Вы находитесь на странице: 1из 95

Development of a novel link-segment model for estimating lower back loading in paramedics

by

Peter Alexander Wetherall Galbraith

A thesis submitted to the School of Kinesiology & Health Studies in conformity with the requirements for the degree of Master of Science

Queens University Kingston, Ontario, Canada September 2011

CopyrightPeter Alexander Wetherall Galbraith, 2011

Abstract
Work conducted as part of this thesis evaluated the lifting techniques of paramedics using a novel link-segment model that was validated against a commercially available software package, 3D Static Strength Prediction Program (3DSSPP). Site visits to four paramedic services across the province were conducted to gain information about bags weights and lifting techniques. Twenty-five paramedics then visited the Biomechanics Lab at Queens University to participate in testing sessions mimicking the daily lifting and carrying tasks performed by paramedics on the job. Participants were outfitted with the Xsens Motion Tracking System and asked to lift and carry bags ranging from 5-20kg. Output from the Xsens system was used in a 3D-inverse dynamic model to estimate loading at the L5/S1 joint. The compressive and shear force estimates at this joint are of particular interest given their correlation with low back pain and injury. Across all conditions the greatest compressive forces were seen during bag pickup and bag release. Additionally, reaching forward 50 cm at pickup increased peak spinal compressive loads by nearly 300N and 500N for a 5kg and 10kg handbag respectively. Not surprisingly, at bag release greater trunk lean values were correlated with higher 1

compressive force estimates. Single-shoulder backpack carries showed similar loading characteristics when compared to double-shoulder backpack carries. Shear force estimates remained well below acceptable levels across all conditions. Based on paramedic feedback, a supplementary testing session was performed with a single participant to evaluate multi-bag carries and stair climbing. The results of this testing session showed that loading was reduced at pickup and release when the load was distributed across two bags. This research led to the development of four recommendations that have been presented to the Association of Municipal Emergency Medical Services of Ontario. 1. Paramedics should not lift single bags or a combination of bags that exceed 20kg. 2. Prior to lifting, bags should be located as close to the paramedic as possible. 3. When placing bags on the ground and when picking bags up off of the ground, paramedics should use a squat lift technique to prevent forward and side bending. 4. When multiple bags are carried the load should be evenly distributed within bags and across sides of the body.

Acknowledgements
I would like to start by thanking my supervisor Dr. Pat Costigan for his help and guidance from start to finish. Not to mention saying that an independent study on longboarding was entirely reasonable. Dr. Joan Stevenson, without whom this project could not have happened. Thank you to my family (Mom, Dad, and Jamie, as well as my extended family in Kingston and Calgary) for encouraging a healthy curiosity in all things and supporting me throughout my life. Rachel for being there at the end of some long days and always being willing to discuss my research. The many friends I have made in the department for always being ready for a celebratory or conciliatory pint, and providing me with so many lessons and experiences that could never take place in a classroom.

Table of Contents
TOC \o "1-3" Abstract Table 1 - Conditions presented to paramedics during in-lab testing. All conditions included a 4m carry.

Figure 1 - XSens Motion Tracking System sensors. Sensors are placed on the lower arms (1&2), upper arms (3&4), scapulae (5&6), upper back (7) and lower back (8).

Chapter 1: Introduction
In August 2010, the Association of Municipal and Emergency Medical Services (AMEMSO) contacted the Queens University Biomechanics Lab in the School of Kinesiology and Health Studies to investigate the weight and design of equipment bags carried by paramedics in the province of Ontario. The 72 certified land ambulance services in Ontario respond to an estimated 1.5 million calls annually. In responding to these calls, paramedics carry their equipment bags over long distances, up and down stairs, and through confined spaces. Given that paramedics can respond to a number of calls per shift, it is not surprising that the various types, size, and weights of these bags are a concern. Despite the fact that common equipment is carried in the bags, there are no standards governing the size and number of bags or the weight carried in any particular bag. The results of this research provides insight into the forces experienced by paramedics while lifting equipment bags on the job and details the aspects of lifting that increase the risk of injury as put forward by the National Institute for Occupational Health and Safety guidelines (Waters et al. 1993). From these findings, the researchers recommend ways to improve the lifting conditions and reduce the strain on paramedics while lifting and carrying bags on the job.

For paramedics, two common scenarios occur: (1) when they are required to lift an individual from the ground while holding a bag in one hand, and (2) when they have to guide themselves and the patient around or over obstructions. The variety of sites that paramedics visit means that paramedics often experience awkward postures. Studies assessing paramedic injuries have shown that low back strains are a major source of time off work and may be one reason why paramedics have such high injury rates (Hogya & Ellis 1990; Okada et al. 2005) Clearly, it is important to understand the loads experienced by paramedics across a variety of lifting conditions. An understanding of the magnitude of these loads while performing paramedic work is essential to determine if the weights and lifting techniques are safe. Biomechanical measures that have been correlated with increased risk of low back pain include: peak compressive force, peak shear force and the cumulative load experienced by the L4/L5 or L5/S1 joint, (Norman et al. 1998; van Dien & Toussaint 1997). Compressive forces act along the craniocaudal axis of the spine and under normal loads the vertebral body withstands most compressive forces. However, the extreme case can lead to disc herniation or prolapse (Roaf 1960). NIOSH guidelines (NIOSH, 1981) have put forward a maximum acceptable limit of 3400N and maximum permissible limit of 6400N of compressive force. The acceptable limit represents a compressive force value that should be safe for 75% of women and 99% of men. Beyond 6400N higher risk for injury is predicted. Shear forces act in 3

the medial-lateral and anterioposterior directions along the spine and are another risk factor for low back pain (Kerr et al. 2001). McGill et. al. (1998) have put forward a maximum acceptable limit of 500N and a maximum permissible limit of 100N for shear force along the anterioposterior axis of the spine. Traumatic events are not always the cause of low back pain. Repetitive loading of even small amounts can lead to low back pain and injury over time. If we consider each of these loading instances a single trauma, the cumulative loading is the accumulation of small traumas over time. As no cumulative exposure limits have been found in the literature, determination of safe and unsafe tasks based on cumulative loading is difficult. Link-segment modeling is often used to quantify these compressive, shear and cumulative loads of the task at hand. Link segment models (LSM) use basic physics equations and represent the body as a series of rigid, connected segments. LSM incorporate anthropometrics and individual measurements of motion as well as estimates of the external forces to provide an estimate of the forces and moments imposed on the joints. Most LSM of the spine also estimate the surrounding musculatures force contribution that is often substantial and, therefore, must be considered. The compressive force contributed by the back extensors, which includes many different muscles, is often estimated by representing all muscles as a single muscle equivalent. Joint load estimates increase when dynamic parameters of motion are included and when the lifting task is asymmetrical (Marras & Granata 4

1995). For better estimates of joint load, it is important to use a fully dynamic three-dimensional model. We hypothesized that heavier loads would lead to higher compressive force estimates and that lifts using backpacks and shoulder bags would produce lower estimates than those conditions using handbags because of the improved load location. Additionally, it was expected that those paramedics that adopted a squat posture when releasing bags would experience lower compressive forces than those who chose to lean forward to release the bag. It was hoped that this study would enhance the bag selection and design criteria as well as producing guidelines for appropriate lifting technique for paramedics. At the conclusion of this research, the results were presented to AMEMSO in the hopes of improving bag weighting and lifting policies across the province.

Chapter 2: Review of Literature


2.1 Review of Modelling Literature

Human link-segment models have been used in video games, rehabilitation, and biomechanical settings to record and gain understanding about human motion. The human body is modeled as a series of connected rigid links based on subject-specific anthropometric parameters. Link-segment models are then used as input to inverse models using basic force and moment equations to quantify the forces and moments experienced by the spine, especially 5

when examining lifting (Kingma et al. 2001; Abdoli-Eramaki et al. 2009; Norman et al. 1998; Potvin, McGill, & Norman 1991). To better estimate loading on the lower back the muscular contributions must be taken into account. To include the inertial components required in a dynamic inverse model, effective motion capture tools are required to measure the motion under investigation. 2.1.1 Capturing Human Motion Before any investigation can be made into the kinetic or kinematic properties of human motion, researchers must be confident that the recorded motions closely represent the actual motions that took place during data collection. Early motion capture techniques relied heavily on film recordings and are time consuming to process, often requiring manual digitization and error checking throughout the entire process. The advent of digital video recordings has sped up many of these processes and is still a key component of modern biomechanical tools such as HU-M-AN (HMA Technology, Canada), 3DSSPP, (University of Michigan, USA) and 3DMatch (University of Waterloo, Canada). Other motion capture systems require participants to be instrumented with light emitting diodes or reflective surfaces. These motion capture systems, such as Vicon Nexus (Vicon Motion Systems, USA) and Optotrak Certus (Northern Digital Inc, Canada), require line of sight and multiple cameras to automatically record 3-dimensional movement, and boast sub-millimeter accuracy. These systems can have large capture volumes but are not particularly portable. Local 2

coordinate systems are created using the sensors attached to each segment. The known orientation between relevant anatomic landmarks and local coordinate system allows the creation of an anatomical coordinate system for each segment. Link-segment models are then built from the anatomical coordinate systems and known anthropometrics. As each segment is located in a global reference frame regardless of the orientation and position of other segments, measurement errors remain relatively constant across all segments. A relatively new technology has emerged in the past decade allowing researchers to capture motion in the field without the need for line of sight or large systems. Such systems rely on accelerometers, gyroposcopes, and magnetometers and advanced software to produce reliable estimates of body segment position and orientation (Roetenberg, Luinge, & Slycke 2009). The Xsens Motion Tracking System (Xsens, The Netherlands) is one of these systems and has been used in motion capture labs and for video game and movie motion capture. These systems are highly valuable for the commercial setting given their real-time capabilities and ease of use. However, the scientific community has questioned the accuracy of these systems for scientific research (Cutti et al. 2006; Luinge & Veltink 2005; Damgrave & Lutters 2009; Brodie, Walmsley, & Page 2008). The main problem is that the measurements drift due to the reliance on magnetometers to determine the sensors heading. Aligning each sensors local coordinate system with the supposed underlying anatomical 2

coordinate system approximates anatomical coordinate systems. Linksegment models developed using inertial sensors must be built sequentially from proximal to distal (or vice-versa). In most cases, the origin of the most proximal segment is assumed to exist at the origin of the global coordinate system. The most proximal segment pivots about the origin based on the orientation of the attached sensor. The most proximal segments endpoint is used as the next segments start point and the process is repeated. Thus errors are accumulated as the LSM is built from proximal to distal, and the greatest position errors are seen in the distal segment. Image-based motion capture systems such as Vicon Nexus are able to avoid cumulative errors in segment orientation because they locate each segment in a global reference frame. The effect of these errors on kinetic and kinematic parameters has been evaluated. Godwin (2009) compared a link-segment model built using Xsens sensors to one using Vicon Nexus and found distal segment endpoint RMS errors of 136 mm, 138 mm, and 101 mm in the anterior-posterior(AP), medial-lateral(ML), and inferior-superior(IS) axes respectively. These errors led to flexion moment RMS errors of 12 Nm, 10Nm, and 4Nm along the AP, ML, and IS axes respectively. These values may seem small but represented between 10% and 30% of peak moments across a variety of trials. Since Godwins work, improvements have been made to the Xsens system by improving the filtering of the accelerometers and 3

gyroscopes. In 2005, Zhou, Hu & Tao showed average error for distal endpoint locations ranging from 10-70mm, and in 2010, Zhou and Hu showed that the system now has RMS position errors of 9mm, and drifts of less than 5 mm/s when used while performing daily activities.These improvements appear to be made by improvements to the proprietary Kalman filter; a technique that combines the predicted and observed values for the accelerometers, gyroscopes and magnetometers. For an explanation of this process see Brodie, Walmsley & Page (2008). These improvements in Xsens output have led to improvements in kinematic accuracy. Cutti et al. (2010) reported RMS joint angle errors of 1.4 and 1.8 for the hip and knee angles when compared with a standard goniometer and errors of approximately 2 at the hip and knee. Basic drawing tests* have shown the Xsens system to be accurate within 0.5 cm using kinematic modeling over periods of 25 seconds and has been deemed acceptable in a neurorehabilitiation setting (Bai et al. 2011). When investigating gait parameters similar repeatability measure values were found when comparing the Xsens system (Cloete & Scheffer 2010) to Vicon Nexus (Kadaba et al 1989) and Polhemus Liberty (Mills et al 2007).
*These drawing tests required participants to repeatedly trace a triangle. Thus when the link-segment model tightly matches the dimensions of the triangle over a series of repetitions we assume that the model is valid.

2.1.2 Modelling the Shoulder Joint Godwin (2009) notes one limitation of her model was that the glenohumeral joint is unable to translate relative to the spine. However, there can be as much as 150 mm of protraction (anterior translation) of the glenohumeral joint itself (Albert et al. 1998), suggesting that a rigid connection is invalid. As noted by Godwin, the inability to represent shoulder protraction may account for some of the errors in distal segment endpoint location. Godwin (2009) and others simply assume that the shoulder joint maintains a constant orientation and position relative to the upper body segment (Cutti et al 2008; Rau, Disselhorst-Klug, & Schmidt 2000; Rab, Petuskey, & Bagley 2002). In these cases the glenohumeral joint is assumed to act as a hinge joint with no shoulder translation. Godwins model is an example of an open loop system where each segment is linked to only one other segment in a chain from back segment to hand segment. Other models (van der Helm 1994a, Dickerson, Chaffin, & Hughes 2007; Maurel et. al. 2010) have constrained shoulder joint translation using a closed link between the scapula, clavicle, and upper back. In this way a triangle is formed between the clavicle, scapula and back and thus shoulder joint translation is limited but not rigid. Yang et al (2010) have suggested that closed-loop models have higher accuracy and fidelity than openloop models.

Three bones make up the shoulder complex: the clavicle, humerus and scapula. The only linkage between the axial skeleton and the upper extremity occurs where the clavicle attaches to the sternum to form the sternoclavicular joint. Sternoclavicular range of motion has been estimated at 20 degrees of protraction (elevation), 60 degrees of forward flexion, and 10 degrees of axial rotation (Inman, Saunders & Abbott 1944; van der Helm 1994b; van der Helm & Pronk 1995). The glenoral fossa of the scapula and the head of the humerus form the glenohumeral joint that permits movement of the upper arm. To represent motions such as shoulder shrugs, and pinching the shoulder blades, LSMs built using inertial systems must instrument the clavicle or scapula to gauge shoulder joint translation. Recommendations exist about how best to model the shoulder complex (Wu et al. 2006); however, problems arise when trying to securely and comfortably attach a sensor over the clavicle (Cutti et al 2008). Additionally, the shoulder joint centre is difficult to determine because it is located deep below the skins surface hiding the bony landmarks that are necessary for joint centre determination (Rau, Disselhorst-Klug, & Schmidt 2000). The effect of shoulder joint translations on inverse dynamic model output should be considered to see if differences between shoulder models lead to substantial differences in force estimates. The only research we found that attempted to quantify this effect showed that shoulder translation had a significant effect on positions and 3

accelerations of arm segments, but did not have a statistically significant effect on L5/S1 moment prediction (Albert et al. 1998). Rab et al. (2000) argued that shoulder joint centre determination errors of 20mm had a negligible effect on shoulder kinematics. It may be the case that shoulder joint centre differences have a small effect on kinetic and kinematic results. 2.1.3 Kinetic Parameters Once link segment models have been built, researchers can investigate the kinetic or kinematic parameters of the model. Kinematic parameters such as: joint angle, segment angle, range of motion, displacement and velocity can be used to investigate differences between groups of individuals or attempt to describe specific aspects of human motion. Alternatively, kinetic parameters can be investigated to gain an understanding of the forces and moment to which the body is subjected. Newtonian physics, built upon the fundamental equations of motion of a rigid body (Zatsiorsky 2002) are used to determine the mechanical forces on various body segments. These parameters are often broken down into their component vectors and expressed in the local coordinate system of the segment of interest. For example force vectors at the L5/S1 disc are often broken down into: a compressive force acting craniocaudally, and two shear forces acting anterior-posteriorly and mediallaterally.

2.1.4

Estimating Muscular Contribution

Calculations based on equations of motion do not provide a true sense of the loading at specific joints. To produce better estimates of joint loading, many models include the force contribution made by the surrounding musculature. As many as 104 different components have been incorporated into spinal models to balance the forward flexion moment using passive and active tissues (Callaghan & McGill 2001), while others use a single back extensor model (Norman et al. 1998). Mathematically -driven optimization models (Brown & Potvin 2005) have also been used to estimate spinal loading. Complex electromyography assisted models (Gagnon et al. 2011; Mientjes et al. 1999; Davis, Marras, & Waters 1998) include the activation of dozens of muscles to estimate loading on the lower back. Potvin et al. (1991) recorded EMG activation of 11 different muscles during symmetrical squat and stoop lifts. They partitioned the reaction moment into 11 muscles and 7 ligaments based on each muscles activation and each ligaments strain based on lumbar flexion angle. Partitioning the reaction moment depended upon assumptions of: ligament stress-strain curves, muscle lines of actions, muscle crosssectional area, and the modeled relationship between muscle contraction velocity, activation and produce muscular force. A large number of assumptions must be made to produce EMG-assisted models.

Given the number of calculations and assumptions required to build EMG-assisted models some researchers have elected to avoid EMG altogether. Optimization models are mathematically-driven functions that partition the reaction moment based on assumed muscular activation patterns. In essence, these models assume the central nervous system is acting to achieve biomechanical equilibrium by minimizing some objective function such as muscular contraction (An et al. 1984, Crowninshield & Brand 1981; Nussbaum, Chaffin, & Rechtien 1995), joint force (Brown & Potvin 2005), metabolic energy consumption (Davy & Audu 1987), or some combination of these factors (Bean, Chaffin, & Schultz 1998; Pel et al. 2008; Seireg & Arvikar 1973). Again, a large number of assumptions are required. One problem with optimization models is their inability to accurately predict muscular co-contraction, which may lead to underestimates of joint loading (Cholewicki, McGill, & Norman 1995). An alternative to optimization models, while still avoiding EMG, is a single equivalent muscle extensor model, in which all muscles that assist in flexion are assumed to act as a single muscle group. This model assumes a single extensor muscle group is the sole means by which the body counterbalances the forward flexion moment, and acts exclusively about the flexion axis. It should be noted that because this muscle group is modeled as acting along the long axis of the spine it cannot counterbalance moments about the inferior-superior (IS) or

anterior-posterior (AP) axes of the spine. However, it does produce a force vector that increases compressive estimates. Decisions regarding moment arm length are important in this case as small length differences can lead to large variations in final joint loading estimates. McGill & Norman (1987) state that 5 cm is a commonly used value for the single extensor muscle moment arm, while Norman et al. (1998) have used a moment arm length of 6cm. These values are in keeping with the real moment arm length for the erector spinae muscle group range of 4.9-6.4 cm (Jorgensen et al. 2001). To factor in some muscular effort to balance IS and AP shear forces, varying the line of action within realistic ranges has been suggested (van Dien & de Looze 1999). These variations can change compressive force estimates by more than 100N and shear force estimates by more than 50N (Nussbaum, Chaffin, & Rechtien 1995). Varied lines of action may not substantially change model output given that heavy lifts commonly exceed the action limit of 3400N of compressive force (NIOSH 1981). These variations may be more relevant for shear force output given that the maximum acceptable limit for shear force is 500N (McGill 1998). It is important to understand how different muscle models influence force output estimates. Potvin et al. (1991) showed that for a 32-kg lift at peak lumbar spine flexion the erector spinae contributed 74% (stoop) and 83% (squat) of all compressive force contributions made by muscular or ligamentous tissues. This indicates that in sagittal 4

plane lifting the erector spinae are the primary contributors to muscular compressive force, and a single extensor muscle model may be an appropriate method in this instance. Cholewicki, McGill, & Norman (1995) found that optimization models predicted approximately 30% lower L4/L5 compressive force values than EMGassisted models, possibly because the optimization model cannot accurately predict co-contraction of antagonistic muscle pairs. There are inherent problems to each model when estimating muscular force contributions. Single muscle extensor models are simplistic and do not attempt to represent the true nature of the back musculature while specifically lacking the muscles acting at oblique angles to the spine that stabilize the spine during twisting and asymmetrical motions. EMG-assisted models rely on a length-strength and forceactivation relationship for each muscle investigated and these relationships may be flawed, especially when considering different strength capabilities across individuals. Additionally, while stability is visibly maintained during most lifts performed during data collections, Brown & Potvin (2005) noted that EMG-assisted models may produce situations where equilibrium is not maintained. As noted before, optimization models rely on mathematical solutions to determine the force contributions of the musculature in the back, and often do not include agonist-antagonist muscular co-activation. In each case some drawbacks are accepted with the goal of achieving more realistic loading estimates. Despite (and possibly because of) their simplicity, 5

single equivalent extensor models continue to be used in biomechanical research.

2.2

Review of Lifting Literature

Lifting is a risk factor for low back pain (Marras et al. 1995; Chaffin & Park 1973). The 2009 U.S. Department of Labors Annual Survey of Occupational Injuries found that 46% of musculo-skeletal disorders were associated with the back and required on average 7 days off work. This survey also found 116,530 injuries requiring time off work were the result of overexertion while lifting. Studies have examined psychosocial, physiological and biomechanical criteria to determine risk factors for low back pain. These factors include: previous instances of low back pain, job satisfaction, job stress, repetitive lifting, heavy lifting, forward flexion, axial twisting, overhead reaching and hand coupling to name a few (Hoogendoorn et al. 2000; Kerr et al. 2001; Marras et al. 1995, 1999; Woolf & Pfleger 2003; Frymoyer et al. 1983; Brinckmann et al. 1998). Jobs requiring frequent and heavy lifting have been associated with increased risk of disc herniation and low back pain in general (Kelsey et al., 1984, Kelsey & White, 1980). Attempts to limit low back injuries have focused on improving postures while lifting and reducing loads (NIOSH 1981; Waters et al 1993). 2.2.1 Compressive Force Guidelines Compressive forces act along the IS axis of the spine and under normal conditions the vertebral body withstands most compressive forces. However, the extreme case can lead to disc herniation or prolapse

(Roaf 1960). Cadaveric research has shown that ultimate compressive forces may been in the range of 2100- 9600N (Brinckmann et al. 1988), having a mean of 4400N with a standard deviation of 1880N (Jger & Luttman 1989). In 1993, the National Institute for Occupational Safety and Health released their lifting equation that identified hazardous lifting tasks based on biomechanical, psychosocial, and physiological factors. Two compressive limits are put forward: the action limit (AL) of 3400N, and maximum permissible limit (MPL) of 6400N. The biomechanical criterion for the equation is based on research showing that: spinal compressive forces of greater than 3400N may increase the risk for low-back injury and injuries may become quite likely beyond 6400N. Waters et al (1993) argue that if the data were normally distributed 21%-30% of lumbar segments would fail when loaded with a force of 3400N given the ultimate compressive force values put forward by Brinckmann et al (1988) and Jger & Luttman (1989). Interestingly, Norman et al (1998) found that the mean peak compressive force of individuals reporting low back pain was 3423N. The AL of 3400N is a conservative estimate for a healthy working population since cadaver lumbar segments may have lower tolerance limits because of declines in lumbar strength with age, as well as changes in bone mineral content (Hansson & Roos 1981). However, when ensuring workplace safety, conservative limits should be used.

2.2.2

Shear Force Guidelines

Shear forces act in the medial-lateral and anterioposterior directions along the spine and are a risk factor for low back pain (Kerr et al. 2001). Krypton et al. (1995) found shear force tolerance limits in cadavers of between 1700N and 2900N. While less work has been done in this area, some guidelines have been developed based on thinking similar to that of the NIOSH Equation. McGill et al. (1998) has put forward an action limit of 500N and a maximum permissible limit of 1000N; these limits are akin to the 3400N action limit and 6400N permissible limit for compressive force. A shear force limit of 500N has been used with reasonable success to predict which workers reported low back pain (Daynard et al. 2001). Tasks that keep compressive forces below 3400N and shear forces below 500N are unlikely to increase the risk of injuries. 2.2.3 Model Complexity Another important point to consider in developing a link segment model is whether to include dynamic components in force and moment calculations. Static models are easier to implement, as fewer calculations are required. Static and quasi-static models have been developed and applied successfully for many years particularly for slower motions, but in cases where inertial contributions are nonnegligible, dynamic models may be more accurate in predicting loading due to the inertial effects of the load and body segments (Marras & Granata 1995; McGill & Norman 1985; Lindbeck & Arborelius

1991; de Looze et al. 1994). Substantially higher predicted loads have been seen when acceleration components are included in force calculations when compared to what is otherwise the same task (Jger & Luttman 1989). Asymmetric lifting is a common occurrence for most tasks due to differences between load origin and destination, movement requirements, obstructions or a variety of other reasons. To that end, postural symmetry is unlikely to happen in working environments. In spite of this many models only take 2D motions into account when analyzing lifting (Albert et al. 1998; Anderson et al. 1985; Waters & Garg 2010). Underestimations of the peak torque have been shown as high as 60% when loads are placed at 90 to the sagittal plane (Kingma et al 1998). Lift asymmetry is identified in the NIOSH lifting equation as a factor that reduces the maximum load that an individual can safely carry (Waters et. al., 1993) It is suggested that researchers may arrive at the wrong conclusions when a 2D model is used for tasks with 30 or more of twisting (Kingma et al. 1998). For these reasons three-dimensional dynamic models should be used when possible.

2.2.4 Paramedic Injury Rates & Lifting Demands Paramedics share similar lifting task demands with a variety of professions such as nurses, nursing aides, and fire-fighters who are required to support a patient and administer some level of care while

simultaneously transporting the tools necessary to provide that care. Studies of the nursing population have shown high levels of low-back pain much of which is related to patient handling, (Jensen 1987, Videman 1984) a task that paramedics are frequently required to perform, with potentially greater strain due to awkward postures and lower lift origins. The compressive force that the spine is subject to while transferring a patient from one space to another has been shown to exceed even the NIOSH maximum permissible limit of 6400N (Marras et al. 1999). It is believed that workers cannot tolerate compressive forces beyond this limit without increasing their risk for injury (Waters et. al. 1993). Patient handling tasks expose paramedics to potentially dangerous low back loads. These loads may be contributing to the high back injury rates seen in paramedics (Crill & Hostler 2005). Back strain injuries (as classified by the International Classification of Diseases, adapted, 8th revision (ICDA 1967)) are very common in the paramedic population and account for 36% of all injuries; of those more than half are caused by lifting activities (Hogya & Ellis 1990). This same study found that each instance of reported back strain led to, on average, 4 days of time off work. High paramedic injury rates are not just a North American phenomenon. A survey of Japanese paramedics and emergency medical technicians found that 25% of respondents had experienced a low back problem in the previous 12 months, and 1/5th had experienced pain for 30 days or more in the 2

same time span (Okada et al 2005). Patient handling may not be the only contributor to injury. Spending as little as 10% of working time in 30 of trunk flexion has been shown to increase the risk for developing low back pain (Hoogendoorn et al. 2000). Paramedics are often required to bend over to lift and interact with the patient and when administering care in the back of the ambulance may spend much of their time in forward flexion. Due to the varied and demanding nature of the job it seems inevitable that paramedics are exposed to those factors that lead to low back pain. Work must be done to improve the working environment for paramedics to prevent injuries and time off work.

Chapter 3: Creating the Link-Segment Model


3.1 Introduction

The goal of this research was to understand the loads generated by individual bag lifts performed by paramedics. Link-segment modeling was used to meet this goal; in this case a novel three-dimensional dynamic hands-down model was developed that incorporated the output from a system of Xsens motion trackers.

3.2

Review

Human link-segment models have been used in video games, rehabilitation, and biomechanical settings to record and gain understanding about human motion. Link-segment models are then used as input to inverse models using basic physics equations to quantify the forces and moments experienced by the spine, especially when examining lifting (Kingma et al. 2001; Abdoli-Eramaki et al. 2009; Norman et al. 1998; Potvin, McGill, & Norman 1991). To include the inertial components required in a dynamic inverse model, effective motion capture tools are required to measure the motion under investigation. In order to better estimate loading on the lower back the muscular contributions must be taken into account. A relatively new technology has emerged allowing researchers to capture motion in the field without the need for line of sight or bulky systems. Systems, like the Xsens Motion Tracking System (Xsens, The Netherlands), rely on accelerometers, gyroposcopes, and magnetometers and advanced algorithms to produce reliable estimates of body segment position and orientation (Roetenberg, Luinge, & Slycke 2009). Body segment positions, and accelerations can be used as input for inverse models to determine kinetic properties such as the net force and moment. Link-segment models developed using inertial sensors must be built sequentially from proximal to distal (or vice-versa). Thus errors are accumulated as the LSM is built from proximal to distal, and the greatest position errors are seen in the distal segment. The effect of these errors on kinetic and kinematic 2

parameters has been evaluated. Godwin (2009) found wrist position RMS errors of 136 mm, 138 mm, and 101 mm in the anteriorposterior(AP), medial-lateral(ML), and inferior-superior(IS) axes respectively, leading to moment RMS errors of 12%, 17%, and 22%, about the AP, ML, and IS axes respectively, as a percentage of peak moment. Recent improvements in Xsens output have led to improvements in kinematic accuracy. Cutti et al. (2010) reported RMS joint angle errors of 1.4 and 1.8 for the hip and knee angles when compared with a standard goniometer and errors of approximately 2 at the hip and knee. Basic drawing tests, requiring subjects to repeatedly trace the same pattern and then checking for fidelity,have shown the Xsens system to be accurate within 0.5 cm using link-segment modeling over periods of 25 seconds and has been deemed acceptable in a neurorehabilitiation setting (Bai et al. 2011). When investigating gait parameters similar repeatability measure values were found when comparing the Xsens system (Cloete & Scheffer 2010) to Vicon Nexus (Kadaba et al 1989) and Polhemus Liberty (Mills et al 2007). Godwin (2009) notes one limitation of their model was that the glenohumeral joint was modeled as a rigid link between the distal end of the humerus and the spine . However, there can be as much as 150 mm of glenohumeral joint protraction (Albert et al. 1998), suggesting that a rigid connection is invalid . The inability of the shoulder to protract may account for some of the errors put forth by 3

Godwin. Thoracoclavicular range of motion has been estimated at 20 degrees of protraction (elevation), 60 degrees of forward flexion, and 10 degrees of axial rotation (Inman, Saunders & Abbott 1944; van der Helm 1994b; van der Helm & Pronk 1995). Because the clavicle is challenging to instrument, some models simply assume that the shoulder joint maintains a constant orientation and position relative to the upper body segment (Godwin 2009; Cutti et al 2008; Rau, Disselhorst-Klug, & Schmidt 2000; Rab, Petuskey, & Bagley 2002). Thus, the humerus rotates about the shoulder joint centre while the shoulder joint centre does not move at all. Other models have sought to monitor clavicular or scapular motion to track shoulder joint translation (Dickerson, Chaffin, & Hughes 2007; Maurel et al 2010; van der Helm 1994a). Yang et al (2010) have suggested that maintaining a closed loop between scapula, clavicle, and back segments is important to improve accuracy and fidelity in shoulder models. The only research we discovered that attempted to quantify this effect showed that shoulder translation had a significant effect on positions and accelerations of arm segments, but did not have a statistically significant effect on L5/S1 moment prediction (Albert et al. 1998). Once link segment models have been built, researchers can investigate kinetic or kinematic parameters of the model. Newtonian physics, built upon the fundamental equations of motion of a rigid body (Zatsiorsky 2002) are used to determine the mechanical forces on various body segments. There have been reported differences in joint 4

load estimates based on the inclusion of dynamic parameters and asymmetrical lifting (Marras & Granata 1995). The muscular force contribution must also be taken into account to produce accurate low back loading estimates. EMG-assisted models, optimization models and single muscle extensor models can be used to estimate this force contribution. Kinetic parameters are often broken down into their component vectors and expressed in the local coordinate system of the segment of interest. Two commonly investigated parameters are the compressive and shear forces at the L5/S1 joint. For these reasons, it is important to use a fully dynamic three-dimensional model including some measure of muscular force contributions to accurately estimate joint loading for these lifting tasks.

3.3
3.3.1

Link Segment Model


Subject Instrumentation

Subjects were outfitted with Xsens Motion Trackers (MTx) sensors on the thoracic spine and lumbar spine, as well as the left and right scapulae, forearm and lower arm (Figure 1). By aligning the MTx sensors with the long axis of the segments the orientation of these body segments can be tracked during lifting tasks. The segment orientations are used to create a representation of the subject under investigation. 5 6 2 1 3 7 8 5 4

Figure 1 - XSens Motion Tracking System sensors. Sensors are placed on the lower arms (1&2), upper arms (3&4), scapulae (5&6), upper back (7) and lower back (8).

Each sensors orientation was recorded using MT Manager (Xsens, Netherlands) and sampled at 100Hz. All data was exported using MT Manager and imported into Matlab (R2009a, The MathWorks) to create the link segment model. All subsequent data processing was also performed in Matlab, except where noted. 3.3.2 Building the Link Segment Model

The LSM is built sequentially from the pelvis to the hands and is designed to accurately represent the body segments. The orientation of each individual segment is determined by the orientation of the attached sensor, except in the case of the head and neck segment, clavicle segment and spine to sternum projection segment (which are discussed later). Anthropometric measurements of each individual are taken to create segments representing: L5-L1,T12-T1, C7 to ear canal, clavicle, upper arm, and lower arm. Back segments are measured by palpating the spinous processes and counting up and down from C7 (the spinous process that protrudes the most in forward flexion). The clavicle is measured from the top of the sternum to the acromion process on the clavicle by palpation. The upper arm is measured from the lateral edge of the acromion process to the lateral epicondyle of the humerus. The lower arm is measured from the estimated joint centre of the elbow to 2

the second knuckle. This helps creates a LSM that represents each individual subject participating in the study. The pelvis, the fixed and unmoving base of the model, is the lowest part and as this is a hands-down model the lower limbs are not included. Attached to the top of the pelvis segment is the lower back segment. Its motion is based on the orientation of the lumbar sensor and its length, like all segments in the model, is based on subject specific measurements. The upper back segment is connected to the top of the lower back segment with its motion based on the output of the upper back sensor. The head and neck segment is assumed to maintain the same orientation as the upper back segment and is projected up from the upper back endpoint. From the top of the upper back segment, a rigid segment is projected forward towards the sternum and represents the subjects trunk depth at this point. This virtual segment helps link the axial skeleton and the humerus. Next, two segments are created, a virtual clavicle segment that runs from the sternum to the shoulder joint and a scapula segment, based on the orientation of the sensor positioned above the scapula, that runs from the top of the upper trunk segment to the shoulder. If we assume that the clavicle, with a fixed length, can pivot freely about the sternum, it will describe a sphere of possible clavicle endpoints. The orientation of the scapula sensor determines the line that pierces the clavicle endpoint sphere. The point of intersection between the line and the sphere is taken as the shoulder joint. 2

Figure 2 - How the orientation of the scapula sensor is used to determine the position of the clavicular endpoint.

Models that have a rigid link between the spine and the shoulder, as has been done, cannot represent shoulder joint translation, as would be seen in reaching forward or reaching overhead. Our model more validly represents these motions. As this model can incorporate asymmetric motions, this procedure is repeated for the other shoulder. On each side of the body, upper arm segments are attached to the shoulder (clavical segment endpoints) and then lower arms (forearms) are attached in turn. The hands are assumed to be rigidly attached to the forearms. All required segment lengths are taken from anatomical measurements of each individual subject. 3.3.3 Model Optimization

This model is optimized in a two-step process. The first step concerns the length of the virtual segment connecting the spine and the sternum (spine-sternum length), which is assumed to be rigid. During an optimization trial, the subject holds a solid object (a piece of wood) 3

and moves their arms around in a series of dynamic pushes, pulls, twists and swings, while maintaining a grip on the object. Thus, the distance between the lower arm endpoints (hands) is maintained throughout the optimization trial. The distance between the hands is calculated for each frame and subtracted from the true distance (the real length of the object). The root mean squared (RMS) error is then calculated. The spine-sternum length is increased and then this calculation is repeated. The spine-sternum length was increased from 0 cm to the measured chest depth of the subject (usually about 1523cm) in 1 cm increments. At the end of this procedure we have calculated RMS errors for each spine-sternum length for a single trial. The calculated distance between hands is shown for 6 different spinesternum lengths in Figure 3. It should be noted that this process improves accuracy when the predicted values are far from the real values, and improves precision when the predicted values vary around the real values.

Figure 3 - The effect of different spine to sternum projection lengths on the calculated 3D distance between lower arm endpoints during a trial where hand-to-hand distance was maintained. The dashed black line represents the true hand-to-hand distance.

For each spine-sternum length, the RMS difference for the entire trial is calculated and the trial with the lowest RMS value is selected as the optimized spine-sternum length. This change is then incorporated into the model as this provides the greatest reduction in segment endpoint

prediction errors. During a series of optimization trials, this length was found to be between 12 and 16 cm, and is in keeping with those values found by measuring that same distance in MRI images from the Visual Human Dataset (US National Library of Medecine, 2011). The RMS difference for each spine-sternum length is presented in Figure 4.

Figure 4 - Finding the spine to sternum projection distance that minimizes error in lower arm endpoint distance calculations.

A second optimization was implemented because of the potential for misalignment between the Xsens sensors and the anatomical coordinate system of the segment to which it is attached. Again, the goal was to reduce the RMS error of the distance between the hands using the same optimization trial used to estimate the spine-sternum length. Using these trials the clavicle segments are rotated in 3 increments from -30 to +30 about their original Z (mediolateral) axis and the RMS error of the 3D distance between hands is calculated for each 3 increment. The segment is then rotated by 1/3 of the angle with the lowest RMS value. For example if the lowest RMS error value is found to exist when the segment is rotated 9, then the segment is rotated 3 in the experimental trials. The same optimization procedure is repeated for the Y-axis (anterioposterior) and again the segment is rotated by 1/3 of the angle with the lowest RMS error value. As the result of rotating each clavicle is calculated separately, not reducing the optimized rotation by 1/3rd would mean that each segment would 2

be rotated to reduce all the error. This would overshoot the necessary rotation and produce link-segment models that would be visibly incorrect. These rotations are applied to the clavicle and then the optimization process is repeated for the upper arms and lower arms in sequence. Thus the reduction by 1/3rd also ensures a degree of optimization sharing across segments. Pilot data showed that 6-9 degrees is usually selected for the clavicles about both axes, and 0-2 degrees for the lower arms. Figure 5 demonstrates this improvement for a trial where the subject maintained a constant distance of 65cm between the hands. The range of measurement error is reduced from 17cm to 15cm and the mean error is reduced from 69 cm to 65 cm, the actual length of the object used in the optimization trials.

Figure 5 - Comparison of the raw and optimized projected distance between lower arm endpoints. The actual length of the object is 0.65m and represented by the thick black horizontal line

This optimization procedure is performed twice in two conditions: hands 55 cm and 5 cm apart. The average of the four trials is used as the optimization value. As this optimization process, like the spine-sternum optimization process, reduces RMS error relative to the real hand-to-hand distance, the process improves accuracy when the predicted values are biased in one direction, and improves precision when the predicted values hover around the real value. The remaining noise in the system may

be due to changes in hand orientation (which was not recorded), sensor drift, and/or the sensors not moving when the segments were (as a result of soft-tissue artifact, or the sensors shifting relative to the clothing underneath).

3.3.4

Determining Loading

The researcher divided every lifting task into distinct phases based on key loading instances. For hand carries these instances are: bagcontact, full-bag-support, bag-leaving-hand, and bag-fully-on-ground; for shoulder carries these instances are: bag-contact, full-bag-support, hand-to-shoulder-transfer, shoulder-to-hand-transfer, bag-leavinghand, and bag-fully-on-ground. These instances define the beginning of increases and decreases in support of the bag to enable a load timing vector to be created for each hand and shoulder. These are called load timing vectors because they change over the course of the trial but are not vectors in the physical sense as they lack direction. These instances are determined by visually inspecting every third frame of an animation of the model in a method similar to watching a video recording. Once important loading instances were observed the animation was stopped and other relevant distance, position, or velocity curves were inspected. Viewing only every third frame was chosen to speed up the process of load timing determination; however, when inspecting distances, positions and velocities all frames surrounding the relevant instance were inspected.

In all trials the participant (and consequently the animation of the model representing the participant) initially reaches forward for the bag. When the researcher determines that the subject has reached furthest forward the animation is paused and the position curve of the lower arm endpoint is inspected around this instance. The instance of maximum reach is determined to be bag-contact. Playback of the animation is resumed. Then the participant picks up the bag and pulls it towards themselves. Again the animation is paused, and the velocities of the relevant hand are inspected. The frame with the highest peak velocity is chosen as full-bag-support. This decision was based on the idea that participants would have full control of the bag when their hand reached its peak velocity after pickup. Additionally, Eger & Stevenson (2004) have shown that peak vertical hand forces occurred between 0.07 and 0.18 seconds after the load has been picked up. Peak vertical hand forces could only occur once the subject is holding the box in their hands and has a high hand velocity. The animation is resumed and in the case of shoulder carries, the animation was again visually inspected up to the point where the hand came close to the shoulder. At this point the 3D distance between the lower arm endpoint (hand) and clavicular endpoint (shoulder) was calculated and the instance of closest proximity was deemed to be the point of hand-to-shoulder-transfer. This same process was repeated for shoulder-to-hand-transfer (offloading). After the load was transferred to the hand the animation was visually inspected for the instance when 3

the participant had a substantial downward reach (putting the bag on the ground). The hand velocities were then inspected for a local maximum to determine the instance when the bag was starting to be released, bag-leaving-hand. The height of the hand was then inspected for the lowest value and this instance was deemed bag-fully-onground. Separate load timing vectors were then created for the hands and shoulders from these load timing instances to create smooth load transitions. From bag-contact to full-bag-support the hand loading vector was linearly increased from 0 to 1, where at 1 the load had been fully transferred to the hand. During sagittal plane box lifts, hand force loading has been shown to increase approximately linearly in instances where the subject does not push down on the box prior to pick up as would be the case for lighter loads (Eger & Stevenson, 2004). In the case of shoulder carries, full-bag-support was maintained in the hand until the load was transferred to the shoulder. The load in the hand is decreased linearly while the load on the shoulder increased linearly during the 3/10ths of a second after the hand-to-shoulder transfer. The transfer length of 3/10ths of a second was chosen during pilot testing because the transfer needed to occur in a relatively short period of time around the hand-to-shoulder transfer instance. Had we elected to transfer the load over a longer period of time we might have observed that some of the load was placed in the hand while it was not near the

shoulder. During the carry phase, the load was either fully in the hand or on the shoulder. The process used to determine the loading vector when picking up the bag was reversed when putting the bag down. For hand carries the load was maintained until bag-leaving-hand and was then decreased linearly until bag-fully-on-ground. For shoulder carries, the load was transferred to the hand at the hand-to-shoulder-transfer and then decreased until bag-fully-on-ground. Specific body segment parameters are then added to the model using subject height, weight, gender, and anthropometric information. Moment of inertia and centre of mass values are determined based on Zatsiorskys anthropometrics tables (Zatsiorsky 2002, pp. 304-305, The Inertial Characteristics of Human Body Segments of 100 Male Subjects). The predicted segment endpoint positions were filtered using a second-order low-pass Butterworth filter with cutoff frequency of 6 Hz (Dickerson, Hughes & Chaffin, 2008). Segment centre of mass (COM) positions were determined as a percentage distance between proximal and distal segment endpoints (Zatsiorsky 2002). Linear COM velocities and accelerations were calculated from the COM displacements, while angular velocities and accelerations were calculated by successive numerical differentiation of the segments angular orientation.

Using the anatomical model and load timing vectors, a threedimensional hands-down, inverse dynamic model calculated the forces and moments at the L5/S1 joint. These calculations are based on Hof (1992). Once the net external moment was computed, a single back extensor model was used to calculate the force required by the spinal extensors to balance the external moment. The required force was calculated by dividing the external moment by a spinal extensor moment arm length of 6 cm (Norman et al. 1998) and the resultant force was used to determine the final forces on the L4/5 disc. For a complete breakdown of the calculations see Appendix A. The entire process to produce compressive and AP shear force estimates is presented below.

Figure 6 - Flow chart representing process by which compressive and AP force estimates are produced. Anthropometric data and optimization trials are used to produce an optimized participant specific link segment model. The participant specific LSM is then used as input to the 3D inverse dynamic model with the data from each lifting trial and the generated load timing vectors. This produced compressive and AP shear force estimates for each trial.

3.4

Methods

A comparison was required against a validated model to ensure that our model could accurately determine the joint loading during the lifting tasks. For this comparison we used the 3-Dimensional Static Strength Prediction Program (3DSSPP) developed by the University of Michigan. 3DSSPP is a software program that uses postural information, relevant anthropometric variables and force input to determine joint forces during a task. The 3DSSPPs output includes: compressive forces and moments at the L5/S1 joint, shoulders, elbows, and hands, as well as reports predicting the percentage of the population that are able to safely complete the task. 3DSSPP is unable to incorporate dynamic components of lifting and is accurate when the rate of lifting is below 3 Hz. 3DSPP has been validated as a strength prediction assessment tool with a correlation greater than 0.85 with actual strength data; however, it is sensitive to postural input errors (Chaffin & Erig 1991; Chaffin 1992). A series of lifting trials was simultaneously recorded using 3DSSPP and our model. These trials consisted of a simple box lift requiring the subject to pick a 10kg box off of a table, touching it to the ground and returning the box to its original position on the table. Two versions of our model are presented, a fully dynamic model that we used for paramedic testing and a 3DSSPP-matched static model. This was necessary since 3DSSPP is a static model and the segment lengths it

uses are based on the entered subject height and cannot be manipulated individually.

3.5

Results

Figure 7 - Presentation of the compressive forces on the L5/S1 joint as a result of a stoop lift as calculated by 3DSSPP, a static-linked segment model designed to match 3DSSPP and a fully dynamic link segment model.

As can be seen in Figure 7 through 10 the 3DSSPP matched link segment model follows the 3DSSPP curve for compressive and anterior-posterior (AP) shear forces. This is particularly critical as these are two of the main outcome measures of this study. One area of obvious disagreement between 3DSSPP and our LSM is in the middle of the trial, when the participant is placing the box on the ground. This is only the case for compressive force estimates as AP shear forces estimates seem to be in close agreement throughout the entire trial. During the middle of the trial the subject is squatting down and touching the box to the ground. At this instance we should see slightly lower compressive force estimates than at pickup as the weight is held closer to the body and the participant is no longer reaching forward. The difficulty in matching the mannequin in 3DSSPP to the observed posture could account for some of this error.

Figure 8 - Presentation of the compressive forces on the L5/S1 joint as a result of a squat lift as calculated by 3DSSPP, a static-linked

segment model designed to match 3DSSPP and a fully dynamic link segment model.

Figure 9 - Anterior-posterior force on the L5/S1 joint as a result of a stoop lift; calculated from 3DSSPP and two link segment models. Positive values represent a tendency to translate anteriorly.

Figure 10 Anterior-posterior force on the L5/S1 joint as a result of a squat lift; calculated from 3DSPP and two link segment models. Positive values represent a tendency to translate anteriorly.

3.6

Discussion

We are pleased with the similarity between the compressive and AP shear force curves from 3DSSPP, the 3DSSPP-matched static link segment model and the dynamic link segment model. One obvious disagreement between 3DSSP and our models occurs near the middle of all trials; this is the instance when the box is closest to the ground. We hypothesize that the difference in this phase of the lift is due to how each model represents the back. 3DSSP has a single segment representing the entire back while our model uses two segments; 3DSSPP therefore does not consider thoracic flexion and as such some information may be lost.

In the case of the stoop lift, the maximum forward flexion measured in our model was approximately 100 of flexion for the lumbar segment and 140 for the thoracic segment, while 3DSSPP measured 100 of flexion for the single trunk segment. For our model, the additional 40 of forward flexion in the upper trunk causes a shorter moment arm thereby reducing the external moment. Therefore, the reaction moment that is required to balance the system is lower and leads to lower force estimates. The differences can be seen in Figure 11, as the participant rounds out their shoulders when touching the box to the ground.

Figure 11 - Three representations of the same position of a trial. On the left the participant touching a 10kg box to the ground as part of a stoop lift. In the centre the same image overlaid with a mannequin from 3DSSPP. On the right the Link Segment Model.

In Figure 11 it is apparent how the differences between the flexion angles come about. The upper back sensor on the participant is tipped very far forward and may be in excess of the actual flexion of the thoracic segment. The 3DSSPP model has no forward flexion at this 3

point and, as a result, the mannequin is closer to 100 of flexion requiring a greater muscular effort to balance the system than the 140 seen in the link-segment model. It would seem that the flexion of the thoracic segment that 3DSSPP lacks may be driving some of the differences between the models. In order to understand how differences in back representations would influence paramedic research we calculated the average flexion angle at the peak loading instance for all paramedic trials. At peak loading no participant was flexed more than 80 so we compared 3DSSP and our LSM during a stoop and squat lift when both models were at 80. Most participants average peak flexion was around 60 so we compared our model against 3DSSPP at this angle as well.

Figure 12 - Compressive and AP shear force LSM estimates of a stoop and squat lift at 60, 80, and 100 degrees of upper back flexion represented as a percentage of those values found using 3DSSPP.

As trunk flexion increases past 60 degrees our dynamic LSM force estimates and 3DSSPP force estimates begin to diverge. As flexion reaches 100 degrees this difference reaches almost a 1000N discrepancy. However, we believe that the true compressive force value lies somewhere between our value and 3DSSPPs at this angle as trunk flexion, especially thoracic flexion, appears decreased in the 3DSSPP model but increased in ours as a result of the shoulders rounding out. Given that most paramedics were not required to go

into high degrees of trunk flexion this is less of an issue for our research than the stoop and squat lift comparisons present. Additionally, the high degree of agreement (greater than 80% for much of the trial) between 3DSSPP and our model shows that we can as validly represent human motion and calculate the resultant forces within the range of 0-60 of forward flexion. It appears that sensor placement can be an issue when it comes estimating low back loading especially in the case of large thoracic flexion such as when touching the ground. If the sensor is placed too low on the thoracic spine then thoracic flexion is underestimated, but if the sensor is placed too high then flexion is overestimated. This may provide a direction for future research to determine the best sensor location for the thoracic segment to achieve best estimates of thoracic flexion. We have shown that our link-segment model is a valid tool to represent human motion. We also believe that our novel shoulder model is more appropriate than static shoulder models in reflecting shoulder joint translation. If we assume that our shoulder model more truly represents what actually happens, and this is then used as input for a dynamic inverse model, then we should expect that more valid loading estimates are the result.

Chapter 4: Paramedic Lifting


4.1 Introduction

The 72 certified land ambulance services in Ontario respond to an estimated 1.5 million calls annually. In responding to these calls paramedics carry their equipment bags over long distances, up and down stairs, and through confined spaces. Given that paramedics can respond to a number of calls per shift it is not surprising that the various types, size, and weights of these bags are a concern. Despite the fact that common equipment is carried in the bags, there are no standards governing the size and number of bags or the weight carried in any particular bag. The Association of Municipal Emergency Medical Services of Ontario (AMEMSO) as part of its responsibilities to its members, wanted to investigate bag lifting as a possible cause for concern regarding low back injury rates. In addition to developing contacts with regional paramedic services, initial meetings were scheduled with AMEMSO President Paul Charbonneau to increase our understanding of the paramedic population based on his opinion of: number of hours worked per week, number of calls per shift, shift timing, general fitness of paramedics, general attitude of paramedics, speed of work, and general complaints made by paramedics regarding their working conditions. These meetings lead to visits to paramedic services in Kingston, Waterloo,

Sudbury-Manitoulin and Greater Sudbury during the month of March to record the range of bag weights, designs and lifting styles. Prior to collecting information, letters of understanding and permission were signed by a representative from each participating service as well as all participating paramedics, as approved by the Queens General Research Ethics Board. Paramedics were asked to fill out a questionnaire as well as answer some questions during an informal interview. During these interviews paramedics were asked which aspects of bag lifting they felt were most demanding, as well as how they felt bag lifting ranked in relation to other demanding tasks associated with the job. Furthermore, some services participated in pilot lifting trials on-site similar to those performed later in the lab. These pilot testing sessions were used to get a sense of the types of motions performed by paramedics as well as learn if the paramedics felt comfortable performing the testing protocol. In total eight subjects participated in pilot testing and 16 participated in interviews or filled out questionnaires. These meetings and pilot testing sessions, along with information gained from a questionnaire sent out to paramedics by Dr. Renee McPhee (personel communication), led to an understanding of the general working demands of paramedics across the province. It was found that:

Paramedics frequently carry bags ranging from 1-20kg, with most bags in the range of 5-12kg.

When responding to calls usually more than one bag is carried. When loading out, bags are often picked up off the stretcher at about waist height, carried to the person in need and then placed on the ground.

When loading in, bags are picked up off the ground then placed on the stretcher or carried by the paramedic.

Heavy bags and bags requiring awkward postures were the biggest complaint, partially due to uneven loading in the bags or carrying too much equipment, some of which is rarely used.

Two common suggestions were received: reducing the amount of weight in the bags and move towards backpack style bags when possible. One participant suggested the use of wheelie bags to reduce the demand on paramedics.

These insights led to the development of an in-lab testing protocol that was designed to reflect the bag lifting demand experienced by paramedics.

4.2

Review

Lifting has been identified as a hazard that can lead to low back pain (Marras et al. 1995; Chaffin & Park 1973). Paramedics share similar lifting task demands with a variety of professions such as nurses,

nursing aides, and fire-fighters who are required to support a patient and administer some level of care while simultaneously transporting the tools necessary to provide that care. Studies of the nursing population have shown high levels of low-back pain much of which is related to patient handling, (Jensen 1987, Videman 1984) a task that paramedics are frequently required to perform with, potentially, greater strain due to awkward postures and lower lift origins. Back injuries are very common in the paramedic population accounting for 36% of total injuries; of those more than half were caused by lifting activities (Hogya & Ellis 1990). Crill & Hostler (2005) surveyed EMS providers and found that almost 20% had reported a back injury while performing EMS work in the previous six months. It is important to understand which aspects of paramedic work lead to low back pain. Measures that have been correlated with increased low back pain include: peak compressive force, and peak shear force experienced by the L4/L5 or L5/S1 joint, (Norman et al. 1998; van Dien & Toussaint 1997). The National Institute for Occupational Safety and Health has identified two limits for compressive force on the lower back based on biomechanical, psychosocial, and physiological research. 3400N has been put forward as an acceptable limit below which most workers should be able to work safely for long periods of time without injury (NIOSH 1981). A maximum permissible limit of 6400N was established beyond which individuals cannot work without injury. Waters et al (1993) argue that if the data were normally distributed 21%-30% of 2

lumbar segments would fail when loaded with a force of 3400N given the ultimate compressive force values put forward by Brinckmann et al (1988) and Jger & Luttman (1989). The limit of 3400N is a conservative estimate for a healthy working population since cadaver lumbar segments may have lower tolerance limits (Hansson & Roos 1981); however, when ensuring workplace safety conservative limits should be used. Shear forces act in the medial-lateral and anterioposterior directions along the spine and are a risk factor for low back pain (Kerr et al. 2001). Krypton et al. (1995) found shear force tolerance limits in cadavers of between 1700N and 2900N. While less work has been done in this area, some guidelines are based on thinking similar to that of the NIOSH Equation. McGill et al. (1998) put forward an action limit of 500N and a maximum permissible limit of 1000N. A shear force limit of 500N has been used with reasonably accuracy to predict which workers reported low back pain (Daynard et al. 2001). Motions that keep compressive forces below 3400N and shear forces below 500N are unlikely to increase the risk of injuries. Clearly, it is important to understand the loads generated by paramedics across a variety of lifting conditions. An understanding of the magnitude of these loads while performing paramedic work is essential to determine if the weights and/or lifting techniques are safe.

4.3
4.3.1

Methods
Participants

Twenty-five participants were recruited from the Kingston paramedic community. All volunteers reported no low back, shoulder, arm, hand and wrist pain in the last year. Informed consent, approved by the Queens University General Research Ethics Board, (Appendix B) was given before testing. Participants were provided a $50 honorarium to compensate for travel and parking. 4.3.2 Instrumentation Prior to testing, subjects were outfitted with the Xsens Motion Tracking System, a wireless motion tracking system with sensors that combine accelerometers, gyroscopes and magnetometers to determine a sensors orientation. Due to hardware difficulties that caused two sensors to be unusable and that no additional sensors were available, the instrumentation setup was altered from that used in model development. It was decided that the scapular sensors were the least important since their range of motion was the smallest. As a result only 6 sensors were used for paramedic testing. Data obtained from pilot testing was reevaluated by reducing the number of sensors used in creating the link-segment model, to test the influence of the novel link-segment model on force outputs during lifting and carrying tasks. Visual inspection showed of force curves showed changes of less than 10% when removing the scapular

sensors. As a result we could assume that our model was still valid when using only 6 sensors. Subjects were instrumented with sensors placed on the lower back, upper back, upper arms, and lower arms. When attached to the aforementioned limbs the motion of the participant can be tracked. Instead of determining clavicular orientation using Xsens sensors, the clavicle segment was projected straight out from the upper back segment based on the direction of the medial-lateral axis of the upper back segment. The clavicle, upper arm, and lower arm segments were all optimized in the same way as explained in Section 3.3.3. 4.3.3 Bags Four EMS bags (provided by the Frontenac Paramedic Service) were acquired and loaded with weights of 5kg, 10kg, 15kg and 20kg. The 5kg and 10kg bags were carried as handbags, the 15kg bag was carried as a shoulder bag, and a backpack was loaded with 20kg. These bags are shown in Figure 13.

Figure 13 - EMS bags used during in-lab testing

4.3.4

Lifting Tasks

Two conditions were developed for the 5kg and 10kg bags, the first condition had the bag placed on the edge of a table as close to the participant as possible while in the second condition the bag was placed approximately 50cm away from the edge of a 78 cm high table, requiring paramedics to reach forward. A height of 78 cm was used as it was between the height of the stretcher and back of a paramedic SUV from which paramedics lift their bags. Two conditions were also developed for the 20kg backpack; the first required the participant to pick up the bag and sling it over one shoulder; the second required the participant to carry the bags like a normal backpack (i.e. using both shoulder straps). In total there were seven lifting conditions (two each for 5kg, 10kg and 20kg bags, one for 15kg bag). For most lifting conditions the participant picked the bag up off of a 78cm high desk, carried it 4m and placed it on the ground. This mimics arriving at the scene, removing the bag from the back of the vehicle or off of a stretcher and carrying it to the site. The exception to this was the double strap shoulder backpack carry where the experimenter helped the subject load the backpack onto their shoulders, which avoided disrupting the sensors placed on the subjects back. Simulating carrying the bags back to the ambulance was not done. A list of these conditions is presented in Table 1.

Bag

Bag Size HxWxD (cm)

Weight

Lift Type

Lift Forward Origin Reach Height Distance 78cm 0cm 50cm 0cm 50cm 0 cm

Lift Destinatio n Floor Floor Floor Floor Floor

(kg) 5 Hand

Yellow ALS Bag Blue OXYGEN Bag Red Duffel Bag Orange Backpac k

23 x 32 x 12

78cm 78cm 78cm

33 x 58 x 24

10

Hand

28 x 52 x 34

15

Single Strap Shoulder Single Strap Shoulder

78cm

78cm N/A

0cm N/A

Floor N/A

55 x35 x20

20

Double Strap

Table 1 - Conditions presented to paramedics during in-lab testing. All conditions included a 4m carry.

4.3.5

Data Collection

Each subject was asked to perform each condition four times, producing 28 trials per subject. Due to a combination of hardware and software problems four subjects performed fewer than 28 trials. In total 21 subjects completed all 28 trials, while all 25 subjects completed a minimum of 22 trials. The participants height, weight, gender and relevant anthropometric measurements were recorded after testing. Participant anthropometrics are listed in Appendix C. Sensor orientation was recorded using MT Manager software (Xsens Technologies, Netherlands). Each trial was recorded and exported individually before analysis.

Figure 14 - Representation of trial setup. Participants were required to lift a bag from a height of 78 cm, at a reach distance of 0 cm or 50cm, and then carry it 4m before placing it on the ground.

After testing, participants were asked to comment on how closely the lifting protocol reflected their day-to-day lifting tasks. Most participants said that the general lifting requirements were accurately represented; however, they felt that the wide variety of lifting demands (i.e. stairs, small doorways, multiple bags) were not present in testing (for all responses see Appendix D). 4.3.6 Data Processing & Statistical Analysis

All trials for all subjects were visually inspected to determine load timing phases and create loading vectors as explained in Chapter 3.3.4. All data processing was performed in Matlab (R2009a, The MathWorks). The output from each trial was time-normalized to create load timing phase consistency across trials and subjects through piecewise normalization. In this process each phase of the trial is allotted a specific percentage of the trial and is linearly interpolated between relevant load timing instances. This process allows subsequent ensemble averaging the phases without any phase shifts. For statistical analysis, relevant data were extracted from curve profiles for each trial. These were: peak compressive force at bag pickup, trunk lean at bag pickup, forward reach at bag pickup, shoulder distance at bag pickup, mean compressive force during load transport, peak compressive force at bag release, trunk lean at bag release, forward reach at bag release, shoulder and distance at bag release. Trunk lean was calculated as the forward rotation of the long axis of 4

the upper back segment. Forward reach was calculated as the horizontal distance between the hand manipulating the bag and the origin of the lower back segment. Shoulder reach was calculated as the horizontal distance between the endpoint of the lower arm manipulating the bag and the endpoint of the clavicle segment on the same side. These measures were averaged by subject across all trials within each condition and then exported for analysis in SPSS (IBM, U.S.A.).

4.4

Results

The average, piecewise, normalized compressive force estimates for all conditions are presented in Figures 15-18. Standard deviations were in the 300-500N range across all conditions. The four hand-carry conditions are presented in Figure 15. For a complete presentation of the mean and standard deviations of compressive and AP shear force curves for each condition across all trials see Appendix E.

Figure 15 Average compressive force at L5/S1 disc during handcarry conditions. MAL is the maximum acceptable limit of 3400N force. Standard deviations were in the range of 300-500N across all conditions.

During a 5kg hand carry, reaching forward increased the peak compressive force by nearly 300 N on average, while during a 10kg hand carry reaching forward increased the peak compressive force by nearly 500 N. During load transport and load leaving the hands similar values were seen regardless of initial reach distance. Increased bag weights led to increased compressive force estimates during shoulder carries (Figure 16). At pickup (load in hands) the 20kg bag increased the compressive force nearly 400N compared with the 15kg bag. During release (load leaving hand) this difference was 300N.

Figure 16 Average compressive force estimates at L5/S1 for shoulder lifts and carry. The maximum acceptable limit is 3400N compressive force. Standard deviations were in the range of 300500N for all conditions.

Peak compressive force for the 20kg backpack during transport was approximately 2000N, which was just slightly higher than the 15kg or 20kg single shoulder carriage during transport. The more asymmetrical body positions seen during single shoulder load transport likely cause uneven loading and, therefore, require greater muscular exertion to maintain upright posture; our model does not account for this additional muscular exertion.

Figure 17 - Average AP Shear force estimates of all 5kg and 10kg hand carries. Positive shear force values represent the tendency to translate anteriorly. The maximum acceptable limit is 500N.

AP shear force estimates showed that reaching forward at pickup increased loading by approximately 80N for both the 5kg and 10kg conditions. Interestingly shear loading was higher at bag release for the 5kg than the 10kg condition. No hand carry condition approached the acceptable limit of 500N.

Figure 18 - Average AP shear force compressive estimates for 15 and 20kg shoulder carry conditions. Positive shear force values represent the tendency to translate anteriorly.

AP shear force curves showed larger estimates for the 20kg than the 15kg condition at pickup as well as throughout much of the trial. Similar AP shear is seen during bag release for both 15kg and 20 kg single shoulder carries. The large negative AP shear values predicted around 20% of the trial may be the result of the bag being swung onto the shoulder and thus for an instance the load (and potentially some of the body) is behind the L5/S1 joint centre, producing a posterior shear. As outlined earlier, double shoulder backpack carrying trials did not include bag pickup or bag release, as a result peaks are not present during the initial and final portions of these curves. Interestingly, during load carriage, higher AP shear forces were seen for the 20kg double shoulder carry than the 20kg single shoulder carry. As before,

no average shear force estimates approached the maximum acceptable limit of 500 N. In fact no subject presented averages higher than 350N of AP shear for any condition. A forward stepwise linear regression (SPSS, IBM, USA) was used to determine the strongest predictors of compressive force at pickup and release, as these points represented the instances of highest loading. For pickup, the strongest predictors were, from most to least importance: bag weight, paramedic body weight, trunk lean angle, and horizontal reach distance from the shoulder (Figure 19). Similar predictors, again in order of decreasing importance, were found at drop off: bag weight, paramedic body weight, trunk lean angle, and reach distance from the hips (Figure 20). All predictors were statistically significant (p< 0.001). While a paramedics body weight cannot be changed, the three other predictors can be manipulated to reduce compressive loading on the lower back.

Figure 19 - The importance of each variable in predicting compressive force at pickup.

At pickup (Figure 19), bag weight is a stronger predictor of compressive force than any other variable. This is not the case at release (Figure 20) as bag weight, subject mass and trunk lean all have similar importance.

Figure 20 - The importance of each variable for predicting compressive force at bag release.

10

See Appendix F for a complete breakdown of the statistical analysis.

4.5

Discussion

Paramedics in Ontario are exposed to a wide variety of loading tasks while performing their job. The physical demands of these tasks are made more challenging by the environment paramedics are required to navigate as well as the cognitive processing required to quickly and effectively help people in need. It is important to pay attention to the health of paramedics to ensure that they perform effectively. As heavy lifting tasks, primarily patient transport, are inherent to the job and cannot be changed, it is necessary to make changes to other aspects to reduce the overall load experienced over the course of the day. This study shows that, as far as equipment bags are concerned, most loads and carrying conditions used by paramedics are currently safe but in some cases are approaching the maximum acceptable limit of 3400N put forward by NIOSH. To reduce the load experienced by paramedics while lifting bags, efforts should be made to: reduce reach distances and trunk lean at pickup and release, decrease bag weight overall, and improve load distribution between bags. Also, loads of 20kg lead to compressive forces approaching acceptable limits. Based on these findings a number of recommendations have been made to keep compressive and AP shear force loading below the maximum acceptable limits:

11

1. Loads below 20kg have been shown to be safe given the lifting tasks investigated here. 2. Prior to lifting, bags should be placed as close to the paramedic as possible. 3. When placing bags on the ground and when picking bags up off of the ground, paramedics should use a squat lift technique to prevent forward and side bending. This research produced results and recommendations similar to what others have found regarding the influence of trunk flexion, symmetry, reach distance and load. Our work as well as the work of Fulmer et al. (2002) has shown that large loads and forward flexion are dictators of low back strain. Research on mail bag design has recommended symmetrical loading across sides of the body as a way to reduce compressive force on the spine, shoulder deviations, and pressure on the shoulder while improving force distribution between the feet (Lin et al. 1996). Waters et al. (1993) recommends reducing forward reach distance and the amount of asymmetry present in lifting tasks due to increasing axial compressive forces and reductions in strength capabilities. Davis & Marras (2003) have put forward that the load weight is the most important factor to control in order to reduce

12

compressive forces at the spine. Our recommendation of 20kg is in keeping with the load limit in the NIOSH equation (Waters et al. 1993). It may be the case that workers are able to safely carry somewhat more than this load; however the degree of twisting motions and the requirement to lower the bag to the ground mean that an ideal situation is not possible for paramedics. It could be dangerous to increase bag loading beyond current measures given that even 13kg loads have been shown to create spinal compressive forces in excess of 5000N (Granata et al. 1999). In regards to load carriage, Straker et al. (1997) found compressive force estimates similar to ours when using self-selected maximums. This provides further evidence that the actions of lifting and lowering rather than carrying should be adjusted to reduce peak loading instances.

Chapter 5: Supplementary Testing


5.1 Introduction
Based on paramedic feedback, a session of supplementary testing was performed with a single participant to gauge the effect of stair climbing as well as multi-bag carries.

5.2
th

Methods

A 50 percentile male was asked to perform 14 different lifting conditions 4 times each; in total 56 trials were performed over the

13

course of 2 hours. A list of the supplementary conditions is presented in Table 2. Conditio n 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) 11) 12) 13) 14) Hand Load (kg) Shoulder Load (kg) 10 15 20 10 15 15 20 25 ------Backpack Load (kg) --------15 15 Over stairs 20 20 Over stairs 25 25 Over stairs Task Descripti on PCD PCD PCD PCD PCD PCD PCD PCD C C C C C C

5 5 5 10 10 ----------

Table 2 - Tasks performed during supplementary testing. PCD = pickup, carry, and drop off. C = Carry

As before, data were recorded and exported using MT Manager and subsequently processed using Matlab. Visual inspection was used to create load timing phases and load timing vectors for each trial. Trials were piecewise normalized and then averaged within each condition.

14

Statistical analysis was not performed because only 1 participant was included in this section of testing.

5.3

Results

Four sets of comparisons can be made when analyzing multi-bag carries: the effect of increasing hand load while maintaining shoulder load, the effect of increasing shoulder load while maintaining hand load, the effect of varying load distribution while maintaining overall load, and the difference between carrying bags across stairs and carrying bags across flat ground. Regardless of load distribution, the compressive forces generated by a 25kg load approached or surpassed the maximum acceptable limit of 3400 N at bag release (Figure 21).

15

Figure 21 - Compressive force estimates of all 25 kg caries. MAL is the maximum acceptable limit of 3400 N of compressive force. Green curves represent 25kg shoulder carries (S25). Red curves represent a 15kg shoulder carry with a 10kg hand carry (S15H10). Blue curves represent a 20kg shoulder carry with a 5kg hand carry (S20H5). Figure 22 - Compressive force estimates across loading conditions, and grouped into categories based on total load. Green bars represent paramedic data, blue bars represent supplementary data. Error bars represent (+/-) 1 standard deviation for paramedic trials.

Compressive force estimates increased with higher loads regardless of distribution. Within loading conditions (i.e. all lifts with 25kg) lower compressive forces were seen when the load was distributed across multiple bags.

No noticeable differences were seen between stair climbing and normal walking across conditions. Mean and peak estimates increased

16

with increased load. Average peak and average mean values for each condition are presented in Figure 23.

Figure 23 - Peak and mean compressive force estimates comparing normal and stair climbing while carrying backpacks with varying loads. Error bars represent 1(+/-) standard deviation.

5.4

Discussion

Supplementary testing showed that regardless of load distribution a 25 kg load presented compressive force estimates approaching the NIOSH limit of 3400 N. Stair climbing and normal walking produced similar loading estimates across all conditions, indicating that these conditions present similar demands to paramedics. Given the low sample size (n=1) more work should be done to understand what differences, if any exist between stair climbing and flat-ground walking. Across all loads used, the lowest force estimates were seen when the load was most evenly distributed across bags. Whether this trend continues beyond more than 2 bags remains to be seen. The main reason for this reduction in compressive force may be primarily due to the reduction in weight of the shoulder bag as shoulder bags in general require more awkward postures to don and doff.Chapter

6:

Conclusions, Recommendations and Limitations


This research has shown that the Xsens Motion Tracking system is a valid tool for representing human motion. It has also been shown that representations of this motion can be improved via an optimization 17

procedure to reduce position errors. This link-segment model can then be used as a framework for a dynamic, inverse model to calculate compressive and shear forces for one-handed and two-handed lifting tasks. Paramedic testing revealed that reaching forward to pick up a bag can increase compressive force estimates even for light loads. Heavier loads lead to high compressive force estimates but not necessarily higher AP force estimates. Finally, loads at or below 20kg did not approach the maximum acceptable shear or compressive limits for the type of bag lifts paramedics perform. It seems unlikely that bag lifting is the primary cause of the high injury rates seen in paramedic populations. With that in mind, statistical analysis revealed that trunk lean angles were correlated with higher compressive force estimates. This indicates that injured paramedics or those want to reduce the chance of back injury should get as close to the bag as possible when picking it up, and use their legs rather than their back to place bags on the ground. Supplementary testing revealed that loading conditions beyond 25kg lead to compressive force estimates that approach the maximum acceptable limits put forward by NIOSH. Additionally, when load was more evenly distributed between a hand and shoulder bags load estimates were less than when the load was localized on one shoulder. With these conclusions in mind we have made several recommendations: 18

1. Paramedics should not lift single bags or a combination of bags that exceed 20kg. 2. Prior to lifting, bags should be placed as close to the paramedic as possible. 3. When placing bags on the ground and when picking bags up off of the ground, paramedics should use a squat lift technique to prevent forward and side bending. 4. When multiple bags are carried the load should be evenly distributed within bags and across sides of the body.

During this research several limitations were identified. First, during all motions it was assumed that the sensors accurately tracked the motion of the segment to which they were attached. In reality some sensors drifted, usually only by a few degrees, but in one trial a sensor drifted 10 degrees. In order to control this drift each subject was required to assume a standardized posture prior to every trial in order to remove any drift that accumulated during or after the previous trial. This drift would influence force estimates but it is difficult to know if it would bias for overestimates or underestimates. Secondly, due to hardware and software complications some conditions could not be collected. This only influenced 4 subjects but meant that the researchers had 25 complete datasets for some lifts but only 21 for others.

19

Third, the testing sessions could not accurately represent all of the bag lifting demands that a paramedic would experience in a normal shift. This is because paramedics experience such a wide range of lifting demands as a result of the varied environments they enter. Many paramedics mentioned this in their exit surveys. However, it would have been difficult to represent such a wide array of environments in a single testing session. Future directions involving improvements to the link-segment model should focus on evaluating the influence of our shoulder model on wrist end-point estimates, as well as its influence on kinetic parameters. Another avenue that should be explored more thoroughly is the influence of load distribution across bags. As found in supplementary testing, distributing the load across bags lead to lower force estimates than when the bag was localized. If additional testing shows this to be a consistent trend then it should provide some interesting bag combination possibilities in the future. Finally, it is expected that the largest loading instances for paramedics occur during patient handling activities. As such interventions should be tested to reduce the either the amount of loading or the frequency at which these activities are performed.

References Appendix A Inverse Dynamic Model Equations


The dynamic inverse model used in this research is built upon the fundamental Newtonian physics equations. For each segment, the

20

sum of all the forces acting on that segment is equal to the mass of that segment multiplied by the net acceleration of that segment.

F = m a

Co M

For each segment, the sum of all the moments acting on that segment is equal to the moment of inertia multiplied by the derivative of the angular momentum.

M = I

For the L5/S1 segment a single extensor muscle model is applied that counterbalances the resulting moment. This force contributes to the net compressive force estimates only. An example of this entire calculation is presented below. Inertial components about each segments centre of mass are not included. As very low segment rotational accelerations are expected, the contribution of the missing terms is assumed to be minimal.

F
M

L 5 / S1

= ( Mas s B a c * a Co M) + (Mas sL b a c * g) + F (Ub a c , He d ,Up p e r s L k k k a A rm

,L o we r s ) A rm

L 5 / S1

= ( I L b a c* L B a c) [ (MassL B a c *g)x CoM L b a c] [(MassL b a c* a Co ML b a c)x CoM L Ba ] + M (Ub a c,He d ,Up p e r s k k k k k k k c k a A rm

,L o we r s ) A rm

M ex t ens or= M L 5 / S1
Fe x t e ns or M e x t e ns or 0.0 6 =

Fc o mp r e s s i v= FL 5 / S1 + Fe x t e ns o r e In order to counterbalance the moment at L5/S1 the back musculature exerts a force at a distance of 6 cm from the L5/S1 joint. The compressive force is then the sum of the force at L5/S1 and the extensor force. Several models have used models with extensor moment arms acting a distance of 5-7cm from the L5/S1 joint (McGill & Norman 1987; Dumas et al 1991; Nemeth & Ohlsen 1986; van Dien & de Looze 1999).

Legend
F: force M: moment a: linear acceleration : angular acceleration : angular velocity

21

CoM: centre of mass

g: force of gravity

I: moment of inertia

Appendix B Information & Consent Form


Ethics Consent for a Research Study entitled:

SCHOOL OF KINESIOLOGY AND HEALTH STUDIES (SK Physical Education Centre Kingston, Ontario, Canada K7L 3N6

Assessment of Emergency Medical Services Bag Lifting and Carrying Techniques


Dear Participant, You are invited to participate in a research study to examine the forces resulting from lifting and carrying bags used by paramedics in the Emergency Medical Services (EMS) industry. This Ethics Consent letter will present key information so you can read about the study and decide whether you wish to become a participant. If you wish to participate, you will be asked to sign this letter and leave a copy of the last page with the researcher. Testing will take place on Queens campus. This study has been granted clearance according to the recommended principles of Canadian ethics guidelines and Queens principles. Purpose of the Study: The purpose of this study is to gain understanding of the forces on the L5/S1 joint resulting from a) lifting, and b) carrying bags used by EMS workers. In addition, we would like your opinion of task discomfort and difficulty under each condition. The study will take about 1 hour and will take place in the Biomechanics Lab at 28 Division St on Queens University campus. Methodology: Personal Information: For your personal safety, we wish to exclude you from the study if you are currently suffering from shoulder, wrist, arm or low back pain. We also want to make sure there are no additional health risks associated with your participation. Therefore, we will ask you to complete the Physical Activity Readiness Questionnaire (PAR-Q). Getting Ready for Testing: Please change into gym clothes for the testing. We will need to measure your body weight and height so that we can determine your anthropometric variables for our calculations. The familiarization task will require dynamic lifting; therefore, we would like you to complete a series of stretching exercises for the back, shoulder girdle and legs. This warm-up will require approximately 5 minutes. Then, we want you to practice each task for 5 minutes so you are aware of what to expect during testing. Testing Protocol:

22

In a single testing session, you will complete the same protocol; lifting and carrying one bag 4 times, and repeating for each of the 5 different bags. The order of bags will be randomized. The bag-lifting and carrying task will begin at erect standing until the investigator requests the lift. Then, bend down or forward in a free style manner to pick up the bag and place it on your shoulder (shoulder bag), back (backpack), or hold it in your hand and return to an erect standing posture. Then please walk as normally as possible across the 5m area outlined by the investigator, this should take less than 5 seconds. Once you have crossed the walkway you will be asked to return to the starting point. This task will be repeated 4 times Subjective feedback: After each task completion, you will be asked to complete a brief questionnaire regarding task difficulty, comfort, and ease of use of the bag. This information will help understand the subjective strengths and weaknesses of each bag. Risks and Benefits of Participation: Research has shown that manual materials handling tasks have high risks for musculoskeletal injuries, especially low back pain. To protect you as much as possible from this risk, we will encourage you to stop at any point if you are experiencing more pain or discomfort than you would anticipate with this task. If you feel extended soreness or pain after completion of your participation, please go to a medical centre for assistance or contact your preferred health care professional. In terms of the benefits of the study, there are no direct personal benefits expected. However, your feedback will contribute to improvements in the bag selection process for paramedics, and potentially other workers in load carriage industries.

Confidentiality:
All information obtained during the course of this study is strictly confidential and your anonymity will be protected at all times. Your identity is only recorded on your signed consent form. Thereafter, you will be assigned a study number that will link your information to this file. We will lock all paper files and computer files so that your identity and feedback is safe and only the principal researcher and supervisor will be granted access. No one in the service will be given your individual information for any aspect of the study. In all cases of publication, only summary data from this will be used so that no individual can be identified. Upon completion of this research project all identifying documents and information will be destroyed. It is likely that individuals in the environment will know that you are participating in this study, however your individual responses will be protected, and as mentioned before only summary data will be presented in publication. We are interested in collecting video and digital photographs for use in this research, as well as for future presentations and publications. In these cases your face will be blurred to protect your anonymity as much as possible. If you are willing to have your photograph taken, we will ask you

23

to sign the section at the bottom of this form pertaining to this information.
Compensation: Each participant will receive $50.00 as compensation for your time. In the event that you choose not to continue you will still receive full compensation. Voluntary Nature of the Study: As a participant, you are a volunteer who may withdraw from the study at any time without coercion or penalty. You may withdraw after hearing about the details of the study or you may also withdraw at any point during the study with no penalty. Any data collected up to that point would be destroyed. Contacts: If at any time you have further questions, problems or adverse events, you can contact: Peter Galbraith (School of Kinesiology and Health Studies) (613)-533-6000 x79019 Dr. Pat Costigan (School of Kinesiology and Health Studies) (613) 5336000 x79037 If you have any questions regarding your rights as a research participant, you can contact: Dr. Joan Stevenson (General Research Ethics Board, Chair)(613) 533-6081 What Does My Signature Mean? I am being asked to sign below. My signature indicates that: I have read the letter of information I was given an Ethics consent letter to read and keep I am aware that the purpose of the study is to assess compressive forces resulting from EMS bag lifting and carrying I realize I can withdraw at any time without penalty or coercion I can contact any of the people identified in the Ethics Consent letter if I have questions, concerns, or complaints I realize that my data will be kept confidential. By signing this consent form, I do not waive my legal rights nor release the investigator(s) and sponsors from their legal and professional responsibilities.

Participants Copy
for your records).

(Please sign and keep this Ethics Consent letter

_____________________________________ Signature of Participant _____________________________________ Witness

__________________ Date ___________________ Date

24

We would like to take photographs or videos of two or three participants. Your signature below indicates that you are willing to allow these photos or videos to be taken for scientific purposes only. In these cases your face will be blurred out to protect your anonymity as much as possible. ____________________________________ Signature of Participant

25

Researchers Copy Ethics Consent

Assesment of Emergency Medical Services Bag Lifting and Carrying Techniques


This page is for the researchers to verify that you are willing to participate in the above study.
I am being asked to sign below. My signature indicates that: I was given a verbal presentation of the above-mentioned research study I have read the letter of information I was given an Ethics consent letter to read and keep I am aware that the purpose of the study is to assess compressive forces resulting from EMS bag lifting and carrying I realize I can withdraw at any time without penalty or coercion I can contact any of the people identified in the Ethics Consent letter if I have questions, concerns, or complaints I realize that my data will be kept confidential. By signing this consent form, I do not waive my legal rights nor release the investigator(s) and sponsors from their legal and professional responsibilities.

(Please sign and return this page ONLY to the researchers) _____________________________________ Print your Name _____________________________________ Signature of Participant _____________________________________ Witness ___________________ Date __________________ Date

We would like to take photographs or videos of two or three participants. Your signature below indicates that you are willing to allow these photos or videos to be taken for scientific purposes only. In these cases your face will be blurred to protect your anonymity.

____________________________________ Signature of Participant

26

Appendix C Paramedic Anthropometrics


These are the anthropometric measurements taken during testing. Age, and the number of years experience as a paramedic were questions asked via email after testing, 2 participants did not respond so these values are based on 23 subjects while the rest are based on 25. Measurement Age Years as a Paramedic Mass Height Lower back segment length Upper back segment length Chest Depth Clavicle segment length Upper arm segment length Lower Arm segment length Mean 32.6 years 7.7 years 82.0 kg 1.73 m 0.16 m 0.33 m 0.17 m 0.19 m 0.32 m 0.34 m Standard Deviation 10.8 years 7.2 years 14.0 kg 0.09 m 0.03 m 0.03 m 0.02 m 0.02 m 0.02 m 0.02 m

Table 3 - Anthropometrics of the participating paramedics.

Appendix D Paramedic Responses to Exit Questions


These are the responses paramedics provided in an email sent out after testing when asked to comment on how accurately the in-lab testing reflected bag-lifting and carrying they do on the job.

27

Participant 1: With regards to how accurate the testing was, I believe it was a good indication on what we do. Most of the time our equipment is on our stretcher, our stretcher is then lowered to a loading height (usually around waist), then we take our equipment off, and when placing them on the floor we would probably bend at hips and knees. Participant 2: The in-lab testing lifts were very realistic lifts that may occur on the job every day, aside from the far reach lift. That particular lift felt awkward to me as I have always practiced safe lifting in order to prolong my career as a medic. Participant 3: There are often multiple lifts of bags and carries of longer duration which may have an effect on stress. Participant 4: I would say the in lab testing did reflect a good amount of accuracy of our bag lifting and carrying on the job. Something that could also be looked at is carrying the bags up and down stairs to reflect peoples homes and apartment buildings with no elevator. As well you could look at the lifting inside of the ambulance, where some people cannot stand up straight in the truck (like myself) and we have to lift the bags off the stretcher hunched over. Otherwise it represents the lifting of bags on and off the stretcher pretty well. Participant 6: The in-lab testing was not realistic to the dynamic conditions of actual bag placement planning and how I lift is not as consistent as a 'perfect lift' might be, depending on those conditions. Also, Paramedics tend to carry more than one bag at once. Participant 8: The weights felt similar to what we actually carry, however, the distances were only a fraction of what we would do on an average call. Testing involving stairs would have been interesting to include, as well, since most calls include at least a few steps. Participant 9: The accuracy of the lift test that you did was fairly accurate in my opinion. The only thing was that, in 99% of cases we aren't lifting our bags off a table or a surface of waist height. We are lifting our bags from the ground up. Other than that I believe it was quite accurate. Participant 11: I thought it was very accurate compare to actual job Participant 12: I think the study reflected our lifting of our bags on the job accurately. It would be nice to somehow also study the affect of late night shifts/tiredness and then repeated bag lifts along with stretcher/patient lifts and what affects that has on our bag lifting abilities. Or, somehow design better, more ergonomic bags. Participant 13: I think for the most part the bag lifting study was pretty accurate, although the shoulder straps I think are rarely used. 28

For the orange first response bag I would use the shoulder straps, but for the monitor and airway bag mostly hand carried, although distance in carrying them would be a factor. Participant 14: I found that the study was quite accurate. The bags were the same as what we carry and pretty true to the weights. When going into a scene it is easy to grab the equipment in a proper manner and distribute the equipment between you and your partner. I feel like the most strain has been on me in times of having to leave a scene in a hurry and having to grab all the equipment tested at one time plus a cardiac monitor, often in an awkward manner. Obviously this isn't what is being tested here but I felt like it is important to bring up. Participant 15: Lifting of bags is not an issue when compared to lifting a heavy patient. I have never seen an occasion where a backpack was used. Participant 16: I personally don't often lift bags or place bags onto a table, I almost always set them on the floor. The only time I lift them from any sort of height is normally off of the stretcher after we have lowered it to about the height of your quads. I've only worked on the first response unit a hand full of times and for no more than a drive to the hospital so I don't have the same experience lifting bags in and out of the back of the truck. I think if I worked on it more often, I could better relate to the study. Participant 17: The lifting was pretty accurate. I'm on the Paramedic Response Unit so I lift a few more bags at one time. Usually I carry the backpack bag in my right hand by the handle and carry the defib and small O2 cylinder in my left hand (if I'm going short distances and know in advance I don't need the big O2 bag). If I carry the larger blue O2 kit, I put the backpack on both shoulders and carry the defib in one hand and the blue O2 bag in the other. Participant 18: I think you guys did a great job on the bag testing, and it definitely did reflect our everyday duties. Although, I don't know how much more it affects our bodies when we carry them for much longer distances (in the residence, through narrow doors, up stairs then back out again), if that in fact would make any difference at all. Participant 19: The pre-hospital environment is very hard to duplicate in the lab as it is constantly changing with multiple variables. We hardly ever work in an open room with lots of space and a flat, smooth floor. For example, we may have lots of space but it is on the side of a hill - that is wet. There may be a flat smooth floor but it is filled with stuff or is a narrow hallway. All of these variables affect how we manoeuvre our bags and equipment around.

29

Participant 20: Bags felt similar in weight, but, the contents were not distributed the same (all of the pockets are stuffed full, making work bags are bulkier, but, easier to lift). Participant 22: I felt that the test fairly accurately reflected our daily lifting. Participant 24: I felt most of it reflected the bag lifting and carrying we do. Participant 25: The bag lifting was close to real - we usually have to lift from a slightly higher surface, with more resistance - the stretcher with a sheet on it. The drag factor there is noticeable. Also, our lifts tend to be more dynamic - due to space (we usually work in peoples homes), furniture, hallways, etc., we are often reaching from different angles and positions, so the lifts you're studying fairly accurately reflect an open space, but don't do justice to a number of the locations we find ourselves in.

Appendix E - Mean +- Standard Deviation Graphs by Condition

30

Figure 24 - Compressive force estimates for all paramedic handbag trials. Solid lines represent average compresssive force while blue bands represent +1 and -1 standard deviations. Dashed lines are used for comparison between graphs.

31

Figure 25 - Compressive force estimates for all 15kg paramedic shoulder lift and carry trials. The solid line represents average compressive force while blue bands represent +1 and -1 standard deviations.

32

Figure 26 Compressive force estimates for all 20kg paramedic backpack lift and carry trials. The solid line represents average compressive force while blue bands represent +1 and -1 standard deviations. The dashed line represents the MAL of 3400N.

33

Figure 27 - Compressive force estimates for all 20kg backpack carry trials. The solid line represents average compressive force while blue bands represent +1 and -1 standard deviations.

Figure 28 AP Shear force estimates for all paramedic hand carries. Solid lines represent mean AP shear force. Blue bands represent +1 and -1 standard deviations.

34

Figure 29 AP shear force estimates for 15kg shoulder carries. Solid lines represent the mean AP shear force. Blue bands represent +1 and -1 standard deviations.

Figure 30 - AP shear force estimates for single shoulder back pack lift and carries. Solid lines represent the mean. Blue bands represent +1 and -1 standard deviations.

35

Figure 31 - AP shear force estimates of 20kg double shoulder backpack carries. Solid lines represent the mean. Blue bands represent +1 and -1 standard deviations.

Appendix F Statistical Analysis

Table 4 - Coefficients, standard error, and significance of variables used as input into a linear model used to predict compressive force at bag pickup. (Taken from SPSS)

36

Table 5 - R values and standard error of linear regression model to predict compressive force at bag pickup. (Taken from SPSS)

Compressive force (N) at bag pickup


Figure 32 - Predicted regression standardized value plotted against real compressive force values at bag pickup. (Taken from SPSS)

37

Compressive force (N) at bag pickup


Figure 33 - Standardized residual plotted against real compressive force values at bag pickup. (Taken from SPSS)

Table 6 - Coefficients and significance of variables used in linear regression to predict compressive force at bag release.

38

Table 7 - R values and standard error of the estimate of the regression equation used to predicted compressive force at bag release.

Compressive force (N) at bag release


Figure 34 - Predicted regression standardized value plotted against real compressive force values at bag release.

39

Compressive force (N) at bag release


Figure 35 - Standardized residual plotted against real compressive force values at bag release.

40

Вам также может понравиться