen that all the alphasreported are in the range 0.53 through 0.84, the lower ones being consistent withthose factors where some original factor loadings were low.Alternate-forms reliability involves comparing two di€erent versions of the samemeasure (Dane, 1990). In the case of the questionnaire, 30 subjects were asked tocomplete the survey and then later, after an interval of at least 18 h, were asked 17 ofthose items orally as part of an interview format. These items were chosen randomlyfrom the larger questionnaire. The inter-item correlations are shown in Table 5.Only three of the inter-item correlations are non-signi®cant, indicating overall goodalternate-forms reliability. The low inter-item correlation in these cases may havebeen due to the relatively small sample size involved and the e€ect of a few outlyingscores on such a sample.3.3. DiscussionThis study has dealt with one of the tools for use in a safety climate assessment, aswell as detailing its development, the instrument's reliability and consistency hasalso been discussed. This tool has been based, wherever possible on the conceptshighlighted by the focus group discussions (Study 1) and a review of relevant literaturepertaining to the measurement and assessment of safety culture and climate.The organisations involved also requested that any questionnaire tool developed beconsistent with what had been used in the past. The ®nal section of this paper brie¯ydescribes the use of the toolkit and discusses some of the issues raised, the pro®ling

124 S.J. Cox, A.J.T. Cheyne / Safety Science 34 (2000) 111±129of the results of the assessment process, and subsequent actions that may be consideredby those assessing safety climate. The con®dential nature of the assessmentprocess makes it dicult to report this study in great detail.4. Illustration of the toolkit in practiceThe developed ``Safety Climate Assessment Toolkit'' process was initially testedin one organisation involved in hydrocarbon production. The pilot organisation isinvolved in the extraction of hydrocarbons, operating on various sites world-wide.This organisation had a strong commitment to safety, health and environmentalexcellence and also has a tradition of `quality'. The initial management characterisationof their culture for safety was one of `total commitment and safety excellence'.The organisation hoped to appreciate a broader perspective of sharedattitudes and communicate these goals through the use of the assessment toolkit, aswell as examining di€erences between occupational groups. The toolkit procedureswere implemented by an in-house team, who only made use of back-up from theresearch team in analysing the di€erences between occupational groups.The full assessment toolkit (Cox and Cheyne, 1999) consists of three sets of tools,as described earlier: climate questionnaire, semi-structured interview schedule and aset of behavioural indicators. It also provides guidance on their application andinterpretation. The organisation using the toolkit in this case considered the followingsources of evidence:Fig. 2. Radar plot for pilot organisation.

S.J. Cox, A.J.T. Cheyne / Safety Science 34 (2000) 111±129 125Table 6Means for the three work groups involved in the study aDependent variable Managers Production DrillingCommunication 6.4 6.24 5.56Supportive environment 6.64 6.02 5.92Involvement 5.75 4.75 4.45Personal priorities 9.2 8.86 7.89Personal appreciation of risk 8.75 7.75 7.71Work environment 9.1 7.95 6.62a Emboldened groups di€er signi®cantly from the others (Tukey tests, P

126 S.J. Cox, A.J.T. Cheyne / Safety Science 34 (2000) 111±129ences between managers, production teams and drilling teams. A series of one-wayanalyses of variance were performed for each factor in the attitude questionnaire.Di€erences in mean scores (standardised to a 10-point scale for each factor) for eachof the three groups are shown in Table 6.Di€erences were found in six of the questionnaire dimensions with drilling teamsshowing signi®cantly lower evaluations than managers in terms of Supportiveenvironment, Involvement, Personal appreciation of risk and Work environmentand lower evaluations of Communication and Personal priorities than both managersand production teams. Production teams, however, only di€ered signi®cantlyfrom managers in terms of Involvement, Personal appreciation of risk and workenvironment. These ®nding are consistent with those of Mearns et al. (1998) suggestingthat drilling teams, who often belong to another `parent organisation' mayhave their own sub-culture and may, consequently, ®nd themselves outside variouscommunication channels.4.2. DiscussionUsers found the process straightforward and found little diculty in constructingtheir safety climate pro®le. Users then examined the initial questionnaire items andinterview responses in order to get some notion of the exact nature of any problemshighlighted by poor scores. This investigation revealed that a large number ofemployees felt that they were not involved, in, or informed of, safety initiatives,indeed it was perceived that a small group of `select' people were the only individualsinvolved in any safety activities, and these individuals gave little, or no, support toothers on safety issues. These feelings were also re¯ected with regard to safetytraining. Once again, the majority of those questioned felt that they were notinformed of relevant training or encouraged to attend courses, other than thoselegally required.Several initiatives are planned as a result of these ®ndings. Firstly, the users havedecided to actively promote, with the help of the training manager, the range ofsafety training courses on o€er. This involves listings of appropriate coursesappearing in the company newsletter, as well as announcements at all team meetingsin an attempt to reach the various `sub-cultures'. Team meetings are also to be used,in the ®rst instance, to encourage individuals to get involved in, and indeed to propose,new safety initiatives. The user will reassess the safety climate pro®le whenthese measures have been in place for at least 1 year.5. DiscussionThis paper has described the development of an innovative approach to assessingculture, and speci®cally safety culture and climate. The development of the ``SafetyClimate Assessment Toolkit'' has exploited the notions of multiple organisationalperspectives and data triangulation (Jick, 1979). It utilises approaches which can becombined to provide an in-depth picture of an organisation's current safety climate.

S.J. Cox, A.J.T. Cheyne / Safety Science 34 (2000) 111±129 129Fitzpatrick, D., 1996. Employee Attitudes O€shore. Unpublished M.Sc. dissertation, LoughboroughUniversity, UK.Flin, R.H., Mearns, K.J., 1994. Risk Perception and Safety in the O€shore Oil Industry. Second InternationalConference on Health, Safety, and the Environment in Oil and Gas Exploration, Jakarta.Flin, R., Mearns, K., Fleming, M., Gordon, R., 1996. Risk Perception and Safety in the O€shore Oil andGas Industry, (O€shore Technology Report OTH 94 454). HSE Books, Sudbury.Hale, A.R., Hovden, J., 1998. Management and culture: the third age for safety. A review of approachesto organizational aspects of health, safety and environment. In: Feyer, A.-M., Williamson, A. (Eds.),Occupational Injury: Risk, Prevention and Intervention. Taylor and Francis, London, pp. 129±165.Hofmann, D.A., Stetzer, A., 1998. The role of safety climate and communication in accident interpretation:implications for learning from negative events. Academy of Management Journal 41, 644±657.Holsti, O.R., 1969. Content Analysis for the Social Sciences and Humanities. Addison-Wesley, Reading,MA.HSE, 1997. Health and Safety Climate Tool Process Guidelines. Health and Safety Exectutive Books,Sudbury.James, L.R., Jones, A.P., 1974. Organisational climate: a review of theory and research. PsychologicalBulletin 81 (12), 1096±1112.Jick, T.D., 1979. Mixing qualitative and quantitative methods: triangulation in action. AdministrativeScience Quarterly 24, 602±611.Krueger, R.A., 1994. Focus Groups, 2nd Edition. Sage, Thousand Oaks.Lee, T.R., 1995. The role of attitudes in the safety culture and how to change them. Conference on`Understanding Risk Perception', O€shore Management Centre, Robert Gordon University, 2 February,Aberdeen.Litwin, M.S., 1995. How to Measure Survey Reliability and Validity. Sage, Thousand Oaks.Marek, J., Tangenes, B., Hellesouy, O., 1985. Experience of Risk and Safety. Work Environment StatfjordField. Oslo University.Marsh, H.W., Balla, J.R., Hau, K., 1996. An evaluation of incremental ®t indices: a clari®cation ofmathematical and empirical properties. In: Marcoulides, G.A., Schumacker, R.E. (Eds.), AdvancedStructural Equation Modeling: Issues and Techniques. LEA, Mahwah, NJ, pp. 315±354.Mearns, K., Flin, R., Gordon, R., Flemming, M., 1998. Measuring safety climate on o€shore installations.Work and Stress 12, 238±254.Rundmo, T., 1993. Risk perception and occupational accidents on o€shore petroleum platforms. Paperpresented at the 2nd O€shore Installation Management Conference, The Robert Gordon University,Aberdeen, April.Schein, E.H., 1985. Organisational Culture and Leadership. Jossey-Bass, San Francisco, CA.Zohar, D., 1980. Safety climate in industrial organisations: theoretical and applied implications. Journalof Applied Psychology 65 (1), 96±102.