Вы находитесь на странице: 1из 65

Health and Safety

Executive

Validation of the HSE Manual handling


Assessment Charts as predictors of
work-related low back pain
Prepared by the Health and Safety Laboratory
for the Health and Safety Executive 2014

RR1026
Research Report
Health and Safety
Executive

Validation of the HSE Manual handling


Assessment Charts as predictors of
work-related low back pain
Andrew DJ Pinder PhD and Gillian A Frost MRes
Health and Safety Laboratory
Harpur Hill
Buxton
Derbyshire SK17 9JN

The aim of this research was to ascertain whether HSE’s ‘Manual handling Assessment Charts’ (MAC tool) could
be used to predict workers losing time from work due to low back pain (LBP). Results from the study suggest
that as the ‘Hand distance from the lower back’ increased, the risk of lost time due to LBP increased. For each
10 cm increase, the rate of lost time increased by approximately 20%. No evidence of relationships between
other risk factors in the MAC and lost time was found. There was no evidence that the rate of lost time due to
LBP increased with either increasing total MAC lifting score or total MAC carrying score.

Due to imprecision in the model estimates (wide confidence intervals), the lack of statistically significant results,
and the limitations of the data, it was decide that it would not be appropriate to alter the scoring system
currently used in the MAC based on these data. Duty holders should be confident in carrying on using the MAC
tool as the risk factors for LBP included were identified as important by earlier studies.

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 2014

First published 2014

You may reuse this information (not including logos) free


of charge in any format or medium, under the terms of the
Open Government Licence. To view the licence visit
www.nationalarchives.gov.uk/doc/open-government-licence/,
write to the Information Policy Team, The National Archives, Kew,
London TW9 4DU, or email psi@nationalarchives.gsi.gov.uk.

Some images and illustrations may not be owned by the


Crown so cannot be reproduced without permission of the
copyright owner. Enquiries should be sent to
copyright@hse.gsi.gov.uk.

ii
KEY MESSAGES

• We tested the ability of the risk factors in the MAC (the Health & Safety Executive’s
‘Manual handling Assessment Charts’) to predict workers losing time from work due to low
back pain (LBP). We found that as the ‘Hand distance from the lower back’ increased, the
risk of lost time due to LBP increased. For each 10 cm increase, the rate of lost time due to
LBP increased by approximately 20%. We did not find evidence of relationships between
the other risk factors in the MAC and lost time due to LBP.
• We looked to see if the boundaries for the individual MAC risk factors should be adjusted.
We did not find evidence that required us to change the existing boundaries.
• We looked at how well the total MAC scores identified high-risk jobs. We did not find
evidence that the rate of lost time due to LBP increased with either increasing MAC lifting
score or MAC carrying score.
• We found it difficult to carry out our analysis because the MAC was designed for assessing
individual tasks, not jobs made up of many tasks and because it uses broad risk categories,
not accurate measurements. This means that our results are not very precise. A larger study
might find relationships that we were not able to find.
• The findings of this project may have implications for the use of the MAC by HSE and local
authority inspectors of health and safety, particularly in relation to guidance on enforcement
action.
• Duty holders should be confident in carrying on using the MAC tool as the risk factors for
LBP included were identified as important by earlier studies.

iii
EXECUTIVE SUMMARY

BACKGROUND

The Health & Safety Executive (HSE) and the Health & Safety Laboratory (HSL) developed the
Manual handling Assessment Charts (‘MAC tool’) for use by inspectors of health and safety
carrying out visits to workplaces, to help them identify high-risk workplace manual handling
activities. It was subsequently made available to the public.

The MAC tool can be used to assess the risks posed by lifting, carrying and team manual
handling activities. It was designed to help the user understand, interpret and categorise the
level of risk of the various known risk factors associated with manual handling activities. It
uses a traffic light colour coding score system to highlight high-risk manual handling tasks.
Numerical weightings have been assigned to the different colour codes and the total score can
be used to rank tasks by severity. All these features are designed to help the user prioritise
interventions to reduce the exposure to risk factors.

HSL carried out an HSE-funded project to collect epidemiological data to test if the 1991
NIOSH Lifting Equation could be used to predict if workers carrying out manual handling
would take time off work due to low back pain (LBP). HSL also began work on developing the
MAC tool after the NIOSH equation project started but before HSL collected any data. HSL
therefore made sure that the data collected could also be used to validate the MAC tool. HSL
recruited 515 subjects and followed them for 18 months to record incident episodes of loss of
time (absence or restricted duties) due to LBP (‘lost time’).

OBJECTIVES

The purpose of this study was to use the prospective data set collected by HSL to validate the
1991 NIOSH Lifting Equation to validate the MAC tool. The specific objectives were:
1. To test the ability of individual risk factors to predict increased risk of lost time
(absence/restricted duties) due to LBP or increased risk of reports of LBP, with or without
lost time.
2. To test for linear trends in parametric factors that predict increased risk of LBP.
3. To test the thresholds between colour bands for individual MAC risk factors.
4. To test the ability of the scoring system of the MAC to identify high-risk jobs by examining
the weights given to individual risk factors.

METHODS

Poisson regression was used to analyse the time to event data by calculating rate ratios for the
different MAC colour bands for each MAC variable. Rate ratios were adjusted for covariates
measured at baseline, either that were of a priori interest or that were shown to be predictors of
lost time.

MAIN FINDINGS
• Data were analysed from 486 of the 515 people recruited. The crude incidence rate of lost
time due to LBP was 29.3 [95% confidence interval (CI) 22.1–38.8] cases per 1,000 person-
days in the study.
• Most study participants (80%) were male, and had not experienced LBP in the 12 months
before the start of the study. The mean age of participants at baseline was 38.9 (SD 10.4)
iv
years, and the median duration of employment was 4.8 (range 0.04–39.1) years. Of the 126
unique jobs held by the 486 study participants, 54 (43%) involved carrying operations, and
24 (19%) involved team-handling operations. Only 80 people in the study (16%) undertook
team-handling operations, and they experienced only eight episodes of lost time due to LBP.
Therefore, the MAC variables for team handling were not investigated.

OBJECTIVE 1:
• There was evidence that ‘Hand distance from the lower back’ was associated with lost time
due to LBP; those coded as Red had more than twice the rate of lost time compared to those
coded as Amber (adjusted RR 2.02, 95% CI 1.05–3.89). There was no evidence that any of
the other risk factors were associated with lost time due to LBP.

OBJECTIVE 2:
• There was a statistically significant increasing linear trend going from Green to Amber to
Red categories for ‘Hand distance from the lower back’. The rate of lost time due to LBP
increased by 20% for each 10 cm increase in hand distance (RR 1.21, 95% CI 1.05–1.40).
The evidence did not support a non-linear relationship between hand distance and lost time
due to LBP. There was no evidence of a linear or non-linear relationship between ‘Carry
distance’ (using the underlying values rather than the colour bands) and lost time due to
LBP.

OBJECTIVE 3:
• Various alternative categorisations were investigated for ‘Load weight/frequency’,
‘Maximum individual load weight’, ‘Maximum effort’, ‘Weighted mean load weight’, ‘Hand
distance from the lower back, ‘Vertical lift region’ and ‘Carry distance’, but no evidence was
found to support any of the alternatives proposed over the current categorisations.

OBJECTIVE 4:
• There was no evidence that the rate of lost time due to LBP increased with either increasing
MAC lifting score or MAC carrying score. Due to imprecision in the model estimates (wide
confidence intervals), the lack of statistically significant results, and the limitations of the
data, it was decided that it would not be appropriate to alter the scoring system currently
used in the MAC based on these data.

v
CONTENTS PAGE

1. INTRODUCTION ..................................................................... 1
1.1 Context of the research 1
1.2 Earlier HSE research 5
1.3 Other relevant research 6
1.4 Research questions 6

2. METHODOLOGY .................................................................... 7
2.1 The design of the study 7
2.2 The sources of data and details on the sample and the response rate 8
2.3 Conversion of recorded data to MAC colour codes 10
2.4 Alternate boundaries for MAC variables 13
2.5 Composite weight/frequency indices 15
2.6 Job-level MAC variables 15
2.7 Analysis techniques 16

3. RESULTS.............................................................................. 20
3.1 Descriptive statistics 20
3.2 Tests of personal variables as risk factors of lost time due to LBP 21
3.3 Statistical analysis – lifting operations 22
3.4 Statistical analysis – carrying operations 25
3.5 Statistical analysis – team-handling operations 27

4. DISCUSSION / CONCLUSIONS........................................... 28
4.1 Predictive ability of individual risk factors 28
4.2 Testing parametric factors for linear trends 28
4.3 Testing colour band thresholds 28
4.4 Testing the scoring system 29
4.5 Limitations of the methodology 29
4.6 Potential interaction with a history of recent LBP 30
4.7 Future use of the MAC tool 31

5. ANNEX: TABLES OF DETAILED RESULTS....................... 32

6. REFERENCES ...................................................................... 52

vii
1. INTRODUCTION

1.1 CONTEXT OF THE RESEARCH

Many methods exist for assessing risk factors for low back pain (LBP) arising from manual
handling operations. After consideration of their usefulness during workplace inspections, HSE
and HSL(1-3) developed the Manual handling Assessment Charts (MAC tool) (Figure 1) for use
by Health and Safety Inspectors carrying out visits to workplaces, to help identify high-risk
workplace manual handling activities. It was subsequently made available to the public and has
proved very popular with those responsible for managing risks from manual handling operations
in the workplace (‘duty holders’). Features of the MAC are shown in Figure 2 to Figure 6.

Figure 1 The front cover of the MAC

1
Figure 2 The lifting flowchart of the MAC

There have been few attempts to validate the various methods of assessing manual handling
operations as predictors of harm to workers. HSL carried out an HSE-funded project to collect
epidemiological data to validate the 1991 National Institute for Occupational Safety and Health
(NIOSH) Lifting Equation(4) as a predictor of workers taking time off work due to LBP caused
by manual handling(5-7). That project was designed to replicate a project already under way in
the USA carried out by the Liberty Mutual Research Institute for Safety(8-16). HSL had begun
work on the NIOSH lifting equation project, but had not started data collection when they also
began work on developing the MAC tool. The opportunity was therefore taken to ensure that
the data collected to validate the NIOSH lifting equation could also be used to validate the
MAC tool.

2
Figure 3 The first page of the MAC assessment guide for lifting operations

Figure 4 The second page of the MAC assessment guide for lifting operations

3
The MAC tool(17) can be used to assess the risks posed by lifting, carrying and team manual
handling activities. It was designed to help the user understand, interpret and categorise the
level of risk of the various known risk factors associated with manual handling activities. It
uses a traffic light colour coding score system to highlight high-risk manual handling tasks.

The different manual handling factors for each of the three types of manual handling operations
that can be assessed with the MAC are presented as separate flow charts. Each flow chart leads
the user through each factor of the manual handling operation, giving guidance on evaluating
and grading the degree of risk. Each operation is supported by an assessment guide (for
example, Figure 4) that gives more information on scoring each factor of the flow chart.
Separate graphs (such as Figure 5) allow the user to consider the combined effects of load and
frequency, both in lifting operations and in carrying operations. Numerical weightings have
been assigned to the different colour codes and the total score can be used to rank tasks by
severity (see Figure 6 for the score sheet). All these features are designed to help the user
prioritise interventions to reduce the exposure to risk factors.

Figure 5 The MAC load weight/frequency graph for lifting operations

4
Figure 6 The MAC score sheet

1.2 EARLIER HSE RESEARCH

The Liberty Mutual project(16) had intended to recruit 2000 subjects and follow them
prospectively for 18 months. Problems with finding sufficient subjects in jobs that met their
inclusion criteria resulted in them only recruiting 453 subjects. They experienced a high drop
out rate and recorded a low number of incident cases. They were therefore unable to test the
predictive ability of the NIOSH equation.

HSL sought to recruit 1000 subjects(18) but also had problems finding subjects in suitable jobs so
the final cohort size was only 515. The annual drop out rate was 19.5%. Because of the time-
consuming nature(19) of reducing task data to the form needed for analysis, the initial analysis(5, 6)
was restricted to task data from a sub-group of 346 subjects. This showed that neither the
maximum NIOSH Single Task Lifting Index (STLI) nor the Composite Lifting Index (CLI)
predicted either loss of time (absence or restricted duties) from work due to LBP (‘lost time’), or
reports of LBP with or without lost time(5, 6). It did find a statistically significant increase in the
risk of lost time as the maximum horizontal hand distance increased (P=0.01). In addition, there
was a statistically significant increase in risk with increasing vertical offset of the hands from
750 mm, but this ceased to be statistically significant after adjustment for weight, age, gender
and LBP in the previous 12 months. The focus of the 1991 NIOSH equation is on the
Recommended Weight Limit (RWL) and the ratio of the actual load handled to it (the Lifting
Index). Values of individual risk factors are not assessed separately. However, the focus of the
MAC is on individual risk factors, with the scoring system as a secondary consideration. It was
therefore important to obtain the maximum statistical power by using task data from all the
subjects in the study; this required additional work to prepare the data from the remaining 169
subjects for analysis. This also provided opportunity to re-analyse the NIOSH data using the
complete data set(7).
5
1.3 OTHER RELEVANT RESEARCH

A recent prospective study has demonstrated relationships between the peak STLI and incident
cases of low back/gluteal pain lasting at least 24 hours in a sub-group of 258 subjects without a
recent history of LBP at baseline(20). They reported, after adjustment for age, BMI and gender, a
Hazard Ratio (HR) of 1.25 (95% CI 1.043–1.502, P=0.016). For the CLI they reported an HR
of 1.13 that approached significance at the 95% level (95% CI 0.996–1.274, P=0.058). They
also reported a significant trend, with the incidence of LBP increasing as the peak load moment
(maximum horizontal hand distance × load) increased (peak load moment≤575 kg-cm vs peak
load moment>1150 kg-cm, HR 2.77, 95% CI 1.49–5.15, P=0.001).

1.4 RESEARCH QUESTIONS

The purpose of this study was to make further use of the prospective data set collected by HSL
to validate the 1991 NIOSH Lifting Equation by seeking to validate the MAC tool. The specific
objectives were:
1. To test the ability of individual risk factors to predict increased risk of lost time
(absence/restricted duties) due to LBP or increased risk of reports of LBP, with or without
lost time.
2. To test for linear trends in parametric factors that predict increased risk of LBP.
3. To test the thresholds between colour bands for individual MAC risk factors.
4. To test the ability of the scoring system of the MAC to identify high-risk jobs by examining
the weights given to individual risk factors.

6
2. METHODOLOGY

2.1 THE DESIGN OF THE STUDY

This study was designed as further analysis of the data collected by HSL to evaluate the 1991
NIOSH lifting equation(6). That study protocol involved baseline measurements of the included
jobs and completion of a baseline questionnaire by participants employed to perform those jobs.
Follow-up of participants occurred every three months for 18 months to record incident cases of
lost time due to LBP. The target of 1000 subjects was chosen(18) as a compromise between the
need to keep the size of the study within manageable bounds and available resources, and the
need to provide meaningful results. The NIOSH equation calculates the RWL by multiplying a
‘Load Constant’ by a series of multipliers with maximum values of 1.0(4). Therefore, the sample
size required to test the ability of each of the constituent multipliers to predict lost time depends
on the range of values the multiplier can have, with larger samples required to evaluate the
multipliers with smaller ranges. It was acknowledged(18) that while the target number would be
sufficient to evaluate the overall equation and might be sufficient to evaluate the more important
individual factors within the equations, it definitely would not be sufficient to evaluate the least
influential.

Pinder and Frost(6) used logistic regression, and Cox regression to test for relationships between
baseline variables and the incidence of lost time due to LBP. They also used Generalised
Estimating Equations to test for relationships between baseline variables and reports of LBP at
three-monthly follow-ups.

The first stage of the current study consisted of completing the coding of task data from the
remaining subjects and ensuring that it was in the correct format for analysis. The second stage
used Poisson regression to examine the ratios of rates of reporting LBP in different conditions.
Poisson regression was chosen in preference to Cox regression because it makes fewer
assumptions, is easier to use, and is easier to interpret.

2.1.1 Selection criteria

To be included in the original study, jobs had to involve manual handling as a regular daily
activity, with each worker performing at least 25 lifts/lowers per day. They had to be expected
to continue in their existing form for at least the 18 months of the follow-up period. Jobs that
required substantial vehicle driving, the handling of people or team handling by teams of more
than two people were excluded. Jobs involving either carrying or team handling in teams of two
people were included in the study.

Both men and women employed in the jobs that qualified for the study were asked to
participate. No age limits or health status restrictions were imposed except that women who
were pregnant or who had had a baby in the previous six months were excluded, since
pregnancy itself can cause LBP. Participants needed to be full-time employees, with at least
one week of experience in the job, be expecting to stay in the job for the following 18 months
and be willing to fill in the baseline questionnaire and a follow-up questionnaire every three
months for 18 months.

2.1.2 Ethical approval of the study protocol and informed consent

The original study was approved by the HSE Research Ethics Committee
(ETHCOM/REG/98/12) in May 2001. Survey Control approval for HSL as part of the UK
Civil Service to carry out a survey of industry was obtained from the HSE Survey Control
Liaison Officer in April 2001. Participating firms and individuals gave informed consent to
7
participation and to the use of photographs of individuals and processes. Because no new data
were being collected, no further ethical approval was needed for this study.

2.2 THE SOURCES OF DATA AND DETAILS ON THE SAMPLE AND THE
RESPONSE RATE

2.2.1 Subject recruitment and baseline questionnaire completion

Subjects read the participant information sheet and gave informed consent to participation and
separate consent to being videoed or photographed. It was emphasised that it was not necessary
to video all participants.

Each subject filled in a four-page questionnaire at baseline. This asked for basic personal data
(gender, date of birth, height, weight and handedness), job-related data (including hours worked,
and history of LBP in the previous 12 months), health data (participation in exercise and
smoking status), history of MSDs in the previous three months (a version of the Nordic
Musculoskeletal Questionnaire) and six psychosocial factors. Five psychosocial scales
(‘Influence on and control over work’, ‘Supervisor climate’, ‘Stimulus from the work itself’,
‘Relations with fellow workers’ and ‘Psychological work load’ were taken from the PAK
(Psykosocial Arbetsmiljökartläggning) questionnaire(21). A sixth scale, ‘Management
commitment to health and safety’, was added.

Subjects usually completed the baseline questionnaire in work time, usually in a group of up to
10 individuals supervised by the HSL researcher. This was often during a planned or natural
break, or at the start of the work shift. The questionnaire was marked as ‘Confidential’ and had
a heading or watermark stating that it should only be returned to HSL staff or HSL. Some
questionnaires were left with workers with return envelopes. Where possible these were
collected from the individuals concerned or returned to union or safety representatives or, as a
last resort, management. Questionnaires and return envelopes were occasionally left for
subjects to return by post. The questionnaires were not anonymous due to the need for repeated
contact with the subjects and the need to link responses on the follow-up questionnaires to the
baseline questionnaire and job measurements.

2.2.2 Baseline measurement of tasks involving manual handling

Where more than one subject was recruited in a job, task measurements were taken only from
individuals who were willing to be videoed. The NIOSH equation does not take account of
differences between individuals performing a task, but focuses solely on the parameters of the
task. This is also true of the MAC tool. Therefore, it was felt that it was justified to limit the
amount of data collected by assuming that intra-participant variation within a job was small
enough to be disregarded, so it was only necessary to gather task data from one representative
individual performing each manual handling operation. If a number of individuals rotated
round workstations, measurements were made on the individual available at a workstation.

Video was used to record each manual handling task carried out as part of the job. The duration
of videoing depended on the nature of the job, how many tasks made up the job, how much
variability there was within it, and the frequency of handling, with high frequency tasks being
videoed for shorter periods than lower frequency tasks. The aim was to gain a representative
sample of the task so that inter-cycle variations in frequency would be averaged out. Linear
dimensions required for the NIOSH analysis were measured with a metal tape measure. Angles
were either measured with a goniometer or estimated from the video. The quality of the hand/
handle coupling was determined from the video. Variables unique to the MAC were coded
directly from the video.

8
Data were recorded on-site in a variety of ways, including on a hand-held computer, on paper
and on video. Sketches of workplaces were made with measured dimensions. Measurements
and verbal statements of measurements were recorded on video for later extraction. Discussions
with managers and workers that occurred during filming of tasks were also recorded in this way.
Where possible, weight information was taken from company records. Otherwise, weights were
recorded either from direct weighing of the item on a set of calibrated electronic bathroom
scales or directly from markings on the object.

The videos were collected with hand-held cameras without additional lighting, at 25 frames per
second on either VHS-C or DV tapes. The tapes were then digitised to MPEG-1 format,
including sound. Coding of the task variables and transcription of comments were carried out
using The Observer software (Noldus Information Technology BV, The Netherlands) version
5.0 and then version XT. This allowed frame-by-frame control of the video to establish timings
of events and so to allow the frequency of lift to be calculated by averaging over multiple
cycles. Different manual handling operations were defined as Activities with the type ‘State
Event’. Modifiers were used to distinguish multiple tasks carried out as part of the same job.
Multiple individuals in a film segment were coded in separate data files. The type of manual
handling operation (such as lifting) was coded at its beginning and either the end of the
operation or the transition to another type of handling (such as carrying) was coded. This made
explicit the complex nature of most actual handling operations. The cycle time for a task was
defined as the duration between one event (such as the initiation of a lift) and the recurrence of
the same event in the next cycle of the task. It therefore included any time between the end of
the handling operation and the recurrence of the marker event in the next cycle.

2.2.3 Follow-up questionnaires

Every three months from the date of entry of a subject to the study, a covering letter and one
page questionnaire were posted to the last known contact address of that person with a Freepost
return envelope. Non-responders were reminded up to three times, starting two weeks after the
initial follow-up letters, and then at weekly intervals. Where possible, this was done by
telephone.

The first section of the follow-up questionnaire asked the individual to check their contact
details. The second asked if he or she was still working in the previously reported job and work
area and for the same company. The third section asked if any LBP had been experienced since
the date of the previous response and three ‘Yes’ options were offered: ‘Work not affected’;
‘Put on light duties/restricted hours’; ‘Taken time off work’. Start and end dates were requested
for the last two options. Dates were not requested for the ‘Work not affected’ option since
multiple episodes could have occurred in the three months, and the ability of subjects to recall
dates of pain episodes would be much less than their ability to recall losing time from work.
The fourth section asked if the subject had been injured at work in the same period and offered
the same response options, including asking for start and end dates for restricted duties or time
off. For the ‘Work not affected’ response, it also asked for the date of injury. Additional
questions asked for the type of injury and the body part injured.

2.2.4 Triangulation of follow-up data

To provide a secondary source of job change/absence data, each company was contacted to ask
about job changes, absences due to LBP and injuries at work that they had records for the
individuals participating in the study. This was done for two periods: the first nine months and
the final nine months of the follow-up. Not all companies responded to these requests. Where
there were discrepancies between these data and the data from individuals, judgements were
made as to which was more reliable, taking into account the tendency of company responses to

9
fail to answer some questions. This proved to be an invaluable means of identifying individuals
who had changed job but had failed to respond to follow-up questionnaires.

HSE databases of reports made to HSE under RIDDOR (1995) (Reporting of Injuries, Deaths
and Dangerous Occurrences Regulations, 1995) were searched by name and study dates for all
participants. The reports retrieved showed that some reports of LBP received from individuals
were due to other causes, such as vehicle impacts, not manual handling. It was noted that many
over three day absences notified by individuals were not found in the RIDDOR database.

2.3 CONVERSION OF RECORDED DATA TO MAC COLOUR CODES

The data collected(6) for the evaluation of the NIOSH equation(4) were recorded as actual
dimensions. It was therefore necessary to define boundaries for each variable to convert them
into the MAC colour bands and MAC scores. Following the NIOSH approach(4), the hand
position was defined as the point mid-way between the left and right hands, and the position of
the low back was approximated as being above the mid-point between the ankles.

2.3.1 Categorisation of task types

To be included in the study, jobs had to include manual handling and the attempt was made to
capture all tasks involving manual handling that were performed regularly as part of the job.
Jobs could also include carrying and/or team handling. In the MAC, a task is only classed as a
carrying operation if the carrying distance is at least two metres. In other words, a task that
involves a carry distance of less than 2 m is treated as only involving lifting. Therefore, any
task that had a carry distance of less than 2 m was classified as a lifting operation. The number
of workers performing a lifting operation was recorded and any two-person tasks were classified
as team handling operations.

2.3.2 Load weight/frequency for lifting operations

The boundaries in the ‘Load weight/frequency chart for lifting operations’ were defined (Table
1) by reference to maximum acceptable weights for lifting reported in Tables 2 and 3 of Snook
and Ciriello(22). These were data for maximum acceptable weights of lift in the floor to knuckle
region, with a compact load (width 340 mm) with handles being lifted over a 760 mm range
close to the body. An upper weight limit of 50 kg was imposed(2). Comparison of these values
with the printed graph in the MAC revealed discrepancies that appear to have crept in during the
production of the final version of the MAC. Almost all of these were only 1 kg; the largest was
3 kg (Table 1). For the analysis, the values in the printed graph were used.
Table 1 Actual (intended) values for the ‘Load weight/frequency graph for lifting
operations’
One lift every (lifts per hour)
Bound- Accept- Day (0) 30 min- 5 min- 2 min- 1 minute 14 sec- 9 sec- 5 sec-
ary able to utes (2) utes (20) utes (30) (60) onds onds onds
(250) (400) (720)
Green/ 50% of 23 kg 19 kg 17 kg 15 kg 14 kg 13 kg 12 kg 10 kg
Amber females (17 kg) (16 kg)
Amber/ 50% of 45 kg 39 kg 38 kg 34 kg 30 kg 22 kg 19 kg 17 kg
Red males (44 kg (38 kg) (37 kg) (16 kg)
Red/ 10% of 50 kg 50 kg 50 kg 50 kg 47 kg 32 kg 28 kg 24 kg
Purple males (44 kg) (33 kg) (29 kg)

10
A look up table was created to convert values of weight and frequency into the MAC colour
band. Linear interpolation was used to determine the boundary points between the points
defined in Table 1.

2.3.3 Load weight/frequency for carrying operations

The principles adopted for lifting operations were used to convert the measured load weights
and frequencies for carrying operations to colour bands using the boundaries in Table 2. The
boundaries had been defined by reference to data from Table 10 of Snook and Ciriello(22) for
maximum acceptable weights of carry over distances of 2.1 m at approximately elbow height
(111 cm above floor level for males and 105 cm above floor level for females). The boxes had
handles and were carried close to the body. Again, an upper weight limit of 50 kg was
imposed(2). Again, there are discrepancies between these values and the printed graph in the
MAC, but all were of 1 kg.
Table 2 Actual (intended) values for the ‘Load weight/frequency graph for carrying
operations’
One carry every (carries per hour)
Boundary Acceptable Day (0) 30 minutes 5 minutes 2 minutes 1 minute 12 seconds
to (2) (12) (30) (60) (300)
Green/ 50% of 25 kg 19 kg 19 kg 19 kg 19 kg 16 kg
Amber females (18 kg) (18 kg) (18 kg) (18 kg)
Amber/ 50% of 44 kg 38 kg 32 kg 30 kg 30 kg 25 kg
Red males (33 kg)
Red/ 10% of 50 kg 50 kg 49 kg 42 kg 42 kg 35 kg
Purple males (48 kg) (43 kg) (43 kg)

2.3.4 Hand distance from low back

The MAC provides photographs to help the user categorise this variable. The original subject
of the photographs replicated the postures, markers were placed on the floor to indicate the
positions of the hands and the mid-ankles and the distances measured (Table 3). Boundaries
between the categories were defined as rounded values of the mid-points between the measured
values (Table 4). These were used to convert the measured values of the horizontal distance of
the hands form the mid-ankles, H, from the NIOSH study into MAC colour bands. As H was
measured at both the origin and destination of the lift, the larger value was used to determine the
MAC colour band.
Table 3 Measured values of ‘Hand distance from the lower back’ for the postures in
the MAC assessment guide for lifting operations
Hand distance from the Arm/back posture Measured horizontal
lower back category distance from mid ankles
Close (Green) Upper arms vertical 372 mm
Moderate (Amber) Trunk bent forward 435 mm
Moderate (Amber) Upper arms angled 533 mm
Far (Red) Upper arms angled & trunk bent 576 mm
Table 4 Numerical boundaries for ‘Hand distance from low back’ colour bands
Hand distance from the Horizontal distance from mid
lower back boundary ankles
Green/Amber 400 mm
Amber/Red 550 mm

11
2.3.5 Vertical lift region

The assessment guide for the MAC defines the colour bands for the ‘Vertical lift region’ in
terms of floor level and anthropometric landmarks – knee height, elbow height, and head height.
The NIOSH equation defines 750 mm as the optimal lift height, linking it to the 50th percentile
male knuckle height. Therefore boundary values (Table 5) were taken from 50th percentile male
data in Table 10.2 of Pheasant and Haslegrave(23) rounded to the nearest 25 mm. Both origin
and destination heights had been recorded so both were converted to MAC colour bands and the
worst-case value used.
Table 5 Numerical boundaries for Vertical lift region
Vertical lift region Mid-hand position Height above
boundary floor level
Red/Amber Head height 1750 mm
Amber/Green Elbow height 1100 mm
Green/Amber Knee height 550 mm
Amber/Red Floor level 0 mm

2.3.6 Trunk twisting and sideways bending

The assessment guide for the MAC does not give precise boundaries to distinguish a neutral,
symmetrical, trunk posture from ones that are twisted and/or bent sideways. It was therefore
decided to adopt the figure of 20° (Table 6) that the QEC (Quick Exposure Check)(24) used for
the neutral/non-neutral boundaries for these aspects of trunk posture, following the precedents
set in the creation of RULA (Rapid Upper Limb Assessment)(25) and in the case-referent study
by Punnett et al.(26).
Table 6 Numerical boundaries for trunk twisting and sideways bending
Trunk twisting/sideways bending Angle of deviation from
boundaries symmetrical posture
No trunk twisting/trunk twisting ±20°
No sideways flexion/sideways flexion 20°

2.3.7 Other MAC variables

Table 7 lists the other MAC variables and how they were coded. It also indicates any
equivalents in the NIOSH equation.
Table 7 Coding of other MAC variables
MAC variable How coded NIOSH equation
equivalent
Postural constraints Coded directly from video None
Grip on the load Recoded directly from the Coupling
equivalent NIOSH category
Floor surface Coded directly from video None
Other environmental factors (extreme temperature, Coded directly from video None
strong air movements, extreme lighting)
Carry distance Coded directly from video None
Obstacles en route (carrying only) Coded directly from video None
Communication, coordination and control (team- Coded directly from video None
handling only)

12
2.4 ALTERNATE BOUNDARIES FOR MAC VARIABLES

In order to explore the effect of changing the boundaries between colour bands for MAC
variables, alternate boundaries were tested for the ‘Load weight/frequency graph for lifting
operations’ and for the ‘Vertical lift region’.

2.4.1 Load weight/frequency for lifting operation alternate boundaries

The approach adopted when testing alternate load weight/frequency boundaries was to reduce
the values used for the boundaries so that they represented weights acceptable to greater
proportions of the population. This increased the severity of the classification of load weight/
frequency combinations. The Purple/Red boundary was changed from acceptable to 10% of
males to acceptable to 25% of males. The Amber/Green boundary was changed from
acceptable to 50% of females to acceptable to 75% of females. Examination of the figures
given by Snook and Ciriello(22) showed that weights acceptable to 25th percentile females and to
75th percentile males were both intermediate between the alternative Purple/Red and
Amber/Green boundaries (Table 8). Therefore, two alternate charts were defined (Table 9),
with one using the 25th percentile female values for the Red/Amber boundaries and the other
using the 75th percentile male values. The charts are illustrated in Figure 7 and Figure 8.
Table 8 Alternate colour band boundaries for ‘Load weight/frequency for lifting
operations’
One lift every (lifts per hour)
Bound- Accept- Day (0) 30 min- 5 min- 2 min- 1 min- 14 sec- 9 sec- 5 sec-
ary able to utes (2) utes (12) utes (30) ute (60) onds onds onds
(250) (400) (720)
Green/ 75% of 19 kg 14 kg 13 kg 13 kg 12 kg 11 kg 10 kg 8 kg
Amber females
Amber/ 25% of 27 kg 20 kg 18 kg 18 kg 17 kg 15 kg 14 kg 12 kg
Red females
Amber/ 75% of 33 kg 28 kg 28 kg 25 kg 22 kg 17 kg 14 kg 12 kg
Red males
Red/ 25% of 50 kg 47 kg 47 kg 42 kg 37 kg 28 kg 24 kg 20 kg
Purple males
Table 9 Alternate boundaries for ‘Load weight/frequency graph for lifting operations’
Boundary Original Alternative 1 Alternative 2
Purple/Red Acceptable to 10% of Acceptable to 25% of Acceptable to 25% of
males males males
Red/Amber Acceptable to 50% of Acceptable to 25% of Acceptable to 75% of
males females males
Amber/Green Acceptable to 50% of Acceptable to 75% of Acceptable to 75% of
females females females

13
50 50

40 40
Weight of load (kg)

30 30

20 20

10 10

0 0
Day (0) 30 minutes (2) 5 minutes (12) 2 minutes (30) 1 minute (60) 14 s econds (250) 9 seconds (400) 5 seconds (720)

One lift eve ry


(lifts pe r hour )

R/P boundary: acceptable to 25% of m ale s A/R boundary: Acceptable to 25% of fe m ales G/A boundary: Acceptable to 75% of fe m ales

Figure 7 Load weight/frequency for lifting operations (with the Red/Amber boundary
set as acceptable to 25% of females - alternative 1)

50 50

40 40
Weight of load (kg)

30 30

20 20

10 10

0 0
Day (0) 30 minutes (2) 5 minutes (12) 2 minutes (30) 1 minute (60) 14 s econds (250) 9 seconds (400) 5 seconds (720)

One lift eve ry


(lifts pe r hour )

R/P boundary: acceptable to 25% of m ale s A/R boundary: acceptable to 75% of m ale s G/A boundary: Acceptable to 75% of fe m ales

Figure 8 Load weight/frequency for lifting operations (with the Red/Amber boundary
set as acceptable to 75% of males - alternative 2)

2.4.2 Vertical lift region alternate boundaries

The boundaries for the ‘Vertical lift region’ are defined in terms of landmarks on the body.
Alternate landmarks were defined by extending the range of the Green zone from between
elbow height and knee height to between shoulder height and mid-lower leg height. These
landmarks relate to the boundaries used in the risk filter in Appendix 1 of the HSE guidance on
the 1992 Manual Handling Operations Regulations(27). Again, the numerical values of the
landmarks were defined (Table 10) using values for 50th percentile males(23).
14
Table 10 Original and alternate boundaries for ‘Vertical lift region’
Boundary Original Height above floor Alternate Height above floor
landmarks landmarks
Red/Amber Head height 1750 mm Head height 1750 mm
Amber/Green Elbow height 1100 mm Shoulder height 1450 mm
Green/Amber Knee height 550 mm Mid lower leg 275 mm
height
Amber/Red Floor level 0.01mm Floor level 0.01 mm

2.5 COMPOSITE WEIGHT/FREQUENCY INDICES

Since most of the jobs in the study involved multiple tasks with different parameters, it was
necessary to find a method of scoring total daily exposure for the MAC load weight/frequency
graphs for jobs involving varying weights and frequencies. Therefore, three variables were
defined (Table 11) in order to capture both the worst-case and the average demands, and the
MAC colour band was obtained for each of these variables.
Table 11 Summary load weight/frequency variables for lifting operations
Maximum individual Maximum effort Weighted mean load
load weight weight
Abbreviation MILW MEFF WMLW
Definition The weight of the The weight handled in Average effect taking all
heaviest item lifted the task with the greatest tasks and frequencies
value of load × into account
frequency
Meaning The worst single load The task that moves The overall demands on
weight most rapidly the worker
Associated frequency The frequency at which The frequency at which Overall frequency of
this load is lifted this load is lifted handling
Multiple values NA Take the worst-case NA
colour band.

2.6 JOB-LEVEL MAC VARIABLES

The MAC was designed as a tool for assessing risk of manual handling operations at the level of
individual tasks. It therefore focuses attention on individual risk factors in order to facilitate
identification of possible changes to reduce exposure to those risk factors. It was not designed
to assess the risk to an individual from all the manual handling tasks carried out by that person.

It would not be possible to analyse the data by using parameters of individual tasks to predict
the incidence of LBP, because if an individual performs multiple tasks it would be impossible to
disentangle the effects of a specific task. Therefore, it was necessary to summarise the task-
level MAC variables to give measures of exposure at the job level. Following the general
approach of the MAC, that considers ‘worst-case scenarios’ when assessing individual risk
factors, the worst-case scenario taken over all tasks within a job was identified for each MAC
variable. Thus, if two tasks were coded Amber for ‘Postural constraints’ and one Red, then the
Red was taken as the job-level colour.

One limitation of using the worst-case scenario approach is that a job involving just one Red
task out of nine tasks is grouped with jobs where all tasks were coded Red. Therefore,
alternative job-level variables were generated for each MAC variable, using the proportion of
tasks within the job that were colour coded Red/Purple for each MAC variable, i.e. fell into the
most severe categories. (The Red and Purple categories were combined because only the load
weight/frequency factor makes use of the Purple category.) In this approach, a job with five out

15
of six tasks coded as Red is considered riskier than one that has just one of six tasks coded Red.
However, though this approach takes into account the number of tasks that make up the job, it
does not take into account the number of repetitions of each task, i.e. it is not a time-weighted
average.

2.7 ANALYSIS TECHNIQUES

2.7.1 Use of incidence data

The follow-up questionnaires allowed individuals to report suffering LBP that had not resulted
in time off work or time at work on restricted duties, as well as incident episodes that involved
losing time from normal duties at work. This meant that individuals could make multiple
reports of suffering LBP within the different three-month intervals between follow-ups. As
most of these reports were not related to lost time, specific dates were not reported. They
therefore constituted ‘repeated panel data’, which can be analysed using Generalised Estimating
Equations(28). An initial examination of this approach showed that it did not add information to
the results already obtained and so it was not pursued any further. Therefore, all the reported
results refer to the analysis of the incidence of lost time using Poisson regression.

2.7.2 Task level worst-case scenarios

In many tasks, one or both of the variables indicating hand height or horizontal reach varied
systematically, for example, when items were being added to a stack. In order to simplify the
analysis for the NIOSH equation in situations where this occurred, the varying values were
replaced by a ‘Single Equivalent Value’ (SEV) that had the same lifting index as the CLI for the
sequence of lifts. Because the MAC has no equivalent to the CLI, it was necessary to identify
the maximum contributing value of the hand height or hand distance from low back in order to
identify the true worst-case scenario.

2.7.3 Use of multiple charts

The selection criteria for inclusion of a job within the study demanded that all tasks assessed
involve manual lifting or lowering. Therefore, all tasks were assessed with the lifting
operations flowchart, but only some tasks also involved carrying operations, and even fewer
were team-handling operations (Table 12). Therefore, the job-level data required for the
carrying operation and team-handling operation flowcharts were based on just the carrying and
team-handling tasks within the job, and the sample sizes were consequently reduced. This
meant that the same job could have two different values for job-level MAC variables, depending
on the type of task being considered.

2.7.4 Outcome of interest

The follow-up questionnaire asked about experience of LBP in the previous three months.
Three options for reporting LBP were given: ‘Work not affected’, ‘Put on light duties/restricted
hours at work’ and ‘Taken time off work’. The questionnaire also asked for dates of episodes of
restricted duties or time off. Because of small numbers reporting restricted duties, and because
absences were often followed by periods back at work on restricted duties, the final two
categories were combined to create a category of ‘lost time’. This was used as the outcome of
interest. Because time to event data and durations were available for this category, it was
possible to carry out longitudinal analysis of the data.

16
2.7.5 Analysis using Poisson regression

Poisson regression was used to investigate the associations between variables in the MAC and
lost time. This technique was chosen since it models the rate of the outcome, taking into
account the length of follow-up for each person and therefore allows for unequal follow-up
times. This method also allows for multiple events per person, time ‘not at risk’, such as when
absent due to an incident episode, and changing exposure status (such as a change in job).

Figure 9 illustrates how incident episodes were dealt with in this study. After 140 days in the
study, a participant began an episode of lost time, which lasted for 40 days. By definition a
participant cannot begin an episode of lost time during an existing episode, so this individual
could not be ‘at risk’ for those 40 days. Once the episode of lost time finished, the participant
was again at risk of a further lost time episode until day 400 when they dropped out of the
study. The person-days at risk (360 days) were therefore fewer than the days in the study – the
duration of follow-up (400 days).

Absent due to LBP

t=0 t=140 t=180 t=400

At risk Not at risk At risk

Figure 9 Example set-up for Poisson regression analysis

In this example, there would be three records for the subject:


• Lost time after 140 days at risk;
• Not at risk for 40 days;
• No lost time after 220 days at risk.

The dependent variable was the variable for whether or not lost time had occurred during an ‘at
risk’ period, with person-days at risk entered as the offset variable for that period.

The output from Poisson regression is presented in terms of rate ratios (RRs). When the
variable of interest is categorical (for example, smoker or non-smoker), the RR compares the
incidence rate experienced by a comparison group (for example, smokers) to that experienced
by the reference group (for example, non-smokers). An RR greater than 1 indicates that the rate
in the comparison group is greater than that in the reference group; an RR less than 1 indicates
that the rate is less than that in the reference group. When investigating continuous variables,
the RR represents the change in rate for a unit increase in the continuous variable.

Robust variance estimation was used throughout to account for possible clustering due to
multiple events/records per participant. Separate analyses were performed for lifting operations,
carrying operations, and team-handling operations, with the same methods followed for each.
All analyses were performed in Stata SE version 11.2 for Windows 1.

1
StataCorp. Stata statistical software SE version: Release 11.2. College Station, TX: StataCorp LP; 2011.
17
Multi-level Poisson regression was used to test personal variables. To do this, job-level MAC
variables were entered into separate Poisson regression models. Rate ratios were calculated
both with and without adjustment for covariates. The personal factors adjusted for were:
• Age;
• History at baseline of recent (within the previous 12 months) LBP;
• Gender;
• Regular exercise;
• Supervisor climate;
• The number of tasks in a job.

The personal variables adjusted for were identified as independent predictors of lost work time
due to LBP through multi-variable Poisson regression. Two other factors that were adjusted for
were:
• The occurrence of carrying tasks within the job (except in the analysis of carrying
operations);
• The occurrence of team handling within the job (except in the analysis of team-handling
operations).

All MAC variables were entered as a series of indicator variables, the significance of which was
tested using a joint Wald test. Reference categories were chosen to be the lowest category. RRs
can be imprecise when the number of observed cases is small. Therefore, categories with fewer
than five observed cases were combined with the ‘worst’ adjacent category where possible.
Since Green is designed to reflect ‘no’ or ‘low’ risk, it was not appropriate to combine this with
any other category so Amber was often combined with Red. Where combination was not
possible, RRs and 95% confidence intervals were not reported. A score test was used to test for
trend across categories, which involved assigning a score to the categories of the MAC variable
(for example, Green=0, Amber=1, Red=2, etc) and entering this into the model as a continuous
variable. Even if categories were combined to avoid categories with few observed cases, the
score test for trend used the original categories. So, if the categories had been combined to give
Green, Amber, and Red/Purple for the analysis, the original scores (0=Green, 1=Amber, 2=Red,
and 3=Purple) would be used in the score test for trend. A continuous test for trend rather than
a score test for trend was performed where appropriate – that is, for the number of Reds/Purples,
the total score, and the proportion of Reds/Purples). This uses the continuous variable rather
than assigning scores to the categorised variable.

The values underlying the MAC colour codes for ‘Hand distance from the lower back’ for
lifting operations, and ‘Carry distance’ for carrying operations, were investigated for non-linear
relationships with the incidence of lost time due to LBP. The Akaike Information Criterion
(AIC) was used to compare linear, quadratic and cubic Poisson regression models. When
comparing models, the lower AIC value indicates the better fitting model, with a difference
greater than 2.0 indicating a marked preference for the model with the smaller AIC.

The alternative categorisations for various MAC variables were also investigated and compared
to the original categories using the AIC.

18
2.7.6 Identification of personal variables that were independent risk factors
of lost time due to LBP

Poisson regression was also used to identify individual-level variables that were independent
risk factors of lost time due to LBP. Personal variables, lifestyle factors and work
characteristics were entered into separate Poisson regression models one at a time. Both crude
(unadjusted) rate ratios and rate ratios adjusted for age at entry, gender and LBP experience
before the study as a priori risk factors. All variables were entered as a series of indicator
variables, the significance of which was tested using a joint Wald test. Reference categories
were chosen to be the lowest category. Where appropriate, variables were also entered as
continuous, rather than categorical, variables.

Only those variables found to be statistically significant (P<0.05) in the adjusted analyses were
considered for inclusion into the multi-variable Poisson regression model. The base multi-
variable model included age at entry, gender and LBP experience before the study as a priori
risk factors. Variables were then included, one at a time, using a forward selection procedure.
The final multi-variable model gave the independent risk factors to be adjusted for in the
analysis of the MAC.

19
3. RESULTS

3.1 DESCRIPTIVE STATISTICS

Altogether, 515 people completed the baseline questionnaire. No follow-up information was
obtained for 17 of these. A further seven people were excluded from analysis since their jobs
were too complex to analyse. Five others were excluded due to their jobs involving little or no
manual handling. The remaining 486 people (94.4% of the 515 recruited) were included in the
analysis. Table 12 shows the number of people included in the analysis of each of the three
flowcharts. Since all tasks required lifting/lowering, this analysis included all 486 people
(100%). Of these, 227 people (47%) took part in carrying tasks, and 80 people (16%)
performed team-handling tasks. Just 57 people (12%) had performed both carrying and team-
handling operations (not necessarily at the same time) during follow-up.
Table 12 Number of participants, follow-up time, and number of episodes of lost time
due to LBP
Total – all Performed Performed team-
performed lifting carrying handling
operations operations operations
People 486 227 80
Follow-up/observation time (days) 223,472
Episodes of lost time due to LBP 65 25 8
Person-days at risk 221,782 106,410 36,127
Crude incidence rate, per 100,000 person- 29.3 (22.1–38.8) 23.5 (15.4–35.8) 22.1 (9.8–50.1)
days (95% CI)
Note: ‘lost time’ refers to work lost due to being placed on restricted duties or absence from work.
250
200150
Frequency
100 50
0

0 200 400 600


Total follow-up/observation time (days)

Figure 10 Histogram of follow-up/observation time for all 486 participants


20
Total follow-up time was 223,472 days, with a median of 549 days (approximately 18 months)
and a range of 13 to 644 days. The distribution of duration of follow-up is provided in Figure
10, which shows that the rate at which individuals dropped out was relatively stable over the
duration of the study. Altogether, there were 65 episodes of lost work time due to LBP
experienced by 52 people during follow-up, with two individuals reporting three episodes each.
The crude incidence rate of lost work time due to LBP was 29.3 [95% confidence interval (CI)
22.1–38.8] cases per 1,000 person-days, and this was similar for those who had performed
carrying operations, and those who had performed team-handling operations (Table 12).

Table 19 and Table 20 in the Annex shows the characteristics of the study participants grouped
by the type of operation they performed. Most study participants (80%) were male, and had not
experienced LBP in the 12 months before the start of the study. The mean age of participants at
baseline was 38.9 (SD 10.4) years, and the median duration of employment was 4.8 (range
0.04–39.1) years. There was evidence that the subgroups that performed carrying and team
handling operations differed to the whole group of study participants. In particular, the
proportion of women was greater among those who had performed carrying operations at some
stage during follow-up compared to those who had not (26% versus 14%, P=0.002). Those who
performed carrying operations tended to be slightly older (P=0.026), shorter (P=0.036), have
been employed for a greater length of time (P<0.001), have a shorter journey to work (P=0.035),
and have lower influence and control over work (P=0.005). Also, a greater proportion of those
who performed carrying operations had experienced some form of musculoskeletal trouble in
the three months before the study, compared to those who had not performed carrying
operations (76% versus 68%, P=0.046). There were similar differences between those who had
undertaken team-handling operations and those who had not. The proportion of women was
greater among those who had performed team-handling operations compared to those who had
not (38% versus 16%, p<0.001). Those who took part in team-handling operations tended to
weigh slightly less (P=0.038), be slightly shorter (P=0.029), and have a smaller proportion of
smokers (29% versus 42%, P=0.031) than those who had not.

The jobs held by the participants are described in Table 21 in the Annex in terms of their MAC
colour coding. In total, there were 126 unique jobs held by the 486 study participants during the
study period. Of those jobs, 54 (43%) involved carrying operations, and 24 (19%) involved
team-handling operations. The number of tasks per job ranged from one to 67, with a median of
four. The distribution of jobs across MAC categories tended to be very unbalanced, with one
category holding the majority of jobs (over 60% in most cases). When considering lifting
operations, for load weight/frequency, postural constraints, floor surface and other
environmental factors, the worst-case scenario over all tasks was Green for the majority of jobs.
From a regulatory perspective, this is encouraging. However, the worst-case scenario was Red
for ‘Hand distance from the lower back’, and ‘Grip on the load’ for the majority of lifting jobs.
Similarly, the worst-case scenario for just the carrying tasks in a job was Red for ‘Grip on the
load’ in the majority of jobs. The worst-case scenario over just the team-handling tasks in a job
was Red for the majority of jobs for ‘Grip on the load’, and for ‘Trunk twisting/sideways
bending’.

3.2 TESTS OF PERSONAL VARIABLES AS RISK FACTORS OF LOST


TIME DUE TO LBP

Table 22 in the Annex shows the crude and adjusted RRs of lost time due to LBP for personal
variables, lifestyle factors and work characteristics. After adjustment for age, gender and
previous LBP experience, both regular exercise and supervisor climate were statistically
significantly associated with lost time due to LBP. Therefore, both were included in the multi-
variable Poisson regression model.

21
Table 13 shows the final multi-variable Poisson regression model, which included both regular
exercise and supervisor climate. All variables in the model would be adjusted for in the analysis
of the MAC.
Table 13 Multivariable Poisson regression model for the relationship between
personal variables and lost time due to LBP
Variable (Participants=465; jobs=126)
Cases Days at risk Rate Ratio (95% CI)/P-value
Sex
Male 57 173,763 1.00 Ref
Female 4 39,627 --- ---
LBP experienced during previous 12 months
No 18 119,921 1.00 Ref
Yes: work unaffected 17 66,195 1.57 (0.75–3.31)
Yes: lost work time 26 27,274 5.91 (3.16–11.06)***
Wald test P<0.001***
Age (years), continuous 61 216,390 1.00 (0.97–1.04)
Wald test P=0.861
Exercise regularly
No 33 87,510 1.00 Ref
Yes 28 125,880 0.49 (0.29–0.84)**
Wald test P=0.009**
Supervisor climate, continuous 61 216,390 0.94 (0.89–0.99)*
Wald test P=0.016*
Ref, reference category
*, statistically significant with P<0.05; **, statistically significant with P<0.01; ***, statistically significant with
P<0.001.

3.3 STATISTICAL ANALYSIS – LIFTING OPERATIONS

3.3.1 Overall results

Table 23 in the Annex shows the crude and adjusted RRs associated with the individual MAC
variables in the lifting operation flowchart. Those whose worst-case scenario was Red for
‘Hand distance from the lower back’ had twice the rate of lost time due to LBP than those with
Amber (adjusted RR 2.02, 95% CI 1.05–3.89), and the score test for trend suggested that the
rate increased linearly from Green to Amber to Red (P=0.017). There were no other statistically
significant results when considering the worst-case scenario variables.

Table 24 in the Annex shows the crude and adjusted RRs associated with the MAC variables
based on the proportion of tasks coded Red or Purple within the job. The ‘<1%’ category
corresponds to all tasks within a job coded as Green or Amber, and the ‘>99%’ category
corresponds to all tasks coded as Red. For variables that have just two categories (‘<1%’ and ‘1
%+’), this is equivalent to comparing the combined Green and Amber category to the combined
Red and Purple category in Table 23 in the Annex. The continuous test for trend was
statistically significant for ‘Hand distance from the lower back’, providing evidence that the rate
of lost time due to LBP increased with the proportion of tasks coded Red (P=0.033). This
suggested that even if it was not possible to remove all Red tasks from a job, removing some
could decrease the risk of lost time due to LBP. There were no other statistically significant
results observed.

3.3.2 Load weight/frequency variables

Load weight/frequency, maximum individual load weight, maximum effort and weighted mean
load weight were all coded using the chart in the MAC and with the alternative graphs for Load

22
weight/frequency for lifting operations (Figure 7 and Figure 8). Table 25 in the Annex
compares the results obtained from the three graphs. None of the categorisations resulted in any
of the variables being statistically significantly associated with lost time due to LBP (all Wald
tests, and score tests for trend were not statistically significant). In addition to this, none of the
alternative categorisations resulted in a better fitting model than the original categorisations
according to the AIC; for example, the difference between the lowest and highest AIC was less
than 2.0 for the load weight/frequency (486.28 versus 486.93). Therefore, there was no
evidence to support altering the chart currently used in the MAC to either of the alternatives.

3.3.3 Hand distance from the lower back

Table 14 shows the results of the investigation into whether the relationship between ‘Hand
distance from the lower back’ and lost time due to LBP was non-linear. The linear trend was
statistically significant, and for every 10 cm increase in ‘Hand distance from the lower back’,
the rate of lost time increased by around 20% (adjusted RR 1.21, 95% CI 1.05–1.40). Neither
the quadratic nor the cubic terms in the quadratic and cubic models were statistically significant.
The AIC indicated that the quadratic model did not improve the model fit (475.96 versus
477.61) and the cubic model made the model fit worse with a difference greater than 2.0
(475.96 versus 478.69). Since the linear model remained the model with the lowest AIC, there
was no evidence that the relationship between ‘Hand distance from the lower back’ and the
incidence of lost time was non-linear, and no obvious alternative categorisations were identified.
Table 14 Relationship between ‘Hand distance from the lower back’ and incidence of
lost time due to LBP, estimated using Poisson regression
Model Adjusted Rate (95% CI) AIC
Ratio a
Model 1 – Linear 475.96
Hand distance from the lower back (10 cm) 1.21 (1.05–1.40)*
Model 2 – Quadratic 477.61
Hand distance from the lower back (10 cm) 0.93 (0.34–2.51)
Hand distance from the lower back (10 cm) squared 1.00 (1.00–1.01)
Model 3 – Cubic 478.69
Hand distance from the lower back (10 cm) 7.12 (0.10–509.2)
Hand distance from the lower back (10 cm) squared 0.97 (0.92–1.03)
Hand distance from the lower back (10 cm) cubed 1.00 (1.00–1.00)
a: adjusted for age, previous LBP experience, gender, regular exercise, supervisor climate, number of
tasks in job, if the job involved carrying, and if the job involved team-handling;
Ref, reference category; *, statistically significant with P<0.05.

Figure 11 shows the distribution of the worst-case scenario values of ‘Hand distance from the
lower back’ over all 126 jobs. This shows the unequal distribution of jobs across categories of
the MAC variable, with most jobs having a worst-case scenario that was Red. An alternative
categorisation was therefore suggested, based on an even distribution of jobs across colour
bands so that a third were coded as Green, a third as Amber, and the final third as Red. The cut-
offs for the alternative categorisation were 50 cm and 75 cm, instead of the current 40 cm and
55 cm (Table 4). The two categorisations for ’Hand distance from the lower back’ are
compared in Table 15. The alternative categorisation resulted in a more balanced distribution of
cases and days at risk across groups as expected, and the score test for trend remained
statistically significant (P=0.032). However, the AIC did not favour one categorisation above
the other, since the difference between the two AICs was less than 2.0. Therefore, the data did
not support changing the current categories for ‘Hand distance from the lower back’ to the
alternative.

23
15 10
Frequency
5
0

20 40 60 80 100 120
Hand distance from the lower back (cm)

Jobs currently coded as:


Green Amber Red

Figure 11 Distribution of jobs for hand distance from the lower back, based on worst-
case scenario over all tasks within a job
Table 15 Comparison of two categorisations of ‘Hand distance from the lower back’
Hand distance from the Cases Days at risk Adjusted Rate (95% CI) AIC or P-
lower back (cm) Ratio a value
Current categorisation AIC=479.62
Green: <40 0 4,346 NA NA
Amber: 40- 12 57,704 1.00 Ref
Red: 55+ 49 149,473 2.02 (1.05–3.89)*
Wald test P=0.035*
Score test for trend P=0.017*
Categorisation based on distribution of jobs AIC=479.97
1st third: <50 12 52,283 1.00 Ref
2nd third: 50- 14 66,338 1.13 (0.53–2.45)
3rd third: 75+ 35 92,902 2.24 (1.05–4.77)*
Wald test P=0.053
Score test for trend P=0.032*
a, adjusted for age, previous LBP experience, gender, regular exercise, supervisor climate, number of
tasks in job, if the job involved carrying, and if the job involved team-handling using separate Poisson
regression models;
Ref, reference category; *, statistically significant with P<0.05.

3.3.4 Vertical lift region

The alternative categorisation for the ‘Vertical lift region’ was based on different body
landmarks. The results of comparing this and the current categorisation are shown in Table 16.
The alternative categorisation resulted in a more even distribution of cases and days at risk
across categories, with some jobs originally coded as Amber moving to the Green category.

24
However, the ‘Vertical lift region’ was not statistically significantly associated with incidence
of lost time due to LBP for either of the categorisations, and there was no evidence that using
the alternative categorisation improved the model fit (since the difference between the two AICs
was less than 2.0). Therefore, there was no evidence to support a change in categorisation for
‘Vertical lift region’.
Table 16 Comparison of two categorisations for ‘Vertical lift region’
Vertical lift region (cm) Cases Days at Adjusted (95% CI) AIC or P-
risk Rate Ratio a value
Current categorisation AIC=485.75
Green: Above knee and below elbow 0 556 NA NA
height (55–109.9)
Amber: Below knee or above elbow 48 177,836 1.00 Ref
height (0.001–54.9 or 110–174.9)
Red: Floor level or below or above 13 33,131 1.25 (0.33–4.79)
head height (<0.001 and/or 175+)
Wald test P=0.747
Score test for trend P=0.721
Alternative categorisation AIC=485.08
Green: Above mid lower leg and below 12 53,665 1.00 Ref
shoulder height (27.5–144.9 cm)
Amber: Below mid lower leg or above 36 124,727 1.36 (0.72–2.58)
shoulder height (0.001–27.4 or 145–174.9)
Red: Floor level or below or above 13 33,131 1.61 (0.38–6.78)
head height (<0.001 or 175+)
Wald test P=0.619
Score test for trend P=0.384
a, adjusted for age, previous LBP experience, gender, regular exercise, supervisor climate, number of
tasks in job, if the job involved carrying, and if the job involved team-handling using separate Poisson
regression models; Ref, reference category

3.4 STATISTICAL ANALYSIS – CARRYING OPERATIONS

Table 26 in the Annex shows the crude and adjusted RRs associated with the MAC variables as
they currently stand for the carrying operation MAC. The score test for trend was statistically
significant for ‘Hand distance from the lower back’ (P=0.031), suggesting that rate of lost time
due to LBP increased linearly from Green to Amber to Red. However, this was no longer
statistically significant when other risk factors were taken into account (P=0.123). There were
no other statistically significant results when considering the worst-case scenario MAC
variables for carrying operations. However, it is worth noting that only 227 people (57%)
undertook carrying tasks, and so the ability of the analysis to detect associations if they truly
existed would be reduced.

Table 27 in the Annex shows the crude and adjusted RRs associated with the MAC variables
based on the proportion of carrying tasks coded Red or Purple. Due to the reduced number of
participants included in this analysis, many variables did not have sufficient numbers of cases to
estimate RRs. There were no statistically significant results.

Table 17 shows the results of the investigation into whether the relationship between ‘Carry
distance’ and lost time due to LBP was non-linear. There was no evidence of a linear or non-
linear relationship with ‘Carry distance’, since no RRs were statistically significant. The
difference in AICs indicated that the model that included a quadratic term for ‘Carry distance’
fitted the data better than the model with just a linear term. However, the quadratic term was
not statistically significant, so there was insufficient evidence to conclude that the relationship is
non-linear.
25
Table 17 Relationship between ’Carry distance’ and incidence of lost time due to LBP,
estimated using Poisson regression
Model Adjusted Rate (95% CI) AIC
Ratio a
Model 1 – Linear 208.28
Carry distance (m) 0.99 (0.92–1.06)
Model 2 – Quadratic 205.64
Carry distance (m) 0.72 (0.47–1.09)
Carry distance (m) squared 1.01 (1.00–1.01)
Model 3 – Cubic 207.46
Carry distance (m) 0.61 (0.29–1.30)
Carry distance (m) squared 1.02 (0.97–1.08)
Carry distance (m) cubed 1.00 (1.00–1.00)
a, adjusted for age, previous LBP experience, gender, regular exercise, supervisor climate, number of
tasks in job, and if the job involved team-handling; Ref, reference category.
20
15
Frequency
10
5
0

0 10 20 30 40 50
Carry distance (m)

Jobs currently coded as:


Green Amber Red

Figure 12 Distribution of jobs for ‘Carry distance’, based on worst-case scenario over
all carrying tasks within a job

Figure 12 shows the distribution of values of the worst-case scenarios for ‘Carry distance’ over
all 54 jobs that involved carrying. This shows the highly skewed nature of the data, with some
carries involving extreme (greater than 20 m) distances. Two alternative categorisations for
‘Carry distance’ were suggested based on the distribution of jobs. The first was based on thirds
of the distribution, so that a third of jobs were coded as Green, a third as Amber, and the final
third as Red. Due to there being a few extreme distances, it was suggested that the Red category
might need to be divided into a Red and Purple category. The second alternative categorisation
was therefore based on thirds for the Green and Amber categories, but the Red and Purple
categories were based on sixths. This meant that a third of the jobs were coded Green, a third
Amber, a sixth Red, and the final sixth as Purple. The proposed alternative categorisations and
26
the original categorisation used in the MAC are compared in Table 18. The differences in AICs
favoured the original categorisation of ‘Carry distance’ over the categorisation that introduced a
fourth (Purple) category, but there was no evidence that the original categorisation fitted the
data substantially better than the categorisation based on tertiles. In addition, there was no
statistically significant association between ‘Carry distance’ and lost time due to LBP for any of
the categorisations used. Therefore, there was no evidence to support changing the current
categories used in the carrying operation flowchart for ‘Carry distance’ to either of the
alternatives.
Table 18 Comparison of alternative categorisations to current categorisation for ‘Carry
distance’, based on the distribution of jobs
Carry distance (m) Cases Days at Adjusted (95% CI) AIC or P-
risk Rate Ratio a value
Current categorisation AIC=206.99
Green: 2- 11 32,383 1.00 Ref
Amber: 4- 9 50,399 0.43 (0.13–1.49)
Red: 10+ 2 19,962 --- ---
Wald test P=0.335
Score test for trend P=0.181
Categorisation based on distribution of jobs AIC=208.28
1st tertile: 2- 9 29,875 1.00 Ref
2nd tertile: 3.25- 6 23,346 0.67 (0.16–2.72)
3rd tertile: 5+ 7 49,523 0.42 (0.09–2.03)
Wald test P=0.544
Score test for trend P=0.278
Categorisation based on distribution of jobs AIC=210.16
1st tertile: 2- 9 29,875 1.00 Ref
2nd tertile: 3.25- 6 23,346 0.67 (0.16–2.79)
5th sextile: 5- 5 29,561 0.46 (0.08–2.59)
6th sextile: 10+ 2 19,962 --- ---
Wald test P=0.725
Score test for trend P=0.273
a, adjusted for age, previous LBP experience, gender, regular exercise, supervisor climate, number of
tasks in job, and if the job involved team-handling using separate Poisson regression models;
Ref, reference category.

3.5 STATISTICAL ANALYSIS – TEAM-HANDLING OPERATIONS

Only 80 people in the study (16%) undertook team-handling operations, and they experienced
just eight episodes of lost time due to LBP. The ability to detect true associations between
MAC variables and lost time due to LBP based on these data would therefore be severely
restricted. Therefore, the MAC variables for team handling were not investigated. The cases
and days at risk are reported in Table 28 in the Annex.

27
4. DISCUSSION / CONCLUSIONS

4.1 PREDICTIVE ABILITY OF INDIVIDUAL RISK FACTORS

The first objective set out in the project proposal was:


• To test the ability of individual risk factors to predict increased risk of lost time
(absence/restricted duties) due to LBP or increased risk of reports of LBP, with or without
lost time.

4.1.1 Lifting operations

There was evidence that the ‘Hand distance from the lower back’ was associated with lost time
due to LBP; those coded as Red had over twice the rate of lost time compared to those coded as
Amber for ‘Hand distance from the lower back’ (adjusted RR 2.02, 95% CI 1.05–3.89).

4.1.2 Carrying operations

There was no evidence that any of the carrying risk factors were associated with lost time due to
LBP. However, fewer people performed carrying tasks, and so the ability of the analysis to
detect associations if they truly existed was less than in the analysis of jobs involving lifting.

4.1.3 Team-handling operations

Only 80 people (16%) undertook team-handling operations and they experienced only eight
episodes of lost time due to LBP. The ability to detect associations between MAC variables and
lost time due to LBP using these data, if they truly existed, was severely restricted, and so this
was not investigated further.

4.2 TESTING PARAMETRIC FACTORS FOR LINEAR TRENDS

The second objective was:


• To test for linear trends in parametric factors that predict increased risk of LBP.

4.2.1 Lifting operations

There was a statistically significant increasing linear trend going from Green to Amber to Red
categories for ‘Hand distance from the lower back’ (score test for trend P=0.017). Investigation
of the underlying values of the hand distance from the lower back (rather than the colour codes)
showed evidence that the rate of lost time due to LBP increased by, on average, 20% for each
10 cm increase in hand distance (RR 1.21, 95% CI 1.05–1.40). The evidence did not support a
non-linear relationship between hand distance and lost time due to LBP.

4.2.2 Carrying operations

There was no evidence of a linear or non-linear relationship between ‘Carry distance’ (using the
underlying values rather than the colour bands) and lost time due to LBP.

4.3 TESTING COLOUR BAND THRESHOLDS

The third objective was:


• To test the thresholds between the colour bands for individual MAC risk factors.

28
4.3.1 Lifting operations

Various alternative categorisations were investigated for ‘Load weight/frequency’, ’Maximum


individual load weight’, ’Maximum effort’, ’Weighted mean load weight’, ’Hand distance from
the lower back, and ‘Vertical lift region’, but no evidence was found to support any of the
alternatives proposed over the current categorisations.

4.3.2 Carrying operations

Alternative categorisations were investigated for ‘Carry distance’, but there was no evidence to
support the alternatives proposed over the current categorisation.

4.4 TESTING THE SCORING SYSTEM

The fourth objective was:


• To test the ability of the scoring system of the MAC to identify high-risk jobs by examining
the weights given to individual risk factors.

4.4.1 Lifting operations

There was no evidence that the rate of lost time due to LBP increased with increasing MAC
lifting score.

4.4.2 Carrying operations

There was no evidence that the MAC score for carrying operations was associated with the rate
of lost time due to LBP.

Due to imprecision in the model estimates (wide confidence intervals), the lack of statistically
significant results, and the limitations of the data discussed below, it was decided that it would
not be appropriate to alter the scoring system currently used in the MAC based on these data.

4.5 LIMITATIONS OF THE METHODOLOGY

4.5.1 Summarising task-level data at the job-level

The MAC tool was designed for use at the task-level, and so it was necessary to summarise the
task-level data to the job-level for the analysis. However, information is lost in doing this,
which could result in potential associations being weakened or attenuated.

4.5.2 Variability of jobs

The nature of most of the jobs included in the study was that they included multiple tasks and
often parameters such as the origin of, or destination for, a lift varied systematically within
otherwise identical tasks. Variability of tasks is seen as beneficial by ergonomists(29) since it
may have protective effects by reducing localised stresses on at-risk tissues in the body.
Dempsey(12) discussed the complexities that arise from task variability when collecting data for
epidemiological studies of manual handling and LBP.

4.5.3 Unequal distribution of jobs

There tended to be an extremely unequal distribution of jobs across categories of the MAC
when defining the job-level variables based on the worst-case scenario, such that some
categories did not have any jobs in them. This unequal distribution of jobs leads to an unequal

29
distribution of cases and days at risk, which can affect the ability of the statistical analysis to
detect associations if they truly exist (reduced statistical power).

4.5.4 Missing data

Missing values in the personal variables meant that 21 people were not included in the analysis
when adjustment for personal variables was made. ‘Multiple imputation’ is a technique that can
be used to fill in the missing values when values are missing completely at random (MCAR). If
this is not the case, the results from analyses that use imputed data can be subject to bias.
Because data were missing from variables such as body weight that are personally sensitive, it
was decided not to use multiple imputation in this instance.

4.6 POTENTIAL INTERACTION WITH A HISTORY OF RECENT LBP

As noted previously, the study by Boda et al.(20) reported a relationship between the NIOSH
peak STLI Index and the risk of low back / gluteal pain in a subgroup of workers free from LBP
for at least 90 days at baseline. By contrast, the original HSL analysis of the 1991 NIOSH
equation(5, 6) had found no such relationship in a cohort containing workers with and without a
history of recent LBP (defined, as in this study, as in the previous 12 months). These differing
findings suggested that the ability of the NIOSH equation to predict LBP may depend on the
recent history of LBP – that is, there could be an interaction present. Re-analysis(7) has shown
that there was a statistically significant interaction (p=0.030) between history of recent LBP and
the NIOSH STLI. The adjusted HR (HR=1.37 (95% CI 1.05-1.78)) for STLI among those
without recent LBP at baseline was statistically significant. For those without previous LBP it
was not (HR=0.88 (95% CI 0.64-1.20).

This raises complex issues as to why a tool such as the NIOSH equation should be able to
predict lost time due to LBP in individuals without recent LBP but not in those who have
suffered LBP in the previous 12 months. It also raises issues of whether this interaction also
applies to other tools such as the MAC and the implications of such interactions for the
usefulness of methods of assessing risk.

Key to these issues is the natural history of LBP, which often occurs for short periods with
spontaneous recovery. As far back as 1969, Rowe commented that LBP is “characteristically
intermittent, episodic and recurrent and can meaningfully be studied only as a continuum which
stretches through the active years of a man’s life”(30). Troup et al.(31) suggested that some backs
have a phase where they are liable to cause pain that can last a year or two and noted that return
to work “does not indicate relief of symptoms, let alone full recovery”. Ferguson et al.(32) found
that patients with long-term muscular LBP generated increase spine loading due to increased
levels of muscle co-activity, meaning they were at greater risk than symptom-free workers.
They suggested that a symptomatic worker with muscular LBP returning to work might need to
be on light duties for a minimum of three months. In a cohort of nursing students followed
throughout training(33), the recent and recurrent history of LBP was a more important risk factor
for new episodes of LBP than just a history of previous LBP.

Given the recurrent nature of LBP, it is necessary to exclude individuals with a history of LBP
from studies trying to elucidate risk factors for first onset of LBP(34). However, such selection
of subjects is more problematic in studies examining the ability of risk assessment methods to
predict LBP or lost time due to LBP since these are designed for use at the task level and do not
distinguish between differential risks to individuals with or without existing LBP. A study of a
sample of 32-33 year olds from the UK 1958 birth cohort(35) excluded 43% who had previously
suffered back pain and the authors raised the possibility that the risk factors they identified
(psychological distress 10 years previously, and smoking) do not apply to individuals with the
most “frail” backs.
30
Troup et al.(36) found that using a person’s history of LBP in combination with a battery of
anthropometric measures, measures of isometric and psychophysical strength and respiratory
function was effective in predicting individuals that remained free of LBP and individuals that
continued to have chronic pain. However, their discriminant analysis model tending to
underestimate occurrences of cases of LBP and was poor at predicting transition to other states.

It therefore seems that the structures of the back of an individual that has recently recovered
from an episode of LBP are at increased risk of generating pain than in individuals without a
history of problems. It can therefore be speculated that this increased risk correlates or overlaps
with the increased risk from manual handling risk factors and that individuals with histories of
recent LBP are as likely to suffer new episodes when not exposed to manual handling as when
exposed to it.

It is therefore possible that, like the NIOSH equation, the predictive ability of the MAC could
be different for individuals with no history of LBP and individuals with recent LBP. However,
the limitations of the data, particularly the unequal distribution of jobs across MAC colour
bands, meant that it was not possible in this study to investigate this issue in more detail.

4.7 FUTURE USE OF THE MAC TOOL

The limitations of the sample size, the complexity of real jobs that rarely consist of a single
repetitive task and the difficulties caused by the broad risk categories the MAC uses in
preference to accurate measurements means that the results reported here are not very precise.
It would therefore be wrong to conclude that the MAC is not a useful method of assessing risk
arising from manual handling operations.

The findings of this project may have implications for the use of the MAC by HSE and local
authority inspectors of health and safety, particularly in relation to guidance on enforcement
action.

Duty holders should be confident in carrying on using the MAC tool as the risk factors for LBP
included were identified as important by earlier studies.

31
5. ANNEX: TABLES OF DETAILED RESULTS

Table 19 Characteristics of the study participants at baseline (i)


All - performed Also performed carrying operations Also performed team-handling operations
lifting operations
No Yes P-value (No No Yes P-value (No
vs. Yes) vs. Yes)
Total 486 259 227 406 406 80 80
Sex, n (%) 0.002** g <0.001*** g
Male 391 (80%) 222 (86%) 169 (74%) 341 (84%) 50 (63%)
Female 95 (20%) 37 (14%) 58 (26%) 65 (16%) 30 (38%)
LBP experienced during 12 months before study, n (%) 0.451 g 0.152 g
No 279 (57%) 147 (67%) 132 (58%) 226 (56%) 53 (66%)
Yes: work unaffected 146 (30%) 83 (32%) 63 (28%) 129 (32%) 17 (21%)
Yes: lost work time 61 (13%) 29 (11%) 32 (14%) 51 (13%) 10 (13%)
Age (years), mean 38.9 (10.4) 37.9 (10.7) 40.0 (10.0) 0.026* h 38.9 (10.5) 39.0 (10.0) 0.925 h
(standard deviation) a
Weight (kg), mean (SD) b 80.1 (14.1) 80.5 (13.6) 79.6 (14.6) 0.510 h 80.7 (13.9) 77.1 (14.7) 0.038* h
Height (kg), mean (SD) c 1.74 (0.09) 1.75 (0.09) 1.73 (0.09) 0.036* h 1.75 (0.09) 1.72 (0.10) 0.029* h
BMI (kg/m2), mean (SD) d 26.3 (3.9) 26.3 (4.0) 26.4 (3.8) 0.816 h 26.4 (3.9) 25.9 (3.9) 0.303 h
Weekly working hours, 40.8 (6.0) 41.2 (7.0) 40.3 (4.5) 0.111 h 40.6 (6.1) 41.7 (5.3) 0.122 h
mean (SD)
Length of employment 4.8 (0.04–39.1) 4.5 (0.04–36.0) 5.9 (0.1–39.1) <0.001*** i 4.8 (0.04–36.4) 5.2 (0.2–39.1) 0.247 i
(years), median (range) e
Daily driving time (min- 25 (0–270) 30 (0–240) 20 (0–270) 0.035* i 30 (0–270) 20 (0–120) 0.308 i
utes), median (range)
Exercise regularly, n (%) 0.356 g 0.722 g
No 197 (41%) 100 (39%) 97 (43%) 166 (41%) 31 (39%)
Yes 289 (59%) 159 (61%) 130 (57%) 240 (59%) 49 (61%)
Current smoker, n (%) 0.153 g 0.031* g
No 294 (60%) 149 (58%) 145 (64%) 237 (58%) 57 (71%)
Yes 192 (40%) 110 (42%) 82 (36%) 169 (42%) 23 (29%)
*, statistically significant with P<0.05; **, statistically significant with P<0.01; ***, statistically significant with P<0.001.
a, one missing value, total n=485; b, 10 missing values, total n=476; c, 4 missing values, total n=482; d, 14 missing values, total n=472; e, 7 missing values, total n=479
g, Chi-squared test; h, T-test; i, Wilcoxon rank-sum test.
32
Table 20 Characteristics of the study participants at baseline (ii)
All- performed Also performed carrying operations Also performed team-handling operations
lifting operations
No Yes P-value (No No Yes P-value (No
vs. Yes) vs. Yes)
Any musculoskeletal trouble in the previous 3 months 0.046* b 0.098 b
No 139 (29%) 84 (32%) 55 (24%) 110 (27%) 29 (36%)
Yes 347 (71%) 175 (68%) 172 (76%) 296 (73%) 51 (64%)
Psychosocial variables, mean (SD) a
Influence on & control over -0.42 (4.84) 0.18 (4.80) -1.09 (4.81) 0.005** c -0.41 (4.85) -0.45 (4.83) 0.946 c
work
Supervisor climate 1.13 (5.02) 1.18 (5.34) 1.07 (4.65) 0.812 c 0.95 (5.13) 2.03 (4.32) 0.087 c

Stimulus from the work -0.02 (5.22) 0.09 (5.39) -0.13 (5.03) 0.655 c 0.01 (5.21) -0.17 (5.31) 0.781 c
itself
Relations with fellow 3.69 (4.27) 3.50 (4.51) 3.90 (3.99) 0.310 c 3.60 (4.31) 4.13 (4.31) 0.322 c
workers
Psychological work load 1.00 (4.77) 0.92 (4.99) 1.09 (4.53) 0.712 c 0.87 (4.84) 1.64 (4.37) 0.201 c

Management commitment 2.12 (5.03) 1.96 (5.21) 2.30 (4.83) 0.473 c 2.02 (5.00) 2.65 (5.21) 0.315 c
to health & safety
Note: psychosocial variables ranged from -10 to +10.
*, statistically significant with P<0.05; **, statistically significant with P<0.01; ***, statistically significant with P<0.001.
a, 20 missing values, total=466
b, Chi-squared test; c, T-test

33
Table 21 MAC variables for study jobs, based on the worst-case scenario over tasks within a job
MAC variable and Description Jobs involving lifting Jobs involving carrying Jobs involving team handling
colour band (worst-case scenario (worst-case scenario (worst-case scenario over
over all tasks) over carrying tasks only) team-handling tasks only)
Total jobs 126 (100%) 54 (43%) 24 (19%)
Total tasks per job, median (range)
Load weight/frequency
Green Read from chart 94 (75%)
Amber Read from chart 28 (22%)
Red Read from chart 3 (2%)
Purple Read from chart 1 (1%)
Hand distance from the lower back
Green CLOSE: upper arm vertical/trunk upright 5 (4%) 8 (15%) 6 (25%)
Amber MODERATE: upper arm angled or trunk bent forward 40 (32%) 25 (46%) 13 (54%)
Red FAR: upper arm angled and trunk bent forward 81 (64%) 21 (39%) 5 (21%)
Vertical lift region a
Green Above knee and/or below elbow height 1 (1%) 0 (0%)
Amber Below knee and/or above elbow height 113 (90%) 22 (92%)
Red Floor level or below and/or above head height 12 (10%) 2 (8%)
Trunk twisting/sideways bending
Green Little or no twisting or sideways bending 33 (26%) 7 (29%)
Amber Trunk twisting OR sideways bending 58 (46%) 5 (21%)
Red Trunk twisting AND sideways bending 35 (28%) 12 (50%)
Postural constraints
Green None 87 (69%) 37 (69%) 21 (88%)
Amber Restricted 37 (29%) 17 (31%) 3 (13%)
Red Severely restricted 2 (2%) 0 (0%) 0 (0%)
Grip on the load
Green Good 1 (1%) 2 (4%) 0 (0%)
Amber Reasonable 2 (2%) 2 (4%) 0 (0%)
Red Poor 123 (98%) 50 (93%) 24 (100%)
Floor surface
Green Dry and in good condition 90 (71%) 41 (76%) 19 (79%)
Amber Dry but in poor conditions or uneven 0 (0%) 0 (0%) 0 (0%)
Red Contaminated, wet, sloping or unstable 36 (29%) 13 (24%) 5 (21%)

34
MAC variable and Description Jobs involving lifting Jobs involving carrying Jobs involving team handling
colour band (worst-case scenario (worst-case scenario (worst-case scenario over
over all tasks) over carrying tasks only) team-handling tasks only)
Other environmental factors
Green No factors present 94 (75%) 45 (83%) 21 (88%)
Amber One factor present 31 (25%) 9 (17%) 3 (13%)
Red Two or more factors present 1 (1%) 0 (0%) 0 (0%)
Carry load weight/frequency
Green Read from chart 48 (89%)
Amber Read from chart 5 (5%)
Red Read from chart 0 (0%)
Purple Read from chart 1 (2%)
Asymmetrical trunk/load
Green Load symmetrical in front/two hands 13 (24%)
Amber Asymmetrical or offset load/hands 36 (67%)
Red One-handed to side or twisting/back bent 5 (9%)
Carry distance (m)
Green 2- 19 (35%)
Amber 4- 23 (43%)
Red 10 m+ 12 (22%)
Obstacles en route
Green No obstacles OR carry route is flat 34 (63%)
Amber Steep slope OR Trip hazards OR steps 19 (35%)
Red Ladders 1 (2%)
Team load weight
Green 2 person <35 kg; 3 person <40 kg 24 (100%)
Amber 2 person 35–50 kg; 3 person 40–75 kg; 4 person 40–100 kg 0 (0%)
Red 2 person 50–85 kg; 3 person 75–125 kg; 4 person 100–170 kg 0 (0%)
Purple 2 person >85 kg; 3 person 75–125 kg; 4 person 100–170 kg 0 (0%)
Communication, co-ordination and control
Green Good 18 (75%)
Amber Reasonable 3 (13%)
Red Poor 5 (13%)
Data are numbers with percentages in parentheses, unless otherwise specified.
a, 21 missing values for lifting jobs, total n=105, 1 missing value for team-handling jobs, total n=23.

35
Table 22 Crude and adjusted RRs of lost time due to LBP for personal variables, estimated using separate Poisson regression models
Personal variable Crude analysis (participants=486; jobs=126) Adjusted analysis (participants=485; jobs=126)
Cases Days at risk Rate Ratio (95% CI)/P- Cases Days at risk Adjusted (95% CI) /P-
value Rate Ratio a value
Gender
Male 57 177,896 1.00 Ref 57 177,329 1.00 Ref
Female 8 43,886 0.57 (0.27–1.18) 8 43,886 0.56 (0.26–1.19)
Wald test P=0.131 P=0.133
LBP in 12 months before study
No 18 125,104 1.00 Ref 18 124,537 1.00 Ref
Yes: work unaffected 19 68,340 1.93 (0.93–4.02) 19 68,340 1.91 (0.92–3.96)
Yes: lost work time 28 28,338 6.87 (3.65–12.93)*** 28 28,338 6.78 (3.61–12.74)***
Wald test P<0.001*** P<0.001***
Age (years) b
<30 11 40,757 1.00 Ref
30- 22 80,971 1.01 (0.48–2.09)
40+ 32 99,487 1.19 (0.59–2.42)
Wald test P=0.835
Continuous variable 65 221,215 1.01 (0.98–1.04) 65 221,215 1.00 (0.97–1.04)
Wald test P=0.517 P=0.859
Weight (kg) c
<70 16 57,405 1.00 Ref 16 57,405 1.00 Ref
70- 17 57,309 1.06 (0.48–2.37) 17 57,309 0.79 (0.38–1.67)
80- 20 53,411 1.34 (0.64–2.81) 20 52,844 1.10 (0.55–2.21)
90+ 11 48,622 0.81 (0.36–1.81) 11 48,622 0.63 (0.28–1.40)
Wald test P=0.665 P=0.498
Continuous variable 64 216,747 1.00 (0.99–1.02) 64 216,180 1.00 (0.98–1.02)
Wald test P=0.687 P=0.751
Height (m) d
<1.70 16 58,952 1.00 Ref 16 58,952 1.00 Ref
1.70- 15 43,783 1.26 (0.55–2.90) 15 43,783 0.93 (0.41–2.12)
1.75- 13 48,487 0.99 (0.48–2.05) 13 48,487 0.65 (0.30–1.43)
1.80- 12 41,644 1.06 (0.41–2.73) 12 41,644 0.79 (0.29–2.16)
1.85+ 9 26,786 1.24 (0.53–2.89) 9 26,219 0.89 (0.37–2.14)
Wald test P=0.965 P=0.835
Continuous variable 65 219,652 1.72 (0.10–28.36) 65 219,085 0.30 (0.01–10.58)
36
Personal variable Crude analysis (participants=486; jobs=126) Adjusted analysis (participants=485; jobs=126)
Cases Days at risk Rate Ratio (95% CI)/P- Cases Days at risk Adjusted (95% CI) /P-
value Rate Ratio a value
Wald test P=0.706 P=0.509
BMI (kg/m2) e
Normal (18.5–24.9) 24 85,582 1.00 Ref 24 85,582 1.00 Ref
Overweight (25.0–29.9) 31 91,973 1.20 (0.65–2.23) 31 91,406 0.97 (0.54–1.74)
Obese (30+) 9 37,062 0.87 (0.38–1.96) 9 37,062 0.99 (0.45–2.19)
Wald test P=0.694 P=0.994
Continuous variable 64 214,617 1.01 (0.95–1.07) 64 214,050 1.00 (0.93–1.08)
Wald test P=0.824 P=0.970
Weekly working hours
<40 25 101,731 1.00 Ref 25 101,731 1.00 Ref
40+ 40 120,051 1.36 (0.77–2.39) 40 119,484 1.28 (0.72–2.26)
Wald test P=0.294 P=0.401
Continuous variable 65 221,782 1.04 (1.00–1.09) 65 221,215 1.03 (0.99–1.07)
Wald test P=0.063 P=0.125
Length of employment (years) f
<1 11 25,230 1.00 Ref 11 25,230 1.00 Ref
1- 17 81,570 0.48 (0.19–1.19) 17 81,003 0.42 (0.18–0.97)*
5- 11 41,580 0.61 (0.22–1.65) 11 41,580 0.37 (0.14–0.98)*
10+ 24 71,217 0.77 (0.30–1.96) 24 71,217 0.56 (0.22–1.40)
Wald test P=0.335 P=0.141
Continuous variable 63 219,597 1.02 (0.98–1.05) 63 219,030 1.01 (0.97–1.05)
Wald test P=0.365 P=0.721
Daily driving time (minutes)
0 4 20,642 ------ 4 20,642 ------
1- 30 93,179 1.00 Ref 30 92,612 1.00 Ref
30- 19 69,577 0.85 (0.44–1.64) 19 69,577 0.65 (0.34–1.25)
60+ 12 38,384 0.97 (0.43–2.19) 12 38,384 0.80 (0.39–1.66)
Wald test P=0.789 P=0.636
Continuous variable 65 221,782 1.00 (0.99–1.01) 65 221,512 1.00 (0.98–1.01)
Wald test P=0.976 P=0.509
Exercise regularly
No 33 89,335 1.00 Ref 33 89,335 1.00 Ref
Yes 32 132,447 0.65 (0.37–1.14) 32 131,880 0.57 (0.33–0.96)*

37
Personal variable Crude analysis (participants=486; jobs=126) Adjusted analysis (participants=485; jobs=126)
Cases Days at risk Rate Ratio (95% CI)/P- Cases Days at risk Adjusted (95% CI) /P-
value Rate Ratio a value
Wald test P=0.136 P=0.034*
Current smoker
No 32 134,562 1.00 Ref 32 113,995 1.00 Ref
Yes 33 87,220 1.59 (0.91–2.79) 33 87,220 1.56 (0.88–2.78)
Wald test P=0.105 P=0.126
Any musculoskeletal trouble in the previous 3 months
No 8 61,047 1.00 Ref 8 61,047 1.00 Ref
Yes 57 160,735 2.71 (1.05–7.01)* 57 160,168 1.59 (0.54–4.71)
Wald test P=0.040* P=0.399
Influence on & control over work g
[-10, -5) 17 35,279 1.00 Ref 17 35,279 1.00 Ref
[-5, 0) 22 63,441 0.72 (0.33–1.58) 22 63,441 0.74 (0.35–1.57)
[0, 5) 17 79,104 0.45 (0.20–0.97)* 17 78,537 0.58 (0.27–1.22)
[5, 10] 5 36,133 0.29 (0.09–0.94)* 5 36,133 0.41 (0.13–1.29)
Wald test P=0.082 P=0.359
Continuous variable 61 213,957 0.92 (0.86–0.98)** 61 213,390 0.94 (0.88–1.01)
Wald test P=0.007** P=0.073
Supervisor climate g
[-10, -5) 13 22,244 1.00 Ref 13 22,244 1.00 Ref
[-5, 0) 21 55,782 0.64 (0.28–1.47) 21 55,781 0.59 (0.27–1.30)
[0, 5) 19 79,963 0.41 (0.18–0.92)* 19 79,963 0.45 (0.22–0.92)*
[5, 10] 8 55,968 0.24 (0.09–0.70)** 8 55,401 0.31 (0.12–0.80)*
Wald test P=0.034* P=0.063
Continuous variable 61 213,957 0.92 (0.87–0.98)* 61 213,390 0.94 (0.89–0.99)*
Wald test P=0.012* P=0.030*
Stimulus from the work itself g
[-10, -5) 15 36,733 1.00 Ref 15 36,733 1.00 Ref
[-5, 0) 13 58,264 0.55 (0.24–1.25) 13 58,264 0.68 (0.31–1.47)
[0, 5) 24 72,655 0.81 (0.36–1.80) 24 72,088 1.12 (0.53–2.38)
[5, 10] 9 46,305 0.48 (0.17–1.37) 9 46,305 0.85 (0.31–2.34)
Wald test P=0.354 P=0.500
Continuous variable 61 213,957 0.96 (0.91–1.02) 61 213,390 0.99 (0.93–1.06)
Wald test P=0.230 P=0.848

38
Personal variable Crude analysis (participants=486; jobs=126) Adjusted analysis (participants=485; jobs=126)
Cases Days at risk Rate Ratio (95% CI)/P- Cases Days at risk Adjusted (95% CI) /P-
value Rate Ratio a value
Relations with fellow workers g
[-10, -5) 2 5,154 --- --- 2 5,154 ------
[-5, 0) 13 32,873 1.00 Ref 13 32,873 1.00 Ref
[0, 5) 22 76,097 0.73 (0.34–1.56) 22 76,097 0.95 (0.45–2.00)
[5, 10] 24 99,833 0.61 (0.28–1.31) 24 99,266 1.07 (0.48–2.39)
Wald test P=0.633 P=0.952
Continuous variable 61 213,957 0.97 (0.90–1.05) 61 213,390 1.02 (0.94–1.10)
Wald test P=0.460 P=0.706
Psychological work load g
[-10, -5) 7 19,103 1.00 Ref 7 19,103 1.00 Ref
[-5, 0) 20 59,336 0.92 (0.38–2.20) 20 59,336 0.84 (0.38–1.86)
[0, 5) 24 85,943 0.76 (0.34–1.69) 24 85,943 0.93 (0.46–1.90)
[5, 10] 10 49,575 0.55 (0.20–1.53) 10 49,008 0.70 (0.26–1.85)
Wald test P=0.656 P=0.894
Continuous variable 61 213,957 0.95 (0.90–1.01) 61 213,390 0.98 (0.93–1.04)
Wald test P=0.109 P=0.516
Management commitment to health & safety g
[-10, -5) 9 17,182 1.00 Ref 9 17,182 1.00 Ref
[-5, 0) 12 44,852 0.51 (0.18–1.43) 12 44,852 0.63 (0.24–1.62)
[0, 5) 26 74,722 0.66 (0.25–1.74) 26 74,722 0.79 (0.33–1.85)
[5, 10] 14 77,201 0.35 (0.12–0.98)* 14 76,634 0.54 (0.21–1.40)
Wald test P=0.184 P=0.570
Continuous variable 61 213,957 0.95 (0.90–1.00) 61 213,390 0.97 (0.921–1.02)
Wald test P=0.068 P=0.258
Ref, reference category;
*, statistically significant with P<0.05; **, statistically significant with P<0.01; ***, statistically significant with P<0.001.
a, adjusted for age (continuous), previous LBP experience, and gender
b, 1 missing value, total n=485 for crude analysis.
c, 10 missing values, total n=476 for crude analysis and total n=475 for adjusted analysis.
d, 4 missing values, total n=482 for crude analysis and total n=481 for adjusted analysis.
e, 14 missing values, total n=472 for crude analysis and total n=471 for adjusted analysis.
f, 7 missing values, total n=479 for crude analysis and total n=478 for adjusted analysis.
g, 20 missing values, total n=466 for crude analysis and total n=465 for adjusted analysis.

39
Table 23 Crude and adjusted rate ratios of lost time due to LBP for MAC lifting variables based on worst-case scenarios over all tasks
within a job, estimated using separate Poisson regression models
MAC variable (worst-case scenario) Crude analysis (participants=486; jobs=126) Adjusted analysis a (participants=465; jobs=126)
Cases Days at risk Rate Ratio (95% CI)/P- Cases Days at risk Rate Ratio (95% CI)/P-
value value
Load weight/frequency (lifting)
Green 52 167,711 1.00 Ref 48 160,022 1.00 Ref
Amber/Red/Purple 13 52,204 0.80 (0.41–1.59) 13 51,501 1.11 (0.54–2.27)
Wald test P=0.530 P=0.785
Score test for trend P=0.613 P=0.696
Hand distance from the lower back
Green 0 4,346 NA NA 0 4,346 NA NA
Amber 12 57,960 1.00 Ref 12 57,704 1.00 Ref
Red 53 157,609 1.63 (0.84–3.15) 49 149,473 2.02 (1.06–3.89)*
Wald test P=0.150 P=0.035*
Score test for trend P=0.063 P=0.017*
Vertical lift region
Green 0 556 NA NA 0 556 NA NA
Amber 52 184,600 1.00 Ref 48 177,836 1.00 Ref
Red 13 34,759 1.33 (0.64–2.76) 13 33,131 1.25 (0.33–4.79)
Wald test P=0.449 P=0.747
Score test for trend P=0.417 P=0.721
Trunk twisting/sideways bending
Green 17 48,631 1.00 Ref 16 47,548 1.00 Ref
Amber 28 100,149 0.80 (0.39–1.64) 25 95,375 1.04 (0.49–2.20)
Red 20 71,135 0.80 (0.37–1.76) 20 68,600 0.67 (0.28–1.59)
Wald test P=0.812 P=0.483
Score test for trend P=0.618 P=0.360
Postural constraints
Green 38 137,451 1.00 Ref 34 131,594 1.00 Ref
Amber/Red 27 82,464 1.18 (0.67–2.11) 27 79,929 1.20 (0.63–2.28)
Wald test P=0.566 P=0.588
Score test for trend P=0.294 P=0.234
Grip on the load
Green 0 138 NA NA 0 138 NA NA
Amber 0 4,864 NA NA 0 4,864 NA NA
40
MAC variable (worst-case scenario) Crude analysis (participants=486; jobs=126) Adjusted analysis a (participants=465; jobs=126)
Cases Days at risk Rate Ratio (95% CI)/P- Cases Days at risk Rate Ratio (95% CI)/P-
value value
Red 65 214,913 NA NA 61 206,521 NA NA
Floor surface
Green 55 167,962 1.00 Ref 51 160,893 1.00 Ref
Amber 0 0 NA NA 0 0 NA NA
Red 10 51,953 0.59 (0.28–1.22) 10 50,630 0.76 (0.34–1.68)
Wald test P=0.156 P=0.497
Score test for trend P=0.156 P=0.497
Other environmental factors
Green 54 170,474 1.00 Ref 50 162,916 1.00 Ref
Amber/Red 11 49,441 0.70 (0.35–1.40) 11 48,607 0.76 (0.37–1.55)
Wald test P=0.318 P=0.444
Score test for trend P=0.499 P=0.672
Maximum individual load weight
Green 48 154,321 1.00 Ref 44 146,714 1.00 Ref
Amber/Red/Purple 17 65,594 0.83 (0.45–1.54) 17 64,809 1.14 (0.59–2.18)
Wald test P=0.558 P=0.698
Score test for trend P=0.628 P=0.624
Maximum effort
Green 59 201,542 1.00 Ref 55 193,404 1.00 Ref
Amber/Red/Purple 6 18,373 1.12 (0.32–3.84) 6 18,119 1.53 (0.45–5.22)
Wald test P=0.862 P=0.497
Score test for trend P=0.742 P=0.750
Weighted mean load weight
Green 55 185,303 1.00 Ref 51 177,065 1.00 Ref
Amber 10 34,612 0.97 (0.48–1.99) 10 34,458 0.98 (0.47–2.06)
Red 0 0 NA NA 0 0 NA NA
Wald test P=0.941 P=0.959
Score test for trend P=0.941 P=0.958
Total number of Reds/Purples for individual task
0–1 12 42,736 1.00 Ref 12 41,323 1.00 Ref
2 25 91,384 0.97 (0.43–2.22) 21 87,737 0.90 (0.41–1.97)
3–4 28 85,795 1.16 (0.51–2.64) 28 82,463 1.28 (0.57–2.88)
Wald test P=0.842 P=0.590

41
MAC variable (worst-case scenario) Crude analysis (participants=486; jobs=126) Adjusted analysis a (participants=465; jobs=126)
Cases Days at risk Rate Ratio (95% CI)/P- Cases Days at risk Rate Ratio (95% CI)/P-
value value
Continuous test for trend P=0.560 P=0.362
Total MAC score for individual task
<10 11 40,569 1.00 Ref 11 40,190 1.00 Ref
10- 19 70,949 0.99 (0.43–2.29) 15 67,449 1.06 (0.44–2.59)
12- 9 25,793 1.29 (0.45–3.70) 9 24,193 1.31 (0.49–3.47)
14- 13 37,078 1.29 (0.49–3.40) 13 34,868 1.01 (0.26–3.92)
16+ 13 45,526 1.05 (0.42–2.66) 13 44,823 1.65 (0.63–4.36)
Wald test P=0.954 P=0.814
Continuous test for trend P=0.604 P=0.221
Ref, reference category; NA, not applicable.
*, statistically significant with P<0.05; **, statistically significant with P<0.01; ***, statistically significant with P<0.001.
a, Adjusted for age, previous LBP experience, gender, regular exercise, supervisor climate, number of tasks in job, if the job involved carrying, and if the job involved
team handling.

42
Table 24 Crude and adjusted rate ratios of lost time due to LBP for the proportion of Reds/Purples observed for each MAC lifting
variable, estimated using separate Poisson regression models
MAC variable (Proportion of tasks with Crude analysis (participants=486; jobs=126) Adjusted analysis a (participants=465; jobs=126)
Red/Purple colour coding)
Cases Days at risk Rate Ratio (95% CI)/P- Cases Days at risk Rate Ratio (95% CI)/P-
value value
Load weight/frequency (lifting)
<1% 62 211,514 1.00 Ref 58 203,145 1.00 Ref
1%+ 3 8,701 --- --- 3 8,378 --- ---
Hand distance from the lower back
<1% 12 62,306 1.00 Ref 12 62,050 1.00 Ref
1–24% 15 42,596 1.83 (0.79–4.22) 13 39,065 2.53 (0.54–11.76)
25–49% 7 54,225 0.67 (0.27–1.69) 5 51,093 0.79 (0.30–2.12)
50–74% 20 33,361 3.11 (1.40–6.91)** 20 32,289 3.63 (1.66–7.98)**
75–99% 5 7,846 3.31 (1.04–10.52)* 5 7,846 2.79 (0.77–10.08)
99%+ 6 19,581 1.59 (0.56–4.54) 6 19,180 1.62 (0.58–4.55)
Wald test P=0.007** P=0.006**
Continuous test for trend P=0.077 P=0.033*
Vertical lift region
<1% 52 185,156 1.00 Ref 48 178,392 1.00 Ref
1%+ 12 34,759 1.33 (0.64–2.77) 13 33,131 1.25 (0.33–4.80)
Wald test P=0.444 P=0.746
Continuous test for trend P=0.355 P=0.569
Trunk twisting/sideways bending
<1% 45 148,780 1.00 Ref 41 142,923 1.00 Ref
1–24% 12 35,942 1.10 (0.52–2.33) 12 33,407 0.69 (0.20–2.39)
25%+ 8 35,193 0.75 (0.31–1.81) 8 35,193 0.63 (0.28–1.41)
Wald test P=0.761 P=0.477
Continuous test for trend P=0.300 P=0.144
Postural constraints
<1% 61 216,708 1.00 Ref 58 208,316 1.00 Ref
1%+ 4 3,207 --- --- 4 3,207 --- ---
Grip on the load
<1% 0 5,002 NA NA 0 5,002 NA NA
1–24% 9 19,489 1.00 Ref 9 19,489 1.00 Ref

43
MAC variable (Proportion of tasks with Crude analysis (participants=486; jobs=126) Adjusted analysis a (participants=465; jobs=126)
Red/Purple colour coding)
Cases Days at risk Rate Ratio (95% CI)/P- Cases Days at risk Rate Ratio (95% CI)/P-
value value
25–74% 15 41,834 0.78 (0.29–2.09) 15 41,834 3.62 (0.17–76.49)
75–99% 5 21,464 0.50 (0.12–2.06) 5 21,464 2.57 (0.13–49.82)
>99% 36 132,123 0.59 (0.25–1.41) 36 132,126 2.35 (0.10–53.78)
Wald test P=0.590 P=0.617
Continuous test for trend P=0.456 P=0.744
Floor surface
<1% 55 167,962 1.00 Ref 51 160,893 1.00 Ref
1%+ 10 51,953 0.59 (0.28–1.22) 10 50,630 0.76 (0.34–1.68)
Wald test P=0.156 P=0.497
Continuous test for trend P=0.125 P=0.457
Other environmental factors
<1% 64 219,804 1.00 Ref 60 211,412 1.00 Ref
1%+ 1 111 --- --- 1 111 --- ---
Ref, reference category; NA, not applicable.
*, statistically significant with P<0.05; **, statistically significant with P<0.01; ***, statistically significant with P<0.001.
a, Adjusted for age, previous LBP experience, gender, regular exercise, supervisor climate, number of tasks in job, if the job involved carrying, and if the job involved
team-handling.

44
Table 25 Comparison of alternative categorisations for load weight/frequency,
maximum individual load weight, maximum effort, and weighted mean load
weight obtained using different load weight/frequency charts
MAC variable (worst- Cases Days at risk Adjusted (95% CI) AIC or P-
case scenario) and chart Rate Ratio a value
Load weight/frequency original chart AIC=486.28
Green 48 160,022 1.00 Ref
Amber 10 43,123 1.01 (0.45–2.28)
Red 3 6,469 2.03 (0.67–6.11)
Purple 0 1,909 NA NA
Wald test P=0.447
Score test for trend P=0.696
Load weight/frequency alternative chart 1 AIC=486.66
Green 29 93,064 1.00 Ref
Amber 22 78,069 0.89 (0.43–1.85)
Red 10 38,481 1.29 (0.54–3.07)
Purple 0 1,909 NA NA
Wald test P=0.781
Score test for trend P=0.799
Load weight/frequency alternative chart 2 AIC=486.93
Green 29 93,064 1.00 Ref
Amber 27 103,100 0.96 (0.50–1.86)
Red 5 13,450 1.36 (0.46–4.01)
Purple 0 1,909 NA NA
Wald test P=0.834
Score test for trend P=0.887
Maximum individual load weight original chart AIC=486.24
Green 44 146,714 1.00 Ref
Amber 14 56,431 1.07 (0.53–2.14)
Red 3 6,469 2.07 (0.68–6.28)
Purple 0 1,909 NA NA
Wald test P=0.438
Score test for trend P=0.624
Maximum individual load weight alternative chart 1 AIC=486.66
Green 29 93,064 1.00 Ref
Amber 22 78,069 0.89 (0.43–1.85)
Red 10 38,481 1.29 (0.54–3.07)
Purple 0 1,909 NA NA
Wald test P=0.781
Score test for trend P=0.799
Maximum individual load weight alternative chart 2 AIC=487.1.2
Green 29 93,064 1.00 Ref
Amber 27 101,610 0.98 (0.50–1.89)
Red 5 14,940 1.27 (0.43–3.76)
Purple 0 1,909 NA NA
Wald test P=0.901
Score test for trend P=0.929
Maximum effort original chart AIC=483.84
Green 55 193,404 1.00 Ref
Amber 3 5,509 --- ---
Red/Purple 3 7,017 --- ---
Purple 0 5,593 NA NA
Wald test P=0.510
Score test for trend P=0.750

45
MAC variable (worst- Cases Days at risk Adjusted (95% CI) AIC or P-
case scenario) and chart Rate Ratio a value
Maximum effort alternative chart 1 AIC=485.42
Green 46 171,833 1.00 Ref
Amber 12 24,591 1.53 (0.27–8.67)
Red 3 9,442 --- ---
Purple 0 5,567 NA NA
Wald test P=0.852
Score test for trend P=0.917
Maximum effort alternative chart 2 AIC=485.21
Green 46 171,833 1.00 Ref
Amber 12 26,652 1.20 (0.20–7.05)
Red 3 7,381 --- ---
Purple 0 5,567 NA NA
Wald test P=0.772
Score test for trend P=0.983
Weighted mean load weight original chart AIC=484.10
Green 51 177,065 1.00 Ref
Amber 10 34,458 0.98 (0.47–2.06)
Red 0 0 NA NA
Purple 0 0 NA NA
Wald test P=0.959
Score test for trend P=0.958
Weighted mean load weight alternative chart 1 AIC=485.17
Green 39 151,843 1.00 Ref
Amber 19 45,568 1.36 (0.61–3.02)
Red 3 14,112 --- ---
Purple 0 0 NA NA
Wald test P=0.671
Score test for trend P=0.869
Weighted mean load weight alternative chart 2 AIC=485.70
Green 39 151,843 1.00 Ref
Amber 19 51,559 1.12 (0.51–2.48)
Red 3 8,121 --- ---
Purple 0 0 NA NA
Wald test P=0.799
Score test for trend P=0.559
a: adjusted for age, previous LBP experience, gender, regular exercise, supervisor climate, number of
tasks in job, if the job involved carrying, and if the job involved team-handling using separate Poisson
regression models; Ref, reference category.

46
Table 26 Crude and adjusted rate ratios of lost time due to LBP for MAC carrying variables based on the worst-case scenario over all
carrying tasks within a job, estimated using separate Poisson regression models
MAC variable (worst-case scenario) Crude analysis (participants=227; jobs=54) Adjusted analysis (participants=220; jobs=54)
Cases Days at risk Rate Ratio (95% CI)/P- Cases Days at risk Adjusted (95% CI)/P-
value Rate Ratio a value
Load weight/frequency (carrying)
Green 25 96,461 NA NA 22 92,795 NA NA
Amber/Red/Purple 0 9,494 NA NA 0 9,949 NA NA
Hand distance from the lower back
Green 1 16,392 NA NA 1 16,392 NA NA
Amber 9 46,445 1.00 Ref 9 45,105 1.00 Ref
Red 15 43,573 1.99 (0.79–5.02) 12 41,247 1.55 (0.62–3.92)
Wald test P=0.145 P=0.351
Score test for trend P=0.031* P=0.123
Asymmetrical trunk load
Green 11 25,673 1.00 Ref 9 24,233 1.00 Ref
Amber/Red 14 80,737 0.40 (0.17–0.94)* 13 78,511 0.65 (0.21–1.97)
Wald test P=0.036* P=0.445
Score test for trend P=0.259 P=0.817
Postural constraints
Green 19 78,274 1.00 Ref 16 74,608 1.00 Ref
Amber 6 28,136 0.88 (0.32–3.40) 6 28,136 0.72 (0.27–1.89)
Red 0 0 NA NA 0 0 NA NA
Wald test P=0.800 P=0.505
Score test for trend P=0.800 P=0.505
Grip on the load
Green 0 1,167 NA NA 0 1,167 NA NA
Amber/Red 25 105,243 NA NA 22 101,577 NA NA
Floor surface
Green 21 87,329 1.00 Ref 18 84,224 1.00 Ref
Amber 0 0 NA NA 0 0 NA NA
Red 4 19,081 --- --- 4 18,520 --- ---
Other environmental factors
Green 24 98,081 1.00 Ref 21 94,415 1.00 Ref
Amber 1 8,329 --- --- 1 8,329 --- ---

47
MAC variable (worst-case scenario) Crude analysis (participants=227; jobs=54) Adjusted analysis (participants=220; jobs=54)
Cases Days at risk Rate Ratio (95% CI)/P- Cases Days at risk Adjusted (95% CI)/P-
value Rate Ratio a value
Red 0 0 NA NA 0 0 NA NA
Carry distance
Green 12 33,380 1.00 Ref 11 32,383 1.00 Ref
Amber/Red 13 73,030 0.50 (0.21–1.15) 11 70,361 0.41 (0.12–1.34)
Wald test P=0.104 P=0.140
Score test for trend P=0.108 P=0.181
Obstacles en route
Green 13 64,184 1.00 Ref 12 62,387 1.00 Ref
Amber/Red 12 42,226 1.40 (0.60–3.27) 10 40,357 1.33 (0.49–3.61)
Wald test P=0.432 P=0.577
Score test for trend P=0.460 P=0.587
Total number of Reds/Purples for individual task
0–1 12 51,748 1.00 Ref 12 50,969 1.00 Ref
2–4 13 54,662 1.03 (0.44–2.37) 10 51,775 1.08 (0.41–2.84)
Wald test P=0.953 P=0.880
Continuous test for trend P=0.871 P=0.596
Total MAC score for individual task
0–7 8 41,078 1.00 Ref 8 40,299 1.00 Ref
8–9 7 31,556 1.14 (0.40–3.24) 5 29,794 1.03 (0.32–3.29)
10–19 10 33,776 1.52 (0.58–3.99) 9 32,651 1.40 (0.51–3.84)
Wald test P=0.691 P=0.774
Continuous test for trend P=0.917 P=0.813
Ref, reference category; NA, not applicable.
*, statistically significant with P<0.05; **, statistically significant with P<0.01.
a, adjusted for age, previous LBP experience, gender, regular exercise, supervisor climate, number of tasks in job, and if the job involved team-handling.

48
Table 27 Crude and adjusted rate ratios of lost time due to LBP for the proportion of Reds/Purples observed for each MAC carrying
variable, estimated using separate Poisson regression models
MAC variable (Proportion of tasks with Crude analysis (participants=227; jobs=54) Adjusted analysis (participants=220; jobs=54)
Red/Purple colour coding) Cases Days at risk Rate Ratio (95% CI) Cases Days at risk Adjusted (95% CI)
Rate Ratio a
Load weight/frequency (carrying)
<1% 25 104,501 NA NA 22 100,835 NA NA
1%+ 0 1,909 NA NA 0 1,909 NA NA
Hand distance from the lower back
<1% 10 62,837 1.00 Ref 10 61,497 1.00 Ref
1–49% 9 22,729 2.49 (1.02–6.06)* 6 20,466 1.68 (0.55–5.16)
50%+ 6 20,844 1.81 (0.56–5.89) 6 1.97 (0.61–6.43)
Wald test P=0.127 P=0.445
Continuous test for trend P=0.372 P=0.304
Asymmetrical trunk/load
<1% 23 101,718 1.00 Ref 20 98,052 1.00 Ref
1%+ 2 4,692 --- --- 2 4,692 --- ---
Postural constraints
<1% 25 106,410 NA NA 22 102,744 NA NA
1%+ 0 0 NA NA 0 0 NA NA
Grip on the load
<1% 2 6,371 --- --- 2 6,371 --- ---
1–49% 8 21,458 1.00 Ref 5 19,220 1.00 Ref
50%+ 15 78,581 0.51 (0.21–1.24) 15 77,153 0.94 (0.31–2.87)
Wald test P=0.139 P=0.911
Continuous test for trend P=0.235 P=0.647
Floor surface
<1% 21 87,329 1.00 Ref 18 84,224 1.00 Ref
1%+ 4 19,081 --- --- 4 18,520 --- ---
Other environmental factors
<1% 25 106,410 NA NA 22 102,744 NA NA
1%+ 0 0 NA NA 0 0 NA NA
Carry distance
<1% 22 84,472 1.00 Ref 20 82,782 1.00 Ref
1%+ 3 21,938 --- --- 2 19,962 --- ---

49
MAC variable (Proportion of tasks with Crude analysis (participants=227; jobs=54) Adjusted analysis (participants=220; jobs=54)
Red/Purple colour coding) Cases Days at risk Rate Ratio (95% CI) Cases Days at risk Adjusted (95% CI)
Rate Ratio a
Obstacles en route
<1% 25 105,847 NA NA 22 102,181 NA NA
1%+ 0 563 NA NA 0 563 NA NA
Ref, reference category; NA, not applicable.
*, statistically significant with P<0.05;
a, for age, previous LBP experience, gender, regular exercise, supervisor climate, number of tasks in job, and if the job involved team handling.

50
Table 28 Cases and days at risk of lost time for MAC team-handling variables based
on the worst-case scenario over all team-handling tasks within a job
MAC variable Crude analysis Adjusted analysis
(participants=80; jobs=24) (participants=76; jobs=24) a
Cases Days at risk Cases Days at risk
Load weight
Green 8 36,127 7 34,496
Amber/Red/Purple 0 0 0 0
Hand distance from the lower back
Green 5 17,710 4 16,713
Amber/Red 3 15,598 3 15,342
Vertical lift region
Green 0 0 0 0
Amber/Red 8 36,127 7 34,496
Trunk twisting/sideways bending
Green 5 21,330 4 19,699
Amber/Red 3 14,797 3 14,797
Postural constraints
Green 6 32,252 5 30,621
Amber 2 3,875 2 3,875
Red 0 0 0
Grip on the load
Green 0 0 0 0
Amber/Red 8 36,127 7 34,496
Floor surface
Green 8 28,637 7 27,006
Amber 0 0 0 0
Red 0 7,490 0 7,490
Other environmental factors
Green 6 30,770 5 29,139
Amber/Red 2 5,357 2 5,357
Communication, co-ordination and control
Green 7 30,213 6 28,582
Amber/Red 1 5,914 1 5,914
Red 0 0 0 0
Total number of Reds/Purples for individual task c
1 6 18,597 5 17,344
2–3 2 17,530 2 17,152
Total MAC score for individual task c
5–9 6 19,567 5 18,314
10–13 2 16,560 2 16,182
Ref, reference category; CI, confidence interval; NA, not applicable.
a, adjusted for age, previous LBP experience, gender, regular exercise, supervisor climate, number of
tasks in job, and if the job involved carrying.

51
6. REFERENCES

1 HSE and HSL (2003). Manual Handling Assessment Charts, INDG383. (Sudbury
Suffolk: HSE Books), INDG383. Available at http://www.hse.gov.uk/pubns/indg383.pdf.

2 Monnington, S.C., Pinder, A.D.J. and Quarrie, C. (2002). Development of an inspection


tool for manual handling risk assessment. (Sheffield: Health and Safety Laboratory),
HSL/2002/30. Available at http://www.hse.gov.uk/research/hsl_pdf/2002/hsl02-30.pdf .

3 Monnington, S.C., Quarrie, C.J., Pinder, A.D.J. and Morris, L.A. (2003). Development of
Manual Handling Assessment Charts (MAC) for Health and Safety Inspectors. In:
McCabe, P.T. (Ed.) Contemporary Ergonomics 2003 (London: Taylor & Francis), pp. 3-
8.

4 Waters, T.R., Putz-Anderson, V. and Garg, A. (1994). Applications Manual for the
Revised NIOSH Lifting Equation. (Cincinnati, Ohio: NIOSH), DHHS (NIOSH)
Publication No. 94-110.

5 Pinder, A.D.J. and Frost, G.A. (2009). The 1991 NIOSH Lifting Equation does not
predict low back pain. In: Marley, R.J., Kumar, A.R., Ware, B.F., and Lockhart, T.E.
(Eds.) Proceedings of the XXIst Annual International Occupational Ergonomics and
Safety Conference, Dallas, Texas, USA, 11-12 June 2009 International Society for
Occupational Ergonomics and Safety), pp. 229-235.

6 Pinder, A.D.J. and Frost, G.A. (2011). Prospective evaluation of the 1991 NIOSH Lifting
Equation. (Sudbury, Suffolk: HSE Books), RR901, 204 pages. Available at
http://www.hse.gov.uk/research/rrhtm/rr901.htm .

7 Pinder, A.D.J. and Frost, G.A. (2012). Evaluation of the ability of the 1991 NIOSH
Lifting Equation to predict loss of time from work due to low back pain: an 18 month
prospective cohort study of industrial workers. Ergonomics, In preparation

8 Ayoub, M.M. (1996). Field Evaluation of NIOSH Lifting Equations. Revised Protocol.
(Lubbock, Texas: Institute for Ergonomics Research, Texas Tech University), 63 pages.

9 Dempsey, P.G. and Westfall, P.H. (1997). Developing explicit risk models for predicting
low-back disability. A statistical perspective. International Journal of Industrial
Ergonomics, 19, (6), 483-497.

10 Dempsey, P.G. and Fathallah, F.A. (1999). Application issues and theoretical concerns
regarding the 1991 NIOSH equation asymmetry multiplier. International Journal of
Industrial Ergonomics, 23, (3), 181-191.

11 Dempsey, P.G. (2001). Field investigation of the usability of the Revised NIOSH Lifting
Equation. In: Proceedings of the Human Factors and Ergonomics Society 45th Annual
Meeting (Santa Monica, CA: The Human Factors and Ergonomics Society), Volume 2,
pp. 972-976.

12 Dempsey, P.G. (2002). Usability of the revised NIOSH lifting equation. Ergonomics, 45,
(12), 817-828.

52
13 Dempsey, P.G. (2003). A survey of lifting and lowering tasks. International Journal of
Industrial Ergonomics, 31, (1), 11-16.

14 Dempsey, P.G., Sorock, G.S., Cotnam, J.P., Ayoub, M.M., Westfall, P.H., Maynard, W.,
Fathallah, F. and O'Brien, N. (2000). Field evaluation of the revised NIOSH lifting
equation. In: Proceedings of the XIVth Triennial Congress of the International
Ergonomics Association and 44th Annual Meeting of the Human Factors and Ergonomics
Society (Santa Monica, CA: Human Factors and Ergonomics Society), Volume 5, pp. 37-
40.

15 Dempsey, P.G., Burdorf, A., Fathallah, F.A., Sorock, G.S. and Hashemi, L. (2001).
Influence of measurement accuracy on the application of the 1991 NIOSH equation.
Applied Ergonomics, 32, (1), 91-99.

16 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.

17 HSE, (2003), Manual handling assessment chart. MAC tool, (Health and Safety
Executive). Available at http://www.hse.gov.uk/msd/mac/index.htm ; accessed on 12-8-
2011.

18 Pinder, A.D.J. (2002). Power considerations in estimating sample size for evaluation of
the NIOSH lifting equation. In: McCabe, P.T. (Ed.) Contemporary Ergonomics 2002
(London: Taylor & Francis), pp. 82-86.

19 Nelson, N.A. and Hughes, R.E. (2009). Quantifying relationships between selected
work-related risk factors and back pain: A systematic review of objective biomechanical
measures and cost-related health outcomes. International Journal of Industrial
Ergonomics, 39, (1), 202-210.

20 Boda, S.V., Bhoyar, P. and Garg, A. (2010). Validation of Revised NIOSH Lifting
Equation and 3D SSP Model to Predict Risk of Work-Related Low Back Pain. In:
Human Factors and Ergonomics Society Annual Meeting Proceedings, pp. 1185-1189.

21 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.

22 Snook, S.H. and Ciriello, V.M. (1991). The design of manual handling tasks: Revised
tables of maximum acceptable weights and forces. Ergonomics, 34, (9), 1197-1213.

23 Pheasant, S. and Haslegrave, C.M. (2006). Bodyspace. Anthropometry, Ergonomics and


the Design of Work. (London: Taylor & Francis), 3rd Edition.

24 David, G., Woods, V., Li, G. and Buckle, P. (2008). The development of the Quick
Exposure Check (QEC) for assessing exposure to risk factors for work-related
musculoskeletal disorders. Applied Ergonomics, 39, (1), 57-69.

25 McAtamney, L. and Corlett, E.N. (1993). RULA: A Survey Method for the Investigation
of Work-Related Upper Limb Disorders. Applied Ergonomics, 24, (2), 91-99.

53
26 Punnett, L., Fine, L.J., Keyserling, W.M., Herrin, G.D. and Chaffin, D.B. (1991). Back
disorders and nonneutral trunk postures of automobile assembly workers. Scandinavian
Journal of Work, Environment and Health, 17, (5), 337-346.

27 HSE (2004). Manual Handling: Manual Handling Operations Regulations 1992 (as
amended). Guidance on Regulations. (Sudbury, Suffolk: HSE Books), L23, Third Edition.

28 van den Heuvel, S.G., Ariens, G.A., Boshuizen, H.C., Hoogendoorn, W.E. and Bongers,
P.M. (2004). Prognostic factors related to recurrent low-back pain and sickness absence.
Scandinavian Journal of Work, Environment and Health, 30, (6), 459-467.

29 Pinder, A.D.J. (2011). Risk assessment of manual handling involving variable loads
and/or variable frequencies: literature review and proposed V-MAC assessment tool.
(Sudbury, Suffolk: HSE Books), RR828, 94 pages. Available at
http://www.hse.gov.uk/research/rrhtm/rr838.htm .

30 Rowe, M.L. (1969). Low back pain in industry: A position paper. Journal of
Occupational Medicine, 11, (4), 161-169.

31 Troup, J.D.G., Martin, J.W. and Lloyd, D.C.E.F. (1981). Back Pain in Industry. A
Prospective Survey. Spine, 6, (1), 61-69.

32 Ferguson, S.A., Marras, W.S. and Burr, D. (2005). Workplace design guidelines for
asymptomatic vs. low-back-injured workers. Applied Ergonomics, 36, (1), 85-95.

33 Feyer, A.M., Herbison, P., Williamson, A.M., De Silva, I., Mandryk, J., Hendrie, L. and
Hely, M.C. (2000). The role of physical and psychological factors in occupational low
back pain: a prospective cohort study. Occupational and Environmental Medicine, 57,
(2), 116-120.

34 Pinder, A.D.J. and Wegerdt, J.F. (2008). Feasibility of carrying out an ergonomics
intervention study to prevent the incidence of musculoskeletal disorders. (Sudbury,
Suffolk: HSE Books), RR626, 104 pages. Available at
http://www.hse.gov.uk/research/rrhtm/rr626.htm .

35 Power, C., Frank, J., Hertzman, C., Schierhout, G. and Li, L. (2001). Predictors of low
back pain onset in a prospective British study. American Journal of Public Health, 91,
(10), 1671-1678.

36 Troup, J.D., Foreman, T.K., Baxter, C.E. and Brown, D. (1987). 1987 Volvo award in
clinical sciences. The perception of back pain and the role of psychophysical tests of
lifting capacity. Spine, 12, (7), 645-657.

54
Published by the Health and Safety Executive 12/14
Health and Safety
Executive

Validation of the HSE Manual handling


Assessment Charts as predictors of
work-related low back pain
The aim of this research was to ascertain whether HSE’s
‘Manual handling Assessment Charts’ (MAC tool) could be
used to predict workers losing time from work due to low back
pain (LBP). Results from the study suggest that as the ‘Hand
distance from the lower back’ increased, the risk of lost time
due to LBP increased. For each 10 cm increase, the rate of
lost time increased by approximately 20%. No evidence of
relationships between other risk factors in the MAC and lost
time was found. There was no evidence that the rate of lost
time due to LBP increased with either increasing total MAC
lifting score or total MAC carrying score.

Due to imprecision in the model estimates (wide confidence


intervals), the lack of statistically significant results, and the
limitations of the data, it was decide that it would not be
appropriate to alter the scoring system currently used in the
MAC based on these data. Duty holders should be confident
in carrying on using the MAC tool as the risk factors for LBP
included were identified as important by earlier studies.

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.

RR1026

www.hse.gov.uk

Вам также может понравиться