Assessment of physical work load in epidemiologic studies: common measurement metrics for exposure assessment

Transcription

1 ERGONOMICS, 1997, VOL. 40, NO. 1, 51±61 Assessment of physical work load in epidemiologic studies: common measurement metrics for exposure assessment R. WELLS, R. NORMAN, P. NEUMANN, D. ANDREWS, J. FRANK², H. SHANNON² and M. KERR² Department of Kinesiology, University of Waterloo, Waterloo ² Institute for Work and Health, Toronto, Canada Keywords: Occupational epidemiology; Exposure assessment; Measurement; Video; EMG; Questionnaire; Low back; Work sampling. There are many possible means of determining exposure ranging from self-reports of physical exposure to measures of muscle activationsand estimated spinal loads. In epidemiologic studies, issues of validity make instrumented measures preferable, however issues of cost and practicability tend to force investigators to less costly but less valid and less reliable measures of exposure, such as self-report questionnaires. This paper presents a method by which estimates of exposure from self-report questionnaires, expert observers, work sampling, video analysis and electromyograms can be reported in a common metric, Newtons of force on a tissue, and show, as an example of its application, estimation of spinal compression on auto workers. A common metric allows a exible approach to selection of measurement methods in occupational settings: no matter which instrument is used the results can be combined to provide an overall picture of exposure. This approach to exposure assessment for the low back allows for comparabilityacross studies and settings. 1. Introduction In epidemiologic studies of work and musculoskeletal health the issue of best quantifying exposure for large numbers of participants performing a wide range of jobs has come to the fore, notably in Stock (1991), Hagberg (1992), Wiktorin et al. (1993, 1996), Kerr et al. (1993), Kumar (1994), Winkel and Mathiassen (1994). Central to these proposals is the concept of a consistent `internal exposure or `dose within the body, suggested as being measured in a metric such as Newtons of force (Winkel and Mathiassen 1994). The optimal measure of exposure would have a number of characteristics de ned: (1) the point within the body being considered; (2) the intensity or magnitude of the exposure; (3) the time variation of the exposure, and (4) the total duration of the exposure (Winkel and Mathiassen 1994, Hagberg et al. 1995). It is necessary to assess all these characteristics to property de ne an individual s `exposure. The use of a `common metric (consistent measure) for exposure across diœerent jobs is one approach to meeting the challenge presented by Winkel and Mathiassen (1994). This paper presents a method by which estimates of exposure from self-report questionnaires, trained observers, work sampling, video analysis and electromyograms can be reported in a common metric, Newtons of force on a tissue, and show, Correspondence, R. Wells. 0014±0139/97 $ Taylor & Francis Ltd

2 52 R. Wells et al. as an example of its application, estimation of spinal compression on automobile assembly workers. 2. Exposure assessment Exposure to physical stressors has received increased scrutiny recently in an attempt to adapt the terms and concepts used in other branches of occupational epidemiology to the study of musculoskeletal disorders (Armstrong et al. 1993, Winkel and Mathiassen 1994, Hagberg et al. 1995). A major challenge is to develop exposure metrics that are related to musculoskeletal disorders, are relevant to occupational settings, and which are capable of exible and cost eœective application across a wide range of occupations and occupational settings. This is currently not done systematically and thus the completeness of many studies and the comparison of studies is rendered problematic (Winkel and Westgaard 1992). In addition, in epidemiologic studies, issues of validity make instrumented measures preferable. However, issues of cost and practicability tend to force investigators to less costly but less valid and less reliable measures of exposure such as self-report questionnaires. The use of the approach presented here could improve the process of comparability of studies as suggested by Leaman (1994). Many studies of the relationship between work and musculoskeletal disorders have concentrated upon `stereotypic work tasks with short cycle times, most likely because of resource and time constraints. One of the more di cult challenges facing ergonomists and epidemiologists is to assess the more commonly occurring jobs with long cycle times or irregular work patterns. This often forces investigators to use diœerent approaches and instruments for assessing these two extremes making comparability of data di cult or impossible. There are many possible exposure measures ranging from job titles (classi cations) and self-reports of physical exposure to direct measures of muscle activations and estimations of muscle loads (Hagberg et al. 1995). Typically, instrumented measures of physical stressors, which might be expected to be the best measures of current exposure, tend to be time consuming and expensive and many investigators have opted for questionnaire instruments instead. Self-reports from questionnaires have the advantage of being cheap to administer, while allowing for measures of a wide range of occupational and non-occupational risk factors both in current and past employment (Wiktorin et al. 1993, 1996). Unfortunately, the results of questionnaire validation studies often have not been encouraging (Buckle et al. 1986, Wiktorin et al. 1993). It appears that respondents can identify whether exposure to some stressor, such as vibration or lifting, has occurred but they do not give reliable information either on the nature or on the magnitude of the exposure (Wiktorin et al. 1993). A single measurement method is unlikely to be feasible given the wide variety of occupational settings and purposes of the investigators. An alternative is to develop a framework that allows a combination of measurement methods to take advantage of the best features of each. This paper will develop such a framework for exposure assessment for low-back pain using, only as an example, a common metric of Newtons of spine compression. This facilitates comparisons between jobs and studies and also allows a `toolbox approach to selection of measurement methods in occupational settings. No matter which instrument is used, the results can be combined to provide an overall picture of low-back exposure in the same units of measurement.

3 Common measurement metrics for exposure assessment 53 There are many reported physical risk factors for low-back pain. Winkel and Mathiassen (1994) describe three categories of variables: physical (Garg 1989), individual, and psychosocial (Bongers 1993). Here physical exposure is the focus, yet even within this category measurement metrics include percent of time in given postures (%), maximum weight lifted (kg), maximum torso angle (degrees), spinal compression (N) and spinal shear (N). In this paper, lumbar compression (N) will be used as a common metric across measurement methods. There is no intention to suggest the pre-eminence of compression as a single risk factor. However, spinal compression as an exposure measure does contain biomechanical justi cation and is, therefore, a reasonable example to present. It encompasses many of the reported risk factors uncovered by studies using questionnaire instruments such as non-neutral trunk postures and heavy lifting. Moreover, lumbar compression has been used extensively as a measure of exposure (Kumar 1990). In addition, Marras et al. (1993) report trunk moment to be a strong risk factor; lumbar compression is highly linearly related to trunk moment. While it is infrequent to see spinal motion unit failure associated with reported occupational low-back pain, high loading of the spine is almost inseparable from high loads on other spinal tissues such as muscle and ligament (McGill and Norman 1986). Notwithstanding, the above rationale for using spinal compression (expressed in Newtons) as an exposure measure, it must be emphasized that this paper is intended to present a method by which exposure information can be obtained from a wide variety of methods; self-report questionnaires, expert observers, work sampling, video analysis and electromyograms. An example of how this approach has been used to structure exposure assessment in a large case-control study of risk factors for low-back pain in automobile workers will then be presented. 3. Estimation of spinal compression from disparate measurement methods The primary problem in arriving at a common metric is to convert inputs from diœerent measurement methods into estimates of exposure in the same units of measurement, in this case spinal compression in Newtons. Five diœerent methods were used to obtain input information. Five separate sources of data were used: postural and load information provided from a selfadministered questionnaire, trained observers using a similar instrument, a posture and load sampling process by trained observers noting worker postures by selection from diagrams at random times during the work period (up to 3h of observation), postural information taken from freeze frame video of the participant s most stressful posture determined by the trained observers and EMG from erector spinae muscles normalized to spinal compression. A short description of each measurement method is presented later. A biomechanical model of the lumbar spine was used as the core of all ve exposure measure methods and therefore merits a brief description. The model is a quasi-dynamic, two-dimensional linked-segment model that comprises 15 segments. It is quasi-dynamic in the sense that inertial forces acting on the hands in lifts, lowers, pulls or pushes, rather than static load weights only, can be input if their magnitudes and directions are known. Body segment accelerations are not included, therefore the model is not completely dynamic. Asymmetric body postures can be input. Anthropometrics for segment masses and locations of mass centres for men and women are taken from Plagenhoef (1971) and Zatsiorsky and Seluyanov (1983) or can be input by the user. The participant s body weight and gender must be speci ed. Postural

4 54 R. Wells et al. input is obtained from digitized x,y coordinates of body joints or via on-screen manipulation of a moveable manikin. The model calculates forces and moments at each joint starting at the wrist of each arm and proceedingto the elbow, shoulder, 7th cervical vertebra and down to L4/L5. Compression and shear forces at L4/L5 are estimated from knowledge of the moment of forceand reaction forcesat this motion unit. A 6cm moment arm length is used to represent the geometry of a single equivalent torso extensor `muscle for the estimation of the compression component. This moment arm length was incorporated as a result of ndings from work with a fully dynamic, much more anatomically detailed, EMG-assisted model (McGill and Norman 1986; Potvin et al. 1991). None of the spine models that have been presented to date in the literature have been directly validated by comparison of model estimates of muscle force, spinal compression or shear with direct measures of these variables in the same units of measurement. Technically this type of validation is currently not possible. Consequently anatomical and physiological content validity in the structure and function of these types of models is important. An attempt has to be made to incorporateasmuch content validity as possible, but assumptions and simpli cations are present in all models. The impact of errors in model output estimates of compression are minimized in the example data presented in this paper since the same spine model was used to process input data from all methods described to provide an estimate of spinal compression. 4. Self-report questionnaire The questionnaire asks respondents to identify their most demanding tasks and their relative durations, and for each task to identify common postures and loads (cf. Kumar 1990). Postures of the trunk and arms arechosen froma number of images and the force exerted is estimated. Software was used to create joint co-ordinates from the indicated postures. Each posture reported, combined with loading parameters, was then input into the biomechanical spine model to determine the lumbar moment of force and spine compression (Andrews et al. 1996). Time and frequency information were gathered to enable assessment of both peak and cumulative loading parameters. The same questions were also answered by trained observers and processed in the same manner. The worker also gave information on many other variables such as vibration and work rates. Of course, worker perception of work remains an important measure and variables such as perceived eœort and pain were assessed as well using Borg-type scaling techniques (Andrews et al. 1996). 5. Posture and load sampling Information on the distribution of low-back compression was also obtained from a work sampling approach. Although a number of similar schemes exist, for example, RULA (McAttamney and Corlett 1993, Keyserling 1986), OWAS (Karhu et al. 1977), they did not give output measures that could be converted to a lumbar compression metric and thus could not be used in this approach. Postures of the back and arms, the load handled and the load direction were sampled for up to 3h at mean intervals of 15s to 1min depending on the cycle time of the job. Once again, the aim was to obtain both postural information and the force data required to estimate low-back demands. The posture was classi ed into one of 13 back postures and the load was classi ed as being close, midway or far from the shoulder in the horizontal direction. Each combination of back posture, arm position,

5 Common measurement metrics for exposure assessment 55 load and load direction was input into a two- or three-dimensional biomechanical model, as appropriate, and a `look-up table of joint moments of force and estimates of spinal compression and shear for 50th percentile male or female created (Wells et al. 1995). The work sampling procedure was thus capable of estimating not only a distribution of postures adopted but also the peak load and the amplitude distribution of spinal loading produced by the task throughout the duration of the sampling period. 6. Trunk electromyography (EMG) Erector spinae EMG recorded during work can be scaled to a standardized isometric reference contraction that is then used to produce a continuous estimate of lumbar compression from the EMG signal. This is a simpli ed version of an EMG processing approach to scale trunk electromyograms to spinal compression developed speci - cally for use in eld settings (Wells et al. 1994). The original method, described by McGill and Norman (1986) and by Potvin and Norman (1993), uses as input a large number of sites of myoelectric activity and lumbar spine kinematics to scale and modulate the myoelectric activity to predict spinal load. The simpli ed technique has been shown to coincide acceptably with biomechanical model estimates of compression when presented in the form of an Amplitude Probability Distribution Function, APDF (Potvin et al. 1990, Mientjes 1996). This study utilized recently available self-contained, portable EMG data collection units (Mega Electronics Inc; ME3000P) that can collect 4 channels of EMG for up to 2h without downloading. The EMG data were collected bilaterally from the erector spinae muscles at the T9 and L3 levels. The EMG was sampled at 1000 Hz, averaged, and stored every 100 ms. The normalization contraction was a standardized, slightly forward exed trunk posture with a weight held in the hands. The load and posture resulted in a spinal compression of about 3000N, close to the action limit (AL) as de ned by NIOSH (1981). The resulting spinal compression was estimated by submitting the joint coordinates, obtained from a video analysis, the workers body weight and the magnitude and direction of the applied forces, to the biomechanical model described earlier (Neumann et al. 1995). The electromyograms recorded during work were downloaded from the datalogger and normalized by the contraction described above. The time history of the normalized EMG was described as an Amplitude Probability Distribution Function (APDF) and reported as `compression normalized EMG in Newtons. The 10th, 50th, 90th and 99th percentiles were chosen to represent static, median and two estimates of peak loading (cf. Jonsson 1982). The basic scheme for generating an exposure measure in a common metric of spinal compression from ve diœerent measurement methods is summarized in table Implementation of the common metric in a eld study of risk factors for low-back pain in automotive assembly workers The automotive assembly environment has many people engaged in well-de ned tasks with cycle times typically of the order 1 to 5 min. Another large group is relief workers who perform dozens of tasks per day or work on repair and rework tasks. Approximately one-third of the employees are engaged in maintenance or other nonrepetitive tasks. To illustrate the implementation of the previously described approach, its use in assessing the exposure of three diœerent jobs in an automotive assembly plant is presented. An example of three types of jobs is presented in table 2.

6 56 R. Wells et al. Table 1. Obtaining a common metric of Newtons of low-back compression from diœerent methods. Spinal compression Measurement method Risk factors estimated from Self-report questionnaire Any risk factor, present Reported posture, load (includes diary) and past and load direction.² Usual and heaviest instant by task Expert observer checklist Current and observable Observed posture, load risk factors and load direction.² Usual and heaviest instant by task Posture and load sampling Observed load and Sampled load, posture posture and load direction.² Continuous sampled record Video with post hoc motion Video record of posture at Digitized posture, load and analysis a number of suspected peak load direction.² Usual and load conditions heaviest instant by task EMG via belt-mounted Muscle activation from Calibrated and processed data logger lumbar and thoracic EMG. musculature and resulting Continuous record spinal load ² Posture and load information input to a biomechanical model, see Andrews et al See Wells et al. (1994) for more details. Table 2. Description of the three jobs used to illustrate the exposure assessment methods. Job type Short-cycle job Long-cycle job Non-cyclic job Job title Assembly operator Paint repair operator Materials handler Cycle time About 1 min About 1 hour Unde ned but repeats most elements daily Typical tasks (1) Runs screws (1) Remove plastic (1) Drive lift truck (2) Pull wires body parts (2) Lift boxes (3) Wait (2) Prepare (3) Open boxes (3) Sand (4) Drag empty (4) Clean pallets (5) Tape (5) Wait (6) Wait Table 3 summarizes the potential methods available for exposure assessment of each job. For more straightforward short cycle jobs (typically of a few seconds to a few minutes duration) there is a choice of a large number of instruments, from questionnaire to biomechanical models, to estimate both peak and cumulative demands. For longer cycle times, of the order of an hour, there still exist a number of instruments to determine peak loads as it is possible to follow workers for periods of up to a day using checklists and video as well as to record EMG on self-contained data loggers for this period. For very irregular jobs it is possible to assess peak

7 Common measurement metrics for exposure assessment 57 Table 3. Suggested use of various instruments to evaluate the three diœerent jobs. Example job Possible methods How used Short-cycle job Peak: Video analysis Little extrapolation from observed data needed Cumulative: EMG or posture sampling Long-cycle job Peak: Self-report questionnaire and Highest peak of those reported posture sampling taken as peak demand Cumulative: EMG or posture Some time-weighted averaging sampling over tasks usually needed Non-cyclic job Peak: `Hardest tasks from diary: Highest peak of those reported estimated posture and load or taken as peak demand simulated for video analysis Cumulative: Diary used to obtain Time-weighting of peak and other time weighting and tasks tasks demands using video or checklists or posture sampling; however, cumulative demands are more problematic. The use of a diary or logbook becomes critical to determine the tasks and the approximate proportion of time spent in each of them. Figure 1 shows how assessments can be structured. For the short cycle job (Job 1) any combination of the assessment methods can be used to obtain a relatively complete picture of the current exposure, often without the necessity of breaking the job down into tasks. For longer cycle jobs (Job 2), a task approach (breaking the job into tasks) becomes necessary; the exposure associated with each task must be measured. The exposure for the job can then be obtained by appropriate time weighting of the task exposures. Task breakdowns may be avoided if representative cycles can be observed and measured directly. For the irregular cycle job (Job 3), a task-based approach is imperative. The challenge then becomes to identify all relevant tasks and obtain a sample of them. If peak load is of interest then the performance or simulation of irregular yet stressful tasks is critical. If cumulative load or static load is of interest then seemingly low demand tasks, such as monitoring via visual display units, are important to capture. Tables 4 and 5 show the bene ts of the approach described; despite using a diœerent measurement method for each job, it is still possible, in principle, to produce an exposure measure (in a common metric and units of Newtons) for comparison of peak and static load demands. Only one example estimate is provided in this paper although each method provides its own estimate of exposure. It is likely that each method has diœerent accuracy and precision, and the estimates of spinal load will diœer on the three example jobs. To avoid misleading impressions of the agreement between the diœerent methods only a single value for each job will be presented at this time. Table 4 presents a peak load from digitized video for the short-cycle job (Job 1), the peak load from posture and load sampling for Job 2 and the peak load from a selfreport questionnaire for the irregular tasks (Job 3). Jobs 1 and 2 show similar peak demands while the materials handling job (Job 3) has considerably higher peak demands. Table 5 illustrates two methods of obtaining static (continuous) loads for the three jobs. For static load (de ned as the 10th percentile of the amplitude probability

8 58 R. Wells et al. Figure 1. Schematic of the integration of four assessment methods for three jobs with widely varying temporal characteristics. Table 4. Exampleof exposuremeasures in the common metric (N) obtained for three diœerent jobs using threediœerent measurement approaches; peak load. The jobs are described in table 2. Exposure measurement method Posture and Biomechanical Exposure in common Self-report load sampling model from video metric, N peak Job 1 ± ± Job 2 ± 2031 ± 2031 Job ± ± 4797

9 Common measurement metrics for exposure assessment 59 Table 5. Example of exposuremeasures in the common metric of spinal load (N) obtained for three diœerent jobs using two diœerent measurement approaches; static (or sustained) load (10th percentile of the amplitude probability distribution function, APDF). Note that the spinal load includes a contribution from the upper body weight (indicated in brackets). The jobs are described in table 2. Exposure measurement method Compression normalized Posture and load Exposure in common metric (N) EMG (N) sampling (N) Static load (contribution of Static Load Static Load upper body weight) Job ± 367 (328) Job ± 878 (355) Job 3 ± (421) distribution function, APDF), processed trunk electromyography and posture and load sampling are possible methods. The long cycle job (Job 2) demonstrates the highest static loads, probably as a result of the long periods of time spent in a exed trunk posture. The short cycle job (Job 1), demonstrates an almost upright trunk position, and as a result the static load is low; just above upper body weight. The common metric approach presented should be contrasted to the more usual measures of exposure for the low back. Even for a given type of method, such as a questionnaire, the output metric is often diœerent between investigators and even between job types studied (e.g. o ce versus construction workers). For example, in diœerent questionnaires, repetitiveness may be quanti ed by the number of lifts or the number of trunk exions; these are quite diœerent estimates of exposure. Similarly, across diœerent occupational settings, risk factors for low-back pain in o ce environments may be recorded as the percentage of time spent in a chair without contact with a lumbar support whereas in a construction setting, the risk factor is more likely to be the percent of time in trunk exion. This makes comparisons between studies and between diœerent job types di cult. It must be acknowledged that while the exposure estimates presented in this paper may share a common metric, the accuracy and precision oœered by each is probably very diœerent. The direct (instrumented) methods will probably give more precise estimates of current exposure, and therefore better estimates of risk, Burdorf (1995). To assess this, a comparison of the estimates produced by each method (intermethod comparison, Burdorf 1995) is underway as part of a large epidemiologic study in automobile workers (Wells et al. 1993). The objective of this paper was to demonstrate that a common exposure metric is feasible from several disparate measurement methods. Acknowledgements This work was supported by the Institute for Work and Health, Toronto, Ontario. References ANDREWS, D. NORMAN, R. W., WELLS, R. P. and NEUMANN, P. 1996, Accuracy and repeatability of low back spine compression force estimates from self reports of body posture during load handling, International Journal of Industrial Ergonomics, in press. ARMSTRONG, T. J., BUCKLE, P., FINE, L. J., HAGBERG, M., JONSSON, B., KILBOM, A., KUORINKA, I. A. A., SILVERSTEIN, B. A., SJOGAARD, G. and VIKARI-JUNTURA, E. R. A. 1993, A conceptual model for work-related neck and upper-limb musculoskeletal disorders, Scandinavian Journal of Work and Environmental Health, 19(2), 73±84.

10 60 R. Wells et al. BONGERS, P. 1993, Psychosocial factors at work and musculoskeletal disease, Scandinavian Journal of Work and Environmental Health, 19, 297±312. BUCKLE, P. W., STUBBS, D. A. and BATY, D. 1986, Musculo-skeletal disorders (and discomfort) and associated work factors, in N. Corlett, J. Wilson, and I. Manenica, (eds), The Ergonomics of Working Postures: Models, Methods and Cases: Proceedings of the First InternationalOccupational Ergonomics Symposium, Z adar, Yugoslavia, 18± 20 April 1985 Taylor & Francis, London, 19±30. BURDORF, A. 1995, Reducingrandom measurement error in assessing postural load on the back in epidemiologic studies, Scandinavian Journal of Work and Environmental Health, 12, 15±23. GARG, A. 1989, Epidemiological basis for manual lifting guidelines. NIOSH report , NIOSH, Cincinnati. HAGBERG, M. 1992, Exposure variables in ergonomic epidemiology, American Journal of Industrial Medicine, 21, 91±10. HAGBERG, M., SILVERSTEIN, B., WELLS, R., SMITH, R., CARAYON, P., HENDRICK, H. P., PERUSSE, M., KUORINKA, I. and FORCIER, L. (eds), 1995, Work-related Musculoskeletal Disorders ( WMSD): A Handbook for Prevention (Taylor & Francis, London). JONSSON, B. 1982, Measurement and evaluation of local muscular strain in the shoulder during constrained work, Journal of Human Ergology, 11, 73±88. KARHU, O., KANSI, P. and KUORINKA, I. 1977, Correcting working postures in industry: a practical method for analysis, Applied Ergonomics, 8(4), 199±201. KERR, M. S., SHANNON, H. S., FRANK, J. W., NORMAN, R. W. K. and WELLS, R. P. 1993, Design of a case-control study for reported occupationallow back pain, in proceedings Work & Health: Scienti c Basis of Progress in the Working Environment, Abstract no. 2/11, Copenhagen. KEYSERLING, W. M. 1986, A computer-aided system to evaluate postural stress in the workplace, American Industrial Hygiene Association Journal, 47(10), 641±649. KUMAR, S. 1990, Cumulative load as a risk factor for low-back pain, Spine, 15, 1311±1316. KUMAR, S. 1994, A conceptual model of overexertion, safety, and risk of injury in occupational settings, Human Factors, 36, 197±209. LEAMAN, T. B. 1994, Research to reality: a critical review of the validity of various criteria for the prevention of occupationallyinduced low back pain disability,ergonomics, 37, 1959± MCATTAMNEY, 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. MCGILL, S. M. and NORMAN, R. W. 1986, Partitioningof the L4/L5 dynamicmoment into disc, ligamentous and muscular components during lifting, Spine, 11, 666±678. MARRAS, W. S., LAVENDER, S. A., LEURGANS, S. E., RAJULLI, S. L., ALLREAD, W. G., FATHALLAH, F. A. and FERGUSON, S. A. 1993, The role of dynamic three-dimensional trunk motion in occupational-relatedlowback disorders: the eœects of workplacefactors trunk position and trunk motion characteristics on risk of injury, Spine, 18, 617±628. MIENTJES, M. 1996, An EMG based continuous low back load estimation technique for three dimensional workplace related jobs and tasks, Masters Thesis, Department of Kinesiology, University of Waterloo, Ontario, Canada. NEUMANN, P., NORMAN, R. W. and WELLS, R. P. 1995, EMG-based estimates of peak and prolonged exposure of the low back to forces and moments during occupational tasks, Proceedings of the International Society of Biomechanics (Finland, IyvaÈ skylaè ), pp. 670±671. NIOSH 1981, Work practices guide for manual lifting, Technical Report No. 81±122. US Department of Health and Human Services, Cincinnati, Ohio. PLAGENHOEF, S. 1971, Patterns of Human Movement (Prentice Hall, New Jersey). POTVIN, J. and NORMAN, R. 1993, Quanti cation of erector spinae muscle fatigue during prolonged dynamic lifting, European Journal of Applied Physiology, 67, 554±562. POTVIN J., MCGILL, S. M. and NORMAN, R. 1991, Trunk muscle and lumbar ligament contributionsto dynamic lifts with varying degrees of trunk exion, Spine, 16, 1099±1107. POTVIN, J., NORMAN, R. and WELLS, R. 1990, A eld method of continuous estimation of compressive load on the L4/L5 disk during performance of repetitive industrial tasks, Proceedings of the 23rd Annual Conference of the Human Factors Association of Canada, Ottawa.

11 Common measurement metrics for exposure assessment 61 STOCK, S. R. 1991, Workplace ergonomic factors and the development of musculoskeletal disorders of the neck and upper limbs: a meta-analysis, American Journal of Industrial Medicine, 19, 87±107. WELLS, R. P., NORMAN, R. and NEUMANN, P. 1994, Long-term monitoringof low back physical exposures, in Proceedings of the 12th Triennial Congress of the International Ergonomics Association, Vol. 2, Toronto, August, 150±152. WELLS, R., MOORE, A. E., NORMAN, R. W., NEUMANN, P. and ANDREWS, D. 1995, Development of SAM: a pen based computer system for work sampling, Proceedings of the 2nd Congress on Prevention of Musculoskeletal Disorders ( PREMUS), 317±319. WELLS, R., NORMAN, R., NEUMANN, P., FRANK J., SHANNON, H. and KERR, M. 1993, Exposure assessment for epidemiologicstudy of reported low back pain, in Proceedings of the 25th Annual Conference of the Human Factors Association of Canada, August. WIKTORIN, C., KARLQUIST, L. and WINKEL, J. 1993, Validity of self-reported exposures to work postures and manual materials handling, Scandinavian Journal of Work and Environmental Health, 19, 208±214. WIKTORIN, C., HJELM, E., WINKEL, J. and KOSTER, M. 1996, Reproducibilityof a questionnaire for assessment of physical load during work and leisure time, Journal of Occupational and Environmental Medicine, 38, 190±201. WINKEL, J. and MATHIASSEN, S.-E. 1994, Assessment of physical work load in epidemiologic studies: concepts, issues and operational considerations, Ergonomics, 37, 979±989. WINKEL, J. and WESTGAARD, R. 1992, Occupational and individual risk factors for shoulderneck complaints, Part II: the scienti c basis(literature review) for the guide. International Journal of Industrial Ergonomics, 10, 85±104. ZATSIORSKY, V. and SELUYANOV, V. 1983, The mass and inertia characteristics of main segments of human body, In H. Matsui and Kobayashi, K. (eds), Biomechanics VIII-B, (Human Kinetics Publishers, Champaign, IL), 1152±1159.