...

Feasibility of carrying out an ergonomics of musculoskeletal disorders

by user

on
Category: Documents
3

views

Report

Comments

Transcript

Feasibility of carrying out an ergonomics of musculoskeletal disorders
Health and Safety
Executive
Feasibility of carrying out an ergonomics
intervention study to prevent the incidence
of musculoskeletal disorders
Prepared by the Health and Safety Laboratory
for the Health and Safety Executive 2008
RR626
Research Report
Health and Safety
Executive
Feasibility of carrying out an ergonomics
intervention study to prevent the incidence
of musculoskeletal disorders
Dr Andrew D J Pinder & Dr Johannah F Wegerdt
Health and Safety Laboratory
Harpur Hill
Buxton SK17 9JN
This work examines the feasibility of assessing the effectiveness of workplace ergonomic interventions to prevent the
onset of musculoskeletal disorders (MSDs). It reviews existing models of causation of MSDs and the scientific literature on
interventions to prevent MSDs. It describes relevant epidemiological methods and research protocols.
Many previous studies of the risk factors for MSDs have not been able to assess causation and the need remains
for intervention studies of high methodological quality to do this. A longitudinal Cluster Randomised Trial is the most
appropriate study design for assessing MSD causation in an occupational setting. Measurement of injury rates generally
requires very large samples and/or long follow-up times to provide adequate statistical power. It is likely that the study
would need to be carried out across multiple employers.
Because of the scale of the MSD problem, it is recommended that HSE consider funding or part-funding a study designed to
test the effectiveness of workplace ergonomics interventions to prevent the onset of episodes of musculoskeletal disorders.
Consideration should be given to making the study a multi-centre, possibly international, collaborative study. Such a study
would be high risk due to the scale and duration needed and the practical and organisational difficulties involved.
This report and the work it describes were funded by the Health and Safety Executive (HSE). Its contents, including any
opinions and/or conclusions expressed, are those of the authors alone and do not necessarily reflect HSE policy.
HSE Books
© Crown copyright 2008
First published 2008
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means (electronic, mechanical, photocopying, recording or otherwise) without the prior written permission of the copyright owner.
Applications for reproduction should be made in writing to:
Licensing Division, Her Majesty’s Stationery Office,
St Clements House, 2-16 Colegate, Norwich NR3 1BQ
or by e-mail to [email protected]
ii
CONTENTS
1 INTRODUCTION......................................................................................... 1
1.1
Purpose of this report .............................................................................. 1
1.2
Why interventions? .................................................................................. 1
1.3
The nature of MSDs................................................................................. 2
1.4
Terminology............................................................................................. 2
1.5
Frameworks for understanding MSDs and identifying possible interventions ....................................................................................................... 4
1.6
Model of an intervention study................................................................. 6
1.7
Purpose of the intervention...................................................................... 8
1.8
Epidemiology of MSDs ............................................................................ 9
1.9
Risk factors (“flags”) for the development of chronicity of MSDS............. 9
2 REVIEW OF EXISTING SCIENTIFIC LITERATURE ................................ 11
2.1
Systematic reviews ................................................................................ 11
2.2
Specific MSD risk factors....................................................................... 12
2.3
Guidelines.............................................................................................. 13
2.4
The value of mechanical intervention in prevention of MSDs ................ 15
2.5
Effectiveness of different interventions .................................................. 16
2.6
Cost effectiveness of interventions ........................................................ 16
2.7
Multi-component interventions............................................................... 18
2.8
Return to work policies .......................................................................... 18
2.9
Recent prospective studies.................................................................... 19
2.10 Recommendations for future work ......................................................... 19
3 EPIDEMIOLOGICAL METHODS.............................................................. 21
3.1
Introduction............................................................................................ 21
3.2
Planning................................................................................................. 22
3.3
Study design.......................................................................................... 33
3.4
Data collection ....................................................................................... 39
3.5
Data analysis and interpretation ............................................................ 43
3.6
Presentation of results ........................................................................... 44
4 EXAMPLE PROJECT PLAN .................................................................... 45
4.1
Introduction............................................................................................ 45
4.2
Assumptions underlying the hypothetical project plan ........................... 45
4.3
Possible Gantt chart for the project with crude estimates of staffing, time and costs .......................................................................................................... 45
4.4
Study implementation issues ................................................................. 51
4.5
Discussion ............................................................................................. 56
5 CHECKLISTS FOR INTERVENTION STUDIES....................................... 58
5.1
Methodological criteria for evaluating occupational safety intervention research ........................................................................................................... 58
5.2
Criteria for reporting randomised trials – the CONSORT statement ...... 59
5.3
Examples of possible study proposals................................................... 66
iii
6 DISCUSSION/RECOMMENDATIONS...................................................... 74
6.1
Introduction............................................................................................ 74
6.2
Study design.......................................................................................... 75
6.3
Main recommendation ........................................................................... 75
6.4
Detailed recommendations .................................................................... 76
7
GLOSSARY .............................................................................................. 78
8
ACRONYMS AND ABBREVIATIONS ...................................................... 80
9
REFERENCES.......................................................................................... 81
iv
EXECUTIVE SUMMARY
Objectives
This work is in response to the Health and Safety Executive (HSE) tender specification entitled:
“Feasibility of Assessing the Effectiveness of Preventative Workplace Ergonomic
Interventions”. This requested an examination of the feasibility of assessing the effectiveness of
such interventions to prevent musculoskeletal disorders (MSDs). Reducing the risk of workers
suffering these problems would reduce the resultant burden of pain, medical treatment and
work-loss.
Preventative (primary) interventions are distinct from secondary interventions, as these are
aimed at preventing chronic disability in individuals already experiencing musculoskeletal
problems, and from tertiary interventions, which are aimed at rehabilitating or getting back to
work individuals suffering from chronic musculoskeletal disorders.
This report reviews existing models of causation of MSDs, reviews the scientific literature on
interventions to prevent MSDs, and summarises the relevant epidemiological methods and
research protocols. It illustrates the methodological issues through an example project plan and
provides checklists for evaluating proposals and reports of randomised intervention studies.
Main Findings
An integrated model for understanding MSD causation has recently been proposed by Karsh
(2006). This integrates physiological and psychological demands of work, the social and
cultural context, the work organisational context and the physical environment with the factors
unique to the individual and proposes pathways and feedback loops through which all these
factors can affect the detection and labelling/attribution of symptoms so that a case of a workrelated MSD occurs.
There is strong evidence (Waddell and Burton, 2001) that:
• Most adults (60-80%) experience low back pain (LBP) at some time, and it is often
persistent or recurrent.
• Physical demands of work (manual materials handling, lifting, bending, twisting and
whole body vibration) can be associated with increased reports of back symptoms,
aggravation of symptoms and ‘injuries’.
• Physical demands of work (manual materials handling, lifting, bending, twisting and
whole body vibration) are a risk factor for the incidence (onset) of LBP, but overall it
appears that the size of the effect is less than that of other individual, non-occupational
and unidentified factors”.
• Care seeking and disability due to LBP depend more on complex individual and workrelated psychosocial factors than on clinical features or physical demands of work
Waddell and Burton (2001) argued that while it is reasonable to seek to reduce the incidence
and prevalence of LBP by reducing exposure to known occupational risk factors, the complexity
of the causal factors may limit the effect of occupational interventions in reducing the societal
impact of LBP. There is evidence that multi-dimensional workplace interventions can be
recommended to reduce some aspects of LBP, but there is insufficient evidence to recommend
specific combinations of interventions (Burton et al., 2004; 2005).
A less despairing view of mechanical issues and back pain (McGill, 2002) criticised the view
that psychosocial variables dominate any biological or mechanical variables. McGill argued
v
that the position that biomechanics plays no role in back health and activity tolerance can be
held only by those who have never performed physical labour and have not experienced first
hand the work methods that must be employed to avoid disabling injury.
The consensus of the guidelines and the many systematic reviews that have been carried out is
that many of the previous studies of risk factors for low back pain are of low methodological
quality and that there is still a need for methodologically high-quality intervention studies.
There are current worldwide efforts to carry out such studies, but there is clear recognition of
the practical difficulties involved, which include:
• Statistically intractable issues in many situations.
• Complex and partially uncontrollable situations in which interventions must be
implemented, thus weakening and confusing the studies.
• The difficulty of recruiting sufficient subjects in jobs that are suitable for intervention.
• The difficulty of measuring the effects of the interventions on health and other possible
outcomes such as productivity changes.
There are published methods of assessing the quality of project proposals and reports of
epidemiological studies. The CONSORT statement (Moher et al., 2001) gives guidance on
reporting Randomised Controlled Trials (RCTs). The Epidemiological Appraisal Instrument
(EAI) (Genaidy et al., 2007) gives rigorously tested criteria and detailed specifications for
assessing epidemiological studies.
Investigating the prevention of MSDs with a workplace intervention requires that the study
design be able to assess causation and the effects of an experiment, not just association. This
requires that the study be a longitudinal RCT where a cohort of workers is split into one or more
intervention groups and one or more control groups and followed over time. It is very likely
that the appropriate type of RCT would be a Cluster Randomised Trial (CRT). This is where
randomisation is done at the group or cluster level instead of the individual level, and is often a
natural study design in an occupational setting.
It is impossible to give figures for sample sizes and study duration before decisions have been
made as to the detailed study design and analysis methods to be used. Measurement of injury
rates generally requires very large samples and/or long follow-up times to provide adequate
statistical power. Therefore such a study is likely to involve several thousand subjects and a
follow-up period of at least a year, and it is likely that 500-1000 individuals will be required per
group. Few employers have enough staff for such a trial to be carried out within one
organisation. It is therefore to be expected that workers will need to be recruited from multiple
employers.
Recommendations
Because of the scale of the MSD problem, HSE should consider funding or part funding a study
designed to test the effectiveness of workplace ergonomics interventions to prevent the onset of
episodes of musculoskeletal disorders. Consideration should be given to making the study a
multi-centre, possibly international, collaborative study. Such a study would be high risk due to
the scale and duration needed and the very significant practical and organisational difficulties
that would be faced.
Detailed recommendations are given in Section 6.4.
vi
1
1.1
INTRODUCTION
PURPOSE OF THIS REPORT
This work is in response to the Health and Safety Executive (HSE) tender specification entitled:
“Feasibility of Assessing the Effectiveness of Preventative Workplace Ergonomic
Interventions”. This is a request for a report looking at the feasibility of assessing the
effectiveness of preventative workplace ergonomic interventions, specifically looking at
musculoskeletal disorders (MSDs).
Musculoskeletal disorders are problems of the
musculoskeletal system, particularly of soft tissues such as tendons or muscles. Symptoms can
range in severity from transient aches or pains through to long-lasting disabling conditions such
as carpel tunnel syndrome or tenosynovitis. The most commonly affected region of the body is
the low back, followed by the upper limbs and then the lower limbs. In some circumstances,
MSDs are caused or made worse by work and are then referred to as “Work-related
Musculoskeletal Disorders” (WMSDs or WRMSDs).
The tender is therefore solely concerned with preventative interventions in the workplace that
are designed to reduce the risk of workers suffering from musculoskeletal disorders. It is not
concerned with secondary interventions (those aimed at preventing problems for individuals
who have already experienced musculoskeletal problems) or tertiary interventions aimed at
rehabilitating or getting back to work sufferers from chronic musculoskeletal disorders.
Intervention studies are known to be difficult, time consuming and costly. Musculoskeletal
disorders are seen as complex, multi-causal, often episodic problems that are often difficult to
diagnose and treat.
This report is intended to review existing studies, provide a description of research protocols
and describe sound methodology for intervention studies.
1.2
WHY INTERVENTIONS?
The purpose of an intervention study in the workplace is to demonstrate that changes in the
workplace can have an effect on health outcomes. Since MSDs are a major source of reports of
ill health, work absence and disability with consequent costs to the economy, it is almost
axiomatic that interventions that can be shown to prevent or alleviate such problems are
desirable in a civilised society. It is therefore of interest to HSE to investigate how effective in
preventing such problems are interventions that apply ergonomics to the workplace.
Moreover, the history of intervention research shows that despite intervention studies being very
costly in terms of time and manpower, they are well worth the effort involved (Kristensen,
2005). They provide two benefits when performed well: theoretical conclusiveness and
practical usefulness.
That there is need for more information about the feasibility of intervention studies is shown by
a complaint of the lack of a clear, prescribed methodology for intervention design (Loisel et al.,
2005). While Loisel et al. are writing specifically in the context of prevention of work
disability rather than from an occupational viewpoint, the complexity of the area means that
each study design will need to be carefully thought out in its own context. This is inevitable in
an area where there are many variables to be considered in many different occupational settings.
Therefore, any ergonomics intervention study must define its methodology carefully at the
planning stage rather than rely on applying a standard methodology unmodified.
1
1.3
THE NATURE OF MSDS
Because of the complexity of the musculoskeletal system, a large number of problems can fall
under the MSD label. These may include:
• Reports of pain (e.g. in the lower back) with no detectable pathology;
• Serious cases of joint or motion dysfunction;
• Problems caused by short-duration life events, such as pregnancy;
• Acute trauma, often involving specific tissues, caused by discrete events, which may
transfer significant amounts of energy to the tissues;
• Serious pathologies, such as Cauda Equina syndrome;
• Life-threatening conditions, such as cancer.
Depending on the factors, exposure to risk factors for MSDs may cause immediate effects (such
as trauma) or may cause symptoms after a long induction period. For example, an exposure
such as lifting may or may not have latent periods before chronic low-back pain results. MSD
symptoms are often transient, intermittent and episodic (Burdorf and van der Beek, 1999a). The
normal metabolic processes of tissue repair will lead to the resolution of many cases. As a
result, the causes of reports of pain are often not investigated in detail, especially if the case is
likely to resolve itself in a short period. However, there is much variability in the prognosis of
individual cases. Risks of recurrence are high and a percentage of cases become chronic. While
it is reassuring to the patient that it is highly likely that an acute episode of low back pain
(LBP), in the absence of indications of serious pathology (“red flags”) will resolve in a
relatively short time-scale, the risk of the individual suffering a future episode is elevated.
Thus, even if an initial episode is due to non-work related causes, the risk is elevated of a
second episode due to exposures at work.
The extent to which MSDs are work-related has been a cause of controversy, particularly in the
USA (Punnett and Wegman, 2004). They noted that: “The presence of one risk factor does not
negate another. Whether occupational factors account for few or many MSDs in the general
population, is not the same question as to what extent people can be protected from preventable
risks at work.” In their overview of the evidence and the controversy, they also remarked that
the relationships between MSDs and workplace risk factors “cannot be represented by a simple
one-to-one mapping”. Many risk factors have been implicated in the causation and maintenance
of MSDs, though the evidence of the sizes of the associations is quite varied. The risk factors
are believed to interact in complex and wide-ranging ways. The multi-factorial nature of MSDs
is also a consequence of many individuals being exposed to risk factors outside the workplace.
1.4
TERMINOLOGY
1.4.1
Primary, secondary and tertiary interventions — prevention and care
Interventions related to MSDs can entail either prevention or clinical care. Both types of
intervention can be subdivided into “primary”, “secondary” or “tertiary”.
Primary prevention “represents interventions with the uninjured worker population and the
workplace to avoid injury (and thus the subsequent need for primary care)” (Gatchel, 2004). It
therefore occurs either in the general population or in workplaces.
Primary care is concentrated on passive modes of treatment of MSDs in the early stages (the
acute phase) of reports of musculoskeletal problems. In other words, it involves symptom
2
control when “acute pain predominates”. Gatchel (2004) suggests this is normally 0-10 weeks
after injury occurrence.
Secondary prevention “refers to interventions with the freshly injured patient to avoid chronic
disability habituation by efforts to return the patient to productivity as soon as possible”
(Gatchel, 2004). According to him, both primary and secondary care can be part of this process,
along with workplace intervention and job-modification programs. It generally occurs first in
health care settings but may also occur in the workplace.
Secondary care (“reactivation care”) helps patients transfer from acute care back into the
workplace. It is usually provided in the first six months after injury or postoperatively as a
limited rehabilitation approach. It is the first level of rehabilitation when primary care does not
resolve the problem. It is “designed to facilitate return to productivity before progressive
deconditioning and psychosocial barriers supervene” (Gatchel, 2004). It is based on the
rationale that early recognition and management of risk factors for developing disabilities can
prevent chronic or permanent disability.
Tertiary prevention “attempts to avoid high costs associated with the permanent loss of
productivity of the small fraction of disabled workers who ultimately become the ongoing
disabled workers” (Gatchel, 2004). It will occur almost exclusively in health care settings.
Tertiary care is designed for the small fraction of individuals who exhibit chronic entrenched
disability. It provides intensive and individualised treatment to help them overcome
biomechanical dysfunction, physical deconditioning and psychosocial stressors (Anagnostis et
al., 2004; Gatchel, 2004).
1.4.2
Biopsychosocial models/interventions and psychosocial factors
The biopsychosocial model of musculoskeletal pain and disability “views pain and disability as
a complex and dynamic interaction among physiologic, psychologic and social factors”
(Gatchel, 2004). Gatchel contrasts it with what he calls “the outdated biomedical reductionistic
approach” in an attempt to emphasise the importance of the psychological and social factors and
that physiological explanations are by themselves inadequate to explain musculoskeletal pain
and its possible sequelae such as disability. The term “biopsychosocial” must be clearly
distinguished from “psychosocial” as they are not synonymous, especially in the context of
MSDs. Engel (1977) coined “biopsychosocial” to highlight the relationships between the
biological aspects, the psychological aspects, and the social aspects of illness that a physician
must consider when deciding how to treat a patient. The use of the term by authors such as
Gatchel (2004) and Waddell (1998) is in precisely this sense.
The term “psychosocial factors” when used in relation to MSDs does not refer to the status of an
injured, ill or disabled individual. Instead, it refers to the combination of psychological and
social factors that occur in the workplace that can influence how individuals perform their jobs
and hence can modify their behaviours in ways that can also affect their risk of suffering from
MSDs and hence of reporting problems or of taking sickness leave. It is “a non-specific term”
that “has served as catch-all in reference to non-physical elements of the job/work environment”
(Sauter and Swanson, 1996). In some circumstances “psychosocial factors” have been linked to
the Demands — Control — Support model of Karasek and Theorell (Engstrom et al., 1999;
Karasek and Theorell, 1990; Theorell, 1996; Theorell, 2004). Since the term “psychosocial”
has such a broad scope, there is clearly overlap with “biopsychosocial”, and considerable scope
for confusion. Where possible this report will use “psychosocial” in the context of work and
“biopsychosocial” in the context of illness/health care.
3
It is clear that both psychosocial and physical risk factors “share a common upstream
determinant” (Punnett and Wegman, 2004) since the organisation of a work process influences
both physical load patterns and psychosocial features such as job demands. Moreover, as
Punnett and Wegman note, “items such as “low job satisfaction” may represent a tautological
outcome of physical and/or psychosocial strain at work and/or the experience of MSD pain
while working”. It is therefore clear that any intervention that considers only one aspect of such
a complex situation is highly likely to be confounded or ineffective. As they also note,
intervening on the up-stream organisational characteristics could be expected to be effective on
both physical and psychosocial pathways, even if they are not independent.
1.4.3
Types of intervention evaluations
The methods that can be used to evaluate workplace interventions have been summarised under
six headings (Robson et al., 2001):
Table 1.
Types of intervention evaluations
Evaluation type
Evaluation purpose
1
Needs assessment
Determines what type of evaluation is needed
2
Process evaluation
Assesses the quality of the intervention delivery and identifies
areas for improvement
3
Effectiveness evaluation
Determines whether an intervention has had the effect intended on
outcomes, and estimates the size of the effect
4
Cost-outcome analysis
Determines the net cost of an intervention relative to its health
effect
5
Cost-effectiveness analysis
Compares different intervention alternatives using cost-effect
ratios
6
Cost-benefit analysis
Compares different intervention alternatives using net benefits
It is the feasibility of the third type, “Effectiveness evaluation”, that is of direct interest in this
report. The first two types must be considered in planning an ergonomics intervention. The
cost considerations of types 4, 5 and 6 will also need to be taken into account. While this will
be possible to a certain extent, before a full effectiveness evaluation is completed,
measurements of effect sizes will be needed from such studies to allow realistic cost-based
evaluations to be performed.
1.5
FRAMEWORKS FOR UNDERSTANDING MSDS AND IDENTIFYING
POSSIBLE INTERVENTIONS
Recently Karsh (2006) has reviewed previous models (Armstrong et al., 1993; Carayon et al.,
1999; Feuerstein, 1996; Hagberg et al., Kuorinka and Forcier, 1995; Kumar, 2001; Moon and
Sauter, 1996; National Research Council and Institute of Medicine, 2001; National Research
Council, 1999; Sauter and Swanson, 1996) for the causation of MSDs.
The major differences between the models related to the specificity of proposed pathways and
proposed mechanisms of action but the theories had the following factors in common:
• Physical and psychological exposures leading to doses, causing responses moderated by
individual factors
4
• Feedback mechanisms or cascading effects
• Several important considerations were not specified, particularly indication of specific
magnitudes, duration of exposure or latency periods.
Figure 1.
Karsh’s integrated model for WMSD causation
5
Based on the previous models, Karsh (2006) has proposed an integrated model that is
reproduced in a slightly simplified form in Figure 1. It links the social/cultural context, the
work organisation and the physical environment to the physical and psychological demands of
the job. The resultant physical and psychological strains depend on the moderating effects of
the physical capacity and psychological coping mechanisms of the exposed individual. These
strains then lead to physical and psychological responses that can then lead to the detection of
MSD symptoms and possibly a diagnosis of a WMSD. Symptom detection and WMSD
diagnosis can lead to modification of work demands, and work organisation, thus creating
feedback loops.
Karsh (2006) noted that the model indicated that many different factors acting simultaneously
can impact both doses and responses. In other words, interactions between factors need to be
considered. He described his model as yielding abundant information for intervention research.
He recommended that known “exposures” (by which he appears to mean “risk factors”) should
be the target of intervention research and that doses, responses and capacity factors should be
measured to the extent possible. He noted that typically it is not known how much exposures
should be reduced. In cases where complete elimination of an exposure is not possible there is a
clear need to examine varying degrees of reduction.
In the previous models “psychosocial risk factors” came within “Work organisation”
(Armstrong et al., 1993; Sauter and Swanson, 1996; Feuerstein, 1996) or were conceptualised as
a profile of the individual (Carayon et al., 1999) or as part of the workplace (National Research
Council and Institute of Medicine, 2001). Because the focus is on the workplace, the
biopsychosocial context of concern to clinicians (Waddell, 1998) is not in view.
Leboeuf-Yde (2004) proposed that instead of considering risk factors for the development of
LBP, the focus should be on “persons at risk”. She suggested that LBP is “but one expression”
of being generally frail and that therefore early identification of high-risk populations would
allow for a “selective preventive approach”. It is not clear how this proposal could be
implemented either at the stage of the identification of high-risk individuals/groups or at the
later stage of “selective intervention”. The main risk factor for an episode of low back pain is a
previous episode of low back pain, so this is a risk indicator that is not useful for primary
prevention. However, if individuals with a recent history of time off work due to back pain
were identified as “persons at risk” of new episodes, the difficulty would then be of selecting
the correct interventions without considering “risk factors”.
Griffiths noted that understanding the mechanisms that mediate successful interventions is a
crucial step to understanding organisational interventions and that such principles will be more
generalisable than the outcome of any particular intervention (Griffiths, 1999).
It is therefore recommended that any tender specification should use the integrated model
proposed by Karsh (2006) as a basis for identifying pathways that should be explored and of
evaluating proposals in response to the tender. It would be up to the drafter of the specification
to decide whether to specify the pathways that proposals should examine or to leave tenderers to
decide which factors to address.
1.6
MODEL OF AN INTERVENTION STUDY
A theoretical model of key elements in occupational intervention studies (Kristensen, 2005)
clarifies the relationships between theory, research and practice (Figure 2). Working down the
Theory column of the diagram shows that the intervention is intended to lead to reduced
exposure and hence to better health. The Practice column shows what the actual course of
events is and that, at all three levels, the actuality can differ from the intention. The Research
6
column in the middle shows how a research project seeks to measure all three levels of an
intervention using valid techniques.
The model highlights a number of quality assurance issues for the intervention process that need
to be addressed at the reporting stage of an intervention study:
• Was the planned intervention actually implemented?
• Did the intervention as implemented lead to the intended change in exposure?
• Did the actual change in exposure have the intended/predicted effect on the study
outcomes?
Figure 2.
Model of occupational intervention studies
Theory
Programme theory
Intended
intervention
Research
Practice
Measurement of
intervention
Actual intervention
Prevention
effectiveness
Intended changes
in
exposure/behaviour
Measurements of
changes in
exposure/behaviour
Actual changes in
exposure/behaviour
Theory of health
Aetiology of
diseases and other
Intended changes
in study outcomes
Measurements of
changes in
outcomes
Actual changes in
outcomes
To illustrate these issues, Kristensen (2005) distinguishes between “programme failure” and
“theory failure” (or success). In other words, it does not help if the intervention is effective if
the target recipient does not receive it (programme failure); nor is the recipient of an actual
intervention helped if the intervention has no effect (theory failure). His point is that the way to
distinguish between the two is to study all steps in the intervention programme.
Another important consideration is the possibility that the intervention model under
consideration is ambiguous — in other words, even if an intervention works, it may not be
known exactly what worked (Lipsey, 1996). It is the realisation that simple interventions are
often not effective in tackling MSDs, which has led to the examination of multi-component
interventions.
A review (Denis et al., 2008) considering whether the “traditional MSD prevention model” was
actually put into practice in workplace interventions found that only slightly more than a third of
studies reviewed actually applied this model. “Shortened” interventions skipped stages
involving work-description and risk factor identification, and “turnkey” interventions went
7
straight from a preliminary analysis to solution development and implementation. “Complete”
interventions were uncommon in changing environments, but it was rare for the intervention
process to be justified. They concluded that intervention processes require a better and more
detailed framework than currently exists, with a need for the application context, and their
advantages and limitations to be defined. They suggested that more complete intervention
processes are preferable when MSDs are the target of the intervention, but that the quick
“turnkey” interventions can be appropriate in well-understood situations.
A final possibility that must be considered is that an implemented intervention, even if
successful, can have negative side-effects or unintended consequences that undermine the
benefits of the intervention (Lipsey, 1996).
1.7
PURPOSE OF THE INTERVENTION
Kristensen (2005) draws a sharp distinction between aetiological intervention studies and
prevention effectiveness intervention studies. He argues that once enough is known about the
association between an exposure and a disease, there is no need to carry out aetiological studies,
and that prevention effectiveness studies are appropriate. Two examples he gives are 1) heavy
lifting and low back pain and 2) low decision latitude and absence from work. His
characterisation of the differences is set out in Table 1. This view is also expressed in one
earlier paper (Skov and Kristensen, 1996). However, as noted in Section 2.1, the consensus of
the others that have considered the need for MSD intervention studies is that the existing studies
have largely been of poor methodological quality and that further (aetiological) intervention
studies are needed to demonstrate the actual effectiveness of ergonomics interventions in
preventing MSDs. He is explicit that large samples are not needed for prevention effectiveness
studies, but that case studies in different settings are suitable instead. These will allow the
practicalities of implementing an intervention programme to be tested.
Table 2.
Characteristics of intervention studies (Kristensen, 2005)
Aetiological interventions
Prevention effectiveness interventions
Large samples
Small samples
Endpoint: health/disease
Endpoint: exposure, behaviour
Randomisation, blinding
No randomisation or blinding
Aetiological theory
Programme theory
Quantitative methods
Quantitative and qualitative methods
Representative groups and workplaces
Case studies
There is a case to be made for carrying out prevention effectiveness studies while aetiological
studies are ongoing rather than waiting until there is universal acceptance of the results of the
definitive aetiological studies. These could be used to help define the variables that should be
studied aetiologically since, if the actual implementation of a suggested intervention is difficult
or impossible, then there is little point in attempting an aetiological study of it.
This said, there is the severe danger that bodies such as HSE when asked to fund intervention
studies will decide on cost grounds to fund prevention effectiveness studies to the exclusion of
aetiological studies. This approach would fail completely to address the vital underlying
scientific issues that only aetiological studies can address and would therefore represent a waste
of resources akin to buying land and building materials for a house without checking whether
the site is on a flood plain and therefore totally unsuitable.
8
1.8 EPIDEMIOLOGY OF MSDS
Burton et al. (2005) summarised the basic epidemiology of non-specific low back pain (i.e. LBP
that cannot be attributed to specific pathologies or lesions) as follows:
• Lifetime prevalence
60-70%
• One year prevalence
15-45%
• Adult incidence per year
5%
• Peak prevalence
Ages 35-55
• Poor correlation between symptoms, pathology and radiological findings.
• In 85% of cases, low back pain cannot be attributed to pathology or neurological
encroachment.
• There is recent evidence of genetic liability to back pain.
• Between 2 and 7% of people with acute low back pain develop chronic low back pain.
• Two thirds of people with an episode of low back pain will suffer another episode
within the next 12 months.
• One third of people absent from work due to low back pain will have another low back
pain related work absence in the next 12 months.
The epidemiology of musculoskeletal problems with specific pathologies can be quite different
to non-specific problems. Prevalence rates in the general population can be quite high, and in
some chronic conditions, such as rheumatoid arthritis, will tend to increase with age.
1.9 RISK FACTORS (“FLAGS”) FOR THE DEVELOPMENT OF
CHRONICITY OF MSDS
Gatchel (2004) has summarised the various “warning flags” that have been described for use in
clinical evaluations of patients with acute episodes of low back pain. These are:
Red
Serious clinical pathology/physiological conditions
Yellow [Bio]Psychosocial risk factors that can increase the risk of an acute case of low back
pain becoming chronic. These are summarised as “ABCDEFW”:
• Attitudes and beliefs about pain — so-called “fear-avoidance”, passivity and
catastrophizing.
• Behaviours of the patient, particularly avoiding activity
• Compensation issues
• Diagnosis and treatment — misunderstanding of the severity and prognosis of their
condition
• Emotions — hopelessness
• Familial factors — particularly to do with social support in the home
• Work related risk factors — perceptions that their work is harmful and dangerous, job
dissatisfaction
Blue Occupational factors [psychosocial factors] believed by patients to impede their
recovery
9
• High demand/low control
• Negative perceptions of management
• Perceived time pressures
• Perceived poor social support
Black Objective occupational factors that may initially lead to the onset of low back pain and
may promote disability once the acute episode has occurred.
• National and local policies such as sickness systems, wage rates, the availability of
modified duties
• Working hours/shift patterns
• Physical factors such as biomechanical demands of the job
10
2
2.1
REVIEW OF EXISTING SCIENTIFIC LITERATURE
SYSTEMATIC REVIEWS
A number of wide-ranging reviews of the science regarding musculoskeletal disorders have
been completed, particularly in the USA. In 1997 one was published by the National Institute
for Occupational Safety and Health (NIOSH), part of the US Department of Health and Human
Services (DHHS) (Bernard, 1997). A report was published as a result of a workshop held by the
National Research Council (NRC) (1999). A subsequent report written jointly by the NRC and
Institute of Medicine (IOM) (2001) in response to a request from the US Congress, examined
more formally the overall patterns of evidence. One paper (Smith et al., 1999) presented at the
NRC workshop was the basis for a more comprehensive systematic review (Karsh et al., 2001).
Karsh (2006) has also summarised and attempted to synthesise the existing theories of the
causation of MSDs.
In the UK and European Union a similar review has been carried out for the Faculty of
Occupational Medicine (Waddell and Burton, 2001) as part of the development of guidelines for
managing low back pain in the workplace (Carter and Birrell, 2000). A more recent review
underpins European guidelines on prevention of low back pain (Burton et al., 2004; 2005).
A series of systematic reviews of treatment methods/secondary prevention methods for back
pain have been published in Spine by the Cochrane Collaboration Back Review Group (Clarke
et al., 2006; Furlan et al., 2002; 2005; Hagen et al., 2000; 2002; 2005; Heymans et al., 2005;
Karjalainen et al., 2001a; 2001b; Niemisto et al., 2003; van Tulder et al., 1999; 2000a; 2000b;
2003b). The Cochrane Collaboration originated with a specific focus on clinical issues and
therefore largely does not seek to address primary prevention (Bouter et al., 2003). A single
study (Jellema et al., 2001) has been found that addresses both primary prevention and
secondary treatment through the use of lumbar supports in seating.
Recently the focus of the Cochrane Back Group seems to have broadened to include primary
prevention where it overlaps with secondary prevention. Thus there is a review of manual
handling advice and assistive devices (Martimo et al., 2007) and there are current protocols for
reviews to be carried out looking at exercise (Choi et al., 2007) and work conditioning
(Schonstein et al., 2003). A protocol for worksite intervention (Aas et al., 2005) is specific to
secondary interventions.
The Cochrane Library (http://www.cochrane.org) contains The Cochrane Central Register of
Controlled Trials that holds details of relevant Randomised Controlled Trials (RCTs) and
Controlled Clinical Trials (CCTs). As of 24 April 2007 there were 1788 records out of a total of
495002 in the database that were identified by the search term SR-BACK, which is used to
identify studies relevant to low back pain (Bouter et al., 2003).
Many more focussed reviews of intervention studies have been carried out (Bongers et al.,
2002; Boocock et al., 2007; Bos et al., 2006; Brewer et al., 2006; Hignett, 2003; Hooftman et
al., 2004; Jellema et al., 2001; Silverstein and Clark, 2004; Sobeih et al., 2006; Tveito et al.,
2004; van Poppel et al., 2004; Village et al., 2005). While there does not appear to be a formal
group systematically reviewing intervention studies seeking to prevent new episodes of MSDs,
many of these reviews have used the Cochrane criteria for systematic reviews (van Tulder et al.,
2003a). The consensus is that many of the studies reviewed are methodologically weak,
especially in their reporting, and that the available evidence is often limited and inconclusive.
11
Tuncel et al. (2006a) carried out a meta-analysis of international studies attempting to prevent
the occurrence/reoccurrence of lower back disorders in manufacturing workplaces. The
inclusion criteria were that:
• The study had to employ a controlled workplace intervention in a manufacturing setting
and had to be aimed at reducing the (re)occurrence of Low Back Disorders (LBDs).
• Participants had to be employees in a manufacturing setting.
• The reported outcome measures had to include one or more of self-reported back pain
episodes, pain intensity, or company records of sick leave due to LBDs.
• The study had to be published in English in a full journal article between January 1965
and July 2004.
• The data had to report data that allowed effect sizes to be calculated.
The meta-odds ratio (OR) they obtained showed an insignificant reduction in LBDs. Their
appraisal showed that the four studies that met their inclusion criteria had at best a marginal
methodological quality and therefore they cautioned against interpreting the low OR they
obtained as evidence of no effect. Their primary conclusion was that further research was
required and that it needed to be rigorous and high quality.
2.2
SPECIFIC MSD RISK FACTORS
A number of risk factors have been associated with a risk of low back pain. The most widely
quoted one is hard physical work but genetic predisposition, co-morbidity, possible weak spinal
structures and weak psychological stamina have also been reported (Leboeuf-Yde, 2004).
However, the evidence does not exist for a sedentary life style, smoking, obesity and alcohol
consumption (Leboeuf-Yde, 2004).
Heritability analysis among twin pairs has shown that life-time prevalence of LBP has a genetic
component (Leboeuf-Yde, 2004). Hestbaek et al. (2004) discuss the “heritability of liability” to
low back pain, but it is not clear how a liable genotype is modified by environmental factors to
produce a phenotypic case of low back pain. Nor is it clear how selection/screening could be
used to help liable individuals avoid the relevant environmental factors.
Individual/demographic factors can be seen as “non-workplace factors” that contribute
alongside work factors to the causation of MSDs. They can also be seen as physiological or
psychological attributes that affect personal responses to workplace factors (Cole and Rivilis,
2004). It has been noted that few individual factors are readily modifiable, especially in the
workplace and therefore there is little use in including them when planning workplace
interventions (Frank et al., 1996). The individual factors identified by Cole and Rivilis are
shown in Table 3. They note that gender and socio-economic status can affect both reporting
and recognising cases of MSDs. They also commented that the impact of combined
interventions can be substantially modified by individual factors such as differential
responsiveness to secondary interventions. However, there are potentially complex interactions
between individual factors and other workplace variables, making separating the contributions
of different factors very complex.
12
Table 3.
Individual factors relevant to MSDs (Cole and Rivilis, 2004)
Usual naming of factor types
Individual factors
Potential construct(s)
Demographic
Gender
Differential responses to stress
Differential labour market
Age
Cumulative exposure
Decreased tolerance
Different skills and experience
Work
Work-style
Different biomechanical exposures
Anthropometry
Height and weight
Mismatch between equipment and
person
Differential tissue demands
Psychological
Personality
Differential kinematics
Differential coping capacity
Lifestyle
Physical activity, hobbies,
sports
Smoking, drugs
Additional loads or physical exposures
Additional exposures
Comorbidity
Diabetes, pregnancy
Distress, depression
Additional internal exposures
Altered biochemistry, different pain
perception threshold
Past history
History of MSD episodes
Lower tolerance
Social
Divorce-widowed
Minority race
Poverty
Lower social support
Discrimination
Complex socio-health contexts
An exploration of the pathways between physical and psychosocial risk factors (Swanson and
Sauter, 2006) used an intervention that provided an alternative keyboard in an office
environment. Their model is based on changes in office technology affecting the physical
demands of the job and the way the work is organised with consequent changes in both
psychological and biomechanical strain. Their results showed only small effects of the
intervention. The relatively small number of subjects involved and the relatively small
intervention may have been a cause of this.
2.3
GUIDELINES
The evidence review (Waddell and Burton, 2001) underlying the Faculty of Occupational
Medicine (FOM) guidelines on the management of low back pain at work (Carter and Birrell,
2000) concluded that:
• Physical demands of work can precipitate individual attacks of low back pain.
• Certain individuals may be more susceptible.
• Certain jobs may be higher risk.
• Overall, physical demands of work account for only a modest proportion of the total
impact of LBP occurring in workers.
Their specific evidence statements contained the following:
13
“Most adults (60-80%) experience LBP at some time, and it is often persistent or
recurrent.” [Strong evidence]
“There is strong epidemiological evidence that physical demands of work (manual
materials handling, lifting, bending, twisting and whole body vibration) can be associated
with increased reports of back symptoms, aggravation of symptoms and ‘injuries’.”
[Strong evidence]
“There is limited and contradictory evidence that the length of exposure to physical
stressors at work (cumulative risk) increases reports of back symptoms or of persistent
symptoms.” [Limited or contradictory evidence]
“There is strong evidence that physical demands of work (manual materials handling,
lifting, bending, twisting and whole body vibration) are a risk factor for the incidence
(onset) of LBP, but overall it appears that the size of the effect is less than that of other
individual, non-occupational and unidentified factors”. [Strong evidence]
“There is strong epidemiological and clinical evidence that care seeking and disability
due to LBP depend more on complex individual and work-related psychosocial factors
than on clinical features or physical demands of work”. [Strong evidence]
When discussing prevention they acknowledged that it is reasonable in principle to seek to
reduce the incidence and prevalence of LBP by interventions designed to reduce exposure to
known occupational risk factors. However, because of the complex set of causal factors for
LBP, of which occupational physical demands are only one, they queried the extent to which
occupational interventions can realistically be expected to reduce the societal impact of LBP.
They considered that
“There is a lack of convincing evidence that it is possible to reduce the incidence or
prevalence of the symptom of LBP substantially.”
The European guidelines (Burton et al., 2004; 2005) on prevention of low back pain took the
same approach of arguing that there is limited scope for preventing its incidence (first time
onset). They therefore noted that primary causative mechanisms remain largely undetermined
and that risk factor modification will not necessarily achieve prevention. They focused
therefore on the prevention of the consequences of LBP through reduction of the impact of
recurrences, care seeking, and disability and work loss.
Their overarching comment was that there was acceptable evidence that the prevention of
various consequences of LBP (e.g. recurrence, care seeking, disability and work loss) is feasible
but that the effect sizes of the interventions are rather modest. They concluded that the most
promising approaches were physical activity/exercise and biopsychosocial education.
In the context of workers, their recommendations were:
• To encourage physical exercise for prevention of incidence and recurrence of LBP and
for prevention of recurrence of sick leave due to LBP;
• Not to use back schools based on biomedical/biomechanical information;
• Not to provide lumbar supports or back belts;
• Not to provide shoe inserts/orthoses;
• To provide temporary modified work and ergonomics workplace adaptations to
facilitate earlier return to work for workers on sick-leave due to LBP.
14
They found that multi-dimensional interventions at the workplace might be recommended to
reduce some aspects of LBP but they did not find sufficient evidence to recommend specific
dimensions and how they should be balanced.
They did not find sufficient consistent evidence in the following areas to make
recommendations:
• Standalone physical ergonomics interventions to prevent LBP;
• Standalone organisational interventions;
• The content of organisational/participative interventions associated with a physical
ergonomics programme.
2.4
THE VALUE OF MECHANICAL INTERVENTION IN PREVENTION OF
MSDS
While these guidelines offer a pessimistic view of the possible effectiveness of ergonomics
interventions, they do make an urgent call for “good quality RCTs” to investigate them further.
A less despairing view of mechanical issues and back pain is espoused by McGill (2002). In a
section in his book entitled “Deficiencies in Current Low Back Disorder Diagnostic Practices”
he writes:
“It is currently popular for many authorities to suggest that back trouble is not a medical
condition. They assert that physical loading has little to do with low back injury
compensation claims; rather they believe workers complain of back problems in order to
benefit from overly generous compensation packages or to convince physicians they are
sick. According to this view, any biomechanically based injury prevention or
rehabilitation program is useless. Variables within the psychosocial sphere dominate
any biological or mechanical variable. If this is true, then this book is of no value—it
should be about psychosocial intervention.”
He is hardly gentle when he comments that:
“The position that biomechanics plays no role in back health and activity tolerance can
be held only by those who have never performed physical labor and have not experienced
first hand the work methods that must be employed to avoid disabling injury. While the
scientific evidence is absolutely necessary, it will only confirm the obvious to those who
have this experience.”
While he does not discuss the epidemiological literature with the approach usually taken in
systematic reviews or meta-analyses, McGill (2002) provides an overview of the
epidemiological literature and highlights studies that support his contention that both
psychosocial and biomechanical factors are important risk factors for LBP. He then provides an
extensive discussion of the anatomy and normal and injury mechanics of the lumbar spine.
Finally, he discusses, in detail, firstly risk reduction guidelines aimed at reducing the
overloading stressors that cause occupational LBP and, secondly, rehabilitation and exercise
programs.
In a subsequent article (McGill, 2004) he lists three new recommendations for workers to
reduce the risk of injury as:
• Avoid repeated full-flexion of the low back;
• Avoid long-duration flexion postures;
• Avoid flexion under acute loads immediately after getting out of bed.
15
2.5
EFFECTIVENESS OF DIFFERENT INTERVENTIONS
Karsh et al. (2001) carried out a comparison of the intervention effectiveness of various study
designs and intervention types from 101 studies. The findings are summarised in Table 4. They
noted that 84% of all the studies had some positive effect, though the majority (55%) had mixed
results. Where mixed results occurred, the vast majority were a combination of mostly positive
results with some non-effects. Only rarely were the results of the interventions actually
negative. However, this could be publication bias or pre-selection of interventions likely to be
effective rather than experimental error or interpretation bias (halo effect). However, only 32%
of studies reviewed used an experimental or quasi-experimental design.
They concluded, with a “qualified ‘yes’” that the review had shown that interventions to control
WMSDs are effective. Restricting the analysis to randomised designs reduced the evidence for
back belts and training and removed it for new tools and technologies. The evidence for exercise
and multiple component interventions was strengthened.
2.6
COST EFFECTIVENESS OF INTERVENTIONS
A study (Lahiri et al., 2005) of the cost-effectiveness of the interventions that have been
attempted estimated that back pain/injury incidence could be reduced by 20% with training, by
56% with engineering/administrative controls, by 60% with a combination of
engineering/administrative controls and training and by 74% by a comprehensive Workplace
Ergonomics Program. They considered cost effectiveness of these programs taking into account
worldwide differences in levels of industrialization. Their findings suggested that full
ergonomics programs would be cost-effective in both developed and developing countries for
their health effects alone. However, they did conclude that training appeared to be the most
cost-effective intervention, despite the impact of training on health outcome being rather
limited. While they were more expensive, engineering and ergonomics interventions had a far
greater impact on total health outcome than training due to the greater reduction in back pain
incidence.
They recommended prospective studies of the recurrence of back pain and studies of workers
who are the “working hurt”. They believed that worker training is a low-cost feasible first step
towards the reduction of work-related back pain in developing countries where resources are
scarce and that it should be encouraged through public policy and regulation. However they
considered it unquestionable that ergonomics programs should be encouraged in highly
developed countries for both health and productivity effects and that when additional resources
become available they should go straight to the full ergonomics programs.
One of the drivers of the current interest in musculoskeletal disorders is the concern about the
economic costs associated with them, particularly in relation to health care, income replacement
by state or insurance benefits, and litigation/compensation costs. The hope is that interventions
will be widely adopted if they can be shown to be effective, either at preventing MSDs or at
getting an individual back into productive paid work more quickly. Though the topic appears to
be relatively unexplored in the MSD field, an economic evaluation will be part of a full
consideration of an intervention, and is best done by comparing two or more interventions
(Korthals-de Bos et al., 2004).
16
1
(2%)
14
(14%)
19
(40%)
29
(29%)
Multiple
component
Total
Job design
1
(100%)
56
(55%)
27
(57%)
8
(57%)
4
(29%)
Exercise
5
(50%)
Tools/technolog 4
ies
(40%)
2
(14%)
14
(67%)
6
(29%)
2
(25%)
Back belt
Training
2
(25%)
4
(50%)
Positive
Number of
studies
(percentage of
studies)
Mixed
No effect
Results
20
(20%)
2
(4%)
7
(50%)
17
12
(12%)
3
(6%)
1
(7%)
1
(10%)
49
(49%)
29
(62%)
1
(100%)
5
(36%)
7
(70%)
3
(3%)
2
(4%)
1
(10%)
5
(5%)
4
(9%)
11
(11%)
7
(15%)
1
(7%)
1
(5%)
5
(24%)
7
(33%)
1
(5%)
2
(25%)
2
(25%)
4
(50%)
7
(33%)
Other
Pre-post one Post only
Post only,
Experimental Quasiexperimental group
nonone group
equivalent
comparison
Study design
Summary of intervention effectiveness (Karsh et al., 2001)
Intervention
Table 4.
101
47
1
14
10
21
8
Total
2.7
MULTI-COMPONENT INTERVENTIONS
Multi-component interventions (also referred to as multi-disciplinary interventions (Tveito et
al., 2004)) are ones that apply a range of different measures to try to prevent low back pain
and/or its consequences. This makes it easy for them to be applied at more than one level of
prevention. Thus, one study (IJzelenberg et al., 2007) applied the biopsychosocial model
through education and training (primary prevention) and through provision of immediate
treatment along with ergonomics advice to LBP cases (secondary prevention).
A multi-component study (Mancini et al., 2005) showed that this kind of intervention was
effective, in this case, in reducing the incidence of work-related eye injuries. Because it was a
“reactive” rather than a “proactive” study they argued that a randomised study would have been
ethically inappropriate. Therefore, they compared the study population with other industry
sectors in the region.
A systematic review (Tveito et al., 2004) concluded that there was limited evidence of the effect
of multidisciplinary interventions on pain outcomes, but that there was no evidence of an effect
on episodes of LBP.
Recent recommendations regarding ergonomic intervention studies have been that multicomponent study designs should be favoured (Silverstein and Clark, 2004; Waters, 2004a). As
Smith et al. (1999) explain, the primary purpose of workplace interventions to control MSDs is
to reduce the stress load to eliminate strain. This can be coupled with increasing the capacity of
the individual to handle greater loads, thereby reducing the possibility for a misfit.
It has been recommended that epidemiological studies be conducted to evaluate the interactive
effects of various risk factors, such as physical and psychosocial stressors, individual and
genetic factors (Hartvigsen et al., 2004; Leboeuf-Yde, 2004), and other factors that may affect
reporting of MSDs (Waters, 2004a). Karsh et al. (2001) reported that 97% of the multiple
component interventions in the studies they reviewed produced some positive results, with the
most effective design being randomised assignment and control groups. Such a study needs to
show that each of the multiple components was successful in meeting its goal, not just that the
whole intervention was successful. This should involve examining intermediate outcomes to
demonstrate that each component was successfully implemented. It must be noted that
considerations of statistical power become more complex for such studies. In the study
(IJzelenberg et al., 2007) that used the biopsychosocial model, incident cases were offered
workplace advice/ergonomic adjustments or further training. The power calculations reported
are for the primary intervention and are based on detectable changes in prevalence rates; they
are not based on utilisation of health care or provision of further advice. The study found that
only 10 subjects utilised the workplace health care while 66 utilised similar care outside the
workplace. Of the 10 subjects utilising the health care, only four accepted the further advice.
Given these tiny sample sizes, there was no chance of the secondary interventions having
sufficient power to demonstrate their effectiveness.
2.8
RETURN TO WORK POLICIES
A “Prevention and Early Active Return-to-Work Safely” (PEARS) program (Badii et al., 2006;
Davis et al., 2004) was a combined primary prevention and return to work intervention. It was
found to be effective at returning to work more quickly employees that had reported
musculoskeletal injuries. Consequently, it showed significant reductions in total days lost and
financial costs. Badii et al. (2006) found an associated increase in the overall incidence of
musculoskeletal injuries and the subset resulting in time-loss. They interpreted this as a shift in
reporting culture specific to musculoskeletal injuries. This raises issues of the difficulty of
18
separating different types of intervention. Their finding of a change in reporting culture with an
increase in musculoskeletal injury reporting and in short duration absences (1-2 days) associated
with an overall decrease in lost time and consequent costs led them to suggest that allowing
injured workers to take short periods of time off might be associated with reduced morbidity
and costs in the long run. Any study therefore that measures lost-time without any kind of
measure of the incidence and prevalence of problems among participants that do not lead to
lost-time is in danger of being confounded by such a change in reporting and short duration
absence culture. The study would need to be carefully designed to control for different
reporting systems and cultures, and different absence management expectations and milieus.
2.9
RECENT PROSPECTIVE STUDIES
A prospective study (van Nieuwenhuyse et al., 2006) of young workers investigated the effect
of work-related factors and individual characteristics on the incidence of LBP in initially painfree individuals working in health care and distribution companies. After one year of follow-up
12.6% of 716 individuals had experienced back pain lasting seven or more days. They found
univariate evidence of a dose-response relationship for pushing and pulling heavy loads but not
for lifting or carrying weights. None of the psychosocial factors they measured was predictive.
Multivariate analysis using Cox regression showed that inability to change posture regularly,
working with the trunk bent or twisted for more than two hours per day, back pain in the
previous year, and pain related fear were significant predictors. They concluded that a more
effective primary prevention of LBP might be achieved by addressing both the ergonomic work
environment and attitudes to pain. They concluded that as a potential preventative measure,
addressing pain-related fear seems as promising as physical work factors. However, as they do
not discuss this conclusion in terms of a model such as that in Figure 1, it is difficult to know
how such an intervention would act as primary prevention rather than secondary or tertiary.
A longitudinal study (Gerr et al., 2005) involved an RCT of a postural intervention among
newly recruited computer users. It showed that the specific workplace postural interventions
used were unlikely to reduce the risk of upper extremity musculoskeletal symptoms among
computer users. They had the problem of relatively low compliance with all aspects of the
intervention due to the inflexibility of workplace configurations
2.10
RECOMMENDATIONS FOR FUTURE WORK
The NRC/IOM (2001) study identified important gaps in the science base on MSDs and
recommended that workplace intervention studies should be carried out directed towards:
• “Conducting rigorous evaluations of workplace interventions including but not limited
to randomized controlled trials or other scientifically valid approaches.
• “Promoting investigation of multi-factorial interventions.
• “Developing effective methods to measure the efficacy and cost-effectiveness of
interventions on the reduction of workplace injuries.
• “Coordinating studies of interventions between the research community and industry.
• “Validating techniques, standards, and manuals for target industries.”
The National Occupational Research Agenda (NORA) of NIOSH in the USA identified
priorities for intervention research (Waters, 2004a) to evaluate the effects of a number of factors
on the development and prevention of MSDs:
•
“Alternative (product and/or tool) design criteria (force, spatial requirements of work);
19
• “Optimization of mechanical (force, movement and posture) work demands and
temporal patterns of exposure;
• “Manual handling alternatives in posture, movement, force, productivity and quality;
• “Ergonomic training and education;
• “Costs and benefits of ergonomics interventions; and
• “Job assignment, selection and choice.”
The consensus of the many systematic reviews referred to in Section 2.1 is that because of the
low methodological quality of the many studies to date, there is still a significant need for
methodologically high-quality intervention studies to be carried out.
The European guidelines on prevention of LBP (Burton et al., 2004; 2005) recommended
further research to address (among others) the following issues:
• Cost-benefit and risk-benefit analysis;
• Good quality Randomised Controlled Trials (RCTs) to study the effectiveness of daily
physical activity for prevention of LBP and recurrence of LBP;
• Good quality RCTs on the role of information orientated to reducing fear avoidance
beliefs and improving coping strategies;
• “Good quality RCTs are urgently needed to study the effectiveness of physical,
psychosocial and organisation ergonomic interventions on a large variety of outcomes,
ranging from prevention of (recurrence of) LBP and prevention of (recurrence of) sick
leave due to LBP up to compensable LBP.”
• Whether interventions can be applied to all workers, irrespective of gender, age,
seniority and/or past history of LBP. If necessary, the optimal approach for each sub
group should be examined.
20
3
3.1
EPIDEMIOLOGICAL METHODS
INTRODUCTION
The ideal process of conducting an epidemiological research study is illustrated in Figure 3. It
is important to be aware of the issues related to these stages. The following sections discuss
these in further detail.
Figure 3.
General sequence of steps in a research project (Altman, 1991).
PLANNING
DESIGN
DATA COLLECTION
DATA PROCESSING
DATA ANALYSIS
PRESENTATION
of summary statistics
INTERPRETATION
PUBLICATION
Guidance had been provided by NIOSH on evaluating the effectiveness of
interventions/strategies to prevent work injuries (Robson et al., 2001). This is written at an
introductory level to encourage students, researchers and practitioners to become involved in
designing and evaluating work place safety interventions. This guidance is generic to all safety
interventions and not specific to ergonomics interventions.
21
3.2
PLANNING
A successful study will have been planned well. The planning includes:
• Deciding the research question;
• Choosing a suitable study design to ensure that the aims of the study are met. For an
intervention study, this means ensuring that the design is able to assess causality;
• Determining study sample size using power considerations;
• Taking great care in deciding what information will be collected;
• Taking into consideration confounding and bias;
• Realistically planning study management and financing;
• Getting ethical approval.
The MRC (2000; Campbell et al., 2007) has produced a discussion document setting out a
possible stepwise framework for developing and evaluating RCTs for complex interventions.
They are explicit that though they draw a comparison with the evaluation of new drugs, they “in
no way intend to imply that the evaluation of a complex package is like the development and
evaluation of a new drug”. Their framework is advice to be applied “to the extent to which it is
relevant” at each stage of the proposed intervention evaluation. The five steps they identify are:
• Theory — why should this intervention work?
• Modelling — How could this intervention work in practice?
• Exploratory trial — Pilot trials to choose optimised study designs;
• Definitive RCT — The central step;
• Long-term implementation — possibly supported by an observational study.
From reviews and theoretical papers such as by Volinn (1999), it is clear that the ergonomics of
MSDs field has reached the stage where definitive RCTs should be attempted though earlier
stages should be considered in the planning of such an RCT.
3.2.1
Research question
The study must be designed around the research question and this must be defined extremely
carefully. It is possible that “an omnibus hypothesis on the effect of an intervention on
musculoskeletal outcomes may in fact be testing something quite different and any inference
drawn misleading” (Dempsey, 2007). It is therefore essential to be clear as to what is being
investigated and what the measured outcomes will be. Dempsey (2007) goes as far as to
suggest that the research question may need to be constrained to one that is answerable, though
this does have the disadvantage that the outcome may be a surrogate measure of effectiveness.
He gives the example of demonstrating reduced mechanical exposure, rather than attempting to
demonstrate reduced morbidity due to decreased mechanical exposure. Unfortunately, until
there is no questioning of the link between mechanical exposure and morbidity, this can only be
described as begging the question and is not a suitable design for demonstrating the
effectiveness of ergonomics interventions. Dempsey (2007) also expresses reticence over the
call for more intervention research in ergonomics for three reasons:
• In many situations, the question is statistically intractable.
• The complex and partially incontrollable situations in which interventions must be
implemented weaken and confuse the studies.
22
• Alternative measures to morbidity such as productivity gains have the potential to
backfire and to lead to further changes to the job, such as increases in work pace that
can be detrimental to employees.
Dempsey considers the kind of evidence accepted in quality engineering where small
incremental changes are routinely implemented and argues that no more should be expected
from ergonomists. The weakness of his argument is that MSDs are a health problem as well as
a work design problem and therefore medical epidemiologists have an interest in the topic. They
are unlikely to be prepared to abandon the sophisticated epidemiological methods they have
developed in clinical trials, particularly of drugs, where many of the same problems are found.
Moreover, there is much to be gained by a consideration of the way that clinical trials are
demanded before new, “alternative” or “complementary” treatments become accepted as part of
standard medicine. The vested interests, emotional commitment and junk science involved in
unproven treatments are all factors that can be seen in the field of ergonomics interventions.
Karsh et al. (2001) noted that “Randomized experimental designs are exceedingly difficult to
carry out in the field”, and of the 47 multiple component interventions they reviewed, only two
used experimental designs. They also reported that of the 101 studies reviewed, only two
reported having conducted a power analysis.
3.2.2
Inferring/detecting causal relationships
Investigating the prevention of MSDs with an intervention study requires detecting causal
relationships. The first stage of investigating causality is to establish an association, usually
with an observational study, and then to consider what the particular association appears to
imply. However, establishing an association, although necessary, is not a sufficient condition to
establish causation. Guidelines have been written on how to evaluate the evidence of a causal
relationship between two associated variables (Hill, 1963) and the major criteria are:
• Temporal relationship;
• Biological plausibility;
• Consistency and alternative explanations (confounding);
• Dose-response relationship;
• Strength of the association;
• Cessation of effects.
To decipher the difference between association and causation requires very careful statistical
analysis. Interpretation from panel studies (longitudinal cross-sectional studies) can often infer
causation, when in fact it is more likely to be association. However, any two quantities changing
over time will show a statistical association. Thus, only in randomised trials and other
experiments can we reasonably describe an observed effect as caused by the preceding change,
because of the controlled nature of the investigation (Altman, 1991). Even with randomised
trial data there must be an assumption that no bias occurred in allocation and compliance, or that
any such bias can be handled by adjustment procedures as such assumptions are not always
correct (Greenland et al., 1999).
Causal pathways can be either direct or indirect and causal diagrams can be used to illustrate
these. On any individual pathway, the relationship can be an accumulation of risk (independent
or clustering) or chain of risks (additive or trigger). A number of methods have been developed
to describe these cause and effect relationships and include Directed Acyclic Graph (DAGs)
(Pearl, 2000a; Hernan et al., 2004) and Influence Networks/Influence Diagrams. Using such
23
techniques allows the study designer to ask which variables need to be controlled to estimate the
causal effect of the intervention on the outcome. Therefore, the study design is more likely to
identify and quantify the causal effects. Causal diagrams can reveal unnoticed shortcomings of
those criteria when used in considering multiple potential confounders (Greenland et al., 1999).
However, it appears that such techniques have not yet been applied to MSDs and additional
work would be needed to develop existing models (Karsh, 2006) to use these techniques.
All studies that assess causation include time in the design and are thus longitudinal studies,
either prospective or retrospective. Prospective studies usually follow individuals forwards
from some point in time, whereas retrospective studies, select individuals and factors that have
occurred in the past.
3.2.3
Determining study sample size and power
As part of the proposal for any epidemiological study, there will be a sample size calculation
identifying how many people need to be included for the study to have sufficient power to
answer the questions posed. The absence of such a calculation can be considered unethical,
since the study then carries a considerable risk of failing to demonstrate a treatment difference
when one is really present (i.e., Type II error) (Pocock, 1983). The inclusion of too many
people in the study is also considered unethical, especially in a clinical trial where there are
considerations of the risks to the well being of the subjects, as well as the unnecessary use of
medical and financial resources. The study should be representative of the population to which
the results will be generalised and calculating the sample size is one part of achieving this.
The magnitude of the intervention is a crucial consideration since, if a genuine dose-response
relationship exists, the dose of the intervention has to be sufficiently large for the effect on the
response variable to be detectable and this needs to be considered in the power calculations that
are used to determine the target sample size. Sample size calculations should be based on the
principles of hypothesis testing and should state:
•
The proposed analysis methods;
•
The size of the increased risk that it is desired to detect, i.e. the effect size;
•
The chosen significance level (or P-value);
•
The probability of achieving this level of significance (power);
•
Standard deviation of the variable (in each group);
•
The follow-up period required for the target significance level and power;
•
Cluster size;
•
Number of clusters;
•
Projected initial participation rates;
•
Projected drop out rates.
The type of power (sample size) calculation used will depend on the precise study design
selected so detailed estimates are not appropriate and are not provided in this report.
The study protocol should define which individuals are to be recruited (inclusion and exclusion
criteria), intervention schedules, data to be collected, analysis methods, contingency plans for
foreseeable problems, and study personnel.
Account must be taken of potential difficulties in finding appropriate individuals prepared to
participate. Given the likely need to use a clustered design in an intervention study, this will
24
start with the need to find workplaces containing clusters of individuals in jobs that are suitable
targets for interventions.
The calculated sample size must take into account the inevitable loss of data because of nonresponding and loss of subjects to follow-up. Further adjustments should be made if the final
analysis is to be adjusted for the effect of confounding variables, if the examination of subgroup
effects is planned (Kirkwood and Sterne, 2006), or if there will be testing of multiple dependent
variables or of interactions between risk factors.
Power calculations for complicated trials including sequential and clustered trials involve
complex statistical methods, for example, CRTs have two components of variation (within
cluster and between cluster). These sources of variation should be estimated separately and
both must be taken into account when calculating sample size for CRTs. This can be done
through using the intra-cluster correlation coefficient (ICC) (IJzelenberg et al., 2007) or the
between-clusters coefficient of variation (CV) (Hayes and Bennett, 1999; Medical Research
Council, 2002; Ukoumunne et al., 1999). Generally, increasing the number of clusters offers
more increase in power than increasing the number of individuals per cluster.
Because of the complexity of power calculations and the ease with which mistakes can be made,
all study design decisions and power calculations will need to involve a statistician. A
recommended software package for performing power calculations is PASS 2005 (Hintze,
2005). It is essential to carry out sample size calculations for several different scenarios, so that
the project team can make an educated decision on the study design, weighing up the logistics
and costs (Kerry and Bland, 1998). An iterative approach will allow testing of a number of
different designs to evaluate the most suitable design, and simulation may be used as part of this
process (Hopkins, 2000).
Non-technical considerations will also need to be taken into account in determining target
sample size (Rothman and Greenland, 1998). In effect, this is a cost-benefit analysis of the
trade-off between increased precision and the cost of the increased sample size. The greater
precision has a value to the beneficiaries of the research but this value is ultimately
indeterminate as the number of beneficiaries is always uncertain. In addition, many potential
benefits involve social, political and biological factors that are almost never quantified.
“Consequently, only informal guesses as to a cost-efficient size for a epidemiologic study are
feasible” (Rothman and Greenland, 1998) and it must take into account unquantified practical
constraints and the practical implications of the study size.
A different and pragmatic approach to sample size calculations termed “sample size on the fly”
has been suggested (Hopkins, 2000). This depends on defining a target confidence interval for
the variable of interest and continuing recruiting subjects until the size of the confidence interval
is reduced to the target value. Hopkins’ argument is that for non-null effects this allows much
smaller sample sizes to be used. While he gives examples of how this approach can be used in
some longitudinal designs, such as pre-post studies, he does not indicate that the approach has
been extended to methods such as Cox regression (Proportional Hazards Models, PHMs) that
are used to evaluate time-to-event data. Any proposal to take such an approach should therefore
be evaluated very carefully to test the robustness of the proposed design.
The following give some indications of the scale of studies that may be needed:
• Lipsey (1996) noted that: “When the number of respondents in a study is less than 5001000 per group (e.g., treatment and control) sampling error can easily be large enough
to obscure meaningful effects”.
25
• Kraus et al. (1997) discussed an example sample size calculation for a hypothetical
back pain cohort study comparing exposed and non-exposed groups and recommended
a conservative estimate of at least 1500 subjects.
• Zwerling et al. (1997) cited a personal communication of a calculation that reduction of
injuries by 25% would take 6 or 7 years follow-up of 3800 workers to achieve power of
80% at 5% significance. They noted that measurement of injury rates “generally
requires very large sample or long follow-up times, or both” and that “few companies
have enough employees to even enter into a trial”.
• IJzelenberg et al. (2007) reported that the design of their CRT was based on a sample
size calculation using an ICC of 0.05, an average of 20 workers per cluster, an initial
participation of 75% and a loss to follow-up of 30%. This was anticipated to be able to
detect a difference of 10% in prevalence between the intervention and control groups
(power 80%, one sided significance of 5%) with 350 workers in nine intervention
clusters. (It appears that the figure of 20 workers per cluster relates to those completing
the study, not the number employed in the workplace).
3.2.4
Controlling confounding and bias
Occupational studies often need complex statistical methods to take account of confounding and
bias. Confounding is not a source of error in an intervention study, but rather a phenomenon
that must be understood. However, failure to take confounding into account in interpreting the
results of a study can lead to errors and can bias the conclusions of the study (Gordis, 2000).
The definition of confounding is that there are alternative explanations for an observed
association between a risk factor and a health outcome, making it difficult to assess the effect of
each risk factor on the outcome variable. Furthermore, most occupations involve exposure to
more than one potential risk factor and the possibility of confounding by other occupational
exposures must be considered in the context of each study (Checkoway et al., 2004). Many
observational epidemiological studies assume a true association has been observed and might
derive a causal inference when, in fact, the relationship may not be causal, but rather a result of
confounding by a third factor that is both a risk factor and associated with the exposure in
question. Therefore, as most causal questions involve the relationships of multiple exposures,
confounding and interaction characterise virtually every situation in which aetiology is
addressed (Gordis, 2000). If confounding cannot be avoided at the design stage of a study, then
disentangling the causal links is often difficult and requires more complex statistical methods
(Mullner et al., 2002). However, if the relevant variables are measured, confounding can be
addressed.
Bias, on the other hand, is a result of an error in the way the study has been carried out and is
defined as “any systematic error in the design, conduct or analysis of a study that results in a
mistaken estimate of an exposure’s effect on the risk of diseases” (Schlesselman, 1982). This
can result in either an overestimate or an underestimate of the difference between interventions.
Possible biases are numerous and need to be assessed and, if possible, eliminated as they affect
the validity of the findings (Gordis, 2000). Sackett (1979) identified 35 different biases that can
occur in case-control studies. Other study designs will have biases typical to that design, for
example, volunteer bias is more common in cross-sectional studies (i.e. questionnaire nonresponders are usually less-healthy) whereas case-control studies are susceptible to recall bias
(Altman, 1991). Bias can also be due to systematic differences between individuals
There are also issues of recruitment and differential attrition biases. Inevitably, there will be
individuals in any workplace who choose not to participate in or cooperate with the study and
this may be a source of recruitment bias. Differential attrition is similar, but results in
26
individuals dropping out of the longitudinal phase of a study due to uncontrolled differences
between clusters that are not relevant to the study.
3.2.5
Randomisation and alternatives
Random allocation can be used to prevent bias or judgement in the selection of individuals to
participate in the different interventions (Altman and Bland, 1999). The technical meaning of
randomisation is for each individual to have a known chance, usually an equal chance, of being
given each intervention, and the intervention to be given cannot be predicted. There are
different ways of conducting randomisation including stratified randomisation, cluster
randomisation and block (or restricted) randomisation, simple randomisation and random
permuted blocks. Block randomisation ensures that similar numbers of subjects in each group
stay close, and stratified randomisation keeps the groups balanced for certain prognostic subject
characteristics (Altman and Bland, 1999). Cluster randomisation is used for numerous reasons,
including administrative convenience for investigators, improving subject compliance, avoiding
contamination of the control group by the intervention group or cross-talk between groups
(Karsh et al., 2001) and because workplace interventions are naturally applied at the cluster
level (Donner and Klar, 2004). Unequal randomisation, although relatively uncommon, is
useful in small intervention studies where there is no prior information on the efficacy of a new
intervention (Pocock, 1983).
In occupational epidemiology, multi-centre studies are often used to aid recruitment, increase
geographical spread and increase the number of staff available to run the study. Balancing for
institutions is a relevant method of randomising and stratification by institutions entering
subjects should be considered since different institutions can show very different response rates
for their subjects for reasons of participant selection and experimental environment. For further
discussion on stratifying with institutions (by using random permuted blocks within strata to
balance for individual factors other than institution), see Zelen (1974; 1979).
With regards to MSD interventions, it has been noted that randomising engineering controls in
multiple workplaces is extremely difficult and that it is much easier to randomise interventions
(such as exercise, education and treatment) aimed at personal behaviours (Silverstein and Clark,
2004). One such study (Lavender et al., 2007) argued that randomisation within workplaces
was necessary to control for potential differences in organisational cultures with regard to injury
reporting despite the risk of cross-contamination of the two study groups by individuals working
alongside each other.
An acceptable alternative to randomisation is minimisation, which aims to balance control and
intervention groups with respect to factors likely to influence the outcome (Altman, 1991).
Differences between groups will almost certainly exist despite the use of random allocation and
this allows direct control of these differences on important variables. Minimisation avoids
randomisation solely by chance, as it reduces any difference in the distribution of known or
suspected determinants of outcome, so that any effect can be attributed to the intervention under
test. Researchers should determine at the planning stage of the study which factors they would
like to see equally represented in the two groups (Treasure and MacRae, 1998). In general,
minimisation is of greatest value in relatively small trials (say with less than 100 subjects)
where several subject factors are known to be of research importance since in a small trial a
large difference can occur in one or more of the prognostic factors purely by chance. However,
if used, it is easy to include institution as another subject factor (stratum) (Pocock, 1983).
Minimisation is a technique mainly used in clinical trials, but is likely to use up too many
resources for occupational experimental trials and is therefore not recommended for
intervention studies.
27
3.2.6
Blinding and prior knowledge
Studies involving human volunteers are required in almost all circumstances to ensure that the
consent (“informed consent”) that volunteers give to participate is informed by knowledge of
the purpose of the study, the procedures involved and the potential risks and benefits. In the
case of MSDs, there is widespread knowledge in the community as to their prevalence and
nature and of the risk factors for them. Moreover, there have been widespread publicity
campaigns to raise the profile of the biopsychosocial approach to managing sickness absence
and return to work. In the UK, these have included publications such as The Back Book (Roland
et al., 2002), media campaigns such as Working Backs Scotland (Waddell et al., 2007), and
inspection and enforcement campaigns such as the joint HSE/LA Better Backs! campaigns in
2005, 2006 and 2007/8 (http://www.hse.gov.uk/betterbacks). It is therefore to be expected that
a significant proportion of any subjects recruited to an intervention study will have heard the
“stay active with back pain” message. A significant proportion will also be aware of the
importance of work-place ergonomic factors as risk factors for back pain.
From this, and from the nature of work-place interventions, it follows that it would be almost
impossible to blind participants as to their status as a member of an intervention group or a
control group. On the same basis, it would be difficult to conceive of a placebo intervention in
such circumstances. This raises an important methodological issue: Is it meaningful to take a
concept such as placebo that was developed in the context of trials involving the administration
of medication to patients, and to apply it to workplace intervention studies? There is almost no
discussion of this issue in the literature on ergonomics interventions, with one notable exception
(Volinn, 1999). There is some discussion in the wider occupational medicine literature by
Kristensen (2005), who noted that “blinding and the use of placebo are not core elements of the
RCT” but that they have special relevance for biomedical research.
Earlier discussion (Skov and Kristensen, 1996) had identified the importance of the effect of the
presence of researchers and their interest in the workers on the behaviour of subjects and their
likelihood to report MSD symptoms. They described these as the “placebo effect” and the
“nocebo effect” which are respectively positive and negative health effects resulting from
patient beliefs about treatments which are in themselves inert or ineffective. However, they
comment that, if the presence of observers increases workers’ awareness of safety measures, the
presence of the observers cannot be inert. They therefore argue that blinding and randomisation
could be superfluous in prevention effectiveness studies aimed at testing methods of reducing
exposure to known risk factors.
Volinn (1999) suggested that “sham” interventions are possible but would clearly need to be
carefully specified to be appropriate for any particular intervention or study design. He noted
that in double-blinded trials of analgesics, the size of the placebo effect depended on how
powerful the active ingredient was perceived to be. The implication is that if a sham ergonomic
workplace intervention is used, it must be perceived by the recipients as potentially beneficial.
Volinn (1999) also suggested that an intervention study should compare three groups: an
intervention group receiving the active intervention that is of interest; a placebo group receiving
a sham inactive intervention; and a control group receiving no intervention.
Volinn (1999) also raises a number of practical difficulties including the need for the
individuals who actually implement ergonomics interventions to be unaware of which
interventions are active and which are the placebo so that they communicate to the recipients a
belief in the effectiveness of the intervention they are delivering. However, the topic of MSDs
is topical and relevant to many people and the biopsychosocial approach and the early return to
work message have been, and will continue to be, widely disseminated and information about
them is available almost instantaneously via the Internet. It is therefore highly likely that a
28
motivated individual could easily find sufficient information to allow them to determine which
was the control and which was the active intervention.
If it is difficult to blind participants to their membership of an experimental group or a control
group, then different attitudes may arise in the two groups. The MRC (2000) framework
guidance warns that ‘The use of a control group in psychosocial and behavioural research can
produce “resentful demoralisation” among the control group triggered by a perception of
differential benefit for the intervention group among participants or providers.’ Related to this,
an employer or workplace may not permit inclusion of employees within a control group on the
basis that the proposed intervention has a perceived benefit and they cannot justify continuing to
expose the control group to the existing risks (Karsh et al., 2001).
One solution would be to move away from using a placebo group to a study which compares the
control group that does not receive an intervention to two intervention groups, without any
prejudgement as to the effectiveness of either. However, unless the effectiveness of one of the
interventions is already known, this makes it impossible to determine the absolute effectiveness
of either intervention compared to no intervention (Lavender et al., 2007). This situation should
be avoided if possible. Another possible solution would be a study based on a crossover design
where the initial control group receive the intervention at a later stage. In a true crossover
study, the intervention group would then revert to the control condition. This is very likely to
be impossible in this kind of study as removing a successful implementation is unlikely to be
acceptable to participants.
Given the nature of the implementation process for an intervention study, it is likely that there
will be not only a period while access to a site or group of workers is negotiated, but also an
extended period while the intervention is being designed and its implementation planned.
Consideration should therefore be given to treating the pre-intervention phase as a control
phase.
The effect of these factors on reporting of cases of MSDs will need to be considered at the
project design stage. If the target outcome is solely lost-time then it will be confounded by
differing expectations about when an individual decides that an MSD is sufficiently severe to
justify taking time off work. If it is taking time off to seek care, then the availability of health
care through the employer or outside working shifts will be confounders that will affect
decisions to take time off. Compensation systems and the early reporting/return-to-work and
sickness absence systems of employers will also be relevant.
3.2.7
Implementation
All intervention studies should include a pilot period, unless this is impractical. In addition, all
variables should be validated in previous studies, since if they are not validated, the study could
end up analysing the efficacy of the new measuring system as opposed to measuring the
outcome of interest.
3.2.8
Study documentation
A large-scale intervention study that runs over an extended period will require careful
documentation, especially to reduce the risk of information loss if members of the study team
leave. Documentation will be required in order to communicate the purpose of the study to
potential stakeholders and participants. The precise form that such documentation takes will
depend on the study protocol and the amount of information that needs to be collected or
communicated. The layout and wording will need to take into account the literacy level of the
target readers and the means of delivery. Where data are being collected using paper
documents, account will need to be taken of the ways in which the data will be recorded and
29
how they will be managed and aggregated for analysis. The accuracy of the data entered will
need to be ensured using error checking techniques and, possibly, double entry. Privacy and
security of data will also need to be planned, especially when data are being collected or
transmitted electronically. Documents are likely to include:
•
Study protocols and training material
•
Submission for ethical approval
•
Information packs
•
Consent forms
•
Baseline questionnaires
•
Follow-up questionnaires
•
Baseline and post-intervention survey protocols
•
Baseline clinical examination protocols
•
Diagnostic criteria
•
Validation of all methods of measurement
•
Follow-up timetables
•
Progress reports and final reports
•
Conference and peer-reviewed scientific papers
3.2.9
Access to the workplace
Carrying out an intervention study will require full and willing cooperation from the employer
where the intervention takes place. The precise design of the study will determine both the
direct and indirect costs to the employer. Direct costs are likely to include management time,
loss of output by production staff while being recruited and studied, and provision of facilities
to the study team. The lower the projected cost to the organisation; the more likely it is to be
willing to cooperate with the study.
Gaining access to working populations is becoming increasingly difficult (Punnett and
Wegman, 2004; Waters, 2004a). Security and commercial issues are very likely to lead to some
organisations being unwilling to allow access to an external study team. While interventions
among small and medium-size employers (SMEs) have been reported (Straker et al., 2004),
many SMEs may decline to participate due to having insufficient resources to be able to
participate without major cost to or disruption of their business. The “gatekeepers” within each
employer (Stephens et al., 2004) will need to be persuaded that participation is worthwhile.
The relationship between the study team and the employer will also need to be considered. The
perceived relationship between HSE and the study team will have a significant effect on the
willingness or otherwise of a firm to participate. If a firm has a positive relationship with HSE
and a proactive approach to health and safety, they might be willing to accept a study team with
direct involvement by HSE employees, such as staff from HSL. If they are afraid that
involvement in such a study could result in enforcement action by HSE inspectors they will be
unlikely to volunteer to allow the study to take place in workplaces under their control even if
the study team were from an outside organisation.
In order to minimise risks to the study, it will be necessary to ensure that support for an
intervention is obtained throughout the organisation it is implemented in. This will need to start
with commitment from senior management and safety management. This commitment will
30
need to be spread down through the organisation so that line managers fully cooperate. In
organisations where they have members, Trade Unions will need to be approached to obtain
support, which is often given and is usually extremely valuable.
It is likely that firms undergoing change for other reasons will not be suitable locations for the
study to take place as such a firm will have as its primary concern implementing its own
changes, and possibly ensuring that the business is additionally compromised.
3.2.10
Participant recruitment and retention
The ideal study would attract interest from large numbers of potential participants, have
inclusion criteria that fit a wide pool of individuals, be of obvious benefit, have minimal
negative aspects and be easy to participate in. Other considerations that will affect the ease of
recruitment of participants include:
• The nature of the intervention;
• Efficiency of contacting potential participants;
• Top-down commitment in the organisation to the project;
• The industry sectors targeted;
• The job types targeted;
• The inclusion/exclusion criteria for organisations, jobs and individuals;
• Individual histories of symptoms/problems;
• Literacy issues.
In occupational studies, it must be expected that there will be turnover of staff employed in jobs
included in the study and that there will be changes to jobs that participants carry out. Some
employment sectors and firms are more prone to change than others and wider macro-economic
factors, such as a decline in economic activity, can lead to individuals losing their jobs.
3.2.11
Follow-up
In order to have a successful study, with limited dropouts and good data, the follow-up period in
a prospective longitudinal study is vital. To maintain a high response rate, the study should
ideally provide suitable incentives to participation and retention, make sure the researchers have
up-to-date and accurate individual details, and consider the follow-up methods to be utilised.
Some of the issues needing to be considered in the follow-up period are:
• Duration of follow-up;
• Frequency of follow-up, hence number of follow-ups;
• Whether to repeat baseline measures;
• Measurement methods for outcome measures;
• Address/contact details checking;
• Multiple contact methods — visit, interview, clinical examination, post, phone, web
page, email, text messages;
• Data triangulation — validating data by seeking confirmatory data from other sources.
Multiple follow-ups involve regular contact with study participants and are advantageous in
improving data quality through more frequent measurement. They also reduce the likelihood of
31
memory limitations making data inaccurate and therefore allow more precise measurements of
when incidents of interest occur and hence better estimation of latency periods (Karsh, 2006).
Every effort should be made to get as high a response rate as possible at each stage of
recruitment and follow-up. However, it must be recognised that there will always be losses to
follow-up in cohort studies as some individuals will not be followed up for the full length of the
study, even with a short follow-up period. Such “right-censoring” will be due to a variety of
reasons, including companies going out of business or laying off workers, or individuals
changing jobs, refusing to continue participating, moving without leaving forwarding addresses,
changing phone numbers or email addresses, leaving the workforce due to illness, death or
pregnancy. On top of these, individuals that suffer the outcome of interest to the study are
thereafter treated as right-censored, so the available pool of individuals still at risk in the study
decreases with every incident case.
3.2.12
Study management and financing
The issue of project management and finance is considered in detail in Section 4 of this report.
However, it is vital that study protocols should consider:
• Probability and consequence of cost and time overrun;
• Budget constraints and cutbacks;
• Changes in staffing of the study team over an extended study duration.
3.2.13
Ethics considerations/approval
All proposals for studies involving human subjects are scrutinised by an ethics committee to
ensure that the potential benefits from the intervention/treatment outweigh the potential risks.
The fundamental ethical principles underpinning research on human beings are described in
national and international guidelines (Council for International Organisations of Medical
Sciences, 1993; 2002; Medical Research Council, 1998; Council of Europe, 2005; The General
Assembly of the International Statistical Institute, 1985). The World Medical Association
(WMA) (2004) developed the Declaration of Helsinki as a statement of ethical principles to
provide guidance in medical research involving human individuals, including research on
identifiable data.
Key issues considered are:
• Respect for the dignity of the subject and his/her well being;
• The free, informed consent of the subject to participation without undue inducement;
• The benefit–harm balance of the study (beneficence and non-maleficience), particularly
where the subject will not benefit directly;
• Distributive justice, i.e., the equitable distribution of both the burdens and the benefits
of participation in research;
• Confidentiality and hence data storage issues;
• The use of appropriate statistical methods in design and analysis.
Moreover, the WMA declared that any study that uses substandard statistical methods should be
deemed unethical. Ethical issues for clustered randomised trials are more complicated (Medical
Research Council, 2002), as consent is often obtained at the group level, the level at which the
intervention is implemented. Even so, where feasible, individual consent should be obtained
and related matters taken into consideration.
32
3.3
STUDY DESIGN
3.3.1
Choice of study type
There are important issues relating to study design in interventions studies. In particular,
“Observers of the medical literature have long noted a particular relationship between study
design quality and the results of intervention studies: study design quality and reported
outcomes are often inversely related”. Even more worryingly, it has recently been argued on
statistical grounds that the majority of published research findings are likely to be false
(Ioannidis, 2005), especially where effect sizes are small, a field uses variable study designs,
definitions and outcome measures and where there are strong vested interests and prejudices in a
field. Therefore, it is essential that the planners of a study designed to test the effectiveness of
ergonomics interventions be realistic about its prospects for success and take proper account of
meeting the requirements for methodological rigour and adequate power.
Epidemiological studies can be divided into two main types: descriptive and analytical.
Descriptive epidemiology describes disease and/or exposure and may consist of calculating
rates, for example incidence and prevalence. Descriptive studies do not use control groups and
can only generate hypotheses, not test them. Analytical epidemiology compares an exposed
group with a control group and usually tests a specific hypothesis. It includes two types of
studies: observational, such as case-control or cohort studies of incidence; and experimental
studies, including randomised controlled trials and intervention studies, where variables are
directly manipulated to test their effects on the outcomes of interest. These are generally
prospective as they monitor the impact of an intervention over time and are therefore able to
give confidence about causation.
The two types of intervention study are clinical trials and community trials (Woodward, 1999).
The key feature of an intervention study is that the allocation of an individual to an intervention
or a control group is planned, even if randomisation and blinding are used to prevent observer
bias. Thus, the investigators initially assign the intervention to whomever they wish and then
observe what happens prospectively. They either apply the intervention to individuals with
health problems to decide on an appropriate clinical treatment or to those presently free of
symptoms in order to decide upon an appropriate preventive strategy in the community or
workplace. There is an important difference between a controlled clinical trial and a workplace
intervention study (Skov and Kristensen, 1996); in a clinical trial the subject group is actively
exposed to a particular treatment method, such as a drug or a surgical method, and the
comparison group are not exposed or are exposed to a non-active placebo. In an intervention
study, generally the purpose is to remove the active exposure from the intervention group and to
investigate what happens to people whose exposure is stopped relative to those whose exposure
continues. As a consequence, this kind of study design is only suitable to conditions where the
exposure has reversible effects so that symptoms regress when the exposure is reduced or
ceases. It is eminently suited therefore to WRMSD studies since MSDs are typically of short
duration and liable to repeated occurrence. It would not be suited to chronic or degenerative
musculoskeletal problems, such as ankylosing spondylitis or to rare serious spinal pathologies
such as cauda equina syndrome.
The main advantage of intervention studies is their efficiency in investigating causality, as they
ensure that the ‘cause’ precedes the ‘effect’. They can also control for confounding and ensure
that interventions are compared efficiently. A reason for avoiding this study type would be if
the intervention were potentially harmful as this would limit recruitment and raise ethical
concerns. Another disadvantage is that the study selection criteria may screen out categories of
individuals as inappropriate, perhaps because of job type, age, or other factor, restricting
extrapolation of the results. Another disadvantage arises because randomisation of subjects to
33
groups is very difficult within the operational constraints encountered by researchers entering
functioning organisations as little more than tolerated guests (Griffiths, 1999).
In order to illustrate the differences between epidemiological studies, the basic designs are listed
in Table 5, and their relative merits for assessing the efficacy of interventions in an occupational
setting on MSDs are listed. The information in the table is not intended to be exhaustive, as, for
example, medical research studies often use mixed versions of basic designs, such as nested
case-controls, case-cohort, case-crossover, and panel study (or repeated cross-over).
Despite the acknowledged difficulties, it is concluded that because of the need to assess
causation and the effects of an experiment, longitudinal randomised controlled trials are the
most appropriate study type for assessing the effectiveness of ergonomics interventions in the
workplace.
3.3.2
Cross-sectional studies
The simplest kind of observational study is the cross-sectional survey, where a set of individuals
are observed or questioned to seek information on their risk factor exposure and/or disease
status. As they only provide a snapshot in time, they can measure the prevalence of disease
only. Cross-sectional studies will, therefore, not inform the researcher if the intervention had
worked or not and have limited usefulness as it is important to consider before and after effects
to assess if the intervention is causal.
Repeated cross-sectional studies can be carried out at different time points to assess trends over
time. However, as these studies can involve different groups of individuals at each time point, it
can be difficult to assess whether apparent changes over time simply reflect differences in the
groups of individuals studied.
It is recommended that cross-sectional study designs are not used to research preventative
intervention MSD questions as they are unable to assess causation.
3.3.3
Case-control studies
In a case-control study, a group of individuals with the disease or condition of interest (cases)
and an unaffected group (controls) are identified and their past exposures to the factors of
interest are compared. This is in contrast to the design of a cohort study, which begins with a
group of exposed people and compares them to a non-exposed group. The advantages of the
case-control approach are practical as it is relatively simple and thus quick and cheap. It is also
valuable when the condition of interest is very rare.
There are several problems with case-control studies. One of the major problems involves
subject recall; some individuals may have limited memory of their exposure, thus resulting in
them being misclassified (i.e. as cases when they were not exposed and should be controls).
Another difficulty is that cases and controls may differ in characteristics or exposures other than
the one that is targeted for study. To avoid the two groups being very different, the cases and
controls can be selected and matched, either on group or individual characteristics (Gordis,
2000). Finally, a case-control study is not appropriate for measuring the effect of an
intervention, as sampling is carried out according to disease rather than exposure status.
However, the most significant restriction of case-control studies in measuring preventative
measures in MSDs is that they are unable to demonstrate causality (Woodward, 1999) and
therefore, should be avoided for assessing preventative interventions in MSDs.
34
Population level (grouped)
Analyse accounting for paired
data
Observational
Observational
Observational
Experimental/
Intervention
Multiple crosssectional (over
time)
Multiple crosssectional (over
time)
Cross-sectional/
Longitudinal
Longitudinal
Panel
Ecological
Cross-over
35 Same individuals studied
repeatedly.
Cross-sectional data are
recorded in a succession of
surveys at two or more points in
time, with new sample on each
occasion.
Association vs. causality
Incidence/prevalence cases
Repeated crosssectional
Observational
Cross-sectional
Main features
Cross-sectional
Form
Timing
Cheap
Limits within individual variation
Smaller sample size Use of readily available data
Useful when risk factor
measurement at individual level
is particularly prone to error
Can detect and establish the
nature of individual change
Permit measurement of
differences or change in variable
over time
Used to located sleeper effects
(connections between events that
are widely separated in time)
Useful for generating hypothesis
Pros Epidemiological study designs
Type of study
Table 5.
Lacks statistical power in
detecting treatment effect
Carry-over effects
Only suitable for long-term
conditions with intervention
providing short-term relief
Takes longer
Can only analyse population level
/not individual level
May encounter ecological fallacy
Panel attrition due to refusals,
changes of residence or death of
respondent
Course of events between
discrete recording points remains
unknown
Conditioning of individuals
Unable to detect changes within
individual as a different group is
measured each time
Unable to establish causality
Cons
Timing
Longitudinal
Longitudinal
Longitudinal
Longitudinal
Type of study
Case-control
Cohort
Nested casecontrol (based
on cohort)
Trial
Intervention
Experimental/
Intervention
36 Exposure/outcome measurement
Exposure/outcome measurement
Retrospective
Selection of cases/controls
Matching
Main features
Observational
Observational
Observational
Form
Loss to follow-up
Expensive and time-consuming
Not suitable for rare diseases
Conditioning of individuals
Exposure may change
Suitable for rare exposures thus
able to study wide range of
diseases
Give sequence/causality
Multiple diseases can be studied
simultaneously
Can demonstrate causality
Compares interventions
efficiently
Control for confounding
Can be costly
Ethical issues with giving
experimental interventions
Selection bias (restricting
generalisability of results)
Not suitable for rare diseases
Conditioning of individuals
Suffer from bias error
Observer/respondent bias
No time sequence
Can investigate only 1 disease
outcome
Provide approximate estimates of
relative risk
Suitable for rare diseases
Quick and cheap
Multiple risk factors can be
studied
Smaller sample sizes than
equivalent cohort studies
Able to evaluate confounding and
interaction
Provides results before the cohort
finishes
Saves resources
Individuals matched
Limits bias error
Cons
Pros
3.3.4
Cohort studies
Cohort studies track the same people over time, and therefore the observed differences in the
members of the cohort are more likely to be the result of genuine changes than differences
between individuals. By doing repeated measures at the individual level, longitudinal studies
have more power than cross-sectional observational studies, by being able to exclude timeinvariant unobserved individual differences and by observing the temporal order of events.
A cohort study could be used to assess the effect of a “found” or uncontrolled intervention and
allow the researcher to control for confounders and measure various outcome measurements at
different time points (thus having the possibility of conducting a nested case-control study).
Because of the need to observe unaffected individuals until a fair proportion develop the
outcome of interest, cohort studies can take a long time and may thus be very expensive. They
are usually unsuitable for studying rare outcomes, as it would be necessary to follow a very
large number of individuals to get an adequate number of events (Altman, 1991).
A number of potential biases must be either avoided or taken into account in conducting cohort
studies. The major biases include information, analytic, non-response and loss to follow-up,
and in assessment of the outcome.
When trying to assess the aetiological effect of a risk factor, individuals recruited to cohorts
should ideally be symptom or disease-free at the start of the study. Ensuring that any exposure
to the risk factor occurs before the outcome enables a causal role for the factor to be postulated.
Advantages of cohort studies include being able to give sequence of events, providing
information on a wide range of outcomes and allowing changes in exposure over time to be
studied. Because of their longitudinal nature, cohort studies would be the most appropriate type
to assess ergonomic interventions (Punnett and Wegman, 2004; Waters, 2004a).
3.3.5
Randomised Control Trials
An RCT would probably be the optimal approach for assessing the effectiveness of preventative
workplace ergonomic interventions (Waters, 2004b) as it can separate the effects of the
intervention from those of extraneous factors such as natural recovery and statistical regression
(Herbert and Bo, 2005). RCTs can manipulate the intervention so that groups are allocated
without bias and so avoid any possible problem due to confounding factors (Woodward, 1999).
A disadvantage of RCTs is that they are generally costly and labour intensive, especially to keep
individuals in the study, and usually have small numbers of participants, which means that they
need to detect large differences between interventions. One way of improving RCTs, is to
increase a study’s sensitivity by carefully selecting individuals, intervention measures and the
study endpoints (Rothman and Greenland, 1998). When conducted well, RCTs have the
advantage of demonstrating causality and compare interventions efficiently.
A group of scientists and editors have developed the CONSORT (CONsolidated Standards Of
Reporting Trials) statement (Begg et al., 1996; Moher et al., 2001; Altman et al., 2001). This
consists of a checklist and flow diagram recommended for use by authors when reporting an
RCT (see Section 5.2).
3.3.6
Clustered Randomised Trials
In order to assess an intervention in an occupational setting, randomisation is likely to be done
at the group or cluster level (i.e. by factory) instead of the individual level and hence the study is
37 termed a Clustered Randomised Trial (CRT). If the intervention involves supplying equipment
or staff to an administrative unit then by randomising these units rather than individuals only a
subset of the units receives the equipment or staff (Medical Research Council, 2002). This
avoids cross-contamination, is cheaper than doing randomisation at the individual level and is
administratively more convenient.
Although CRTs retain many of the essential features of individually randomised trials, they do
have extra logistical, ethical and statistical issues to consider (Elbourne and Campbell, 2001;
Kerry and Bland, 1998), and should usually be avoided unless RCTs are practically impossible.
However, it appears to be a practical study design for a workplace intervention study.
IJzelenberg et al. (2007) recently reported a CRT of an intervention to prevent low back using
the CONSORT recommendations.
3.3.7
Crossover studies
In a crossover study design, the individuals receive different interventions during different
periods of time. For example, the effect of intervention 1 can be individually compared with the
effect of intervention 2 on each subject, allowing within-individual differences to be calculated.
Unfortunately, there are several disadvantages to crossovers that restrict their application. They
are particularly vulnerable to the effects of subject withdrawal, since individuals that withdraw
after the first period cannot be included in the analysis because they do not receive the other
intervention (Altman, 1991). They are also less appropriate when the efficacy of the first
intervention continues for a prolonged interval (Rothman and Greenland, 1998). If this is likely
to occur, the trial may incorporate a ‘washout’ period between the intervention periods.
Crossover studies are ideally suited for long-term conditions for which treatment only provides
short-term relief (Woodward, 1999). Compared to parallel group studies, crossover studies are
more complex to analyse.
The main advantage of crossover studies is that by accounting for between-subject variability in
the outcome, they may be more efficient than a parallel group trial (Kirkwood and Sterne, 2006;
Rothman and Greenland, 1998). Depending on the specific research question, a crossover study
is potentially good for assessing an intervention to prevent MSDs.
3.3.8
Quasi-experimental studies
Quasi-experiments are a variety of experimental design where a quasi (almost)-experiment is
done instead of a full experiment. Typically they are controlled studies in which exposure is
assigned, but not according to a randomised experimental protocol (Cook and Campbell, 1979;
Rothman and Greenland, 1998). They are frequently used by ergonomists to investigate
research questions (Goldenhar and Schulte, 1996; Zwerling et al., 1997). Recent examples
include studies of participatory interventions (Laing et al., 2005; Rivilis et al., 2006) and a
comparison of two groups with different exposures to biomechanical load (Bonfiglioli et al.,
2007). They are suitable for “found” experiments or observational studies but are not as
powerful as a true randomised experiment.
One of their purposes is to capture longer time-periods and a sufficient number of different
events to control for various threats to reliability and validity. Instead of investigating cause,
they tend to report trends and instead of randomisation they tend to match subjects or worksites.
Zwerling et al. (1997) describe them as being appropriate in circumstances where it is neither
necessary nor ethical to carry out a randomised controlled trial. The comparison group serves to
provide an estimate of what the injury incidence would be without the intervention. If selection
or matching of subjects is not possible confounding can be controlled either by multivariate
analysis or by stratification.
38 3.4
DATA COLLECTION
3.4.1
Data quality
In order for any study to be successful, consideration must be given to a number of issues
regarding the data collected because the quality of any study is determined by the quality of the
resulting data. This depends on the types of information to be collected and the methods used to
obtain them and to ensure their accuracy. Moreover, the chosen data collection method will
affect response rate, bias and costs. As well as outcome data, epidemiological studies of MSDs
usually involve measurements of the workplace and the physical exposures experienced there.
Some studies rely on data from records, although self-administered questionnaires and interview
methods are very common means of data collection (Bowling, 2002). If an interview method is
preferred, the issue of structured, semi-structured or in-depth needs to be addressed as well as
whether the interview is to be personal or by telephone. Each method has its advantages and
disadvantages, and each has implications for bias.
In an intervention study, it is important that the intervention should not evolve over time
according to the experience of those providing it (Medical Research Council, 2000). This is to
ensure that the same changes are implemented and that comparable data are collected.
Therefore, quality checks and monitoring of the status of the implementation will need to be
planned into the study. However, once an intervention has been shown to be effective and is
being widely implemented, it is desirable that it should be allowed to evolve to reflect local
conditions.
Whichever data collection method is chosen, it should be validated internally and externally.
Failure to validate can introduce bias.
3.4.2
Variable definition and selection
Exposures for MSDs are complicated to measure because very large numbers of factors have
been implicated as risk factors. The problem with many measurement methods is that they
measure instantaneous exposure, not ongoing or historical exposure and are time-consuming to
use. The methods that have been used for exposure assessment include (Hernberg, 1992):
•
Static biomechanical models
•
Dynamic biomechanical models
•
Measurements of energy consumption
•
Questionnaires
•
Ergonomics assessments
•
Registration of work load
•
Posture measurement
Moreover, there is a need to understand better the specific effects of a number of factors
(Punnett and Wegman, 2004):
•
Dynamic forceful motions
•
Prolonged low-effort exertions
•
Extreme postures
•
Repetitive motion close to the centre of the normal range
•
Non-cyclical work
39
Finally, the measures chosen should satisfy the following criteria (Lipsey, 1996):
• Valid measurement of the variables of interest;
• Reliability of measurements over irrelevant variation in the occasion and circumstances
of measurement;
• Practicality of measurement in the circumstances of the study;
• Sufficient sensitivity to respond to changes in/distinguish levels of the variable;
• Multivariate measures of outcomes to capture the range of aspects of the outcomes.
3.4.3
Baseline data
Data will be required that are sufficient to adequately characterise the differences between the
intervention group and the control group before the start of intervention (Friedman et al., 1998).
These baseline data are likely to be of a range of types, both parametric and non-parametric.
They are also likely to include both objective measures and subjective responses from
participants, such as:
• Basic demographic and anthropometric data on individual participants;
• Contact details;
• Job details;
• Details of the employer;
• Details about the employer;
• Health status details, especially history of MSDs, measured with a tool such as the
Nordic Musculoskeletal Questionnaire (NMQ);
• Relevant medical history;
• Results of a clinical examination;
• Psychosocial measures;
• Safety management systems in place;
• Safety climate measures;
• Details of exposure to the risk factors being studied, including measures of posture,
activity and biomechanical loading;
• History of previous exposure;
• Measures of confounders and covariates.
Measurements of exposure to risk factors are likely to include photographic or video evidence.
Although they have been widely used, (e.g., Bergstrom et al. (2007)), there are concerns about
the lack of precision of questionnaire methods for obtaining physical exposure data, especially
if a measure of cumulative load is being sought (Waters et al., 2006). In fact, Waters et al.
(2006) note the pressing need for a valid, reliable and practical method for estimating
cumulative spinal loading.
Assessing physical exposure using surrogate measures such as job title, is an unacceptable
method as the variability of exposures to individuals within job titles is very great (Punnett and
Wegman, 2004; Gardner et al., 2000). This is especially so where generic job titles, such as
“production worker”, are used to indicate position within a hierarchy, not the specific demands
and activities of the job.
40 Baseline data need to be collected before and immediately after the implementation of the
interventions. Data will also need to be collected on the process of implementing the
intervention and on how well the final status of the intervention reflected the target. In other
words, was the intervention implemented as planned? This will permit the measurement of the
precise dose of the intervention as initially applied and an assessment of whether the
intervention as implemented led to the intended change in exposure.
3.4.4
Outcome measures/case definitions
The target outcomes of the intervention need to be clearly specified at the outset, as do case
definitions. This is fundamental to any intervention study to reduce the risk of reading
previously accepted findings/conclusions into the data. In the context of MSDs it is necessary
because there is a progression of health outcomes of increasing severity, and different
interventions may have different effects on symptoms, symptom reports, ‘injuries’, sickness
absences or the incidence of long-term disability (Waddell and Burton, 2001).
Careful selection of outcome measures/case definitions will enable the efficient capture of the
information needed to test the intervention by focusing data collection on the outcomes of
interest and avoiding the collection of unnecessary data. Outcome measures must be chosen
carefully to maximise response rates at follow-up. It is strongly recommended that a hierarchy
of case definitions be used to capture the range of possible outcomes. These can range from
temporary discomfort, through acute injury, to long-term disabling work loss. Where possible,
they should use accepted definitions in order to facilitate comparison with other studies.
Indeed, standardisation of definitions is particularly important where clinical outcomes are
being used and in other situations where data collection is to be carried out by a range of
individuals.
The lack of data on latency periods mean that intervention studies can appear to fail if dose and
response measurements are taken at the wrong times (Karsh, 2006). There are difficulties in
measuring doses in a practical and economical manner and of measuring apparent cascading
doses and responses. As a result, multiple outcomes should be measured, both short and longterm.
3.4.5
Follow-up data
The purpose of a follow-up phase of a longitudinal study is to collect data marking the transition
of an individual to becoming a case of the outcome of interest. This can be done by measuring
status at fixed time intervals and/or by recording when the change of state/event of interest
occurs. Once time-to-event data or incident status data are available, they can be related to the
exposure status of the individual. However, in a longitudinal study, there is always the risk that
the exposure of the individual will change over the duration of the study because of various
factors, including:
• Changes in the workplace, such as job redesign;
• Changes in working patterns;
• Changes of job;
• Changes in the individual, such as injury or health problems;
• Changes in the psychosocial context at work, such as changes in management or
staffing in a work team;
• Changes in personal circumstances altering the amount of social support received.
41 It cannot be assumed, therefore, that measurements at baseline will adequately characterise the
exposure throughout the follow-up period and appropriate monitoring or re-measurement of
exposure will be needed. Possible methods for such monitoring include:
•
Repeat of baseline questionnaires at follow-up;
•
Questions about changes in the job at follow-up;
•
Repeat of physical exposure measurements at follow-up.
A system for immediate reporting of incidents alongside a regular follow-up system is
recommended in order to increase the probability of capturing incident events. An incident
reporting system would probably attempt to capture:
•
Date of data report;
•
Date of incident;
•
Nature of incident;
•
Severity of incident;
•
Results of a clinical examination and any first aid/medical treatment given;
•
Repeat measures of the status of the intervention and its precise dose;
•
Repeat measurements of health status and psychosocial status;
•
Repeat measures of confounders and covariates;
•
Any changes in employment and hence exposure status;
•
Changes in health status that affect the ability to continue in the study (e.g. pregnancy);
•
Changes in contact details.
A programme of regular follow-ups would collect data on a sub-set of these , i.e.,
•
Date of data report;
•
Dates of any incidents in the previous period;
•
Nature of incident;
•
Severity of incident, including amount of time off work;
•
Repeat measures of the status of the intervention and its precise dose;
•
Repeat measurements of health status and psychosocial status;
•
Repeat measures of confounders and covariates;
•
Any changes in employment and hence exposure status;
•
Changes in health status that affect the ability to continue in the study (e.g., pregnancy);
•
Changes in contact details.
3.4.6
Project management data
Finally, project management data will also need to be collected, including:
•
Informed consent from participating firms and individuals;
•
Contact details for firms and individuals;
•
Evidence of validation and calibration of data collection methods;
42 3.5
•
Participation rates;
•
Response to follow-up rates.
DATA ANALYSIS AND INTERPRETATION
This section will discuss issues related to the analysis of the data. The analyses undertaken will
depend on the original research question, the study design used and the types of variables
measured.
Specific to occupational epidemiology, analyses will need to take into consideration worker
turnover, job redesigns, exposure stability and consistency, non-work-related exposures
(concurrent exposures) and the healthy worker effect. Moreover, analyses should consider
missing data, non-compliance with the study protocol, and subject drop-out (Pearl, 2000b).
Analysis of the effect of non-compliance with the protocol needs to consider the effect on
exposure to risk factors, as well as health outcomes (Silverstein and Clark, 2004).
During many randomised controlled trials, participants are lost to follow-up, which can result in
bias if the characteristics of these individuals differ between the randomised groups. Such
differential attrition prevents a full intention-to-treat analysis being carried out and can
introduce bias (Tierney and Stewart, 2005; Hollis and Campbell, 1999). It is therefore
important to describe the missing data and take appropriate steps in the analysis (Omar et al.,
2004; Peng et al., 2004; Collins et al., 2001; Schafer and Graham, 2002; Carpenter et al., 2002).
Model developers should clearly state the extent of missing observations in their data and how
these were dealt with in the modelling process. Before fitting the model, it is important to
determine if there are systematic differences in the characteristics of subjects with missing risk
factors as this could introduce bias. Furthermore, if the extent of missing data is large,
appropriate methods should be used to substitute missing values and the results examined to see
if the results remain consistent. Many methods are available such for substitution and, before
fitting the model, a preferred approach should be selected in the light of the objectives of the
study (Roberts et al., 2002).
Analysis of intervention trials can be complex and should examine the size of the effect of the
intervention, both before and after adjustment for baseline variables. Analysis of subgroups
may also be needed. Analyses will need to be more specialised for cluster randomised trials or
crossover trials (Kirkwood and Sterne, 2006) since clustered data are treated as repeated
measurements in longitudinal studies. When analysing clustered data, summary measures for
each cluster with robust standard errors should be reported and random effects models and
Generalized Estimating Equations (GEE) should be used. Analyses of crossover trials take the
design into account by using methods for paired data; for numerical outcomes, the mean
difference between each subject’s results on the first and second intervention is analysed and the
standard deviation of the mean differences reported.
As mentioned previously, the ability to assess causation and association between exposure and
outcome depends on the study design, which will in turn affect the type of analysis undertaken.
Table 6 summarises the measures used to test association and the impact of an exposure
(Kirkwood and Sterne, 2006).
43 Table 6.
Measures of association and impact
Measures of association
Risk ratios
Assess the strength of association between an exposure and an outcome
Odd ratios
Rate ratios (or hazard
ratios)
Comparisons of the Risk,
Odds and Rate ratios
Used for rare outcomes
Measures of the impact of an exposure
Attributable risk
Gives the magnitude of the excess risk in absolute terms (sometimes a
percentage: proportional attributable risk)
Comparing attributable and Gives the measure of strength of an association compared to the excess
relative measures
risk.
Population attributable risk
The impact at the population level is assessed by the excess overall risk (or
rate) in the population, as compared with the risk among the unexposed.
Potential impact of
reducing exposure
Measure of impact that would be achieved by a completely successful
intervention, which managed to eliminate the exposure.
Measures of the impact of an intervention
Efficacy
The efficacy of an intervention is measured using the risk ratio to
determine how many cases it prevents.
Number needed to treat
The number of individuals who must be treated to prevent one adverse
event.
3.6
PRESENTATION OF RESULTS
Systematic reviews of ergonomics intervention studies have typically assessed them as having
poor to marginal methodological quality (Tuncel et al., 2006a; 2006b), mainly due to
incomplete reporting of methods and findings. It is recommended that attention is paid to
reporting in order to ensure that the results are presented as comprehensively as required, which
will aid interpretation, application and future meta-analysis. Results should be presented in at
least two stages; firstly with an initial report describing and summarising the data, and secondly
presenting the results from all statistical analyses. The CONSORT statement (Moher et al.,
2001) includes a checklist and flow diagram recommended for use by authors reporting an RCT
(see Section 5.2).
44 4
4.1 EXAMPLE PROJECT PLAN
INTRODUCTION
This section of the report gives an example of a possible project plan and illustrates the
implications in terms of staffing, and duration, and hence cost.
It cannot be stressed too strongly that this example is hypothetical and has many untested
assumptions included. All estimates of time required are crude. The estimates of costs are
based on typical rates at HSL so should be seen as merely indicative of the scale of costs
required for such a study. The detail of the interventions to be implemented is deliberately not
specified, but it is assumed that engineering interventions are included.
As a result, any attempt to take the time or cost indications and to use them as a basis for a
tender specification or as a benchmark to evaluate submitted tenders against would be a gross
misuse of the example.
4.2 ASSUMPTIONS UNDERLYING THE HYPOTHETICAL PROJECT PLAN
The hypothetical project is based round a longitudinal clustered randomised study design where
randomisation is done at the workplace level. The time estimates are based on the recruitment
of five substantial groupings of clusters. Timings in the project plan are estimates in threemonth blocks so it is assumed that each cluster grouping could be recruited within one three
month period. This assumption may be untestable until the recruitment process is started. Each
cluster grouping could relate to a particular type of intervention or to a particular industry
sector. It is assumed that each cluster grouping would have at least 1000 participants, equally
split between an intervention group and a control group. A follow-up period of one year is
used.
Staffing descriptions are based on roles within the project. Staff time estimates for each role in
each three-month block or phase are of the proportion of time required from a full-time team
member. Full-time equivalent (FTE) numbers of staff are calculated by multiplying percentage
time by duration and summing. The FTE estimates show that multiple individuals would be
required at some stages of the study for some roles. At other stages, these roles are not required.
4.3 POSSIBLE GANTT CHART FOR THE PROJECT WITH CRUDE
ESTIMATES OF STAFFING, TIME AND COSTS
A large scale, complex intervention study must inevitably have a significant duration. There are
therefore complex logistical issues that must be managed effectively for it to be successful. An
illustrative Gantt chart for the example project is shown in Figure 4. It assumes no overlap
between project phases/stages, except in Phase 3, Data collection. Table 7 shows possible
timings for the stages of the example project. Table 8 shows a breakdown by role of the
possible staffing requirements of the hypothetical project. It includes reviews of the viability of
the project at a number of break points and identifies the point of no return after which
cancellation would be unwise. It also indicates the stakeholders that will need to be consulted at
the different stages. Table 9 shows a breakdown of time estimates for the hypothetical project
by phase and by job role.
The full timescale shown includes two years for publication of the results at scientific
conferences and in peer-reviewed journals. The assumption is that the results of the study
would be presented to HSE at the start of this period, in a report format but would then be
submitted for peer-reviewed publication. While publishers are increasingly using electronic
45 submission and review, which are helping to reduce the period from submission to publication,
the vagaries of the peer-review process and publication schedules mean that a substantial period
must be allowed before final publication occurs. Because there would be little activity during
this period, the staff time involved and associated costs would be relatively small.
On the basis of the example timings given, and assuming no rescheduling, HSE could expect to
have final results that could inform its policy decisions 5.75 years after the project planning
began. If a longer follow-up period was used this would increase the project duration directly.
If recruiting clusters were to take, say, an average of 3 months longer than the assumed 3
months, this would increase the project duration by that amount.
The time estimates have been used to calculate estimates of cost. This involved using HSL
charge out rates for appropriate grades of staff. As almost all the work would occur within the
first six years of the project plan, total costs were adjusted by making an allowance for inflation.
This was done by first allowing approximately a year for HSE to decide to fund the project and
then averaging the costs to the mid point in the project. This resulted in making an adjustment
for inflation over four years. An allowance of 25% was then added to cover project specific
expenditure such as travel and equipment costs. Because of the very great uncertainty in such
estimates at this stage, many of which would remain even at the end of Phase 1, a 100%
contingency was then added.
The estimated total cost of this project, given the assumptions stated above, is approximately
£11.5 million. This figure must be seen as only indicative. No breakdown of costs is provided,
as one would be misleading. When comparing tenders received, HSE will want to take into
account the different charging regimes that different organisations have. It is therefore
recommended that in such a situation HSE compares the amount of staff time allocated in the
submitted tenders. It is anticipated that such tenders would provide more detail of how staff
time would be allocated to the various phases of the project.
The final plan for any project funded by HSE should be similar in outline. Any tenders received
should be compared with this outline to ensure that all necessary activities are included. It
should be borne in mind that tenders could differ in detail from this outline, particularly in
relation to sequencing or overlapping of project activities.
46 0 0.25
1
0.5
2.1
0.75
Time in years
8 Publication
7 Interpretation
0 0.25
0.5
0.75
6 Presentation of summary statistics
5 Data analysis
4 Final data processing
3 Data collection – cluster E
3 Data collection – cluster D
3 Data collection – cluster C
3 Data collection – cluster B
3 Data collection – cluster A
2 Approval
Planning/
1 statistical design
Phase
Time in years
1
2.2
1
1.25
2.3
1.25
1.5
3.1
1.5
1.75
3.1
3.2
1.75
2
3.1
3.2
2
Figure 4.
2.25
3.1
3.2
3.3
2.25
2.5
3.1
3.2
3.3
3.4
2.5
2.75
3.2
3.3
3.4
3.5
2.75
3
3.3
3.4
3.5
3
47
3.25
3.3
3.4
3.5
3.25
3.5
3.4
3.5
3.6
3.5
3.75
3.5
3.6
3.75
4
3.6
4
4.25
3.6
4.25
4.5
3.6
4.5
4.75
4.75
Gantt chart for the hypothetical project
5
4
5
5.25
5
5.25
5.5
6
5.5
5.75
7
5.75
6
8.1
6
6.25
6.25
6.5
8.2
6.5
6.75
8.3
6.75
7
7
7.25
7.25
7.5
8.4
7.5
7.75
7.75
Table 7.
Possible timings for the example project
Phase/stage
Stage duration
Time elapsed
during phase
Total time elapsed
1
Planning/statistical design
6 months
6 months
6 months
2
Approval
2.1 Gaining ethical approval
3 months
3 months
9 months
2.2 Engagement of stakeholders
3 months
6 months
1 year 0 months
2.3 Staff training
3 months
9 months
1 year 3 months
3.1 Recruitment of firms in first cluster
3 months
3 months
1 year 6 months
3.2 Identification of suitable jobs and
locations in first cluster
3 months
6 months
1 year 9 months
3.3 Designing, piloting and economic
6 months
evaluation of interventions in first cluster
1 year 0 months
2 years 3 months
3.4 Recruitment of individuals, collection and 3 months
processing of pre-intervention baseline
measures for first cluster
1 year 3 months
2 years 6 months
3.5 Implementation of interventions,
collection and processing of postintervention baseline measures for first
cluster
3 months
1 year 6 months
2 years 9 months
3.6 Collection and processing of follow-up
measures for first cluster
1 year
2 years 6 months 3 years 9 months
3.6 Collection and processing of follow-up
measures for final cluster
1 year
3 years 6 months 4 years 9 months
4
Final data processing
3 months
3 months
5 years 0 months
5
Data analysis
3 months
3 months
5 years 3 months
6
Presentation of summary statistics
3 months
3 months
5 years 6 months
7
Interpretation
3 months
3 months
5 years 9 months
8
Publication
2 years
2 months
7 years 9 months
8.1 Preparation of scientific papers and
submission to conferences/journals
3 months
3 months
6 years 0 months
8.2 Wait for results of peer review
6 months
9 months
6 years 6 months
8.3 Revision of papers and resubmission
3 months
1 year 0 months
6 years 9 months
3
Data collection
8.4 Making conference presentations/waiting 1 year
for journal publication
48 2 years 0 months 7 years 9 months
Table 8.
Possible staffing of the example project
Stage
Project staff directly
involved
1
Planning/statistical
design
Project manager
MSD specialist
Epidemiologist
Psychologist
Trials manager
2
Approval
2.1
Gaining ethical approval Project manager
Epidemiologist
MSD specialist
Psychologist
Ethics committee
Project manager
Epidemiologist
MSD specialist
Psychologist
Funding body;
Collaborating organisations
Project manager
Trials manager
Facilities team
Employer organisations;
Chief Executives; Trade
Unions; Royal Colleges;
Other professional bodies
PBP2 Project break point
Project manager
Epidemiologist
MSD specialist
Psychologist
Funding body;
Collaborating organisations
2.3
Staff training
Project manager
MSD specialist
Epidemiologist
Psychologist
Trials manager
Call centre
Field ergonomists
Clinicians
Clerical team
Facilities team
Intervention topic
specialists/engineers
Economist
3
Data collection
3.1
Recruitment of firms
Clerical team
Facilities team
Boards and safety advisors
of companies
3.2
Identification of suitable Trials manager
Field ergonomists
jobs and locations
Intervention topic
specialists/engineers
Project manager
MSD specialist
Facilities team
Local managers/safety
officers; Trade Unions;
Safety reps
Designing, piloting and
economic evaluation of
interventions
Project manager
Facilities team
Local managers, engineers
and safety officers; Trade
Unions; Safety reps;
Intervention providers (e.g.
equipment, processes or
training)
PBP1 Project break point
2.2
3.3
Engagement of
stakeholders
Other project staff External stakeholders
involved
Funding body;
Collaborating organisations
Project manager
Trials manager
Call centre
MSD specialist
Psychologist
Trials manager
Field ergonomists
Intervention topic
specialists/engineers
Economist
49 PBP3 Project break point
Project manager
Epidemiologist
MSD specialist
Psychologist
Funding body;
Collaborating organisations
3.4
Recruitment of
individuals, collection
and processing of preintervention baseline
measures
MSD specialist
Epidemiologist
Psychologist
Trials manager
Field ergonomists
Clinicians
Clerical team
Project manager
Facilities team
Local managers and safety
officers; Trade Unions;
Safety reps
3.5
Implementation of
interventions, collection
and processing of postintervention baseline
measures.
MSD specialist
Epidemiologist
Psychologist
Trials manager
Field ergonomists
Intervention topic
specialists/engineers
Clerical team
Project manager
Facilities team
Local managers, engineers
and safety officers; Trade
Unions; Safety reps;
Intervention providers (e.g.
equipment, processes or
training)
Project manager
Epidemiologist
Trials manager
MSD specialist
Psychologist
Funding body;
Collaborating organisations
PBP4 Project break point
3.6
Collection and processing Trials manager
of follow-up measures
Call centre
Field ergonomists
Clinicians
Clerical team
Project manager
Facilities team
PBP5 Project break point —
point of no return
Project manager
MSD specialist
Epidemiologist
Psychologist
Trials manager
4
Final data processing
MSD specialist
Epidemiologist
Psychologist
Clerical team
Project manager
5
Data analysis
Epidemiologist
MSD specialist
Psychologist
Economist
Project manager
6
Presentation of summary MSD specialist
Epidemiologist
statistics
Psychologist
Economist
Project manager
Clerical team
7
Interpretation
MSD specialist
Epidemiologist
Psychologist
Economist
Project manager
8
Publication
8.1
Preparation of scientific MSD specialist
papers and submission to Epidemiologist
conferences/journals
Psychologist
Economist
Project manager
Clerical team
Funding body
8.2
Wait for results of peer
review
Project manager
MSD specialist
Journal editors/reviewers
Funding body;
Collaborating organisations
None
50
Epidemiologist
Psychologist
Economist
8.3
Revision of papers and
resubmission
MSD specialist
Epidemiologist
Psychologist
Economist
Project manager
Clerical team
8.4
MSD specialist
Making conference
presentations/waiting for Epidemiologist
journal publication
Psychologist
Project manager
Economist
Table 9.
Conference organisers;
Journal editors; Scientific
community
Time estimates (FTE years) for the hypothetical project
Phase
Role
1
2
3
4
5
6
7
8
Total
Project manager
0.250
0.525
1.250
0.025
0.025
0.025
0.025
0.050
2.175
MSD specialist
0.250
0.250
1.125
0.250
0.175
0.250
0.250
0.325
2.875
Epidemiologist
0.250
0.250
1.000
0.250
0.250
0.250
0.250
0.325
2.825
Psychologist
0.250
0.250
1.000
0.250
0.175
0.250
0.250
0.325
2.750
Trials manager
0.250
0.500
3.313
4.063
Call centre
0.100
1.125
1.225
Field ergonomists
0.250
4.875
5.125
Clinicians
0.250
1.750
2.000
Clerical team
0.250
2.375
Facilities team
0.025
1.250
1.275
Topic specialists
1.875
1.875
Economist
0.250
0.050
0.050
0.050
0.075
0.475
21.188 1.025
0.675
0.825
0.825
1.100
29.538
Total
1.250
2.650
0.250
2.875
4.4
STUDY IMPLEMENTATION ISSUES
4.4.1
Staffing
Suitably qualified and experienced staff will be required to manage and implement the study
over its duration. Skills will be required in project management and specifically in trials
coordination, in addition to the need for technical skills in musculoskeletal ergonomics and in
epidemiology. If, as suggested in this report, clinical examinations of participants are proposed,
suitably medically qualified, experienced and trained staff will be required to carry them out. In
addition, administrative requirements such as contacting participants for follow-up purposes
will need to be allowed for. The staff and skill mix required will vary over the duration of the
project. While some roles will be required throughout the project, some will only be required at
specific stages. While there will be clear advantages in senior staff being involved for the
duration of the project, this cannot be guaranteed and the possible need to find replacements
must be considered. In some roles, individuals will be more easily replaced, especially where
multiple individuals are required. Planning will need to ensure that suitable individuals are
51 available at the necessary stages of the project for the duration of the study and that
replacements can be brought in, possibly at short notice, when that proves necessary.
4.4.2
Finance
A commitment to providing sufficient finance to fund the whole study will be required before
the project is started. This should be made contingent upon the project being shown to be viable
at each of the break points/review stages proposed in the flow chart in Table 8. An adequate
contingency allowance will need to be available to take account of both foreseeable problems
such as staff turnover or difficulties recruiting subjects and other unforeseen circumstances.
Such a complex study will almost certainly be carried out at multiple sites and there will be
significant amounts of travel involved as a result. Sufficient time and funding will be needed to
allow proper reporting of the results of the study in the appropriate scientific fora and journals.
Because of the potential scale and duration of such a project, one single funding body, such as
HSE, may feel that its budget is not sufficient to cover the entirety of the project. If this is the
case, the possibilities of collaborative funding from suitable partners will need to be explored.
It is beyond the scope of this work to consider who such partners might be.
One approach to funding that could be used to help manage the risks of the project would be to
allocate money in budgets for the full proposed life of the project but to only commit money to
future stages in the light of the on-going monitoring of the progress of the project.
A specific funding issue that will need addressing at the project design phase is how the actual
implementation of interventions will be funded. The possible interventions vary in cost
implications and in potential payback to the employer. Thus, a training intervention would have
costs associated with the provision of the trainer and training materials, and the lost
opportunity/lost production costs of the staff being trained. Even with on-going training or
refresher training, such costs could easily be one or two orders of magnitude smaller than the
costs of an engineering intervention where processes and equipment were modified.
If an employer can see a clear financial case for the intervention and has sufficient resources
then they are likely to be willing to fund it themselves, especially if they can introduce
associated changes that also reduce costs. Thus, a training course that claims to be able to
significantly reduce lost-time and associated compensation claims, or an engineering change
that also increases productivity, reduces labour costs or improves quality will tend to be viewed
favourably.
However, if an employer is presented with a suggested intervention that has benefits that are not
clear cut, or if they are in financial difficulties, or are financially risk averse, then they are
unlikely to be willing to allow the intervention to happen, even if the costs to them are relatively
minor.
If an employer is funding an intervention, they will have a financial interest in its success. This
will have beneficial effects in creating pressure to implement it successfully, possibly leading to
ad hoc modifications intended to improve it. It may also lead to attempts to portray the
intervention as more successful than it was in reality. However, if there is a perception that an
intervention is failing, then whether or not the perception is justified, there may be a business
decision to terminate it before the study has run its course. This last situation is likely to occur
whether or not the employer is funding the intervention. Decisions will therefore need to be
made before ethics approval is sought as to when and how the outcome of an intervention will
be communicated to the employer and employees concerned.
52 The net effect of this will be to add another potential confounder to the study design. HSE
might therefore wish to explore the possibility of part-funding selected interventions, possibly
on a match-funding basis, with the employer providing the rest of the funding. Further
consideration of this approach, particularly its policy implications, is also beyond the scope of
this work.
4.4.3
Monitoring of project progress
An intervention study is likely to be a project of very significant scale and duration. The
funding bodies will want to monitor progress and spending. It will be impossible to produce
estimates of duration and cost for the whole project before the planning phase is complete.
Therefore, the planning phase should be funded before a firm decision is made about proceeding
with the whole project. It is possible that the project proposal will need to be revised in the light
of concerns raised by the ethics committee that considers it, and therefore commitment to the
full project should be delayed until the proposal has been cleared ethically.
Because of the inevitable, and probably significant uncertainties that there will be at the start of
the project, initial estimates of cost and duration are likely to need revising in the light of the
progress achieved. Rigid limits on expenditure and timing that cannot be adjusted in the light of
the progress of the project are likely to cause problems. If it becomes clear at a project break
point that the project cannot meet its scientific goals then the project should be terminated early.
4.4.4
Identification of intervention sites
One of the most demanding parts of an intervention study is likely to be the logistic one of
recruiting sufficient suitable workplaces, jobs and individuals. A number of criteria will need to
be met:
• Management and workforce are willing to cooperate with the study.
• Sufficient jobs and potential participants are available within the workplace to form a
cluster within the study and to justify implementation of the intervention. Recruiting
very small numbers in each cluster will increase project costs significantly.
• A suitable control cluster can be identified for each participating cluster.
• Comprehensive job analysis shows that the job is suitable for intervention.
• The sector, workplace and workforce are sufficiently stable to allow a reasonable
expectation of low drop out during the follow-up period. Contrasting annual drop out
rates of approximately 43% (Dempsey et al., 2002) and 16% (Lavender et al., 2007)
have been reported. A high turnover may itself be an indication that the job is high risk
for MSDs. Conversely, a successful intervention may result in turnover diminishing
significantly.
Tuncel et al. (2006a) recommended that interventions should be designed and implemented in
accordance with the specific needs of the workplaces where they are happening. This is
consistent with the findings of a study that sought to identify key issues requiring intervention in
the printing industry (Brown et al., 2006). That study argued for an in-depth exploration of “the
working practices, beliefs, and attitudes within a workforce” before the interventions to be
tested are selected.
Given the trend towards automation and to overseas manufacturing, attempting to find UK
workplaces in manufacturing where a common intervention can be implemented across large
numbers of workers doing identical jobs is likely to be very difficult. Other economic sectors
will have larger numbers of workers with similar exposures, particularly in office-based
53 environments. However, if physical risk factors are the target of the intervention, then such
sectors are likely to create problems with a lack of variability of exposure between exposed
individuals. While some specialised workforces, such as emergency services, will have high
exposures, there is also a distinct possibility that there would be very high variability of the
exposures that an individual experiences, and this would also create problems with
measurement and analysis.
It is therefore likely that the only practical solution would be to identify a class of intervention
and implement it appropriately in a range of workplaces. If the intervention were to be an
engineering intervention to, say, optimise the biomechanics of a workstation, then given that
workstations will be of variable quality before the intervention, the target standard of the
intervention would have to be specified so that all workstations in the intervention group were
brought up to the same standard. In this case, the magnitude of the change made would need to
be included in the analysis.
This approach would at least ensure constant post-intervention exposure among the recipients of
the intervention, which would simplify analysis. However, it is inevitable that there would be
considerable variability of the exposure among any control group that did not receive an
intervention. There would also be significant variability in the exposure history of the
intervention group. These confounding factors will need to be taken into consideration at the
design stage.
4.4.5
Identifying suitable interventions
A clear conceptual basis is required for interventions and as mentioned previously (Section 1.5)
the model proposed by Karsh (2006) is a suitable framework. However, defining the
intervention is likely to be difficult, especially if the decision is made to tailor the intervention
to the circumstances in the workplace. Griffiths (1999) gives the example of “Control” as a
psychosocial factor that might be targeted and points out that perceived lack of control is likely
to be the result of a set of factors that are unique to the organisation and also to the moment in
time . She also points out that it is important to seek to establish how change in a variable
causes a change in the response variable.
For an intervention to be accepted in a workplace it has to be one that the management and
workforce are willing to at least try. Therefore, the intervention sites are likely to be ones that
have not recently had similar interventions and perceive problems that a suitable intervention
might address. This is a selection effect and it will need to be taken into account in the design
and analysis. It is therefore possible that such a study would have to be carried out on a
pragmatic “as found” basis.
Another factor that has been identified as relevant to the success of interventions is the role of
“gatekeepers” with access to and control over resources (Stephens et al., 2004). If such
stakeholders are not fully engaged with a project, it is highly likely that it will not be
implemented successfully in their organisations, even if it has been approved or endorsed by
senior managers.
Related to the last point is the cost-benefit balance to the organisation that is a potential location
for the intervention. If significant costs, either direct or indirect, are involved in an intervention
then it will be harder to gain approval for it to be implemented. There will therefore be pressure
to choose interventions that are “low-cost” and “easy” to implement. This may result in
“behavioural” interventions such as training being preferred by employers over engineering
changes. Moreover, there may be costs to the individual employee. If either the employer or
the employees do not see benefits that outweigh the costs, then the intervention may simply not
happen or may be fatally compromised.
54 4.4.6 Recruitment of individual participants
The biggest threat to the study, once approved, is likely to be a problem in recruiting sufficient
participants to give adequate statistical power. The recent CRT by IJzelenberg et al. (2007)
reported that power calculations showed that an initial sample of 350 workers would be required
in nine intervention clusters. With an equal control group, this implies that they were seeking to
ask 700 to participate. In fact, they were only able to invite 590 to participate. There is every
indication that this was a well-planned study. It is therefore demonstrates the difficulties of
such studies that they were unable to ask more than 85% of their relatively modest target to
participate. It is therefore essential that at the planning stage detailed consideration is given to
recruitment and contingency plans put in place should recruitment prove more difficult than
anticipated. In a clustered trial, this is likely to include the inclusion of additional clusters.
Allowance must be made in the power calculations for the likely refusal rates in clusters.
Consideration should be given to the possibility of a multi-centre study, possibly an
international one, which would allow access to much greater populations for recruitment. Such
an approach would create additional management and communication problems, particularly if
subjects are recruited from multiple language groups.
4.4.7 Problems implementing interventions
Workplace interventions are often not straightforward, with weak, inconsistent, or even nonexistent implementation of intervention plans occurring in practical settings (Lipsey, 1996;
McCluskey et al., 2006; Griffiths, 1999). “A pressing problem that has plagued ergonomic
intervention research is the lack of understanding as to why seemingly identical interventions
work in some instances and not in others.” (Karsh et al., 2001) This observation led to the
recommendation that research should pay special attention to the effect of a variety of
implementation approaches to ergonomic interventions to determine the effect that
implementing the intervention in different ways has on the outcomes. In other words, studies of
the implementation process are needed.
4.4.8 Barriers to change and “unforeseen organisational obstacles” to
interventions
“Successful implementation, where the key players are onside and organisational obstacles are
overcome, is difficult to achieve” (McCluskey et al., 2006). Barriers can be cognitive,
behavioural, organisational, socio-cultural or financial (Campbell et al., 2007). The following
factors have recently been reported (Whysall et al., 2006):
• Inability to generate behaviour change among employees due to resistance by
employees or failure by managers to promote behaviour change;
• Getting managerial authorisation and/or commitment, which can involve multiple levels
of approval;
• Managerial perceptions of the importance of tackling MSD, that may result in them
seeing managing or cooperating with an intervention as merely another task to be fitted
into an already overcrowded schedule;
• Management failure to appreciate the value of taking preventative action resulting in
action only happening in response to specific problems;
• Perceptions by management that health and safety initiatives originating from higher up
the management chain reflect badly on their competence;
• Lack of resources, particularly staff time and appropriate skills;
55
• Conflicting priorities of production and health and safety resulting in the potential longterm benefits of the intervention being sidelined by the immediate need to maintain
production at the target rate;
• Problems finding appropriate equipment to implement the intervention;
• Industrial relations issues leading to opposition to the change from the workforce.
These authors also reported a number of other factors that were seen as facilitating the change
process:
• Supportive managers
• Changes in management
• Good awareness of health and safety and/or communication
• Local control of budgets simplifying the approval process
IJzelenberg et al. (2007) found problems in implementing later stages of their intervention. The
initial intervention of training was delivered to 258 individuals. The second stage intervention
of rapid access to physical therapy was used by only 10 of this group and 66 used external
therapists. Of the seven workers that consulted the in-company physical therapist about LBP,
only three used the further option of a workplace examination and consequent ergonomic
adaptation. While IJzelenberg et al. offered a possible explanation in terms of difficulties of
implementation, it is apparent that the failure of the workers to use the in-company provision
was the key difficulty limiting the power of this stage of their study.
There is a danger of changes in workplaces occurring that are unexpected by the researchers and
beyond their control and that can overwhelm a study. These changes may be driven by
commercial or even safety concerns and if driven by a level of an organisation that is not
actively supporting the study, are likely to ignore the needs of the study and hence to cause
major problems.
4.5
DISCUSSION
The example project plan gives an indication of the possible scale, duration, staffing and costs
of a well-planned intervention study. It cannot be emphasised strongly enough that the
estimates are crude and would need significant refining in any tender submission as all of them
would depend on the precise details of the project specification. The nature of the proposed
intervention, its theoretical basis and likely incidence rates will need to be examined in detail.
In particular, the basis on which the follow-up duration is determined in any tender submitted to
HSE should be examined very carefully.
As noted earlier, it is easier to enhance statistical power in a CRT by increasing the number of
clusters, rather than increasing the size of each cluster. However, once a cluster has been
identified then recruiting extra individuals within that cluster is less labour intensive than
identifying another suitable cluster. There is therefore a trade-off between cluster number and
cluster size. The relative advantages of a narrow study of a few large clusters and a broad study
of many small clusters are set out in Table 10. The ultimate balance to be obtained is
impossible to predict at this stage so it will be up to the study designers to specify, in the light of
their power calculations and their knowledge of the likely availability of clusters, the target
number and size of clusters to be sought.
56 Table 10.
Cluster size/number trade-off
Narrow study
Broad study
Target sample size N = 4000
Target sample size N = 4000
Few very large firms, e.g. F=4
Many large firms, e.g. F = 40
8 clusters in matched pairs
80 clusters in matched pairs
Cluster size = 500
Cluster size = 50
Limited range of exposures
Wide range of exposures
Easy to design interventions
Extensive work to design suitable interventions
Inherent variability within clusters so hard to
characterise jobs accurately
Less variability within clusters so easier to
characterise jobs accurately
Limited number of stakeholders
Many stakeholders
Relatively few contact people at intervention sites
Much more management of contacts with
intervention sites and stakeholders
Increasing the follow-up period would be an efficient way of increasing the power of the study.
However, drop-outs would increase with time so sample size would need adjusting to account
for that. Lengthening the follow-up phase would have relatively small impact on the amount of
staff time required to run the project as it would increase the numbers of follow-up contacts
which are relatively low effort. It would also have an impact on the amount of follow-up data
acquired and the consequent effort required to manage, process and interpret it. Therefore, a
trade-off will exist between the sample size and the need to control the duration of the project.
It must be borne in mind that there is no simple relationship between the sample size desired to
give the target power and the cost of implementing the project. While approximately 72% of
the time is allocated to the Data collection phase, a significant proportion of this is assigned to
preparations that would be necessary before an intervention could be implemented. As noted
above, it would be relatively cheap to increase the number of individuals within a cluster or to
extend the duration of the follow-up phase. Therefore increasing or decreasing subject numbers
would have a noticeable effect on effort required at some stages of the project but would have
little or no effect at other stages,
For the suggested design, pairs of matched workplaces would be required for a comparison
between an intervention group and a control group. Blinding would be very difficult in these
circumstances. If matched triplets were available, then comparisons could be made between a
control group, a sham intervention group and an intervention group. Blinding of the sham and
intervention groups would be less difficult than blinding a single intervention group.
It is possible that the staffing of the roles could be adjusted so that team members fulfilled more
than one role or fulfilled different roles as the study progressed. Time estimates are made on the
basis that when a particular role is not required the individual staff concerned would be
occupied on other work so would not expect funding from this project. This would also allow
part-time staff to be involved in the project.
In the example plan, a twelve-month period is allocated to Stages 3.1 to 3.4, which involve
preparation for implementation of the actual intervention at Stage 3.5. Such a lengthy period is
provided because of the need to negotiate access to suitable sites and to ensure that if the aim is
to make engineering changes to the job or the workplace, then the planned interventions are
properly designed and piloted. This period could almost certainly be reduced if it was decided
to implement interventions requiring less preparation, such as organisational or psychosocial
interventions.
57 5
5.1 CHECKLISTS FOR INTERVENTION STUDIES
METHODOLOGICAL CRITERIA FOR EVALUATING OCCUPATIONAL
SAFETY INTERVENTION RESEARCH
Shannon et al. (1999) listed methodological criteria suitable for evaluating occupational safety
intervention research. These have been added to by Karsh et al (2001) who took into
consideration other recommendations. Based on earlier work (Silverstein and Clark, 2004),
Genaidy et al. (2007) have developed the EAI, an “Epidemiological Appraisal Instrument”,
which can be used to evaluate the methodological quality of proposed ergonomic
epidemiological studies. It is designed for evaluating both observational epidemiological
studies and intervention studies. It consists of 43 items forming five measurement scales:
•
Reporting (17 items)
•
Subject/record selection (7 items)
•
Measurement quality (10 items)
•
Data analysis (7 items)
•
Generalization of results (2 items)
The existence of these checklists provides methods by which a proposed study can be evaluated.
The EAI has been developed by a team with expertise in research in both MSDs and
epidemiology and is more comprehensive than the list compiled by Karsh et al. (2001) though
they provide additional material criteria that will be valuable in evaluating any proposal for an
intervention study. Moreover, the EAI team has sought to provide rigorously tested criteria
with detailed specifications for levels of answers for each question. It is therefore
recommended that the EAI is used, with the indicated additions from Karsh et al. (2001), to
evaluate any proposed study in response to a tender by HSE. In order for this to happen,
detailed criteria, similar to those already in the EAI, will need to be developed for the additional
questions.
Table 11 summarises the methodological criteria originally specified by Shannon et al. (1999)
(*) and added to by Karsh et al. (2001) (**). The criteria are cross-referenced to the criteria of
the EAI. Similarly, the detail of the questions to be answered when using the EAI is set out in
Table 12 and cross-referenced to the criteria from Karsh et al. (2001). Inevitably, the criteria in
the two lists cannot be matched precisely due to overlaps, gaps and the use of different
phraseology. The less precise identifications are indicated by question marks. Questions given
by Karsh et al. (2001) that are not in the EAI are indicated by dashes.
In Table 11, for one question (Q 24) under “Statistical analysis”, there is a significant difference
between the wording of Shannon et al. (1999) and Karsh et al. (2001) who rightly remove the
restriction that power and confidence intervals should be calculated only if findings are nonsignificant. One question in the EAI (Question 10) is specific to intervention studies. The
detailed criteria describe some questions as not applicable to intervention studies; these are
indicted by an asterisk against the question number and by Strikethrough of the question text.
This is also done where equivalent criteria are included in the list provided by Karsh et al.
(2001). As the EAI list has been through a much more detailed and rigorous development
process than the list from Karsh et al. (2001), the judgement of the creators of the EAI (Genaidy
et al., 2007) should be accepted.
58 When using the EAI question set it is essential to consult the full paper (Genaidy et al., 2007)
which sets out detailed criteria for assessing each response as “Yes”, “Partial”, “No”, “Not
applicable” or “Unable to determine”.
An example of the detailed criteria is as follows:
1.
Is the hypothesis/aim/objective of the study clearly described?
Yes — clearly described
The objective is clearly stated in one or two statements in the introduction. The relationship to be examined between the exposure/intervention and outcome
variables is clearly stated. Partial — somewhat described
There is sufficient information to be able to infer the objective in the introduction.
The relationship to be examined between the exposure/intervention and outcome
variables has to be inferred. No — Not described
The study objective is not described in the introduction, and there is insufficient information
provided to even ‘infer’.
5.2
CRITERIA FOR REPORTING RANDOMISED TRIALS – THE CONSORT
STATEMENT
Assessment of the quality of the intervention relies on sufficient detail in trial reports, but many
reports provide only superficial descriptions of complex interventions. This poor level of
reporting underlies the conclusions of systematic reviews that the methodological quality of
existing studies is poor. In fact, it is possible that good quality studies have been inadequately
reported, resulting in their evidence being discounted.
In order to improve the quality of reporting, the CONSORT statement (Moher et al., 2001) on
reporting of clinical trials recommends that reports of clinical trials include "precise details of
the interventions intended for each group and how and when they were actually administered."
Interventions should be described in sufficient detail to enable readers to assess if the
intervention was administered well (Genaidy et al., 2007). The CONSORT checklist and
flowchart are reproduced in Table 13 and Figure 5. It is recommended that the CONSORT
statement be used alongside the EAI to evaluate any tender received by HSE. It is further
recommended that it be specified by HSE that the EAI and CONSORT statement and flowchart
will be used to evaluate the progress and final reporting of any intervention project funded or
part-funded by HSE.
In order to provide practical solutions and to be able to interpret the results from ergonomics
intervention studies, efforts should be made to provide dose-response and time-response
relationships as the bases for creating threshold limit values for prevention of MSDs.
59 Table 11.
Intervention evaluation criteria proposed by Karsh et al. (2001)
Karsh et Source Program objectives and conceptual basis
al. (2001)
criterion
EAI criteria
1
*
Were the program objectives stated?
1
2
**
Does the study identify a gap in the existing literature?
—
3
*
Was the conceptual basis of the program explained and sound?
1,4
Study design
4
*
Was an experimental or quasi-experimental design employed instead 4
of a non-experimental design?
5
**
Did the interventions appear to be long enough to find the desired
effect?
38
6
**
Did the subjects have a reasonable amount of exposure to the
intervention?
38?
7
**
Was the intervention powerful enough to cause change?
17?
8
**
Were the subjects randomly selected or volunteers?
23
9
**
Was the outcome assessment blind to the intervention status?
29,30
External validity
10
*
Were program participants/study population fully described?
8
11
*
Was the intervention explicitly described?
2
12
*
Were contextual factors described?
11,12
13
**
Could the reader repeat the study based on the description in the
methods section?
2,4,6,13
14
**
Is the study population representative of the end-user population?
42,43
15
**
Is the study setting representative of the workplace to which the
results will be applied?
42,43
Outcome measurement
16
*
Were all relevant outcomes measured?
—
17
*
Was the outcome measurement standardised by exposure?
40
18
*
Were the measurement methods shown to be valid and reliable?
25,26
19
**
Were baseline measures of the outcome collected?
35
Qualitative data
20
*
Were qualitative methods used to supplement quantitative data?
—
Threats to internal validity
21
*
Were the major threats to internal validity addressed in the study?
23,24,31,32,
36, 37
22
**
Were confounding factors controlled for?
36,37
Statistical analysis
23
*
Were the appropriate statistical analyses conducted?
36,37,39
24
*
If study results were negative, were statistical power or confidence
intervals calculated?
13,17
25
**
Was survivor bias avoided?
9,21
26
**
Were response or recruitment rates adequate?
19
27
**
Was attrition a problem?
—
60 Conclusions
28
*
Did conclusions address program objectives?
—
29
*
Were the limitations of the study addressed?
—
30
*
Were the conclusions supported by the analysis?
—
31
*
Was the practical significance of the results discussed?
42,43
32
**
Are recommendations for workers discussed?
—
33
**
Are areas of further research discussed?
—
Source = *
Source = **
Criterion originally specified by Shannon et al. (1999)
Additional criterion specified by Karsh et al. (2001)
61 Table 12.
The EAI question set with additions from Karsh et al. (2001)
EAI
Study description
criterion
Karsh et
al. (2001)
criteria
1
Is the hypothesis/aim/objective of the study clearly described?
1,2
—
Does the study identify a gap in the existing literature?
20
2
Are all the exposure variables/intervention(s) clearly described?
5,21
3
Are the main outcomes clearly described?
—
4
Is the study design clearly described?
2,3,21
5
Is the source of subject population (including sampling frame) clearly described?
—
6
Are the eligibility criteria for subject selection clearly described?
21
7
Are the participation rate(s) reported? Are ascertainments of record availability
described?
—
8
Are the characteristics of study participants described?
4
9
Have the characteristics of subjects lost after entry into the study or subjects not
30
participating from among the eligible population been described? Have the details of
unavailable records been described?
10+
Have all important adverse effects been reported that may be consequences of the
intervention(s)?
—
11
Are the important covariates and confounders described in terms of individual
variables?
6
12
Are the important covariates and confounders in terms of environment variables
described?
6
13
Are the statistical methods clearly described?
21
14
Are the main findings of the study clearly described?
—
15
Does the study provide estimates of the random variability in the data for the main
outcomes or exposures (i.e. confidence intervals, standard deviations)?
13
16
Does the study provide estimates of the statistical parameters (e.g. regression
coefficients or parameter estimates such as odds ratio)?
—
—
Did the conclusions address programme objectives?
14
—
Were limitations addressed?
15
—
Were the conclusions supported by the results?
16
17
Are sample size calculations performed and reported?
13,28
—
Are recommendations for workers discussed?
24
—
Are areas of further research discussed?
25
Methodological quality
18 Is the comparison/reference group comparable to the exposed/intervention/case
group?
—
19 Is the participation rate adequate? Is the ascertainment of record availability
adequate?
31
—
Were all relevant outcomes measured?
8
—
Were qualitative methods used to supplement quantitative data?
10
20
Are the study subjects from different groups recruited over the same period of time? —
21
Are subject losses or unavailable records after entry into the study taken into
account?
30
22*
Are newly incident cases taken into account?
—
62 23
Are the study subjects randomized to groups?
24
Is the randomized assignment to groups concealed from both subjects and observers 11
until recruitment is complete and irrevocable?
11,29
Measurement quality
25*
Are the exposure variables reliable?
9
26*
Are the exposure variables valid?
9
27*
Are the methods of assessing the exposure variables similar for each group?
—
28*
Is exposure conducted at a time prior to the occurrence of disease or symptoms?
—
29
Are the observers blinded to: subject groupings when the exposure/intervention
assessment was made or the disease status of subjects when conducting exposure
assessment?
33
30
Are the subjects blinded to their grouping when the exposure/intervention assessment 33
was made?
31
Are the main outcome measures reliable?
11
32
Are the main outcome measures valid?
11
33
Are the methods of assessing the outcome variables standard across all groups?
—
34
Are the observations taken over the same time for all groups?
—
Data analysis
35
Is prior history of disease and/or symptoms collected and included in the analysis?
18
—
Was attrition a problem?
32
36
Is there adequate adjustment for covariates and confounders in terms of individual
variables in the analyses?
11,12,19
37
Is there adequate adjustment for covariates and confounders in terms of environment 11,12,19
variables (other than exposure) in the analyses?
38
Is the minimum follow-up time since initial exposure sufficient enough to detect a
relationship between exposure/intervention and outcome?
26,27?
39
Do the analyses adjust for different lengths of follow-up of subjects in
cohort/interventions studies; is the time period between the exposure and the
outcome the same for cases and controls?
12
40*
Are outcome data reported by levels of exposure?
7
41
Are the outcome/exposure data reported by subgroups of subjects?
—
Generalization of results
42
Can the study results be applied to the eligible population?
17,22,23
43
Can the study results be applied to other relevant populations?
17,22,23
63 Table 13.
The CONSORT checklist for reporting a randomised trial
Paper section & topic Item Description
Title & Abstract
1
How participants were allocated to interventions (e.g., "random allocation",
"randomised", or "randomly assigned").
Introduction
2
Scientific background and explanation of rationale.
Methods
Participants
3
Eligibility criteria for participants and the settings and locations where the
data were collected.
Interventions
4
Precise details of the interventions intended for each group and how and
when they were actually administered.
Objectives
5
Specific objectives and hypotheses.
Outcomes
6
Clearly defined
primary and secondary outcome measures and, when
applicable, any methods used to enhance the quality of measurements (e.g.,
multiple observations, training of assessors).
Sample size
7
How sample size was determined and, when applicable, explanation of any
interim analyses and stopping rules.
Randomisation -Sequence generation
8
Method used to generate the random allocation sequence, including details of
any restrictions (e.g., blocking, stratification)
Randomisation -Allocation
concealment
9
Method used to implement the random allocation sequence (e.g., numbered
containers or central telephone), clarifying whether the sequence was
concealed until interventions were assigned.
Randomisation -Implementation
10
Who generated the allocation sequence, who enrolled participants, and who
assigned participants to their groups?
Blinding (masking)
11 Whether or not participants, those administering the interventions, and those
assessing the outcomes were blinded to group assignment. When relevant,
how the success of blinding was evaluated.
Statistical methods
12
Statistical methods used to compare groups for primary outcome(s); Methods
for additional analyses, such as subgroup analyses and adjusted analyses.
Results
Participant flow 13 Flow of participants through each stage (a diagram is strongly
recommended). Specifically, for each group report the numbers of
participants randomly assigned, receiving intended treatment, completing the
study protocol, and analysed for the primary outcome. Describe protocol
deviations from study as planned, together with reasons.
Recruitment
14
Dates defining the periods of recruitment and follow-up.
Baseline data
15
Baseline demographic and clinical characteristics of each group.
Numbers analysed
16 Number of participants (denominator) in each group included in each
analysis and whether the analysis was by "intention-to-treat". State the
results in absolute numbers when feasible (e.g., 10/20, not 50%).
Outcomes and
estimation
17
For each primary and secondary outcome, a summary of results for each
group, and the estimated effect size and its precision (e.g., 95% confidence
interval).
Ancillary analyses
18
Address multiplicity by reporting any other analyses performed, including
subgroup analyses and adjusted analyses, indicating those pre-specified and
those exploratory.
Adverse events
19
All important adverse events or side effects in each intervention group.
Discussion
Interpretation
20
Interpretation of the results, taking into account study hypotheses, sources of
potential bias or imprecision and the dangers associated with multiplicity of
analyses and outcomes.
Generalisability
21
Generalisability (external validity) of the trial findings.
Overall evidence
22
General interpretation of the results in the context of current evidence.
64
Figure 5.
The CONSORT flowchart Aug. 2005
Assessed for eligibility
(n= )
Excluded (n= )
Enrolment
Is it Randomized?
Allocated to intervention
(n= )
Received allocated intervention
(n= )
Did not receive allocated intervention
(n= )
Give reasons
Lost to follow-up (n=
Give reasons
)
)
Follow-Up
)
Analysis
65
)
Discontinued intervention (n=
Give reasons
Analyzed (n=
)
Excluded from analysis (n=
Give reasons
Allocated to intervention
(n= )
Received allocated intervention
(n= )
Did not receive allocated intervention
(n= )
Give reasons
Lost to follow-up (n=
Give reasons
Discontinued intervention (n=
Give reasons
Analyzed (n=
Allocation
Not meeting inclusion criteria
(n= )
Refused to participate
(n= )
Other reasons
(n= )
)
)
Excluded from analysis (n=
Give reasons
)
5.3
EXAMPLES OF POSSIBLE STUDY PROPOSALS
5.3.1
Introduction
This section gives some examples of possible study designs.
It starts by listing
inclusion/exclusion criteria that can be used to select the MSDs that are to be studied. It then
lists the types of workplace interventions available for reducing workplace exposure to MSD
risk factors.
Table 14 lists details of possible primary intervention types and gives specific examples of each
type. It also gives example measurement methods and specific objectives that could be used to
evaluate the effectiveness of each intervention component.
The focus of the possible tender will be on primary workplace interventions aimed at preventing
musculoskeletal pain, injury and disability. Interventions aimed at not only the primary level
but also at secondary and/or tertiary levels are also within scope. Interventions only targeted at
the secondary (return to work) or tertiary (treatment) levels should be excluded.
Before any specific intervention is selected for implementation, the existing literature and other
evidence about its effectiveness should be considered.
Tables 15, 16 and 17 give three examples of possible specific protocols that could be
implemented. Table 16 is a mixed secondary and primary prevention study. Table 18 gives
details of features that would be common to all three protocols.
5.3.2
Inclusion/exclusion criteria for MSDs
Possible work-related musculoskeletal disorders that could be studied are:
• Low back pain, including disk prolapses, sciatica and other forms of nerve root pain,
and simple mechanical low back pain;
• Upper limb disorders, such as Carpel Tunnel Syndrome, Epicondylitis, Rotator Cuff
Syndrome and Non-Specific Arm Pain;
• Lower limb problems such as bursitis;
• Problems associated with seated postures;
• Problems associated with whole-body vibration.
The following musculoskeletal disorders should be excluded from any studies:
• Back problems due to non-work related causes/pre-existing medical conditions such as
arthritis, ankylosing spondylitis, spondylisthesis, traumatic impacts or pregnancy.
• Hand-arm vibration syndrome;
• Whiplash and other impact injuries to the neck;
• Traumatic tendon/ligament injuries in the lower limb such as sports injuries to the
anterior cruciate ligaments.
5.3.3
Type of interventions under consideration
The types of ergonomics interventions that have been carried out in the past to reduce the
exposure in the workplace to risk factors for MSDs have been categorised into six groups
(Tuncel et al., 2006a; Karsh et al., 2001; Volinn, 1999). A multiple component intervention
would combine two or more interventions or types of intervention.
66 • Organisational environment changes, including psychosocial interventions, changes to
work organisation/methods and participatory approaches;
• Job design (ergonomics) including engineering redesigns and providing assistive
devices;
• Job placement/worker selection;
• Education/training (including providing advice on ergonomics);
• Physical exercise (including work hardening);
• Back supports.
67
Table 14.
Intervention
type
Possible primary interventions to reduce the incidence of episodes
of MSDs
Intervention details/examples Example baseline and post Specific outcome
intervention measurements, (measurable aim) =
(implementation, follow-up Objective of intervention
or prospective event data)
• Agreement between labour • Effectiveness of
All
intervention
interventions and management on best
interests of employees
implementation
• Pain scales/diagrams
• Reports of musculoskeletal
trouble
• Clinical examination
• MSD injury rates
• Time to injury
• Injury severity
• MSD absence rates
• Absence duration
• Staff turnover
• Intervention implemented
according to plan
• Reduced exposure to MSD
hazards
• Reduced incidence of
clinical diagnoses
• Reduced pain
intensity/duration
• Reduced prevalence of
musculoskeletal trouble
• Reduced injury rates
• Reduced recurrence rates
• Increased survival time
• Decreased injury severity
• Reduced absence rates
• Reduced days lost
• Reduced disability
• Reduced medical care
costs
• Reduced compensation
claims
• Reduced staff turnover
Training/edu • Understanding of body
posture, anatomy and
cation biomechanics
• Training in ergonomic risk
analysis
• Correcting erroneous
beliefs
• Training in lifting, job and
work techniques • Posture • Understanding of risk
factors
• Quality of risk
assessments
• Techniques in normal use
•
•
•
•
• Understanding of correct
use of PPE
• Acceptability of PPE to
users
• Ongoing use of PPE
• Posture
• Repetitive movements
• Number of manual
handling operations
• Lifting index or similar
metrics
• Joint loading/percent
capable
• MAC red scores
• Improved body posture
• Elimination of some
handling operations
• Reduced repetition &
forces
• Reduced manual handling
operations
• Reduced daily physical
workload
PPE
• Back belts
• Wrist supports
Engineering • Reduce load moment
• Mechanise handling
redesigns
• Reduce frequency
• Decrease unit load and
increase frequency
• Reduce vertical movement
• Reduce force requirements
• Improve postures – move
force requirements to
stronger muscle groups
• Replace carrying with
pushing/pulling
68 Improved posture
Increased knowledge
Improved risk assessments
Altered attitudes to and
beliefs about pain
• Increased job-specific
skills
Changes in • Eliminate “job and finish”
• Rebalance machine paced
work
tasks
organisation
• Rotate employees among
tasks with different
exposures
• Alter patterns of rest
breaks
• Introduce work pauses:
passive, active, diverted
• Repetitiveness of tasks
• Time spent on each task
and number of tasks
• Total exposure
• Actual working patterns
• Increased variation in jobs
• Ongoing implementation
of job rotation
• Ongoing use of rest breaks
and compliance with
changes
Psychosocial • Improve management
commitment to health and
safety
• Improve communication
about health and safety
• Improve reporting
mechanisms for health and
safety problems,
particularly MSDs
• Reduce excessive demands
• Increase worker control
• Improve worker support
• Perceptions of
management commitment
• Use made of reporting
mechanisms
• General Health
Questionnaire (GHQ)
• Psychosocial factor
questionnaires
• Other psychological scales
• Improved perceptions of
management commitment
to health and safety
• Reporting mechanisms in
place and seen to be
effective.
• Improved psychological/
psychosocial status
• Increased job satisfaction
•
•
•
•
• Increased strength
• Improved flexibility
• Decreased resting heart
rate
• Increased estimated
maximal oxygen
consumption
• Decreased percent body fat
• Decreased body mass
index
Exercise
• Implement coordination,
strength, & aerobic fitness
programme
• Provide progressive
resistance exercise
programme
• Encourage active microbreaks with stretching
exercises
Health
advice/care
to prevent
recurrence
• Teach pain management • Worker expectations of
• Reduced levels of fear,
techniques
interventions and beliefs
pain & pain-related
about MSD prognosis
disability/fear
• Spread “remain active”
message
• Immediacy of care-seeking • Improved understanding of
keep active message
• Encourage return to
normal activities as
• Reduced catastrophic
quickly as possible
thinking and depression
• Provide mini back schools
• Improved functional status
• Provide easy access to
• Uptake of mini back
health care providers.
schools
• Increased use of health
care providers
Muscle strength
Joint flexibility
Resting heart rate
Estimated maximal
oxygen consumption
• Percent body fat
• Body mass index
69
Table 15.
Preventing ULD problems in newly employed workers
Objective:
To reduce the incidence of musculoskeletal disorders affecting the upper limb
(ULDs) amongst workers in the first 6 months of employment
Study design
Longitudinal experimental prospective case cross-over design with randomisation
Follow-up for 12 months (each individual)
Individuals can enter at any time (dynamic)
Target recruitment period: 1-2 years
Length of study: 3-4 years? (Depending on sample size and rate of new jobs)
Intervention type
Multi-component
PPE
Engineering redesigns
Administrative tools
Psychosocial interventions
Measurement/
outcome
Clinical examination and diagnoses
Incident rate of injuries
Pain scores
Amount of lost/restricted time
Subject selection/
entry criteria
Exclude subjects with musculoskeletal trouble in the previous week
Exclude subjects with musculoskeletal trouble lasting longer than 24 hours in the
previous month
Exclude subjects with ‘disability’ due to musculoskeletal trouble in the previous
three months
No previous exposure to similar types of work (preferable)
Upper age limit
Exclude those returning to the job or those with experience in similar jobs
Control group
selection/entry
criteria
Subjects who did the same or a similar job within the previous 12 months (> two
years of work-experience?)
Or case-crossover (subject acts as own control in second half of study)
Possible target
employment
sectors
Army, fire-fighters
Newly opened workplaces
Keyboard workers – call centres or large offices
Study strengths Reduced likelihood of subjects having a history of MSD problems
Using clinical diagnoses (as opposed to relying on subjective measurements)
Subjects act as own controls (reduces confounding)
Study weaknesses Sites with large numbers of new recruits are likely to have high staff turnover
A study involving clinical diagnoses will be more expensive than one relying on
self-reports
It is difficult to compare working environment and jobs for new recruits (except
at new sites) as companies rarely do mass new recruiting
Relevant references (Breslin and Smith, 2006; Hakkanen et al., 2001a; 2001b; Harkness et al., 2003a;
2003b; Heuer et al., 1996; Jones et al., 2006; Macfarlane et al., 1997; Melhorn et
al., 1999; 2001; Nahit et al., 2001a; 2001b; Parenmark et al., 1988; Park et al.,
1994; 1996; Thompson et al., 1951; Waersted and Westgaard, 1991; Westgaard
and Aaras, 1984)
70 Table 16.
Preventing disability through return to work interventions
Objectives: To prevent disability through return to work (RTW) intervention
To reduce the severity and frequency of episodic lower back pain (LBP)
To reduce the incidence of LBP re-occurrence
To prevent a progression from an acute condition to chronic disability)
To reduce pain and disability and decrease period before return to work.
Study design Cohort (experimental longitudinal) or RCT (single blind)
Follow-up for 3-5 years (each subject)
Subjects can enter at any time (dynamic) – but when they go off work with LBP or
within a fixed period (i.e. 2 weeks) of going on sick leave.
Intervention type
Clinical intervention: Cognitive-behavioural interventions during rehabilitation
(within 2 months of reporting sick) until worker makes full return to regular work.
Occupational (ergonomics) intervention: Provide tailored workplace redesign
interventions
Measurement/
outcome
Baseline clinical diagnoses and quarterly clinical assessments
Industrial records/Days off work (sick leave)
Questionnaires (self-assessment)
Return to Work perception survey
Pain (intensity), discomfort and disability
Psychosocial factors
Health care utilization
Physical function
Subject selection/ Exclude subjects with constant pain at baseline
entry criteria Include subjects with recent history of low back disorders (≥2 episodes within last
12 months) with diagnosed occupational back pain
Subject enters study when absent from work for ≥24 weeks due to LBP
Control group
selection/entry
criteria
Subjects with occupational low back pain without intervention
Possible target
employment
sectors
Airline industry
Police force
Study strengths
Using clinical diagnoses (as opposed to relying on subjective measurements)
Long duration follow-up: few studies have looked at effects beyond 1 year
The minority of LBP patients with long-duration work absenteeism account for a
large proportion of socio-economic burden of LBP.
RTW is a complex social phenomenon, not well characterized by measures
collected at a single point in time.
Few studies have addressed the impact of interventions targeting outside
individual psychosocial risk factors.
Study weaknesses More expensive involving clinical diagnoses
Difficulties in comparing/matching different industries
Need to recruit psychologists to provide cognitive-behavioural interventions
Relevant references (Anderson, 1987; Elders et al., 2000; Franche et al., 2005a; 2005b; Lagerstrom et
al., 1998; Loisel, 2005; Loisel et al., 2005; Mahmud et al., 2000; Martocchio et
al., 2000; Mayer et al., 2001; Ostelo et al., 2003; Pransky et al., 2005; Scheer et
al., 1997; Snook, 2004; Staal et al., 2002; 2005; Sullivan et al., 2005; Troup and
Videman, 1989; van den Heuvel et al., 2005; Von Korff and Saunders, 1996;
Wasiak et al., 2003)
71 Table 17.
Prevention of LBP among high mileage drivers
Objectives:
Primary prevention of episodes of LBP among individuals driving high
mileages as part of their work
Study design
Cluster randomised trial
Follow-up for two years
Subjects recruited as part of clusters
Target recruitment period: 3-6 months
Length of study 2-3 years (depending on sample size and number of clusters)
Intervention
type
Provision of customised specialist seats
Provision of anti-vibration seating
Postural advice/training
Advice on rest breaks
Advice on manual handling
Provision of manual handling aids for delivery drivers
Measurement/
outcome
Clinical examination/diagnosis
Subject
selection/entry
criteria
Exclude drivers with existing LBP
Exclude drivers with LBP within the previous 12 months
Exclude drivers commuting more than 50 miles per week between home and
work
Drivers of work vehicles travelling more than 500 miles per week or for more
than 12 hours per week
Include drivers of a range of vehicles from cars to HGVs
Control group
selection/entry
criteria
Drivers in matched occupational clusters
Possible target
employment
sectors
Businessmen
Taxi/private hire drivers
Light goods van drivers
HGV and PSV drivers
Study strengths
Using clinical diagnosis
Study
weaknesses
Clinical examinations are a more costly measurement method
Relevant
references
(Chen et al., 2005; Gyi and Porter, 1998; Magnusson et al., 1996; Pope et al.,
2002; Porter and Gyi, 1995; 2002; Okunribido et al., 2006; Skov et al., 1996;
van der Beek et al., 1994)
72 Table 18.
Common features of the three example proposals
Measuring tools
Self-reported frequency of symptoms
Self-reported work restriction/loss
Company books (days off work)
Clinical visits
Baseline clinical examination
Medical history
Physicians diagnosis of MSDs
Main outcome measure
• Measured in comparison with baseline clinical
diagnoses
• Questionnaires (self-assessment)
• Incident cases of MSDs
• Psychosocial factors
Assess effectiveness /accuracy of subjective
questionnaires
Secondary outcomes Confounders Age
Gender
Anthropometry
Cigarette smoking
Non-occupational physical activities
Healthy worker effect
Main risk factors to study Failure to recruit
Failure to implement interventions
Loss to follow-up;
Study contamination,
Sample size (90% power and 5% significance
level)
Standard deviation
Mean difference
N (considering response rate)
Analysis Survival analysis
73
6
6.1
DISCUSSION/RECOMMENDATIONS
INTRODUCTION
There is a view that primary interventions to prevent initial onset of MSDs are not likely to be
effective because of the very high lifetime prevalence of MSDs. On this view, the concentration
should be on secondary interventions to return injured individuals to work as quickly as
possible, despite them still suffering pain, particularly with the aim of preventing long-term
work loss and disability (Burton et al., 2004; 2005).
However, this worldview has a number of inherent difficulties. Firstly, it is counterintuitive to
tell people that the pain they are feeling doesn’t really matter, especially when the associated
guidance is very vague as to the use of pain-relief to control the pain and when first-aid advice
on dealing with new episodes of MSDs is effectively non-existent. Secondly, it is perfectly
rational for an individual suffering pain to avoid situations or activities that they think could
exacerbate or cause a new episode of the pain. The whole system of risk assessment and control
of manual handling risks is based on the premise that MSD causing situations can often be
eliminated or at least ameliorated. The third problem with the early return to work message is
that it denies by implication the value of workplace modification in prevention but then
explicitly calls for workplace modification to ease reintegration of injured workers. Such a
gloomy prognosis of the possibility of prevention is not universal, with, for example, Volinn
(1999) noting that rapid progress in medicine has been associated with explanatory studies. He
therefore argued that despite most previous studies of workplace MSD risk factors having been
pragmatically oriented (i.e. observational), “explanatory workplace intervention studies may
come to prevail and, assuming they do, rapid progress in preventing low back disorders may be
expected”.
Evidence from such a study or studies would add to the body of knowledge about the prevention
of episodes of MSDs. It could then be integrated with the other knowledge via meta-analysis
and systematic reviews and then influence guidance and standards for the prevention of MSDs.
Burdorf has been calling for some time for a shift towards intervention research and is involved
in carrying it out (Burdorf, 2007; Burdorf et al., 1997a; 1997b; IJzelenberg et al., 2007). An
article by Dempsey (2007) to which Burdorf (2007) refers is more circumspect, pointing out the
practical difficulties involved. Dempsey’s experience in attempting an epidemiological
evaluation of the NIOSH lifting equation (Dempsey et al., 2002; Dempsey, 2002) clearly has
influenced his thinking. Despite this understandable caution, there is a demand within the
ergonomics scientific community for such trials. However, the individuals who are calling for
such studies are, with the exception of Burdorf, approaching the issue from a technical
ergonomics end rather than as epidemiologists. If Burdorf, who is an epidemiologist/statistician
with considerable experience in the MSD field, was not calling for intervention studies, the
inclination would be to highlight comments betraying limited knowledge of the epidemiological
methods required and to argue that the practical difficulties are so large as to make the success
of an intervention study very doubtful.
However, there are worldwide efforts to carry out these studies. Moreover, it is the nature of
science that when a previously unaddressed area becomes topical or potentially solvable, then
multiple research groups will be working on it at the same time. It is also unlikely that one
single intervention study will prove to be definitive, however large and comprehensive it is.
74 6.2
STUDY DESIGN
In principle, it would be possible to carry out a cohort study of initial onset of MSDs by
following a cohort from birth to early adulthood but this would be a very long-term project.
Given a pragmatic acceptance that back pain is episodic (Eisen, 1999; Burdorf and van der
Beek, 1999b; Cassidy et al., 2005) and that the majority of cases resolve with conservative
treatment in a relatively short time frame, then a credible aim would be the prevention of new
episodes in individuals who have been symptom free for a period long enough to demonstrate
complete recovery from any previous episodes. (This may require close examination of the
value of a previous episode of back pain as a predictor of a subsequent one to see if the
relationship weakens with time). Such an approach would require a careful back pain history to
be taken from each individual that might conveniently be combined with a detailed clinical
examination.
Any robust intervention study needs careful design and thorough planning. Implementation will
require dedication and relevant experience from the study managers in order to overcome the
significant logistic and organisational challenges that it will face. The ethical issues must be
addressed to the satisfaction of an ethical committee that will include lay members representing
the public as well as scientists and topic experts.
In order to allow criteria for assessing causality to be met, it is necessary to use an experimental,
as opposed to observational, study design to look at interventions to prevent MSDs. This means
that studies must be prospective and longitudinal, not cross-sectional or retrospective.
Interventions are most easily implemented at group level (e.g., work team, factory or company
level) and so the best study design would be a clustered randomised trial.
There are so many variables to be considered that it would be counterproductive to specify at
this stage a precise study design listing target MSDs, industry sectors to be involved,
intervention characteristics, follow-up periods, or subject numbers. Instead, it is recommended
that HSE write any tender document in such a way as to ensure that potential contractors have to
demonstrate that their proposals are scientifically sound while giving them sufficient flexibility
to produce innovative proposals that will overcome the practical problems. The example project
plan in Section 5 and the three outline protocols in Section 5.3 are written from this viewpoint.
6.3
MAIN RECOMMENDATION
The conclusion of this feasibility study is that there are very significant practical obstacles to the
successful execution of a scientifically robust study to demonstrate the effectiveness of an
ergonomics intervention to prevent the onset of episodes of musculoskeletal disorders.
However, because of the scale of the MSD problem, there are important calls for such studies to
be done despite the difficulties.
1. It is therefore recommended that HSE should consider funding or part funding a
longitudinal study designed to test the effectiveness of interventions designed to prevent
the onset of new episodes of MSDs, and that consideration be given to making it a multicentre, possibly international, study.
2. It is also recommended that HSE write any tender document in such a way as to ensure
that potential contractors have to demonstrate that their proposals are scientifically sound
while giving them sufficient flexibility to produce innovative proposals that will
overcome the practical problems.
A hard-headed approach is needed to the carrying out of such a study to maximise the chances
of it being successful. A study that demonstrates either the effectiveness or ineffectiveness of
75 such interventions will have succeeded. A study that fails to produce clear results due to
methodological limitations or insufficient power would have been better unattempted.
3. It is recommended that any study that HSE funds have an integrated management team to
lead the project throughout its life. At the minimum, this would need to consist of an
epidemiologist experienced in intervention studies, an ergonomist with significant
understanding of MSDs, and an experienced project manager. No single individual is
likely to be able to fill all of these roles in any large-scale project. It is also recommended
that an occupational psychologist or similar person familiar with the psychosocial and
biopsychosocial aspects of MSDs be involved throughout the study.
4. It is recommended that the viability of any project should be rigorously reviewed,
particularly at the design stage, and throughout the subject recruitment phase. If it
becomes clear at any of these stages that the project has a low probability of success then
it should be terminated.
5. The recommended study design is a longitudinal cluster randomised trial because it is a
natural design to implement in workplaces. This design will require control clusters to
which the intervention is not implemented. Case-control and cross-sectional studies
cannot demonstrate causality and are completely inappropriate. Analysis methods will
need to be selected at the study design stage.
6.4
DETAILED RECOMMENDATIONS
1. Any tender specification should use the integrated model proposed by Karsh (2006) as a
basis for identifying potential intervention points that should be explored.
2. Proposals in response to the tender should be evaluated against their ability to test the
effectiveness of interventions at these points. At this stage, use should be made of
methods for mapping causal relationships.
3. Any tender specification should require that any proposal is written against the criteria of
the Epidemiological Assessment Instrument (EAI) (Genaidy et al., 2007) augmented by
the additional material in the Karsh et al. (2001) checklist. In order for this to happen,
detailed criteria, similar to those already in the EAI, will need to be developed for these
additional questions. Tenders will need to demonstrate awareness of the requirements of
the CONSORT statement on the reporting of RCTs (Moher et al., 2001).
4. When comparing tenders received, HSE should compare the amount of staff time
allocated in order to form a judgement on the ability of the project team to deliver the
specified project.
5. Any tenders received should be compared with the outline project plan in Section 5 to
ensure that all necessary activities are included. It should be borne in mind that tenders
could differ in detail from the example outline, particularly in relation to sequencing or
overlapping of project activities.
6. Multi-component/multi-factorial study designs that attempt to intervene in several ways
should be favoured.
7. Known risk factors should be the targets of the intervention research and doses, responses
and capacity factors should be measured to the extent possible. Possible interactions
between risk factors need to be considered.
76 8. All study design decisions and power calculations will need to involve a statistician and
must be documented. The PASS 2005 software package can be recommended for
performing power calculations.
9. Except where completely impossible, any intervention should be piloted to check that it
can be implemented successfully.
10.The study will need to measure not only lost-time due to MSDs, but will also need a
hierarchy of case definitions to capture the range of possible adverse outcomes that do not
lead to lost-time. A system for immediate reporting of incidents alongside a regular
follow-up system is recommended in order to increase the probability of capturing
incident events.
11.The study will need to control for different workplace cultures, and different absence
management expectations and milieus. If a multi-national study is chosen then cultural
differences will need to be considered at the design stage.
12.It is recommended that economic evaluations are carried out when choosing the
interventions to be attempted and in the final analysis of the results of the study. While a
variety of methods are available (Korthals-de Bos et al., 2006), it is likely that cost-utility
analysis and cost-benefit analysis would be suitable. Both broad societal perspectives
and the narrow perspective of the employer would need to be considered.
13.Once the study is complete, careful attention must be paid to proper reporting of the
study. This will aid interpretation, application and future meta-analysis. Results should
be presented in at least two stages; firstly with an initial report describing and
summarising the data, and secondly presenting the results from all statistical analyses. It
is recommended that any tender specification state that the CONSORT checklist and flow
diagram (Moher et al., 2001) will be used alongside the augmented EAI (Genaidy et al.,
2007; Karsh et al., 2001) to evaluate not only the initial tenders but also the progress and
final reporting of the project.
77 7
GLOSSARY
Term
Definition
Aetiology The factors that cause a disease or health problem
Attributable Fraction
(AF) The proportion of incident cases that can be attributed to a particular exposure or risk factor. Association A relationship between two variables so that a change in one is linked to a change in the second. Bias The effect of external or uncontrolled factors on the conclusions of a study. Blinding Allocation of subjects to a treatment condition so that they, or they experimenter, or both, are unaware of whether the treatment is the active intervention or the placebo. Causality
The determination if a change in one variable causes a change in a second
variable. Confounding
A confusion of effects so that the apparent effect of an exposure is
distorted by an extraneous factor. Differential attrition
A difference in drop-out rates between intervention and control groups
due to factors extraneous to the study. Ecological study
A study that compares results from groups rather than individuals. Ecological fallacy
Assuming that the results of an ecological study apply to all individuals
within the group
Effect size
The size of a difference in outcome measures, e.g., between an intervention group and a control group.
Healthy worker effect
The tendency for a workforce to consist of “survivors” who can cope with
the exposures in the workplaces. They remain after other individuals have left the workforce due to morbidity or mortality. Incidence rate
The frequency of new cases of the outcome of interest. Intention to treat analysis
Analysis of outcomes based on how subjects were assigned to
intervention or control groups, rather than on the basis of the actual intervention received.
Intervention study
A study where the exposure to a risk factor is manipulated to examine its effect. Matching
Allocation of subjects to groups so that individuals or groups are matched for certain variables, such as age and gender. Non-specific low back
pain
Low back pain that cannot be attributed to a specific pathology or lesion. Non-specific arm pain
Arm pain that cannot be attributed to a specific pathology or lesion. Nocebo effect
A negative health outcome associated with negative beliefs about the effects of an inactive treatment. Placebo
An inactive or ineffective treatment or intervention given to a control group. Placebo effect
A positive health outcome associated with positive beliefs about the effects of an inactive treatment. Power
The ability of a statistical test to detect a genuine difference or effect Prevalence rate
The frequency of existing cases of the outcome of interest Randomisation
Allocation of subjects to study conditions so that the probability that each subject is allocated to a particular condition is known.
78
Randomised controlled
trial
A longitudinal study where subjects are randomised to control and
intervention groups.
Statistical significance An observed outcome being less probable than a pre-determined level,
typically 5%.
Stratification
Creating subject groups that fit into separate levels or categories on a
particular variable, such as age.
79 8
ACRONYMS AND ABBREVIATIONS
Acronym/abbreviation
Meaning
CCT
Controlled Clinical Trial
CRT
Clustered Randomised Trial
CV
Coefficient Of Variation
DAG
Directed Acyclic Graph
DHHS
Department of Health and Human Services
EAI
Epidemiological Appraisal Instrument
FOM
Faculty of Occupational Medicine
GEE
Generalized Estimating Equations
HSE
Health and Safety Executive
HSL
Health and Safety Laboratory
ICC
Intra-cluster Correlation Coefficient
IOM
Institute of Occupational Medicine
LBD
Low back disorders
LBP
Low back pain
MRC
Medical Research Council
MSD
Musculoskeletal Disorder
NIOSH
National Institute for Occupational Safety and Health (U.S.)
NORA
National Occupational Research Agenda
NRC
National Research Council
NMQ
Nordic Musculoskeletal Questionnaire
OR
Odds Ratio
PHM
Proportional Hazards Model
RCT
Randomised Controlled Trial
WMA
World Medical Association
WMSD
Work-related Musculoskeletal Disorder
WRMSD
Work-Related Musculoskeletal Disorder
80 9
REFERENCES
1 Moon, S. D. and Sauter, S. L. (Ed.) (1996). Beyond Biomechanics: Psychosocial Aspects
of Musculoskeletal Disorders in Office Work. (London: Taylor & Francis)
2 Carter, J. T. and Birrell, L. N. (Ed.) (2000). Occupational health guidelines for the
management of low back pain at work. Evidence, review and recommendations. (London:
Faculty of Occupational Medicine of the Royal College of Physicians)
3 Aas, R.W., Holte, K.A. and Moller, A. (2005). Worksite intervention for neck and back
disorders in workers. Cochrane Database of Systematic Reviews: Protocols
4 Altman, D. (1991). Practical Statistics for Medical Research. (London: Chapman & Hall/CRC), 1st Edition. 5 Altman, D.G. and Bland, J.M. (1999). Statistics notes. Treatment allocation in controlled
trials: why randomise? British Medical Journal, 318, (7192), 1209
6 Altman, D.G., Schulz, K.F., Moher, D., Egger, M., Davidoff, F., Elbourne, D., Gotzsche,
P.C. and Lang, T. (2001). The revised CONSORT statement for reporting randomized
trials: explanation and elaboration. Annals of Internal Medicine, 134, (8), 663-694.
7 Anagnostis, C., Gatchel, R.J. and Mayer, T.G. (2004). The Pain Disability Questionnaire:
A New Psychometrically Sound Measure for Chronic Musculoskeletal Disorders. Spine,
29, (20), 2290-2302.
8 Anderson, J.A. (1987). Dorsopathies. Baillieres Clinical Rheumatology, 1, (3), 561-582.
9 Armstrong, T.J., Buckle, P., Fine, L.J., Hagberg, M., Jonsson, B., Kilbom, A., Kuorinka,
I., Silverstein, B.A., Sjogaard, G. and Viikari-Juntura, E.R. (1993). A conceptual model
for work-related neck and upper-limb musculoskeletal disorders. Scandinavian Journal
of Work, Environment and Health, 19, (2), 73-84.
10 Badii, M., Keen, D., Yu, S. and Yassi, A. (2006). Evaluation of a Comprehensive
Integrated Workplace-Based Program to Reduce Occupational Musculoskeletal Injury
and Its Associated Morbidity in a Large Hospital. Journal of Occupational and
Environmental Medicine, 48, (11), 1159-1165.
11 Begg, C., Cho, M., Eastwood, S., Horton, R., Moher, D., Olkin, I., Pitkin, R., Rennie, D.,
Schulz, K.F., Simel, D. and Stroup, D.F. (1996). Improving the quality of reporting of
randomized controlled trials. The CONSORT statement. Journal of the American
Medical Association, 276, (8), 637-639.
12 Bergstrom, G., Bodin, L., Bertilsson, H. and Jensen, I.B. (2007). Risk factors for new
episodes of sick leave due to neck or back pain in a working population. A prospective
study with an 18-month and a three-year follow-up. Occupational and Environmental
Medicine, 64, (4), 279-287.
13 Bernard, B. (1997). Musculoskeletal Disorders and Workplace Factors. A Critical
Review of Epidemiologic Evidence for Work-Related Musculoskeletal Disorders of the
Neck, Upper Extremity, and Low Back. (Cincinnati, Ohio: NIOSH), DHHS (NIOSH)
Publication No. 97-141
81 14 Bonfiglioli, R., Mattioli, S., Fiorentini, C., Graziosi, F., Curti, S. and Violante, F.S.
(2007). Relationship between repetitive work and the prevalence of carpal tunnel
syndrome in part-time and full-time female supermarket cashiers: a quasi-experimental
study. International Archives of Occupational and Environmental Health, 80, (3), 248253.
15 Bongers, P.M., Kremer, A.M. and Laak, J. (2002). Are psychosocial factors risk factors
for symptoms and signs of the shoulder, elbow, or hand/wrist? A review of the
epidemiological literature. American Journal of Industrial Medicine, 41, (5), 315-342.
16 Boocock, M.G., McNair, P.J., Larmer, P.J., Armstrong, B., Collier, J., Simmonds, M. and
Garrett, N. (2007). Interventions for the prevention and management of neck/upper
extremity musculoskeletal conditions: a systematic review. Occupational and
Environmental Medicine, 64, (5), 291-303.
17 Bos, E., Krol, B., Der Star, A.V. and Groothoff, J. (2006). The effects of occupational
interventions on reduction of musculoskeletal symptoms in the nursing profession.
Ergonomics, 49, (7), 706-723.
18 Bouter, L.M., Pennick, V. and Bombardier, C. (2003). Cochrane back review group.
Spine, 28, (12), 1215-1218.
19 Bowling, A. (2002). Research methods in health: investigating health and health
services. (Buckingham: Open University Press), 2nd Edition.
20 Breslin, F.C. and Smith, P. (2006). Trial by fire: a multivariate examination of the
relation between job tenure and work injuries. Occupational and Environmental
Medicine, 63, (1), 27-32.
21 Brewer, S., van Eerd, D., Amick III, B.C., Irvin, E., Daum, K.M., Gerr, F., Moore, S.J.,
Cullen, K. and Rempel, D. (2006). Workplace interventions to prevent musculoskeletal
and visual symptoms and disorders among computer users: A systematic review. Journal
of Occupational Rehabilitation, 16, (3), 317-350.
22 Brown, T.P., Rushton, L., Williams, H.C. and English, J.S.C. (2006). Intervention
development in occupational research: an example from the printing industry.
Occupational and Environmental Medicine, 63, (4), 261-266.
23 Burdorf, A. (2007). Invited Point of View: The art of conducting workplace intervention
studies. International Journal of Industrial Ergonomics, 37, (2), 175-176.
24 Burdorf, A., Rossignol, M., Fathallah, F.A., Snook, S.H. and Herrick, R.F. (1997a).
Challenges in assessing risk factors in epidemiologic studies on back disorders. American
Journal of Industrial Medicine, 32, (2), 142-152.
25 Burdorf, A., Sorock, G.S., Herrick, R.F. and Courtney, T.K. (1997b). Advancing
epidemiologic studies of occupational injury - approaches and future directions.
American Journal of Industrial Medicine, 32, (2), 180-183.
26 Burdorf, A. and van der Beek, A. (1999a). Exposure assessment strategies for workrelated risk factors for musculoskeletal disorders. Scandinavian Journal of Work,
Environment and Health, 25, (Suppl 4), 25-30.
82 27 Burdorf, A. and van der Beek, A.J. (1999b). In musculoskeletal epidemiology are we
asking the unanswerable in questionnaires on physical load? Scandinavian Journal of
Work, Environment and Health, 25, (2), 81-83.
28 Burton, A.K., Balague, F., Cardon, G., Eriksen, H.R., Henrotin, Y., Lahad, A., Leclerc,
A., Muller, G. and van der Beek, A.J. (2004). European Guidelines for Prevention in
Low Back Pain. (Brussels: European Commission Research Directorate General), 53
pages.
29 Burton, A.K., Balague, F., Cardon, G., Eriksen, H.R., Henrotin, Y., Lahad, A., Leclerc,
A., Muller, G. and van der Beek, A.J. (2005). How to prevent low back pain. Best
Practice & Research - Clinical Rheumatology, 19, (4), 541-555.
30 Campbell, N.C., Murray, E., Darbyshire, J., Emery, J., Farmer, A., Griffiths, F., Guthrie,
B., Lester, H., Wilson, P. and Kinmonth, A.L. (2007). Designing and evaluating complex
interventions to improve health care. British Medical Journal, 334, (7591), 455-459.
31 Carayon, P., Smith, M.J. and Haims, M.C. (1999). Work organization, job stress, and
work-related musculoskeletal disorders. Human Factors, 41, (4), 644-663.
32 Carpenter, J., Pocock, S. and Lamm, C.J. (2002). Coping with missing data in clinical
trials: a model-based approach applied to asthma trials. Statistics in Medicine, 21, (8),
1043-1066.
33 Cassidy, J.D., Cote, P., Carroll, L.J. and Kristman, V. (2005). Incidence and course of
low back pain episodes in the general population. Spine, 30, (24), 2817-2823.
34 Checkoway, H., Pearce, N. and Kriebel, D. (2004). Research Methods in Occupational
Epidemiology. (Oxford: Oxford University Press), 2nd Edition.
35 Chen, W.Q., Yu, I.T.S. and Wong, T.W. (2005). Impact of occupational stress and other
psychosocial factors on musculoskeletal pain among Chinese offshore oil installation
workers. Occupational and Environmental Medicine, 62, (4), 251-256.
36 Choi, B.K.L., Verbeek, J.H., Jiang, Y. and Tang, J.L. (2007). Exercises for prevention of
recurrences of low-back pain. Cochrane Database of Systematic Reviews: Protocols
37 Clarke, J., van Tulder, M., Blomberg, S., de Vet, H., van der Heijden, G. and Bronfort, G.
(2006). Traction for low back pain with or without sciatica: an updated systematic review
within the framework of the Cochrane collaboration. Spine, 31, (14), 1591-1599.
38 Cole, D.C. and Rivilis, I. (2004). Individual factors and musculoskeletal disorders: a
framework for their consideration. Journal of Electromyography and Kinesiology, 14,
(1), 121-127.
39 Collins, L.M., Schafer, J.L. and Kam, C.M. (2001). A comparison of inclusive and
restrictive strategies in modern missing data procedures. Psychological Methods, 6, (4),
330-351.
40 Cook, T.D. and Campbell, D.T. (1979). Quasi-experimentation: Design and analysis
issues for field settings. (Chicago: Rand-McNally)
83 41 Council for International Organisations of Medical Sciences (1993). International
guidelines for ethical review of epidemiological studies. (Geneva: CIOMS)
42 Council for International Organisations of Medical Sciences (2002). International ethical
guidelines for biomedical research involving human subjects. (Geneva: CIOMS)
43 Council of Europe (2005). Additional Protocol to the Convention on Human Rights and
Biomedicine concerning Biomedical Research. (Strasbourg: Council of Europe)
44 Davis, P.M., Badii, M. and Yassi, A. (2004). Preventing Disability From Occupational
Musculoskeletal Injuries in an Urban, Acute and Tertiary Care Hospital: Results from a
Prevention and Early Active Return-to-Work Safely Program. Journal of Occupational
and Environmental Medicine, 46, (12), 1253-1262.
45 Dempsey, P.G. (2002). Usability of the revised NIOSH lifting equation. Ergonomics, 45,
(12), 817-828.
46 Dempsey, P.G., Sorock, G.S., Ayoub, M.M., Westfall, P.H., Maynard, W., Fathallah, F.
and O'Brien, N. (2002). Prospective investigation of the revised NIOSH lifting equation.
In: McCabe, P.T. (Ed.) Contemporary Ergonomics 2002 (London: Taylor & Francis), pp.
77-81.
47 Dempsey, P.G. (2007). Effectiveness of ergonomics interventions to prevent
musculoskeletal disorders: Beware of what you ask. International Journal of Industrial
Ergonomics, 37, (2), 169-173.
48 Denis, D., St-Vincent, M., Imbeau, D., Jette, C. and Nastasia, I. (2008). Intervention
practices in musculoskeletal disorder prevention: A critical literature review. Applied
Ergonomics, 39, (1), 1-14.
49 Donner, A. and Klar, N. (2004). Pitfalls of and controversies in cluster randomization
trials. American Journal of Public Health, 94, (3), 416-422.
50 Eisen, E.A. (1999). Methodology for analyzing episodic events. Scandinavian Journal of
Work, Environment and Health, 25, (4), 36-42.
51 Elbourne, D.R. and Campbell, M.K. (2001). Extending the CONSORT statement to
cluster randomized trials: for discussion. Statistics in Medicine, 20, (3), 489-496.
52 Elders, L.A., van der Beek, A.J. and Burdorf, A. (2000). Return to work after sickness
absence due to back disorders--a systematic review on intervention strategies.
International Archives of Occupational and Environmental Health, 73, (5), 339-348.
53 Engel, G.L. (1977). The need for a new medical model: a challenge for biomedicine.
Science, 196, (4286), 129-136.
54 Engstrom, T., Hanse, J.J. and Kadefors, R. (1999). Musculoskeletal symptoms due to
technical preconditions in long cycle time work in an automobile assembly plant: a study
of prevalence and relation to psychosocial factors and physical exposure. Applied
Ergonomics, 30, (5), 443-453.
55 Feuerstein, M. (1996). Workstyle: definition, empirical support, and implications for
prevention, evaluation, and rehabilitation of occupational upper-extremity disorders. In:
84 S.D. Moon and S.L. Sauter (Eds.), Beyond Biomechanics: Psychosocial Aspects of
Musculoskeletal Disorders in Office Work, (London: Taylor & Francis), pp. 177-206.
56 Franche, R.L., Baril, R., Shaw, W., Nicholas, M. and Loisel, P. (2005a). WorkplaceBased Return-to-Work Interventions: Optimizing the Role of Stakeholders in
Implementation and Research. Journal of Occupational Rehabilitation, 15, (4), 525-542.
57 Franche, R.L., Cullen, K., Clarke, J., Irvin, E., Sinclair, S. and Frank, J. (2005b).
Workplace-Based Return-to-Work Interventions: A Systematic Review of the
Quantitative Literature. Journal of Occupational Rehabilitation, 15, (4), 607-631.
58 Frank, J.W., Kerr, M.S., Brooker, A.S., DeMaio, S.E., Maetzel, A., Shannon, H.S.,
Sullivan, T.J., Norman, R.W. and Wells, R.P. (1996). Disability resulting from
occupational low back pain. Part 1: What do we know about primary prevention? A
review of the scientific evidence on prevention before disability begins. Spine, 21, (24),
2908-2917.
59 Friedman, L.M., Furberg, C.D. and DeMets, D.L. (1998). Fundamentals of Clinical
Trials. (New York: Springer-Verlag), 3rd Edition.
60 Furlan, A.D., Brosseau, L., Imamura, M. and Irvin, E. (2002). Massage for low-back
pain: a systematic review within the framework of the Cochrane Collaboration Back
Review Group. Spine, 27, (17), 1896-1910.
61 Furlan, A.D., van Tulder, M., Cherkin, D., Tsukayama, H., Lao, L., Koes, B. and Berman,
B. (2005). Acupuncture and Dry-Needling for Low Back Pain: An Updated Systematic
Review Within the Framework of the Cochrane Collaboration. Spine, 30, (8), 944-963.
62 Gardner, L.I., Landsittel, D.P., Nelson, N.A. and Pan, C.S. (2000). Misclassification of
physical work exposures as a design issue for musculoskeletal intervention studies.
Scandinavian Journal of Work, Environment and Health, 26, (5), 406-413.
63 Gatchel, R.J. (2004). Musculoskeletal disorders: primary and secondary interventions.
Journal of Electromyography and Kinesiology, 14, (1), 161-170.
64 Genaidy, A.M., Lemasters, G.K., Lockey, J., Succop, P., Deddens, J., Sobeih, T. and
Dunning, K. (2007). An epidemiological appraisal instrument - a tool for evaluation of
epidemiological studies. Ergonomics, 50, (6), 920-960.
65 Gerr, F., Marcus, M., Monteilh, C., Hannan, L., Ortiz, D. and Kleinbaum, D. (2005). A
randomised controlled trial of postural interventions for prevention of musculoskeletal
symptoms among computer users. Occupational and Environmental Medicine, 62, (7),
478-487.
66 Goldenhar, L.M. and Schulte, P.A. (1996). Methodological issues for intervention
research in occupational health and safety. American Journal of Industrial Medicine, 29,
(4), 289-294.
67 Gordis, L. (2000). Epidemiology. (Philadelphia: W.B. Saunders Company), 2nd
Edition.
68 Greenland, S., Pearl, J. and Robins, J.M. (1999). Causal diagrams for epidemiologic
research. Epidemiology, 10, (1), 37-48.
85 69 Griffiths, A. (1999). Organizational interventions: facing the limits of the natural science
paradigm. Scandinavian Journal of Work, Environment and Health, 25, (6), 589-596.
70 Gyi, D.E. and Porter, J.M. (1998). Musculoskeletal problems and driving in police
officers. Occupational Medicine, 48, (3), 153-160.
71 Hagberg, M., Silverstein, B., Wells, R., Smith, M.J., Hendrick, H.W., Carayon, P. and
Perusse, M. Kuorinka, I. and Forcier, L. (Ed.) (1995). Work Related Musculoskeletal
Disorders (WMSDs): A Reference Book For Prevention. (London: Taylor and Francis)
72 Hagen, K.B., Hilde, G., Jamtvedt, G. and Winnem, M.F. (2000). The Cochrane review of
bed rest for acute low back pain and sciatica. Spine, 25, (22), 2932-2939.
73 Hagen, K.B., Hilde, G., Jamtvedt, G. and Winnem, M.F. (2002). The Cochrane review of
advice to stay active as a single treatment for low back pain and sciatica. Spine, 27, (16),
1736-1741.
74 Hagen, K.B., Jamtvedt, G., Hilde, G. and Winnem, M.F. (2005). The Updated Cochrane
Review of Bed Rest for Low Back Pain and Sciatica. Spine, 30, (5), 542-546.
75 Hakkanen, M., Viikari-Juntura, E. and Martikainen, R. (2001a). Incidence of
musculoskeletal disorders among newly employed manufacturing workers. Scandinavian
Journal of Work, Environment and Health, 27, (6), 381-387.
76 Hakkanen, M., Viikari-Juntura, E. and Martikainen, R. (2001b). Job experience, work
load, and risk of musculoskeletal disorders. Occupational and Environmental Medicine,
58, (2), 129-135.
77 Harkness, E.F., Macfarlane, G.J., Nahit, E.S., Silman, A.J. and McBeth, J. (2003a).
Mechanical and psychosocial factors predict new onset shoulder pain: a prospective
cohort study of newly employed workers. Occupational and Environmental Medicine,
60, (11), 850-857.
78 Harkness, E.F., Macfarlane, G.J., Nahit, E.S., Silman, A.J. and McBeth, J. (2003b). Risk
factors for new-onset low back pain amongst cohorts of newly employed workers.
Rheumatology (Oxford), 42, (8), 959-968.
79 Hartvigsen, J., Lings, S., Leboeuf-Yde, C. and Bakketeig, L. (2004). Psychosocial factors
at work in relation to low back pain and consequences of low back pain; a systematic,
critical review of prospective cohort studies. Occupational and Environmental Medicine,
61, (1), 2e
80 Hayes, R.J. and Bennett, S. (1999). Simple sample size calculation for clusterrandomized trials. International Journal of Epidemiology, 28, (2), 319-326.
81 Herbert, R.D. and Bo, K. (2005). Analysis of quality of interventions in systematic
reviews. British Medical Journal, 331, (7515), 507-509.
82 Hernan, M.A., Hernandez-Diaz, S. and Robins, J.M. (2004). A structural approach to
selection bias. Epidemiology, 15, (5), 615-625.
83 Hernberg, S. (1992). Introduction to Occupational Epidemiology. (Chelsea, Michigan:
Lewis Publishers Inc.)
86 84 Hestbaek, L., Iachine, I.A., Leboeuf-Yde, C., Kyvik, K.O. and Manniche, C. (2004).
Heredity of low back pain in a young population: a classical twin study. Twin Research,
7, (1), 16-26.
85 Heuer, H., Klimmer, F., Kylian, H., Seeber, A., Schmidt, K.-H., Hoffmann, G. and
Luttke-Nymphius, M. (1996). Musculoskeletal problems in bricklayers as a function of
length of employment: the role of secondary selection by low-back pain. Work and
Stress, 10, (4), 322-335.
86 Heymans, M.W., van Tulder, M.W., Esmail, R., Bombardier, C. and Koes, B.W. (2005).
Back Schools for Nonspecific Low Back Pain: A Systematic Review Within the
Framework of the Cochrane Collaboration Back Review Group. Spine, 30, (19), 21532163.
87 Hignett, S. (2003). Intervention strategies to reduce musculoskeletal injuries associated
with handling patients: a systematic review. Occupational and Environmental Medicine,
60, (9), E6
88 Hill, A.B. (1963). Medical ethics and controlled trials. British Medical Journal, 1,
(5337), 1043-1049.
89 Hintze, J. (19-4-2005). PASS 2005. (Kaysville, Utah: NCSS and PASS. Number
Cruncher Statistical Systems.)
90 Hollis, S. and Campbell, F. (1999). What is meant by intention to treat analysis? Survey
of published randomised controlled trials. British Medical Journal, 319, (7211), 670-674.
91 Hooftman, W.E., van Poppel, M.N., van der Beek, A.J., Bongers, P.M. and van Mechelen,
W. (2004). Gender differences in the relations between work-related physical and
psychosocial risk factors and musculoskeletal complaints. Scandinavian Journal of Work,
Environment and Health, 30, (4), 261-278.
92 Hopkins, W.G. (2000). A new view of statistics.
(http://www.sportsci.org/resource/stats/: Internet Society for Sport Science)
93 IJzelenberg, H., Meerding, W.J. and Burdorf, A. (2007). Effectiveness of a back pain
prevention program: a cluster randomized controlled trial in an occupational setting.
Spine, 32, (7), 711-719.
94 Ioannidis, J.P. (2005). Why most published research findings are false. PLoS Medicine,
2, (8), e124
95 Jellema, P., van Tulder, M.W., van Poppel, M.N., Nachemson, A.L. and Bouter, L.M.
(2001). Lumbar supports for prevention and treatment of low back pain: a systematic
review within the framework of the Cochrane Back Review Group. Spine, 26, (4), 377386.
96 Jones, G.T., Harkness, E.F., Nahit, E.S., McBeth, J., Silman, A.J. and Macfarlane, G.J.
(2006). Predicting the onset of knee pain: results from a 2-year prospective study of new
workers. Annals of the Rheumatic Diseases, 66, (3), 400-406.
97 Karasek, R. and Theorell, T. (1990). Healthy work: stress, productivity, and the
reconstruction of working life. (New York: Basic Books)
87 98 Karjalainen, K., Malmivaara, A., van Tulder, M., Roine, R., Jauhiainen, M., Hurri, H. and
Koes, B. (2001a). Multidisciplinary Biopsychosocial Rehabilitation for Neck and
Shoulder Pain Among Working Age Adults: A Systematic Review Within the Framework
of the Cochrane Collaboration Back Review Group. Spine, 26, (2), 174-181.
99 Karjalainen, K., Malmivaara, A., van Tulder, M., Roine, R., Jauhiainen, M., Hurri, H. and
Koes, B. (2001b). Multidisciplinary Biopsychosocial Rehabilitation for Subacute Low
Back Pain in Working-Age Adults: A Systematic Review Within the Framework of the
Cochrane Collaboration Back Review Group. Spine, 26, (3), 262-269.
100 Karsh, B.T.(2006). Theories of work-related musculoskeletal disorders: Implications for
ergonomic interventions. Theoretical Issues in Ergonomics Science, 7, (1), 71-88.
101 Karsh, B.T., Moro, F.B.P. and Smith, M.J. (2001). The efficacy of workplace ergonomic
interventions to control musculoskeletal disorders: a critical analysis of the peer-reviewed
literature. Theoretical Issues in Ergonomics Science, 2, (1), 23-96.
102 Kerry, S.M. and Bland, J.M. (1998). Sample size in cluster randomisation. British
Medical Journal, 316, (7130), 549
103 Kirkwood, B.R. and Sterne, J.A.C. (2006). Essential Medical Statistics. (Oxford:
Blackwell), 2nd Edition.
104 Korthals-de Bos, I., Gerritsen, A.A., van Tulder, M.W., Rutten-Van Molken, M.P., Ader,
H.J., de Vet, H.C. and Bouter, L.M. (2006). Surgery is more cost-effective than splinting
for carpal tunnel syndrome in the Netherlands: results of an economic evaluation
alongside a randomized controlled trial. BMC Musculoskeletal Disorders, 7, 86
105 Korthals-de Bos, I., van Tulder, M., van Dieten, H. and Bouter, L. (2004). Economic
evaluations and randomized trials in spinal disorders: principles and methods. Spine, 29,
(4), 442-448.
106 Kraus, J.F., Gardner, L., Collins, J., Sorock, G. and Volinn, E. (1997). Design factors in
epidemiologic cohort studies of work-related low back injury or pain. American Journal
of Industrial Medicine, 32, (2), 153-163.
107 Kristensen, T.S. (2005). Intervention studies in occupational epidemiology.
Occupational and Environmental Medicine, 62, (3), 205-210.
108 Kumar, S. (2001). Theories of musculoskeletal injury causation. Ergonomics, 44, (1),
17-47.
109 Lagerstrom, M., Hansson, T. and Hagberg, M. (1998). Work-related low-back problems
in nursing. Scandinavian Journal of Work, Environment and Health, 24, (6), 449-464.
110 Lahiri, S., Markkanen, P. and Levenstein, C. (2005). The cost effectiveness of
occupational health interventions: Preventing occupational back pain. American Journal
of Industrial Medicine, 48, (6), 515-529.
111 Laing, A.C., Frazer, M.B., Cole, D.C., Kerr, M.S., Wells, R.P. and Norman, R.W. (2005).
Study of the effectiveness of a participatory ergonomics intervention in reducing worker
pain severity through physical exposure pathways. Ergonomics, 48, (2), 150-170.
88 112 Lavender, S.A., Lorenz, E.P. and Andersson, G.B. (2007). Can a new behaviorally
oriented training process to improve lifting technique prevent occupationally related back
injuries due to lifting? Spine, 32, (4), 487-494.
113 Leboeuf-Yde, C. (2004). Back pain--individual and genetic factors. Journal of
Electromyography and Kinesiology, 14, (1), 129-133.
114 Lipsey, M.W.(1996). Key issues in intervention research: A program evaluation
perspective. American Journal of Industrial Medicine, 29, (4), 298-302.
115 Loisel, P. (2005). Intervention for return to work - what is really effective? Scandinavian
Journal of Work, Environment and Health, 31, (4), 245-247.
116 Loisel, P., Buchbinder, R., Hazard, R., Keller, R., Scheel, I., van Tulder, M. and Webster,
B. (2005). Prevention of work disability due to musculoskeletal disorders: the challenge
of implementing evidence. Journal of Occupational Rehabilitation, 15, (4), 507-524.
117 Macfarlane, G.J., Thomas, E., Papageorgiou, A.C., Croft, P.R., Jayson, M.I. and Silman,
A.J. (1997). Employment and physical work activities as predictors of future low back
pain. Spine, 22, (10), 1143-1149.
118 Magnusson, M.L., Pope, M.H., Wilder, D.G. and Areskoug, B. (1996). Are Occupational
Drivers at an Increased Risk for Developing Musculoskeletal Disorders? Spine, 21, (6),
710-717.
119 Mahmud, M.A., Webster, B.S., Courtney, T.K., Matz, S., Tacci, J.A. and Christiani, D.C.
(2000). Clinical management and the duration of disability for work-related low back
pain. Journal of Occupational and Environmental Medicine, 42, (12), 1178-1187.
120 Mancini, G., Baldasseroni, A., Laffi, G., Curti, S., Mattioli, S. and Violante, F.S. (2005).
Prevention of work related eye injuries: long term assessment of the effectiveness of a
multicomponent intervention among metal workers. Occupational and Environmental
Medicine, 62, (12), 830-835.
121 Martimo, K., Verbeek, J., Karppinen, J., Furlan, A.D., Kuijer, P., Viikari-Juntura, E.,
Takala, E. and Jauhiainen, M. (2007). Manual material handling advice and assistive
devices for preventing and treating back pain in workers. Cochrane database of
systematic reviews (Online)
122 Martocchio, J.J., Harrison, D.A. and Berkson, H. (2000). Connections between lower
back pain, interventions, and absence from work: a time-based meta-analysis. Personnel
Psychology, 53, 595-624.
123 Mayer, T., Gatchel, R.J. and Evans, T. (2001). Effect of age on outcomes of tertiary
rehabilitation for chronic disabling spinal disorders. Spine, 26, (12), 1378-1384.
124 McCluskey, S., Burton, A.K. and Main, C.J. (2006). The implementation of occupational
health guidelines principles for reducing sickness absence due to musculoskeletal
disorders. Occupational Medicine, 56, (4), 237-242.
125 McGill, S. (2002). Low Back Disorders. Evidence-Based Prevention and
Rehabilitation. (Champaign, IL: Human Kinetics)
89 126 McGill, S.M. (2004). Linking latest knowledge of injury mechanisms and spine function
to the prevention of low back disorders. Journal of Electromyography and Kinesiology,
14, (1), 43-47.
127 Medical Research Council . (1998). MRC guidelines for good clinical practice in
clinical trials. (London: Medical Research Council)
128 Medical Research Council . (2000). A Framework For Development And Evaluation Of
RCTs For Complex Interventions To Improve Health. (London: Medical Research
Council)
129 Medical Research Council (2002). Cluster randomised trials: methodological and ethical
considerations. (London: Medical Research Council)
130 Melhorn, J.M., Wilkinson, L., Gardner, P., Horst, W.D. and Silkey, B. (1999). An
outcomes study of an occupational medicine intervention program for the reduction of
musculoskeletal disorders and cumulative trauma disorders in the workplace. Journal of
Occupational and Environmental Medicine, 41, (10), 833-846.
131 Melhorn, J.M., Wilkinson, L. and Riggs, J.D. (2001). Management of musculoskeletal
pain in the workplace. Journal of Occupational and Environmental Medicine, 43, (2), 8393.
132 Moher, D., Schulz, K.F. and Altman, D. (2001). The CONSORT Statement: revised
recommendations for improving the quality of reports of parallel-group randomized trials.
Journal of the American Medical Association, 285, (15), 1987-1991.
133 Mullner, M., Matthews, H. and Altman, D.G. (2002). Reporting on statistical methods to
adjust for confounding: a cross-sectional survey. Annals of Internal Medicine, 136, (2),
122-126.
134 Nahit, E.S., Macfarlane, G.J., Pritchard, C.M., Cherry, N.M. and Silman, A.J. (2001a).
Short term influence of mechanical factors on regional musculoskeletal pain: a study of
new workers from 12 occupational groups. Occupational and Environmental Medicine,
58, (6), 374-381.
135 Nahit, E.S., Pritchard, C.M., Cherry, N.M., Silman, A.J. and Macfarlane, G.J. (2001b).
The influence of work related psychosocial factors and psychological distress on regional
musculoskeletal pain: a study of newly employed workers. Journal of Rheumatology, 28,
(6), 1378-1384.
136 National Research Council . (1999). Work-Related Musculoskeletal Disorders. Report,
Workshop Summary, and Workshop Papers. (Washington DC: National Academy Press)
137 National Research Council and Institute of Medicine . (2001). Musculoskeletal
Disorders and the Workplace: Low Back and Upper Extremities. (Washington, D.C.:
National Academy Press)
138 Niemisto, L., Kalso, E., Malmivaara, A., Seitsalo, S. and Hurri, H. (2003).
Radiofrequency denervation for neck and back pain: a systematic review within the
framework of the Cochrane Collaboration Back Review Group. Spine, 28, (16), 18771888.
90 139 Okunribido, O.O., Magnusson, M. and Pope, M. (2006). Delivery drivers and low-back
pain: A study of the exposures to posture demands, manual materials handling and wholebody vibration. International Journal of Industrial Ergonomics, 36, (3), 265-273.
140 Omar, R.Z., Ambler, G., Royston, P., Eliahoo, J. and Taylor, K.M. (2004). Cardiac
surgery risk modeling for mortality: a review of current practice and suggestions for
improvement. Annals of Thoracic Surgery, 77, (6), 2232-2237.
141 Ostelo, R.W., de Vet, H.C., Waddell, G., Kerckhoffs, M.R., Leffers, P. and van Tulder,
M. (2003). Rehabilitation following first-time lumbar disc surgery: a systematic review
within the framework of the Cochrane collaboration. Spine, 28, (3), 209-218.
142 Parenmark, G., Engvall, B. and Malmkvist, A.K. (1988). Ergonomic On-the-Job Training
of Assembly Workers. Arm-Neck- Shoulder Complaints Drastically Reduced amongst
Beginners. Applied Ergonomics, 19, (2), 143-146.
143 Park, R., Krebs, J. and Mirer, F. (1994). Mortality at an automotive stamping and
assembly complex. American Journal of Industrial Medicine, 26, (4), 449-463.
144 Park, R.M., Krebs, J.M. and Mirer, F.E. (1996). Occupational disease surveillance using
disability insurance at an automotive stamping and assembly complex. Journal of
Occupational and Environmental Medicine, 38, (11), 1111-1123.
145 Pearl, J. (2000a). Causality: Models, Reasoning, and Inference. (Cambridge:
Cambridge University Press)
146 Pearl, J. (2000b). Imperfect Experiments: Bounding Effects and Counterfactuals. In: J.
Pearl (Ed.), Causality: Models, Reasoning and Inference, (Cambridge: Cambridge
University Press), Chapter 8, pp. 259-282.
147 Peng, Y., Little, R.J. and Raghunathan, T.E. (2004). An extended general location model
for causal inferences from data subject to noncompliance and missing values. Biometrics,
60, (3), 598-607.
148 Pocock, S. (1983). Clinical Trials: A Practical Approach. (Chichester: John Wiley &
Sons Ltd)
149 Pope, M.H., Goh, K.L. and Magnusson, M.L. (2002). Spine ergonomics. Annual Review
of Biomedical Engineering, 4, 49-68.
150 Porter, J.M. and Gyi, D.E. (1995). Low Back Trouble and Driving. In: 2nd International
Scientific Conference on Prevention of Work-Related Musculoskeletal Disorders,
PREMUS 95, Montreal, Canada, September 24-28, 1995 (Montreal, Quebec, Canada:
Institut de Recherche en Sante et en Securite du Travail du Quebec), pp. 117-119.
151 Porter, J.M. and Gyi, D.E. (2002). The prevalence of musculoskeletal troubles among car
drivers. Occupational Medicine, 52, (1), 4-12.
152 Pransky, G., Gatchel, R., Linton, S.J. and Loisel, P. (2005). Improving return to work
research. Journal of Occupational Rehabilitation, 15, (4), 453-457.
91 153 Punnett, L. and Wegman, D.H. (2004). Work-related musculoskeletal disorders: the
epidemiologic evidence and the debate. Journal of Electromyography and Kinesiology,
14, (1), 13-23.
154 Rivilis, I., Cole, D.C., Frazer, M.B., Kerr, M.S., Wells, R.P. and Ibrahim, S. (2006).
Evaluation of a participatory ergonomic intervention aimed at improving musculoskeletal
health. American Journal of Industrial Medicine, 49, (10), 801-810.
155 Roberts, L., Little, P., Chapman, J., Cantrell, T., Pickering, R. and Langridge, J. (2002).
The back home trial: general practitioner-supported leaflets may change back pain
behavior. Spine, 27, (17), 1821-1828.
156 Robson, L.S., Shannon, H.S., Goldenhar, L.M. and Hale, A.R. (2001). Guide to the
Effectiveness of Strategies for Preventing Work Injuries; How to Show Whether a Safety
Intervention Really Works. (Cincinnati, OH: NIOSH), DHHS (NIOSH) Publication No.
2001-119, 139 pages.
157 Roland, M., Waddell, G., Moffett, J.K., Burton, A.K. and Main, C. (2002). The back
book: the best way to deal with back pain; get back active. (Norwich: TSO), 2 Edition.
158 Rothman, K.J. and Greenland, S. (1998). Modern Epidemiology. (Philadelphia, PA:
Lippincott Williams and Wilkins), 2nd Edition.
159 Sackett, D.L. (1979). Bias in analytic research. Journal of Chronic Diseases, 32, (1-2),
51-63.
160 Sauter, S.L. and Swanson, N.G. (1996). An ecological model of musculoskeletal
disorders in office work. In: S.D. Moon and S.L. Sauter (Eds.), Beyond Biomechanics:
Psychosocial Aspects of Musculoskeletal Disorders in Office Work, (London: Taylor &
Francis), pp. 3-21.
161 Schafer, J.L. and Graham, J.W. (2002). Missing data: our view of the state of the art.
Psychological Methods, 7, (2), 147-177.
162 Scheer, S.J., Watanabe, T.K. and Radack, K.L. (1997). Randomized Controlled Trials in
Industrial Low Back Pain. Part 3. Subacute / Chronic Pain Interventions. Archives of
Physical Medicine and Rehabilitation, 78, (4), 414-423.
163 Schlesselman, J.J. (1982). Case-Control Studies: Design, Conduct, and Analysis. (New
York: Oxford University Press)
164 Schonstein, E., Kenny, D.T., Keating, J. and Koes, B.W. (2003). Work conditioning,
work hardening and functional restoration for workers with back and neck pain.
Cochrane Database of Systematic Reviews: Protocols
165 Shannon, H.S., Robson, L.S. and Guastello, S.J. (1999). Methodological criteria for
evaluating occupational safety intervention research. Safety Science, 31, (2), 161-179.
166 Silverstein, B. and Clark, R. (2004). Interventions to reduce work-related
musculoskeletal disorders. Journal of Electromyography and Kinesiology, 14, (1), 135152.
92 167 Skov, T., Borg, V. and Orhede, E. (1996). Psychosocial and physical risk factors for
musculoskeletal disorders of the neck, shoulders, and lower back in salespeople.
Occupational and Environmental Medicine, 53, (5), 351-356.
168 Skov, T. and Kristensen, T.S. (1996). Etiologic and prevention effectiveness intervention
studies in occupational health. American Journal of Industrial Medicine, 29, (4), 378381.
169 Smith, M.J., Karsh, B.T. and Moro, F.B.P. (1999). A review of research on interventions
to control musculoskeletal disorders. (Washington DC: National Academy Press), WorkRelated Musculoskeletal Disorders: Report, Workshop Summary, and Workshop Papers,
229 pages.
170 Snook, S.H. (2004). Work-related low back pain: secondary intervention. Journal of
Electromyography and Kinesiology, 14, (1), 153-160.
171 Sobeih, T.M., Salem, O., Daraiseh, N., Genaidy, A. and Shell, R. (2006). Psychosocial
factors and musculoskeletal disorders in the construction industry: a systematic review.
Theoretical Issues in Ergonomics Science, 7, (3), 329-344.
172 Staal, J.B., Hlobil, H., van Tulder, M.W., Koke, A.J., Smid, T. and van Mechelen, W.
(2002). Return-to-work interventions for low back pain: a descriptive review of contents
and concepts of working mechanisms. Sports Medicine, 32, (4), 251-267.
173 Staal, J.B., Rainville, J., Fritz, J., van Mechelen, W. and Pransky, G. (2005). Physical
exercise interventions to improve disability and return to work in low back pain: current
insights and opportunities for improvement. Journal of Occupational Rehabilitation, 15,
(4), 491-505.
174 Stephens, P., Hickling, N., Gaskell, L., Burton, M. and Holland, D. (2004). Occupational
health and SMEs: Focused intervention strategies. Part 1 Overview Report. (Sudbury,
Suffolk: HSE Books), Research Report 257, 28 pages.
175 Straker, L., Burgess-Limerick, R., Pollock, C. and Egeskov, R. (2004). A randomized
and controlled trial of a participative ergonomics intervention to reduce injuries associated
with manual tasks: physical risk and legislative compliance. Ergonomics, 47, (2), 166168.
176 Sullivan, M.J., Feuerstein, M., Gatchel, R., Linton, S.J. and Pransky, G. (2005).
Integrating psychosocial and behavioral interventions to achieve optimal rehabilitation
outcomes. Journal of Occupational Rehabilitation, 15, (4), 475-489.
177 Swanson, N.G. and Sauter, S.L. (2006). A multivariate evaluation of an office ergonomic
intervention using longitudinal data. Theoretical Issues in Ergonomics Science, 7, (1), 317.
178 The General Assembly of the International Statistical Institute (1985). Declaration of
professional ethics. (Voorburg, The Netherlands: International Statistical Institute)
179 Theorell, T. (1996). Possible mechanisms behind the relationship between the demandcontrol-support model and disorders of the locomotor system. In: S.D. Moon and S.L.
Sauter (Eds.), Beyond Biomechanics: Psychosocial Aspects of Musculoskeletal Disorders
in Office Work, (London: Taylor & Francis), pp. 65-73.
93 180 Theorell, T. (2004). Keynote 5: Work-related musculoskeletal disorder related to stress how could they be prevented? Changing individual behaviour or developing
organisation? In: Premus 2004: Fifth International Scientific Conference on Prevention
of Work-Related Musculoskeletal Disorders (Zurich: Institute of Hygiene and Applied
Physiology, Swiss Federal Institute of Technology), Volume 1, pp. 45-46.
181 Thompson, A.R., Plewes, L.W. and Shaws, E.G. (1951). Peritendinitis crepitans and
simple tenosynovitis; a clinical study of 544 cases in industry. British Journal of
Industrial Medicine, 8, (3), 150-158.
182 Tierney, J.F. and Stewart, L.A. (2005). Investigating patient exclusion bias in metaanalysis. International Journal of Epidemiology, 34, (1), 79-87.
183 Treasure, T. and MacRae, K.D. (1998). Minimisation: the platinum standard for trials?
Randomisation doesn't guarantee similarity of groups; minimisation does. British
Medical Journal, 317, (7155), 362-363.
184 Troup, J.D.G. and Videman, T. (1989). Inactivity and the Aetiopathogenesis of
Musculoskeletal Disorders. Clinical Biomechanics, 4, (3), 173-178.
185 Tuncel, S., Iossifova, Y., Ravelo, E., Daraiseh, N. and Salem, S. (2006a). Effectiveness
of controlled workplace interventions in reducing lower back disorders. Theoretical
Issues in Ergonomics Science, 7, (3), 211-226.
186 Tuncel, S., Lotlikar, H., Salem, S. and Daraiseh, N. (2006b). Effectiveness of behaviour
based safety interventions to reduce accidents and injuries in workplaces: critical
appraisal and meta-analysis. Theoretical Issues in Ergonomics Science, 7, (3), 191-209.
187 Tveito, T.H., Hysing, M. and Eriksen, H.R. (2004). Low back pain interventions at the
workplace: a systematic literature review. Occupational Medicine, 54, (1), 3-13.
188 Ukoumunne, O.C., Gulliford, M.C., Chinn, S., Sterne, J.A. and Burney, P.G. (1999).
Methods for evaluating area-wide and organisation-based interventions in health and
health care: a systematic review. Health Technology Assessment, 3, (5), iii-92
189 van den Heuvel, S.G., Boshuizen, H.C., Hildebrandt, V.H., Blatter, B.M., Ariens, G.A.
and Bongers, P.M. (2005). Effect of sporting activity on absenteeism in a working
population. British Journal of Sports Medicine, 39, (3), e15
190 van der Beek, A.J., Oort-Marburger, D. and Frings-Dresen, M.H. (1994). The relations
between work demands and health complaints in lorry drivers. A model tested by means
of LISREL. International Archives of Occupational and Environmental Health, 66, (3),
179-184.
191 van Nieuwenhuyse, A., Somville, P.R., Crombez, G., Burdorf, A., Verbeke, G., Johannik,
K., Van den Bergh, O., Masschelein, R., Mairiaux, P., Moens, G.F. and The BelCoBack
Study Group (2006). The role of physical workload and pain related fear in the
development of low back pain in young workers: evidence from the BelCoBack Study;
results after one year of follow up. Occupational and Environmental Medicine, 63, (1),
45-52.
94 192 van Poppel, M.N.M., Hooftman, W.E. and Koes, B.W. (2004). An update of a systematic
review of controlled clinical trials on the primary prevention of back pain at the
workplace. Occupational Medicine, 54, (5), 345-352.
193 van Tulder, M., Furlan, A., Bombardier, C. and Bouter, L. (2003a). Updated method
guidelines for systematic reviews in the Cochrane Collaboration Back Review Group.
Spine, 28, (12), 1290-1299.
194 van Tulder, M., Malmivaara, A., Esmail, R. and Koes, B. (2000a). Exercise therapy for
low back pain: A systematic review within the framework of the Cochrane collaboration
back review group. Spine, 25, (21), 2784-2796.
195 van Tulder, M.W., Cherkin, D.C., Berman, B., Lao, L. and Koes, B.W. (1999). The
effectiveness of acupuncture in the management of acute and chronic low back pain. A
systematic review within the framework of the Cochrane Collaboration Back Review
Group. Spine, 24, (11), 1113-1123.
196 van Tulder, M.W., Scholten, R.J., Koes, B.W. and Deyo, R.A. (2000b). Nonsteroidal
anti-inflammatory drugs for low back pain: A systematic review within the framework of
the Cochrane Collaboration Back Review Group. Spine, 25, (19), 2501-2513.
197 van Tulder, M.W., Touray, T., Furlan, A.D., Solway, S. and Bouter, L.M. (2003b).
Muscle relaxants for nonspecific low back pain: a systematic review within the
framework of the Cochrane Collaboration. Spine, 28, (17), 1978-1992.
198 Village, J., Rempel, D. and Teschke, K. (2005). Musculoskeletal disorders of the upper
extremity associated with computer work: A systematic review. Occupational
Ergonomics, 5, (4), 205-218.
199 Volinn, E. (1999). Do workplace interventions prevent low-back disorders? If so, why? A
methodologic commentary. Ergonomics, 42, (1), 258-272.
200 Von Korff, M. and Saunders, K. (1996). The course of back pain in primary care. Spine,
21, (24), 2833-2837; discussion 2838-2839.
201 Waddell, G. (1998). The Back Pain Revolution. (Edinburgh: Churchill Livingstone)
202 Waddell, G. and Burton, A.K. (2001). Occupational health guidelines for the
management of low back pain at work: evidence review. Occupational Medicine, 51, (2),
124-135.
203 Waddell, G., O'Connor, M., Boorman, S. and Torsney, B. (2007). Working Backs
Scotland: a public and professional health education campaign for back pain. Spine, 32,
(19), 2139-2143.
204 Waersted, M. and Westgaard, R.H. (1991). Working Hours as a Risk Factor in the
Development of Musculoskeletal Complaints. Ergonomics, 34, (3), 265-276.
205 Wasiak, R., Pransky, G.S. and Webster, B.S. (2003). Methodological challenges in
studying recurrence of low back pain. Journal of Occupational Rehabilitation, 13, (1),
21-31.
95 206 Waters, T., Yeung, S., Genaidy, A., Callaghan, J., Barriera-Viruet, H., Abdallah, S. and
Kumar, S. (2006). Cumulative spinal loading exposure methods for manual material
handling tasks. Part 2: methodological issues and applicability for use in epidemiological
studies. Theoretical Issues in Ergonomics Science, 7, (2), 131-148.
207 Waters, T.R. (2004b). National efforts to identify research issues related to prevention of
work-related musculoskeletal disorders. Journal of Electromyography and Kinesiology,
14, (1), 7-12.
208 Waters, T.R. (2004a). National efforts to identify research issues related to prevention of
work-related musculoskeletal disorders. Journal of Electromyography and Kinesiology,
14, (1), 7-12.
209 Westgaard, R.H. and Aaras, A. (1984). Postural muscle strain as a causal factor in the
development of musculo-skeletal illnesses. Applied Ergonomics, 15, (3), 162-174.
210 Whysall, Z., Haslam, C. and Haslam, R. (2006). Implementing health and safety
interventions in the workplace: An exploratory study. International Journal of Industrial
Ergonomics, 36, (9), 809-818.
211 Woodward, M. (1999). Epidemiology: study design and data analysis. (Florida:
Chapman & Hall/CRC)
212 World Medical Association (2004). World Medical Association Declaration of Helsinki:
Ethical Principles for Medical Research Involving Human Subjects, 7th Edition
213 Zelen, M. (1974). The randomization and stratification of patients to clinical trials.
Journal of Chronic Diseases, 27, 365-375.
214 Zelen, M. (1979). A new design for randomized clinical trials. The New England Journal
of Medicine, 300, (22), 1242-1245.
215 Zwerling, C., Daltroy, L.H., Fine, L.J., Johnston, J.J., Melius, J. and Silverstein, B.A.
(1997). Design and conduct of occupational injury intervention studies: a review of
evaluation strategies. American Journal of Industrial Medicine, 32, (2), 164-179.
Published by the Health and Safety Executive
05/08
Health and Safety
Executive
Feasibility of carrying out an ergonomics
intervention study to prevent the incidence
of musculoskeletal disorders
This work examines the feasibility of assessing the
effectiveness of workplace ergonomic interventions to prevent
the onset of musculoskeletal disorders (MSDs). It reviews
existing models of causation of MSDs and the scientific
literature on interventions to prevent MSDs. It describes
relevant epidemiological methods and research protocols.
Many previous studies of the risk factors for MSDs have
not been able to assess causation and the need remains
for intervention studies of high methodological quality to
do this. A longitudinal Cluster Randomised Trial is the most
appropriate study design for assessing MSD causation in an
occupational setting. Measurement of injury rates generally
requires very large samples and/or long follow-up times to
provide adequate statistical power. It is likely that the study
would need to be carried out across multiple employers.
Because of the scale of the MSD problem, it is recommended
that HSE consider funding or part-funding a study designed to
test the effectiveness of workplace ergonomics interventions
to prevent the onset of episodes of musculoskeletal disorders.
Consideration should be given to making the study a multicentre, possibly international, collaborative study. Such a study
would be high risk due to the scale and duration needed and
the practical and organisational difficulties involved.
This report and the work it describes were funded by
the Health and Safety Executive (HSE). Its contents, including
any opinions and/or conclusions expressed, are those of the
authors alone and do not necessarily reflect HSE policy.
RR626
www.hse.gov.uk
Fly UP