A Review of Research on Interventions to Control Musculoskeletal Disorders

Michael J. Smith, Ben-Tzion Karsh, Francisco B. P. Moro

Dept. of Industrial Engineering, University of Wisconsin-Madison

A. Introduction

The purpose of our paper is to address the question posed by the National Academy of Sciences (NAS)—'What is the state of available scientific evidence on interventions to control musculoskeletal disorders?' Toward this end we will also be answering the following four questions: (1) What kinds of interventions have been assessed for their effectiveness in controlling the incidence and/or severity of musculoskeletal disorders of the back and/or upper extremities? (2) What do the overall results from these studies reveal about the effectiveness of these interventions? (3) How trustworthy is the research basis for drawing conclusions on intervention effectiveness? (4) Do studies show the relative contributions of biomechanical and other factors to intervention effectiveness? We have intentionally limited the analysis in this paper to peer reviewed journal articles. Although we also reviewed proceedings documents and book chapters, in our opinion, the latter do not meet the criteria for ''scientific evidence" as well. Therefore these were excluded from the analysis. We did not review trade journal articles or articles in the popular press, nor did we review the NIOSH or OSHA documents on successful ergonomic interventions.

This paper begins with a description of the model that served as our conceptual framework. We then describe the methodology used to select papers for this review. The research evidence relating to the efficacy of laboratory interventions, field interventions with healthy subjects, and field interventions with injured subjects is then presented, followed by concluding remarks about the state of the scientific knowledge on interventions to control musculoskeletal disorders.

We propose a model to examine interventions to control musculoskeletal disorders based on the balance theory of Smith & Carayon-Sainfort (1989, 1995). This model states that working conditions (and other environmental features outside of work) can produce a "stress load" on the person. That load can have biomechanical, physiological and psychological consequences such as forces on the joints, increased blood pressure and/or perceptions of pain. The load can produce a negative influence on the person which leads to "strain" if it exceeds the person's capacity. This has been called a "misfit" between the environmental demands and the personal resources. If exposure continues for a prolonged time period, then this strain can produce serious musculoskeletal disorders.

Figure 1 illustrates a system's model for conceptualizing the various elements of a work system, that is, the loads that working conditions can exert on workers. In this model these various elements interact to determine the way in which work is done and the effectiveness of the work in achieving individual and organizational needs and goals. At the center of this model is the individual with his/her physical characteristics, perceptions, personality and behaviors. The



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 200
--> A Review of Research on Interventions to Control Musculoskeletal Disorders Michael J. Smith, Ben-Tzion Karsh, Francisco B. P. Moro Dept. of Industrial Engineering, University of Wisconsin-Madison A. Introduction The purpose of our paper is to address the question posed by the National Academy of Sciences (NAS)—'What is the state of available scientific evidence on interventions to control musculoskeletal disorders?' Toward this end we will also be answering the following four questions: (1) What kinds of interventions have been assessed for their effectiveness in controlling the incidence and/or severity of musculoskeletal disorders of the back and/or upper extremities? (2) What do the overall results from these studies reveal about the effectiveness of these interventions? (3) How trustworthy is the research basis for drawing conclusions on intervention effectiveness? (4) Do studies show the relative contributions of biomechanical and other factors to intervention effectiveness? We have intentionally limited the analysis in this paper to peer reviewed journal articles. Although we also reviewed proceedings documents and book chapters, in our opinion, the latter do not meet the criteria for ''scientific evidence" as well. Therefore these were excluded from the analysis. We did not review trade journal articles or articles in the popular press, nor did we review the NIOSH or OSHA documents on successful ergonomic interventions. This paper begins with a description of the model that served as our conceptual framework. We then describe the methodology used to select papers for this review. The research evidence relating to the efficacy of laboratory interventions, field interventions with healthy subjects, and field interventions with injured subjects is then presented, followed by concluding remarks about the state of the scientific knowledge on interventions to control musculoskeletal disorders. We propose a model to examine interventions to control musculoskeletal disorders based on the balance theory of Smith & Carayon-Sainfort (1989, 1995). This model states that working conditions (and other environmental features outside of work) can produce a "stress load" on the person. That load can have biomechanical, physiological and psychological consequences such as forces on the joints, increased blood pressure and/or perceptions of pain. The load can produce a negative influence on the person which leads to "strain" if it exceeds the person's capacity. This has been called a "misfit" between the environmental demands and the personal resources. If exposure continues for a prolonged time period, then this strain can produce serious musculoskeletal disorders. Figure 1 illustrates a system's model for conceptualizing the various elements of a work system, that is, the loads that working conditions can exert on workers. In this model these various elements interact to determine the way in which work is done and the effectiveness of the work in achieving individual and organizational needs and goals. At the center of this model is the individual with his/her physical characteristics, perceptions, personality and behaviors. The

OCR for page 200
--> Figure 1:  Balance Model of Work System Misfit individual has technologies available to perform specific job tasks. The capabilities of the technologies affect performance and also the worker's skills and knowledge needed for its effective use. The task requirements also affect the skills and knowledge needed. Both the tasks and technologies affect the content of the job and the physical demands the job makes on the person. The tasks with their technologies are carried out in a work setting that comprises the physical and the social environment. There is also an organizational structure that defines the nature and level of individual involvement, interaction and control. The purpose of interventions to control musculoskeletal disorders is to reduce the stress load to eliminate strain. As discussed below, this can be done by modifying the elements of the work system shown in Figure 1. Another tactic to control musculoskeletal disorders is to increase the capacity of the individual to handle greater loads, thereby reducing the possibility of a misfit. B. The Nature of Interventions to Control Musculoskeletal Disorders There are a variety of actions that have been applied in the workplace for eliminating or reducing the occurrence of occupational musculoskeletal disorders. These include engineering redesigns, changes in work methods, administrative controls, training, organized exercise, work hardening, personal protective equipment, and medical management to reduce exposures. Some of these have been evaluated in research studies using both laboratory and field settings. The purpose of this paper is to characterize the nature of these research studies, evaluate their methodological soundness, and determine conclusions that can be made based on their strengths and weaknesses. We will first comment on the types of actions. Engineering redesign aims to control exposures to the biomechanical risk factors for musculoskeletal injury. Engineering redesign has three main directions: (1) redesign of machinery, (2) providing assistive devices, and (3) tool redesign. Redesign of machinery deals with reducing exposure to biomechanical risk factors through modification of the machinery or the workstation. An example would be the use of adjustable tables to improve the postures of body parts (neck, shoulders, arms, hands, wrists, and back). Another would be the realignment of controls for better access that promotes less forceful activation with better body part postures.

OCR for page 200
--> Assistive devices provide mechanical advantages when dealing with loads. An example would be a lifting device such as a lift table, a hoist or a patient lifter. Tool redesign could be thought of as a sub-set of machinery design, but we have separated them due to the extensive efforts in tool redesign. Examples would be a reduction in the weight of powered hand tools, improved grip designs, alternative keyboard designs, and alternative mouse designs. The primary risk factors addressed with engineering redesign are loads (forces, weights) and body part postures. Work methods improvement is also aimed primarily at biomechanical risk factor control, but can also influence the psychosocial work environment. The improvement requires changes in employee behavior to achieve risk reduction. This approach is often accompanied by employee training to provide a basis for the behavior change. The main direction of work methods improvement is to modify the task design to reduce or eliminate risk factors. An example would be changing the techniques used in cutting meat to reduce the frequency of cutting motions, and to improve body part postures while reducing loads. Administrative controls are aimed at reducing the time of exposure to biomechanical and psychosocial risk factors. The two main directions are rotating employees among jobs with differing exposures and the use of rest breaks. Improved medical management activities could also be considered as administrative controls although the OSHA Ergonomic Guidelines for the Red Meat Industry considers medical management as a separate category. Training is aimed at informing employees about the risk factors of musculoskeletal injury, and/or changing behavior to reduce risk. An example would be an ergonomic education program that provides employee orientation to risk factors. Another would be providing on-the-job instruction in revised task methods. A third would be providing instruction on how to use specific capabilities of workstation adjustments such as how to properly adjust a chair or a work table. Exercise and work hardening programs increase the capacity of the employee. That could mean increasing strength, or flexibility, or tolerance for pain, or skills to conduct tasks. Personal protective equipment typically blocks employee contact with a hazard. For musculoskeletal injury, an example would be gloves to dissipate the energy from hand tool vibration. However, for musculoskeletal injury there is another type of personal protective device that serves as a "support" for the musculature to reduce/balance forces: the back belt (or similar devices). C. Methods Used to Review the Literature The methodology used to evaluate the state of intervention research for the control of musculoskeletal disorders proceeded in five phases as outlined in Table 1. Phase 1 consisted of a comprehensive search to find any article related to musculoskeletal disorder interventions. To accomplish this, the on-line databases PsychLit (1974—present), Engineering Index (1987—present), and Medline (1966—present) were searched using 21 different search terms related to musculoskeletal disorder interventions, put into various combinations. At the same time, we looked through 14 different publications which included NIOSH publications, text and reference books in Ergonomics, and National Safety Council publications. The combined efforts yielded 720 articles.

OCR for page 200
--> In phase 2, abstracts from all 720 articles were read. If the article was a review of many other research articles, then the entire article was read and the bibliographies were examined for additional relevant articles. This brought the total number of articles to 768. In phase 3 all empirical studies that employed what could be liberally considered an intervention for controlling musculoskeletal disorders were obtained from the library. There were 198 such articles. In phase 4, 186 of the 198 articles were read and categorized. Twelve of the articles were unavailable through our library or interlibrary loan services within the review time scope. The categories consisted of laboratory vs. field studies, which were further broken down by intervention type: engineering, administrative, work method, training/exercise, and personal protective equipment. Phase 5 consisted of selecting articles from the pool of 186 that met the following criteria: Peer reviewed journal article Directly related to musculoskeletal disorder interventions Representative of the research Methodologically sound (relative to the other articles) For an article to have been considered methodologically sound, the following characteristics were considered: Control condition Accounting for confounds Relevancy of measures Randomized trials Blinded evaluators Based on the steps in this phase, we selected 43 articles for in-depth review and analysis. Not all articles selected met all the conditions for methodological soundness. D. General Discussion of the Strengths and Limitations of Research on the Effects of Interventions on Musculoskeletal Disorders Hersey, Collins, Gershon, and Owen (1996) described the main challenges to all intervention research: use a theoretical basis, have sensitive measures, use "sound" research design, have appropriate statistical power, and apply interventions that can provide "interpretable" results. Similar issues have also been discussed by Lipsey (1998). Whereas these challenges are present in all types of intervention research, they are tailored here to interventions for the control of musculoskeletal disorders. The first challenge is for intervention research to be based on theory. Thus there is the need to have research questions, hypotheses, and/or a conceptual idea of the issue under study. In some disciplines the observation of phenomenon is the basis for generating concepts, and naturally occurring experiments (passive interventions) are examined. In others there is the need for manipulation of the variables (active interventions).

OCR for page 200
--> The second challenge is to use sensitive measures. Ergonomic intervention studies can fall short of this by using measures somewhat removed from musculoskeletal injury or by using measures that are not sensitive to the intervention. An example of the former would be a laboratory study that measured posture changes as an outcome. Such posture changes may influence the risk of injury, but may represent a low probability of injury. An example of a measure that may not be sensitive to a given intervention would be using the total number of workplace injuries in a facility as the measured outcome when the intervention was only geared toward preventing back injuries. Another way ergonomic intervention studies try to accomplish the goal of sensitive measures is by using multiple related measures. While this provides redundant and even contrasting information about an issue, there is no guarantee that multiple measures will provide any better characterization or precision. The next challenge is to use a sound research design. In traditional experimental design this includes having sufficient observations, random assignment to conditions, using control groups, controlling for confounds, and having multiple observation points for each participant. In "naturally occurring interventions" many of these conditions are not met. For instance, random assignment is unlikely, as are control groups which received no treatment. An important issue is statistical power. Many intervention studies seem to have low power, either because of poor contrast between the intervention and control groups (i.e. the control group received a treatment not unlike the intervention group), contamination effects (i.e. the intervention and control subjects worked together making it easy for the controls to be exposed to the intervention), insensitive dependent variables, or small (often inadequate) sample sizes. These problems are most frequent in field intervention studies, but are also observed in the laboratory studies. For instance, the laboratory studies may suffer from small sample size and few repeated trials. The fifth challenge is to use an intervention that can yield interpretable results. Studies fail this challenge when, for example, the intervention used is composed of multiple inter-related components such as training, exercise, organizational changes, and ergonomic improvements. In such cases, the contribution of any one component of the intervention cannot be assessed. The strengths and limitations of laboratory and field methods can be best discussed in terms of threats to internal and external validity. (The scope of this paper prevents a full discussion of threats to statistical conclusion and construct validity, but see Cook and Campbell (1979) for a detailed discussion of these issues.) Internal validity concerns the ability to make causal statements, whereas external validity is the extent to which a study is generalizable. Laboratory studies are characterized by random assignment to conditions, a high degree of control over the study environment and independent variable manipulations, and very often control groups. These characteristics give laboratory studies a high degree of internal validity because in controlled settings, there is little that can influence the dependent variable except for the independent variable. On the other hand, laboratory studies are often criticized because of low external validity. Dipboye (1990) listed several reasons for this criticism. First, there is an unrealistic nature to laboratory studies. This stems from the controlled artificial environment. Second, there is a lack of representative sampling (i.e. the typical use of college students, trained athletes). Field studies, as typically carried out, present a tradeoff in terms of threats to validity. Since field studies are carried out in real world settings, they are considered to have high external validity. This same characteristic is the reason that they also often have low internal validity. In

OCR for page 200
--> real world settings, there are many things that are "naturally occurring" which produce changes at the same time that interventions are instituted. This produces confounds and/or unexplained variance that cannot be controlled, and may dilute, or incidentally enhance, the intervention effects. In addition, field interventions do not always use random assignment to conditions or have control groups. When "control" groups are present, they are more often than not "quasi control groups". They are often a convenient group of employees who are not getting the intervention, but for whom measures are available to contrast to the intervention group. Another tactic in field ergonomic interventions is the use of "comparison'' groups which are employees who receive an alternative treatment who are then compared to the main treatment. Any of these problems can create confounding, which limits causal interpretations. Given the tradeoffs between laboratory and field studies, it is clear that both are necessary to gain a complete picture of the effectiveness of interventions to control musculoskeletal disorders. Considering the inherent limitations discussed above, the results of any single study in isolation must be interpreted carefully. Rather, a broader view examining the entire literature as a whole provides some insight into the potential effectiveness of interventions to control musculoskeletal disorders. It must be recognized that many of the limitations and problems described above for intervention research to control musculoskeletal disorders is true for almost all types of the intervention research studies, and are not unique to musculoskeletal intervention research. E. Review of Select Research We selected 43 research papers to represent the intervention literature on controlling musculoskeletal disorders. The majority of studies dealt with the risks for back injuries, but there were some which dealt with upper extremity musculoskeletal disorders. These studies were put into three categories: (1) laboratory experimental studies, (2) field studies using previously injured employees as subjects, and (3) field studies using healthy employees. Each category will be discussed separately, and then the entire group will be assessed in total. 1. Conclusions from Examples of Laboratory Intervention Research: The laboratory ergonomic intervention research can be classified into studies which examined improved procedures for carrying out tasks (lifting technique), improved equipment designs (keyboards, hand tools) and the use of personal protective equipment (back belts, gloves). We have selected fifteen studies we believe characterize the laboratory ergonomic intervention evaluations. Table 2 provides highlights of these studies. 1.1. Methodological Strengths and Weaknesses of the Laboratory Intervention Research Studies: Strengths: (1)   There was substantial control over the exposures so that the consistency, level and frequency of exposures were constant across subjects in all of the laboratory studies. For example, Lin, Radwin and Snook (1997), using a special device, were able to ensure that each subject in their

OCR for page 200
--> study grasped the handle the same way, and had identical forearm-upper arm angles. Similarly, Schoenmarklin and Marras (1989) used a device to ensure that all subjects began with their right arm at a 45-degree angle to the hammering fixture. They also used a computer to pace each subject at 57 strikes/minute. (2)   The majority of the studies (12/15) had some objective response measurement, and the precision of the objective response measurements was high. The three studies that did not use "objective" measures were Lavender and Kenyeri (1995), who measured acceptable lifting weights, Smith, Karsh, Conway, Cohen, James, Morgan, Sanders and Zehel (1998), who measured perceived discomfort and did not use goniometers for the posture measures, and Swanson, Galinsky, Cole, Pan, and Sauter (1997), who measured perceived discomfort. An example of a study that used "objective" measurements is Lander, Hundley and Simonton (1992), who used a force platform, pressure transducer, and EMGs to measure some of their outcomes. (3)   The majority of the studies indicated that they randomized or counterbalanced the order of conditions. Nine of the studies randomized the order of conditions, while three counterbalanced the order presentation. Weaknesses: (1)   The exposures were focused on a small aspect of a larger process, and their application to the "bigger picture" is debatable. For instance, Schoenmarklin and Marras (1989) examined arm angle and force when pounding nails with a hammer where the subjects had a required trunk and shoulder posture. (2)   The exposures were not representative of the "real world". They may not "generalize" or even extrapolate to the "real world". For example, Lander, Simonton and Giacobbe (1990) and Woodhouse, Heinen, Shall and Bragg (1990) both used trained athletes in their lifting technique experiments. (3)   The range and time of exposures in the laboratory studies were very limited when compared to the field studies, and most were "constrained" by the apparatus or procedures such that the subjects' responses were constrained or limited to a small range. For example, Schoenmarklin and Marras (1989) had their subjects pound nails for only three minutes per condition. Resnick and Chaffin (1997) only used 30-second trials. Lander, Simonton and Giacobbe (1990) had subjects make only six lifts. (4)   For many, the sample size was small (8/15 had ten or fewer subjects), as was the number of repeated trials (7/15 only used one repetition per condition). This may have limited the ability to detect differences in conditions, and could explain the mixed results. (5)   Except for a few studies (4 out of 15) the participants were not workers (and may not be representative of workers). In one of these studies (Oh and Radwin, 1993), where workers were

OCR for page 200
--> participants along with students, there were differences in the physical capacity of the workers as well as their outcomes. (6)   The outcomes examined were intermediate states or surrogates, not measures of disorder symptoms, diagnostic criteria or endstates. Some endpoints (posture, force) dealt with "risk factors", while others dealt with short-term pain, discomfort and fatigue. None dealt with diagnostic tests or disorders. The connection to the reduction of "musculoskeletal disorders" is at a distance. However, all outcomes were "theoretically" consistent with a relationship to musculoskeletal disorders. For example, Woodhouse, Heinen, Shall and Bragg (1990) measured lifting force and muscle work and power, while Swanson, Galinsky, Cole, Pan and Sauter (1997) and Smith, Karsh, Conway, Cohen, James, Morgan, Sanders and Zehel (1998) measured short term discomfort. (7)   For eleven out of fifteen (11/15) studies, the findings were mixed and thus not clear enough to serve as the basis for a recommendation about the effectiveness of an intervention. As an example, Nakaseko, Grandjean, Hunting and Gierer (1985) found that there was less ulnar wrist deviation when using an experimental keyboard (vs. a standard keyboard), but there were no differences between the keyboards in reports of pain/discomfort. (8) Thirteen out of 15 laboratory studies used within subject designs and Anova or t-test techniques for their analyses, yet none of them mentioned that they corrected or even tested for violations of sphericity. B. General Conclusions about the Findings from the Laboratory Intervention Research Studies: (1)   The results of studies on "proper" lifting posture and technique are unclear. It is not possible to define the "best" lifting postures and techniques. All three of the laboratory lifting technique studies found mixed results. Leskinen, Stalhammer, Kuorinka and Troup (1983) found, when comparing four different lifting styles, that the squat lift resulted in the highest peak forces in the feet, but the lowest peak L5/S 1 compression forces. Hart, Stobbe and Jareidi (1987) compared three lifting postures and found that the lowest trunk flexion moments occurred in the lordotic posture. They also found that the greatest abdominal muscle activity occurred with the kyphotic lumbar posture, while the least amount was found in the straight back posture. (2)   The evidence on the effectiveness of back belt use for reducing back injury risk is inconclusive. There were no differences found between wearing and not wearing a belt for maximum acceptable weight limits, joint angles, peak lifting force, total muscle work, or average muscle power (Lavender and Kenyeri, 1995; Marley and Duggasani, 1996; Woodhouse, Heinen, Shall and Bragg, 1990). Experienced athletes doing limited weight lifting activities showed some benefits (for example increased intra-abdominal pressure—IAP), but other results were not consistent (e.g. inconsistent differences between belt and no-belt conditions for external oblique and erector spinae mean EMGs) (Lander, Simonton and Giacobbe, 1990; Lander, Hundley and Simonton, 1992).

OCR for page 200
--> (3)   The evidence is mixed whether alternative keyboards designed to improve hand/wrist postures can provide benefits of reduced risk factors for upper extremity musculoskeletal disorders. Nakaseko, Grandjean, Hunting and Gierer (1985), Smith, Karsh, Conway, Cohen, James, Morgan, Sanders and Zehel, (1998), and Swanson, Galinsky, Cole, Pan and Sauter, (1997) found no differences in reported pain between alternative and standard keyboards. Smith, Karsh, Conway, Cohen, James, Morgan, Sanders and Zehel (1998) found that there was less pronation when using a split keyboard compared to when using a traditional keyboard. (4)   There is evidence that alternative hand tools designed to improve hand/wrist postures and/or to reduce forces on the palm/fingers can provide benefits for the reduction of risk factors for upper extremity musculoskeletal disorders. Oh and Radwin (1993) found benefits for an extended trigger on a pneumatic nutrunner and Schoenmarklin and Marras (1989) found some postural benefits for angled hammers. (5)   There is some evidence that the use of weight handling devices such as hoists can reduce the risk factors for upper extremity musculoskeletal disorders. Resnick and Chaffin (1997) found that using an articulated arm, resulted in less peak push forces compared to a hoist with an overhead rail or hoist with a fixed pivot. 2. Conclusions from Examples of Field Intervention Studies: The field intervention studies have been classified into those studies that used injured subjects or subjects suffering from pain, and those using healthy subjects. The studies using injured subjects can be further broken down into exercise, back school, early intervention, and physical therapy interventions. The studies that used healthy subjects can be further broken down into ergonomic improvement, training, back education, exercise, and weight belt interventions. 2.1. Examples of Intervention Studies of Injured Employees: Fifteen injured-employee intervention studies were selected as representative of the literature that used injured employees as subjects. Table 3 provides highlights of these studies. 2.1.1. Methodological Strengths and Weaknesses of the Field Studies using Injured Subjects: Strengths: (1)   Thirteen out of 15 studies used random assignment to conditions. (2)   Eleven of the 15 studies used a control or comparison group. Six studies had control groups (i.e. groups that did not receive any treatment), and five studies had comparison groups which received the standard treatment for the disorder of interest. The four other studies had a pre-post treatment design that compared different types of interventions.

OCR for page 200
--> (3)   Seven out of fifteen studies compared interventions. For example, Moffett, Chase, Portek and Ennis (1986) compared a back school intervention to an exercise intervention, Bergquist-Ullman and Larsson (1977) compared a back school intervention, physical therapy intervention, and a placebo intervention (heat treatment), and Bru, Mykletun, Berge and Svebak (1994) compared a cognitive intervention, a relaxation intervention, and a combined cognitive-relaxation intervention. (4)   All of the studies had multiple outcome measures. As examples, Lindstrom, Ohlund, Eek, Wallin, Peterson, Fordyce and Nachemson (1992) measured pain, return to work rates, sick leave days, and recurrence of pain; Kellet, Kellet and Nordholm (1991) measured flexion, strength, sick leave days, and presence of symptoms; and Harkappa, Mellin, Jarvikoski and Hurri (1990) measured pain, disability, compliance with treatment, and sickness days. (5)   As compared to the laboratory studies, some of the exposures were for a long time period (several months). Donchin, Woolf, Kaplan and Floman (1990), for example, had their intervention group attend back exercise classes bi-weekly for three months, while subjects in Alaranta, Rytokoski, Rissanen, Talo, Ronnemaa, Puukka, Karppi, Videman, Kallio and Slatis (1994) received treatments for three weeks. (6)   As compared to the laboratory studies, these studies had longer-term follow-up measures. Seven out of fifteen (7/15) studies had follow-up times of one year or more, and the rest had follow-up times of less than 1 year. For example, Harma, Ilmarinen, Knauth, Rutenfranz and Hanninen (1988) conducted a follow-up assessment at 4 months; Greenwood, Wolf, Pearson, Woon, Posey and Main (1990) conducted follow-up assessment after 18 months; and Linton, Hellsing and Anderson (1993) conducted a follow-up assessments after 3 weeks, 6 months and 12 months. (7)   The studies tested interventions aimed at "real life" situations using "actual workers". Studies were not constrained to a small focus, which enhances the generalizability to the workplace. (8)   Seven out of fifteen (7/15) studies used statistical techniques that analyzed multiple independent variables simultaneously. Such techniques are advantageous because they allow for statistical control of confounders. (9)   The studies had large sample sizes. Only two of the fifteen studies had less than 100 subjects at pre-intervention. Three out of fifteen had more than 400 subjects at pre-intervention, while the majority of the studies (10/15) had sample sizes between 100 and 400 subjects at pre-intervention. (10)   Fourteen of the studies measured subjective perceptions of symptom presence or pain. Eight of the studies measured disorder endstates (such as diagnosed injury, sick days due to injury). Instances of the latter include Greenwood, Wolf, Pearson, Woon, Posey, and Main (1990), who measured days of disability and Bergquist-Ullman and Larsson (1977), who measured days absent from work.

OCR for page 200
--> Weaknesses: (1)   The subjects in these studies were previously injured workers. It is possible that the results would not be generalizable to healthy workers, because injured workers may behave differently than healthy workers. (2)   Ten studies reported participant attrition at follow-up assessment. The other five studies did not report whether or not subjects were lost to follow-up. For example, Kellet, Kellet and Nordholm (1991) started with 111 subjects at pre-intervention and had a final count of 85 subjects at follow-up, and Cooper, Tate, Yassi and Khokhar (1996) had 158 subjects at pre-intervention and ended up with 128 at follow-up. (3)   In eleven of the fifteen studies subjects received multiple treatments within an intervention group, and it was not possible to identify which treatments were responsible for the outcomes. For example, the intervention group in Linton, Bradley, Jenson, Spangfort and Sundell (1989) received physical therapy, training, and pain management for their intervention. (4)   With the exception of one study, none of the other fourteen adjusted the alpha level to correct for multiple testing. B. General Conclusions about the Findings from the Field Intervention Research Studies using Injured Subjects: (1)   When evaluated as a whole, there appear to be benefits due to the interventions for reduced musculoskeletal pain and symptoms, earlier return to work, and for reduced use of sick leave. Five studies found positive results, nine found mixed results, and only one found no results. (2)   Looking solely at exercise interventions, there appear to be positive effects for greater trunk flexion, reduced risk of re-injury, earlier return to work, and reduced use of sick leave. (3)   Early intervention right after a current injury does not provide benefits over later intervention post injury for persons with a prior musculoskeletal injury. 2.2. Examples of Intervention Studies with Healthy Workers: Thirteen intervention studies were selected as representative of the literature that used healthy employees as subjects. Table 4 provides highlights of these studies.

OCR for page 200
--> Table 1. Phases of literature review and analysis Phase 1   Comprehensive search yielded 720 articles. Phase 2   Read abstracts. Examined review articles for additional titles. This brought the total number of articles to 768. Phase 3   Obtained relevant articles from the library. There were 198 such articles. Phase 4   Twelve of the articles were unavailable. Read and categorized 186 articles. Phase 5   Selected 43 articles to represent the literature.

OCR for page 200
--> Table 2. Methodological characteristics of the laboratory intervention studies Authors Subjects Intervention1 Random assignment (RA)or random order (RO) or counter- balancing (CB)2 Use of control condition3 Dependent measures Statistics Results4 Comments Leskinen et al., 1983. • 20 male subjects. • Screened for recent back trouble or spinal surgery. Lifting technique. Not indicated. Not applicable. Force at feet, spinal compression. • Paired t-tests. Force at feet (+), spinal compression (0)5. • Within-subjects design. Stubbs et al., 1983. • Eight female student nurses (w/9 month experience). Age range 19 to 23 years. • None had taken sick leave for back pain and there was no recent history of illness or abdominal operations. Lifting technique. RO. Not applicable. Intra- abdominal pressure (IAP), comfort. • 2-way Anova. IAP (+), comfort (0)6. • Within-subjects design. • Forty lifts for each procedure. Hart et al., 1987. • 20 male subjects (mean age = 32.9 years) currently lifting and carrying weights. • No sign or symptoms of acute low back pain. Recruited from local industries. Lifting technique. RO. Not applicable. Flexion, muscle activity (EMG). • Mixed effects 3-way Anova. • Duncan's multiple range test used for post-hoc analysis. Trunk flexion (> lordosis), abdominal muscles (+), external oblique (+), erector spinae (< lordosis, > kyphosis)7. • Within-subjects design. Woodhouse et al., 1990. • 10 well-conditioned male athletes aged 21-35. • Subjects could not participate if they had any one of a number of medical conditions diagnosed during a physical exam provided as part of the study. Weight belt. CB. Yes. Force, work, power. • 1-way repeated measures Anova with Scheffe tests for post-hoc comparisons. Force (0), work (0), power (0). • Within-subjects design. • All lifts were squat-style lifts, at maximum contractions. Lander et al., 1990. • 6 skilled male adults who regularly weight lifted (mean age = 23.4). Weight belt. RO. Yes. Force, intra-abdominal pressure (IAP), muscle EMG, joint moments. • 2-way repeated measures Anova or 2-way repeated measures Anova, as needed. Absolute and relative joint angles (0), IAP (+), L5/S I moment (-), rectus abdominus (0), external obliques(+), erector spinae (+). • Within-subjects design. • The results of the mean EMG values in the proceeding column were divided by the L5/S I moment. Lander et al., 1992. • 5 skilled male adults who regularly weight lifted (mean age = 23.4). Weight belt. RO. Yes. Force, intra- abdominal pressure (IAP), muscle EMG. • 2-way repeated measures Anova with planned comparisons. Force platform (0), joint angles (0), IAP (+), back extensor (0), abdominal constrictor (0), knee extensor (-), hip extensor (-). • Within-subjects design. • Lifts done at maximum effort. Lavender and Kenyeri, 1995. • 11 males and 5 females (age 18-33). Weight belt. CB. Yes. Maximum acceptable weight. • Repeated measures Anova. Maximum acceptable weight (0). • Within-subjects design. • 2 lifts/minute for 40 minutes/condition.

OCR for page 200
-->   Subjects Intervention1 Random assignment (RA)or random order (RO) or counter- balancing (CB)2 Use of control condition3 Dependent measures Statistics Results4 Comments Marley and Duggasani 1996. • 8 college-aged males in good health (age 22-39). Weight belt. RO. Yes. Seventeen physiological, kinematic, and psychological variables. • Full factorial Anova. Blood pressure (greater with belt), no difference (0) on all other variables. • Within-subjects design. • Lifting style allowed to vary. Nakaseko et al., 1985. • 30 female and 1 male trained typist (age 17-52) typing at least 100 strokes/min. Alternative keyboard and wrist rest. RO. Yes. Pain, force, body posture. • Anova and t-tests. Smaller wrist rest = up right posture, lower elbow position. Split keyboard + large wrist rest = greater inclination, arm elevation, elbow angle. Shoulder flexion and abduction: large wrist rest > small wrist rest. Ulnar abduction: traditional keyboard > split keyboard. Neck/Shoulder (0), Arm/Hand (0). • Within-subjects design. • 30 minutes of typing per trial. • Subjects used their preferred workstation settings. Swanson et al., 1997 • 50 female clerical workers (age 18-38) in good health with a minimum of 6 months experience with keyboard work and typing a rate of 40-55 words/min. Alternative keyboard. Not indicated. Yes. Discomfort. • Anova Overall musculoskeletal discomfort (0), fatigue (0). • Keyboard conditions were between-subjects. • Typed 300 minutes per day for 2 days. • Workstations adjusted so that all subject body postures were equivalent. Smith et al., 1998 • 18 professional touch typists from a temporary agency who typed at least 55-words/min with five of fewer typing errors in a 5-min. test. • Subjects were screened for any history of musculoskeletal cumulative trauma disorders (age 18-49, typing experience 6-32 years) Alternative keyboard and wrist rest. CB. Yes. Posture, discomfort. • Wilcoxon signed rank test for repeated measures variables. • Mann-Whitney for between-subject variables. Musculoskeletal pain (0), hand pronation (traditional keyboard > split keyboard), shoulder and elbow pain (without wrist rest > with wrist rest). • Mixed design. • Typed 2 days with the alternative keyboard (8 hours) and 1 day with the standard keyboard (4 hours) • workstations adjusted so that all subject body postures were equivalent. Schoenmarklin and Marras, 1989 • 8 healthy right-handed men who were novice hammer users and had no hand or wrist injuries (age 23-29) Hammer handle angles. Not indicated. Yes. Wrist angle deviations. • Manova. • Follow-up Anova if the Manova was significant. • Used Duncan's test for mean comparisons. Ulnar deviation less with angled hammers, radial deviation less with straight hammer, driving force (0). • Within-subject design. • 57 hammer strikes per minute for 3 minutes per condition.

OCR for page 200
--> Authors Subjects Intervention1 Random assignment (RA)or random order (RO) or counter- balancing (CB)2 Use of control condition3 Dependent measures Statistics Results4 Comments Oh and Radwin, 1993 • 7 make, 11 female students and 8 male, 3 female factory workers. • The factory workers were experienced hand tool users. Trigger and handle spans of pneumatic power hand nut runners RO. Not applicable. Finger/hand forces. • Regression, Anova with Tukey post-hoc. Grip strength affected by handle span. Peak finger and palmar forces increased as handle span increased. Finger and palmar holding exertions (extended trigger < conventional trigger). • Within-subjects design. Lin and Radwin, 1997 • 6 male, 1 female. Pace, force, angle RO. Not applicable. Perceived discomfort • Anova Discomfort ratings increased with increased pace, force, and angle. • Within-subjects design. • All subjects used the same arm/hand positioning. Resnick and Chaffin, 1997 • 5 young healthy males and 5 young healthy females who did not report any musculoskeletal problems (mean age = 20) Manual material handling devices RO. Not applicable. Push and pull forces • Repeated measures Anova. Peak pull force (0). Peak push force: articulated arm < hoist with overhead rail < hoist with fixed pivot. • Within-subjects design. • 30 second trials. 1 If the interventions listed are separated by an ''or", that means there were more than one intervention group. If several interventions are separated by commas, it means that a single intervention group received all of those treatments. 2 Some of the studies used random assignment after stratification. 3 A group that received no treatment, whether randomly assigned or not. This column may contain a description of the control condition. 4 (+) means that the intervention had better scores on the DV, compared to the control/placebo/treatment as usual. (-) means that the intervention had worse scores on the DV, compared to the control. (0) means that the intervention and control did not differ. 5 Squat lifting technique compared to the other lifting techniques. 6 Australian (shoulder) lifting technique compared to the other lifting techniques. 7 Straight back compared to lifting with a lumbar lordosis or kyphosis.

OCR for page 200
--> Table 3. Methodological characteristics of the injured-subject field intervention studies Authors Subject Intervention1 Random assignment2 Use of control group3 Dependent measures Statistics Results4 Comments Lindstrom et al., 1992. • 103 (pre-intervention) blue-collar workers from a care assembly plant sick-listed for 8 weeks with sub-acute low back pain. 98 subjects remained post-intervention. • Exclusion criteria included specific diagnoses. Exercise. Yes. Treatment as usual. Return to work, sick leave, recurrence of pain. • t-tests and Log likelihood. • Assessed whether the groups differed on a number of potential confounds. Return to work (+), sick leave (+), and recurrence of low back pain (+) • Comparisons of 1-year pre- intervention, year of the intervention, and I-year post. • The physicians who made the return to work decision were not blinded to their patients' experimental condition. Alaranta et al., 1994. • 293 patients (pre-intervention) with back disease without inflammation, pain duration at least 6 months, 30-47 years old, no claims, one back surgery maximum, no other rehabilitation. 287 patients at post. Exercise. Yes. Traditional treatment. Flexion, strength, pain, sick leave, symptom presence. • Chi-square, t-tests, paired t-tests or Wilcoxon. Subjective back pain (+), sick leave days (0). For males in intervention group, flexion (+) and rotation (+) • 3 and 12 months follow-up evaluations. • Groups stratified by age and gender. Kellet et al., 1991. • One hundred eleven (85 at post- intervention) employees of a company. • Inclusion criteria: self-reported current or previous back pain; written communication; willingness to exercise at least once a week outside working hours for 1.5 years. • Exclusion criteria: any period of sick leave greater 50 days during 1.5 year prior to study; other medical reasons affecting the employees ability to participate. Exercise. Yes. Yes. Number of sick leave days, cardiovascu- lar fitness, self-reported back pain. • Paired t-test, t-test for independent groups. • Tested for between group differences at pre- intervention. Change score in sick days leave (+) and in episodes of back pain (+). Within exercise group, # sick days (+), # of episodes of back pain (0), and cardiovascular fitness (0). Within control group, # sick days (0), # of episodes of back pain (0), and cardiovascular fitness (-). • Follow-up at 1.5 years. • Prospective study. Harma et al., 1988. • 119 women volunteered for the physical training intervention study. Only 75 at post-intervention. • Criteria: at least 1.5 years of experience in shift work, age 20-49 years, and working as a nurse or nursing aide in a specific hospital. Physical training. No Yes. Musculoskeletal symptoms, physical fitness. • Wilcoxon test and Mann-Whitney U-test. Physical fitness (+) and musculoskeletal (+) between groups. Within physical training group, physical fitness (+) and musculoskeletal (+). • Follow-up at 4 months. • Groups were formed by matching subjects. Moffet et al., 1986. • 92 patients (pre-intervention) aged 18-67 (both genders) with more than 6 months low back pain in a clinic. 78 patients post-intervention. • Excluded for: history of spinal surgery, attending physiotherapy, evidence of an underlying disease. Back school or exercise. Yes. No. Pain, functional disability, activity limitations. • Change scores with t-test and multiple regression. • Assessed whether the groups differed on a number of potential confounders. At 6 weeks: activity (back school >), pain (0), disability (0). At 16 weeks: activity (0), pain (0), and disability (back school >). • Follow-up at 6 and 16 weeks. • The physiotherapists and rheumatologists who assessed patients were blinded to the study conditions.

OCR for page 200
--> Authors Subjects Intervention1 Random assignment2 Use of control group3 Dependent measures Statistics Results4 Comments Donchin et al., 1990. • 142 hospital employees with at least 3 annual episodes of low back pain. Back school or exercise. Yes. Yes. Flexion, strength, pain, back extension, back muscle endurance. • Paired t-tests, Ancova, multiple regression. • Assessed whether the groups differed on a number of potential confounders. Trunk flexion (exercise > back school > control). Abdominal strength (exercise > back school > control) at 3 months, and (0) at 9 months. Back extension and muscle endurance (0). Back pain (exercise > back school or control). • Post-intervention assessments done after 3 and 9 months. Bergquist- Ullman and Larsson, 1977. • 217 patients with low-back pain (pre- intervention). 197 patients post- intervention. Back school or physical therapy. Yes. Short waves of lowest possible intensity heat. Pain, absence from work, duration of symptoms, number and duration of recurrences and absence due to recurrences. • Chi-square tests, Anova and Ancova. Days between first treatment and recovery: back school and physical therapy faster than placebo. At 6 weeks, pain (0). At I year, incidence of recurrences (0), length of recurrence (0), and absence due to recurrences (0). • Assessed the effects of covariates. • 6-weeks and I-year follow-up reported (follow-ups occurred 10 days, 3 weeks, 6 weeks, 3 months, 6 months and I year after). • Subjects were stratified and randomly assigned to groups. Greenwood et al., 1990. • Worker's compensation fund population. • Coal industry sample. Sample of 284 claims. Early rehabilita- tion. Yes. Cases handled in the usual way. Days of disability, amount of medical and disability, benefits. • Two tailed studentized test, chi-square tests. • Test for between group differences at pre-intervention. Length of disability (days) (0), disability benefits paid (0), medical benefits paid (0). • Follow-up at 18 months. Cooper et al., 1996. • All registered or licensed practical nurses employed at a hospital that sustained a compensable soft-tissue back injury. • Sample was screened for concomitant non-occupational musculoskeletal lesion or confounding treatment. Pregnant subjects and those with absence leave of more than 5 weeks were excluded. • 40 (38) in the nurses intervention group, 118 (90) nurses in the control group [pre (post)]. Education and early rehabilita- tion. No. Yes. Perceived pain and disability • 1-way Anovas, 3-way Anovas for repeated measures, 2-way Anovas, regression models. • Compared groups for demographic character- istics at pre-intervention. • Used adjusted p-level of .01. Perceived pain [levels low, mid, high] (0), perceived disability [levels low, mid) (0), perceived disability [level high) (+) Within intervention and control groups, perceived pain [high] (+), perceived disability [mid, high] (+). • Intervention group drawn from high-risk wards. Control group from other wards. • Nurses were classified by blocking characteristics. • Follow-up at 6 months. Mellin et al., 1989. 288 men, 168 women. Rehabilita- tion or back treatment or exercise and ergonomic instruction. Yes. Written and oral instructions in exercise and ergonomics Index of physical measurements (IPM). • One-way Anova and t-tests. Multiple linear regression. • Tested for between group differences at pre-intervention. Change in IPM: better for inpatients vs. outpatients, better for inpatients vs. control, (0) outpatients vs. control. • Same physiotherapist did the measurements at pre and post intervention. • 3-month follow-up reported (study had 3, 8, and 18 months follow up, a 2nd intervention at 1.5 yrs, and another follow up at 3 and 12 months after).

OCR for page 200
--> Authors Subjects Intervention1 Random assignment2 Use of control group3 Dependent measures Statistics Results4 Comments Harkappa et al., 1989. • 476 at pre-intervention (459 post). • Selection criteria: physically strenuous or moderately strenuous work for atleast 10 years; suffered from chronic or recurrent back pain for at least two years; working and physical capacity was affected; sick leave during last two years; low back pain was the major health problem, no other severe long-term illness present. Rehabilita- tion or back treatment or exercise and ergonomic instruction. Yes. Written and oral instructions in exercise and ergonomics. Pain index, disability index, compliance with treatment. • 3 and 2 Way Anovas for repeated measures. 2-way Anovas and Chi-square analysis. • Tested for between group differences at pre- intervention. Pain scores and disability scores were less for inpatients and outpatients compared to controls. Pain was less for inpatients compared to outpatients. • 3-month follow-up reported. (study had 3, 8, and 18 months follow up, a second intervention at 1.5 years, and another follow up at 3 and 12 months after). Harkappa et al., 1990. • 476 at pre-intervention (402 post). • Selection criteria: physically strenuous or moderately strenuous work for at least 10 years; suffered from chronic or recurrent ack pain for at least two years; working and physical capacity was affected; sick leave during last two years; low back pain was the major health problem, no other severe long-term illness present. Rehabilita- tion or back treatment or exercise and ergonomic instruction. Yes. Written and oral instruction in exercise and ergonomics. Pain index, disability index, compliance with treatment, days of sickness allowance. • 1-way Ancovas. One- way Anovas and Chi- square tests. • Tested for between group differences at pre- intervention. Pain index (0), disability index (0), compliance better for inpatients vs. outpatients and controls. Days of sickness allowance greater for controls vs. inpatients and outpatients. • 2.5 year follow-up reported. (study had 3, 8, and 18 months follow up, a second intervention at 1.5 years, and another follow up at 3 and 12 months after). Bru et al., 1994. • 111 subjects. • Selection criteria: females; different professions; availability; reported pain in the neck, shoulder and/or low back over the last seven days; reported pain in the neck, shoulder and/or low back that caused leave of absence for some period over last 12 months; back pain had to be reported for at least 2 periods over the last six months. • Criteria to drop subjects: Medical conditions (e.g., rheumatoid arthritis, Bechterew's disease, epilepsy, previous surgery of spine, osteoporosis, breast cancer, fibromyalgia, pregnancy). Cognitive or relaxation or cognitive and relaxation. Yes. Yes. Neck pain, shoulder pain, low-back pain. • Manova treating data as doubly multivariate, with repeated measures, Mancovas. • Used pre-test scores as covariates. Directions of the changes between the groups were not provided. • Follow-up immediately after changes between the groups were not provided. and 4 months after intervention. Linton et al., 1989. • 66 female LPNs or nursing aids. Screening: had to have been sick-listed for back pain at some time during the previous -year period and had to be currently working. Physical therapy, training, pain. manage- ment Yes. Yes. Pain, activities of daily living, absen- teeism • Separate 2 X 2 (treatment group x assessment period) analyses of covariance for repeated measures. • Tested for group differences at pre- intervention. • Pre-test value of each measure served as covariate. Pain intensity (+), pain behavior (+), activities of daily living(+). Pain- related absenteeism (+) • Follow-up every 6 weeks for six months (no assessments during these) and at 6-months (assessment made).

OCR for page 200
--> Authors Subjects Intervention1 Random assignment2 Use of control group3 Dependent measures Statistics Results4 Comments Linton et al., 1993. • 240 patients [pre] complaining of MSP (musculoskeletal pain). • Study 1: 106 patients [post] with history of MSP during the past 2 years but not sick listed during the most recent three months. • Study 2: 92 patients [post] who had not been sick listed for MSP during the past 2 years. Early intervention. Yes. Treatment as usual. Pain, days off of work. • t-test, chi-square tests. • Tested for between group differences at pre- intervention. Study 1: Within intervention and control group: Pain today (+), pain/week (+), pain-free days (+), and activity index (+). Pain control (+) within controls. Between groups: Pain today (0), pain/week (0), pain-free days (0), activity index (), pain control (0). Chronic pain (0) and sickness absenteeism (0). Study 2: Within intervention and control group: Pain today (+), pain/week (+), pain-free days (+), and activity index (+) Pain control (+) within controls. Between groups: Pain today (0), pain/week (0), pain-free days (0), activity index (0), pain control (0). Chronic pain (+) and sickness absenteeism (+). • Follow- up 3 weeks, 6 and 12- months. 1 If the interventions listed are separated by an "or", that means there were more than one intervention group. If several interventions are separated by commas, it means that a single intervention group received all of those treatments. 2 Some of the studies used random assignment after stratification. 3 A group that received no treatment, whether randomly assigned or not. This column may contain a description of the control condition. 4 (+) means that the intervention had better scores on the DV, compared to the control/placebo/treatment as usual . (-) means that the intervention had worse scores on the DV, compared to the control. (0) means that the intervention and control did not differ.

OCR for page 200
--> Table 4: Methodological characteristics of the healthy-subject intervention studies. Authors Subjects Intervention1 Random assignment2 Use of control group3 Dependent measures Statistics Results4 Comments Orgel et al., 1992. • 23 (of 34) grocery store cash register employees at pre- and 19 at post-intervention. Multiple changes at workstation. No. No. Musculoskeletal discomfort. • Wilcoxon signed rank test. Medication use (+), recovery days (+), discomfort on low back/buttock/leg (+), on neck/upper back/ shoulder (+), and arm/forearm/wrist (0). • Post-intervention measures taken 4 months post. • Controlled for hours of work at register. May and Schwoerer, 1994. • 800 production employees in 2 shifts of a meatpacking plant whose primary tool is a knife. Multiple ergonomic improve- ments. No. No. CTDs/ employee, # physician referred CTD cases, production days lost, restricted working days. • Wilcoxon signed ranks. # of CTDs (+), # of doctor referred cases (+), production days lost (0), restricted duty days (+). • Compared measures from 1 year prior to the interventions to 1 year post-intervention. Lanoie and Tavenas, 1996. • About 90 packers in a warehouse. Multiple ergonomic improve- ments. No. No. Accidents, back related injuries. • Poisson regression. • Statistically controlled for various possible # of accidents (0), # back related injuries (+) confounders. • Assessed measures over the 3 years it took to complete all of the interventions. Keyserling et al., 1993. • Subset of 151 jobs with problems—taken from 335 jobs. • To be selected, jobs must have had at least one potentially hazardous ergonomic exposure. Multiple ergonomic improve- ments. No. No. Posture. • Paired t-tests. Trunk posture (+), shoulder posture (+), neck posture (-) • Study lasted 42 months. • Participative union-management program. • Workstation redesign most often used. Garg and Owen, 1992. • 38 of 57 nursing assistants in one nursing home. • 95% female, age 19-61, .5-20 years experience, • 75% had suffered low back pain. Training. No. No. Injury incidence, severity rates, physical stress, and L5/SI compression. • 2 sample t-tests and Anova. Hand force (+), L5/S1 force (+), Incidence back injury (+), severity back injury (+). • Post-intervention assessments done after 4 and 8 months for some employees, and only after 4 months for the rest. Wickstrom et al., 1993. • 88 planers (age range 24 to 55 years) and 125 sheet metal workers (age range 19 to 56 years). Training in biomecha- nics, physical training, ergonomic evaluation. No. All employees at a different metal industry. Occurrence of low back pain, registered sick leave. • Chi-square, t-tests. Only for sheet metal workers: Fitness of low back tissues (+), occurrence low back pain (0), sick leave due to back pain (+) • Follow-up I year after intervention. • Used participatory groups. Feldstein et al., 1993. • 45 nurse aides and orderlies on 2 surgical units (age range 19-62 years) at pre-intervention and 37 at follow-up. Back education. No. Yes. Pain. • Chi-square, t-tests, or Ancova. • Statistically controlled for various possible confounders. Back pain (0) for within and between groups. • Post-intervention assessment after 1-month. • Control group was a unit not getting the intervention.

OCR for page 200
--> Authors Subjects Intervention1 Random assignment2 Use of , control group3 Dependent measures Statistics Results4 Comments Daltroy et al., 1997. • 4000 postal workers at 2 mail processing facilities, Back education, pain manage- ment, lifting training, ergonomics. Yes. Yes. Injury rates, likelihood of repeat injury. • Extended log-linear models, Wilcoxon. • Statistically controlled for between group differences. Low back injury rates (0), back injuries attributed to lifting and handling (0), other musculo-skeletal injuries attributed to lifting and handling (0). • Study lasted 5.5 years. • Control group members could be re-assigned into the intervention group if they were injured. • Matched on job title and job characteristics. Versloot et al., 1992. • 500 bus drivers. Back education. Yes. Yes. Absenteeism. • Manova. • Statistically controlled for between group differences. Absenteeism (0). • Compared 2 years pre-, 2 years during, and 2 years post- intervention. • Intervention and control groups came from 2 different geographic locations. Parenmark et al., 1988. • 33 newly hired assembly workers (1945 years old) without arm, neck, or shoulder complaints. • 60 assembly workers with more than 1 year experience. None of them reported being ill. Movement pattern training. Not indicated. Yes. Sick days. • Wilcoxon. For new hires, total sick days (+) and upper extremity sick days (+). For experienced workers, total sick days (0) and upper extremity sick days (0). • 48-week post-intervention assessment. Gundewall et al., 1993. • 69 nurses and nurse's aides between 18 and 58 years of age (1 male). 60 remaining at post-intervention. Subjects comprised people with and without back pain. Exercise. Yes. Yes. Strength, lost work days, pain. • 2-sample t-tests, paired t-tests, Mann-Whitney test. • Tested for between group differences at pre- intervention. Lost workdays (+), days with complaints (+), intensity of pain (+), back muscle strength (+). • Subjects were stratified. • Investigators were not blinded • Follow-up at 13 months. Mitchell et al., 1994. • 1316 warehouse workers in 5 different areas of an air force airport (mean age = 41.3, 974 males). Use of back belt. No. Yes. Back injury. • Used chi-square and logistic regression. • Controlled for several possible confounders in the ogistic regression. Use of back belt (0). • Retrospective study. Reddell et al., 1992. • 642 baggage handlers working for a major airline (572 males, age 19-67) in five different job types (though all manually handled baggage, mail, or supplies). Use of back belt, training. Yes. Yes. Lumbar injuries, lost work days, restricted work days, worker's compensation costs, hours worked. • Anova. Total case injury incidence rate (0), restricted workday case injury incidence rate (0), # lost orkdays (0), # restricted workdays (0), worker's comp. rates (0). • Experiment lasted 8 months. • No statistical comparison of possible between groups differences. 1 If the interventions listed are separated by an "or", that means there were more than one intervention group. If several interventions are separated by commas, it means that a single intervention group received all of those treatments. 2 Some of the studies used random assignment after stratification. 3 A group that received no treatment, whether randomly assigned or not. This column may contain a description of the control condition. 4 (+) means that the intervention had better scores on the DV, compared to the control/placebo/treatment as usual. (-) means that the intervention had worse scores on the DV, compared to the control. (0) means that the intervention and control did not differ. "Mixed" means that the intervention group differed from the control group on some of the outcome measures.

OCR for page 200
This page in the original is blank.