PROTOCOL: Safety Interventions for the Prevention of Accidents at Work

1.1 DESCRIPTION OF THE CONDITION

Accidents at work are estimated to kill more than 300,000 workers worldwide every year (Concha-Barrientos, Nelson, Fingerhut, Driscoll, & Leigh, 2005). In the European Union, the number of fatalities amounts to nearly 5,000 cases annually, and accidents at work cause many more cases of disability each year (Eurostat, 2004b). Aside from the human cost, workplace accidents also represent a significant economic burden to society (Eurostat, 2004a). Although the risks of accidents at work have been reduced over the last 20-30 years, the number of accidents remains unacceptably high and continues to receive much attention from a wide spectrum of policy and decision-makers.

In this review, an ‘accident at work’ is understood as an accident which causes physical harm (injury) to people at work; that injury may be minor, major or fatal. We use the term ‘accident’ for the causal event(s) leading to the harmful exposure of an individual, whereas we reserve the term ‘injury’ for physical harm as the consequence(s) of such an event. In the health and safety field it is important to distinguish ‘accidental injuries’ from chronic injuries related to long term exposures, as the aetiologies are different. We found the European Commission definition of a work accident the most exhaustive and clear (European Commission 1999). In this review an accident at work is thus defined as “a discrete, sudden and unexpected occurrence in the course of work which leads to physical harm (injury)”. This includes cases of acute poisoning and wilful acts of other persons, but excludes deliberate self-inflicted injuries and accidents on the way to and from work (commuting accidents). The phrase “in the course of work” is taken to mean whilst engaged in an occupational activity, or during the time spent at work and includes road traffic accidents occurring in the course of work.

However, we exclude accidents where the resulting injury is mental harm alone. Occupational diseases, such as occupational dermatitis and repetitive strain injuries, are not included in this review, because they have longer exposure periods, are not discrete, sudden and unexpected, and thus fall outside the definition of an accident. However, harmful exposures, such as acute poisoning and chemical burns are included, as the exposure period is short, and the event is usually discrete, sudden and unexpected, and thus accidental. It should be emphasized, that in this review the term ‘accidental’ does not mean unpredictable or unpreventable, even the specific event can be unforeseen by the victim. Most accidental injuries and their precipitating events are predictable and preventable (Davis & Pless, 2001).

The safety science literature has identified a number of measures influencing risk, safety behaviour and accidents at work (Haddon, 1968; Guldenmund, 2010b; Zohar, 2010; Kjellén, 2000; Tuncel, Lotlikar, Salem, & Daraiseh, 2006; Heinrich, 1931; Rivara & Thompson, 2000b; Spangenberg, 2010). The development in the research field has improved the predictive power of theories and conceptual models over the last four decades, and these have informed authorities, companies and employees in developing more reliable and efficient measures for the prevention of accidents at work (Hale, 2006).

In the last 10-15 years, the multidimensional characteristics of risk to workers, and not least the understanding of how to prevent accidents at work, have been widely emphasized in the safety science literature (Reason, 1997; Kjellén, 2000; Shannon, Robson, & Sale, 2001; Robson, Shannon, Goldenhar, & Hale, 2001; Guastello, 1993; Lund & Aarø, 2004). This development is referred to as the “third age of safety” (Hale & Hovden, 1998). Where accidents previously were seen from a technical, legal or human factors perspective, in recent years cultural and organisational factors have become important additional perspectives included in accident prevention programs at work (Spangenberg, 2010; Mearns, Whitaker, & Flin, 2003; Parker, Lawrie, & Hudson, 2006; Grote, 2007), and so represent a key perspective to understand the complex and multifaceted approaches to reduce harm to workers (Rasmussen, 1997; Spangenberg, Mikkelsen, Kines, Dyreborg, & Baarts, 2002). Six categories of approach can be identified:

Multifaceted approaches: safety interventions for the prevention of accidents at work are characterised as a complex process which usually integrates a number of components. Research has emphasised the importance of integrating these various components in order to achieve a high level of safety at work (Guastello, 1993; DeJoy, 2005).

1.2 DESCRIPTION OF THE INTERVENTION

In this review we focus on primary safety interventions, defined as any attempt deliberately applied to promote safety and decrease the frequency or severity of accidental injuries at work (Robson et al., 2001), that affect risks and safety at work and which can have consequences for the occurrence of accidents causing injury to people at work. Such accidents may subsequently have consequences in terms of disability, work absence and other personal and economic costs.

As safety interventions are rarely based on one component alone, it is expected that some of the safety interventions for the studies included in this review will consist of more than one type of intervention component, such as safety training, a safety campaign, goal setting, safety feedback or machine safeguarding. It is recognised in the field that there is a certain lag phase (latency period) in the implementation of safety interventions, and that this is dependent on the type of safety intervention and the context (Robson et al. 2007). For example the introduction of a new provision or safety legislation will have a longer lag phase compared to the introduction of a new safeguard on a machine. This will be taken into consideration when effects of interventions are evaluated. A safety intervention may consist of one or more components, and can run for a shorter or longer period of time, or involve a permanent change, such as new regulations or orders.

A safety intervention can be initiated at work, e.g., by the employer or the employees, or initiated externally by public authorities, social partners or other stakeholders, but the intervention attempts to improve safety and prevent accidents at work. Interventions aimed at the prevention of accidents at work can operate at different levels, namely at the micro-, meso- or macro-level, i.e., individual, group or organisational, or the broader societal-industry level, respectively (Landeweerd, Urlings, De Jong, Nijhuis, & Bouter, 1990; Hofmann, Jacobs, & Landy, 1995; Lund and Aarø, 2004; Spangenberg, 2010; Haslam et al., 2005; Dyreborg, 2011).

This review adopts an inclusive approach to the meaning of a safety intervention, and recognises the various theoretical approaches and distinctions made in the scientific literature and in practice in this field. It is not expected that studies evaluating safety interventions in general are explicit about how the intervention might work, but on the basis of the information we will seek to classify the types of interventions. Lund and Aarø (2004) have suggested that safety interventions for prevention of accidents at work might work by applying three main types of measures: attitude modification, behaviour modifications and structural modifications. Additionally, a fourth type, safety climate modifications, is suggested by a large body of research (Christian et al., 2009; Spangenberg, 2010; Nahrgang et al., 2007; Zohar & Luria, 2003; Zohar, 2002). In practice, these types of measures can involve a number of different components to form integrated safety interventions.

In this review we define a safety intervention component as an independently operating entity in the intervention, which can stand-alone. This means, for example, that the instruction to the use of a specific item of fall protection equipment is not a component in itself, but is part of the component ‘introduction of fall protection equipment’, since the introduction to the use of the equipment would be meaningless without this equipment. However, in the case where a general introduction or campaign about fall risks at a workplace is given to raise awareness among workers besides the introduction of fall protection equipment, it will be classified as two separate components, as both components can be stand-alone interventions.

Below is listed the main types of components of safety interventions directed at the individual or the group/organisational level, respectively.

Main types of components directed at the individual level

Attitude modification: This may be achieved by means of information and persuasive messages in campaigns, leaflets, booklets, films, posters, direct mail, guidelines, or by teaching or various counselling approaches.

If other types of safety interventions are found in the included studies, they will be added to the classification scheme (see detailed classification in section 12.3).

The review excludes secondary and tertiary interventions, such as on-site injury treatment, rehabilitation and return to work programs. Public safety campaigns directed at the general population and community-based safety interventions are also excluded from the review, as they are not primarily implemented at workplaces.

1.3 HOW THE INTERVENTION MIGHT WORK

Even though the working environment, the nature of work, and the workforce vary from one industry to another, we assume that the different types of safety interventions work in similar ways across various settings, although the effect may be modified by contextual factors, such as whether the industry is static or dynamic (Sadayappan & Moaued, 2011; Cooper, 2007). Work settings that experience transient workforces, such as in construction work, can be a barrier to the implementation of safety interventions as steps taken are easely lost when staff disappear and new ones come in. Also peer pressures resulting from the social dynamics within a group could be weaker in dynamic settings with unstable workgroups, thus making safety interventions more difficult to maintain, for example, in improvement to the safety climate in construction work (Dedobbeleer & Béland, 1991; Kines et al., 2010).

In the following we describe how the four main types of safety interventions might work with references to the most relevant theory in the field.

Attitude modification: The underlying assumption of this type of measures is that attitudes can be modified by the provision of information and knowledge to the relevant persons, and that this in turn can influence behaviour. The theoretical or conceptual support of such approaches is the KAP (Knowledge-Attitudes-Practices) model (Lund and Aarø, 2004). According to this model, safety-related behaviour is determined by an individual's knowledge and attitudes. Within this perspective safety information, e.g., provided by pamphlets, safety campaigns, or safety courses, can change peoples' behaviours, through providing them with the right information or knowledge about the hazards at the workplace, and the consequences these can have on their health.

In particular, social psychology research has provided theoretical knowledge on the relationship between attitudes and human behaviour (Hofmann & Tetrick 2003). This attitude behaviour relationship has provided the basis for a number of practical approaches for the prevention of accidents and injuries at work, where knowledge and information related to risk and safety at work has been disseminated or taught to workers with the intention to modify their attitudes and promote safer behaviour (Burke et al. 2006).

To our knowledge Wicker's (1969) review is the first to review research on the relationship between attitudes and behaviour. He concluded that attitudes probably do not predict behaviour. Since this review, social psychology researchers have sought to develop models that increase the predictive power of attitudes. The main approach within the psychology research area has been to build more complex models of the relationship between attitudes and behaviour, by inclusion of a number of additional components assumed to affect behaviour (Armitage & Conner 2001). The two most known models are the Theories of Reasoned Action (TRA) (Ajzen & Fishbein 1980; Fishbein & Ajzen 1975) and Theory of Planned Behaviour (TPB) (Ajzen 2012). The assumption is that behavioural intention is the best predictor of behaviour. Behavioural intention is regarded as a product, not only of attitudes, but also of social influences (including social norms) and self-efficacy (or perceived behavioural control).

Behaviour modifications are based on an environmental approach. Where the attitude modification approaches mainly explains behaviour in terms of internal mental states and cognitive processes (knowledge-attitudes-behaviour), behaviour modification approaches represent an external focus that explains behaviour in terms of environmental consequences (Luthans & kreitner 1985).

The theoretical framework for the behaviour modification programs is based on behaviourism, specifically operant (learned) conditioning that can be traced back to B.F. Skinner (1969), who contributed with the distinction between learned and unlearned behaviour. Skinner suggested that humans choose various responses to receive a particular consequence. This contingency is framed as the Antecedent-Behaviour-Consequence (A-B-C) model, where both antecedents and consequences are responsible for affecting the behaviour of an individual. Where antecedents serve to define or signal the behaviour, then the consequences of behaviour serve in influencing and reinforcing the behaviour more directly and also encourage the occurrence of behaviour (Krause et al. 1999). Skinners theoretical framework has been expanded by inclusion of the mediating role of cognition, and the term organisational behaviour modification has been suggested for this approach (Luthans & kreitner 1985). Another important expansion of the Skinnerian approach is the social learning theory (Bandura 1971). These theoretical frameworks have informed safety intervention research and practice. Some commonly used components in behaviour-based safety interventions are safety training, goal-setting and feedback, observation and feedback, verbal feedback, data analysis, and problem solving (McAfee & Winn 1989, Krause 1999). A number of behaviour based safety studies have been conducted since the 1970's (Tuncel et al. 2006), which includes a number of various models and components (Cooper 2007; Laitinen & Ruohomäki 1996).

Structural modifications are approaches comprised of a number of models and components related to changes in the physical, organisational or regulatory environment. A common feature of the structural approaches is that environmental factors are changed, often over longer time and more profound. One type of structural modification is the engineering control, e.g., introduction of machine safeguards, walkways, elimination of hazardous substances or materials, and other changes in the physical environment, that directly influences individuals' safety without necessarily affecting their behaviour. Another type of structural change is social control, e.g., changes in laws and regulations that provide coercive power or incentives for people or organisations to change behaviour.

Engineering control approaches are based on the hierarchy of controls (Lingard & Holmes 2001) stating that the most effective preventive measures are taken by eliminating the risk at the source. The most common models used are based on a simple linear model, where a chain of multiple events culminate in an injury. The safety prevention is then directed at removing one or more elements involved in this chain of events, to prevent the occurrence of the injury. One such model is Heinrichs ‘Domino theory’ (Heinrich, 1931), which has had a tremendous effect on practical safety interventions, and still is much in use. A further development of these models are the complex linear models, such as ‘Haddons Matrix’ (Haddon 1968) and the Swiss cheese model (Reason 1997).

Another type of structural modification is social control, which is related to legislation, enforcement and legal sanctions of rules and regulations to achieve behaviour on a non-voluntary basis, or marketing, which is a voluntary exchange, e.g., insurance related benefits for low risk companies. Regulation may serve as a potentially powerful institutional force to promote the adoption of occupational health and safety policies and practices (Chambers et al. 2013). The basic idea is that such instruments provide an incentive for companies or people at work to stick to certain (legal) standards, either because of the risk of penalties in case of noncompliance, or because a benefit can be achieved in exchange for an appropriate behaviour (Rothschild 2000).

Safety climate and culture modifications: Safety culture has long been a subject of interest for safety science, and particularly so following the Chernobyl nuclear meltdown in 1986. However, the theoretical framework for safety culture is generally underdeveloped and the link to research on organizational culture has been weak or even non-existent (Choudhry et al. 2007; Guldenmund 2000). There is, for instance, no widely accepted definition of an organization's safety culture or any consensus on how to change a safety culture to improve safety. Therefore, the concept of safety culture is vague and not easily translated into safety prevention efforts, which may explain why there is a noticeable lack of (safety) culture change intervention studies in the safety literature (DeJoy 2005; Hale et al. 2010; Nielsen 2014). The most elaborated theory of safety culture is based on Edgar Schein's Theory of organizational culture (Schein 2004), where the essence of culture is its core of basic assumptions that manifest as values and in turn defines behavioral norms, e.g., norms that influence safety behavior. The basic assumptions and values are taken for granted and maintained by members of a group, and will be taught to new members as the correct way of thinking and feeling in relation to specific aspects, such as safety. Following this, safety culture might be defined as those aspects of the organizational culture which will impact on attitudes and behavior related to increasing or decreasing safety or risk (Guldenmund, 2010a).

A related concept of culture is climate, which describes the shared perceptions of organizational policies, practices, and procedures, both formal and informal (Reichers & Schneider 1990). Safety climate is supported by a wide range of consistent theoretical frameworks, relating to organizational sense making processes (Weick, 1993; Zohar & Luria, 2004; Weick, 1995), social interactions (Morgeson & Hofmann, 1999), social exchange and climate theory (Christian, Bradley, Wallace, & Burke, 2009; Nahrgang, Morgeson, & Hofmann, 2007; Mearns, Hope, Ford, & Tetrick, 2010; Zohar, 2003; Zohar & Luria, 2004).

Safety climate has been a widely researched approach to accident prevention since the term was introduced by Zohar (1980). Although there is no general consensus on a definition of safety climate, and the literature has been plagued by conceptual ambiguity (Zohar 2010), meta-analyses have identified some common ground and reveal that safety climate is a robust predictor of safety performance (Nahrgang et al. 2007). The assumption underlying this approach is that the safety climate of an organization/group, which can be understood as a socially constructed phenomenon as it emerges as a group-level property through shared cognitions and social consensus, informs workers on how they are expected to act under different circumstances. Thus, the safety climate reflects the shared priority of safety in an organization/group compared to other competing goals such as productivity or quality. It has been suggested, that safety climate can be understood as a surface manifestation (espoused values) of the deeper cultural levels (Guldenmund 2000; Guldenmund 2010b; Schein 2004). Furthermore, the culture and climate approaches have brought focus on the role of leaders and leadership in creating general organizational change and the prevention of accidents at work. These approaches are often centered on the commitment to and priority of safety demonstrated by supervisors and top management (Hofmann & Tetrick 2003; Zohar 2002).

1.4 WHY IT IS IMPORTANT TO DO THIS REVIEW

Accidents at work are prevalent in society, but there are several gaps in our knowledge on the efficiency of various safety prevention components designed to be used by policy makers and practitioners to prevent accidents at work.

Despite earlier attempts to summarize the evidence, the effectiveness of strategies for preventing accidents at work remains unclear. One limitation is that previous reviews have restricted their focus to one type of injury, for example, eye injuries, or one type of prevention measure (Tuncel et al., 2006; Robson et al., 2005; Robson et al., 2007; Cameron & Duff, 2007), one type of event, e.g. falls (Hsiao & Simeonov, 2001; Rivara & Thompson, 2000a), or on one industry, e.g. agriculture or construction industry (Lisa & Risto, 2000; Lehtola et al., 2008; Rautiainen et al., 2008). Specificity of focus can decrease the likelihood of misclassification and may therefore be useful when addressing very narrow review questions. On the other hand this would also severely restrict the number of studies eligible for review. Also, too specific a focus produces results that are ultimately not useful in broadly addressing the types of safety interventions that aim to prevent injuries at work, i.e., identify a ‘best’ approach to safety intervention in a particular context. For this reason this review will consider all work settings and the contextual factors reported, and also consider the relevant follow-up times, as the latency period for when to expect an effect will differ depending on the type of safety intervention. The current evidence may be summarised as follows:

Multifaceted approaches: A review of accident prevention programs by Lund and Aarø (2004) concludes that the greatest effect is obtained in a combination of attitudinal, behavioural and structural approaches (multi-faceted interventions). Even though the review is comprehensive, and provides a very useful categorization and modelling of the level and type of intervention, the study did not establish summarized effect sizes for the various prevention measures. Moreover, the study also included non-occupational accidents such as those occurring during leisure time, in traffic and at homes. The review also suggested that it might not be possible to influence an organisation's safety culture directly, and this has also been supported by other studies (Richter & Koch, 2004; G. Grote, 2007). A review by Guastello (1993) compared reductions in accidents for ten types of workplace safety interventions, and showed that individual approaches had smaller effect sizes compared to more comprehensive (multi-faceted) programs. However, the review did not assess the statistical significance for effect sizes; thorough and systematic assessment of the methodological quality of the included studies was lacking; and finally, an appropriate conceptual categorization of the workplace safety programs was missing. A recent update of a Cochrane systematic review (van der Molen et al., 2008) assessed the effectiveness of preventing injuries in the construction industry (van der Molen et al., 2013). Of the 17 studies included in the review, 12 were Interrupted Time Series studies (ITS) and one was a controlled before-and-after (CBA) study. The authors reported limited evidence for the effectiveness of a multifaceted safety campaign (Spangenberg et al., 2002) and a multifaceted drug program (Wickizer, Kopjar, Franklin, & Joesch, 2004) in preventing injuries, and no evidence was found that legislation is effective in preventing nonfatal or fatal injuries in the construction industry (Lipscomb, Li, & Dement, 2003). The methodology of this review has been criticised recently for a lack of sensitivity to context and a lack of flexibility in using the Effective Practice of Organisation of Care (EPOC), which makes it difficult to address challenges faced by workplace interventions (Lipscomb, Pompeii, Myers, Schoenfisch, & Dement, 2009). In order to avoid such criticism in the present review we have taken some important methodological steps: firstly, we do not restrict the review to the construction industry; secondly, we implement various methods for the control of ITS designs; thirdly, we distinguish between various types of safety interventions by classifying them theoretically and conceptually; fourthly, we consider the contextual factors reported; and finally, we consider the relevant follow-up times, as the latency period for when to expect an effect will differ depending on the type of safety intervention.

Even though a number of reviews and evaluations of safety interventions have been conducted, reviews of workplace safety interventions using systematic approaches are limited in number, not up to date, and not comprehensive, particularly in systematic reviews that include programs covering different levels and types of components. This review aims to evaluate the effects of the main types of safety interventions for preventing injuries at work and to synthesize best practices that are widely applicable.

This systematic review will summarize the most recent scientific evidence on the effectiveness of the main types of safety interventions and their components in preventing accidents at work; the process will be based on the conceptual model of Lund and Aarø (2004). A main type of safety intervention could be, for example, attitude/knowledge modification at the individual level or structural modification at the organizational level (Figure 1). The review aims to fill the gap in the extant knowledge on safety interventions at work by evaluating the effects of the main types of interventions for preventing injuries at work and to synthesize best practices that are widely applicable. The review will not be confined by language, geographical location of the study, age or gender. It will only consider accidents at work; it will not include those which occur in the home, or during leisure activity, or which involve third parties such as hospital patients or pedestrians passing a construction site.

OPEN IN VIEWER

1:

Possible pathways for preventing injuries at work

3.1 CRITERIA FOR CONSIDERING STUDIES FOR THIS REVIEW

A variety of research designs have been used to evaluate workplace safety interventions. Random allocation is not always feasible for all types of safety interventions in workplace settings because of a number of practical issues. Moreover, workplaces are often highly dynamic and complex social entities and accidents are rare events, thus not lending workplace safety interventions to be evaluated using highly controlled studies. For the same reason, observational studies are included in the review which allows for the assessment of longer term effects and opportunities to incorporate larger samples at times. Thus, for the purposes of this review we will be including a broader range of studies than is typically found in Campbell reviews. In addition, because safety interventions are often directed at the workplace level, this review will not only consider studies measuring effects at the level of the individual, but also consider studies measuring effects at the organizational level. We will, however, keep studies measuring effects at the individual level separate from studies measuring effects at the organizational level because they are asking different questions and are analysing data at two different levels.

3.2 SEARCH METHODS FOR IDENTIFICATION OF STUDIES

3.3 DATA COLLECTION AND ANALYSIS