A Randomised and Controlled Trial of Participative Ergonomics for Manual Tasks (PErforM)

A participative ergonomics approach to reducing injuries associated with manual tasks is widely promoted, however only limited evidence from uncontrolled trials was available to support the efficacy of such an approach. This paper reports on a randomised and controlled trial of PErforM, a participative ergonomics intervention designed to reduce the risks of injury associated with manual tasks. One hundred and seventeen small to medium sized food, construction, and health workplaces in South-East Queensland were audited by government inspectors using a manual tasks risk assessment tool (ManTRA). Forty-eight volunteer workplaces were then randomly assigned to Experimental and Control groups with the Experimental group receiving the PErforM program. Inspectors audited the workplaces again, nine months following the intervention. A significant decrease in estimates of manual task risk was observed in the intervention group, as was an increase in risk assessment activity and improved management systems for manual task risk assessment and control. These data provide strong evidence for the efficacy of participative ergonomics approaches to reducing the risk of musculoskeletal injury associated with manual tasks. A Randomised Controlled Trial of Participative Ergonomics for Manual Tasks (PErforM) Burgess-Limerick, Egeskov, Pollock & Straker 2/30 Introduction Musculoskeletal injuries related to the performance of manual tasks have been recognised as a source of significant pain, disability and disadvantage for the injured person and a substantial burden on modern societies. Statistics across a range of jurisdictions suggest that more than 30% of all occupational injuries are musculoskeletal injuries associated with manual tasks (eg., NOHSC 1998, Liberty Mutual 2002). One approach to reducing the burden from musculoskeletal injuries is participative ergonomics. Participative ergonomics developed from quality circles in Japan (Noro 1991), industrial democracy and social participation in Europe and Scandinavia (Jensen 1997), and the failure of traditional corporate control models to bring economic growth to USA companies in the 1970s (Brown 1993). Whilst there are many variations in the models and techniques used in participative ergonomics (see Haines and Wilson 1998 for a review), the basic concept is to involve workers in improving their workplaces to reduce injury and increase productivity. In this way the expert knowledge workers have of their own tasks is utilised to assist in risk assessment and control. Potential benefits of participative ergonomics are thought to include an improved flow of useful information within an organization, an improvement in the meaningfulness of work, more rapid technological and organisational change, enhanced performance and reductions in work related health problems (Brown 1993, Haims and Carayon 1998). Participative ergonomics has been used to create more human centred work (Imada 2000), to improve work organisational climate (Maciel 1998), to reduce mental workload (Vink et al. 1995) and to rehabilitate workers with back pain (Loisel et al. 2001). Participative ergonomics has also been used to try to prevent musculoskeletal disorders associated with manual tasks across a range of industries including electrical manufacturing (St-Vincent et al. 1998), car manufacturing (Halpern and Dawson 1997), meat processing (Moore & Garg 1997, NIOSH 1994), print media (Rosecrance and Cook 2000), office computer work (Westlander 1995), construction (de Jong and Vink 2000, Vink et al. 1997) and health (Straker 1990); and it is now the internationally recommended approach to reducing musculoskeletal diorders associated with manual tasks (Carrivick et al. 2001, DOL 1999, Jensen 1997, NIOSH 1997, Stubbs 2000). Despite this, there is only limited evidence to support the efficacy of such an approach. Evidence for the Efficacy of Participatory Ergonomics Many uncontrolled case studies of participative ergonomics have been reported, with some showing improvements in health outcomes. For example, Koda et al. (1997) reported an increase in risk control actions and a decrease in compensable back pain incidence in an uncontrolled trial in the Tokyo waste disposal organization. Conversely some case studies have found a deterioration in health outcomes following a participative ergonomics intervention. For example, Moore and Garg (1997) reported an increase in musculoskeletal disorder incidence rate and lost time incidence rate in an uncontrolled trial in a meat processing plant. (Their latter paper [1998] reported a marked decrease in lost time incidence rate.) However the absence of a comparison group means that the cause of any changes found in these and similar studies may not have been due to the intervention alone. A Randomised Controlled Trial of Participative Ergonomics for Manual Tasks (PErforM) Burgess-Limerick, Egeskov, Pollock & Straker 3/30 Wickstrom et al. (1993) conducted a more strongly designed study in the metal fabrication industry in Finland. A participative ergonomics intervention was delivered in one factory, with a similar factory acting as a comparison. This eliminated some confounding factors, such as the impact of general economic conditions over time. A decrease in back pain in the intervention factory was reported, while no change was observed in the comparison factory. A study with even better control was reported recently by Carrivick et al. (2001, 2002). This study involved an intervention for cleaners in one hospital in Australia and compared outcomes with orderlies in the same hospital, cleaners in a nearby comparable hospital, and cleaners in the whole of Western Australia. The intervention group experienced substantial reductions in injury rate, injury duration and injury costs, compared to increases in the comparison groups. The Wickstrom et al. and Carrivick et al. studies provide reasonable levels of evidence for the efficacy of participative ergonomics to improve musculoskeletal injury outcomes. However, as the intervention and comparison groups were not randomly assigned, confounding factors may have caused observed differences. This limitation in the available evidence has impeded the implementation of participative ergonomics at an organisational level and at a government level (eg the difficulties associated with the introduction of an “ergonomics standard” in the USA). Randomised and controlled trials are recognised as the highest level of evidence (Sackett et al. 1997). Whilst the need for randomised and controlled trials in ergonomics has been recognised (Straker et al. 2001), only two such trials of participative ergonomics interventions have been reported. Morkden et al. (2002) recently reported on a randomised and controlled trial in the Norwegian aluminium industry evaluating worker versus supervisor and manager training forms of participative ergonomics. Employees from eight plants were involved with supervisors and managers only, whole work teams, or work teams without their supervisors, in two departments assigned to participative ergonomics training or no training. Within the intervention groups, some groups had the supervisor only trained whilst others had the whole work team trained. Control work teams were on different shifts but were aware of the study and were consulted about control changes. Other departments in the plants formed a secondary non-randomised control group. The authors reported an increase in coping strategies in the intervention group but no changes in job control or musculoskeletal symptoms. Earlier, Loisel et al. (1997) reported on a randomised and controlled trial using participative ergonomics to return people with back pain to their work. They found that the use of a group (including ergonomist, injured worker, supervisor and management and union representatives) to evaluate and generate solutions to hazards in the injured worker's worksite resulted in a more rapid return to full work duties than medical management. No randomised and controlled trial of any ergonomics approach to reduce injuries at a workplace level has previously been reported. This is at least partly due to the resource and logistic difficulties in mounting a randomised and controlled trial involving multiple organisations (Straker 2000a). A Randomised Controlled Trial of Participative Ergonomics for Manual Tasks (PErforM) Burgess-Limerick, Egeskov, Pollock & Straker 4/30 Measures of Participative Ergonomics Efficacy Studies attempting to evaluate the efficacy of participative ergonomics interventions to reduce musculoskeletal disorders have used a plethora of outcome measures ranging from ratings of efficacy by a participative ergonomics committee (Rosecrance and Cook 2000), through physical measures such as heart rate (Pohjonen et al. 1998), to productivity measures (Maciel 1998) and cost-benefit analysis (Lanoie and Tavenas 1996). Straker (2000b) classified the possible measures as either short or long term. Short term measures included measures of procedural activity, physical risk measures and psychosocial measures. Long term measures included productivity and health outcomes. Procedural measures could include direct risk assessment and control activities as well as more indirect activities such as the development of appropriate management systems and policies. Physical risk measures could include observational ratings, physical measurement of hazards (eg. weight of box), measurement of biological responses (eg. heat rate) and estimates of biological stresses (eg. low back moment). Psychosocial measures could include ratings by individuals of their discomfort, exertion, workload and satisfaction and estimates of the organisational environment such as safety culture and team cohesion. Productivity measures could include work unit output, product quality/failure rate, system down time, absenteeism. Efficacy has also been evaluated using cost-benefit analysis. Long term health outcomes commonly used include musculoskeletal injury incidence rates, durations and associated costs. The limitations of the various measures have been discussed previously (see Burdorf 1992, Straker 1991 and 2000b, Li and Buckle 1999a and b) and it is clear that no single measure is adequate. Aim The aim of the current research was to evaluate the efficacy of a participative ergonomics intervention aimed at reducing injuries associated with manual tasks through a randomised and controlled trial using a battery of outcome measures. This paper describes the physical risk estimates, legislative compliance, procedural activity, and organisational environment outcome measures obtained for randomly allocated Experimental and Control workplaces. The project also evaluated productivity and workers’ compensation outcomes and used in depth interviews to investigate the process of implementing a participative approach, and these data will be reported elsewhere. A Randomised Controlled Trial of Participative Ergonomics for Manual Tasks (PErforM) Burgess-Limerick, Egeskov, Pollock & Straker 5/30 Method Design A randomised and controlled trial was conducted to evaluate the efficacy, in terms of a range of outcome measures, of a participative ergonomics intervention aimed at reducing injuries associated with manual tasks in small to medium sized workplaces in three diverse industry sectors. An initial manual task audit of 117 randomly selected workplaces was undertaken by government inspectors between October and December 2000. These workplaces were invited to participate in the evaluation of the intervention, and 48 workplaces volunteered. Thirty-one of these workplaces were randomly assigned to the Experimental group and the remainder formed a Control group. Workplaces in the Experimental group received the intervention between March and July 2001. All workplaces, and an additional 30 similar workplaces which were not audited initially, were again audited between April and July 2002. The intervention was made available to Control workplaces in August to December 2002. This paper reports data obtained from the volunteer workplaces only (Experimental and Control). Sample Queensland is a large state in the North-East of Australia, covering 1.7 million km, with a population of 3.5 million people. The South-East corner of the state is relatively densely populated, with 65% of the population (2.3 million people) in an area equal to 1.3% of the state (22,339 km). Workplace health and safety legislation in Australia is largely a State government responsibility, and the relevant authority in Queensland is the Division of Workplace Health and Safety in the Department of Industrial Relations. This authority maintains a database of all workplaces in the State, and inspectors employed by the authority have right of entry to inspect any workplace, without notice, to assess compliance with the Workplace Health and Safety Act (DIR 2000). Under the Act, workplaces employing 30 or more staff are required to manage the risk associated with manual tasks, have a Workplace Health and Safety Officer and must establish a health and safety committee. To assist employers to meet their obligations under the Act, the Department provides advisory standards, including a Manual Tasks Advisory Standard enacted in February 2000 (DWHS 2000). The Advisory Standard also advocates a consultative, participatory approach to risk assessment and control as the main approach to reducing associated musculoskeletal injuries. The workplaces chosen for the initial audit employed 30-100 employees, were single workplace employers (ie, not part of a larger organisation), located within South-East Queensland, and in one of three industry sectors. Small to medium sized workplaces were chosen as these have been identified as an area of need because they employ a large proportion of the workforce yet have limited ability to provide in house ergonomics expertise. Independent organizations were required so the workplace management could make decisions not constrained by remote, higher level management. The industry sectors were chosen in consultation with Division of Workplace Health and Safety staff after a review of relevant compensation statistics. The workplaces chosen for initial audit conducted business in: food processing other than meat (Australia and New Zealand Standard Industry Classification [ANZSIC] codes 2112-2190); construction A Randomised Controlled Trial of Participative Ergonomics for Manual Tasks (PErforM) Burgess-Limerick, Egeskov, Pollock & Straker 6/30 related manufacturing and wholesaling (ANZSIC 2323, 2741, 2742, 4531, and 4539); or health and community services, specifically nursing homes and accommodation for the aged (ANZSIC 8613 and 8721). For the sake of brevity, the industry sectors are hereafter referred to as “Food”, “Construction” and “Health”. Procedure Following identification of all workplaces which met the inclusion criteria (N=162), 120 workplaces were randomly selected for initial audit. Three workplaces were not able to be audited due to closure or some other reason. Seventeen government Workplace Health and Safety inspectors were trained by the investigators in the use of a Manual Tasks Risk Assessment tool (ManTRA) which incorporates assessment of manual task risk levels, manual tasks related safety activity and organisational environment (further details are provided in the following section). The Queensland government publicised that their inspectors would be conducting a ‘blitz’ on manual tasks, following the release of a revised Manual Tasks Advisory Standard in February 2000 (DWHS 2000). As required by the Workplace Health and Safety Act, where inspectors observed instances of tasks, which in the opinion of the inspector posed a significant risk of injury, the inspector was required to take action. This enforcement action involved either: a) issuing a prohibition notice which mandates immediate cessation of performance of the task; b) issuing an improvement notice which requires the employer to comply with the details of the notice within a given timeframe; or c) providing formal written advice regarding the nature of the risk. Following the initial audit, all audited workplaces (N=117) were offered the opportunity to participate in the evaluation of a participative ergonomics intervention aimed at reducing injury risks associated with manual tasks. Due to project timing constraints, random allocation of volunteers was initially weighted towards the Experimental group. The expected number of volunteer workplaces did not eventuate and consequently the Control group was smaller than planned. Forty-eight workplaces volunteered and 31 of these workplaces were randomly assigned to the Experimental group. The remaining 17 workplaces formed the Control group. Workplaces allocated to the Control group were informed that, due to the design of the evaluation, their participation would be delayed until the following year. Table 1 provides a summary of the breakdown of workplaces by industry sector. Table 1: Summary of workplace numbers by Group and Industry sector. Food Construction Health Total Experimental 8 7 16 31 Control 3 5 9 17 Refused 21 31 17 69 Total 32 43 42 117 Three ergonomists were trained in the delivery of a participatory ergonomics for manual tasks (PErforM) intervention designed by the investigators. The intervention aimed to improve each workplace’s management systems to support participation in a risk assessment and control process; and provide supervisors and work teams with sufficient knowledge and skills to enable them to perform manual task risk assessment and control. The intervention was delivered to each workplace over a series of four sessions. In most cases these sessions were held on separate days, however some flexibility existed to accommodate individual workplace circumstances. An outline of the PErforM program is provided in Appendix A. A Randomised Controlled Trial of Participative Ergonomics for Manual Tasks (PErforM) Burgess-Limerick, Egeskov, Pollock & Straker 7/30 The intervention was delivered by the three consultants to each of the 31 workplaces over a six month period, although the involvement of each workplace typically spanned a period of less than three months. Workplaces were typically visited over several weeks to provide ongoing encouragement and support. The investigators retrained five of the original inspectors and trained a further six inspectors in the use of ManTRA, and the second audits occurred between April to July 2002 (at least nine months after the delivery of the intervention was complete). The government again publicised the conduct of a manual tasks “blitz” in the mass media. The inspectors were not informed of which workplaces had volunteered for the intervention, or which had received the intervention. Audit Tool The audit tool used by the inspectors (ManTRA) was designed by the investigators to serve a dual purpose: (a) to assist inspectors to form an opinion regarding the compliance of the workplace with the requirements of the relevant advisory standard (DWHS 2000), and consequently whether any formal action was required; and (b) to provide the investigators with a method of assessing the level of manual task related safety activity, relevant organisational environment variables, and an assessment of the level and nature of manual task related injury risk present in the workplace. The usability, reliability and validity of the tool was tested with government inspectors and found to be good. A copy of the complete tool, and accompanying explanatory notes, is available from the UQ Ergonomics website (ergonomics.uq.edu.au/download/mantra.pdf – password available from [email protected]). The initial sections of the audit tool involved identifying the jobs performed by staff at the workplace and the number of staff in each category. The inspectors then selected staff to interview with the intention of selecting a 10% representative sample of nonadministrative or managerial staff. The number of staff actually selected for interview varied from 1 to 13 depending on the size of the workplace and the inspectors’ perceptions of the degree of risk present at the workplace. At least one supervisor, and at least one person with responsibility for purchasing, were also interviewed using questions provided in the audit tool. The questions asked of employees included a measure of procedural activity based on Cheyne, Cox, Oliver and Tomas’ (1998) Safety Activity measure. Employees were asked to indicate whether they had participated in certain safety activities in the past twelve months. Cheyne et al.’s list of activities was modified to reflect the focus on manual tasks (eg, Cheyne et al.’s “Seen a safety video” was re-written as “Seen a manual task safety video”). Activities not related to manual tasks were omitted (eg “took place in fire evacuation practice”). The final list consisted of nine activities. Managers were also asked about the number of manual tasks risk assessments carried out at the workplace, and the number of controls implemented. The quality of the safety management system for manual tasks was indicated by the extent to which the organisation followed the guidelines for management of safety provided by the Division of Health and Safety (DWHS 1999). The Division’s guidelines require management of: • Health and safety policy • Allocation of responsibilities and accountabilities for health and safety A Randomised Controlled Trial of Participative Ergonomics for Manual Tasks (PErforM) Burgess-Limerick, Egeskov, Pollock & Straker 8/30 • Sub-contractor and purchasing controls • Health and Safety consultation with employees • Risk management • Provision of information • Training • Incident and injury investigation, reporting and recording Supervisors, purchasing officers, and the employees who were doing tasks being analysed by the inspectors were asked questions targeting each of the elements of the management system relevant to their level. For example, supervisors were asked to indicate the extent to which their allocation of health and safety responsibilities was clear, specific, in writing and known to other employees (relating to the guideline on “allocation of responsibilities for health and safety”). Purchasing officers were asked if risk assessments were conducted prior to purchasing new equipment (guideline on “sub-contractor and purchasing control”). Employees were asked if they had received information on the health and safety policy and if so the source of the information (guideline on “health and safety policy”). A confirmatory factor analysis of the data subsequently confirmed that the different questions designed to gain information on each guideline related to each other in the expected way. Information on the following factors was derived from the audits. Factors from supervisors: • Allocation of responsibilities for general health and safety • Allocation of responsibilities specifically for manual tasks • Training (for health and safety and risk assessment responsibilities) Factors from purchasing officers: • Sub-contractor controls • Purchasing controls Factors from employees: • Communication of health and safety policy via line management • Communication of health and safety policy via health and safety official • Health and safety consultation representation • Manual task risk management • Training for manual tasks • Incident and injury investigation Further information on the organisational environment was also collected in a parallel qualitative study involving interviews with participating workplaces and ergonomics consultants. The results of the qualitative study will be reported separately. The audit process also involved the inspectors asking employees questions about body discomfort, the tasks undertaken, and their opinions regarding the links between particular tasks and discomfort. On the basis of this information, the inspectors’ observations, and knowledge of the relevant industry, the inspector selected tasks for observation and coding for risk exposure using a matrix developed for this purpose by the investigators. The inspectors were requested to select a sample of two-three tasks for every worker and a sample of one worker for every 10 workers at a workplace, however the A Randomised Controlled Trial of Participative Ergonomics for Manual Tasks (PErforM) Burgess-Limerick, Egeskov, Pollock & Straker 9/30 number of tasks chosen for examination was influenced by the inspectors’ perceptions of the risk associated with tasks performed in each workplace. The manual tasks risk matrix (see Appendix B) involves a rating of severity (on a five point scale) of risk factors for each of five body regions. For each task inspectors rated five body regions (lower limb, back, neck, shoulder/arms, wrist/hand) independently for seven types of risk. The types of risk were the total task time, duration of continuous performance without a break, cycle time, force, speed, awkwardness and vibration. Cycle time and duration of continuous performance scores were combined into a “repetition” risk factor (on a five point scale) using the table presented in Appendix B. Force and speed scores were similarly combined in an “exertion” risk factor. In this way the risk matrix was aligned with the direct risk factors for manual tasks identified in the advisory standard relevant to the jurisdiction (DWHS 2000). For each task which was assessed, inspectors also made a decision about whether the task complied with legislative requirements or whether formal action was justified. If an inspector believed there was non compliance the inspector could: a) issue a notice prohibiting the task from being performed until changes were made; b) issue a notice requiring a workplace to improve the task in some way within a specified time-frame; or c) issue formal advice regarding a manual task risk. Issued notices placed a legal obligation on the obligation holder (eg. employee, employer, supplier), with the threat of prosecution and significant fines for lack of compliance. Whilst the inspectors’ decisions regarding whether to take action were based on the inspectors’ opinions, guidance was provided to the inspectors that action may be indicated on the basis of the completed risk matrix when one or more of the following criteria were met: a) an exertion score of five for any body region; b) the sum of exertion and awkwardness was eight or greater for any body region; or c) the cumulative risk score (sum of total time, exertion, awkwardness, vibration and repetition) for any body region was 15 or greater. Analysis Four groups of dependent variables are reported in this paper: manual task risk estimates, inspector actions, procedural activity and organisational environment. Manual task risk estimates included: • total assessed risk exposure (TARE) – the sum of cumulative risk scores for all tasks assessed at a workplace • number of tasks assessed at each workplace • workplace mean cumulative risk (total task risk across all body regions) • workplace mean level of exposure to each risk factor (summed across body regions) • workplace mean level of risk for each body region (summed across risk factors) • workplace mean number of tasks which exceeded one or more action criteria Inspector actions measures included the number of enforcement actions taken by an inspector at each workplace. Procedural activity measures included: • number of safety activities reported by employees • number of manual tasks risk assessments undertaken, as reported by managers A Randomised Controlled Trial of Participative Ergonomics for Manual Tasks (PErforM) Burgess-Limerick, Egeskov, Pollock & Straker 10/30 • number of manual task risk controls implemented, as reported by managers Organisational environment measures included information from supervisors, purchasing officers and employees based on the factor analysis as described above. The independent variables for statistical analysis were group (Experimental, Control), Industry (Food, Construction, Health) and Time (Audit 1 pre intervention, Audit 2 post intervention). Three-way mixed model ANOVAs (Time as a repeated measure) were computed for the variables defined above using SPSS v10. Partial eta was calculated as a measure of effect size (by convention, a partial eta of 0.02 indicates a small, 0.15 a medium, and 0.35 a large effect). A Randomised Controlled Trial of Participative Ergonomics for Manual Tasks (PErforM) Burgess-Limerick, Egeskov, Pollock & Straker 11/30 Results Total Assessed Risk Exposure Figure 1 illustrates the workplace Total Assessed Risk Exposure for the Experimental and Control groups before and after the intervention for each industry sector and combined industries. Whilst the Experimental group appeared to start with a higher risk exposure, following intervention there was a clear trend for the Experimental group workplace risk exposure to reduce while the Control group risk exposure tended to increase or remain stable. A Time x Group x Industry mixed model ANOVA confirmed the Time x Group interaction (F1,31 = 5.40, p = .027, partial eta 2 = .148). There were no significant main effects of Time (F1,31 = 2.12, p = .155, partial eta 2 = .064), Group (F1,31 = 1.07, p = .309, partial eta = .033) or Industry (F2,31 = 0.38, p = .690, partial eta 2 = .024) nor were the remaining 2 and 3 way interactions significant (Group x Industry F2,31 = 0.49, p = .618, partial eta = .031, Time x Industry F2,31 =1.08, p =.352, partial eta 2 = .065, Time x Group x Industry F2,31 = 0.20, p = .822, partial eta 2 = .013). Figure 1: Workplace mean (+SEM) Total Assessed Risk Exposure by Group, Time and Industry. T o ta l A s s e s s e d R is k E x p o s u re 0 100 200 300 400 500