Effect of handrail height on balance recovery reactions in younger and older adults

Transcription

1 Effect of handrail height on balance recovery reactions in younger and older adults by Victoria (Vicki) Komisar A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Institute of Biomaterials and Biomedical Engineering University of Toronto Copyright by Victoria Komisar, 2018

2 Effect of handrail height on balance recovery reactions in younger and older adults Abstract Vicki Komisar Doctor of Philosophy Institute of Biomaterials and Biomedical Engineering University of Toronto 2018 Handrails can significantly reduce the risk of a fall, provided that their design enables fast, accurate and effective grasping reactions for balance recovery. This thesis characterizes the influence of handrail height on balance recovery reactions in younger and older adults. Common strategies for balance recovery and factors that are likely to influence a person s ability to use a handrail for balance recovery are identified through a literature review. Lab-based studies involving perturbations of upright stance and during ongoing gait (level-walking and slope descent) are then used to explore the relationship between handrail height and balance recovery. Following perturbations of upright stance, higher handrails (within the range tested) increased the highest perturbation magnitude that younger adults could withstand without stepping or falling, while using the rail for balance recovery. Higher rails also resulted in reductions to peak center of mass and trunk angular displacement, velocity and momentum during balance recovery. The concomitant peak vertical and resultant handrail forces, and resultant handrail impulse, were reduced as rail height increased. ii

3 Handrail height significantly affected handrail contact and movement time during slope descent, with movement time optimized for older adults around 64% of participant height (roughly elbow height). In general, increased handrail height also resulted in increased hand speeds, and reduced vertical handrail overshoot. For both slope descent and level gait, increased rail height led to reduced peak trunk displacements and velocities, and applied handrail forces. However, the peak thoracohumeral elevation angles required for recovery also increased with handrail height, clearly surpassing the range of motion of individuals with severe osteoarthritis in the shoulder. My results indicate that handrails installed at 56% to 66% of individual height (approximately midforearm- to elbow-height) may be advantageous for balance recovery, particularly in home environments where handrail installation height can be customized to the user. iii

4 Acknowledgments I am indebted to everyone on the Home & Community and Mobility Teams at Toronto Rehab past and present for the friendship and support over the course of my PhD. In particular, I am grateful to my advisors Drs Alison Novak and Brian Maki for their unwavering support, valuable insights, and extreme patience in guiding me through the development of this work. To my committee members Drs Karl Zabjek and Milos Popovic thank you for the mentorship, guidance and questions that were critical for advancing the development of my thesis. Finally, I am grateful to Dr Geoff Fernie, for giving me the opportunity to contribute to the field of falls prevention not simply through thesis studies, but also through product development opportunities that matter. My thesis studies would not have been possible without the contributions of many. I am grateful to Konika Nirmalanathan and Dr Emily King for their support at all stages of Chapter 3 of this thesis. Susan Gorski, Dr Bruce Haycock, Roger Montgomery, Gary Evans, Dan Smyth, Larry Crichlow, Steve Pong and Peter Goshulak were instrumental in making it physically-possible to run my thesis studies. Aaron Marquis, David Houston, Bella Boyaninska, Mazhar Jabakhanji, Sining Qin, Angela Chen, Angela Lam, Ivan Martinovic and Jane Wheeler were invaluable in data collection and processing. I am also grateful to Jiaqi Huang and Alex Florea whose work was beyond the scope of this thesis, but who made valuable contributions to our team as my summer students in previous years. Finally, infinite thanks to the many of you who generously agreed to be on-call for supporting collections and addressing last-minute crises of all sorts with my thesis work: Philippa Gosine, Dr Carolyn Duncan, Dr Tilak Dutta, Carmen Baker, Jenny Bautista, Rebecca Greene, Dr Dan Vena, Dr Azadeh Yadollahi, Shoshana Teitelman, Lois Ward, Pam Holliday, Keara Maguire, and Alanna Komisar. I am indebted to everyone who volunteered to participate in my thesis studies. Thank you for your time and commitment, particularly for protocols that were not always easy to complete. I would also like to acknowledge project funding from the Canadian Institutes of Health Research Operating Grant (CIHR MOP ) and the Ontario Research Excellence Fund (ORF-RE4-023). iv

5 Further, I am grateful to many sources for providing me with financial support at various stages of my PhD, including scholarships from: AGE-WELL Network of Centres of Excellence Graduate and Postdoctoral Awards in Technology and Aging; Ontario Graduate Scholarships; OSOTF- Toronto Rehabilitation Institute Graduate Student Scholarships; OSOTF-Lipton/Unilever Graduate Scholarships in Neurosciences Research; Ontario Centres of Excellence TalentEdge Awards (partially funded through Andrew J. Hart Enterprises); NSERC CREATE:CARE Fellowships in Rehabilitation Engineering; University of Toronto s Institute of Biomaterials and Biomedical Engineering and Faculty of Applied Science and Engineering; and the St. George s Society, Toronto. I am also grateful to the AGE-WELL Network of Centres of Excellence in Technology and Aging, the International Society of Biomechanics, University of Toronto s School of Graduate Studies, and University of Toronto s Institute of Biomaterials and Biomedical Engineering, for providing me with travel and conference awards to share my findings with others. Finally, to my family and friends who have supported me throughout my PhD: thank you, tremendously, for everything. This work would not have been possible without you, and I am so thankful to you for it. v

6 vi Table of Contents Acknowledgments... iv Table of Contents... vi List of Tables... x List of Figures... xi List of Appendices... xiv Chapter 1: Introduction Motivation Objectives and Scope... 2 Chapter 2: Review of the Literature Overview of strategies for balance recovery Compensatory leg and trunk activity: stiffening and stepping Compensatory upper-limb activity: movement, light contact and grasping Factors that are likely to influence the success of balance recovery via compensatory grasping Initial conditions value of proactive handrail contact Speed and appropriateness of balance response initiation Speed and accuracy of the reaching trajectory Grasp completion Sufficiency of forces and moments applied to the handrail to stabilize the center of mass Synthesis Chapter 3: Study 1: Use of handrails for balance recovery Characterizing loading profiles in younger adults and effects of handrail height Preface Abstract Introduction Materials and Methods Participants Testing environment Protocol Data processing Statistical analysis Results Descriptive statistics of maximum withstood perturbations (MWPs) Characteristic handrail force profiles and peak handrail forces Influence of user weight on peak handrail forces Discussion Peak compensatory handrail forces and influence of starting position Influence of handrail height on applied forces Implications for handrail design Study limitations Conclusions... 51

7 Chapter 4: Study 2: Influence of handrail height and falling direction on center-of-mass control and the physical demands of reach-to-grasp balance recovery reactions Preface Abstract Introduction Materials and Methods Participants Experimental setup Perturbation design Protocol Data processing Data analysis Results Balance recovery strategy forward-directed versus backward-directed falling Effect of fall direction and handrail height on the physical demands of reactive grasping handrail impulse Effect of fall direction and handrail height on COM and trunk control Discussion Backward-directed falling resulted in poorer COM control and greater physical demands of grasping, compared to forward-directed falling As handrail height increased, COM and trunk control improved and the physical demands of grasping decreased Limitations and future work Conclusions Chapter 5: Study 3: Effect of handrail height and age on the speed and accuracy of reach-tograsp balance recovery reactions during slope descent Preface Abstract Introduction Methods Participants Experimental setup Experimental protocol Data processing Metrics Statistical analysis Results Time course of reach-to-grasp movement Peak hand speeds during the reach trajectory trajectory Accuracy of reaching trajectory Discussion Influence of aging Influence of handrail height Study limitations Conclusions vii

8 Chapter 6: Study 4: Effect of handrail height and age on trunk and shoulder kinematics following perturbation-evoked grasping reactions during level-ground walking and slope descent Preface Abstract Introduction Methods Kinematic data collection, processing and modeling Handrail force data collection and processing Outcome measures Data analysis Results Shoulder elevation and plane of elevation angles Trunk angular kinematics Handrail forces Discussion Influence of handrail height on trunk and shoulder kinematics Influence of aging on balance recovery Influence of walking surface incline Comparison to findings during upright-stance perturbations Limitations Conclusions Chapter 7: Thesis discussion Summary of dissertation Applicability of models of whole-body balance recovery and compensatory grasping Inverted pendulum model Ballistic-and-corrective phase model of reach-to-grasp movements Impact of testing paradigms Implications for handrail design Handrail height Handrail structural strength Limitations and future directions The sample population consisted of healthy younger and older adults only The experimental environments were limited to level ground and slope descent Only one handrail cross-sectional shape was tested The perturbations used in this study were expected and repeated Ease of grasp completion was not evaluated Perturbations during the ongoing gait studies occurred at different stages of the gait cycle The safety harness likely limited participant movement, particularly for Studies 3 and References viii

9 Appendix A: Synchronization of Data Sources Preface Abstract A-1 Introduction A-2 Part A: Development of Synchronization Method A-2.1 System Implementation A-2.2 Verification of synchronization method A-3 Part B: Application of synchronization method during balance recovery reactions A-3.1 Methods A-3.2 Results A-4 Discussion A-5 Conclusions Appendix B: Error analysis and justification for filter parameters Preface B-1 Error analysis and justification for filter parameters - kinematics B-1.1 COM kinematics impact of possible cluster shifting B-1.2 Shoulder elevation angle estimates impact of marker shifting due to skin movement B-1.3 Justification for motion capture filter cut-off frequencies Study B-1.4 Justification for motion capture filter cut-off frequencies Study B-1.5 Justification for motion capture filter cut-off frequencies Study B-2 Justification for filter parameters kinetics B-2.1 B-2.2 Justification for 20 Hz cut-off frequency for load cell signals during the upright stance perturbation studies Justification for 8Hz cut-off frequency for load cell signals during the ongoing gait perturbation studies ix

10 Chapter 3: List of Tables Table 3.1: Participant demographics Table 3.2: Descriptive statistics (Mean (SD)) of the Maximum Withstood Perturbation magnitudes for each fall direction, initial hand position, and handrail height Table 3.3: Descriptive statistics of peak handrail forces during balance reactions in the forward falling direction Table 3.4: Descriptive statistics of peak handrail forces during balance reactions in the backward falling direction Table 3.5: Pearson s r-values (p-values in parentheses) and linear regressions describing the relationship between individual weight (kg) and peak handrail forces (N) for the loading axes and fall directions that resulted in the highest mean peak loads applied to each loading axis Chapter 5: Table 5.1: Participant demographics Table 5.2: Summary of metrics Table 5.3: Overview of the physical interpretation of the statistical model Table 5.4: Summary of timing variable performance with respect to handrail height for both younger and older adults Table 5.5: Summary of speed variable performance with respect to handrail height for both younger and older adults Table 5.6: Summary of accuracy variable performance with respect to handrail height for both younger and older adults Appendix A: Table A1.1: Comparison of time delays derived using the analog reference signal or through comparing position signals x

11 Chapter 2: List of Figures Figure 2.1: Summary of important steps that affect the likelihood of successful balance recovery via handrail grasping Figure 2.2: The inverted pendulum model applied to balance recovery via handrail grasping on level ground Figure 2.3. Visual representation of balance recovery via handrail grasping when the axis of rotation is no longer fixed in terms of length or position] Chapter 3: Figure 3.1: Experimental environment Figure 3.2: Summary of starting positions, fall directions and handrail heights tested in this study.. 35 Figure 3.3: Definitions of handrail loading axes Figure 3.4: Typical patterns of force generation along all three Cartesian axes Figure 3.5: Peak handrail forces for each condition and load axis Figure 3.6: The relationship between participant mass and peak handrail forces for the worst-case loading conditions Chapter 4: Figure 4.1: Testing environment and axis conventions Figure 4.2: Characteristic COM kinematics, trunk angular kinematics and resultant handrail force profiles during forward and backward falling Figure 4.3: Resultant handrail impulse for each condition Figure 4.4: COM range, and peak displacement, velocity and momentum magnitudes Figure 4.5: Trunk angular kinematics for each fall direction and handrail height Chapter 5: Figure 5.1: Experimental setup Figure 5.2: Time course of the reach-to-grasp reaction for a characteristic young participant Figure 5.3: Global timing variables with respect to handrail height Figure 5.4: t_onset-to-peak variables Figure 5.5: t_peak-to-contact variables Figure 5.6: Peak hand speed variables with respect to handrail height xi

12 Figure 5.7: Accuracy variables with respect to handrail height Figure 5.8: Observed strategies during balance recovery Chapter 6: Figure 6.1: Laboratory setup and definitions of kinematic outcome measures Figure 6.2: Peak shoulder elevation angles and concomitant plane of elevation angles during balance recovery Figure 6.3: Trunk roll kinematics Figure 6.4: Trunk pitch kinematics Figure 6.5: Peak handrail forces, normalized to %BW Chapter 7: Figure 7.1: Relationship between handrail height and unsuccessful grasping likelihood during slope descent Appendix A: Figure A-1: Time reference signal: a sawtooth wave with added bias and scaling to make use of a reasonable range of the analog data acquisition systems Figure A-2: Schematic diagram of system architecture Figure A-3: Timing reference signal recording compared to a loopback reference from an analog input on the main collection system connected directly to the original outgoing signal, with the added scaling and bias removed Figure A-4: Two sample trials of the time delay comparison between the position command angle, the angle as derived through motion capture and the angle measured by the encoder Figure A-5: The Challenging Environment Assessments Laboratory Figure A-6: Sample vertical hand position (a), with and without synchronization technique applied, and lateral handrail forces (b) during a single trial in which the participant reaches to grasp a handrail following a balance perturbation Figure A-7: Histogram of motion capture recording time delays across 612 trials Appendix B: Figure B-1: Sample estimates of the vertical position of the COM for a shorter (159cm) participant in Study 2, with unaltered data and a simulated 5cm upward shift in the pelvis cluster Figure B-2: Sample estimates of the vertical position of the COM for a different participant in Study 2 (height = 173cm), with unaltered data and a simulated 5cm upward shift in the pelvis cluster xii

13 Figure B-3: Thoracohumeral elevation angle estimates for one trial of one older adult, collected during slope descent when the handrail was 44 inches high Figure B-4: Relationship between peak thoracohumeral elevation angle (unaltered estimate) and the change in angle magnitude when shoulder markers are shifted 3cm downward and 3cm upward. 159 Figure B-5: Effect of filtering noisy COM estimates with a 6Hz filter Figure B-6: Sample effects of filtering on vertical hand position and velocity traces, for two trials collected from a young participant Figure B-7: Sample vertical force signal for a reach-to-grasp trial in Study Figure B-8: Effect of filter cut-off frequency on the limiting case for posterior loading Figure B-9: Frequency spectra of the handrail load cell signals when the handrail was 44 inches high, for the ongoing gait paradigm xiii

14 List of Appendices Appendix A: Synchronization of Data Sources Appendix B: Error analysis and justification for filter parameters xiv

15 Chapter 1 Introduction 1.1 Motivation Failure of balance recovery mechanisms to counteract a balance disturbance can lead to a fall. Falls are a leading cause of unintentional injury and death in North America, particularly among adults over the age of 65 [1, 2]. Handrails are common safety features that can significantly enhance a person s ability to recover from balance loss and avoid a fall [3], provided that their design allows the person to quickly and accurately reach to grasp the rail, and then apply sufficient forces and torques to the rail to stabilize their centre of mass (COM) [4]. This is particularly pertinent to contexts where the balance disturbance is too large for hip or ankle strategies (i.e. stiffening the muscles in the hips and ankles) [5] to be adequate for balance recovery, and where the stepping reactions that individuals would normally use to recover from large balance disturbances [4] are not reliable, such as on stairs or on icy, slippery ramps. Despite the widespread presence of handrails in our built environment, our understanding of how design and installation parameters (e.g. height) affect their utility for balance recovery is limited. In the 1980s, Maki and colleagues conducted pioneering biomechanical studies on the influence of handrail design on its potential effectiveness for balance recovery (e.g. [6, 7, 8, 9]). These studies explored how features such as height [6, 7, 8] and cross-sectional shape, size and surface texture [9] affected the highest volitional forces and moments that younger and older adults could apply to stairway handrails while standing still. Specific to handrail height, this research found that as handrail height increased (within the range of 32 to 42 inches), participants applied higher maximum volitional forces and moments in the anterior and posterior directions [6, 7], which were advantageous for countering translational and rotational forces acting on the person s COM in these directions during a fall. However, higher rails also reduced the maximum volitional forces applied to the rail in the downward direction [6, 7], which was disadvantageous for counteracting downward forces acting on the COM. This work recommended 36 inches (91 cm) as an optimal height for stairway handrails, as it allowed most participants to apply high forces and moments to the rail in all tested directions, and was perceived to be the most comfortable based on heuristic feedback. These studies substantially advanced our understanding of how the ability of younger and older adults to apply high voluntary stabilizing forces and moments to a handrail depends on the rail s 1

16 2 design. However, this work took place while participants stood still, and focused on the slow application of high forces with participant-selected grip locations on the rail [6, 7, 8]. The extent to which the findings translate to balance recovery contexts where applying high stabilizing forces and moments to the rail can only occur if the person has first executed a fast and accurate reach-tograsp reaction with their COM in motion is unknown. Further questions arise with respect to whether high handrail forces are even necessary for balance recovery, and the relationship between applied handrail forces and a person s ability to stabilize their COM in a balance recovery situation. Ongoing gait contexts that better represent common circumstances of falls [10] must also be considered. Finally, effects of aging should be explored, due to common physiological changes of aging such as loss of strength and muscle mass [11, 12], and decreased ability to coordinate effective protective responses such as stopping a fall with the limbs [13]. In older adults, these factors both increase the risk of falling, and increase the risk of injury should a fall occur. 1.2 Objectives and Scope This thesis investigates the effect of handrail height on recovery from balance perturbations. After a literature review (Chapter 2), this objective is addressed through the following four studies: 1) Use of handrails for balance and stability: Characterizing loading profiles in younger adults and effect of handrail height In Study 1 (Chapter 3), we focused broadly on the influence of handrail height on compensatory grasping effectiveness in fifty young adults. Specifically, we characterized the influence of handrail height on the maximum perturbation magnitude that young adults could withstand without stepping or falling, while using the handrail for balance recovery. We also quantified the concomitant peak forces that participants applied to the rail during balance recovery. Starting position (i.e., not contacting the rail before perturbation onset, versus contacting the rail with one or two hands) was considered. 2) Influence of handrail height and falling direction on center of mass control and the physical demands of reach-to-grasp reactions In Study 2 (Chapter 4), we characterize detailed kinematics of COM and trunk control during balance recovery and the impact of handrail height, using controlled perturbation magnitudes. We also quantified the impulse that participants applied to the handrail at multiple stages of the recovery reaction, as a proxy measure of the physical demands of balance recovery. This work builds on Study 1 in two major ways. First, by extracting COM and trunk control kinematics

17 3 during balance recovery, we gain more detailed insights into individual stability during balance recovery than were possible by looking solely at the maximum perturbation that participants would withstand without stepping or falling. Second, by considering impulse shortly after rail contact (within the first two seconds), we gain a stronger understanding of the handrail forces that must be sustained beyond the instant that a peak force is observed. 3) Effect of handrail height and age on the speed and accuracy of reach-to-grasp balance reactions during slope descent In Study 3 (Chapter 5), we characterize the timing features, hand speeds, and vertical handrail overshoot during reach-to-grasp reactions in 14 younger adults and 13 older adults, following perturbations during slope descent. Regression models are used to relate performance in outcome measures with respect to individual height, and quantify significant effects of age and handrail height. This study builds on my first two studies in three respects. First, it considers perturbations during ongoing gait (slope descent), rather than the upright stance perturbations used in Studies 1 and 2. Evaluating balance recovery reactions during ongoing gait is important because arm reactions have been observed to be slower during gait or other leg movement, compared to stationary stance [14, 15, 16]. The delay in reaction may have consequences to balance recovery for handrails that cannot be reached quickly. Second, it explores effects of aging. Third, it considers a wider range of handrail heights (30 to 44 inches high). 4) Effect of handrail height and age on trunk and shoulder kinematics following perturbation-evoked grasping reactions during level-ground walking and slope descent. In Study 4 (Chapter 6), we evaluate the influence of handrail height on trunk control and arm kinematics following perturbations during level-ground walking and slope descent. This study characterizes trunk angular roll and pitch in terms of displacement and velocity after balance perturbations, and how they vary with handrail height. It also quantifies angular characteristics of the thoracohumeral joint during balance recovery and discusses findings relative to those of patient populations with shoulder pathologies in the published literature. Thoracohumeral joint angles are of interest because reaching for handrails at high heights may exceed the range of motion in many individuals with shoulder conditions, such as arthritis or rotator cuff injuries. Chapter 7 serves as a general discussion for the thesis, in which I summarize the primary contributions from each study; describe the implications for handrail design in the context of balance recovery; and discuss limitations and future directions. Appendix A details the method used

18 4 to synchronize data sources, while Appendix B provides a selection of analyses on possible sources of error and justification for filter parameters.

19 Chapter 2 Review of the Literature This literature review begins with an overview of strategies for balance recovery, focusing on the role of the legs, trunk and arms in recovering from perturbations. It then discusses the factors that are likely to influence compensatory grasping, including: (1) whether the person is initially touching the handrail; (2) the speed and appropriateness of the initial reaction to balance loss; (3) the speed and accuracy of the reaching trajectory; (4) the ability to complete a successful grasp; and (5) the ability to apply sufficient forces and moments to the handrail to stabilize the centre of mass. 2.1 Overview of strategies for balance recovery Balance recovery is a complex task that often requires substantial coordination of the legs, trunk and arms to enable humans to regain control of the position and velocity of their COM with respect to their base of support. In this thesis, base of support is defined by the boundaries of (a) the person s feet in contact with the ground, and (b) the person s hand if in contact with an external surface. While this thesis focuses primarily on compensatory grasping, activities in the trunk, legs and non-grasping arms are also included in the literature review for completeness Compensatory leg and trunk activity: stiffening and stepping Broadly, balance recovery strategies involving the legs and trunk have been classified as fixedsupport and change-in-support strategies [4]. While handrail grasping has also been classified within the fixed-support and change-in-support framework, discussion of the role of handrails in balance recovery is reserved for Section (on the role of the arms in balance recovery) Fixed-support strategies Fixed-support strategies are often used in response to low-velocity balance perturbations, and are characterized by the person s base of support remaining in place (i.e., no stepping or reaching to contact nearby handholds), while muscles in the trunk and legs quickly activate to stiffen the body and control the displacement and velocity of the COM [4, 5, 17]. Horak and Nashner discuss fixedsupport strategies in depth in their 1986 work [5]. Briefly, they include: Ankle strategies, which involve fast, sequential activation of agonist muscles starting near the ankle joint (within roughly 100ms of the perturbation), then upward through to the trunk [5]. 5

20 6 An ankle strategy allows individuals to increase torques around the ankle joints that compensate for the translational and rotational forces acting on their COM, and thus stabilize the position and velocity of their COM with respect to their base of support. Hip strategies, which involve the simultaneous activation of agonist thigh and trunk muscles. Hip strategies have been observed during balance perturbations on shorter support surfaces (relative to foot length) [5], resulting in horizontal shear force between the feet and support surface, with minimal ankle torques [5]. Fixed-support balance control strategies exhibiting activation patterns in the legs and trunk that differ from those described above have been observed in other balance recovery situations, such as following lateral perturbations [18], and during perturbations that are continuously-delivered over longer periods (>1s) [19]. In platform perturbation settings, perturbation magnitude has also been shown to influence fixed-support balance recovery responses. In particular, peak agonist muscle magnitude has been reported to increase as perturbation magnitude increases (albeit not linearly), for both forward and backward platform movement [19]. Similarly, co-contraction indices of the tibialis anterior and medial gastrocnemius muscles and torque generation in the hips and ankles have been noted to increase with perturbation magnitude for transient backward platform movement [19, 20] Compensatory stepping Fixed-support strategies are helpful for balance recovery from small disturbances. However, changein-support strategies, such as compensatory stepping, are often necessary for recovery from large balance disturbances (assuming that the person is not already grasping a handhold) [4]. Stepping allows the person to modify the position and size of their base of support relative to the position and velocity of their COM, and can thus help the person to maintain or regain stability. Even young adults have been observed to step following low-to-medium magnitude platform perturbations (acceleration <1.5 m/s 2 ) when not explicitly instructed to avoid stepping, and when balance recovery was possible without stepping or handhold grasping [21]. In this thesis, compensatory stepping refers to rapid steps evoked by balance disturbances, while volitional stepping describes steps that participants are instructed to execute as quickly as possible following cues that do not threaten balance, such as light or sound. While the shape of vertical ground reaction force profiles for volitional and compensatory stepping have been observed to share common features, compensatory stepping is distinguished through

21 7 reduced latency of initiation, and reduced duration of the preparatory phase (i.e., time between perturbation and foot lift-off), swing phase, and overall step [22]. In the absence of environmental constraints to step execution, the swing phase of compensatory stepping was reduced by over 50%, occurring in less than 200ms on average [22]. Beyond apparent similarities in the shape of vertical ground reaction force profiles, perturbation-evoked stepping has been shown to be preceded by frequent symmetric activation of the tibialis anterior in both legs, irrespective of the side of the swing foot [23]. This activation was termed an automatic postural response to platform motion that was most frequently observed during the first trial, and served to stabilize the COM following the perturbation [23]. Despite the rapid nature of compensatory stepping and possible presence of common, automatic muscle activity, young adults have shown the ability to modulate their step trajectory and timing based on environmental constraints, while still recovering balance with a single step. For example, Zettel and colleagues noted that young adults could step over an obstacle at 15% of their height following platform perturbations, even with a barrier mounted to the side to constrain lateral foot placement [24, 25]. While the presence of an obstacle roughly doubled the swing duration (~150ms to ~350ms for no-obstacle and obstacle respectively) and increased the anterior step length by roughly 15cm, participants were still generally able to recover balance [24, 25]. More recently, Scovil and colleagues observed that young adults could often execute compensatory steps through a narrow (20cm-wide) slot in front of them, even when the precise position of the slot (on the left or right side of the body) varied, and even without being able to see the position of the slot until onset of the platform perturbation [26]. However, seeing the position of the slot before perturbation onset significantly reduced the likelihood that participants would contact the barriers around the slot with their foot; reduced the time to compensatory step initiation; and reduced the frequency of needing to take a second compensatory step, presumably to arrest the forward motion of the trunk [26]. Compensatory stepping has also been explored in older adults. Of note, the temporal and spatial characteristics of the first step following antero-posterior platform perturbations in healthy young and older adults have been observed to be similar [27]. Brauer and colleagues noted delays in initiation of gastrocnemius muscle activation following platform perturbations in older adults (both healthy and those with balance impairments, characterized by scoring <50/56 on the Berg Balance Scale) compared to younger adults [28]. However, the magnitude of average delay was only 6ms on average in healthy older adults, and 22ms in those with balance impairments [28].

22 8 Despite these similarities, more pronounced age-related differences in other aspects of compensatory stepping behavior have been observed, including: Reduced destabilization threshold for initiating a compensatory step: Even following low perturbations, evidence from a number of studies suggests that older adults may favour stepping responses for balance recovery, compared to fixed-support strategies. Jensen and colleagues reported that many older adults executed compensatory steps to recover from low-magnitude antero-posterior platform perturbations, whereas younger adults in the same study maintained balance with their feet in-place [29]. This occurred even though the difference in momentum (whole body, head-arms-trunk, and limb segments) leading into step initiation was not statistically-significant between the two populations [29]. Separately, in a study involving waistpull perturbations of varying magnitudes, Mille and colleagues noted that compared to younger adults, older adults initiated stepping with lower anterior excursion of the pelvis relative to the base of support, and with reduced pelvis velocities [30]. Increased tendency to take multiple steps: A number of platform perturbation studies have reported that participating healthy older adults were significantly more likely than younger adults to execute more than one compensatory step after antero-posterior and lateral perturbations (e.g., [31, 27]). Maki and colleagues further noted that the tendency of both younger and older adults to take more than one step following lateral perturbations was increased when walking in place, compared to perturbations during quiet stance [31]. Increased frequency of collisions between feet following lateral perturbations: Older adults have been shown to experience more frequent collisions between the swing and stance leg after platform perturbations, particularly for perturbations delivered while walking in place [31]. This suggests that older adults may be compromised in their ability to control their lower-limb trajectory following perturbations during ongoing leg movement [31] Compensatory upper-limb activity: movement, light contact and grasping The upper limbs can also play a critical role in balance control and recovery. In a pilot study of healthy young adults, simply being able to move the arms resulted in significant improvements to clinical balance assessments, including increased maximal voluntary step length and reduced time to complete the voluntary step [32]. Beyond gross clinical assessments, the upper limbs can (a) move

23 9 quickly to provide protective functions or counteract COM momentum [33]; (b) provide tactile feedback during contact with a rigid surface, which can enhance stability [34]; and (c) grasp nearby handrails, to anchor the body with respect to the ground and stabilize the COM [4] Upper-limb movement to counteract COM momentum or protect from injury Numerous studies have reported engagement of the upper limbs following balance perturbations in the absence of reachable handholds, in both upright stance conditions (e.g., [35, 36, 37]) and following simulated trips and slips while walking (e.g., [38, 39, 33, 40]). While a few studies have proposed that upper-limb muscle engagement reflects a common corrective response [41] or startle response [42], evidence from other studies suggests that rapid arm movements are unlikely to be generic, and are often highly-modulated to the balance recovery task. For example, Pjinappels and colleagues noted asymmetrical muscle activity on the shoulders during balance recovery from trips, which depended on the tripping leg (left versus right) [33]. Activation of upper-limb muscles has also been shown to occur more quickly as perturbation magnitude increases from low to mediummagnitude perturbations, though response latencies did not differ with further increases in perturbation magnitude [37, 35, 3]. More recent work suggests that arm motion may serve as a strategy to counteract the momentum of the trunk and COM following low-magnitude forward or backward perturbations, thereby helping individuals to regain stability. Supporting evidence is provided in studies involving trips [39, 33], forward platform translations [36], and lean-and-release perturbations [43, 44], which report that participating younger adults moved their arms in the direction opposite that of the trunk (i.e., arm movement upward with forward/downward trunk rotation during forward falling [39, 33, 43, 44]; arm movement forward with backward COM movement during backward falling [36]). Cheng and colleagues further noted that while moving the arms forward during lean-and-release perturbations was initially destabilizing, the forward linear momentum of the arms later in balance recovery appeared to help to reduce the forward momentum of the trunk, and contribute to stability for low to moderate lean angles (< 10 o ) [44]. Participants generally needed to step to recover from higher lean angles, even with arm swing [44]. In contrast with young adults, arm motions generated by older adults during trips [39] and lateral platform translations [36] were not consistent with strategies that would have counteracted COM momentum. Instead, participating older adults moved their arms in directions that were consistent

24 10 with protective strategies observed during actual falls, which served to reduce the impact between the hips and the ground [45, 36, 39] Light contact with support-surfaces to enhance tactile feedback Contacting support surfaces with the hands has been demonstrated to reduce postural sway and trunk velocity during quiet standing [46, 47], likely due to enhanced tactile feedback between the surface and the finger [34]. The stability benefits of hand contact with a support surface are less apparent when recovering from externally-imposed perturbations to the trunk and feet. In a platform perturbation study of individuals with diabetes mellitus and healthy controls (mean age of groups: 59.7 years and 61.1 years respectively), Dickstein and colleagues demonstrated that neither center of pressure (COP) response latencies nor medial gastrocnemius muscle latencies were significantly affected by light or heavy hand contact (contact force <1N and as high as desired, respectively) with a hand-support surface to the right of participants [48]. However, the touch mode significantly affected mean COP velocity in both populations: antero-posterior (A-P) COP velocity was minimized during heavytouch and largest during no touch (implying a stability advantage with touching), though mediallateral (M-L) COP velocity was maximized with heavy touch and minimized with no touch [48]. Note that the increase to M-L COP velocity with light and heavy touch may not necessarily translate into reduced stability, as participants also had a larger base of support in the M-L direction due to hand-surface contact on the right. On the whole, however, the applicability of these findings to major balance disturbances cannot be evaluated because the perturbation magnitudes were low (peak platform velocity: 30cm/s; displacement: 6cm) [48]. More recently, Martinelli and colleagues explored the effect of light touch on balance in young adults during backward waist-pull perturbations, with participants standing on malleable and rigid surfaces, and with both vision and no-vision conditions tested [49]. Light touch resulted in statisticallysignificant reductions to both gastrocnemius activation and COP amplitude during the perturbation (indicating a stability advantage with lower muscular demands) [49]. However, the low magnitudes of differences in mean COP excursion amplitude between light-touch and no-touch conditions (<0.5cm, or under 5% of the total mean COP amplitude (~13 to 15cm)) make the functional relevance difficult to evaluate [49], particularly for larger perturbations that would have required stepping, grasping, or high forces between the hand and contact surface for balance recovery.

25 Handrail grasping to stabilize the COM Handrails (and other handholds, including grab bars and grab poles) can contribute to balance control and recovery through three major ways (adapted from [4]): 1. By allowing the user to apply stabilizing forces and moments to the handrail, so as to counteract the downward translational and rotational forces acting on the body following balance loss; 2. By providing rigid support for the user to grasp and anchor the body; and 3. By allowing the user to extend the base of support (beyond their feet), thereby increasing the range of COM displacement that can occur without further loss of stability. The utility of handrails in enhancing balance control and recovery is well-supported by empirical data in young adults, following both small and large balance disturbances. In a study of continuous postural perturbations during quiet stance, Camernik and colleagues observed significant reductions to COP excursions when participants grasped a vertical handhold, compared to no handhold contact [50]. Conversely, in a study involving backward waist pulls during treadmill walking, Misiaszek and colleagues noted that holding an anteriorly-positioned handrail allowed participants to continue walking after perturbations with reduced disruption to their gait cycle timing (i.e., stance and swing phase were prolonged by ~75ms when participants arms were at their sides, but were only prolonged by ~12ms when participants held the rail) [51]. The reduced alteration to gait timing suggests that handrail grasping minimized the extent to which the perturbation disrupted balance. Finally, by using a three-step mock staircase mounted to a robotic platform, Maki and colleagues demonstrated that being able to grasp the handrail after a balance perturbation significantly reduced the likelihood that participants would contact the crash pad at the base of the staircase (i.e., fall ), compared to when no handrail was present ( fall in 7/13 trials with no handrail, compared to 6/80 trials with a handrail present) [3]. Importantly, participants in this study were perturbed in conditions that made effective compensatory stepping nearly impossible, and at magnitudes from which successful balance recovery was extremely challenging without handrail grasping [3]. In older adults, empirical evidence on the functional utility of handrails for balance recovery is limited, due to the absence of studies on balance recovery with and without a handrail in this population. However, the value of handrails for balance control and recovery in older adults is supported by studies involving balance-challenging activity (e.g., unexpected platform perturbations [15], treadmill walking [52], stair ascent and descent [53, 54], and corridor walking among assisted living facility residents [55]), in which participants either used surrounding handrails without having

26 12 been explicitly instructed to do so, or expressed preferences for using a handrail when walking. Most notably, in a study involving surprise perturbations (i.e., participants were told that the robotic platform would not move during the trial of interest), older adults were over twice as likely to reach to grasp the nearby handrail as young adults (10/12 versus 4/12, respectively) after balance loss [15]. 2.2 Factors that are likely to influence the success of balance recovery via compensatory grasping A number of factors can affect whether a person can quickly apply high grasping forces to a handrail to stabilize their COM following balance loss. This section will be framed by examining key steps in the process of using a handrail for balance recovery, including: 1. Initial conditions (i.e., whether the person is touching the handrail before balance loss) 2. The speed and appropriateness of the initiated balance recovery response 3. The speed and accuracy of the reaching trajectory, leading into handrail contact 4. The success of grasp completion 5. The sufficiency of the forces and moments that the person applies to the handrail to stabilize their center of mass The extent to which factors related to the person, task and environment [56] could influence the successful execution of the steps outlined above will be considered, with emphasis on plausible effects of aging and handrail design features. These factors are described in detail below, and summarized in Figure 2.1. Note that successful balance recovery is likely to be affected by concomitant reactive stepping, engagement of muscles in the trunk and legs, and activity in the nonreaching arm. However, detailed discussion of these activities is beyond the scope of this section Initial conditions value of proactive handrail contact Proactive handrail contact (i.e., the person is touching the handrail before a balance disturbance) is helpful because it eliminates the need for the person to quickly and accurately reach for the handrail before grasping can occur. This reduces the probability that balance recovery via handrail grasping will fail due to insufficient speed or accuracy in the reaching trajectory. It also reduces the time lag of the reach-to-grasp movement that allows the COM to gain momentum following a destabilization, thereby increasing the magnitude of forces that the person may need to apply to the handrail to counteract this added momentum [6]. Further, proactive handrail contact allows users to benefit from the tactile feedback in the hands (as described in Section ) to enhance balance control during small perturbations [48, 49].

27 13 Select factors that have been shown to, or are likely to, affect the response or situation Balance loss Aging Awareness of handrail position Perceived handrail comfort: Height Surface texture Initial distance from hand to handrail Height Horizontal distance from handrail Aging, fall status, and concurrent attentional demands Leg movement Magnitude of destabilization Cueing Balance confidence and fear of falling Resemblance to a handrail Aging Presence of handhold in visual field Cueing Holding an object Fast initiation of appropriate reaction Hand initially touching rail? Fast and accurate reach to contact the rail Grasp completion Handrail height Aging Handrail cross-sectional design Other concomitant activity Engagement of lower-limb and trunk muscles Compensatory stepping Movement of the non-reaching arm Application of high stabilizing forces & torques to the rail Successful balance recovery Unsuccessful balance recovery Figure 2.1: Summary of important steps that affect the likelihood of successful balance recovery via handrail grasping. Green arrows signify that the step of interest (filled, turquoise boxes) was executed successfully; red arrows signify that it was not. If the person fails to execute any step of interest, the likelihood of failed balance recovery increases. The unfilled boxes on the right depict factors that are likely to (or that have been shown to) affect performance in executing the step of interest. Note that handrail cross-sectional design and other concomitant activities in the trunk, legs and non-reaching arm are also likely to affect balance recovery. However, they are indicated in lighter text as they are beyond the scope of this literature review. Note also that this list of factors is not comprehensive. A number of factors related to aging, task and environment could influence the likelihood of proactive contact, including: 1. Perceived handrail comfort: Handrails that are not comfortably touched while walking are unlikely to encourage proactive contact. Handrail height is likely to affect comfort: high installation heights that require large thoracohumeral joint elevation angles or substantial elbow

28 14 flexion may be uncomfortable or difficult to contact, particularly for individuals with reduced range of motion in these joints due to arthritis [57], long rotator cuff tears (particularly among older adults) [58], a history of shoulder dislocations [59], or other conditions. Conversely, low handrails (below fingertip-height) may require leaning to touch, which could be destabilizing. In Maki and colleagues evaluations of the effect of handrail height on maximum volitional force generation in healthy younger and older adults, the mean preferred handrail height (reported in heuristic feedback) was 36, or 91cm [7]. Other handrail design features beyond the scope of this thesis also affect handrail comfort. Of note, various surface textures may discourage contact if they are painful to contact, or lead to high friction between the rail and the hand sliding along it. In prior evaluations of handrail surface texture on comfort, heuristic feedback from younger and older adults revealed that rough and foam surfaces resulted in poor sliding comfort between the hand and the rail [9]. 2. Decreased balance confidence and fear of falling: In lab-based studies of stair gait in healthy older adults, decreased balance confidence and fear of falling have been associated with increased proactive handrail use, particularly during stair descent [60, 54]. 3. Cueing: In an unexpected-perturbation study involving 160 community-dwelling older adults, McKay and colleagues demonstrated that verbal cueing (i.e., being instructed to use a nearby handrail) increased the tendency of participants to contact the rail proactively (37/40 participants), compared to no cueing (5/40 participants), visual cueing (i.e., flashing green LEDs within the translucent black handrail) (0/40 participants), or combined verbal and visual cueing (23/40 participants) [61]. The high compliance with verbal cueing is notable, given that participants were told that they would not be perturbed in the analyzed first trial. (For details on the handrail cueing system, see Scovil et al [62]). 4. Resemblance to a handrail: In focus groups of residents in six assisted living facilities in Texas, good corridor handrail systems were generally characterized by continuity and easiness of grasping [55]. Further, in the one facility that installed chair rails along the walls (to ideally function as handrails), many residents mistook the rails for interior decoration (e.g., molding), and did not use them as handrails [55]. This thesis does not evaluate the effect of handrail resemblance on balance control, but it remains an important consideration for handrail design.

29 15 Published literature comparing balance recovery with both proactive handrail contact and reach-tograsp reactions appears to be limited to one study, which reported the peak forces that four young men applied to a handrail following balance loss on stairs [3]. Compared to reach-to-grasp reactions, gripping the handrail before perturbation approximately halved the mean applied peak lateral forces (16% of body weight (%BW) versus 8% BW), and nearly doubled the applied peak upward normal forces (12%BW versus 21% BW) [3]. The concomitant effects of proactive contact on kinematic control during balance recovery, or on the likelihood of a fall, were not reported [3]. This work provided helpful insights on the effect of proactive contact on handrail forces during balance recovery; however, the small sample limits the generalizability of the findings Speed and appropriateness of balance response initiation During a fall, hand, hip or head impact with the ground can occur in less than a second after balance loss [63, 64]. This may limit the time available for executing balance recovery reactions to avoid a fall. Accordingly, the speed with which individuals detect and react to balance perturbations and the appropriateness of the reaction to the detected perturbation may be important for balance recovery. In this sub-section, I describe physiological elements of balance loss detection, followed by factors that are likely to affect balance response initiation Physiology of balance loss detection Humans integrate visual, vestibular and proprioceptive information to detect the position and movement of body segments in space, and then execute appropriate motor actions to control balance. Vision is important for creating visuospatial maps of the environment, which allow us to identify hazards [65] and potentially-destabilizing situations (e.g., patches of ice on the sidewalk), as well as integrate with proprioceptive information of the limbs to avoid obstacles and execute rapid, targeted limb movements [66]. While our ability to detect and respond to a vast array of balance disturbances is likely optimized with combined visual, vestibular and proprioceptive input (assuming that the input is accurate and can be integrated appropriately by the central nervous system), proprioceptive information from the legs appears to be the most sensitive of the three systems for detecting postural sway in young adults [67]. The role of plantar mechanoreceptors in detecting and guiding the subsequent response to balance loss has been established experimentally. Of note, Perry and colleagues explored the effect

30 16 of attenuation of plantar-surface sensation via cooling (submersion in ice water) on balance recovery in younger adults from translational, multi-directional platform perturbations [68]. For backwarddirected falling, cooling resulted in higher COM excursion toward the heels (i.e., the posterior edge of the base of support) before foot-lift occurred for compensatory stepping, indicating delayed detection of when the COM approached the posterior limits of stability while standing [68]. Similar trends were observed in full-vision and blindfolded conditions, suggesting that impediments to plantar surface sensation could not be overcome by vision alone [68]. Whereas proprioceptive information from the legs is important for detecting postural sway and imbalances initiated at the feet (e.g., weight shifting, slipping), information from the vestibular system may be more important for detecting imbalances initiated closer to the head (e.g., being pushed by others, as observed in some videos of falling in care homes [10]). Declines to the visual, vestibular and somatosensory systems have been observed with aging [69, 70, 71] each of which could compromise ability to detect and respond to balance loss. Loss of vision is an established risk factor for falls (see review in [72]). In older adults (>65 years), declines to plantar-surface sensitivity to vibratory and touch stimuli have been observed [73]. The increased thresholds of detecting plantar stimuli in older adults could delay detection of the onset of balance disturbance, and the subsequent initiation of the balance recovery response. More recent work by Wingert and colleagues revealed that age-related declines in hip proprioception were not associated with changes in postural sway, but did result in reduced mini-bestest scores related to dynamic balance control [71]. This suggests that mechanoreceptors in the hips are unlikely to be as sensitive as those in the feet for detecting balance loss, but remain important for the successful execution of limb movements important to balance control Factors affecting balance response initiation Numerous factors influence the speed and nature of balance response initiation, including: 1. Aging: A number of studies report age-related delays to upper-limb muscle activation after balance loss leading into reach-to-grasp reactions, following translational platform perturbations of upright stance (~13ms to 16ms) [74, 75] and ongoing gait (~100ms, though data from only four younger adults are reported in the comparison) [15]. The experimental paradigm may be important in comparing age-related changes in arm muscle onset. During repeated translational

31 17 perturbations of upright stance (i.e., averaged over multiple trials per person), observed differences were statistically-significant but low in magnitude [74, 75]. Conversely, following platform perturbations during ongoing gait, analyses of first trials (involving deceptions) revealed greater age-related delays in upper-limb muscle activation time, which may have been exacerbated by the unexpected nature of the perturbations [15]. 2. Leg movement: Both active and passive (i.e., motor-driven pedalling) movement of the legs have been observed to delay reaching movement initiation following sudden chair tilts, compared to no leg movement (~43ms to 47ms) [14]. Comparisons of activation timing for compensatory reaching between walking and standing are perhaps less appropriate, because published studies involving reach-to-grasp reactions during over-ground gait are largely limited to first-trial analyses of unexpected perturbations (e.g., [15, 16, 61]). Accordingly, it is difficult to know if apparent delays to upper-limb muscle initiation stem from walking, or from the perturbation being a surprise. Regardless, for reach-to-contact reactions following forward platform translations (resulting in backward-directed falling) with the handrail on the participant s right side, mean arm muscle activation onsets in young adults have been reported around 123ms to 131ms [75, 76] following upright stance perturbations, versus 186ms 1 during ongoing gait (among young participants who reached for and contacted the nearby handrail) [16]. In older adults, the apparent differences increase: Weaver and colleagues report mean posterior deltoid onsets of 147ms for older adults following upright stance perturbations [75], while McKay and colleagues report 213ms for upper-limb onset following surprise perturbations during ongoing gait with an older adult group (in the absence of handrail cueing) [61]. 3. Awareness of handrail position: A few platform perturbation studies have explored how prior knowledge of the position of a handrail (i.e., to the left or right of the body) before perturbation onset influenced reactive grasping (e.g., [77, 74, 76, 75]), with varied results. Cheng and colleagues characterized reach-to-grasp reactions with a small, moveable handhold [78], 1 The source paper reports EMG onset in the middle deltoid and biceps for the reaching arm. To calculate reaction onset from the results reported in the paper, I considered participants who reached for and grasped or contacted the laterally-positioned handrail, and took the minimum onset between the biceps and middle deltoid. 186ms represents the mean of these reaction onset values.

32 18 positioned anterior to the body and to the left or right of participants. Reaction time was fastest when participants (both younger and older adults) relied on stored visuospatial information only (i.e., they could see the position of the rail before the perturbation, but vision was occluded after perturbation onset), compared to full-vision and online-only vision (i.e., participants could not see the location of the handrail before perturbation). Online-only vision increased the likelihood that participants would raise both arms immediately after the perturbation [77, 74]. Further, in the case of older adults, online-only vision resulted in more frequent use of the wrong arm, resulting in contra-lateral reaching [74]. Conversely, Weaver and colleagues varied the predictability of handrail position (i.e., always on the right side, versus on the right or left side) [75, 76]. Participant vision was occluded leading into every perturbation trial, irrespective of the predictability condition [75, 76]. The unpredictable condition led to slight delays in EMG latency (<20ms) for both forward and backward falling, and in both younger and older adults [75, 76]. 4. Magnitude of destabilization: As discussed in Section , upper-limb muscle activation has been observed to occur more quickly as perturbation magnitude increases from relativelylow to moderate perturbation levels [37, 35, 3] Speed and accuracy of the reaching trajectory Following movement initiation, the person must quickly and accurately reach to contact the handrail before grasping can occur. The speed and accuracy of the reach trajectory may be co-dependent, with speed-accuracy trade-offs well-established in the volitional reaching literature (e.g., Fitt s Law) [79] Strategies for fast and accurate reaching: volitional and compensatory The control of compensatory reaching may be better considered if discussed with reference to models of volitional reaching. In this thesis, compensatory reaching describes rapid reaching movements evoked by balance perturbations. Note that this does not cover: (a) corrective actions to arm movements following mid-reach perturbations (as described in a number of studies of reaching using robotic systems); or (b) corrective strategies to overcome limitations in range of motion due to conditions such as stroke (e.g., during distal reaches, a person with limited range of motion in the elbow might compensate by leaning further forward).

33 19 Volitional reaching has been previously characterized as a two-phase process [80, 81], comprising: 1) An initial, higher-velocity, ballistic phase, during which much of the reaching distance is covered; and 2) A subsequent, lower-velocity, corrective phase leading into contact with the object, during which adjustments to the direction of the reach trajectory and orientation of the hand can be performed to increase the likelihood of successful grasp completion. The use of this model to describe volitional reach-to-grasp movements is well-established (e.g., Jeannerod, 1984 [80]; Haaland and Harrington, 1989 [81]). However, its use to characterize compensatory reaching is relatively nascent. Accordingly, while results from the comparativelylimited volume of published papers exploring detailed kinematics of compensatory reaching are not inconsistent with this model (e.g., Ghafouri et al, 2004 [82]), and evidence is emerging to support its appropriateness for describing elements of compensatory reaching (e.g., Gage et al, 2007 [66]; Cheng et al, 2012 [77]), it should not be treated as a quintessential representation of compensatory reaching. In Study 3 of this thesis (Chapter 5), I use it to characterize aspects of compensatory reaching trajectories, as it reveals age-related differences in compensatory reaching that are consistent with literature on volitional reaching. Nevertheless, two aspects of this model are highlighted in the context of compensatory reaching: (1) Directional control of the reach trajectory during the ballistic phase; and (2) The possible relationship between the duration of the corrective phase and the need to modulate the reach trajectory leading into handrail contact. 1) Directional control of the reach trajectory during the ballistic phase: The nomenclature may imply that the ballistic phase involves little directional control of the hand trajectory. In compensatory reaching, however, numerous studies of young adults have reported that the early stages of reach-to-grasp movements are heavily-modulated, based on the relative movement of the person, and of the object being grasped. Of note, Gage et al compared control strategies during volitional (sound-cued) versus compensatory (evoked by sudden, multi-directional chair tilts) reach-to-grasp movements [66], while both Ghafouri et al, and Corbeil et al explored earlystage wrist trajectories following multidirectional perturbations of upright stance [82, 36]. Differences between characteristic wrist trajectories based on perturbation direction and amplitude were apparent within the first 100ms of the perturbation [82, 66, 36] even for trials

34 20 where vision was occluded for the first 200ms after perturbation onset [82]. Different (and more direct) trajectories were observed with volitional reaching [66]. The observed, rapid modulation of reaching movement with respect to the falling direction (particularly in the absence of online vision) implies that individuals can quickly integrate vestibular or somatosensory information about the motion of the body with prior visual knowledge of the location of the handhold, and execute relatively fast and accurate reach-to-grasp reactions. Gage and colleagues study further noted that while the temporal features differed between compensatory and volitional reaches (in that muscle onset latencies and handhold contact time were slower during volitional movement), the consistency in the relative sequence of muscle activation suggested that volitional and compensatory reaching shared common control elements [66]. 2) Relationship between the duration of the corrective phase and need for precise directional control: While directional control of the reaching trajectory may occur during the ballistic or early phase of the movement, the possible relationship between duration of the corrective phase and the need for more precise modulation of the reaching trajectory should also be considered. Specifically, prolonging the time between peak hand velocity and the time to object contact has been proposed as a strategy to execute greater corrective functions, in both volitional and compensatory reach-to-grasp movements [77, 66]. Supporting evidence for this apparent speed-accuracy trade-off in volitional reaching activity stems from a study that characterized the magnitude and temporal characteristics of peak hand velocity and aperture when reaching to grasp wooden blocks of varying width (0.5, 1.0, 1.5 and 2.0 cm) as quickly as possible [83]. The work revealed that the duration of the acceleration phase (i.e., before peak hand velocity) was consistent with object width [83]. However, the duration of the deceleration phase was prolonged substantially as the object width decreased (by roughly 50ms from widest to narrowest) [83], indicating a longer period of slower movement for objects that required more precision in the trajectory. In compensatory reach-to-grasp movements, Cheng and colleagues noted that the time to handrail contact after peak hand velocity showed statistically-significant increases (by ~10ms) when participants were relying purely on online visuo-spatial information (VSI) to grasp a handhold in an unpredictable location, compared to normal or stored VSI [77]. This was consistent with the need to correct for the reduced modulation of trajectory direction with handhold position early in the reach-to-grasp reaction with only online-vsi, and the increased deviation from straight-line trajectory [77].

35 Factors affecting the speed and accuracy of compensatory reaching The speed and accuracy of compensatory reach-to-grasp reactions may be affected by a number of factors, including: 1. Initial distance between the hand and the handhold: In rapid volitional and compensatory reaching, a few studies have reported increased movement or object contact times with increasing reach distance (e.g., [83, 3, 84]). In particular, peak reach velocity has also been observed to increase with reach amplitude during volitional reach-to-grasp movements [83, 84]. During compensatory reaching, Maki and colleagues noted that standing at a greater lateral distance from a handrail increased time to rail contact after perturbation onset, following destabilizations on a mock staircase [3]. For a person standing with their arms at their sides, handrail height will affect the initial distance between the hand and the handrail. If individuals adopt a reach trajectory with little deviation from a straight line (as has been observed in studies of volitional reaching, such as [85], [66] and [86]), one may reasonably expect that increased handrail height (relative to wrist height) would lead to increased time to handrail contact and possible increases to peak reach speed. However, previously-observed tendencies of individuals to quickly raise their arms after balance perturbations (akin to a generic, startle -like response) [64] could contribute error to the reach trajectory, and potentially delay time to handrail contact for lower handrails. Indeed, while the ability of young adults to quickly modulate their reach trajectory based on falling direction relative to the handrail is well-established, rapid upward hand movement was observed in these studies, at least in the initial stages of reaching [82, 66]. Accordingly, handrail height could plausibly influence the speed and trajectory error associated with compensatory reaching, but the specific effects are unknown. 2. Aging: While the speed and accuracy of compensatory reach-to-grasp movements are likely to vary with age, the specific effects are again difficult to predict. In a Fitts-like task of volitional reach-to-grasp movements with pens of varying diameters and distances away from participants, older adults demonstrated increased movement times, reduced peak velocities, and increased time from peak hand velocity to movement contact, compared to younger adults [84]. Pratt and colleagues reported similar increases in movement time and time from peak velocity to target contact in older adults compared to younger adults, following reaching movements to a visual target [87].

36 22 The extent to which these results from volitional reaching are reflected in age-related changes in compensatory reaching is less clear. Following upright stance perturbations, Cheng and colleagues noted time to handrail contact was slower in older adults than in younger adults (by ~30ms), independent of whether vision was occluded at some point in the trial [74]. However, a separate study exploring unexpected perturbations during gait reported that older adults did not differ significantly from younger adults in terms of time to grasp completion, despite demonstrating slower onset latencies [15]. This suggests that older adults may attempt to compensate for delayed arm muscle onset by reducing the error or increasing the velocity of some part of their reach trajectory. Indeed, Weaver and colleagues observed that older adults demonstrated significantly higher peak vertical wrist velocities than younger adults during compensatory grasping, following backward-directed falling from perturbations of upright stance [75]. 3. Presence of the handhold in the visual field: In a platform perturbation study of rapid reaching for an anteriorly-positioned handhold at varying lateral distances from the right arm, King and colleagues noted that handrail contact time and movement time were both generally faster when participants (all young adults) were looking directly at the handhold, compared to when they were looking at a screen in front of them (and thus reliant on peripheral vision) [88]. Further, greater deviation from a direct trajectory to the handhold was observed as the lateral distance of the handhold from the participant increased, in the peripheral vision conditions [88]. These results suggest that looking at the handhold can reduce handhold contact time (likely by improving trajectory accuracy), compared to relying on peripheral vision Grasp completion Following handrail contact, the hand must be oriented appropriately, with an aperture that allows the grasp to be completed. Grasp completion can be delayed as a result of collisions between the hand and the handrail, as well as by reaching past the handrail (i.e., overshooting ) before grasping [15]. A selection of factors that could affect grasp completion and grasping errors (e.g., collisions, reaching too far, or not reaching far enough (undershooting)) include: 1. Aging, fall status, and concurrent attention demands: A number of perturbation studies have reported increased frequency of handrail grasping errors in healthy older adults compared to younger adults (e.g., [89], [74]), with the caveat that statistically-significant differences were

37 23 not always observed due to the low number of errors in general [89]. Further, Westlake and colleagues noted that older adults who were classified as fallers (i.e., they had experienced at least one unintended fall in the last 12 months) demonstrated grasping errors in roughly three times as many trials as healthy older adults in the study while engaging in concurrent verb generation and verb-recall tasks [89]. Given that task and participant classification did not affect movement onset, the increased frequency of grasp errors in older fallers may imply limits in ability to shift from one cognitive task to another in this population [89]. 2. Cueing: In one study that explored the effect of handrail cueing types on handrail grasping in older adults, McKay and colleagues reported that the percentage of collision errors was halved when light-based cues (flashing green LED lights built into the rail) were used before perturbation onset, compared to no cue [61]. Further, a greater percentage of participants achieved a full power grip with the light-based cue [61]. 3. Holding an object: In a platform perturbation study that examined how holding an object (cane, cane top) in the right hand (proximal to the rail) affected handrail grasping in young adults, Bateni and colleagues observed that participants were significantly less likely to contact the rail when holding the object (compared to having their right hand free), even when the object did not contribute to balance recovery and the consequences of not grasping included falling into a safety harness [90]. Further, in cases where hand contact did occur, participants often only completed partial grasps (contacting the rail with one to three fingers, while continuing to hold the object) [90] Sufficiency of forces and moments applied to the handrail to stabilize the center of mass Once a person has grasped the handrail, they must then apply high forces and moments to the handrail to counteract the translational and rotational forces acting on the body, and stabilize their center of mass. This subsection begins with an overview of the inverted pendulum model of balance, which will help to explain how handrail height could affect grasping force and moment generation. Prior work on exploring handrail force and moment generation is then discussed. Finally, factors that could limit a person s ability to apply high forces and moments to a handrail are highlighted.

38 Mechanical models of balance control: the inverted pendulum The potential for handrail height to affect a person s ability to apply stabilizing forces and moments to the rail can be explored by considering an inverted pendulum model of balance control [91]. Maki and Fernie provide a detailed explanation in the context of stairway falls [6]. Briefly, when characterizing forward or backward falling, we consider three major components of motion [6]: 1. Rotation about the COM 2. Translation of the COM in the horizontal plane (mostly forward or backward) 3. Vertical translation of the COM (generally downward) Note that lateral translation of the COM also contributes to instability during falling in the forward, backward and lateral directions [92, 93], and previous characterizations of peak handrail forces following perturbations on stairs revealed high lateral force components (up to 45% BW) [3]. However, to simplify this explanation of how the inverted pendulum model applies to forward or backward falling, lateral movement components are ignored for now. With these components of motion in mind, we first consider early stages of falling where a stance foot is in place. For example, this could occur following a trip that results in forward or backward falling [10]. In this case, we treat the body as an inverted pendulum, with the fulcrum approximately around the ankle joint (see Figure 2.2). The role of the handrail is then to provide a stabilizing moment to counteract the destabilizing rotational force acting on the COM, as well as the existing forward and downward momentum of the COM [6]. It follows that a high stabilizing moment can be achieved by some combination of: (a) increasing the force applied to the handrail; or (b) increasing the moment arm by raising the handrail height. Conversely, in latter stages of falling (or during different types of falls, such as slips), it is no longer appropriate to model falling motion based on a fixed axis of rotation [6]. At this point, the feet may be moving (e.g., during slipping or stepping), or the knees and hips may have bent substantially. While ability to use the handrail to counteract rotational forces may remain important, the role of the handrail in counteracting the translational momentum of and forces acting on the COM increases. See Figure 2.3 for further details. Accordingly, factors such as a person s ability to apply high stabilizing forces to the handrail and the extent to which the handrail s design facilitates high force generation become critical for balance recovery.

39 25 point of rotation (assume position is fixed relative to the rail) W gravity acting on COM F A2 F A1 d W moment arm for downward force acting on COM W F V W F V h 1, h 2 handrail heights 1 and 2 R d W h h 2 1 R d W F A1, F A2 handrail reaction forces 1 and 2 (axial) d F_V, moment arm for handrail reaction forces (vertical) d F_V d F_V F V handrail reaction forces (vertical) destabilizing moment = W x d W stabilizing moment 1 = F A1 x h 1 + F V x d F_V stabilizing moment 2 = F A2 x h 2 + F V + d F_V R ground reaction force vector (vertical and horizontal components merged for clarity) To recover balance, the stabilizing moment from the handrail about the foot must counteract the destabilizing moment (due to body weight) and the rotational momentum of the body. Assuming that the person s anterior hand placement and vertical reaction forces (from the handrail) are consistent (for simplicity), the stabilizing moment can be modified by increasing the axial handrail force or by increasing the height of the handrail. Since h 2 > h 1, it follows that the destabilizing moment can be counteracted with a reduced F A2 compared to F A1 Figure 2.2: The inverted pendulum model applied to balance recovery via handrail grasping on level ground. Note that for a given destabilizing moment (assume the same in both images), the axial (anteroposterior) handrail force needed to counteract the destabilizing moment may decrease if the moment arm contributed by the handrail is longer. Only the stance leg is shown to simplify the image. Adapted from [6]. The magnitude and direction of the horizontal components of the ground reaction force vector will depend on the type of destabilization, the time during the gait cycle that the destabilization is measured, and other factors [94]. W R F V F H notes: F H and F V (components of handrail forces) counteract the downward translational and rotational force due to body weight, and the forward momentum of the COM. R represents the ground reaction force vector (not to scale), with horizontal and vertical components direction of BOS displacement Figure 2.3. Visual representation of balance recovery via handrail grasping when the axis of rotation is no longer fixed in terms of length or position, due to factors such as the foot sliding or knee collapsing. In this context, a person s ability to apply high forces to the handrail (impacted by posture and cross-sectional design) become important. The magnitude and direction of the ground reaction force vector will similarly depend on many factors, including shear forces between the foot and ground, the slope of the surface, and other factors [95]. Adapted from [4] and [6].

40 Factors that could affect ability to apply high stabilizing forces and moments to the handrail, and stabilize the COM In light of the representations of falling described above, a number of factors may influence the forces and moments that individuals can apply to handrails. 1. Handrail height: Younger and older adults have been previously reported to apply significantly-higher forward and backward volitional forces and moments to a handrail as its height increased, whereas ability to apply high downward pushing forces to the handrail decreased with increasing handrail height [6, 7, 8]. That handrail height affected maximum volitional force generation across the 10 inch range that was tested (32 to 42 inches) is likely unsurprising, and consistent with other work that explored isometric pull and push forces applied to handles with varying arm posture or handle height (e.g., [96, 97]). Balance control and recovery contexts should also be considered. To my knowledge, only two studies have explored forces that young adults applied to a handhold with varying handhold height: following simulated bus accelerations [98], and following constant-velocity perturbations (0.16m/s, 0.33m/s and 0.5m/s) with a duration of 200ms [99]. In the case of simulated bus accelerations, Sarraf and colleagues explored concomitant peak resultant handhold forces, COP excursions, and muscle activation in the legs and arms, while holding: (a) a vertical handhold at shoulder height, or (b) a horizontal handhold overhead [98]. They noted that grasping a handhold at shoulder height resulted in increased peak handhold forces and decreased COP displacement compared to grasping overhead [98]. Further, integrated muscle activity in the biceps and triceps was significantly higher for overhead grasping during forward and backward perturbations [98]. This implies a stability advantage with the rail at shoulder height, which allowed participants to reduce COP excursion with lower muscle activity in the arms, and apply higher handrail forces. Separately, in a study that explored hand-in-place reactions with three handhold heights (approximately elbow, shoulder and eye-height), the lowest handhold resulted in the highest peak applied handhold forces for backward-directed falling [99]. COP excursions were not affected by handhold height, demonstrating that the higher tested heights enabled comparable COP stability with reduced handhold forces [99]. Taken together, the findings described above suggest possible stability advantages with handholds at shoulder-height (compared to elbow-height or overhead) following mild balance perturbations. However, results should be considered with caution due to the apparent absence

41 27 of accounting for handhold forces in the COP calculation, and due to the relatively-low handhold force magnitudes in [99] (mostly < 20N for posterior perturbations) that make the functional role of the handhold in balance control difficult to discern. Further, these studies did not address lower handrails (i.e., below elbow-height in adults) that are required in many building codes (e.g., [100, 101]) and that are more common in the community. Despite these limitations, results are consistent with studies of isometric pulling forces applied to handles that revealed that volitional force generation ability decreases with very-high handles (1.75m, compared to 1m) [102]. 2. Posture and strength: The forces that persons can apply to handrails during balance recovery likely depend on interactions between reaching posture (e.g., shoulder flexion and abduction during grasping) and individual strength. Of note, Hughes and colleagues explored how normal, isometric shoulder strength relates to posture, considering the maximum volitional torques that men and women (20 to 78 years) could apply to a modified Cybex II dynamometer [103]. They noted that the highest shoulder flexion torques were achieved when the shoulder was closer to a neutral position (i.e., upper arm parallel to the major axis of the trunk) [103], which may favour lower handrails for applying high stabilizing push forces to the rail in the anterior direction during forward falling. However, this hypothesis should be considered with caution, due to the increased possible need to apply high shoulder flexion torques with lower handrails if the elbow is fully extended in order to stabilize the COM of the body, as the length of the moment arm between the handrail and the shoulder will necessarily be higher with the elbow fully-extended. It follows that higher torque capacity in the shoulder may be necessary to stabilize the COM with a low handrail, compared to higher rails with a shorter moment arm between the handrail and the shoulder. In contrast with flexion torques, shoulder extension torques increased as shoulder flexion angle increased, and shoulder adduction torques increased as shoulder abduction angle increased [103]. This suggests that ability to apply high pulling forces to the handrail (in the posterior direction, and pulling toward the body in the medial-lateral axis) may be favored with higher rails that can be reached with high flexion and abduction angles. Beyond the shoulder, elbow torque generation has been explored in a few studies. Of note, Mukhopadhyay and colleagues characterized the influence of forearm rotation and elbow angle on elbow torque capacity in 36 male university students [104]. They found that forearm

42 28 pronation (60%) was disadvantageous for elbow torque generation, compared to when the forearm was in a neutral position which may be problematic for applying upward forces to a handrail [104]. However, elbow torque generation with the forearm pronated was maximized with the elbow at 90 o, compared to 135 o or 45 o [104]. In cases where forearm rotation was neutral, elbow flexion angles of 135 o resulted in slight (6%) increases to maximum voluntary torque generation, compared to elbow flexion angles of 90 o [104]. However, substantial elbow flexion (45 o angles) decreased torque production by nearly 25%, relative to elbow flexion angles of 90 o [104]. This suggests that the benefits of higher handrails for maximising shoulder torque capacity may be offset by reduced elbow pulling torque capacity, though the extent to which these are offset is unknown. 3. Sex: Hughes and colleagues work also revealed that, on average, participating women applied lower maximum volitional isometric shoulder torques to the dynamometer than did men, when controlling for age and mass [103]. 4. Aging: In the maximum volitional force generation studies described above, Maki and colleagues noted that, on average, younger adults applied forces to the handrail that were approximately 1.5 to 2 times as high as those applied by older adults, when normalized to %BW [6, 7]. This result is consistent with age-related declines in strength, muscle power, and muscle mass [105, 106, 107], which would be expected to limit ability to both apply high volitional forces, and sustain these forces over the 4s trial period in Maki et al s study protocol [6, 7]. Hughes and colleagues similarly noted age-related declines to isometric shoulder torques in each of flexion, extension, abduction and adduction [103]. It should be noted that Maki and colleagues early work focused on the slow application of high volitional forces. In balance recovery contexts via rapid handrail grasping, the speed with which high stabilizing forces can be applied to the handrail is also important. While both type-1 ( slowtwitch ) and type-2 ( fast-twitch ) muscle fibres have been reported to decline in number with age, the cross-sectional area of remaining type-2 fibres has been observed to also decline in the vastus lateralis in men [108, 109]. More recent work involving biopsies of deltoid muscles revealed similar age-related declines in total type-2 muscle fibre area, with decreases in the number of fibres in men, and decreases in individual fibre diameter in women [110]. Accordingly, the consequences of loss of muscle mass and strength may be amplified for older adults when recovering balance from large perturbations, should they be unable to quickly apply

43 29 the high forces to the handrail to counteract the momentum of the COM. Age-related reductions to rate of torque generation have been observed in the ankles and hips [111, 112], and one may reasonably expect that such declines could manifest in the upper limbs as well. 2.3 Synthesis It is clear that many factors can influence a person s ability to use a handrail for balance recovery. The inverted pendulum model provides strong theoretical support for possible stability advantages at a given force level with increasing handrail height; however, the extent to which this support translates to compensatory contexts and to reach-to-grasp reactions, in particular has not yet been explored. To evaluate the applicability of the inverted pendulum model to compensatory grasping, we need to consider both kinematic (e.g., COM and trunk control) and kinetic (e.g., concomitant forces applied to the handrail) measures. COM control can provide insights into the role of the handrail in arresting translational movement of the body, while trunk angular kinematics can reveal how the rail helps to reduce rotational movement of the body. While the head-arms-trunk segment does not include the lower limbs (and is thus not a perfect representation of the pendulum), it still accounts for well over half of body mass [113] in many adults. Accordingly, a person s ability to control the trunk is likely important in successful recovery from balance loss. Conversely, a person s tendency to take a step serves as a global measure of balance recovery with the handrail. Taken together, these three kinematic measures should provide us with a detailed picture of the role of the handrail in stabilizing the body, while accounting for potential differences in balance recovery strategy that stem from perturbations in different directions. Quantifying the concomitant handrail forces will help us to understand individual reliance on the handrail to regain stability. Despite the high prevalence of falls while walking [10], the absence of studies evaluating compensatory grasping during ongoing gait was notable. Beyond a limited number of published papers on first-trial, unexpected perturbations (e.g., [16, 15, 61]), studies related to compensatory grasping appear to be constrained to upright-stance and seated perturbations. Given the apparent delays to initiating arm reactions during leg movement [14], and the possibility of compromised ability of older adults to control limb trajectories when walking [31], an evaluation of the effect of handrail height on balance recovery during gait is warranted. On this basis, this evaluation should consider age-related changes in the speed and accuracy of the reaching trajectory, as well as the

44 30 impact on trunk kinematics during balance recovery. Reach trajectory kinematics with varying handrail height are of particular interest, given the possible conflict between previously-observed generic, startle -like responses to perturbations that consist of raising the hands very quickly [64] (which would favour higher handrails from a speed perspective), versus other studies that report early modulation of the reach trajectory following perturbations (e.g., [36, 66, 82]) that may increase a person s ability to quickly reach for rails at a lower initial distance from the rail (i.e., wrist-height). Conversely, quantifying the relationship between handrail height and trunk stability following perturbations during gait will help to shed light on whether findings from upright stance perturbations are observed in broader contexts. Finally, the paucity of studies that report handrail forces during compensatory grasping was apparent. To my knowledge, only one published study depicted peak handrail forces in multiple axes during reach-to-grasp reactions (four young men experienced balance perturbations on a mock staircase), with one participant applying up to 61% of his body weight to the rail in the highest observed resultant handrail force in the study [3]. A handful of other papers report handhold forces during reactions where the person s hand was already touching the handhold (e.g., [99, 98]). However, the low force magnitudes (150N being the highest resultant force in [98], and reported handhold forces mostly < 20N in [99]) suggests that either contacting the handrail can help to reduce handrail forces for maintaining balance, or that the protocols may not have sufficiently threatened balance to necessitate high handrail forces for stabilizing the COM. Given the significant relationship between handrail height and maximum volitional force generation ability [7], and the plausible importance of applying high handrail forces in multiple Cartesian axes to recover balance [3], more rigorous exploration of handrail forces during balance recovery and the dependence of these forces on handrail height and starting posture (reach-to-grasp versus proactively contacting the rail) is warranted.

45 31 Chapter 3 Study 1: Use of handrails for balance recovery Characterizing loading profiles in younger adults and effects of handrail height Preface The content of this chapter was submitted to the journal Applied Ergonomics in May I have made minor modifications to the framing of the original manuscript to improve its fit in my thesis. The co-authors are Konika Nirmalanathan, Dr Emily King, Dr Brian Maki, and Dr Alison Novak. I conceptualized the study, collaborated to develop the protocol and collect data, analyzed data, and wrote the first draft of the manuscript. Abstract Well-designed handrails significantly enhance balance recovery by allowing users to apply high forces to the rail and stabilize their center of mass. However, data on user-applied handrail forces during balance recovery are limited. We characterized peak forces that 50 young adults applied to a handrail in response to forward and backward platform perturbations; quantified effects of handrail height (34, 38, 42 inches) and position prior to balance loss (standing beside the rail with or without hand contact, or facing the handrail with two-handed contact); and examined the relationship between handrail forces and individual mass. We also characterized the highest perturbation magnitude (up to the physical limits of the platform) that participants could withstand without stepping or falling ( maximum withstood perturbation, MWP), while using the handrail for balance recovery. Participant MWP increased significantly as handrail height increased for both falling directions. The highest handrail forces were applied when participants faced the handrail and grasped with two hands. In these cases, increased handrail height was associated with increased anterior forces and decreased downward, upward and resultant forces. Peak handrail forces in most axes correlated strongly with individual weight. Implications of these findings for handrail design are discussed.

46 Introduction Falls remain the leading cause of unintentional activity-limiting injury among North American adults [2, 114]. Handrails can significantly improve people s ability to recover from balance loss and avoid a fall [3]. By grasping a handrail, a person can anchor their body and apply substantial forces and moments to stabilize their centre of mass [4]. Despite the widespread presence of handrails, our understanding of the forces that users apply to them during balance recovery is limited. Knowledge of these forces and how they relate to kinematic balance control is important for understanding how handrail forces contribute to balance recovery. Previously, Maki and colleagues characterized the highest volitional forces and moments that younger and older adults could apply to stairway handrails, and quantified how these forces varied across a range of handrail heights [7, 8]. Notably, they found that as handrail height increased (within the range of ), participants applied higher maximum volitional forces in the anterior and posterior directions, and lower forces in the downward direction [7]. However, these studies focused on the slow application of voluntary handrail forces while participants stood still forces during dynamic balance recovery scenarios are likely to be higher. Applied handhold forces following balance perturbations have been explored in only two published studies: simulating handhold use while riding a bus [98], and during a simulated stair slip [3]. Sarraf et al. reported the peak resultant forces that younger adults applied to proactively-grasped overhead and shoulder-height handholds during simulated bus accelerations [98]. Maki and colleagues reported the peak forces that participants applied to a 36 inch (91cm) handrail during reach-to-grasp and proactive-contact balance reactions following simulated oversteps on a mock staircase [3]. They demonstrated that participants quickly applied high forces during reach-to-grasp reactions (up to 61% of body weight), and that mean upward-normal and medial-lateral forces differed between reach-to-grasp and proactive-contact reactions. While this work provided valuable data on handrail forces during balance recovery, the single handrail height and small sample (four men) limit the generalizability of the findings. A further key consideration lies in relating applied handrail forces to a user s weight. While voluntary forces applied to handrails and similar aids (e.g. grab bars, grab poles) are often reported as a percentage of individual body weight (e.g. [7, 115, 116]), to our knowledge the relationship between individual weight and compensatory handrail forces has not yet been established. For fixtures like these,

47 33 designed to provide substantial balance support, it is important to understand the extent to which a user s weight predicts the forces that the device must be able to support. It is clear that handrails and other handholds sustain substantial forces exerted by users during challenging balance recovery situations, including both reach-to-grasp and proactive-contact reactions. More comprehensive data on applied handrail forces during balance recovery are needed. To address this gap, this study characterizes the forces that younger adults applied to a handrail while recovering from substantial forward and backward balance disturbances specifically, the maximum withstood perturbation (MWP), or the highest perturbation magnitude that the participant can withstand during balance recovery with a handrail without stepping for falling. This study also quantifies the effects of handrail height and the participant s starting posture on peak handrail forces and MWP. Finally, it determines the extent to which peak applied forces correlate with body weight. 3.2 Materials and Methods Participants Fifty young adults (25 males) participated. Participant demographics are presented in Table 3.1. All participants reported being free of neurological, vestibular and musculoskeletal disorders. Approval was granted by both university and hospital Research Ethics Boards. Participants provided informed consent prior to study commencement. Table 3.1: Participant demographics Age (years) Height (cm) Weight (kg) Grip strength (kg) Mean (SD) 25.0 (3.8) (9.8) 75.0 (18.1) 43.8 (13.1) Range Testing environment Data were collected in the Challenging Environments Assessment Laboratory (CEAL) at Toronto Rehabilitation Institute-University Health Network. CEAL consists of a 5m x 5m laboratory secured to a robotic platform that translated in the horizontal plane to disrupt participant balance (Figure 3.1a). A height-adjustable horizontal handrail (unpainted aluminum tube; outer diameter 1.5 inches

48 34 (38mm)) was instrumented with two tri-axial load cells (AMTI MC3A-1000; Advanced Medical Technology, Inc, Watertown, MA) to collect forces applied during balance recovery. The handrail was braced to minimize vibration (Figure 3.1b). For safety, participants wore an overhead harness that prevented a fall to the ground, although sufficient slack was left in the harness line to allow participants to move naturally. Participants wore knee guards as added protection from possible impact with the floor, and an elbow guard on the dominant arm to reduce impact in case of contact with the testing apparatus. (a) (b) Figure 3.1: Experimental environment. (a) The Challenging Environments Assessment Laboratory: a lab mounted on top of a robotic platform that can deliver perturbations. (b) Inside the lab: a participant is standing beside the braced handrail while wearing a safety harness, knee guards, and an elbow guard. Foam blocks were positioned in front of and behind participants to discourage reactive stepping Protocol After they provided informed consent, participants height and weight were measured. Dominant hand grip strength was measured using a handheld dynamometer (Jamar Hydraulic Hand Dynamometer; Patterson Medical, Mississauga, ON) according to a protocol recommended by the American Society of Hand Therapists [117]. Mean values from three trials are reported in Table 3.1. Participants wore standardized athletic shoes to mitigate possible effects of footwear on balance Perturbation design Balance was perturbed by sudden platform translations. Both backward and forward falling motions were tested. Backward falling was induced with forward platform translations; forward falling resulted from backward platform translations. Perturbation timing and direction were randomized to mitigate the potential for anticipatory behavior and pre-planning of movements. Each perturbation consisted of a square-wave acceleration profile (300ms acceleration pulse, then an equal and opposite deceleration). Perturbation magnitudes are described in Sections and

49 Facing- Rail Hand-in- Place Reach-to- Grasp 35 Participants experienced several perturbation trials prior to data collection to accustom them to the platform movement and protocol Perturbation protocol The testing conditions included three starting positions, three handrail heights, and two perturbation directions (Figure 3.2a). Perturbation magnitude was varied systematically to find the Maximum Withstood Perturbation (MWP, the highest-magnitude perturbation for each combination of falling direction, handrail height and starting position that the participant could withstand without failure see below for additional details). Participants were instructed to use the handrail to recover balance and to avoid stepping or falling into the harness. The three starting positions were: 1) Reach-to-Grasp: Participants stood erect beside the handrail with their arms at their sides. Their feet were approximately shoulder-width apart, with their centre-line 58% of their arm length (measured from the middle fingertip to the acromion) away from the handrail (Figure 3.2b). This distance approximates the distance between the tip of the middle finger and the elbow (adapted from [118]). Participants were instructed to reach to grasp the handrail with their dominant hand as quickly as possible following initiation of the platform perturbation. 2) Hand-in-Place: Participants stood as described for Reach-to-Grasp trials, except that their dominant hand rested lightly on the handrail. 3) Facing-Rail: Participants stood facing the handrail, with their heels positioned 58% of their arm length away from the handrail (Figure 3.2c). Both hands rested lightly on the handrail. (a) Forward fall (backward platform translation) (b) d Backward fall (forward platform translation) (c) LOW MED HIGH d Figure 3.2: (a) Summary of starting positions, fall directions and handrail heights tested in this study. The arrows signify direction of platform movement. (b) Schematic diagram (top view) of participants distance from the handrail for Reach-to-Grasp and Hand-in-Place trials, with d = 58% of the participant s arm length. (c) Schematic diagram (top view) of participants distance from the handrail for Facing-Rail trials, with d = 58% of the participant s arm length.

50 36 For each starting position, three handrail heights were tested: 34 inches/86 cm (LOW), 38 inches/96 cm (MED), and 42 inches/106 cm (HIGH), all measured from the floor to the top of the handrail. These heights were based on current building code provisions, to match the heights at which handrails were likely to be installed in the community: 34in and 38in approximate the lower and upper boundaries of the International Building Code provisions for handrail height on stairs and ramps [101] and the Ontario Building Code provisions for stairs, ramps and care facility corridors [100]; 42in approximates the maximum height of handrails built into guards on stairway landings in Ontario [100]. This resulted in a total of nine blocks of trials, comprising the different starting positions and handrail heights. Within each block of trials, both forward and backward falling motions were elicited in a randomized sequence. The presentation order of the handrail heights was randomized. For each height, all participants began with Reach-to-Grasp trials. While standing beside the handrail as described above, participants completed a concurrent counting task (counting backwards by a number from two to nine, from a randomly-selected start number) to distract their attention. Foam blocks positioned in front of and behind their feet discouraged reactive stepping. All participants initially experienced a low-magnitude perturbation (platform acceleration 2.5m/s 2 ). Following each successful trial, the perturbation magnitude was increased in 0.5 m/s 2 increments until either (a) the physical limits of the platform were reached (5 m/s 2 ), or (b) participants failed two trials for a given perturbation magnitude and direction. Failure was defined as taking a step in the direction of falling motion, or contacting the rail or foam blocks with the non-reaching hand. If participants were unable to successfully complete Reach-to-Grasp trials with the starting acceleration (2.5 m/s 2 ), the perturbation magnitude was lowered in 0.5 m/s 2 increments until participants were able to complete Reach-to-Grasp reactions successfully. The perturbation magnitude was then increased again until participants failed two trials for each fall direction. Note that if the step or handrail contact with the non-reaching hand occurred before the participant experienced twelve perturbations, the trial was not identified as failed ; participants simply repeated the trial at the same perturbation level. Participants were given up to twelve perturbations before trials were identified as failed, to become accustomed to the perturbation protocol [77]. After completing Reach-to-Grasp trials, participants completed the Hand-in-Place and Facing-Rail trials in a similar manner to the protocol described above. However, for the Hand-in-Place and Facing-Rail conditions, the initial perturbation magnitude was 1 m/s 2 lower than the magnitude of

51 37 their last failed trial in each direction during the previous testing condition. For example, if a participant failed at 4 m/s 2 for forward falls and 3.5 m/s 2 for backward falls during Reach-to-Grasp trials at a given rail height, (s)he began the corresponding Hand-in-Place trials with perturbation magnitudes of 3 m/s 2 for forward falls, and 2.5 m/s 2 for backward falls. Perturbation magnitudes were increased in 0.5 m/s 2 increments, as described above. After completing the above protocol for each starting position at a given handrail height, participants sat and rested while the handrail height was changed. Participants also rested between trial blocks (starting positions) to mitigate fatigue Data processing Force data were collected at 1000Hz. Inertial artifacts due to platform movement were removed by subtracting load cell data collected from platform accelerations in which there was no handrail contact. Force signals were digitally filtered with a second-order, zero-lag Butterworth filter with a cut-off frequency of 20Hz. Only data from each participant s MWP trial for each testing condition were analyzed. The MWP represents the situation of highest demand for each individual when using the handrail in each condition. Dependent variables computed from each MWP trial were the highest force magnitudes that participants applied to the rail in each direction: anterior, posterior, medial (pulling toward the body), lateral (pushing away from the body), downward, upward, and resultant (Figure 3.3, in the Results sub-section). Forces are reported as a percentage of the participant s body weight (%BW) or in Newtons Statistical analysis Descriptive statistics (means, standard deviations) were calculated for all variables of interest. Repeated measures ANOVAs with two within-subject factors (handrail height, starting position) were performed to characterize the effects of these factors on peak handrail forces (SAS version 9.4, Cary, NC). Post hoc comparisons with Tukey adjustments were performed where significant main or interaction effects were identified. A significance threshold of p<0.05 was used for all analyses. Forward and backward falling motions were explored separately. To examine the influence of body mass on handrail forces, the correlation between participant weight and peak applied force was analyzed for conditions in which the mean peak forces were substantial (>20% BW). The relationship between participant weight and peak horizontal and

52 38 vertical forces was modeled using linear regressions, for the conditions that resulted in the highest peak horizontal and vertical forces. The two heaviest participants (135.0 and kg versus 45 to 106 kg for the remaining 48 participants) disproportionately biased these estimates, and were thus excluded from the regressions. 3.3 Results All participants completed testing without incident. Five of the tallest participants (heights cm, corresponding to >95 th percentile female and >85 th percentile male heights [117]) were excluded from the Facing-Rail, LOW handrail, forward falling motion for safety reasons: they could not reach the LOW handrail position while standing erect, so the risk of injury (due to collision with the apparatus) if they failed to grasp the rail was deemed to be unnecessarily high. In the Facing-Rail, HIGH handrail condition, analysis was restricted to 49/50 participants due to problems with the data acquisition system Descriptive statistics of maximum withstood perturbations (MWPs) Descriptive statistics of participant MWPs are presented in Table 3.2. Notably, mean MWP magnitudes generally increased with increasing rail height for Reach-to-Grasp and Hand-in-Place reactions. The highest MWPs were with Facing-Rail trials, where only one participant failed to reach the acceleration limits of the platform in one condition (LOW, backward falling). Hand-in-Place MWP magnitudes generally exceeded those of Reach-to-Grasp reactions. Table 3.2: Descriptive statistics (Mean (SD)) of the Maximum Withstood Perturbation magnitudes for each fall direction, initial hand position, and handrail height. FF represents the forward falling motion; BF represents the backward falling motion. All units are in m/s 2. Fall Reach-to-Grasp Hand-in-Place Facing-Rail direction LOW MED HIGH LOW MED HIGH LOW MED HIGH FF 3.88 (0.79)* BF 3.70 (0.88)* 4.20 (0.75)* 4.03 (0.85)* 4.26 (0.80) 4.17 (0.81) 4.55 (0.62)* 4.45 (0.70) 4.80 (0.40)* 4.66 (0.61) 4.85 (0.37) 4.88 (0.34) 5.00 (0.00) 4.99 (0.07) 5.00 (0.00) 5.00 (0.00) 5.00 (0.00) 5.00 (0.00) N.B. Repeated measures ANOVAs with two within-subject factors (handrail height, initial hand position) were performed to quantify the effect of these factors on MWP magnitude. Facing-Rail trials were excluded from this analysis, as participants reached the acceleration limits of the platform for all but one trial. Significant main effects of handrail height and hand position were found for both fall directions (p<0.001 in all cases). Pairwise comparisons with Tukey corrections were performed to quantify the effect of handrail height on MWP within each hand position and fall direction combination. * p<0.05 for LOW-MED comparison within a given hand position and fall direction p<0.05 for LOW-HIGH comparison within a given hand position and fall direction

53 Characteristic handrail force profiles and peak handrail forces Characteristic handrail force profiles for each perturbation direction and starting position are presented in Figure 3.4. Figure 3.5 depicts peak handrail forces as a percentage of body weight (%BW) in each axis; Table 3.3 and Table 3.4 provides precise numbers for each condition, for forward and backward falling respectively. Repeated-measures ANOVA of the %BW data for both falling directions demonstrated statisticallysignificant main effects due to hand position for all six force components and resultant forces (p s<0.001). There were also significant main effects of handrail height in all cases (p s<0.036) except posterior force during forward falling (p=0.92). However, just over half (8/14) of the analyses also showed significant position-height interactions (p's<0.018). A more detailed description of the findings, including pairwise comparisons of means (where appropriate), is provided below. Exact p- values and descriptive statistics can be found in Table 3.3, Table 3.4 and Figure Influence of starting position Starting position significantly affected peak handrail forces (Figure 3.5). Of note, Facing-Rail reactions consistently resulted in significantly higher peak handrail forces than Reach-to-Grasp or Hand-in-Place reactions. For forward falling motions, participants applied significantly higher peak anterior, downward and resultant forces to the rail for Reach-to-Grasp reactions than for Hand-in- Place reactions (Figure 3.5a, c, g). In contrast, backward falling motions resulted in significantly higher peak handrail forces in the posterior, medial and resultant directions for Reach-to-Grasp compared to Hand-in-Place reactions (Figure 3.5b, f, g). While starting position significantly affected MWPs in the medial and lateral directions, the forces are quite low (<20%BW) in all cases except medial forces during backward falling.

54 to o handrail (N) to o handrail (N) Force applied to handrail (N) Force applied to handrail (N) Force applied to handrail (N) Handrail force (N) Handrail force (N) Handrail force (N) Force applied to handrail (N) 40 Fz (+) Fz (+) Fy (+) Fx (+) (a) (b) (c) Fx (+) Figure 3.3: Definitions of handrail loading axes. (a) and (b) illustrate handrail force axis conventions for the Reach-to-Grasp and Hand-in-Place trials, while (c) provides the conventions for the Facing-Rail trials. All force axis conventions denote the force applied by the participant to the rail. In all cases, Fx (+) represents anterior (forward) force; Fx (-) represents posterior (backward) force; Fy (+) represents lateral (pushing away from torso) force; Fy (-) represents medial (pulling toward torso) force; Fz (+) represents downward force; and Fz (-) represents upward force. 300 Forward falling motion (a) 200 Backward falling motion (b) Reach-to-Grasp - FW fall Reach-to-Grasp - BW fall Fx A-P A-P Fy Lateral Late Fz Vertical Vert (c) (d) Time relative to perturbation onset (ms) 200 Time relative to perturbation onset (ms) Hand-in-Place - FW fall 100 Hand-in-Place - BW fall A-P A-P Lateral Late Vertical -200 Vert (e) (f) Time relative to perturbation onset (ms) 200 Time relative to perturbation onset (ms) Facing-Rail - FW fall -100 Facing-Rail - BW fall A-P -700 A-P Lateral Late -700 Time 0relative 1000 to perturbation 2000 onset 3000 (ms) Vertical -700 Time 0relative 1000 to perturbation 2000 onset 3000 (ms) Vert Figure 0 3.4: Typical patterns of force generation along all three Cartesian 0 axes, during (a) Reach-to- Grasp Forward falling motion; (b) Reach-to-Grasp Backward falling motion; (c) Hand-in-Place Forward falling motion; (d) Hand-in-Place Backward falling motion; (e) Facing-Rail Forward falling motion; and (f) Facing-Rail Backward falling motion. All traces were collected from one participant, while the rail was LOW.

55 Resultant force (%BW) Lateral force (%BW) Downward force (%BW) Upward force (%BW) Anterior force (%BW) Posterior force (%BW) 41 (a) (b) 100 α β γ β γ 100 β γ α β γ 80 * (c) RG HP FR RG HP FR RG HP FR RG HP FR Forward Falling Backward Falling Forward Falling Backward Falling p(low-high)=0.01 (d) α β γ β γ β γ β γ * 100 * 80 * * RG HP FR RG HP FR 0 RG HP FR RG HP FR (e) Forward Falling Backward Falling (f) Forward Falling Backward Falling α α 100 α α RG HP FR RG HP FR Forward Falling Backward Falling p(low-high)=0.05 p(low-high)=0.01 Medial force (%BW) RG HP FR RG HP FR Forward Falling Backward Falling p(low-high)=0.01 (g) α β γ α β γ * RG HP FR RG HP FR Forward Falling Backward Falling p(low-high)<0.01 Handrail height legend LOW MED HIGH Starting position effects (non-interacting comparisons) α p<0.05 for RG-HP comparison β p<0.05 for RG-FR comparison γ p<0.05 for HP-FR comparison Handrail height effects (interacting comparisons) * p<0.05 for LOW-MED interaction comparison p<0.05 for LOW-HIGH interaction comparison p<0.05 for MED-HIGH interaction comparison Figure 3.5: Peak handrail forces for each condition and load axis: a) anterior; b) posterior; c) downward; d) upward; e) lateral; f) medial; and g) resultant forces. Individual bars denote mean forces for the condition; error bars denote standard deviations. RG, HP and FR represent Reach-to-Grasp, Hand-in-Place and Facing-Rail reactions respectively. Medial and lateral force directions were not defined in the FR condition, as all sideways forces in this condition were negligible. Statistical analyses indicate the effect of starting position on peak handrail forces for each fall direction and load axis. Where significant interaction effects were not found, the significant handrail height comparison p s are indicated in the graphs under the horizontal axis. Where significant interactions were present, interaction pairwise comparisons revealed effects of handrail height within each starting position (significant comparisons indicated with symbols in graphs above).

56 Table 3.3: Descriptive statistics of peak handrail forces during balance reactions in the forward falling direction. All force units are in %BW. Reach-to-Grasp Hand-in-Place Facing-Rail p p p (Height x Load axis Low Med High Low Med High Low Med High (Height) (HandPos) HandPos) Anterior Mean (SD) 20 (6) 21 (6) 22 (6) 19 (5) 20 (4) 19 (3) 42 (10) 45 (10) 50 (9) <0.001 <0.001 <0.001 Range Posterior Mean (SD) 3 (2) 4 (3) 4 (3) 3 (3) 5 (4) 6 (5) 62 (21) 61 (21) 59 (18) < Range Down Mean (SD) 17 (7) 17 (6) 17 (6) 10 (6) 10 (4) 9 (3) 71 (13)* 53 (13)* 33 (12) <0.001 <0.001 <0.001 Range Up Mean (SD) 6 (8) 6 (6) 4 (5) 9 (9) 6 (5) 4 (3) 62 (20)* 44 (17)* 28 (13) <0.001 <0.001 <0.001 Range Lateral Mean (SD) 6 (4) 6 (5) 7 (4) 5 (3) 5 (3) 5 (3) ǂ < Range Medial Mean (SD) 10 (6) 11 (7) 12 (7) 9 (5) 9 (4) 9 (4) < Range Resultant Mean (SD) 27 (9) 27 (8) 28 (8) 24 (7) 22 (5) 21 (4) 95 (21)* 80 (20)* 70 (18) <0.001 <0.001 <0.001 Range NB: The medial and lateral load axes were not defined for the Facing-Rail condition, and are thus excluded from analysis. Mean sideways forces in the Facing-Rail condition were negligible (<3% BW). ǂ p=0.05 for Low-High comparison (where main effect of handrail height observed but no significant interaction effects found). In these cases, no significant comparisons between Low-Medium or Medium-High were found. * p<0.05 for Low-Medium interaction comparison within the hand position p<0.05 for Low-High interaction comparison within the hand position p<0.05 for Medium-High interaction comparison within the hand position 42

57 43 Table 3.4: Descriptive statistics of peak handrail forces during balance reactions in the backward falling direction. All force units are in %BW. Reach-to-Grasp Hand-in-Place Facing-Rail p p p (Height x Load axis Low Med High Low Med High Low Med High (Height) (HandPos) HandPos) Anterior Mean (SD) 2 (2) 2 (2) 2 (1) 3 (3) 4 (4) 4 (4) 25 (15)* 34 (15)* 40 (15) <0.001 <0.001 <0.001 Range Posterior Mean (SD) 38 (11) 40 (12) 40 (11) 31 (8) 32 (7) 34 (6) 60 (17) 62 (13) 64 (12) ǂ < Range Down Mean (SD) 13 (8) 14 (7) 13 (6) 14 (12) 13 (9) 11 (6) 43 (24) 40 (18) 28 (14) <0.001 <0.001 <0.001 Range Up Mean (SD) 24 (9) 20 (8) 12 (7) 25 (8)* 19 (7)* 13 (6) 46 (18)* 38 (12)* 25 (10) <0.001 <0.001 <0.001 Range Lateral Mean (SD) 9 (15) 7 (6) 5 (4) 14 (16) 12 (14) 9 (6) ǂ < Range Medial Mean (SD) 27 (7) 29 (9) 29 (8) 19 (6) 20 (6) 21 (5) ǂ < Range Resultant Mean (SD) 53 (13) 52 (13) 50 (12) 44 (11) 43 (12) 41 (8) 78 (20) 73.3 (14.6) 70 (13) <0.001 ǂ < Range NB: The medial and lateral load axes were not defined for the Facing-Rail condition, and are thus excluded from analysis. Mean sideways forces in the Facing-Rail condition were negligible (<3% BW). ǂ p<0.05 for Low-High comparison (where main effect of handrail height observed but no significant interaction effects found). In these cases, no significant comparisons between Low-Medium or Medium-High were found. * p<0.05 for Low-Medium interaction comparison within the hand position p<0.05 for Low-High interaction comparison within the hand position p<0.05 for Medium-High interaction comparison within the hand position

58 Influence of handrail height The most substantial effects of handrail height on peak handrail forces were observed during Facing-Rail reactions. In these cases, and in both fall directions, peak anterior force increased significantly with increasing handrail height (Figure 3.5a). Conversely, peak downward, upward and resultant handrail forces decreased significantly as handrail height increased (Figure 3.5c, d, g). For Reach-to-Grasp and Hand-in-Place reactions, the most notable influence of handrail height was for upward forces during backward falling motions, which decreased significantly and substantially as handrail height increased (Figure 3.5d). Although significant main effects showed that handrail height affected almost all components of peak handrail forces in both fall directions, the remaining effects, summarized in Table 3.3, Table 3.4 and Figure 3.5, were comparatively small Influence of user weight on peak handrail forces We characterized the influence of user weight on peak handrail forces for the worst-case loading conditions (i.e. the handrail height, starting position and fall direction combination that resulted in the highest mean peak forces; Table 3.3 and Table 3.4) for each force axis of interest. Table 3.5 shows Pearson correlation r-values and linear regression equations to model the relationship between participant mass and peak handrail force for the worst-case loading conditions. All of the worst-case loading scenarios for the handrail resulted from Facing-Rail reactions (note that the handrail loading directions of the medial and lateral forces recorded in the other starting positions correspond to anterior and posterior forces during Facing-Rail reactions). In general, participant mass and peak handrail forces were strongly correlated (r >0.6) [120] in all of the worstcase loading scenarios, with the exception of anterior, posterior and upward loads when the handrail was low (r = 0.547, and 0.491, respectively). Figure 3.6 highlights the relationship between peak handrail forces and individual weight. Comparisons to the recent ISO 17966:2016 structural strength standards for grab bars and removable grab rails [121] can be seen in Figure 3.6. Some participants applied peak posterior forces that exceeded the ISO standard for horizontal loading of grab bars at all tested handrail heights (Figure 3.6). Many participants also applied peak downward and upward forces that 44

59 45 exceeded the ISO static strength standard for vertical loading of removable grab rails, although this was predominately for LOW trials (Figure 3.6 c,d). No participants applied peak vertical forces that surpassed this standard when the rail was HIGH. Table 3.5: Pearson s r-values (p-values in parentheses) and linear regressions describing the relationship between individual weight (kg) and peak handrail forces (N) for the loading axes and fall directions that resulted in the highest mean peak loads applied to each loading axis. All relationships below are for Facing-Rail trials, as this condition resulted in the highest applied loads compared to Reach-to-Grasp and Hand-in-Place conditions. The worst-case loading conditions for each force axis (i.e. the handrail height, initial hand position and fall direction combination that resulted in the highest mean peak forces for the force axis) are indicated in bold. Force axis Fall direction condition Pearson s r (p-value) LOW MED HIGH Linear regression equation for worst-case loading condition Anterior Forward (<0.001) (<0.001) (<0.001) F = 5.43m Posterior Backward (0.001) (<0.001) (<0.001) F = 6.38m Downward Forward (<0.001) (<0.001) (<0.001) F = 7.74m Upward ǂ Forward (0.001) (<0.001) (<0.001) F = 6.06m N.B. The medial and lateral force magnitudes during Reach-to-Grasp and Hand-in-Place trials represent the same axis as the Anterior and Posterior peak force magnitudes during the Facing-Rail trials (i.e. pushing towards or pulling away from the wall). Since the anterior and posterior forces generated during Facing-Rail reactions exceeded those of the medial and lateral forces applied during Reach-to-Grasp and Hand-in-Place reactions, only Facing-Rail reactions were included in this table, representing the worst-case for the loading axes of interest. ǂ Peak upward forces occurred following platform deceleration during the forward fall direction condition, resulting in backward falling motion F corresponds to predicted peak force in N; m corresponds to individual mass in kg

60 Peak downward force (N) Peak anterior force (N) (a) (c) (d) * data1 200 data data2 data Participant 120 mass 140 (kg) 0 data3 data3 Participant mass (kg) data4 data4600 0data1 Peak handrail force - LOW Peak Peak handrail force - HIGH handrail force - MED data5 data data2 Linear regression - LOW Linear regression - MED data6 Linear regression - HIGH data data3 data7 ISO structural 100 strength 120 standard 140 data7 for grab bars (horizontal loading) data4 and removable 40 grab 60 rails (vertical loading) data5 Figure 3.6: The relationship between participant mass and peak handrail forces for the worst-case loading conditions. All worst case data were recorded during Facing-Rail reactions. (a) Peak anterior forces following perturbations in the forward falling direction. (b) Peak posterior forces following perturbations in the backward falling direction. (c) Peak downward forces following perturbations in the forward falling direction. (d) Peak upward forces following perturbations in the forward falling direction. The asterisk (*) for (d) indicates that the peak forces were generally recorded during the second phase of the balance reaction, after the platform had stopped moving and participants pulled upward and backward to maintain stability from platform deceleration. 3.4 Discussion Handrails and other handholds are common assistive tools that an individual can use to recover from balance loss and avoid a fall. We have characterized the peak forces that young adults apply to handrails in response to perturbations of upright stance, and the effect of starting position and handrail height. The relationship between individual weight and peak handrail forces was also evaluated. Finally, we quantified participants MWP and its relationship to starting position and handrail height (b) Peak posterior force (N) Peak upward force (N) 200

61 Peak compensatory handrail forces and influence of starting position Participants in this study applied substantial forces to the handrail during balance recovery. These forces exceeded previously-reported compensatory handhold forces (61% of individual body weight during reach-to-grasp balance reactions on stairs [3], and 150N for hand-in-place reactions following simulated bus accelerations [98]). In general, the Facing-Rail trials in which participants were contacting the rail before perturbation resulted in the highest peak handrail forces in the anteriorposterior, vertical, and resultant directions, with resultant forces reaching an average of 95% BW for LOW rails. That this condition led to the highest forces is unsurprising, as this is the only condition in which participants grasped the rail with both hands. Facing the handrail may also have been advantageous: isokinetic and isometric pull strength advantages have been observed when participant arms are in the sagittal plane (reaching directly forward), compared to cases where the shoulder was abducted or adducted (reaching to the side) [122, 123, 124]. The combination of advantageous arm posture and two-handed grasping thus appear to have enabled higher handrail force generation during the Facing-Rail trials. Correspondingly, the maximum withstood perturbations for the Facing-Rail trials exceeded those of the other conditions; these high perturbation magnitudes would also contribute to high compensatory forces. Although Facing-Rail reactions resulted in the largest applied handrail forces in our study, participants also generated considerable handrail forces during Reach-to-Grasp and Hand-in-Place reactions. When the handrail was low, the peak resultant forces averaged 53% BW and reached a maximum of 77% BW for Reach-to-Grasp responses to backward falling motions. To our knowledge, applied handrail forces following reach-to-grasp reactions have only been reported for forward falling motions during simulated stairway descent [3]. We found comparable force values for this scenario. However, our results for other fall directions and starting positions substantially exceed the previously-reported mean and highest resultant handrail forces, which were 38% BW and 62% BW, respectively [3]. When investigating the individual force components, the posterior and upward vertical forces during reach-to-grasp reactions to backward falling motions were substantially larger than those for forward falling, both in the present dataset and in prior work (38-40% BW and 12-24% BW, compared to <6%BW, or the <10N reported previously [3]). In contrast, simulated stair falls demonstrated higher mean peak anterior and downward forces [3] when compared to our findings. These differences in applied forces can likely be attributed to differences in the direction and nature of falling motion.

62 Influence of handrail height on applied forces When used in practice, handrails are often installed within a prescribed range of heights. The International and Ontario Building Codes stipulate that required handrails on stairs, ramps and (in Ontario) corridors of care facilities be installed at heights between mm [101, 100] (34-38 inches LOW and MED in this study), while handrails built into guards on stairway landings in Ontario must not exceed a height of 1070mm (42 inches HIGH in this study) [100]. The installation height for rails on level ground in private homes, including horizontally-oriented grab bars, is not regulated in these codes [101, 100]. Our findings demonstrate the influence of handrail height on peak applied forces. While anterior and posterior forces rose somewhat with rail height, vertical forces during Facing-Rail reactions and upward forces during Reach-to-Grasp and Hand-in-Place reactions decreased substantially as rail height increased. Peak resultant forces also decreased with higher handrails for both fall directions. Collectively, these data demonstrate decreased physical demands on the rail with higher handrails. Reductions in applied handrail forces with increasing rail height are especially notable for Reach-to- Grasp and Hand-in-Place reactions during backward falling, given that these lower peak forces came from higher maximum withstood perturbation (MWP) magnitudes than those which participants could withstand at using the lower handrail. This seeming contradiction may be explained by considering an inverted pendulum model of balance [91, 17]: higher handrails provide a greater lever arm for generating stabilizing moments compared to lower rails, allowing smaller handrail forces to effectively counteract perturbation-induced translational and rotational forces acting on the COM with respect to the feet and ankles [6]. Similarly, Maki et al. [7, 8, 6] reported larger maximum volitional moments generated using higher handrails, up to 42 inches. During compensatory movements, the moment advantage afforded by a higher rail likely helped participants to sustain higher MWPs before stepping or falling, while applying lower forces to regain stability Implications for handrail design The findings from this study have several implications for the design of handrails and similar products, such as grab bars and removable grab rails that are used to enhance balance control and recovery. In light of the substantial decrease in vertical and resultant peak rail forces with higher rails particularly for Facing-Rail reactions, which resulted in the highest peak forces along the vertical axes a possible strategy to reduce the likelihood of handhold structural failure (i.e. detaching from

63 49 the wall) could involve installing it higher. This may be particularly useful for removable grab rails, such as suction-cup handholds that support transfers into bathtubs, where applied vertical forces would be expected. Higher rail installations may also be beneficial for stability in persons with reduced strength, given that participants withstood greater perturbation magnitudes when the rail was high, usually while applying lower resultant forces to the handrail. Maximum volitional force generation ability on handrails has been observed to be significantly lower for older adults compared to younger adults [7, 9], and age-related declines in grip strength are common [125, 126, 12]. Grip strength may also be reduced in individuals with diabetes [127] or arthritis [128]. Higher rails that reduce grasping force demands are therefore likely to be more useful to many adults in balance control and recovery situations. This research reinforces the importance of handrail designs that are comfortable for users to contact proactively while walking or standing, before losing balance. Participants sustained higher perturbation magnitudes during Hand-in-Place reactions than Reach-to-Grasp reactions, yet applied comparable or lower handrail forces during Hand-in-Place reactions. Further, previous research has established that lightly touching a support-surface reduces postural sway [46] and velocity [47], even following mild perturbations [48]. To promote proactive contact, handrails should be installed within a height range that does not require substantial abduction or flexion in the shoulder or elbow, to avoid causing discomfort in users with limited upper-limb range of motion. The mean preferred stairway handrail height in [7] among both younger and older adults, based on heuristic comfort data, was 36 (0.91m). Further, handrail surface textures should be comfortable when sliding the hand along the surface, while enabling sufficient hand-handrail friction to support balance recovery. Interestingly, participants in this study applied substantially higher posterior and upward compensatory forces to the handrail than reported in the literature (e.g. [3]), which had not previously evaluated the handrail forces exerted while recovering from backward falling. Our findings emphasize the importance of accounting for potentially-high posterior and upward forces in the design of handrails and components for coupling rails with the wall or ground, such as in grab bar mounts. These must be designed with sufficient structural strength to support high posterior and upward forces, particularly in locations where backward falls (e.g. slips) are likely. To our knowledge, this is the first study to demonstrate that the peak forces that participants applied to handrails during balance recovery were often strongly correlated with individual weight, within the range of 45 to 106 kg. While this relationship may be expected in the downward loading

64 50 direction, it is less intuitive in other loading directions, such as upward pulling forces. While standards and guidelines exist to help inform structural strength requirements for handrails, grab bars and similar aids (e.g. CSA B-651 for handrails [129]; ADA Standards for Accessible Design for grab bars [130]), ISO is the only standard that, to our knowledge, includes a provision for estimating user weight limits for handrails, grab bars and removable grab rails based on static strength testing in the vertical and horizontal axes [121]. Our results suggest that the assumption in [119] that individual weight can predict applied handrail forces is reasonable within the range of participant weights considered, although further research is needed to check this assumption among users with greater mass Study limitations While this study provides important data on the forces that younger adults apply to handrails during balance recovery, there are several limitations. Only two perturbation directions were tested, using one general waveform (square wave acceleration profile for the platform). Other waveforms, such as the sudden deceleration in [3], would likely result in different balance recovery responses [131] and concomitant differences in loading profiles. This is especially important for the Facing-Rail trials, in which participants appeared to demonstrate distinct responses to the acceleration and deceleration phases of the perturbation. Further, we focused on the context of balance recovery after perturbations of upright stance, and did not consider the sit-to-stand use case in this study. Applied forces during sit-to-stand must also be considered in product design as a common use of rails, grab bars, and grab poles, as the loading profiles are likely to differ from those of balance recovery. This study considered three fixed handrail heights based on building code benchmarks, rather than scaling to individual participant height. This may limit the extent to which correlations between handrail height and applied forces can be generalized to participants who are much taller or shorter than our sample. However, the alignment of testing conditions with heights at which handrails are likely to be installed in the community (i.e. within Code boundaries) increases the applicability of our findings to existing installations. This study investigated perturbations during quiet standing on level ground and only considered trials in which participants successfully recovered balance without stepping or falling. Real falling scenarios outside the lab often occur in more complex environments where a change of level may be present (e.g. stairs), and where individual responses to balance loss are more varied. For example,

65 51 factors such as reactive stepping and ongoing gait prior to balance loss are likely to affect handrail force profiles. However, by not permitting stepping reactions, we simulated a fall situation where increased handrail reliance, and thus larger physical demand, is necessary. This paper reports only peak forces during maximum withstood perturbations, and thus represents only very high-demand scenarios for handrails. However, it should be noted that all participants reached the highest perturbation levels (5.0m/s 2 ) that the platform could deliver for most Facing-Rail conditions. Despite the limited fall scenarios included in this study, our results provide substantial novel information related to how applied handrail forces are influenced by handrail height, and initial body position and user bodyweight. This information may be useful for guiding product development and installation decisions for handrails and other handholds in the community. Finally, the study population did not fully correspond to the weight distribution of the North American population. Among the participants whose data were used in our correlation and regression estimates, the highest individual weight (106 kg) was between the 90 th -95 th percentile adult female, and the 85 th -90 th percentile male [119]. While this represents a reasonable range of the population, caution should be exercised in extrapolating our handrail force-weight relationships to individuals outside this range. Grip strength and individual weight were moderately-correlated in this study (r=0.542); however, the weight-strength coupling may be lower with a greater range of individual weights, ages, health conditions and lifestyle patterns that could influence strength (e.g. football, weightlifting). A participant sample where weight and strength are less tightly-coupled may demonstrate weaker relationships between weight and handrail forces, particularly along the axes where participants pulled on the rail (i.e. upward, medial) to recover from balance loss. 3.5 Conclusions We have characterized the peak forces that young adults apply to handrails in response to forward and backward perturbations; the effect of handrail height and starting position; and the relationship with body weight. Handrail height and starting position significantly affected peak handrail forces during forward and backward falling motions, with the highest peak forces applied during Facing- Rail reactions. Peak handrail forces generally correlated strongly with individual weight for the testing conditions that resulted in the highest mean vertical and horizontal forces. The reduction in peak vertical and resultant handrail forces with increased rail height, combined with increased maximum withstood perturbation magnitude, implies a possible stability advantage with higher rails, particularly for individuals with reduced hand strength.

66 52 Chapter 4 Study 2: Influence of handrail height and falling direction on center-of-mass control and the physical demands of reach-tograsp balance recovery reactions Preface The content of this chapter has been accepted for publication in the journal Gait and Posture. The coauthors are Konika Nirmalanathan and Dr Alison Novak. It comprises a secondary analysis of the dataset collected in the previous chapter, focusing on reach-to-grasp reactions at a fixed (3.5 m/s 2 ) perturbation magnitude. I conceptualized the study, collaborated to collect data, performed data analysis, and wrote the first draft of the manuscript. Abstract The ability to maintain and recover center of mass (COM) and trunk control after a destabilization is critical for avoiding falls and fall-related injuries. Handrails can significantly enhance a person s ability to recover from large destabilizations, by enabling the person to grasp and apply high forces to the rail to stabilize their COM. However, the influence of handrail height and falling direction on COM control and the demands of grasping are unknown. We investigated the effect of handrail height (34, 38, 42 inches) and fall direction (forward, backward) on COM and trunk control, and the corresponding physical demands of reach-to-grasp balance reactions. Thirteen young adults were destabilized with platform perturbations, and reached to grasp a nearby handrail to recover balance without stepping. COM kinematics and applied handrail forces were collected. COM control was evaluated in terms of: (1) COM range and peak displacement, velocity and momentum in all Cartesian axes; and (2) trunk angular displacement, velocity and momentum in the roll and pitch axes. The physical demands of grasping were estimated via resultant handrail impulse. Compared to forward-directed falling, backward-directed falling was generally associated with greater peak COM and trunk angular displacement, velocity and momentum, along with greater handrail impulse. Higher handrails generally resulted in reduced peak COM and trunk angular displacement, velocity and momentum, as well as reduced handrail impulse. These results suggest that higher handrails (within the range of heights tested) may provide a stability advantage within the range of handrail heights tested, with better COM control achieved with lower physical demands of grasping.

67 Introduction Many falls result from activities that challenge control of a person s center of mass (COM) with respect to their base of support (BOS), such as walking, incorrect weight shifting, tripping, stumbling, or bending [10, 132]. The position and velocity of the COM with respect to the BOS can influence fall risk during slipping [133]. Accordingly, the ability to maintain and recover COM and trunk control from destabilizations is critical for avoiding falls and fall-related injuries. Individuals employ many strategies to control their COM and trunk following balance disturbances. However, the effectiveness of these strategies is heavily context-dependent. For small perturbations, fixed-support strategies, such as quickly contracting muscles in the trunk and lower limbs [5, 134], can provide stabilizing torques to counteract the rotational forces acting on the COM. Conversely, change-in-support strategies (e.g. stepping; reaching to grasp nearby handholds) are often required to recover from large destabilizations [4]. In situations where stepping reactions may not be reliable (e.g. on stairs or icy walkways), grasping reactions are important for balance recovery. It follows that handrails can significantly enhance ability to recover from balance loss [3], provided that their design enables users to quickly and accurately reach to grasp the rail, and then apply sufficient grasping forces to stabilize their COM [4]. While the value of grasping reactions for balance recovery is well-established, our understanding of how handrail height affects both COM control, and the physical demands of achieving this control during reach-to-grasp reactions, is limited. When considering the inverted pendulum model of balance control, individuals may be able to stabilize their COM while applying lower forces to the handrail when the handrail is high, due to the stabilizing moment advantage gained from higher handrails [6]. Unknown to date is how handrail height impacts COM control following forward and backward balance loss when a reach-to-grasp reaction is executed. This study investigates the effect of handrail height and fall direction on COM control and the corresponding physical demands of reach-to-grasp balance reactions, following forward and backward platform perturbations. We hypothesized that handrail height and fall direction would affect key COM control measures, along with the physical demands of reach-to-grasp reactions in terms of the impulse applied to the rail in response to balance loss.

68 Materials and Methods Participants A secondary analysis of a previously-collected dataset was performed [135]. Fifty healthy young adults participated in the larger study, of which 13 participants met our inclusion criteria for analysis in this study (nine males; 18 to 28 years old). To be included in this analysis, participants wore a fullbody motion capture marker set and completed all six testing conditions at perturbation magnitudes of 3.5m/s 2 without falling or stepping (see Sections and for protocol details). For this study, participant heights ranged from 159 to 193cm (mean height: cm); weights ranged from 58.1 to 95.3kg (mean weight: kg). All participants reported being free of neurological, vestibular and musculoskeletal disorders. The institutional and university s Research Ethics Boards approved this work. All participants provided informed consent Experimental setup Data were collected using a 5m x 5m laboratory, secured to a robotic platform that can deliver balance perturbations (Figure 4.1a). An overhead safety harness protected participants from contact with the floor; slack in the harness line allowed participants to move naturally after balance loss. Participants wore knee guards to minimize possible impact with the floor, and a guard on the right elbow to minimize impact from potential contact with the handrail. Participants recovered balance using a height-adjustable, horizontal handrail (outer diameter: 3.8cm) (Figure 4.1b). Fourteen passive motion capture cameras collected kinematic data (Motion Analysis Inc, Santa Clara, CA). Two load cells (one at either end of the handrail) collected handrail loading data (AMTI MC3A-1000; Advanced Medical Technology, Inc, Watertown, MA) Perturbation design Sudden platform translations disrupted balance: backward platform movements simulated forwarddirected falling (Figure 4.1d); forward platform movement simulated backward-directed falling (Figure 4.1e). Perturbations consisted of square-wave acceleration profiles with a 300ms acceleration pulse, followed immediately by an equal and opposite deceleration pulse (acceleration magnitude=3.5m/s 2 ; peak velocity=1.1m/s; displacement=0.32m).

69 Backward falling Forward falling 55 (a) (b) (c) d (d) (e) (f) A-P (+) (BF) A-P (+) (FF) (g) M-L (+) Vertical (+) (h) Roll (+) (i) Pitch (+) Figure 4.1: Testing environment and axis conventions. (a) The Challenging Environments Assessment Laboratory. (b) Inside the lab: a participant stands beside the handrail while wearing a safety harness. Foam blocks were used to discourage participants from stepping. (c) Participants stood beside the rail during testing, with their centerlines a distance of d=58% of their arm length away from the rail. (d) Sample screen captures of forward falling (induced via backward platform translations) with the LOW handrail. (e) Sample screen captures of backward falling (induced via forward platform translations). (f) COM conventions for the antero-posterior (A-P) axis for forward falling (FF) and backward falling (BF): A-P variables were calculated in the falling direction. (g) COM conventions for the medial-lateral (M-L) and vertical axes: M-L variables were calculated based on movement toward the handrail. (h) Trunk roll angular convention. (i) Trunk pitch angular convention.

70 Protocol Individual height, weight and right arm length (acromion to fingertip) were measured. Reflective motion capture markers were used to track whole-body COM (see Section for details on COM estimation calculations). Rigid marker clusters were secured to the pelvis (cluster at the sacrum), upper body (cluster near the thoracic level of T12), and bilaterally mid-thigh and mid-shank. The distal and proximal ends of these segments with respect to tracking markers were identified in a neutral, stationary pose. Upper limb movement was approximated with tracking markers on the hands (base of the second and fifth metacarpals), wrists (ulnar and radial joints), elbows (medial and lateral epicondyles) and shoulders (acromion; front and back of the glenohumeral joint). Participants wore standardized athletic shoes during testing. To begin testing, participants stood erect beside the handrail with their arms relaxed at their sides and feet approximately shoulder-width apart. Their body s center-line was 58% of their right arm length away from the handrail (Figure 4.1c), which approximates the distance between the elbow and middle fingertip (adapted from [118]). Foam blocks in front of and behind participants feet discouraged compensatory stepping. Upon experiencing a perturbation, participants were instructed to reach to grasp the handrail as quickly as possible to re-stabilize, and to avoid stepping or falling into the harness. All participants received at least four lower-magnitude perturbations for each handrail height, which allowed participants to gain familiarity with the protocol for each handrail height. These perturbations were initially delivered at 2.5 m/s 2 in each direction (forward and backward) and increased in 0.5 m/s 2 intervals before the trials analyzed in this study were reached (at 3.5 m/s 2 ). If the participant took a step or fell during a familiarization trial, the trial was repeated. This resulted in a minimum of four lower-magnitude perturbations for each handrail height before data was included for analyses. To minimize pre-planning of movements, perturbation timing and falling direction were randomized. Participants counted backward from a randomly-selected start number by an integer between two and nine to distract attention. For each fall direction, three handrail heights were tested: LOW (86.5cm/34 in); MED (96.5cm/38 in); and HIGH (106.5cm/42 in). 34 inches and 38 inches approximate the lower and upper boundaries of the International Building Code handrail height requirements on stairs and ramps [101], while 42 inches approximates the maximum height of handrails built into stairway landings in the Ontario

71 57 Building Code [100]. This resulted in six testing conditions, with one trial per testing condition for each participant. The testing order of rail heights and fall directions was randomized Data processing Motion capture and handrail force data were sampled at 250Hz and 1000Hz respectively, and synchronized offline [136]. Inertial artifacts in the handrail force signals due to platform motion were removed by subtracting force recordings collected without a participant contacting the handrail. COM and trunk angular kinematics were estimated with a twelve-segment, link-segment model (Visual 3D; C-Motion Inc, Germantown, MD). COM kinematics were calculated from a weighted average of trunk, pelvis, upper- and lower-limb segments, with individual segment COMs approximated with existing anthropometric models [137, 138]. All kinetic and kinematic signals were filtered with zero-lag, low-pass Butterworth filters with the following orders and cut-off frequencies: 1) load cell signals: 2 nd -order/20hz; 2) COM kinematics: 2 nd -order/6hz; 3) trunk angular kinematics: 4 th -order/6hz. Power analyses revealed that 99% of the signal power was under 6Hz for all of our analyzed COM and trunk angular position signals. Visual inspection of all filtered kinematic signals further confirmed that the overall signal shape was preserved, particularly where peak position and velocity values were extracted. Force and COM data filters were applied in MATLAB (The Mathworks, Inc, Natick, MA); trunk angular data filters were applied in Visual 3-D. COM position data were differentiated to calculate velocity, and multiplied by individual mass to calculate momentum Data analysis To evaluate balance control, COM and trunk angular kinematics were analyzed (Figure 4.1d/e). Peak COM displacement, velocity and momentum were calculated 1) along the A-P axis, in the fall direction (Figure 4.1f); 2) along the M-L axis, toward the handrail (Figure 4.1g); and 3) along the vertical axis, downward (Figure 4.1g). COM positional range was calculated along the anteriorposterior (A-P), medial-lateral (M-L) and vertical axes. In contrast with displacement, the COM range describes the difference between the highest and lowest COM position recorded during a trial; the measurement of range accounts for when participants moved in both directions along the same axis. Similar to COM kinematics, the trunk kinematic metrics included determination of both peaks and ranges: 1) Trunk roll angle range, and peak trunk roll angular velocity and momentum (toward

72 58 the rail); and 2) Peak forward trunk pitch angular displacement, velocity and momentum. Trunk roll and pitch angles were defined with respect to the vertical axis (Figure 4.1h,i), in the coronal (roll), and sagittal (pitch) planes. All kinematic metrics were calculated after perturbation onset (platform acceleration > 0.1m/s 2 [136]). The exception was displacement metrics, which were calculated with respect to participants standing position immediately before perturbation onset. Note that the coordinate system translated with the platform. The resultant handrail impulse was calculated as a proxy measure of the physical demands of reactive grasping. Handrail impulse was defined as the time integral of the normalized resultant handrail force curve, calculated over 1) 500ms, 2) 1000ms, 3) 1500ms, and 4) 2000ms after initial handrail contact (where initial contact is defined by handrail force along any axis > 25N). Handrail force data were normalized to a percentage of the participant s body weight (%BW) to facilitate comparisons between participants, resulting in units of %BW*s for impulse. Impulse was calculated up to 2000ms after initial handrail contact, to capture key elements of balance recovery with the handrail, including slowing of COM velocity and restoration of COM position for at least 1s after a participant s highest COM velocity and displacement were observed. Statistical analyses were performed using 2x3 repeated measures ANOVAs (SAS Enterprise Guide version 9.1, Cary, NC), with fall direction (forward, backward) and handrail height (LOW, MED, HIGH) comprising the within-subject factors. Data were rank-transformed to meet ANOVA normality assumptions [139]. Post-hoc pairwise comparisons with Tukey adjustments to account for multiple comparisons were performed following identification of significant main effects. Where interaction effects were identified, only pairwise comparisons of handrail height within each fall direction were considered (i.e., forward-low was not compared to backward-high). Sphericity was confirmed with Mauchly s test in SPSS (IBM, Armonk, NY). Significance levels were p<0.05 for all analyses. 4.3 Results Balance recovery strategy forward-directed versus backwarddirected falling Balance recovery strategies from one participant, during forward-directed and backward-directed falling, are shown in Figure 4.1d and Figure 4.1e. Characteristic COM kinematic, trunk angular kinematic and resultant handrail force profiles are shown below in Figure 4.2.

73 Resultant handrail force (%BW) Trunk angular velocity ( o /s) Trunk angle ( o ) COM velocity (m/s) COM position (m) 59 (a) Forward falling 0.3 A-P M-L Vertical* Backward falling 0.3 A-P* M-L Vertical* 0 0 (b) A-P M-L Vertical* A-P* M-L Vertical* (c) Pitch Roll Pitch Roll (d) 150 Pitch Roll 150 Pitch Roll (e) Force Force Time relative Time after to perturbation onset (s) onset (s) Time relative Time after to perturbation onset (s) onset (s) Figure 4.2: Characteristic COM kinematics, trunk angular kinematics and resultant handrail force profiles during forward and backward falling. All traces are from one participant, when the handrail was LOW. (a) COM position in the anterior-posterior (A-P), medial-lateral (M-L) and vertical axes. Note that the sign convention for the A-P axis during backward falling was not flipped, to more clearly indicate backward COM movement. (b) COM velocity in all three Cartesian axes. (c) Trunk pitch and roll angles. (d) Trunk pitch and roll angular velocity. (e) Resultant handrail force. Momentum traces were excluded as the shape of these traces was the same as that of velocity. The asterisk (*) for A-P COM traces during backward falling, and all vertical COM traces, indicates that the axis conventions are opposite to those shown in Figures 1f/g. The sign convention for the A-P axis during backward falling was not flipped to more clearly indicate backward COM movement, while the vertical axes were not flipped to highlight initial upward COM movement during forward falling, versus downward movement during backward falling.

74 Resultant handrail impulse (%BW*s) 60 Forward-directed falling was generally characterized by the trunk pitching forward, with little downward displacement of the hips (Figure 4.1d; Figure 4.2c/d). Conversely, backward-directed falling often involved dropping the hips, while the trunk remained relatively upright (Figure 4.1e). Backward-directed falling generally demonstrated greater COM displacement and velocity magnitudes (Figure 4.2a/b) and handrail forces (Figure 4.2e) compared to forward-directed falling. On average, participants contacted the handrail more quickly during forward falling (338+41ms) than backward falling (351+32ms) (F(1,12)=8.17;p=0.014). Handrail contact time did not vary significantly with rail height (F(2,24)=1.24; p=0.306) Effect of fall direction and handrail height on the physical demands of reactive grasping handrail impulse Handrail impulse (Figure 4.3) during backward-directed falling significantly exceeded that of forward-directed falling (all F(1,12) s>16.51; all p s<0.002), with the greatest increase to impulse occurring in the first 500ms after handrail contact. Further, impulse decreased significantly as handrail height increased (all F(2,24) s>3.49; all p s<0.047). However, only the increase from LOW to HIGH was significant (500ms: p L-H <0.001; 1000ms: p L-H =0.014; 1500ms: p L-H =0.039; all other pairwise p s (LOW-MED, and MED-HIGH) > 0.083). The increase from LOW to HIGH was not significant at 2000ms, following Tukey corrections (p L-H =0.072). Significant interaction effects between fall direction and handrail height were not observed for impulse (all interaction F(2,24) s<0.98; all interaction p s >0.115) ms 1500ms ms 2000ms 30 α 20 * * * LOW MED 2.5 HIGH LOW 4.5 MED HIGH 6.5 Forward falling Backward falling Rail height and fall direction - significant main effect of fall direction (p<0.002 at all time points) * - pairwise comparison p<0.039 between LOW and HIGH α - pairwise comparison p=0.072 between LOW and HIGH after Tukey corrections (though significant main effect of rail height present) Figure 4.3: Resultant handrail impulse for each condition, measured 500ms, 1000ms, 1500ms and 2000ms after initial handrail contact. Mean values are plotted; error bars represent standard deviation.

75 COM Range Vertical (m) Downward COM displacement (m) Downward COM velocity (m/s) Downward COM momentum (kg*m/s) COM Range ML (m) ML COM displacement toward rail (m) ML COM velocity toward rail (m/s) ML COM momentum toward rail (kg*m/s) COM Range AP (m) AP COM displacement in fall direction (m) AP COM velocity in fall direction (m/s) AP COM momentum in fall direction (kg*m/s) Effect of fall direction and handrail height on COM and trunk control COM kinematics Compared to forward-directed falling, backward-directed falling resulted in significantly worse COM control (i.e., higher COM range, displacement, peak velocity and peak momentum) in all axes, with the exception of COM range in the A-P axis (A-P COM range: fall direction F(1,12)=1.23; p=0.082; all other COM metrics: fall direction F(1,12) s>15.38; p s<0.002) (Figure 4.4). (a) LOW MED HIGH LOW MED HIGH Interaction p=0.029 LOW MED HIGH LOW MED HIGH (b) * * * * * * LOW MED HIGH 0 LOW MED HIGH 0 LOW MED HIGH 0 LOW MED HIGH (c) 0.15 * 0.15 * 0.6 * * * LOW MED HIGH 0 LOW MED HIGH 0 LOW MED HIGH 0 LOW MED HIGH * - significant pairwise comparison between handrail heights (p<0.05) N.B. All falling direction p s < 0.002, except for COM Range AP (p=0.29) Falling direction legend: Forward Falling Backward Falling Figure 4.4: COM range, and peak displacement, velocity and momentum magnitudes in the (a) antero-posterior (A-P), (b) medial-lateral (M-L), and (c) vertical axes. Bars represent mean values; error bars represent standard deviation. Significant main effects of handrail height on COM control were not found in the A-P axis (F(2,24) s<0.77; p s >0.475). In the M-L and vertical axes, all COM control variables decreased significantly (indicating better COM control) as handrail height increased (handrail height main effect F(2,24) s >5.77; p s<0.009). Pairwise comparisons revealed that the decrease in COM variables

76 62 as rail height increased from LOW to HIGH was significant in both M-L and vertical axes (all LOW- HIGH p s < 0.007). The decrease from MED to HIGH was also significant for COM range and displacement in the ML axis, and for downward COM displacement (p s<0.029). The only observed significant interaction effect was for peak COM velocity in the A-P axis (F(2,24)=4.13; p=0.029), which rose as handrail height increased during forward-directed falling, but decreased as handrail height increased for backward-directed falling. Significant interaction effects were not observed for other COM control variables (F(2,24) s<2.69; p s>0.089) Trunk angular kinematics Trunk roll range was significantly higher for backward falling compared to forward falling (F(1,12)=10.56; p=0.007), while peak trunk roll velocity and momentum did not differ significantly between falling directions (F(1,12) s<3.28; p s>0.095) (Figure 4.5a,b,c). Conversely, peak forward trunk pitch displacement, velocity and momentum were significantly higher for forward falling than for backward falling (F(1,12) s>137.37; p s<0.001) (Figure 4.5 d,e,f). All trunk angular metrics decreased significantly as handrail height increased (F(2,24) s>10.88; p s<0.001). Pairwise comparisons revealed that all trunk angular metrics decreased significantly as handrail height increased from LOW to HIGH (p s<0.001). Trunk pitch metrics further decreased from LOW to MED (p s<0.008). For roll, all decreases from LOW to MED, and MED to HIGH, were significant, with the exception of the roll angle range LOW-to-MED decrease (roll angle range p L- M=0.124; all other roll p L-M s and p M-H s<0.039). Significant interaction effects between fall direction and handrail height were not found (all interaction F(2,24) s<1.51; all interaction p s>0.346).

77 Peak roll angular momentum toward rail (kg o /s) Peak forward pitch angular momentum (kg o /s) Peak roll angular velocity toward rail ( o /s) Peak forward pitch angular velocity ( o /s) Roll angle range ( o ) Peak forward pitch angle ( o ) 63 Trunk roll angular kinematics Trunk pitch angular kinematics (a) Fall direction p=0.002 * * (d) * Fall direction p<0.001 * LOW MED HIGH 0 LOW MED HIGH (b) Fall direction p=0.095 * * * (e) * Fall direction p<0.001 * LOW MED HIGH 0 LOW MED HIGH (c) 1 * Fall direction p=0.188 * (f) * Fall direction p<0.001 * 0.8 * LOW MED HIGH 0 LOW MED HIGH * - significant pairwise comparison between handrail heights (p<0.039) Fall direction legend: Forward Backward Figure 4.5: Trunk angular kinematics for each fall direction and handrail height. Bars represent mean values; error bars represent standard deviations. (a) Trunk roll angle range. (b) Peak roll angular velocity (toward the handrail). (c) Peak roll angular momentum (toward the handrail). (d) Peak forward pitch angle. (e) Peak forward pitch angular velocity. (f) Peak forward pitch angular momentum.

78 Discussion To avert a fall after balance loss, one must regain control of the position and velocity of their COM and trunk. The results of this study indicate that both falling direction and handrail height significantly affected trunk and COM control following platform perturbations, along with the impulse that participants applied to the handrail to regain stability Backward-directed falling resulted in poorer COM control and greater physical demands of grasping, compared to forwarddirected falling Young adults have been previously observed to fall more frequently from backward destabilizations than from forward destabilizations [64], with higher wrist impact velocities [140]. In this study, backward-directed falling resulted in less-controlled COM kinematics than forward-directed falling, even though the stabilizing impulse that participants applied to the handrail during backwarddirected falling consistently exceeded that of forward-directed falling. The heightened handrail impulse during backward-directed falling may be explained in part by how both the real and perceived instability of backward-directed falling may exceed that of forward-directed falling. Reduced stability during backward falling may have stemmed from the center of pressure being closer to the posterior edge of the base of support while standing, demanding a larger response during a backward-directed fall particularly in the absence of stepping. Poorer visual perception of the body relative to the floor in the falling direction may have further increased the perception of instability, and contributed to larger applied handrail forces in response to backward-directed falling [141, 142], over the course of the balance recovery response. Backward-directed destabilizations have been shown to elicit stronger responses compared to forward-directed destabilizations, including greater dorsiflexor and plantarflexor co-contraction [19] and increased likelihood of reaching for handrails [90] even with reduced perturbation magnitudes for backward-directed falling [90]. Taken together, these factors may have led to more aggressive handrail use during backward-directed falling, compared to forward-directed falling As handrail height increased, COM and trunk control improved and the physical demands of grasping decreased As handrail height increased, participants demonstrated consistent or better COM and trunk control in response to perturbations, even though handrail impulse decreased significantly as increases in height. An inverted pendulum model of balance [91] may help to explain the better COM and trunk

79 65 control with increased rail height, without a concomitant increase to handrail forces. In this context, the handrail enables users to generate high stabilizing forces and moments to counteract the translational and rotational forces acting on the COM with respect to the ankles [6]. Accordingly, the higher rails evaluated in this study afforded greater stabilizing moments than did the lower handrails for a given applied handrail force, due to the increased moment arm between the user s ankles and the rail. These findings are consistent with past research, where higher maximum voluntary moment generation ability with increased handrail height has been observed in both younger and older adults [7, 8]. Building on these past studies, our findings suggest that the stabilizing moment advantage may have enabled participants to achieve better COM and trunk control with the higher rails evaluated in this study, while applying lower impulse to the handrail during reactive grasping. While our results suggest a stability advantage with higher handrails tested in this study, our findings would not necessarily apply outside of the tested range. The mechanical advantage of a larger moment arm with increasing handrail height would be offset at some point by other factors, such as substantial reductions to volitional strength with handholds surpassing shoulder height [97], or the user eventually not being able to reach the handrail altogether. Further research is required to determine this optimal height for balance recovery across various populations. We note that while statistically-significant effects of handrail height and falling direction on impulse and COM/trunk control were observed, the functional importance of these differences is unknown because participants did not step or fall in the included trials. The stability advantages with the higher handrails evaluated in this study may be more important to individuals with reduced trunk and upper-limb strength due to conditions such as in persons with stroke [143], who may both (a) demonstrate worsened trunk and COM control, and (b) be unable to apply the greater handrail forces needed for balance recovery with lower rails. For example, reduced upper-limb strength and trunk flexion-extension isometric strength (while standing) have been observed in post-stroke patients [143, 144], who may experience greater challenges in restoring their trunk to upright stance with the greater pitch angular displacements observed with lower handrails in this study. Conversely, higher handrails may be problematic for users with upper-limb arthritis or other conditions that constrain range of motion. Further evaluations of handrail height on balance recovery should consider individuals with reduced strength and range of motion, including older adults with demonstrated age-related declines in handrail force generation ability [7], and speed and accuracy of reactive grasping [74].

80 Limitations and future work This study enhances our understanding of how handrail height and falling direction influence balance recovery. However, several limitations should be acknowledged. First, we focused on reachto-grasp reactions following perturbations of upright stance; ongoing gait was not studied. While both contexts are important, ongoing gait testing may reveal different results due to delays in arm movement onset and handrail contact with leg movement [14]. Reduced speed of reach-to-grasp reactions may result in increases to peak COM and trunk kinematic variables and thus worsened control due to the increased time after balance loss before the handrail can be used to generate stabilizing forces. However, the potentially-negative COM control implications of delayed reactive grasping during ongoing gait may be countered by being able to step in response to perturbations. Second, the high number of outcome measures in this study increases the risk of a Type I error within our main effects, although applying Tukey corrections reduces the likelihood of false positives within our post hoc comparisons. Third, this study evaluated three handrail heights only. While the tested handrail heights were selected to coincide with existing building standards (which increases the applicability of our findings to handrails in the community), performance in metrics may differ for handrails outside of our tested range. Finally, our sample was limited to healthy young adults. Further research should include other populations, including individuals with reduced trunk and upper-limb strength and range of motion, whose balance recovery reactions may be affected differently by varying handrail height. 4.5 Conclusions We have characterized the influence of falling direction and handrail height on COM and trunk control and the corresponding physical demands of reactive grasping in younger adults. Backwarddirected falling resulted in poorer COM control than did forward-directed falling, despite higher impulse applied to the handrail during backward-directed falling. Trunk control was generally worse during forward-directed falling. As handrail height increased, COM and trunk control improved, and impulse applied to the handrail decreased. Our findings suggest a possible stability advantage with increased handrail height in both falling directions within the range of tested handrail heights, demonstrated by participants having achieved greater COM and trunk control while applying lower impulse to the handrail during balance recovery.

81 67 Chapter 5 Study 3: Effect of handrail height and age on the speed and accuracy of reach-to-grasp balance recovery reactions during slope descent Preface This chapter includes a manuscript that has been drafted for submission to the journal Experimental Brain Research, with co-authors Dr Brian Maki and Dr Alison Novak. I conceptualized the study; collaborated to develop the protocol and collect data; performed all statistical analyses (with guidance from statistical consultant, Dr Marguerite Ennis), and wrote the first draft of the paper. Abstract Rapid reach-to-grasp reactions are common responses to balance loss, and can help individuals to regain stability and avoid a fall. Given the short time scales over which falls occur (with hip impact often within 1500ms of balance loss), the speed and accuracy with which handrails can be grasped after a destabilization may be important for a person s subsequent ability to recover. However, the influence of handrail height on the speed and accuracy of reach-to-grasp reactions is unknown. We investigated the effect of handrail height (41% to 72% of participant height, h) on the speed and accuracy of reach-to-grasp reactions during slope descent, in younger and older adults. Thirteen younger and fourteen older adults walked up and down a 4.8m, 8 o slope mounted to a robotic platform. During randomly-selected walks, sudden platform translations destabilized participants and evoked reach-to-grasp reactions. Middle deltoid onset, hand trajectory kinematics, and handrail contact time (measured with an optical system mounted to the handrail) were collected. Regression models quantified the effect of age and handrail height on metrics of interest. Aging was associated with slower deltoid onset, slower peak lateral and downward hand speed, reduced time to peak upward and lateral hand speed after deltoid onset, and increased time to handrail contact from peak hand speeds. Handrail contact and movement time varied with handrail height, with regressionestimated movement time minimized in older adults when the rail was at 64%h (close to elbow height). As handrail height increased, peak upward and lateral hand speeds generally increased, while vertical handrail overshoot decreased. Our findings suggest a possible speed and accuracy advantage with increased rail height (up to elbow height) in older adults, as demonstrated by reduced movement time and vertical handrail clearance in these conditions.

82 Introduction Reach-to-grasp reactions are common responses to balance loss when handrails are present [15, 101], and can help individuals to regain stability and avoid falls [3]. However, the brief time scales over which falls occur often with hip or hand impact taking place within 1500ms of balance loss [64, 63] may limit the time available for reach-to-grasp reactions to be executed successfully. Consequently, the speed and accuracy with which a person reaches to contact a nearby handrail after balance loss is likely important for their subsequent ability to grasp the handrail and restabilize. While handrails are helpful for balance recovery, our understanding of how installation height affects the speed and accuracy of reactive grasping is limited. Previous work reported that individuals tend to quickly raise their arms after balance loss (akin to a generic, startle -like response) [64], which may increase the length of the reaching trajectory and delay the time to handrail contact. In contrast, during volitional reaching, individuals adopt near-straight trajectories [145, 85], which could reduce the time required to contact the handrail. While young adults have been observed to alter their reach trajectory direction based on their falling direction relative to a handhold [82], the extent to which handrail height affects trajectory path in balance recovery contexts is unknown. The effects of aging on the speed and accuracy of reactive grasping are of interest, because of the increased fall-related injury risk in older adults [2, 1]. Slower upper-arm muscle onset latencies and handrail contact time in older adults have been observed following perturbations of upright stance [74, 75], though age-related changes in handrail contact time during ongoing gait perturbations have not yet been reported. Since many falls occur while walking [10], grasping reactions during gait should be considered. Further to this, leg movement has been observed to result in slower muscle onset latencies and reaching movement time following balance perturbations, compared to no leg movement [14, 15, 16]. Given the reduced reach-to-grasp reaction speed that occurs during ongoing gait and in older adults, the consequences to balance recovery may be amplified with handrails that cannot be reached quickly or accurately. Insights into more detailed kinematics of compensatory reaching strategies with varied age and handrail height are also needed. One study observed that the time to grasp completion in older adults did not differ significantly from younger adults, despite slower arm muscle onset latencies in older adults [15]. This suggests that older adults may compensate for delayed muscle onset through some combination of reaching at a higher velocity, or reducing the length of their reach trajectory.

83 69 Indeed, increased peak vertical wrist velocity magnitudes in older adults have been observed during grasps evoked by forward and backward platform perturbations, though the increase was only statistically-significant for forward platform movements, and limited to upright stance rather than gait [75]. Alternatively, trajectory overshoot reduction could occur by increasing the time to handrail contact after a rapid initial phase of reaching, once the peak hand speed has been achieved. In volitional and compensatory contexts with young adults, this has been proposed as a strategy to enable more precise modulation of both the reach trajectory and hand aperture, to increase the likelihood of successful grasping once the rail is contacted [66, 80, 77]. However, we have not yet established if these reaching strategies are present following ongoing gait perturbations, or if they depend on age or handrail height. This paper investigates the influence of handrail height on the speed and accuracy of reach-to-grasp reactions in younger and older adults during gait. Specifically, this study aimed to understand differences in reach timing and velocity between younger and older adults during simulated slippery slope descent and, if present, the potential impact (or dependence) of these differences on reach trajectory accuracy. We further aimed to understand if these measures of speed and accuracy vary with handrail height for both age groups. Slippery slope descent is of interest, as the sloped surface may increase the challenge of executing effective compensatory stepping following balance loss, due to the reduced friction between the foot and the sloped walking surface [146]. This increases the possible importance of compensatory grasping for balance recovery when walking down slopes. We expected that handrail height would affect both the speed and the accuracy of reactive grasping, and that older adults would execute slower and less accurate reaching than younger adults. 5.2 Methods Participants Eighteen healthy younger and eighteen healthy older adults participated. Of these, 14 younger adults (7 males) and 13 older adults (8 males) were analyzed, due to technical issues with the motion capture system for four young and five older adults. All participants analyzed in this study were right-handed and had normal or corrected-to-normal vision (self-reported). Participant demographics are presented in Table 5.1. Mean participant height and weight did not differ significantly between groups (p s > 0.45), based on two-tailed t-tests. The study was approved by institutional research ethics boards.

84 70 Table 5.1: Participant demographics Younger adults Older adults Mean + SD Range Mean + SD Range Age Height Weight years cm kg 19 to 33 years to cm 50.8 to kg years 60 to 77 years cm 155 to cm kg 46.3 to kg Experimental setup A computer-driven platform was used to change the walking surface incline and perturb balance (Figure 5.1a). The laboratory (mounted to the platform) was inclined at 8 o with respect to level ground. For safety, an overhead, robotic safety harness followed participants during testing. A twelve camera, three-dimensional passive motion capture system (Motion Analysis Inc, Santa Clara, CA) recorded kinematic data. (a) (b) (c) (d) wood tiles trigger force plates non-trigger force plates handrail (44mm outer diameter) laser (beam mounted 1 cm away from surface of rail) 1.2m 1.2m (e) 0.6m z x 8 o y Figure 5.1: Experimental setup. (a) Outside the laboratory, mounted to a robotic platform. (b) Inside the lab. The optical system mounted to the rail to estimate contact time is outlined in the yellow rectangle. (c) The walking surface layout (top: side view; bottom: overhead view). Perturbations occurred during randomlyselected walks down-slope, when either trigger force plate was loaded with 100N (downward force). (d) A schematic of the optical system on the handrail. Three lasers were mounted around the handrail at one end; a signal receiver was mounted at the other end. This allowed the system to detect when a hand was within 1cm of the rail, to estimate handrail contact time. (e) The coordinate system in the laboratory, which translated with the lab during perturbations. The walkway consisted of eight 1.2m x 1.2m fiberboard panels (Flakeboard, USA), in a 4x2 arrangement (4.8m x 2.4m) (Figure 5.1 b, c). The middle four fiberboard panels were mounted on top of force plates of the same dimensions (AMTI, Natick, MA). The outer panels were mounted on wooden boxes, to unify walkway height. All participants wore standardized running shoes with a layer of felt and cotton fabric under the sole (the cotton fabric was on the outside, and interacted with the walking surface). This increased the challenge of executing rapid reactive steps after balance

85 71 perturbations, due to reduced shoe-floor friction [147]. A height-adjustable handrail (steel tube; 44mm outer diameter) was positioned beside participants on the side of their dominant hand, at a distance of 60cm from the center of the walkway. Participants were instructed to walk such that their mid-coronal plane was approximately aligned with the middle of the walkway. Handrail contact time was estimated with an optical system mounted to the rail (Figure 5.1d). The system consisted of three lasers mounted 1cm away from the rail, which detected when the hand interfered with the beams. Hand trajectory kinematics were estimated from a reflective marker secured to the head of the second metacarpal ( hand marker ). Middle deltoid onset latencies were measured with surface electromyographic sensors (Telemyo DTS; Noraxon Inc., Scottsdale, AZ). Video cameras mounted to the laboratory walls recorded participant behavior during each trial (sample rate=30hz) Experimental protocol After providing informed consent, participants height and weight were measured. Participants walked up and down the slope continuously beside the handrail (self-selected pace), with their arms at their sides. They completed a concurrent secondary task (counting backward from a researcherselected number between 100 and 1000 by a number between two and nine; listing animals that start with every letter of the alphabet; or other conversation with the researcher inside the lab). As the purpose of the secondary task was to distract participants, performance in the secondary task was not evaluated. We thus offered a variety of tasks to accommodate differences in skill. One older adult was unable to complete the secondary task in tandem with the protocol, and thus walked without talking throughout the study. However, visual inspection of the data did not reveal this participant to be an outlier. Balance was disrupted with backward platform translations, which consisted of a 300ms squarewave acceleration pulse, followed by an equal and opposite deceleration pulse (acceleration=3.75 m/s 2 ; peak velocity=1.13 m/s; displacement=0.34m). Perturbations were triggered during randomlyselected walks, when either trigger force plate was loaded with 100N (see Figure 5.1c). Participants were instructed to reach to grasp the handrail as quickly as possible when perturbed. The following handrail heights were tested: 30, 32, 34, 36, 38, 40, 42 and 44 inches (76, 81, 86, 91, 97, 102, 107 and 112 cm, respectively). Rail heights were selected to exceed the prescribed range for handrails on ramps by various building codes and accessibility standards, including the 86cm to

86 72 97cm permitted by the Ontario Building Code in Canada [100] and the ADA Standard for Accessible Design [130], and the 90cm to 100cm permitted in the New Zealand Building Code [148]. Handrail height was defined as the vertical distance between the top of the handrail and the top of the walking surface. The order of testing handrail heights was randomized for each participant. All participants completed a practice trial at the start of testing, to minimize first-trial effects and allow participants to gain familiarity with the protocol. This trial was not analyzed. After the practice trial, participants completed four trials per handrail height, of which the first three per height are analyzed in this paper. Trials were repeated if participants did not reach to grasp the rail, or if their hands were raised leading into the perturbation beyond normal arm swing (e.g. talking with hands ). Participants reported perceived exertion after each block of trials (Borg ratings [149]). No participants perceived their exertion to exceed moderate intensity (rating of 14). Participants were allowed to rest at any point, with a mandatory 10 minute rest period halfway through testing Data processing Data sources (motion capture, EMG, optical system on rail, accelerometer that measured platform movement) were synchronized offline [136]. All kinematic signals were oriented such that the vertical axis was aligned with gravity (Figure 5.1e). Kinematic data were sampled at 250Hz; all other sources were sampled at 1000Hz. Hand velocity estimates were calculated by filtering the hand marker position signal with a zero-lag, second-order, low-pass Butterworth filter (30Hz), and then differentiating the filtered signal with respect to time. EMG data were filtered with zero-lag, secondorder Butterworth filters, using the following procedure: 1) band-pass filtered (10 to 499Hz) to address signal drift and reduce the potential for aliasing; 2) full-wave rectified to establish linear envelope wave forms; and 3) low-pass filtered (30Hz) to attenuate high-frequency noise Metrics Metrics are broadly grouped into three categories, related to the timing aspects of the movement, and the speed and accuracy of the hand trajectory.

87 Timing metrics The key global timing metrics were: 1) middle deltoid onset latency (t_onset); 2) handrail contact time (t_contact); and 3) movement time (t_movement). t_onset was defined as the time after perturbation onset (platform acceleration>0.1 m/s 2 [90]) that the EMG signal magnitude surpassed the onset threshold, and was sustained above that level for 20ms or more [66]. The onset threshold was defined as: where baseline is the middle deltoid EMG signal recorded over the 200ms leading into perturbation onset, for each trial. All deltoid onsets were confirmed via visual inspection. t_contact was defined as the time after perturbation onset at which one of the sensors on the optical system detected hand contact. t_movement was defined as the difference between t_contact and t_onset [16]. We also determined the timing of peak hand speeds and position, extracted: (a) relative to middle deltoid onset (t_onset-to-peak), and (b) relative to handrail contact time (t_peak-to-contact) Speed variables The hand trajectory speed metrics included peak hand speed in the upward, lateral and downward directions (v_up, v_lateral and v_down, respectively). These metrics were measured between perturbation onset and 20ms after t_contact Accuracy variables The accuracy metrics included: 1) vertical handrail overshoot (Overshoot); 2) peak vertical hand position (Peak_Position). Overshoot was defined as the peak vertical distance between the participants hand marker and the top of the rail, and served as a measure of trajectory deviation from a more direct path. The top of the rail was landmarked with two motion capture markers (one at each end of the rail). Peak_Positon represented the vertical position of the hand marker relative to the walking surface. All metrics are summarized in Table 5.2

88 74 Table 5.2: Summary of metrics Type of metric Metric Abbreviation Timing metrics Middle deltoid onset t_onset Handrail contact time t_contact Movement time t_movement Time from deltoid onset to peak hand position or speed t_onset-to-peak Time from peak hand position or speed to handrail contact t_peak-to-contact Speed metrics Peak upward hand speed v_up Peak lateral hand speed (toward the rail) v_lateral Peak downward hand speed v_down Accuracy metrics Vertical handrail overshoot Overshoot Peak hand position Peak_Position Statistical analysis Effects of handrail height and age on metrics of interest were estimated with a mixed, linear, repeated measures regression model (SAS 9.4, Cary, NC). We used this regression approach to minimize the influence of individual anthropometry on our findings, as it allowed us to normalize handrail height to a percentage of individual height (% individual height, h). The model was fitted with the maximum likelihood estimation method. Data from participants three trials for each handrail height were averaged. Handrail height and age were modeled as a within-and betweensubject factors, respectively. The following general model was used: Where: Height is a continuous variable that represents normalized, zero-mean handrail height; and Age is a categorical variable Model coefficients (β s) were evaluated for statistical significance (p<0.05). In cases where higherorder terms (Height 2, Age*Height, Age*Height 2 ) were not significant, the model was re-calculated

89 75 with the term eliminated. The physical interpretation of the significance levels of model coefficients is described in Table 5.3. Table 5.3: Overview of the physical interpretation of the statistical model. Coefficient Physical interpretation if the coefficient is significant (p<0.05) β 0 The model intercept (i.e. mean value of the metric) is significantly different from 0 β 1 β 2 β 3 β 4 β 5 Participant age significantly affected performance in the metric Handrail height affected performance in the metric, with the slope of the model being significant Handrail height affected performance in the metric, with the curvature of the model being significant Handrail height and age differentially affected participant performance, with the difference in slopes of participant performance with respect to handrail height being significant Handrail height and age differentially affected participant performance, with the difference in curvature of participant performance with respect to handrail height being significant 5.3 Results All analyzed participants completed the data collection protocol. However, for two participants, EMG data from four handrail heights were not collected successfully, due to issues with collection apparatus. For two handrail heights with one participant, t_onset was >1s exceeding reported middle deltoid onsets following perturbation-evoked reach-to-grasp reactions during gait [16, 15]. Confirmation with videos suggested that the arm movement was not compensatory for these trials. We thus removed those data points from analysis, resulting in incomplete datasets for these participants Time course of reach-to-grasp movement The time course of the reach-to-grasp movement for a representative young adult, including key muscle and kinematic signals, is shown in Figure 5.2. Detailed statistics for timing variables are summarized in Table 5.4.

90 down 76 t_onset-to- Peak 0.08 t_contact t_peak-to-contact t_movement t_onset 0.02 EMG vertical hand position top of handrail toward rail up vertical hand velocity lateral hand velocity Time relative to perturbation onset (ms) Figure 5.2: Time course of the reach-to-grasp reaction for a characteristic young participant. The handrail was 36 inches (96.5 cm) high. Data were collected from a single trial. Perturbation onset (solid black line) occurs at 0ms. Middle deltoid onset (first dotted line) and handrail contact time (third dotted line) occur 183ms and 488ms after perturbation onset respectively, resulting in a movement time of 305ms. Time to handrail contact approximately aligns with peak downward hand speed and peak aperture width. After handrail contact, the participant s hand rotated around the rail, leading to hand velocity continuing in the lateral direction. The marker on the participant s hand settled approximately 2.5 cm above the top of the rail, indicated by the dashed line showing handrail position. The middle dotted line represents the time of peak upward hand velocity (v_up). Important variables are also defined, with t_onset-to-peak and t_peak-to-contact shown for v_up.

91 Table 5.4: Summary of timing variable performance with respect to handrail height for both younger and older adults. Note that significant interactions between Age and Height 2 were not found, so they are excluded from the table. Significant effects and interactions are bolded. Variable Younger adult performance (mean (SD)) Older adult performance (mean (SD)) Age p Height p Height 2 p Age*Height p Global timing variables Middle deltoid onset (t_onset) (ms) 174 (38) 200 (40) Handrail contact time (t_contact) (ms) 531 (54) 552 (72) Movement time (t_movement) (ms) 358 (52) 355 (62) (YA) (OA) Time to peaks after deltoid onset Time to peak upward hand speed from deltoid onset (t_onset-to-peak for v_up) (ms) Time to peak lateral hand speed from deltoid onset (t_onset-to-peak for v_lateral) (ms) Time to peak downward hand speed from deltoid onset (t_onset-to-peak for v_down) (ms) Time to peak vertical hand position from deltoid onset (t_onset-to-peak for Peak_Position) (ms) 219 (41) 189 (43) < (52) 265 (53) (59) 350 (57) (52) 272 (50) < Time from peaks to rail contact Time from peak upward hand speed to rail contact (t_peak-to-contact for v_up) (ms) Time from peak lateral hand speed to rail contact (t_peak-to-contact for v_lateral) (ms) Time from peak vertical hand position to rail contact (t_peak-to-contact for Peak_Position) (ms) 132 (37) 161 (49) < (38) 88 (55) <0.001 < (37) 78 (41) <0.001 <

92 ct (ms) o d contact vel to contact (ms) (ms) vel ontact to contact (ms) (ms) t_onset (ms) t_contact (ms) t_movement (ms) Global timing metrics: deltoid onset, handrail contact time, movement time Mean t_onsets were significantly faster in young adults than in older adults (174ms versus 200ms; p=0.003; Figure 5.3a). Significant main effects of age were not found for t_contact or t_movement (p s>0.217; Figure 5.3b and c). Handrail height significantly affected both t_contact and t_movement (Height 2 p s<0.037). However, rail height affected t_movement in younger versus older adults differently (Age*Height p=0.032; Height and Height 2 p s<0.037 for both populations). Based on the regression models, t_movement was minimized for younger and older adults with the rail at 56% individual height (h) (~mid-forearm height [116]) and 64%h (~elbow height [116]) respectively. Note that the mean initial resultant distance between the hand and the handrail at t_onset was comparable between age groups (34cm) (a) YA: t = h N (b) YA: t = h N h 2 N (c) YA: t = h N h 2 N OA: t = h N 600 OA: t = h N h 2 N OA: t = h N h 2 N Peak upward vel 100 to contact (ms) Peak lat vel to contact (ms) YA 40 data Peak upward vel to contact (ms) Peak lat vel to contact (ms) YA data Rail height 400 (% individual height) OA Rail data height (% individual 400 height) Rail height OA data (% individual height) peak opening speed (ms) Time from peak opening speed 300 to contact (ms) 300 YA model data YA model 300 data 600 Younger adults - data YA data YA data OA Younger model data adults regression line OA model data OA data Older OA 300 data adults - data YA Older model adults regression 200 line YA model YA model YA model OA model OA model OA model OA model Figure 5.3: Global timing variables with respect to handrail height. (a) Middle deltoid onset latency (t_onset). (b) Time to handrail contact relative to perturbation onset (t_contact). (c) Movement time (handrail contact time relative to deltoid onset, t_movement). All points represent the mean data from three trials at the rail height of interest for a given participant. Equations to generate the regression lines are above the respective figures. h N represents handrail height normalized to a percentage of individual height, and zeromean. 78

93 ms) g speed (ms) t (ms) ntact eed to contact (ms) Time to peak hand position (ms) Time to peak downward speed (ms) Time to peak upward speed (ms) Time to peak lateral speed (ms) Time from deltoid onset to peak hand speeds and position Compared to younger adults, older adults generally showed reduced t_onset-to-peak for v_up and v_lateral (p s<0.002; Figure 5.4). Older adults also showed reduced t_onset-to-peak for Peak_Position, though the differences were not significant (298ms versus 272ms for younger and older adults respectively; p=0.063). The two groups converged for t_onset-to-peak for v_down (p=0.344), which occurred approximately at t_contact (362ms and 350ms for younger and older adults respectively). t_onset-to-peak for Peak_Position, v_up and v_lateral all increased with rail height (p s<0.001). (a) YA: t = h N (b) YA: t = h N 600 OA: t = h N 600 OA: t = h N (c) Rail height (% individual height) YA: t = h N Rail height (% individual height) (d) 600 YA: t = h N OA: t = h N OA: t = h N YA data 100 YA data Time to peak opening speed (ms) Time from peak opening speed to contact (ms) Time to peak opening speed (ms) 0 0 YA data OA data YA model OA model 200 Time to peak opening speed (ms) YA Rail height (% individual height) YA Rail 450 OA data OA data height (% individual height) 450 Time from peak opening speed OA data to contact (ms) OA data YA OA Younger model model adults -250 data YA OA Younger model modeladults regression line 400 YA data OA Older OA model dataadults - data 200 OA Older model adults regression line YA model Figure 5.4: t_onset-to-peak OA model variables: (a) upward hand speed (t_onset-to-peak for v_up); (b) lateral hand speed (t_onset-to-peak for v_lateral); (c) hand position (t_onset-to-peak for Peak_Position); and (d) downward hand speed (t_onset-to-peak for v_down). Equations to generate the regression lines are above the respective figures (t represents the relevant timing variable). h N represents handrail height normalized to a percentage of individual height, and zero-mean Time from peak opening speed to contact (ms) 300

94 ct (ms) o d contact vel to contact (ms) (ms) vel ontact to contact (ms) (ms) Time to contact from peak upward hand speed (ms) Time to contact from peak lateral hand speed (ms) Time to contact from peak upward hand position (ms) Time to handrail contact from peak hand speeds and position Older adults prolonged t_peak-to-contact for v_up, v_lateral and Peak_Position (p s<0.028; Figure 5.5). As handrail height increased, t_peak-to-contact for v_up, v_lateral and Peak_Position generally decreased (Height and Height 2 p s<0.002) (a) YA: t = h 600 N h 2 N (b) YA: t = h N h 2 (c) N 400 OA: t = h N h 2 N OA: t = h N h 2 N YA: t = h N h 2 N 400 OA: t = h N h N Peak upward vel 100 to contact (ms) Peak lat vel to contact (ms) Peak upward 60 vel to 70contact YA (ms) data Peak lat vel 70 to contact (ms) YA data Rail 400height (% individual height) OA data Rail height 400(% individual height) OA datarail height (% individual height) peak opening speed (ms) Time from peak opening speed 300 to contact (ms) 300 YA model data YA model 300 data 600 Younger adults - data YA data YA data OA Younger model data adults regression line OA model data OA data Older OA data adults - data YA Older model adults regression 200 line YA model YA model YA model OA model OA model OA model OA model Figure 5.5: t_peak-to-contact variables: (a) upward hand speed (t_peak-to-contact for v_up); (b) lateral hand speed (t_peak-to-contact for v_lateral); and (c) hand position (t_peak-to-contact for v_down). Equations to generate the regression lines are above the respective figures (t represents the relevant timing variable). h N represents handrail height normalized to a percentage of individual height, and zero-mean Peak hand speeds during the reach trajectory trajectory On average, younger adults demonstrated higher peak hand speeds than younger adults, though the differences were only significant for v_lateral and v_down (p Lateral =0.026; p Down =0.014; p Up =0.274; Table 5.5 and Figure 5.6). The most notable influence of handrail height on peak hand speed was for v_up, which increased with handrail height (Height p<0.001). v_lateral and v_down also varied significantly with rail height (v_lateral: Height and Height 2 p s<0.001); v_down: Height 2 p=0.002).

95 Table 5.5: Summary of speed variable performance with respect to handrail height for both younger and older adults. Significant effects and interactions are bolded. Variable Peak upward hand speed (v_up) (m/s) Younger adult performance (mean (SD)) Older adult performance (mean (SD)) Age p Height p Height 2 p Age*Height p Age*Height 2 p 1.81 (0.74) 1.74 (0.53) < Peak lateral hand speed (v_lateral) (m/s) 2.37 (0.54) 2.10 (0.47) <0.001 <0.001 (both age groups) Peak downward hand speed (v_down) (m/s) 1.78 (0.58) 1.45 (0.45) Table 5.6: Summary of accuracy variable performance with respect to handrail height for both younger and older adults. Note that significant interactions between age and rail height were not found for any of these variables, so the terms are not reported in the table. Significant effects are bolded. Variable Younger adult performance (mean (SD)) Older adult performance (mean (SD)) Age p Height p Height 2 p Vertical handrail overshoot (Overshoot) (cm) Peak hand position (Peak_Position) (cm) 11.8 (4.6) 12.5 (4.5) <0.001 < (9.3) (9.3) <

96 ct (ms) o d contact vel to contact (ms) (ms) vel ontact to contact (ms) (ms) Peak hand speed upward (m/s) Peak hand speed lateral (m/s) Peak hand speed downward (m/s) (a) (b) (c) YA: Speed = h N YA: Speed = h N h 2 N YA: Speed = h N h 2 N OA: Speed = h N 600 OA: Speed = h N h 2 N OA: Speed = h N h 2 N Peak upward vel 100 to contact (ms) Peak lat vel to contact (ms) Peak 60 upward vel 70to contact YA (ms) data Peak 70lat vel to contact 40 (ms) YA data Rail 400 height (% individual height) OA data Rail height (% 400 individual height) Rail OA height data (% individual height)z k opening speed (ms) Time from peak opening speed 300 to contact (ms) 300 YA model data YA model 300 data 600 Younger adults - data YA data YA data OA Younger model data adults regression line OA model data OA data Older OA 300 data adults - data YA Older model adults regression 200 line YA model YA model YA model OA model OA model OA model OA model Figure 5.6: Peak hand speed variables with respect to handrail height: (a) upward (v_up); (b) lateral (v_lateral); and (c) downward (v_down). All points represent the mean data from three trials at the rail height of interest for a given participant. Equations to generate the regression lines are above the respective figures. h N represents handrail height normalized to a percentage of individual height, and zero-mean Accuracy of reaching trajectory Age did not significantly affect Overshoot or Peak_Position (p s>0.073; Figure 5.7a/b and Table 5.6). Both Overshoot and Peak_Position varied significantly with handrail height (Height and Height 2 p s<0.020). Notably, Overshoot decreased as handrail height increased, which suggests that participants reached more accurately for higher handrails. The concomitant increase in Peak_Position with increasing handrail height suggests that the reduced Overshoot did not stem from a generic response to perturbations (i.e. participants quickly raising their hands to a common height). Instead, participants modulated their hand trajectories based on the height of the rail. Traces showing the position of the hand marker with respect to time in the lateral and vertical direction are presented in Figure 5.1c. Pearson correlations revealed that Overshoot was strongly- and negatively-associated with the initial vertical distance between the hand and the handrail at t_onset (r = ), though very weakly- and negatively-associated with the initial lateral distance between the hand and the rail (r = ) [120]. 82

97 contact vel to contact (ms) (ms) ntact l to contact (ms) (ms) Vertical handrail overshoot (cm) Peak hand position (cm) Vertical hand position (cm) 83 (a) 600 (b) YA: Overshoot = h N h 2 N YA: Peak = h N h 2 N OA: Overshoot = h N h 2 N OA: Peak = h N h 2 N Peak upward vel 100 to contact (ms) Peak lat vel to contact 90 (ms) in Peak 60upward 70 vel to contact YA (ms) data Peak lat vel to contact (ms) YA data Rail 400 height (% individual height) OA data Rail height (% individual 400 height) 80 OA data om peak opening speed to contact (ms) model YA data 1 0 YA 0.2 model data Younger adults - data YA data OA Younger model data adults regression line Time OA relative model data to perturbation Older OA 300 data adults - data YA Older model adults regression 200 line YA model onset (ms) YA model OA model OA model Figure 5.7: Accuracy variables with respect to handrail height. (a) Vertical handrail overshoot (Overshoot). (b) Peak hand position (Peak_Position). (c) Sample hand trajectory traces with respect to time for a characteristic older adult participant, in the vertical direction. Three handrail heights are included: 32 inches (45.5%h), 38 inches (54.0%h), and 44 inches (62.6%h) All three trials for each handrail height are shown. Note the decrease in vertical handrail overshoot with increasing rail height. (c) in 38 in Videos from a selection of trials were reviewed to gain further insights into whether a startle -like response was present in the early stages of balance recovery, with emphasis on the behavior of both the reaching and non-reaching arms (Figure 5.8). We did not find evidence of a generic response to balance loss. Instead, balance recovery strategies varied. For example, one young participant quickly raised her right (reaching) hand to reach for the rail (Figure 5.8a-iii), but took longer to raise her left hand (Figure 8a-iv), which ultimately moved away from her trunk and likely contributed to stability by increasing her moment of inertia relative to her feet. In a latter trial, however, the same participant raised both arms quickly (albeit asymmetrically), and did not reach laterally toward the rail until well after her reaching hand achieved its peak position (Figure 5.8b). In Figure 5.8c, an older participant initiated reaching with the grasping hand well before lifting his other arm in the upward and leftward directions. This behavior is consistent with a strategy that could counteract his momentum when leaning downward and toward the rail, akin to that observed following upright stance and lean-and-release perturbations without handholds [36, 44]. Finally, a separate older participant did not appear to raise her non-reaching arm at all, even in the early trial shown in Figure 5.8d, and despite being destabilized to the extent that one foot slipped off the floor. Instead, she moved her non-reaching arm posterior to her body, consistent with a protective response that would help to break a fall by reducing the impact between the hips and the ground [45].

98 84 (a) i ii iii iv v vi vii viii ix (b) i ii iii iv v vi vii viii ix (c) i ii iii iv v vi vii viii ix (d) i ii iii iv v vi vii viii ix Figure 5.8: Observed strategies during balance recovery. Perturbation onset was approximately at frame i. (a) A young participant begins to reach for the rail (frame iii), but takes longer to move her left hand upward and away from her trunk. (b) The same participant, in a later trial. Both arms rise quickly after perturbation onset, but asymmetrically. The participant does not reach laterally with her right hand until frame vii well after peak hand position occurs. (c) A later trial for an older participant. Rapid movement initiation of the left hand appears to start after the right hand (frames iv and ii respectively). The left hand continues to rise well after grasp completion. (d) A different older participant, in an early trial. The left arm does not appear to rise; instead, the participant reaches downward and backward. This may be a protective strategy in anticipation of a fall, given the participant s apparent difficulty in restabilizing (as evidenced by the left foot sliding off the walking surface from frames vii to ix). 5.4 Discussion This study characterized the effect of handrail height and aging on the speed and accuracy of reachto-grasp balance reactions during slope descent. Aging was associated with slower deltoid onset, slower peak lateral and downward hand speed, reduced time to peak upward and lateral hand speed after deltoid onset, and increased time to handrail contact from peak hand speeds. Handrail contact time and movement time did not differ significantly between groups. As handrail height increased, time to peak upward and lateral hand speed from deltoid onset increased, while time from peak

99 85 speeds to rail contact decreased. Increasing rail height generally resulted in greater trajectory accuracy, indicated by reduced vertical handrail overshoot in both age groups Influence of aging In this study, older adults exhibited longer deltoid onset latencies than young adults, which suggests a longer response time in this population following the platform perturbation. Despite this, the older group took less time to reach peak hand velocities after deltoid onset occurred (i.e. reduced t_onsetto-peak by ~30 to 40ms). The delays in deltoid onset are consistent with age-related differences in arm muscle activation following translational [75] and rotational [35] perturbations of upright stance, and may reflect decreased ability to detect support-surface translations due to age-related declines in plantar-surface cutaneous sensation or other sources of lower-limb proprioceptive feedback [73, 68]. Conversely, the reduced t_onset-to-peak in older adults in this study may imply heightened anxiety, potentially due to increased fear of falling from a perturbation. Although we are not aware of published literature on anxiety in older adults during balance recovery, modulation of muscle activity in response to increased fall anxiety has been shown during quiet stance [150, 151], and fear of falling has been associated with increased postural sway in older adults [152]. In conditions with increased perceived injury risk from falling, younger adults have also demonstrated shorter latencies to the first peak center-of-mass velocity following balance perturbations, as well as heightened muscle activation in the middle deltoid and medial gastrocnemius [153, 141]. Regardless of the underlying reason, the tendency of older adults in this study to shorten t_onset-to- Peak may also have allowed them to prolong t_peak-to-contact (by ~30ms), without consequence to overall t_contact or t_movement. In volitional reaching, this has been proposed as a strategy to increase the time available to correct the hand trajectory leading into object contact [80]. Prolonged t_peak-to-contact been observed in young adults who lacked prior visual information on the handhold position before balance perturbations [77] or who relied purely on peripheral vision during reactive grasping [88]; authors hypothesized that this strategy helped to improve grasping accuracy in those contexts. For older adults, this strategy may also reflect the combination of possible increased reliance on visual feedback to guide the reach-to-grasp movement (due to age-related declines in upper-limb proprioceptive acuity [154, 155]), along with age-related declines in the speed of processing visual information [156]. This may have contributed to the reduced v_down (by ~0.35m/s) observed in older adults, leading into handrail contact. Alternatively, the reduced v_down in older adults may simply reflect a fear of injury from colliding with the handrail.

100 Influence of handrail height Our findings show that handrail height affected t_contact and t_movement for both young and older adults, with significant interaction effects with age further observed for t_movement. The fastest regression-predicted t_movement was at approximately 56%h for young adults, but at 64%h for older adults (Figure 5.3). These heights approximate mid-forearm and elbow height in adults [118]. Notably, handrails with the lowest initial vertical distance from the hand (around 46%h in this study) were not contacted more quickly than higher rails, which require the participant to reach further. More direct trajectories have been observed in studies with rapid volitional reaching [66, 85, 145]), which would support minimizing the vertical trajectory distance (i.e., overshoot) between the handrail and the hand. However, our data demonstrate that rather than adopting a near-linear trajectory towards the rail (with some minimum vertical clearance to allow the fingers to wrap around the rail), participants often raised their reaching hands quickly following perturbation onset, even when the handrail was at or below the initial position of the hand (Figure 5.7c). This response required individuals to reach downward to grasp for all handrail heights, and likely contributed to the increased vertical overshoot observed with lower handrails in this study. While participants tended to raise their reaching hands quickly after perturbation onset, we did not find evidence that this comprised a generic or startle -like response to balance disturbances, as observed previously [64]. Instead, the reaching and non-reaching arms often moved asymmetrically (Figure 5.8), with rapid wrist movement generally appearing sooner in the reaching arm. Further, the behavior of the non-reaching arm was often consistent with stabilizing strategies (e.g., increasing the moment of inertia of the rotating trunk relative to the feet; counteracting linear momentum of the trunk [36, 44]), or protective strategies (e.g., using the arms to reduce potential impact between the hips and the ground [45]). When combined with the significant increase in peak hand position with rail height, this demonstrates that the increase in vertical handrail overshoot with rail height is unlikely to have resulted from a generic response to balance loss (which would have been characterized by arms being raised to a common height). On this basis, the raising of the reaching arm above the handrail (which led to vertical overshoot) may well have a functional role, such as providing sufficient clearance for the fingers leading into grasping. An alternate possibility is that participants were more reliant on peripheral vision to see lower rails leading into perturbation onset, and anecdotally participants described higher handrails as easier to see. Greater deviations from direct path trajectories and prolonged t_peak-to-contact from maximum wrist velocity have also been previously observed where participants relied solely on peripheral vision during reactive grasping

101 87 [88]. While we do not have eye-tracker data to confirm participant gaze behavior, the general increase in both vertical handrail overshoot and t_peak-to-contact (for each of Peak_Position, v_up and v_lateral) for lower handrails in the present study is consistent with this prior work [88]. However, further research is required to understand if handrail height affects gaze behavior and the presence of the rail in a person s visual field, and any concomitant effects on the reach trajectory. While statistically-significant effects of handrail height on t_contact and t_movement were observed for both populations, the functional importance of these effects is difficult to evaluate. Based on the regressions (Figure 5.3), the slowest modeled t_movement for older adults within the range of heights tested (occurring around 42%h) only exceeded the fastest modeled t_movement (at 64%h) by 50ms. Further, the variability in actual t_contact and t_movement was high, relative to the difference between the slowest and fastest regression-modeled times (Figure 5.3). This suggests that the observed effects of handrail height on the speed and accuracy of reactive grasping may not have substantially affected overall balance recovery in this study. However, the reduced trajectory accuracy and increased movement times with lower rail heights may have greater consequences to grasping effectiveness in older adults in real-life situations, where the balance disturbances are less predictable, or when environmental conditions preclude effective reactive stepping (e.g. icy ramps; wet bathrooms). Further research is needed to evaluate the influence of rail height on grasping effectiveness, particularly for lower heights that reduced reaching speed and accuracy for older adults in this study Study limitations This study provides important insights into the effect of handrail height and aging on the speed and accuracy of reactive grasping. However, limitations should be addressed. First, this work explored handrail heights within 30 to 44 inches (41 to 72% of participant height). This range exceeds the limits of many building code requirements for handrail height on ramps and other walkways (e.g., [101, 100]), and is thus likely to apply to many handrails in the community. However, performance in our variables of interest may differ with handrails that are higher or lower than our tested range. While we evaluated detailed trajectory kinematics, we did not assess the effectiveness of each response in terms of achieving a full power grip following rail contact. Further, participants were told in advance to grasp the rail and experienced repeated trials; hence, they may have pre-planned aspects of their responses. As well, this study was limited to a single handrail shape. However, more complex shapes may require more precise control of the trajectory to achieve a functional grip.

102 88 Finally, our sample only included healthy adults (i.e. free of neurological, cardiac or musculoskeletal conditions such as arthritis or osteoporosis), due to the challenging nature of the protocol. Using balance perturbations to elicit reaching reactions was important to our study context, as other cues to initiate reach-to-grasp movements (e.g., sound) have been reported to result in slower deltoid activation [66, 157] and handhold contact time ( [66]). However, the use of balance perturbations also constrained the study sample to relatively-healthy participants. Further to this, persons with reduced upper-limb range of motion (due to arthritis, stroke, or other conditions [158, 159]) may be differentially affected by the handrail height due to pain associated with upper limb elevation, challenges in quickly reaching upward, or limited ability to quickly extend their elbow to reach for lower heights. Further research should explore how rail height affects the speed and accuracy of reactive grasping in these populations. 5.5 Conclusions We have quantified the influence of handrail height on the speed and accuracy of perturbationevoked reach-to-grasp reactions in younger and older adults during ramp descent. Handrail height significantly affected time to handrail contact and movement time in both populations, with movement time optimized at higher handrail heights for older adults. Younger adults took longer than older adults to reach their peak upward and lateral hand speeds after middle deltoid onset. However, younger adults contacted the rail more quickly than older adults after their respective peak upward and lateral hand speeds, peak aperture opening speed, and peak hand position were achieved. The delayed time to contact from peak upward and lateral hand speed and peak aperture opening speed in older adults may imply a reaching strategy that allows for greater corrective actions to the fingers and overall trajectory leading into handrail contact. As handrail height increased, peak upward and lateral hand speed generally increased, though both populations took longer to reach peak speeds in those directions after deltoid onset. Younger adults demonstrated significantly higher peak reaching speeds in the lateral and downward directions than did older adults. For both populations, vertical handrail overshoot generally decreased as rail height increased. When considered with how peak hand position increased with rail height, the reduced vertical handrail overshoot and increased peak upward and lateral hand speeds suggests substantial modulation of the reach trajectory quickly after deltoid onset. Our findings imply a possible speed and accuracy advantage with increased rail height (up to elbow height) in older adults, as demonstrated by reduced movement time and vertical handrail overshoot in these conditions.

103 89 Chapter 6 Study 4: Effect of handrail height and age on trunk and shoulder kinematics following perturbation-evoked grasping reactions during level-ground walking and slope descent Preface This chapter is currently in preparation for targeted submission to the journal Human Factors. At present, it is written as a thesis chapter based on a secondary analysis of the dataset for Study 3 (in the previous chapter). I conceptualized the study, collaborated to develop the protocol and collect data, analyzed the data, and wrote the chapter herein. Abstract Prior studies in this thesis suggest that higher handrails (up to 42 inches) may be advantageous for stability, following perturbations of upright stance. However, we have not yet established if these advantages are present following perturbations during gait, or if rails that are 42 inches high are likely to be reachable by individuals with limited range of motion in the shoulder. In this secondary analysis of data collected in the previous chapter, we characterize age-related differences in trunk and shoulder kinematics following perturbation-evoked reach-to-grasp reactions, during both levelground walking and slope descent, and the impact of handrail height. The peak resultant handrail forces during balance recovery were also evaluated, to provide insights into the concomitant demands on the handrail during grasping, and possible differences between conditions of the contribution of the handrail to stability. Increased handrail height was generally associated with reduced trunk angular displacements and velocities, indicating greater ability to remain upright. However, these stability gains with increasing rail height may be offset by the increased peak shoulder elevation angles, which surpassed the active range of motion of many with osteoarthritis in the shoulder. Slope descent mostly resulted in reduced forward trunk pitch displacement and velocity, and reduced handrail forces. Older adults generally demonstrated comparable or reduced trunk angular displacements and velocities. Collectively, the results from this study demonstrate benefits to trunk control with increasing handrail height during both level walking and slope descent, but provide an important caveat that these benefits may not extend to those with limited range of motion in the shoulder.

104 Introduction The results presented in Study 1 and Study 2 (Chapters 3 and 4, respectively) have shown a stability advantage with higher handrails (among rail heights of 34, 38 and 42 inches), following perturbations of upright stance. Study 3 (Chapter 5) revealed that increasing handrail height generally helped to reduce reach-to-grasp movement time in older adults (up to 64% of individual height, or approximately elbow height [118]), following perturbations during slope descent. Further, increasing handrail height largely resulted in higher peak upward and lateral hand speeds and lower vertical handrail overshoot, among both younger and older adults. Despite this support for higher handrails to enhance balance recovery, many outstanding questions remain. In particular, we have not yet established whether the stability advantages with increasing handrail height are present following perturbations during ongoing gait. While many of our detailed stability measures focused on trunk and centre of mass (COM) kinematics with feet in place, we have not explored whether the observed effects of handrail height on these measures are present when reactive stepping is available following balance perturbations. The testing environment is also of interest. To our knowledge, Maki and colleagues perturbation study on stairs with four young participants [3] is the only published study to have explored balance recovery via handrail grasping in an environment other than level ground. Thus, the extent to which balance recovery findings on level ground translate to other locations where handrails are common, such as slopes, should be evaluated. Finally, detailed kinematics of the grasping arm during balance recovery should be characterized with respect to handrail height. While higher handrails may appear to be advantageous for balance recovery, they may also require substantial elevation of the upper arm with respect to the thorax in order to be reached. Such grasping postures may well exceed the range of motion of many with upper-limb arthritis, hemiplegia due to stroke, or other conditions associated with aging that limit movement in the joints [160, 161]. Quantifying the effect of handrail height on thoracohumeral (shoulder) elevation angles during reach-to-grasp balance reactions where the user cannot pick their grip location, and where the ongoing motion of the trunk must be considered can help to illuminate the extent to which higher handrails may be problematic (if at all) for users with limited range of motion in the upper-arm.

105 91 The primary purpose of this study is to characterize age-related differences in trunk and shoulder kinematics following perturbation-evoked reach-to-grasp reactions, during both level-ground walking and slope descent, and the impact of handrail height. The peak resultant handrail forces during balance recovery were also evaluated, to provide insights into the concomitant demands on the handrail during grasping, and possible differences between conditions of the contribution of the handrail to stability. We expected that shoulder elevation angles would increase with handrail height. We also expected that greater trunk control would be observed with increased handrail height. Finally, we expected that older adults would demonstrate poorer stability than younger adults, which would manifest in higher trunk angular displacements and velocities following balance perturbations. 6.2 Methods This study comprises a secondary analysis of the dataset collected in the previous chapter, including the same participants, experimental setup, and protocol. Briefly, data from fourteen younger adults and thirteen older adults were analyzed. All analyzed participants were right-handed. Participants walked back and forth inside the laboratory, mounted to the robotic platform in the Challenging Environments Assessment Laboratory. During randomly-selected walks when the handrail was to the right of participants, the platform quickly translated backward to disrupt participant balance. Participants were instructed to reach to grasp the handrail as quickly as possible upon experiencing a perturbation. See Figure 6.1 for a schematic diagram of the testing setup. A total of eight handrail heights were tested: 30, 32, 34, 36, 38, 40, 42 and 44 inches (76, 81, 86, 91, 97, 102, 107 and 112 cm, respectively). In addition to the slope descent trials described in the previous chapter, data were also analyzed during level-ground walking (i.e. two slope conditions). After completing a familiarization trial, participants completed four perturbation trials per handrail height and walking surface incline. The first three of these trials are analyzed in this study, for a total of 48 trials per participant.

106 92 (a) (b) 1.2 m elevation angle (+) 1.2 m 0.6m (c) plane of elevation angle (+) 4.8 m plane of elevation angle (-) (d) gait speed estimated when participants walked along this row of force plates perturbation triggered when participant stepped on this row of force plates wood tiles roll (+) pitch (+) Figure 6.1: Laboratory setup and definitions of kinematic outcome measures. (a) Schematic diagram of the laboratory transverse (top) and sagittal (bottom) plane views. Participants walked back and forth along the 4.8m-long walking surface, with the centre-line of the surface approximately aligned with participants sagittal plane. (b) Thoracohumeral (shoulder) elevation angle definition comprising the elevation of the upper arm with respect to the thorax. (c) Shoulder plane of elevation angle definition comprising the angle between the humerus and the participant s coronal plane (at the level of the thorax). Anterior positioning of the humerus is characterized by a positive angle (effectively horizontal flexion), while posterior positioning of the humerus is characterized by a negative angle (effectively horizontal extension). (d) Trunk roll and pitch angular definitions comprising the roll and pitch of the trunk with respect to the vertical axis (i.e. parallel to gravity) Kinematic data collection, processing and modeling Reflective motion capture markers were used to track trunk and upper-limb kinematics. Rigid marker clusters were secured to the pelvis (cluster at the sacrum) and upper-body (cluster near the thoracic level of T12). The distal and proximal ends of these segments were identified in a neutral, stationary pose. Movement of the right upper-arm was tracked via markers on the elbows (medial

107 93 and lateral epicondyles) and shoulders (acromion; front, back and distal side of the glenohumeral joint). Kinematic and kinetic data sources were synchronized offline [136]. Kinematic data were collected with twelve motion capture banks (Motion Analysis Corporation, Santa Clara, CA), at a sample rate of 250 Hz. Trunk and shoulder angular kinematics were estimated with a link segment model, consisting of three segments (pelvis, trunk, right upper arm (Visual 3D; C-Motion Inc, Germantown, MD). All motion capture markers were filtered with second-order, zero-lag, low-pass Butterworth filters prior to model estimation (MATLAB; The Mathworks, Inc, Natick, MA). Markers on the trunk and shoulders were filtered with a cut-off frequency of 10Hz, while markers on the epicondyles were filtered at 30Hz. Filter parameters were selected based on a combination of power analysis and visual inspection. Specifically, 99% of the signal power for all markers used in analysis was below 10 Hz (based on data from three representative trials). However, the position signals from the epicondyles were found to be quite sensitive to filtering around the time that participants contacted the handrail (likely due to the sudden deceleration of the lower-arm segment when the hand contacted the handrail), with >1cm of position error imposed by the filter in some cases. Filtering at 30 Hz allowed the key features of these signals to be preserved, and thus enabled more accurate estimation of shoulder angles Handrail force data collection and processing Handrail force data were collected via tri-axial load cells (AMTI MC3A-1000; Advanced Medical Technology, Inc, Watertown, MA) mounted to the handrail posts (one on either end of the rail), at a sample rate of 1000 Hz. Inertial artifacts in the handrail force signals (resulting from platform motion) were removed by subtracting force recordings collected without a participant contacting the handrail. Handrail force signals were then filtered with a second-order, dual-pass, low-pass Butterworth filter (8Hz), to attenuate noise due to the vibration of the handrail. The filter cut-off frequency was based on the lowest natural frequency of the handrail (approximately 10Hz, based on Fourier analysis), which occurred when the rail was 44 inches high Outcome measures Upper-limb kinematics The primary upper-limb variables of interest were the (1) peak thoracohumeral ( shoulder ) elevation angles, and (2) peak shoulder plane of elevation angles (see Figure 6.1b and c for illustrated

108 94 definitions). Shoulder angles were defined based on the joint coordinate systems recommended by the International Society of Biomechanics [162], The shoulder elevation angle represents the elevation of the humerus with respect to the thorax, where 0 o was defined as when the humerus was parallel with the individual s sagittal and coronal planes [163]. Conversely, the peak shoulder plane of elevation angle represents the transverse-plane rotation of the humerus relative to the participant s coronal plane (Figure 6.1c). The shoulder plane of elevation angle had a value of 0 o when the upper arm coincided with the participant s coronal plane; a value of 90 o occurs when the upper arm is anterior to the thorax and coincident with the participant s sagittal plane [163]. Prior biomechanical studies have described positive plane of elevation angles to be analogous to horizontal flexion, while negative plane of elevation angles are analogous to shoulder extension [163]. Peak shoulder elevation angles were measured (1) between perturbation onset and grasp completion (i.e. Elevation_to_Grasp), and (2) between perturbation onset and completion of using the handrail for recovery (i.e. Elevation_to_Recover). Note that for some trials, Elevation_to_Grasp and Elevation_to_Recover were equal, if the highest observed elevation angle during the trial occurred before grasp completion. The shoulder plane of elevation angles concurrent with Elevation_to_Recover were also estimated (i.e., Plane_to_Recover), to provide insights into the postures that participants adopted when high elevation angles occurred Trunk kinematics Trunk angular kinematic data were analyzed to evaluate balance control. Metrics were defined with respect to the vertical axis in global coordinates, in the individual s coronal (roll) and sagittal (pitch) planes (Figure 6.1d). Variables of interest included peak forward trunk angular displacement and velocity, trunk roll range, and peak angular displacement and velocity toward and away from the handrail. Peak angular displacement variables were defined as the difference between peak trunk angular position along the axis of interest, and the participant s trunk angular position in the frame immediately before perturbation onset. Conversely, roll range was defined as the difference between the highest and lowest trunk roll angles (i.e. toward the rail and way from the rail) Handrail forces The peak resultant forces that participants applied to the handrail were extracted, to quantify the contribution of the handrail to balance recovery. Handrail forces were analyzed as a percentage of participant body weight (%BW).

109 Gait speed Participant gait speed leading into the perturbation was estimated to characterize possible differences in behavior due to age or walking surface incline. Gait speed was defined as the quotient between (a) the total anterior displacement of a marker on the participant s pelvis cluster, starting from 1.2m away from the force plate that triggered perturbations (Figure 6.1a), through to the frame before perturbation onset, and (b) the time required for the participant to walk that distance Data analysis Analyses for shoulder angles were based on the highest observed Elevation_to_Grasp and Elevation_to_Recover, recorded from the three trials collected for each slope and handrail height condition for each participant. As the associated research objective sought to understand if high elevation angles (surpassing the range of motion for many with osteoarthritis in the shoulder) were observed during balance recovery, we preserved the worst-case outcome for each condition. The concomitant Plane_to_Recover angles (i.e., those measured when the highest Elevation_to_Recover angles were observed) were analyzed. Conversely, analyses for trunk angular kinematics, peak handrail forces, and gait speed, involved the mean value of each participant s performance across the three trials for each slope and handrail height condition. This was done to minimize the impact of spurious trials on evaluations of walking surface incline and handrail height. Statistical analyses were performed for each variable of interest using a 2x2x8 repeated measures ANOVA (SAS Enterprise Guide version 9.1, Cary, NC), with walking surface slope (level ground, 8 o slope descent) and handrail height (30, 32, 34, 36, 38, 40, 42 and 44 inches) comprising the withinsubject factors. Age was modeled as a between-subject factor. All outcome measures were rank-transformed to meet the normality assumptions required for the ANOVA [139]. Main effects (age, slope, handrail height) and interactions (age*slope, age*height, slope*height) were modeled, with significance levels of p<0.05 used for all analyses. Post hoc comparisons with Tukey adjustments were performed following identification of significant main effects and interactions, to identify differences in participant walking strategy between age groups and slopes.

110 Results All participants analyzed in this study completed the full data collection protocol without adverse events. For one participant, data from two handrail heights were excluded due to the timing of middle deltoid onset exceeding 1s, which suggested that the reach-to-grasp movements were not compensatory (as described in Chapter 5). Those data points were thus eliminated from analyses, resulting in incomplete datasets for that participant. On average, gait speed in young adults was 0.77m/s for both level walking and slope descent. In older adults, gait speed was on average 0.78m/s during level walking, and 0.72m/s during slope descent Shoulder elevation and plane of elevation angles Elevation_to_Grasp did not reveal significant main or interaction effects (p s > 0.081) (Figure 6.2a, b). However, Elevation_to_Recover demonstrated significant main effects of age (p<0.001), handrail height (p=0.011) and slope (p<0.001), as well as a significant age*slope interaction (p=0.039) (Figure 6.2c, d). Specifically, older adults generally exhibited higher Elevation_to_Recover angles than did younger adults. Further, slope descent resulted in higher Elevation_to_Recover angles than did level walking. This was likely due to participants continuing to step while maintaining handrail grasping during balance recovery, which resulted in greater arm elevation (with the hand posterior to the trunk) during slope descent. Elevation_to_Recover angles generally increased as handrail height increased, particularly among older adults during slope descent (Figure 6.2d). Interactions effects between slope*height, and age*height, were not significant (p s > 0.229) for Elevation_to_Recover angles. Plane_to_Recover angles were also analyzed (Figure 6.2e, f). In general, participants were reaching in the posterior direction when the Elevation_to_Recover angles were observed, indicated by the prevalence of negative Plane_to_Recover angles. On average, older adults demonstrated lower Plane_to_Recover than young adults (p<0.001), indicating more frequent posterior reaching in older adults. Further, Plane_to_Recover was significantly lower during slope descent compared to level ground walking (p=0.001), indicating that grasping during slope descent more frequently involved shoulder extension during high elevation angles. Significant main effects of handrail height, and significant interaction effects, were not found (p s>0.289).

111 Shoulder plane of elevation angle ( o ) Shoulder plane of elevation angle ( o ) Peak shoulder elevation angle to recover ( o ) Peak shoulder elevation angle to recover ( o ) Peak shoulder elevation angle to grasp ( o ) Peak shoulder elevation angle to grasp ( o ) 97 (a) (b) young adults older adults young adults older adults Handrail height (inches) Handrail height (inches) (c) (d) young adults older adults young adults older adults Handrail height (inches) Handrail height (inches) (e) (f) young adults older adults young adults older adults Handrail height (inches) Handrail height (inches) 60 flexion extension flexion extension Figure 6.2: Peak shoulder elevation angles and concomitant plane of elevation angles during balance recovery. Top row: Peak shoulder elevation angles between perturbation onset and handrail grasp completion (Elevation_to_Grasp), during (a) level-ground walking, and (b) slope descent. Middle row: Peak shoulder elevation angles between perturbation onset and balance recovery with the handrail (Elevation_to_Recover), during (c) level-ground walking, and (d) slope descent. Bottom row: Shoulder plane of elevation angles measured during Elevation_to_Recover (i.e., Plane_to_Recover), during (e) level-ground walking, and (f) slope descent. Positive angles indicate shoulder horizontal flexion; negative angles indicate shoulder horizontal extension. Younger adults are indicated on the left of each plot (grey); older adults are indicated on the right (black). Median values are indicated by the horizontal lines in the middle of each box, while the boxes represent the 25 th and 75 th percentiles. Whiskers indicate maxima and minima, up to 1.5 times the length of the interquartile range (25 th to 75 th percentile). Data values beyond this range were considered outliers. The red line (82.2 o ) in plots a-d indicates the mean, maximum active shoulder forward elevation angle that could be achieved by individuals with osteoarthritis in the shoulder, before shoulder arthroplasty operations. The area shaded in red represents mean + 1 standard deviation (33.3 o ), within the range of angles represented on the graph.

112 98 The observed shoulder elevation angles have been compared to literature, which reports the average active maximum forward elevation angles achieved by 62 older adults with osteoarthritis in the shoulder, before undergoing total shoulder arthroplasties (mean + 1SD: 82.2 o o ; see Figure 6.2) [164]. All testing conditions resulted in some participants exhibiting Elevation_to_Grasp and Elevation_to_Recover angles that were within 1SD of the mean (48 o ). However, the general increase in Elevation_to_Recover with increasing handrail height particularly for older adults during slope descent emphasizes a plausible challenge to balance recovery that higher handrails could create among individuals with osteoarthritis in the shoulder, who may not have the range of motion to apply grasp and apply high stabilizing forces to handrails at that height Trunk angular kinematics Trunk roll displacement, range and peak velocity toward and away from the rail Handrail height significantly affected trunk roll displacement and velocity toward the handrail, and trunk roll range, which decreased as handrail height increased (rail height p s <0.001) (Figure 6.3a, c, d). In contrast, trunk roll displacement away from the rail increased significantly as handrail height increased, where participants generally had to lean away from the rail in order to reach it when at higher heights (p<0.001) (Figure 6.3b). Although statistically significant, it should be noted that the increased magnitudes of roll displacement away from the rail at the highest height (~8-10 o ) were much lower than displacement toward the rail when at the lowest heights (~20 o ). Younger adults demonstrated significantly higher peak roll velocities toward the rail, compared to older adults (p=0.038; Figure 6.3d). The most notable effect of slope was for peak roll velocity toward the rail, which was reduced by roughly 4 o /s during slope descent (p=0.008). In contrast, peak roll displacement and velocity away from the rail were higher during slope descent (p s<0.047; Figure 6.3b, e), though the actual difference between magnitudes was negligible (0.5 o for displacement; 0.8 o /s for velocity). Trunk roll kinematics, including all p-values, are summarized in Figure 6.3.

113 Peak roll velocity away from rail ( o /s) Peak roll velocity toward rail ( o /s) Peak roll range ( o ) Peak roll displacement away from rail ( o ) Peak roll displacement toward rail ( o ) 99 (a) (b) young adults older adults Level Slope p<0.001 for rail height p=0.043 for age*rail height p>0.250 for age, slope, age*slope, and slope*rail height (c) (d) (e) Level Slope Level Slope Level Slope Level Slope Handrail height and walking surface incline condition p<0.001 for rail height p=0.047 for slope p>0.198 for age, age*slope, slope*rail height, and age*rail height p<0.001 for rail height p=0.077 for slope p=0.090 for age*rail height p>0.312 for age, age*slope, and slope*rail height p<0.001 for rail height p=0.008 for slope p=0.038 for age p>0.525 for rail height*slope, age*slope, and age*rail height p<0.001 for rail height p=0.021 for slope p=0.009 for age*slope p>0.144 for age, slope*rail height, and age*rail height Figure 6.3: Trunk roll kinematics. (a) Peak displacement toward the rail. (b) Peak displacement away from the rail. (c) Trunk roll range. (d) Peak roll velocity toward the rail. (e) Peak roll velocity away from the rail. Bars represent mean values for each population and condition; error bars represent 1 standard deviation.

114 Peak forward pitch velocity ( o /s) Peak forward pitch angle ( o ) Trunk pitch displacement and peak velocity in the forward direction Peak forward trunk pitch displacement and velocity both varied significantly with handrail height (height p s <0.001), and generally decreased as handrail height increased (Figure 6.4). They were also generally reduced in older adults (age p s < 0.001). Slope descent resulted in reduced forward trunk pitch displacement and velocity compared to level gait (p s<0.004). Significant interaction effects between age and slope for both pitch metrics (p s<0.001) were also found. Post hoc comparisons revealed that younger adults demonstrated significantly greater pitch displacements and velocities following perturbations during level walking (p s < 0.001), but the differences between age groups were not significant during slope descent (p s > 0.096). Further, younger adults demonstrated significantly reduced pitch metrics for balance recovery during slope descent compared to level gait (p s<0.001). However, the reduction to pitch metrics was only significant for velocity in older adults (displacement p=0.853; velocity p=0.017). Significant interactions between rail height and slope, and rail height and age, were not found (p s > 0.145). (a) (b) Level Level Slope Slope Handrail height and walking surface incline condition young adults older adults p<0.001 for age, rail height, and age*slope p=0.004 for slope p>0.376 for rail height*slope and rail height*age p<0.001 for age, slope, rail height, and age*slope p>0.145 for rail height*slope and rail height*age Figure 6.4: Trunk pitch kinematics. (a) Peak forward trunk angular displacement. (b) Peak forward trunk pitch angular velocity. Bars represent mean values for each population and condition; error bars represent 1 standard deviation.

115 Peak resultant handrail forces (%BW) Handrail forces Peak handrail forces were included as an explanatory variable for the observed differences in performance in trunk angular kinematics with respect to age and rail height. Peak resultant handrail forces showed slight, but statistically-significant differences with handrail height (p<0.001; Figure 6.5), and generally decreased as rail height increased, from 32 inches (17.9%BW) to 44 inches (15.4%BW). However, post hoc comparisons revealed that the only significant differences between individual handrail heights were for 32 inches versus 44 inches, and 36 inches versus 44 inches (p s = 0.042). All other pairwise comparisons between rail heights were non-significant (p s>0.076). Handrail forces showed slight but statistically-significant reductions during slope descent compared to level-ground walking (17.8%BW for level versus 16.2%BW for slope; p<0.001). Interpretation should consider the significant age*slope interaction (p=0.006). Post hoc comparisons revealed that younger adults demonstrated lower peak handrail forces during slope descent compared to level walking (p<0.001), though performance in older adults did not differ significantly between slope versus level conditions (p=0.704). Significant main effects of age on peak handrail forces were not observed (p=0.896), nor were significant interactions between slope and rail height (p=0.494) or age and rail height (p=0.101). The highest handrail forces observed during level walking and slope descent were 51.3%BW and 46.9%BW, respectively Level Slope Handrail height and walking surface incline condition young adults older adults p<0.001 for slope p=0.005 for age*slope p=0.013 for rail height p>0.101 for age, rail height*slope and rail height*age Figure 6.5: Peak handrail forces, normalized to %BW. Bars represent mean values for each population and condition; error bars represent 1 standard deviation.

116 Discussion We have characterized the effect of handrail height and age on trunk and shoulder kinematics following perturbation-evoked grasping reactions during level-ground walking and slope descent. In general, increased handrail height was associated with greater trunk stability, indicated by reduced peak pitch displacements and velocities in the forward direction, and reduced peak roll velocities and displacements toward the rail. However, increased handrail height also resulted in greater peak shoulder elevation angles during balance recovery, which surpass the active range of motion of many with osteoarthritis in the shoulder [164]. Compared to level-ground gait, slope descent was generally associated with reduced forward trunk pitch displacement and velocity, and reduced handrail forces. Relative to younger adults, older adults generally demonstrated reduced roll velocity toward the rail in both slope conditions, and reduced pitch displacement and velocity on level ground only. Peak resultant handrail forces generally decreased (slightly) as handrail height increased, and were slightly lower in older adults than in younger adults Influence of handrail height on trunk and shoulder kinematics Similar to the Study 2 findings (Chapter 4) during perturbations of upright stance [165], increasing rail height mostly resulted in significant reductions to roll and pitch displacements and velocities in the present study. The exception was roll displacement away from the rail, which increased as handrail height increased. However, the magnitudes of roll displacement away from the rail were low (less than half than observed for roll-toward, and generally under 10 o ), and were not accompanied by significant effects of rail height on peak roll velocity away from the rail. When considered with the marginal decrease in peak handrail forces with increasing handrail height, this suggests that stabilizing moment advantage for higher handrails predicted by the inverted pendulum model (i.e., increased lever arm between the handrail and the person s ankles or the floor allows for greater stabilizing torques at a given applied handrail force [6]) may still be relevant during ongoing gait. While increased handrail height helped to enhance trunk stability in the present study, without the need for increased handrail forces, it also resulted in higher shoulder elevation angles during balance recovery. As demonstrated through comparisons to established, active ranges of motion of individuals with severe osteoarthritis in the shoulder (see Figure 6.2 and [164]), increasing the Elevation_to_Recover angle particularly if coupled with an awkward posture, such as shoulder horizontal extension (i.e., negative Plane_to_Recover angle) is likely to be problematic for persons

117 103 with shoulder pathologies. Indeed, while those with limited range of motion in the shoulder may compensate by relying more heavily on elbow flexion during volitional reaching, this option may not be present in reach-to-grasp balance reactions, where the person cannot choose their grip location on the rail, and must account for the dynamic movement of the trunk. Consequently, the apparent stability advantage with increased handrail height should be interpreted with caution. While higher rails may be advantageous for trunk stability in healthy adults, individuals with limited range of motion in the shoulder may benefit from lower rails that are less likely to result in high shoulder elevation angles for balance recovery. Further studies should investigate the influence of handrail height on balance recovery in populations with conditions that limit range of motion in the arms Influence of aging on balance recovery Contrary to expectation and to established age-related declines in trunk control during gait [166], older adults in the present study generally demonstrated comparable or reduced trunk displacements and velocities than younger adults during balance recovery. Further, current results did not provide evidence that older adults were more reliant on the handrail for balance recovery. In fact, the mean peak resultant forces that they applied to the handrail were slightly below those applied by young adults. We also note that gait speed did not differ between groups or conditions, with the exception of older adults walking slightly slower during slope descent. Possible differences in recovery strategy between young and older adults may help to explain why older adults demonstrated a more upright trunk, leading to potentially better control, in the present study. For example, while we did not characterize stepping behavior in this study, the role of stepping in arresting trunk and COM movement is well-established [92]. Further, previous studies have observed older adults to initiate stepping with lower perturbation thresholds compared to young adults [30, 29], as well as with reduced pelvis excursions and velocities relative to their base of support following waist-pull perturbations [30]. Older adults have also been previously observed to be more likely than young adults to execute multiple steps following forward and backward perturbations of upright stance [27]. Extending these prior findings to the present study, trunk behavior could plausibly differ between age groups if older adults executed more steps than younger adults, even with handrail use. Beyond possible age-related differences in stepping behavior, our findings of reduced trunk roll and pitch angular kinematics are consistent with other work on age-related differences in trunk control during balance-challenging activities. In particular, Gill and colleagues investigated trunk sway during quiet stance and gait in younger (15 to 25 years), middle-aged (45 to 55 years), and older adults (65

118 104 to 75 years) [167]. They noted that while older adults demonstrated greater sway during quiet stance than younger adults, they reduced trunk pitch sway when walking five steps with their eyes closed, and during stair gait [167]. Older adults also demonstrated lower roll and pitch velocity than younger adults during all gait activities [167]. Authors hypothesized that older adults may have been more apprehensive in this study, which could lead to increased trunk stiffness and reduced sway. Alternatively, the reduced sway during gait activities may have related to stiffness in joints or the possibility of hip flexion contractures. Indeed, Kerrigan and colleagues observed greater pelvic sagittal motion during gait in young adults compared to older adults, independent of gait speed [168] Influence of walking surface incline In this study, the most notable effect of slope was on reductions to peak forward trunk displacement and velocity during slope descent, compared to level gait (Figure 6.4). The most likely reason for this stems from how the effective anterior edge of the base of support in the feet is reduced when the feet are inclined downward. This would, presumably, reduce the allowable anterior trunk excursion and velocity before a stepping response is initiated Comparison to findings during upright-stance perturbations In general, the effects of handrail height on trunk displacements and velocities were similar between perturbations evoked during gait in the present study, and those observed in Study 2 (Chapter 4): increased rail heights were generally accompanied by decreases to roll range and velocity toward the rail, and forward pitch displacement and velocity. This is encouraging, as it suggests that effects identified through testing during upright stance perturbations are also likely present in gait contexts, even with the increased variability in balance recovery strategy due to factors such as stepping. For roll range and velocity, the magnitudes observed in young adults during gait (for rail heights of 34, 38 and 42 inches) were generally comparable to those observed during forward falling following upright stance perturbations (roll range averages within 2.5 o for each rail height; roll velocity-towardrail averages within 3 o /s for the rail at 34 and 38 inches, though reduced during gait by roughly 8 o /s for the rail at 42 inches). In the case of peak forward pitch displacement and velocity, the magnitudes during gait were well under those observed for forward falling during upright stance perturbations, though closer to those observed for backward falling. For pitch displacement, mean magnitudes were slightly reduced during gait compared to backward falling, by ~1.9 o to 4.6 o.

119 105 Conversely, mean magnitudes for peak pitch velocity were slightly increased during gait relative to backward falling, by 10.2 o /s to 13.5 o /s. In the present study, peak resultant handrail forces were much lower than those observed for reachto-grasp reactions following upright stance perturbations in Study 1 (Chapter 3) approximately 18% BW in this study for younger adults, versus 27% BW for forward falling and 52% BW for backward falling [135]. Part of the discrepancy may be attributed to the lower perturbation magnitudes in the present study (3.75 m/s 2, versus an average of around 4.1m/s 2 and 3.97m/s 2 for forward and backward falling respectively in Study 1). Lower perturbation magnitudes would be expected to result in reduced handrail forces, due to the reduced momentum that would need to be transferred to the rail to regain stability. The lower filter cut-off frequencies in the present study (to attenuate noise due to vibration in the handrail) may have also contributed to reduced peak handrail forces. However, a more plausible explanation for the difference likely lies in participants ability to execute multiple steps after balance perturbations in this study. This would have allowed participants to increase their base of support at the feet, thereby increasing the capacity to control high trunk and COM displacements and velocities [169] with less reliance on the handrail. Further research is needed to understand how handrail forces vary based on whether or not participants can step, and whether these forces depend on the conditions in which participants are stepping Limitations Limitations of the present study should be acknowledged. First, participants walked a fixed lateral distance (60cm) away from the handrail leading into perturbation onset. For comparison, the average preferred lateral position among participants in prior work evaluating the effect of handrail height on maximum volitional force generation was approximately 33cm [7], suggesting that participants would likely have stood closer to the rail if given the opportunity to select their lateral distance away. Given previous comparisons of reach-to-grasp reactions in stair descent conditions (where lateral distances of 32cm and 61cm were tested) [3], we would expect a few differences in findings had participants stood at a self-selected distance away from the rail. First, standing closer to the rail would likely decrease time to handrail contact (in line with previous findings during reach-to-grasp reactions [3]), thereby reducing the time for the trunk and COM to accelerate before the handrail is used to arrest the person s movement. This would likely result in reductions to observed trunk angular displacements and velocities. Second, postures used during grasping would differ with lateral distance from the rail. This would likely affect shoulder elevation angles leading into handrail contact

120 106 and throughout recovery (given previously-observed differences in the angle with which the hand approached the rail based on initial stance position [3]). The magnitude of resultant handrail forces during recovery would likely differ as well, in light of previously-reported reductions to lateral handrail force when standing closer to the rail [3]. Further research is needed to understand possible interaction effects between lateral distance from the rail and handrail height during balance recovery. Second, our assessment of balance control was based on metrics related to trunk angular displacement and velocity only. More comprehensive exploration of other balance recovery metrics, included quantitative characterization of stepping behavior, would help to strengthen any assessment of the effect of handrail height on balance recovery during ongoing gait. Third, this study incorporated only one perturbation design, with expected perturbations (i.e., participants were told that perturbations would occur). This may have enabled participants to preplan their responses to the destabilizations, which would not be possible in situations where the perturbations were truly unexpected. Given prior observation that individuals adapt their gait patterns when anticipating destabilizing situations (e.g., slippery floors, unlocked rollers) [170, 171], we can reasonably expect that participants adopted more conservative behaviour in this study. Further, different perturbation designs would be expected to reveal different balance recovery strategies, as evidenced by past studies [131]. Fourth, the shoulder range of motion data provided in Figure 6.4 were derived from a review of range of motion in persons with shoulder osteoarthritis before total arthroplasties [164]. This does not represent the breadth of conditions that limit upper-limb range of motion, nor did participants in this study adopt identical postures to those used during the controlled, active range of motion assessments in [164] (varied plane of elevation angles observed in this study, versus only forward elevation measured in [164]). Consequently, our comparison of shoulder elevation angles with existing data should be interpreted with some caution. Nevertheless, this study provides evidence that shoulder elevation angles during balance recovery can increase with rail height, and that the balance control advantages with increased rail height are unlikely to be appropriate for all users. Finally, our sample included healthy younger and older adults only, due to the challenging nature of the protocol. We expect that a more frail older adult population would use the handrail differently for balance control and recovery, which could impact our assessment of upper body control. Conversely, in individuals with other pathologies that have been shown to limit range of motion in

121 107 the shoulder (e.g., stroke [161]), we may well see different reach-to-grasp strategies for balance recovery that differ from those observed in our sample. Further research is needed to understand the impact of handrail height on balance recovery in these populations. 6.5 Conclusions We have characterized the effect of handrail height and age on trunk and shoulder kinematics following perturbation-evoked grasping reactions during level-ground walking and slope descent. Increased handrail height was generally associated with reduced trunk angular displacements and velocities, indicating greater ability to remain upright. However, these stability gains with increasing rail height may be offset by the increased peak shoulder elevation angles, which surpassed the active range of motion of many with osteoarthritis in the shoulder [164]. Slope descent mostly resulted in reduced forward trunk pitch displacement and velocity, and reduced handrail forces. Relative to younger adults, older adults generally demonstrated comparable or reduced trunk angular displacements and velocities. This is likely indicative of reduced thresholds for stepping in older adults, which affected trunk kinematics. Taken together, the results from this study demonstrate benefits to trunk control with increasing handrail height during both level walking and slope descent, but provide an important caveat that these benefits may not extend to those with limited range of motion in the shoulder.

122 Summary of dissertation Chapter 7 Thesis discussion The overall objective of this thesis was to investigate the effect of handrail height on recovery from balance perturbations. Effects of aging were evaluated, and testing conditions using perturbations during upright stance, level-ground walking, and slope descent were considered. Starting position (i.e., not contacting the rail before perturbation onset, versus contacting the rail with one or two hands) was also examined. Studies 1 and 2 (Chapters 3 and 4) both demonstrated that increasing handrail height allowed participating young adults to withstand higher perturbation magnitudes without stepping or falling and improve COM and trunk control, while reducing the peak forces and impulse that they applied to the rail. This suggests that higher handrails (within the range tested 34, 38 and 42 inches) can enhance recovery from perturbations of upright stance. Further, balance recovery was improved with the hand contacting the rail proactively (compared to reach-to-grasp reactions), as evidenced by the significant increase in maximum withstood perturbation in the proactive rail contact conditions. Study 3 (Chapter 5) demonstrated that, following perturbations during slope descent, handrail height showed statistically-significant effects on the movement time of compensatory reaching in younger and older adults, and was not compromised by high handrails that were initially further from participants hands. Data-based regression models suggest that movement time in older adults may be minimized with the rail at 64% of individual height (approximating elbow height [118]), within the tested range of 41%h to 71%h though this optimum should be interpreted with caution given the high variability in observed movement times. In both populations, peak hand speeds in the upward, lateral and downward directions varied significantly with handrail height, while vertical handrail overshoot decreased as handrail height increased. While older adults demonstrated slower deltoid onset latencies (by ~26ms, on average), they took less time to achieve peak hand velocity in the upward and lateral direction than did young adults. This likely helped older adults to exhibit comparable movement times to young adults, despite exhibiting reduced peak lateral and downward hand speeds. Consistent with observations during volitional reach-to-grasp movements, older adults exhibited longer peak-to-contact times than younger adults, which may have helped them to increase trajectory accuracy leading into handrail contact.

123 109 Finally, Study 4 (Chapter 6) revealed significant effects of handrail height on trunk control following perturbations during level walking and slope descent, with trunk control generally improving as handrail height increased. This is consistent with findings of Studies 1 and 2. Surprisingly, older adults generally demonstrated reduced forward trunk pitch displacements and speeds, and reduced roll speeds toward the rail. This may stem from reduced thresholds in older adults to execute reactive steps [30], which can help to arrest the forward movement of the trunk [26]. Peak resultant handrail forces were generally lower than those observed following the upright stance perturbations in Study 1 (Chapter 3). Increasing handrail height generally resulted in greater peak thoracohumeral elevation angles to recover balance, particularly among older adults during slope descent. 7.2 Applicability of models of whole-body balance recovery and compensatory grasping This thesis incorporated two key models for describing whole-body balance recovery and reach-tograsp movements: (1) the inverted pendulum model of balance control; and (2) the ballistic-andcorrective phase model of reach-to-grasp movements Inverted pendulum model The findings of Studies 1, 2 and 4 provide strong support for the applicability of the inverted pendulum model [91, 6] to balance recovery via handrail grasping and the influence of handrail height. In particular, the reductions to maximum withstood perturbation and improvements to trunk and COM control with increasing handrail height, combined with either comparable or reduced concomitant peak handrail forces and handrail impulse, suggest that the longer lever arm afforded by higher handrails (relative to the ground) provided a stability advantage. Notably, the general increases in trunk and COM stability with increased rail height were consistent across falling in multiple directions (forward, backward), and in different contexts (upright stance perturbations, level-ground walking, and slope descent). While this may have been predictable for forward falling, the relationship was less expected for backward falling, where the inverted pendulum model may be a less appropriate representation due to the observed collapse of participant knees (per Figure 4.1e and as outlined in the literature review) [6]. As handrail contact time did not vary significantly with rail height in that study, this difference cannot be attributed to increased time lags in the reach-to-grasp movement that would have affected the momentum of the COM before balance recovery. However, an alternate possibility not explored in my thesis, but

124 110 worth considering, is posture. Based on visual inspection of participant videos, it appears that higher handrails (across the different falling directions and conditions) could generally be grasped with the elbow bent, whereas lower rails were often grasped with the elbow fully extended. Grasping with the elbow extended would have the effect of lengthening the moment arm between the point of application of handrail force (i.e., the hand) and the shoulder, which must then stiffen to transfer the person s momentum to the handrail and stabilize the COM. Lengthening this moment arm (between the shoulder and the hand) would thus increase the shoulder torques necessary for this transfer of momentum to occur quickly and to slow COM movement. This may help to explain the increased handrail impulse with lower handrails during backward falling over the first 1500ms after handrail contact (well after the platform had stopped moving) and overall reductions to COM stability. As highlighted in the discussion of Study 2, the verification of the inverted pendulum model only applies within the range of handrail heights tested. If we consider increasing handrail heights beyond the tested range, the stability gains afforded by the longer lever arm are likely offset by compromised force generation ability during overhead reaching [97], and (eventually) by not being able to reach the handrail at all. In compensatory contexts, two published studies report compromises to balance control during overhead reaching during hand-in-place reactions [98, 99]. The precise point at which increasing handrail height ceases to be advantageous for stability in healthy adults is beyond the scope of this thesis, but may be valuable to explore in future work Ballistic-and-corrective phase model of reach-to-grasp movements In Study 3, the two-phase model of reach-to-grasp movements (involving the initial, high-velocity ballistic phase, and the subsequent, reduced-velocity corrective phase [80]) was helpful for identifying an age-related change in the control of compensatory reaching. Specifically, older adults achieved peak hand velocities more quickly than young adults, but also took longer to contact the rail once peak hand velocities were achieved effectively prolonging the corrective phase. This may have helped older adults to reach more accurately, and may help to account for why movement times between the two age groups did not differ. Theoretical support for the value of prolonging the corrective phase in older adults is provided by age-related declines in upper-limb proprioceptive acuity [154, 172], coupled with delays in integrating visual feedback [156]. One aspect of this model not addressed in Study 3 is the actual velocity profile of the hand leading into handrail contact. A number of studies of both volitional and compensatory reaching report

125 111 approximate bell-shaped velocity profiles (e.g., [66, 173, 174]), which demonstrate a general slowing of the hand leading into contact with the object being grasped. The results from Study 3 are not entirely consistent with these profiles. Instead, many participants (particularly young participants) appeared to increase vertical hand velocity in the downward direction leading into handrail contact, as shown in Chapter 5, Figure 5.2. As such, while the idea of a reduced-velocity, corrective phase of compensatory reaching is plausible and supported by some evidence (e.g., [77]), detailed kinematics of the reach trajectory leading into handrail contact should be explored in greater depth to more substantively demonstrate (or refute) its applicability. For example, testing on handrail cross-sections with complex geometry that require more precise modulation of the fingers for grasp completion may better reveal a possible relationship between the duration of the corrective phase and the need for increased trajectory and prehension control during compensatory reaching. 7.3 Impact of testing paradigms The studies in this thesis involved perturbations during both upright stance and ongoing gait. As demonstrated by comparisons of handrail forces and trunk behaviour between Studies 1 and 2 (Chapters 3 and 4, involving upright stance) and Study 4 (Chapter 6, involving gait), the different protocols likely impacted the findings. In particular, testing during ongoing gait resulted in trunk pitch behavior among young adults that was close in magnitude to that observed for backward falling during the perturbations of upright stance in Study 2 (where the trunk was generally more upright). Trunk roll behavior following perturbations while walking was comparable to that observed for forward falling, though demonstrated a lower range and velocity toward the rail than that observed for backward falling. More notably, peak resultant handrail forces in the ongoing gait context (~18% BW for younger adults) were much lower than those observed for reach-to-grasp reactions following the upright stance perturbations in Study 1 (~27%BW for forward falling and 52%BW for backward falling). Further, significant effects of age were not found from ongoing gait testing. This differs from prior research on maximum volitional force generation on stairway handrails, in which the force magnitudes that young adults applied were roughly twice as high as those applied by older adults [7, 8]. One possible reason for this convergence in peak force generation between age groups in the ongoing gait condition may be due to the additional confound of stepping, which often provided participants with means beyond the handrail to recover balance. Indeed, the reduction in applied

126 112 handrail forces in young adults during gait, compared to upright stance, suggests that participants were not as reliant on the handrail for balance recovery in the ongoing gait protocol. Despite the variability in recovery strategies in the ongoing gait protocol, and the potential for confounding due to stepping, the effects of handrail height on metrics of interest were comparable across testing conditions. Notably, performance in trunk control metrics generally improved as handrail height increased, while the concomitant handrail forces decreased, following forward and backward perturbations of upright stance as well as perturbations during gait. This similarity in effects of rail height across conditions provides some support for the generalizability of our findings across different modes of balance loss. It may also support the ecological reasonableness of the upright stance perturbations (with constrained stepping) as a method for evaluating the effect of handrail height on balance recovery. This is worth highlighting for many reasons. First, while many falls occur while walking [10], data collection and analysis with upright stance perturbations is far simpler it can occur with a smaller platform and reduced capture volume, and does not require a safety harness with tracking capabilities. Second, shorter handrails can be used for upright stance perturbations, thereby increasing the stiffness and reducing the mass. In Studies 1 and 2, this simplified the process of designing a stiff collection apparatus, which increased the natural frequency of the rail and reduced the need for low filter cut-off frequencies (< 20Hz), and improved data quality from the load cells. Accordingly, while upright stance perturbations with constrained stepping do not comprehensively represent all types of balance loss, they appear to allow us to differentiate between handrail heights with a less complex setup than that needed for perturbations during ongoing gait, with concomitant increases to data quality. Finally, comparisons between compensatory handrail forces recorded in this thesis (largely Study 1) can be made with (a) the compensatory handrail forces reported from Maki and colleagues 1998 work on reach-to-grasp reactions during perturbation-evoked stairway falls [3], and (b) the maximum volitional handrail forces reported in Maki and Fernie s 1983 evaluation of handrail parameters [6]. Generalizations are difficult, likely due to differences in handrail force application strategy between the volitional versus compensatory experimental paradigms. The unconstrained COM movement in the compensatory studies also likely impacted findings, particularly for handrail forces applied in the anterior and posterior directions. For anterior forces applied to the handrail in Study 1 (Chapter 3), participants applied roughly 20% of their body weight to the handrail during forward falling with reach-to-grasp

127 113 and hand-in-place reactions, with negligible effects of handrail height. This is consistent with the maximum volitional anteriorly-directed forces observed in Maki and Fernie s work (20%BW) [6], though below the mean anterior handrail forces during reach-to-grasp reactions reported in the 1998 study, in the highest perturbation condition (28%BW, though only 12%BW and 21% BW for the low and medium perturbation conditions respectively). The higher peak compensatory handrail forces observed in the stairway study may stem from a greater COM momentum component being transferred to the rail, due to participants being forced to overstep a trick stair and thus move in the forward and downward direction. Supporting evidence for the COM momentum hypothesis stems from the consistent increase in applied handrail forces with perturbation magnitude [3]. Alternatively, recorded forces in the 1998 study may be have been disproportionately impacted by 25% of the sample (1 of 4 participants) consisting of a weightlifter [3]. For posterior forces applied to the handrail in Study 1, participants applied peak forces of roughly 40% BW to the rail during backward falling involving reach-to-grasp reactions roughly double the maximum volitional handrail forces that young adults applied to the rail in the backward direction in Maki and Fernie s work (23% BW) [6]. Compensatory posterior forces were negligible (<10N) in the 1998 stairway perturbation study [3], likely due to the falling direction being consistently forward. The increased posterior compensatory forces in Study 1 are likely due to the increased potential for COM momentum to be transferred to the handrail in the compensatory context. Unlike posterior forces, participating young adults generally applied higher volitional forces to the rail in [6] (40% BW) than those observed in either Study 1 (17% BW for reach-tograsp reactions; 10% BW for hand-in-place reactions) or in the 1998 stairway perturbation study (11%BW on average) [3]. Balance recovery strategy is likely the most substantial contributor to this discrepancy. Specifically, the highest downward handrail forces typically occurred right at hand contact, which was often followed by pulling movements in the other direction. Accordingly, high downward forces were typically only applied briefly (a few ms see Figure 3.4a in Chapter 3), and were often sensitive to filtering. In the maximum volitional force paradigm, conversely, participants were asked to push on the handrail in the downward direction as hard as possible, though were also instructed to apply the forces slowly [6].

128 114 In summary, compensatory handrail forces in the anterior direction (applied with one hand while falling forward) sometimes exceeded maximum volitional handrail forces. In the posterior direction, compensatory handrail forces applied during reach-to-grasp reactions while falling backward were nearly double those of the maximum volitional forces applied to the handrail in the backward direction. This is likely due to the presence of a COM momentum component in the compensatory context; the fact that participants were not restricted in their posture and did not depend solely on arm musculature during compensatory grasping; and that the peak compensatory forces were not sustained over the multiple seconds that the maximum volitional handrail forces were applied for. Conversely, participants were observed to apply much lower average compensatory forces to the handrail in the downward direction than they were likely capable of (as evidenced by downward compensatory forces in both Study 1 and the 1998 stairway falls paradigm being less than half of those observed in the maximum volitional force paradigm (< 17% BW for compensatory; 40%BW for volitional) [3, 6]. This likely stemmed from participants not needing to apply high downward forces to the handrail to recover balance, and from the sensitivity of the downward force peak to filtering. However, higher downward compensatory forces may well occur in other contexts, particularly among users who may be unable to generate high torques in their ankles, knees and hips that would be needed for balance control and recovery with less reliance on a handhold. 7.4 Implications for handrail design While the objective of this thesis was to explore the effect of handrail height on balance recovery, the findings have a number of practical implications that others may benefit from when installing supportive rails to prevent falls in the home. Below, I summarize a few of these implications for handrail design, with emphasis on handrail height and structural strength Handrail height Stability gains were observed with increased rail height, for those without limited range of motion in the shoulder The results from Studies 1, 2 and 4 collectively point to stability gains with increased handrail height, within the range of handrail heights tested. Further, Study 3 demonstrated that higher handrails did not result in increased movement times, as would have been expected had participants adopted trajectories that more closely resembled straight lines (which would favour lower rails, near wrist height). Accordingly, individuals installing supportive handrails in their homes may benefit from

129 115 higher rails or similar devices (e.g. grab bars to assist with transfers in and out of the bathtub). This may be especially important for older adults with declines to strength due to muscle loss [108]. The major qualifier is that this suggestion is unlikely to apply to individuals with limited range of motion in the shoulder due to conditions such as stroke or osteorarthritis [144, 164]. As evidenced by the increased thoracohumeral elevation angles with rail height during balance recovery in Study 4, the stability gains may be muted among those who cannot reach high handrails without experiencing pain. For these populations, handrail designs that encourage proactive contact (and thus reduce the overall handrail force needed for balance recovery, as discussed below) may be especially important Handrail designs that encourage proactive contact are important Study 1 provides clear evidence of the balance recovery advantages when users are contacting the handrail before perturbations, including significant increases to maximum withstood perturbation and, in the case of one-handed hand-in-place reactions, concomitant decreases to peak handrail forces. Presumably, hand-in-place reactions allowed participants to arrest the movement of their COM more quickly, without the time lag from reach-to-grasp reactions that allows the COM to gain momentum after perturbation onset (and thereby increases the magnitude of handrail forces needed to counteract the greater momentum, in the absence of stepping). The stability gains with hand-inplace reactions are likely to be especially important for individuals with reduced upper-limb strength (due to conditions such as diabetes [127] or sarcopenia [175]), who may be unable to apply the higher handrail forces needed to arrest falls. For this reason, handrail designs should encourage proactive contact. Beyond important design considerations such as resembling handrails, and having cross-sectional designs that are comfortable to slide the hand along, handrails should be installed at heights that can be comfortably reached. For stairway handrails, participants in the early maximum volitional force studies reported that handrail heights of 36 maximized comfort of the rail heights tested [7]. However, a range above and below 36 presumably exists such that the installed rail is still reasonably-comfortable for proactive contact, and the specific boundaries of this range across varied users may be worth exploring in future work.

130 For non-compulsory supportive handrails in the home, individuals may benefit from rails installed around 56% to 66% of their height, corresponding roughly to mid-forearm- to elbow-height In this thesis, compulsory handrails are those required by local building codes (e.g., on most staircases), while non-compulsory handrails are those installed in addition to those that are required (e.g., in corridors or near bathtubs), generally to provide further stable support for balance and walking. I handle these two separately, as compulsory handrails are generally subject to local building code requirements for handrail height, which may not permit a sufficient range to install handrails at heights recommended in this section. I begin by considering non-compulsory handrails. Each of Studies 1, 2 and 4 demonstrated improved balance recovery with increased handrail height, up to a maximum of 42 inches/106.7 cm (Studies 1 and 2) and 44 inches/111.7 cm (Study 4). The range tested in Studies 1 and 2 was not sufficient to reveal a plateau in participant performance with increased rail height, while 44-inch rails showed marginal improvements (compared to 42-inch rails) in most measures of trunk stability. For the sample population of Studies 3 and 4, 44 inches/111.7 cm corresponds approximately to 66% of mean participant height (168 cm for older adults; 170 cm for younger adults). 66% of individual height roughly corresponds to slightly above elbow-height [118]. Accordingly, individuals who are installing non-compulsory supportive handrails in the home may consider installation heights around their elbow height. (Handrails required by local building codes are generally subject to the height restrictions stated in these codes, which I address in the next sub-section). The proposed installation height above assumes that: 1) The rail is being installed primarily for the use of one person, which justifies customizing to individual anthropometry. If the rail is intended to be shared among individuals with a greater height disparity, compromise will likely be necessary. 2) Rails at 66% of individual height can still be contacted comfortably, so as to avoid discouraging proactive contact. Should the user in question be unable to contact handrails at 66% of their height comfortably, the rail should be installed lower. When considering the results from Studies 1, 2 and 4, handrails at 38 inches high (96.5 cm) still provided significant stability advantages over rails at 34 inches high. 38

131 117 inches is approximately 56% of the height of the sample population of Studies 1, 2 and 4 (and roughly 57% of the height of the older adult participants in Study 4). 56% of individual height corresponds roughly to mid-forearm-height (adapted from [118]). This will still provide improved stability over lower rails (e.g., those installed at wrist-height), with reduced likelihood of requiring high shoulder elevation angles or elbow flexion to contact proactively For compulsory handrails subject to existing building code constraints in Ontario, balance recovery will likely be enhanced by installing above the minimum-permitted height in the current Ontario Building Code (86.5 cm) Studies 1, 2 and 4 further demonstrated that the poorest balance recovery performances (in terms of maximum withstood perturbation, COM control, trunk control, and physical demands of grasping) occurred with lower handrails. In Studies 1 and 2, the worst balance recovery performances were observed with handrails at 34 inches (86.5 cm) the current lower boundary of the Ontario Building Code [100]. While this thesis did not explore balance recovery on stairs, we can reasonably expect that the balance recovery improvements observed with increased handrail height may be present on stairs as well particularly during stair descent conditions, where the handrail height is effectively lower for anterior reaching. It is worth highlighting that Maki and Fernie s work on the effect of stairway handrail height on maximum volitional force and moment generation found that, based solely on kinetic considerations, rails between 36 and 38 inches (inclusively) were considered acceptable for younger adults, while rails between 36 and 40 inches (inclusively) were considered acceptable for older adults [6]. This resulted in a collective acceptable range of 36 to 38 inches, from which 36 inches was recommended based on heuristic data of handrail comfort. Within the constraints of the Ontario Building Code (86.5 to 96.5 cm [100], or 34 to 38 inches), the findings from this thesis support the acceptable range as preferred over 34-inch handrails. In particular, installing toward the upper boundary (38 inches, or 96.5 cm) will likely be advantageous for balance recovery for adults without upper-limb range of motion limitations, though further research is needed to verify that this height does not discourage proactive contact. As discussed further in the Limitations section, I have avoided recommending specific handrail heights for populations, and for implementation in building codes. While local jurisdictions could certainly use findings from this thesis in guiding policy decisions, they must also consider other

132 118 handrail users, such as children. However, the findings from this thesis can still be helpful for persons who are installing supportive rails at home, for the purpose of enhancing balance recovery Handrail structural strength A further contribution of my thesis research is the quantification of forces that individuals apply to handrails in the tested balance recovery contexts. My results have implications for setting structural strength requirements for handrails and similar devices, such as grab bars and removable grab rails Handrail designs should resist forces in the upward direction, particularly where the installation height is low Study 1 revealed that participants indeed apply high pulling forces to handrails during balance recovery, particularly during backward falling with lower (34 ) handrails. To my knowledge, forces of the magnitudes that we observed (24%BW on average, for reach-to-grasp reactions during backward falling with 34 rails) have never been previously reported. Given these findings, handrails in the community should be designed to resist upward forces. This is especially important in contexts where backward falls are likely, and where the handrail is installed at lower heights The observed peak handrail forces exceeded existing structural strength standards for grab bars and removable grab rails Study 1 also demonstrated that participating young adults applied peak handrail forces that exceeded the ISO standard for structural strength of grab bars and removable grab rails (e.g., handholds mounted to bathroom walls via suction cups) [121]. In the case of vertical forces, however, this only occurred when the handrail was at 34 and 38, but not at 42. Individuals installing these products in their homes may wish to consider the reduction in vertical forces with increased handrail height, particularly if their selected products are not designed to resist high vertical shear forces between the handhold and the wall. With this said, Study 1 results should be interpreted with caution as only peak handrail forces are reported. However, other mechanical factors are likely to play a role in the structural integrity or failure of a handhold, including the energy absorption capability of the combined handhold, wall, and mounting fixture. Other considerations beyond the scope of this thesis, such as the duration of high force application during a fall, and the frequency of repetition (i.e., how often the rail is used for balance recovery), must also be considered.

133 119 Decisions on handhold strength requirements must also consider handhold forces from a greater range of balance-challenging activities in which the handhold would be used, such as supporting sitto-stand transfers. Despite these limitations, data on the relationship between handrail height and applied forces and on the loading directions in which high handrail forces are observed during balance recovery remain important contributions that may help to inform handrail design and installation decisions in the community, and help to inform the development of future standards. 7.5 Limitations and future directions While the findings in this thesis provide important information on the influence of handrail height on balance recovery in younger and older adults, many limitations remain. Some of these limitations are discussed below, along with possible future directions where relevant The sample population consisted of healthy younger and older adults only This thesis focused on balance recovery in younger and older adults; other cohorts of handrail users were not tested. Groups who were not considered, but whose perspectives would be needed in any larger-scale implementation of our findings to populations, include: 1) Children: The absence of children in this thesis is most relevant to Studies 1 and 2, which both support 42 handrails as providing stability advantages over 34 and 38 -high handrails. These stability advantages are unlikely to apply to many children, particularly those who must reach overhead to grasp or who cannot grasp 42 handrails. 2) Adults with neurological or musculoskeletal conditions, or who did not self-report normal or corrected-to-normal vision: The studies in this thesis included only healthy individuals, who were able to complete demanding balance recovery protocols without high perceived exertion (based on Borg assessments [149]) or risk of injury. The effect of handrail height on balance recovery is likely to differ in those with conditions that constrain shoulder elevation, or elbow flexion or extension (e.g., stroke, rotator cuff injuries, arthritis). In particular, the balance control advantages with higher (42 ) handrails in Studies 1 and 2, and the movement time speed advantages with rails at elbow height in Study 3, may not apply to individuals who cannot comfortably achieve the high thoracohumeral elevation angles observed in Study 4 to reach handrails at those heights. Given the real and perceived importance of handrails for safety

134 120 and accessibility to many in these populations [55, 176, 177, 178], they should be considered in future evaluations of the effect of handrail design on balance control and recovery. 3) Wheelchair users: While this thesis focuses on balance recovery while standing or walking, any implementation of findings to level or sloped surfaces in public spaces must consider wheelchair users as well. In a ramp accessibility study of adults who used wheelchairs in Ottawa, 9 of 11 participants reported using handrails when navigating ramps in the winter [179]. During wheelchair ramp ascent and descent in packed-snow with freezing-rain conditions, two-railing propulsion was preferred as a strategy for navigating steep ramps, as it allowed the less-active participants to control their speed and extract their wheelchair when stuck [179]. Handrails in that study were installed at a relatively-high height (1m) [179], though the small sample and absence of children limits the confidence with which 1m-high handrails can be recommended for this (rather-diverse) population The experimental environments were limited to level ground and slope descent The experiments in this thesis covered a limited set of environments: level ground and slope descent, in controlled laboratory conditions. The effect of handrail height on balance recovery in other contexts where handrails are common, such as stairs, may differ. Specific to stairs, differences that could plausibly affect individual balance recovery with respect to handrail height include: 1) During stair descent, the handrail height is effectively lower if the person is reaching forward, thereby lessening the stabilizing moment advantage compared to handrails installed at the same height on level ground. Based on a study of the influence of stairway pitch and handrail height on maximum volitional force generation (with stair descent conditions simulated), Maki and colleagues noted that the estimated acceptable range of handrail heights increased with stairway pitch (i.e., higher rails were preferable on steeper stairs for enhancing volitional force and moment generation) [8]. 2) Body mechanics during balance loss and recovery can differ for stair descent versus level ground [180]. Observations of large forward and backward destabilizations (simulated oversteps and slips) during stair descent have indicated a posterior grip location relative to the person s COM [180], due to the downward and anterior movement of the body following balance loss. This is similar to some balance recovery strategies observed during gait, (even after handrail grasping),

135 121 as supported by the negative thoracohumeral plane of elevation angles observed during balance recovery in Study 4 (Chapter 6). In contrast, participants grasped the handrail anterior to the body during both forward and backward falling in Studies 1 and 2 in this thesis (see recovery strategies in Figure 4.1, in Chapter 4). The highest volitional anterior push forces that young women could exert on a handle have been reported to increase as the angle of shoulder flexion from the horizontal increases [181] (i.e., participants applied higher pushing forces to handles in front of them, compared to beside them and, presumably, behind them). This suggests a possible grasping force advantage for the level-ground recovery strategies involving anterior grasping with constrained stepping, compared to stairway falls requiring posterior grasping. The difference in posture during balance recovery on stairs could similarly influence the relationship between handrail height and a person s ability to apply high stabilizing handrail forces. We have not yet established how the combination of reduced effective handrail height during stair descent and altered body mechanics during stair falls would influence the effect of handrail height on balance recovery on stairs versus level ground or gradual slope descent. Accordingly, the results in this thesis apply to level ground and 8 o slopes only (i.e., the conditions tested in this thesis), though can be combined with other studies (e.g. Maki and Fernie s prior stairway handrail assessments [6]) to be considered in other environments. The effect of handrail height on balance recovery on stairs merits investigation in future studies Only one handrail cross-sectional shape was tested The experiments described in this thesis were conducted with one handrail cross-sectional shape only (round), using a narrow range of diameters (38mm diameter for Studies 1 and 2; 44mm diameter for Studies 3 and 4). While this geometry is recommended in a number of accessibility standards (e.g., CSA [129], ADA [130]) for allowing high volitional force generation, it does not represent the full breadth of handrail cross-sections that are present or permitted in the community [100, 101]. The effects of handrail height and age on compensatory grasping that were observed in this thesis could differ with handrails with more complex cross-sectional geometry. In particular: 1) The observed effects of handrail height on whole-body kinematics of balance recovery (maximum withstood perturbation; detailed COM and trunk metrics) and the concomitant applied handrail forces could differ with handrail cross-sections that do not easily permit high forces to be applied to the handrail along axes of interest. Of note, Study 1 revealed that

136 122 participants applied high forces to the handrail in the upward direction during backward falling nearly 25%BW, on average, for one-handed grasping when the handrail was low (34 ). Accordingly, handrail cross-sectional geometry that does not allow for the power grips needed to apply high pulling forces [102] (e.g., 2x6 s, 4x4 s, or large handrails with difficult-to-access (or non-existent) finger purchases) could reduce a person s ability to recover from backward balance loss via handrail grasping. It follows that the consequences to recovery from backward falling with lower handrails are likely to be amplified with complex handrail cross-sectional geometry. 2) The observed age-related delays from time of peak hand velocities to handrail contact could be amplified with complex handrail cross-sectional geometry that requires precise modulation of the fingers. If the observed delays indeed represent a prolonging of the corrective phase that has been used to characterize volitional reaching, then presumably more complex cross-sectional designs could increase the duration of the corrective phase. Further, age-related proprioceptive declines in the upper-limb observed in volitional activities (particularly the elbow and fingers) [172, 182] could increase the time needed for accurate handrail grasping in older adults. Further research on the relationship between handrail cross-sectional geometry and balance recovery is clearly warranted, particularly in older adults The perturbations used in this study were expected and repeated This thesis did not involve deceptions; instead, all participants were told that perturbations would occur, and the instructions to participants drew explicit attention to the handrail (i.e., participants were instructed to reach to grasp the handrail as quickly as possible when they felt a perturbation). Further, perturbations were repeated. While the protocol for Studies 1 and 2 included perturbations of multiple directions and magnitudes interspersed among the analyzed trials, the protocol for Studies 3 and 4 included one perturbation magnitude, direction and waveform only. Consequently, participants may have pre-planned aspects of their responses to perturbations, as well as adapted their stance, gait and recovery strategy to the perturbation protocol. This thesis did not characterize adaptation or pre-planning; however, the tendency of individuals to alter their gait in anticipation of a destabilization, and to adapt to repeated platform perturbations, is well-established [183, 23, 170]. Given established evidence of adaptation, it is likely that results could differ if we only used first-trial analyses with individuals naive to perturbation platforms, particularly if deceptions were involved. The most probable impact would be on the speed and accuracy findings in Study 3, as unexpected

137 Unsuccessful grasping likelihood over all 4 trials per rail height (%) 123 perturbations may well lead to behavior that more closely approximates a startle -like response proposed in other literature [42, 64]. Further research is needed to understand detailed kinematic differences in behavior between unexpected versus expected and repeated perturbations, and the impact of these differences on studies of reactive grasping Ease of grasp completion was not evaluated While this thesis explored the speed and accuracy of the reaching component of the reach-to-grasp reaction, as well as trunk and COM balance control and the concomitant forces applied to the handrail, the success of grasp completion at initial handrail contact (i.e., whether a person could grasp the rail without colliding, undershooting, or making other errors) was not evaluated. This should be characterized in a comprehensive evaluation of how handrail height affects the speed, accuracy and effectiveness of balance recovery via handrail grasping. Accordingly, the accuracy findings from Study 3 (Chapter 5) should be interpreted with caution, as they only focused on the reach trajectory and did not explore grasp completion or corresponding errors. We have started to characterize the effect of handrail height on the likelihood of unsuccessful grasping for the dataset collected for Studies 3 and 4 (during ongoing gait), as well as a breakdown of which errors are prevalent at different rail heights. However, analysis is incomplete. Initial results suggest that handrail height may affect the likelihood of unsuccessful grasping (see Figure 7.1 below), and a complete analysis of our grasp-coded data will be a priority in the future young adults older adults Handrail height (inches) Figure 7.1: Relationship between handrail height and unsuccessful grasping likelihood during slope descent. All four collected trials per participant are included. Adapted from [184].

138 Perturbations during the ongoing gait studies occurred at different stages of the gait cycle In Studies 3 and 4, perturbations were initiated when a trigger force plate was loaded by 100N in the downward direction. In some cases, this occurred when a participant contacted the trigger plate with their toe; in other cases, it occurred during heel contact. This difference in timing of perturbation initiation serves as a source of variability in data from the ongoing gait studies The safety harness likely limited participant movement, particularly for Studies 3 and 4 All participants wore a safety harness attached to the laboratory ceiling, to avoid actual falls. While this was necessary for safety, it may also have limited participant movement during experiments. This is unlikely to have substantially affected findings in Studies 1 and 2, since the harness was loose enough to allow the hips to come within ~30cm of the ground thereby allowing a considerable drop leading into handrail grasping. However, the harness permitted a smaller drop in Studies 3 and 4 (such that participant knees would not contact the ground). This may have affected our findings in two major ways: 1) The reduced ability to drop in Studies 3 and 4 may have resulted in less extreme balance recovery reactions. Supporting evidence for this is provided by comparing time to handrail contact in Studies 2 and 3 (~350 ms and 530 ms respectively, for young adults even though the perturbation magnitudes in Study 3 were slightly higher). Peak handrail forces were similarly lower in Study 4 compared to Study 1, for young adults. The instructions ( reach to grasp the handrail as quickly as possible when you feel a perturbation ) were identical for both studies. Part of this difference may be attributed to participants standing further away from the rail (in the lateral direction) in Study 3, as well as to the general ongoing gait paradigm and ability of participants to step which may have delayed reactive reaching. However, the possible reduced need to use the handrail for balance recovery (on account of the harness) cannot be ruled out as having delayed metrics such as handrail contact time. 2) Trunk angular displacements and peak velocities may have been reduced due to the harness in Study 4. Younger adults in Study 4 demonstrated different trunk behavior than those in Study 2. The harness may well have reduced trunk movement. However, it is worth noting that peak trunk pitch angles in Study 4, for example, were maximized with lower

139 125 handrails even though this is the context where the impact of the harness would expected to be the greatest. However, the differences in trunk performance across handrail heights may well be amplified in other contexts, where no harness is present to restrict trunk movement. On this basis, further research may be helpful for exploring the effect of handrail height on balance recovery, without a harness to confound motion. Lining the floor with crash pads would allow such studies to be done safely, at least for younger adults.

140 126 References [1] Public Health Agency of Canada, "Seniors' Falls in Canada: Second Report," Public Health Agency of Canada, Ottawa, [2] S. K. Verma, J. L. Willetts, H. L. Corns, H. R. Marucci-Wellman, D. A. Lombardi and T. K. Courtney, "Falls and Fall-Related Injuries among Community-Dwelling Adults in the United States," PLoS ONE, vol. 11, no. 3, [3] B. E. Maki, S. D. Perry and W. E. McIlroy, "Efficacy of handrails in preventing stairway falls: a new experimental approach," Safety Science, vol. 28, no. 3, pp , [4] B. E. Maki and W. E. McIlroy, "The Role of Limb Movements in Maintaining Upright Stance: The "Change-in-Support" Strategy," Physical Therapy, vol. 77, pp , [5] F. B. Horak and L. M. Nashner, "Central Programming of Postural Movements: Adaptation to Altered Support-Surface Configurations," Journal of Neurophysiology, vol. 55, no. 6, pp , [6] B. E. Maki and G. R. Fernie, "Biomechanical Assessment of Handrail Parameters with Special Consideration to the Needs of Elderly Users," Toronto, [7] B. E. Maki, S. A. Bartlett and G. R. Fernie, "Influence of Stairway Handrail Height on Ability to Generate Stabilizing Forces and Moments," Human Factors, vol. 26, no. 6, pp , [8] B. E. Maki, S. A. Bartlett and G. R. Fernie, "Effect of Stairway Pitch on Optimal Handrail Height," Human Factors, vol. 27, no. 3, pp , [9] B. E. Maki, "Influence of handrail shape, size and surface texture on the ability of young and elderly users to generate stabilizing forces and moments," [10] S. N. Robinovitch, F. Feldman, Y. Yang, R. Schonnop, P. M. Leung, T. Sarraf, J. Sims-Gould and M. Loughin, "Video capture of the circumstances of falls in elderly people residing in long-term care: an observational study," The Lancet, vol. 381, no. 9860, pp , [11] D. A. Kallman, C. C. Plato and J. D. Tobin, "The role of muscle loss in the age-related decline of grip strength: Cross-sectional and longitudinal perspectives," Journal of Gerontology: Medical Sciences, vol. 45, no. 3, pp. M82-88, [12] B. H. Goodpaster, S. W. Park, T. B. Harris, S. B. Kritchevsky, M. Nevitt, A. V. Schwartz, E. M. Simonsick, F. A. Tylavsky, M. Visser and A. B. Newman, "The loss of skeletal muscle strength, mass, and quality of aging in older adults: The health, aging and body composition study," Journal of Gerontology: Medical Sciences, vol. 61A, no. 10, pp , 2006.

141 127 [13] R. Schonnop, Y. Yang, F. Feldman, E. Robinson, M. Loughin and S. N. Robinovitch, "Prevalence of and factors associated with head impact during falls in older adults in longterm care," Canadian Medical Association Journal, vol. 185, no. 17, pp. E803-E810, [14] S. Quant, B. E. Maki, M. C. Verrier and W. E. McIlroy, "Passive and active lower-limb movements delay upper-limb balance reactions," NeuroReport, vol. 12, no. 13, pp , [15] E. C. King, S. M. McKay, T. A. Lee, C. Y. Scovil, A. L. Peters and B. E. Maki, "Gaze behavior of older adults in responding to unexpected loss of balance while walking in an unfamiliar environment: A pilot study," Journal of Optometry, vol. 2, pp , [16] E. C. King, T. A. Lee, S. M. McKay, C. Y. Scovil, A. L. Peters, J. Pratt and B. E. Maki, "Does the "eyes lead the hand" principle apply to reach-to-grasp movements evoked by unexpected balance perturbations?," Human Movement Science, vol. 30, pp , [17] Y.-C. Pai and J. Patton, "Center of mass velocity-position predictions for balance control," Journal of Biomechanics, vol. 30, no. 4, pp , [18] S. M. Henry, J. Fung and F. B. Horak, "EMG responses to maintain stance during multidirectional surface translations," Journal of Neurophysiology, vol. 80, pp , [19] B. E. Maki and G. Ostrovski, "Scaling of postural responses to transient and continuous perturbations," Gait and Posture, vol. 1, pp , [20] S. Park, F. B. Horak and A. D. Kuo, "Postural feedback responses scale with biomechanical constraints in human standing," Experimental Brain Research, vol. 154, pp , [21] W. E. McIlroy and B. E. Maki, "Task constraints on foot movement and the incidence of compensatory stepping following perturbation of upright stance," Brain Research, vol. 616, no. 1-2, pp , [22] W. E. McIlroy and B. E. Maki, "Influence of destabilization on the temporal characteristics of "volitional" stepping," Journal of Motor Behavior, vol. 28, no. 1, pp , [23] W. E. McIlroy and B. E. Maki, "Adaptive changes to compensatory stepping responses," Gait and Posture, vol. 3, no. 1, pp , [24] J. L. Zettel, W. E. McIlroy and B. E. Maki, "Can stabilizing features of rapid triggered stepping reactions be modulated to meet environmental constraints?," Experimental Brain Research, vol. 145, no. 3, pp , [25] J. L. Zettel, W. E. McIlroy and B. E. Maki, "Environmental constraints on foot trajectory reveal the capacity for modulation of anticipatory postural adjustments during rapid triggered stepping reactions," Experimental Brain Research, vol. 146, no. 1, pp , 2002.

142 128 [26] C. Y. Scovil, J. L. Zettel and B. E. Maki, "Stepping to recover balance in complex environments: Is online visual control of the foot motion necessary or sufficient?," Neuroscience Letters, vol. 445, no. 1, pp , [27] W. E. McIlroy and B. E. Maki, "Age-related Changes in Compensatory Stepping in Response to Unpredictable Perturbations," Journal of Gerontology, vol. 51A, no. 6, pp. M289-M296, [28] S. G. Brauer, M. Woollacott and A. Shumway-Cook, "The influence of a concurrent cognitive task on the compensatory stepping response to a perturbation in balance-impaired and healthy elders," Gait and Posture, vol. 15, no. 1, pp , [29] J. L. Jensen, L. A. Brown and M. H. Woollacott, "Compensatory stepping: The biomechanics of a preferred response among older adults," Experimental Aging Research, vol. 27, no. 4, pp , [30] M.-L. Mille, M. W. Rogers, K. Martinez, L. D. Hedman, M. E. Johnson, S. R. Lord and R. C. Fitzpatrick, "Thresholds for inducing protective stepping responses to external perturbations of human standing," Journal of Neurophysiology, vol. 90, no. 2, pp , [31] B. E. Maki, M. A. Edmonstone and W. E. McIlroy, "Age-related differences in laterally directed compensatory stepping behavior," The Journals of Gerontology: Series A, vol. 55, no. 5, pp. M270-M277, [32] M. Milosevic, K. M. Valter McConville and K. Masani, "Arm movement improves performane in clinical balance and mobility tests," Gait and Posture, vol. 33, pp , [33] M. Pijnappels, I. Kingma, D. Wezenberg, G. Reurink and J. H. Van Dieen, "Armed against falls: the contribution of arm movements to balance recovery after tripping," Experimental Brain Research, vol. 201, pp , [34] M. Kouzaki and K. Masani, "Reduced postural sway during quiet standing by light touch is due to finger tactile feedback but not mechanical support," Experimental Brain Research, vol. 188, pp , [35] J. H. J. Allum, M. G. Carpenter, F. Honegger, A. L. Adkin and B. R. Bloem, "Age-dependent variations in the directional sensitivity of balance corrections and compensatory arm movements in man," Journal of Physiology, vol. 542, no. 2, pp , [36] P. Corbeil, B. R. Bloem, M. van Meel and B. E. Maki, "Arm reactions evoked by the initial exposure to a small balance perturbation: A pilot study," Gait and Posture, vol. 37, pp , [37] W. E. McIlroy and B. E. Maki, "Early activation of arm muscles follows external perturbation of upright stance," Neuroscience Letters, vol. 184, pp , [38] D. S. Marigold, A. J. Bethune and A. E. Patla, "Role of the unperturbed limb and arms in the reactive recovery response to an unexpected slip during locomotion," Journal of

143 129 Neurophysiology, vol. 89, pp , [39] P. E. Roos, M. P. McGuigan, D. G. Kerwin and G. Trewartha, "The role of arm movements in early trip recovery in younger and older adults," Gait and Posture, vol. 27, pp , [40] Z. Merrill, A. J. Chambers and R. Cham, "Arm reactions in response to an unexpected slip - Impact of aging," Journal of Biomechanics, vol. 58, pp , [41] J. E. Misiaszek, "Early activation of arm and leg muscles following pulls to the waist during walking," Experimental Brain Research, vol. 151, pp , [42] O. P. Sanders, D. N. Savin, R. A. Creath and M. W. Rogers, "Protective balance and startle responses to sudden freefall in standing humans," Neuroscience Letters, vol. 586, pp. 8-12, [43] K. B. Cheng, Y.-C. Huang and S.-Y. Kuo, "Effect of arm swing on single-step balance recovery," Human Movement Science, vol. 38, pp , [44] K. B. Cheng, K.-M. Wang and S.-Y. Kuo, "Role of arm motion in feet-in-place balance recovery," Journal of Biomechanics, vol. 48, pp , [45] F. Feldman and S. N. Robinovitch, "Reducing hip fracture risk during sideways falls: Evidence in young adults of the protective effects of impact to the hands and stepping," Journal of Biomechanics, vol. 40, pp , [46] J. J. Jeka and J. R. Lackner, "Fingertip contact influences human postural control," Experimental Brain Research, vol. 79, no. 2, pp , [47] R. Dickstein, C. L. Shupert and F. B. Horak, "Fingertip touch improves postural stability in patients with peripheral neuropathy," Gait and Posture, vol. 14, pp , [48] R. Dickstein, R. J. Peterka and F. B. Horak, "Effects of light fingertip touch on postural responses in subjects with diabetic neuropathy," Journal of Neurology, Neurosurgery and Psychiatry, vol. 74, no. 5, pp , [49] A. R. Martinelli, D. B. Coelho, F. H. Magalhaes, A. F. Kohn and L. A. Teixeira, "Light touch modulates balance recovery following perturbation: from fast response to stance restabilization," Experimental Brain Research, vol. 233, pp , [50] J. Camernik, Z. Potocanac, L. Peternel and J. Babic, "Holding a handle for balance during continuous postural perturbations - Immediate and transitionary effects on whole body posture," Frontiers in Human Neuroscience, vol. 10, pp. 1-8, [51] J. E. Misiaszek, M. J. Stephens, J. F. Yang and K. G. Pearson, "Early corrective actions of the leg to perturbations at the torso during walking in humans," Experimental Brain Research, vol. 131, pp , 2000.

144 130 [52] A. P. Marsh, J. A. Katula, C. F. Pacchia, L. C. Johnson, K. L. Loury and W. J. Rejeski, "Effect of treadmill and overground walking on function and attitudes in older adults," Medicine and Science in Sports and Exercise, vol. 38, no. 6, pp , [53] S. M. Reid, A. C. Novak, B. Brouwer and P. A. Costigan, "Relationship between stair ambulation with and without a handrail and centre of pressure velocities during stair ascent and descent," Gait and Posture, vol. 34, pp , [54] K. A. Hamel and P. R. Cavanagh, "Stair performance in people aged 75 and older," Journal of the American Geriatric Society, vol. 52, pp , [55] Z. Lu, S. D. Rodiek, M. M. Shepley and M. Duffy, "Influences of physical environment on corridor walking among assisted living residents: Findings from focus group discussions," Journal of Applied Gerontology, vol. 30, no. 4, pp , [56] A. Shumway-Cook and M. H. Woollacott, Motor Control: Translating Research into Clinical Practice, Philadelphia: Wolters Kluwer, [57] G. M. Gartsman, T. S. Roddey and S. M. Hammerman, "Shoulder arthroplasty with or without resurfacing of the glenoid in patients who have osteoarthritis," Journal of Bone and Joint Surgery, vol. 82, no. 1, pp , [58] R. H. Cofield, J. Parvizi, P. J. Hoffmeyer, W. L. Lanzier, D. M. Ilstrup and C. M. Rowland, "Surgical repair of rotator cuff tears," Journal of Bone and Joint Surgery, Vols. 83-A, no. 1, pp , [59] C. M. Robinson, J. Howes, H. Murdoch, E. Will and C. Graham, "Functional outcome and risk of recurrent instability after primary traumatic anterior shoulder dislocation in young patients," Journal of Bone and Joint Surgery, Vols. 88-A, no. 11, pp , [60] T. Herman, N. Inbar-Borovsky, M. Brozgol, N. Giladi and J. M. Hausdorff, "The dynamic gait index in healthy older adults: The role of stair climbing, fear of falling and gender," Gait and Posture, vol. 29, no. 2, pp , [61] S. M. McKay, J. E. Fraser and B. E. Maki, "Effects of uni- and multimodal cueing on handrail grasping and associated gaze behavior in older adults," Accident Analysis and Prevention, vol. 59, pp , [62] C. Y. Scovil, P. Corbeil, T. A. Lee, S. M. McKay, A. L. Peters and B. E. Maki, "A novel handrail cueing system to prevent falls in older adults," Gerontechnology, vol. 6, no. 4, pp , [63] W. J. Choi, J. M. Wakeling and S. N. Robinovitch, "Kinematic analysis of video-captured falls experienced by older adults in long-term care," Journal of Biomechanics, vol. 48, pp , 2015.

145 131 [64] E. T. Hsiao and S. N. Robinovitch, "Common protective movements govern unexpected falls from standing height," Journal of Biomechanics, vol. 31, pp. 1-9, [65] D. L. Sturnieks, R. St George and S. R. Lord, "Balance disorders in the elderly," Clinical Neurophysiology, vol. 38, pp , [66] W. H. Gage, K. F. Zabjek, S. W. Hill and W. E. McIlroy, "Parallels in control of voluntary and perturbation-evoked reach-to-grasp movements: EMG and kinematics," Experimental Brain Research, vol. 181, pp , [67] R. Fitzpatrick and D. I. McCloskey, "Proprioceptive, visual and vestibular thresholds for the perception of sway in humans," Journal of Physiology, vol. 478, no. 1, pp , [68] S. D. Perry, W. E. McIlroy and B. E. Maki, "The role of plantar cutaneous mechanoreceptors in the control of compensatory stepping reactions evoked by unpredictable, multi-directional perturbation," Brain Research, vol. 877, pp , [69] V. Heyl and H.-W. Wahl, "Psychosocial adaptation to age-related vision loss: A six-year perspective," Journal of Visual Impairment and Blindness, vol. 95, no. 12, pp , [70] S. Iwasaki and T. Yamasoba, "Dizziness and Imbalance in the Elderly: Age-related Decline in the Vestibular System," Aging and Disease, vol. 6, no. 1, pp , [71] J. R. Wingert, C. Welder and P. Foo, "Age-related hip proprioception declines: Effects on postural sway and dynamic balance," Archives of Physical Medicine and Rehabilitation, vol. 95, no. 2, pp , [72] S. R. Lord, "Visual risk factors for falls in older people," Age and Ageing, vol. 35, no. S2, pp. ii42-ii45, [73] S. D. Perry, "Evaluation of age-related plantar-surface insensitivity and onset age of advanced insensitivity in older adults using vibratory and touch tests," Neuroscience Letters, vol. 392, pp , [74] K. C. Cheng, S. M. McKay, E. C. King and B. E. Maki, "Does aging impair the capacity to use stored visuospatial information or online visual control to guide reach-to-grasp reactions evoked by unpredictable balance perturbation?," Journal of Gerontology: Medical Sciences, pp. 1-8, [75] T. B. Weaver, L. E. Hamilton and C. D. Tokuno, "Age-related changes in the control of perturbation-evoked and voluntary arm movements," Clinical Neurophysiology, vol. 123, pp , [76] T. B. Weaver and C. D. Tokuno, "The influence of handrail predictability on compensatory arm reactions in response to a loss of balance," Gait and Posture, vol. 38, pp , 2013.

146 132 [77] K. C. Cheng, S. M. McKay, E. C. King and B. E. Maki, "Reaching to recover balance in unpredictable circumstances: Is online visual control of the reach-to-grasp reaction necessary or sufficient?," Experimental Brain Research, vol. 218, pp , [78] K. C. Cheng, S. M. McKay, E. C. King, J. Y. Tung, T. A. Lee, C. Y. Scovil and B. E. Maki, "The moveable handhold: A new paradigm to study visual contributions to the control of balance-recovery reactions," Gait and Posture, vol. 29, no. 2, pp , [79] P. M. Fitts, "The information capacity of the human motor system in controlling the amplitude of movement," Journal of Experimental Psychology, vol. 47, pp , [80] M. Jeannerod, "The timing of natural prehension movements," Journal of Motor Behavior, vol. 16, no. 3, pp , [81] K. Y. Haaland and D. L. Harrington, "Hemispheric control of the initial and corrective components of aiming movements," Neuropsychologia, vol. 27, no. 7, pp , [82] M. Ghafouri, W. E. McIlroy and B. E. Maki, "Initiation of rapid reach and grasp balance reactions: is a pre-formed visuospatial map used in controlling the initial arm trajectory?," Experimental Brain Research, vol. 155, pp , [83] R. J. Bootsma, R. G. Marteniuk, C. L. MacKenzie and F. T. J. M. Zaal, "The speed-accuracy trade-off in manual prehension: effects of movement amplitude, object size and object width on kinematic characteristics," Experimental Brain Research, vol. 98, pp , [84] C. J. Ketcham, R. D. Seidler, A. W. A. Van Gemmert and G. E. Stelmach, "Age-related kinematic differences as influenced by task difficulty, target size, and movement amplitude," Journal of Gerontology: Psychological Sciences, vol. 57B, no. 1, pp , [85] C. G. Atkeson and J. M. Hollerbach, "Kinematic features of unrestrained vertical arm movements," The Journal of Neuroscience, vol. 5, no. 9, pp , [86] P. Morasso, "Spatial Control of Arm Movements," Experimental Brain Research, vol. 42, pp , [87] J. Pratt, A. L. Chasteen and R. A. Abrams, "Rapid aimed limb movements: Age differences and practice effects in component submovements," Psychology and Aging, vol. 9, no. 2, pp , [88] E. C. King, S. M. McKay, K. C. Cheng and B. E. Maki, "The use of peripheral vision to guide perturbation-evoked reach-to-grasp balance-recovery reactions," Experimental Brain Research, vol. 207, pp , [89] K. P. Westlake, B. P. Johnson, R. A. Creath, R. M. Neff and M. W. Rogers, "Influence of non-spatial working memory demands on reach-grasp responses to loss of balance: Effects of age and fall risk," Gait and Posture, vol. 45, pp , 2016.

147 133 [90] H. A. Bateni, A. Zecevic, W. E. McIlroy and B. E. Maki, "Resolving conflicts in task demands during balance recovery: does holding an object inhibit compensatory grasping?," Experimental Brain Research, vol. 157, no. 1, pp , [91] D. A. Winter, "Human balance and posture control during standing and walking," Gait and Posture, vol. 3, pp , [92] M. W. Rogers and M.-L. Mille, "Lateral Stability and Falls in Older People," Exercise and Sport Sciences Reviews, vol. 31, no. 4, pp , [93] W. E. McIlroy and B. E. Maki, "The control of lateral stability during rapid stepping reactions evoked by antero-posterior perturbation: Does anticipatory control play a role?," Gait and Posture, vol. 9, no. 3, pp , [94] M. Pijnappels, M. F. Bobbert and J. H. van Dieen, "How early reactions in the support limb contribute to balance recovery after tripping," Journal of Biomechanics, vol. 38, no. 3, pp , [95] M. S. Redfern, R. Cham, K. Gielo-Perczak, R. Gronqvist, M. Hirvonen, H. Lanshammar, M. Marpet, C. Y.-C. Pai and C. Powers, "Biomechanics of Slips," Ergonomics, vol. 44, no. 13, pp , [96] N. J. Seo, T. J. Armstrong and J. G. Young, "Effect of handle orientation, gloves, handle friction and elbow posture on maximum horizontal pull and push forces," Ergonomics, vol. 53, no. 1, pp , [97] D. B. Chaffin, R. O. Andres and A. Garg, "Volitional postures during maximal push-pull exertions in the sagittal plane," Human Factors, vol. 25, no. 5, pp , [98] T. A. Sarraf, D. S. Marigold and S. N. Robinovitch, "Maintaining standing balance by handrail grasping," Gait & Posture, vol. 39, pp , [99] J. Babic, T. Petric, L. Peternel and N. Sarabon, "Effects of supportive hand contact on reactive postural control during support perturbations," Gait and Posture, vol. 40, no. 3, pp , [100] Queen's Printer for Ontario, , "Ontario Regulation 332/12: Building Code," 1 January [Online]. Available: [Accessed 6 September 2017]. [101] International Code Council, "2015 International Building Code: Chapter 10 Means of Egress," International Code Council, [Online]. Available: Codes/2015%20IBC%20HTML/Chapter%2010.html. [Accessed 12 January 2017]. [102] D. M. Fothergill, D. W. Grieve and S. T. Pheasant, "The influence of some handle designs and handle height on the strength of the horizontal pulling action," Ergonomics, vol. 35, no. 2,

148 134 pp , [103] R. E. Hughes, M. E. Johnson, S. W. O'Driscoll and K.-N. An, "Age-related changes in normal isometric shoulder strength," The American Journal of Sports Medicine, vol. 27, no. 5, pp , [104] P. Mukhopadhyay, L. O'Sullivan and T. J. Gallwey, "Estimating upper limb discomfort level due to intermittent isometric pronation torque with various combinations of elbow angles, forearm rotation angles, force and frequency with upper arm at 90 abduction," International Journal of Industrial Ergonomics, vol. 37, no. 4, pp , [105] D. A. Skelton, C. A. Greig, J. M. Davies and A. Young, "Strength, power and related functional ability of healthy people aged years," Age and Ageing, vol. 23, pp , [106] I. Janssen, S. B. Heymsfield, Z. Wang and R. Ross, "Skeletal muscle mass and distribution in 468 men and women aged yr," Journal of Applied Physiology, vol. 89, no. 1, pp , [107] E. J. Metter, R. Conwit, J. Tobin and J. L. Fozard, "Age-associated loss of power and strength in the upper extremities in women and men," Journal of Gerontology: Biological Sciences, vol. 52A, no. 5, pp. B , [108] J. Lexell, "Human aging, muscle mass, and fiber type composition," The Journals of Gerontology: Series A, vol. 50A, pp , [109] J. Lexell, C. C. Taylor and M. Sjostrom, "What is the cause of the ageing atrophy? Total number, size and proportion of different fiber types studied in whole vastus lateralis muscle from 15- to 83-year-old men," Journal of the Neurological Sciences, vol. 84, no. 2-3, pp , [110] G. Fayet, A. Rouche, J.-Y. Hogrel, F. M. S. Tome and M. Fardeau, "Age-related morphological changes of the deltoid muscle from 50 to 79 years of age," Acta Neuropathologica, vol. 101, no. 4, pp , [111] M. E. Johnson, M.-L. Mille, K. M. Martinez, G. Crombie and M. W. Rogers, "Age-related changes in hip abductor and adductor joint torques," Archives of Physical Medicine and Rehabilitation, vol. 5, no. 4, pp , [112] D. C. Mackey and S. N. Robinovitch, "Mechanisms underlying age-related differences in ability to recover balance with the ankle strategy," Gait and Posture, vol. 23, no. 1, pp , [113] P. de Leva, "Adjustments to Zatsiorsky-Seluyanov's Segment Inertia Parameters," Journal of Biomechanics, vol. 29, no. 9, pp , 1996.

149 135 [114] J.-M. Billette and T. Janz, "Injuries in Canada: Insights from the Canadian Community Health Survey," Statistics Canada, [115] D. Vena, A. C. Novak, E. C. King, T. Dutta and G. R. Fernie, "The evaluation of vertical pole configuration and location on assisting the sit-to-stand movement in older adults with mobility limitation," Assistive Technology. [116] D. M. O'Meara and R. M. Smith, "Differences Between Grab Rail Position and Orientation During the Assisted Sit-to-Stand for Able-Bodied Older Adults," Journal of Applied Biomechanics, vol. 21, pp , [117] O. Shechtman and B. S. Sindhu, "American Society of Hand Therapists Grip Strength: Key Recommendations for Outcome Evaluation of Grip Strength," [Online]. Available: [Accessed 12 February 2016]. [118] D. A. Winter, Biomechanics and Motor Control of Human Movement, Hoboken, New Jersey: John Wiley and Sons, [119] U.S. Department of Health and Human Services, "Anthropometric Reference Data for Children and Adults: United States, ," Centers for Disease Control and Prevention, National Center for Health Statistics, Hyattsville, [120] British Medical Journal Publishing Group Ltd, "Correlation and Regression," British Medical Journal Publishing Group Ltd, [Online]. Available: [Accessed 15 May 2017]. [121] International Standards Organization, "ISO : Assistive products for personal hygiene that support users - Requirements and test methods," International Standards Organization, Geneva, Switzerland, [122] A. Mital and H. F. Faard, "Effects of sitting and standing, reach distance, and arm orientation on isokinetic pull strengths in the horizontal plane," International Journal of Industrial Ergonomics, vol. 6, pp , [123] S. Kumar, "Arm lift strength in work space," Applied Ergonomics, vol. 22, no. 5, pp , [124] C. M. Haslegrave, M. F. Tracy and E. N. Corlett, Force exertion in awkward working postures - strength capability while twisting or working overhead, Ergonomics, vol. 40, no. 12, pp , [125] H. Fredriksen, J. Hjelmborg, J. Mortensen, M. Mcgue, J. W. Vaupel and K. Christensen, "Age Trajectories of Grip Strength: Cross-Sectional and Longitudinal Data Among 8,342 Danes Aged 46 to 102," Annals of Epidemiology, vol. 16, no. 7, pp , 2006.

150 136 [126] T. Rantanen, K. Masaki, D. Foley, G. Izmirlian, L. White and J. M. Guralnik, "Grip strength changes over 27 years in Japanese-American men," Journal of Applied Physiology, vol. 85, no. 6, pp , [127] E. Cetinus, M. A. Buyukbese, M. Uzel, H. Ekerbicer and A. Karaoguz, "Hand grip strength in patients with type 2 diabetes mellitus," Diabetes Research and Clinical Practice, vol. 70, no. 3, pp , [128] A. Fraser, J. Vallow, A. Preston and R. G. Cooper, "Predicting 'normal' grip strength for rheumatoid arthritis patients," Rheumatology, vol. 38, pp , [129] Canadian Standards Association, "CSA B651-12: Accessible design for the built environment," Canadian Standards Association, Mississauga, Canada, [130] Department of Justice, United States of America, "2010 ADA Standards for Accessible Design," Department of Justice, United States of America, [131] L. A. Brown, J. L. Jensen, T. Korff and M. H. Woollacott, "The translating platform paradigm: perturbation displacement waveform alters the postural response," Gait and Posture, vol. 14, no. 3, pp , [132] M. E. Tinetti, M. Speechley and S. F. Ginter, "Risk factors for falls among elderly persons living in the community," New England Journal of Medicine, vol. 319, no. 26, pp , [133] M. J. Pavol and Y. C. Pai, "Feedforward adaptations are used to compensate for a potential loss of balance," Experimental Brain Research, vol. 145, pp , [134] S. N. Robinovitch, B. Heller, A. Lui and J. Cortez, "Effect of strength and speed of torque development on balance recovery with the ankle strategy," Journal of Neurophysiology, vol. 88, no. 2, pp , [135] V. Komisar, K. Nirmalanathan, E. C. King, B. E. Maki and A. C. Novak, "Use of handrails for balance and stability: Characterizing force profiles in younger adults," [136] V. Komisar, A. C. Novak and B. C. Haycock, "A novel method for synchronizing motion capture with other data sources for millisecond-level precision," Gait and Posture, vol. 51, pp , [137] W. Dempster, "Space Requirements of the Seated Operator," Wright-Patterson Air Force Base, [138] E. J. Havanan, "A Mathematical Model for the Human Body," Wright-Patterson Air Force Base, [139] W. J. Conover and R. L. Iman, "Rank transformations as a bridge between parametric and

151 137 non-parametric statistics," The American Statistician, vol. 35, no. 3, pp , [140] J.-S. Tan, J. J. Eng, S. N. Robinovitch and B. Warnick, "Wrist impact velocities are smaller in forward falls than backward falls from standing," Journal of Biomechanics, vol. 39, no. 10, pp , [141] L. A. Brown and J. S. Frank, "Postural compensations to the potential consequences of instability," Gait and Posture, vol. 6, pp , [142] A. L. Adkin, J. S. Frank, M. G. Carpenter and G. W. Peysar, "Fear of falling modifies anticipatory postural control," Experimental Brain Research, vol. 143, pp , [143] S. Tanaka, K. Hachisuka and H. Ogata, "Muscle strength of trunk flexion-extension in poststroke hemiplegic patients," American Journal of Physical Medicine and Rehabilitation, vol. 77, no. 4, pp , [144] R. W. Bohannon, "Adequacy of simple measures for characterizing impairment in upper limb strength following stroke," Perceptual and Motor Skills, vol. 99, pp , [145] T. Pozzo, P. J. Stapley and C. Papaxanthis, "Coordination between equilibirum and hand trajectories during whole body pointing movements," Experimental Brain Research, vol. 144, no. 3, pp , [146] A. N. Lay, C. J. Hass and R. J. Gregor, "The effects of sloped surfaces on locomotion: A kinematic and kinetic analysis," Journal of Biomechanics, vol. 39, pp , [147] V. Komisar, A. C. Novak, E. C. King, B. E. Maki, K. F. Zabjek and G. R. Fernie, "Influence of Balance Disturbance Method on the Timing and Accuracy of Reach-to-Grasp Balance Recovery Reactions During Level-Ground Walking," in International Society of Gait and Posture Reseach World Congress, Seville, [148] Department of Building and Housing, "Compliance Document for New Zealand Building Code - Clause D1: Access Routes - Second Edition," 10 October [Online]. Available: access-routes/asvm/d1-access-routes-amendment-5.pdf. [Accessed 9 September 2017]. [149] G. A. V. Borg, "Psychophysical bases of perceived exertion," Medicine and Science in Sports and Exercise, vol. 14, no. 5, pp , [150] M. G. Carpenter, J. S. Frank, C. P. Silcher and G. W. Peysar, "The influence of postural threat on the control of upright stance," Experimental Brain Research, vol. 138, pp , [151] L. A. Brown, M. A. Polych and J. B. Doan, "The effect of anxiety on the regulation of upright standing among younger and older adults," Gait and Posture, vol. 24, pp , 2006.

152 138 [152] B. E. Maki, P. J. Holliday and A. K. Topper, "Fear of Falling and Postural Performance in the Elderly," Journal of Gerontology, vol. 46, no. 4, pp. M123-M131, [153] T. W. Cleworth, R. Chua, J. T. Inglis and M. G. Carpenter, "Influence of virtual height exposure on postural reactions to support surface translations," Gait and Posture, vol. 47, pp , [154] D. E. Adamo, B. J. Martin and S. H. Brown, "Age-related differences in upper-limb proprioceptive acuity," Perceptual and Motor Skills, vol. 104, pp , [155] T. Kalisch, J.-C. Kattenstroth, R. Kowalewski, M. Tegenthoff and H. R. Dinse, "Age-related changes in the joint position sense of the human hand," Clinical Interventions in Aging, vol. 7, pp , [156] R. P. Di Fabio, C. Zampieri and J. F. Greany, "Aging and saccade-stepping interactions in humans," Neuroscience Letters, vol. 339, no. 3, pp , [157] D. A. E. Bolton, R. Patel, W. R. Staines and W. E. McIlroy, "Transient inhibition of primary motor cortex suppresses hand muscle responses during a reactive reach to grasp," Neuroscience Letters, vol. 504, pp , [158] M. T. Vogt, E. M. Simonsick, T. B. Harris, M. C. Nevitt, J. D. Kang, S. M. Rubin, S. B. Kritchevsky and A. B. Newman, "Neck and shoulder pain in 70- to 79-year-old men and women: Findings from the Health, Aging and Body Composition Study," The Spine Journal, vol. 3, pp , [159] J. G. Broeks, G. J. Lankhorst, K. Rumping and A. J. H. Prevo, "The long-term outcome of arm function after stroke: Results of a follow-up study," Disability and Rehabilitation, vol. 21, no. 8, pp , [160] M. Frankle, S. Siegal, D. Pupello, A. Saleem, M. Mighell and M. Vasey, "The reverse shoulder prosthesis for glenohumeral arthritis associated with severe rotator cuff deficiency: A minimum two-year follow-up study of sixty patients," The Journal of Bone and Joint Surgery, Vols. 87-A, no. 8, pp , [161] M. R. Mouawad, C. G. Doust, M. D. Max and P. A. McNulty, "Wii-based movement therapy to promote improved upper extremity function post-stroke: A pilot study," Journal of Rehabilitative Medicine, vol. 43, pp , [162] G. Wu, F. C. T. van der Helm, H. E. J. Veeger, M. Makhous, P. Van Roy, C. Anglin, J. Nagels, A. R. Karduna, K. McQuade, X. Wang, F. W. Werner and B. Buchholz, "ISB recommendation on definitions of joint coordinate systems of various joints for the reporting of human motion - Part II: shoulder, elbow, wrist and hand," Journal of Biomechanics, vol. 38, pp , [163] J. Aizawa, T. Masuda, T. Koyama, K. Makamaru, K. Isozaki, A. Okawa and S. Morita, "Three-dimensional motion of the upper extremity joints during various activities of daily

153 139 living," Journal of Biomechanics, vol. 43, pp , [164] D. Bryant, R. Litchfield, M. Sandow, G. M. Gartsman, G. Guyatt and A. Kirkley, "A comparison of pain, strength, range of motion, and functional outcomes after hemiarthroplasty and total shoulder arthroplasty in patients with osteoarthritis of the shoulder," The Journal of Bone and Joint Surgery, Vols. 87-A, no. 9, pp , [165] V. Komisar, K. Nirmalanathan and A. C. Novak, "Influence of handrail height and fall direction on center of mass control and the physical demands of reach-to-grasp balance recovery reactions," Gait and Posture, [166] M. H. Woollacott and P.-F. Tang, "Balance control during walking in the older adult: Research and its implications," Physical Therapy, vol. 77, no. 6, pp , [167] J. Gill, J. H. Allum, M. G. Carpenter, M. Held-Siolkowska, A. L. Adkin, F. Honegger and K. Pierchala, "Trunk sway measures of postural stability during clincal balance tests: Effects of age," Journal of Gerontology: Biological Sciences and Medical Sciences, vol. 56, no. 7, pp. M , [168] D. C. Kerrigan, M. K. Todd, U. Della Croce, L. A. Lipsitz and J. J. Collins, "Biomechanical gait alterations independent of speed in the healthy elderly: Evidence for specific limiting impairments," Archives of Physical Medicine and Rehabilitation, vol. 79, pp , [169] Y.-C. Pai and J. Patton, "Center of mass velocity-position predictions for balance control," Journal of Biomechanics, vol. 30, no. 4, pp , [170] R. Cham and M. S. Redfern, "Changes in gait when anticipating slippery floors," Gait and Posture, vol. 15, no. 2, pp , [171] D. S. Marigold and A. E. Patla, "Strategies for dynamic stability during locomotion on a slippery surface: effects of prior experience and knowledge," Journal of Neurophysiology, vol. 88, no. 1, pp , [172] W. R. Ferrell, A. Crighton and R. D. Sturrock, "Age-dependent changes in position sense in human proximal interphalangeal joints," NeuroReport, vol. 3, no. 3, pp , [173] N. Yang, M. Zhang, C. Huang and D. Jin, "Synergic analysis of upper-limb target-reaching movements," Journal of Biomechanics, vol. 35, pp , [174] J. Wang and G. E. Stelmach, "Coordination among body segments during reach-to-grasp action involving the trunk," Experimental Brain Research, vol. 123, no. 3, pp , [175] E. M. Castillo, D. Goodman-Gruen, D. Kritz-Silverstein, D. J. Morton, D. L. Wingard and E. Garrett-Connor, "Sarcopenia in elderly men and women: The Rancho Bernardo Study," American Journal of Preventive Medicine, vol. 25, no. 3, pp , 2003.

154 140 [176] W. C. Mann, D. Hurren, M. Tomita, M. Bengali and E. Steinfeld, "Environmental problems in homes of elders with disabilities," The Occupational Therapy Journal of Research, vol. 14, no. 3, pp , [177] D. E. Rosenberg, D. L. Huang, S. D. Simonovich and B. Belza, "Outdoor built environment barriers and facilitators to activity among midlife and older adults with mobility disabilities," The Gerontologist, vol. 53, no. 2, pp , [178] G. Chen, C. Patten, D. H. Kothari and F. E. Zajac, "Gait deviations associated with poststroke hemipariesis: Improvement during treadmill walking using weight support, speed, support stiffness, and handrail hold," Gait and Posture, vol. 22, no. 1, pp , [179] E. D. Lemaire, P. A. O'Neill, M. M. Desrosiers and D. G. Robertson, "Wheelchair ramp navigation in snow and ice-grit conditions," Archives of Physical Medicine and Rehabilitation, vol. 91, pp , [180] P. Gosine, "PhD Committee Update - July 2017," Toronto, [181] A. Y. Chow and C. R. Dickerson, "Shoulder strength of females while sitting and standing as a function of hand location and force direction," Applied Ergonomics, vol. 40, no. 3, pp , [182] D. E. Adamo, N. B. Alexander and S. H. Brown, "The Influence of Age and Physical Activity on Upper Limb Proprioceptive Ability," Journal of Aging and Physical Activity, vol. 17, no. 3, pp , [183] A. J. Chambers and R. Cham, "Slip-related muscle activation patterns in the stance leg during walking," Gait and Posture, vol. 25, no. 4, pp , [184] M. Jabakhanji, V. Komisar, A. C. Novak and G. R. Fernie, "Classification of grasping reactions with varying handrail height during level walking and ramp descent," in University of Toronto Undergraduate Engineering Research Day, Toronto, [185] V. Komisar, A. C. Novak, E. C. King, B. E. Maki, K. F. Zabjek and G. R. Fernie, "Effect of handrail height on the speed and accuracy of reach-to-grasp balance recovery reactions during ramp descent: A pilot study," in Proceedings of the 19th Triennial Congress of the International Ergonomics Association, Melbourne, Australia, [186] C. M. O'Connor, S. K. Thorpe, M. J. O'Malley and C. J. Vaughan, "Automatic detection of gait events using kinematic data," Gait and Posture, vol. 25, pp , [187] O. Tirosh and W. A. Sparrow, "Identifying heel contact and toe-off using forceplate thresholds with a range of digital-filter cutoff frequencies," Journal of Applied Biomechanics, vol. 19, no. 2, pp , 2003.

155 141 Preface Appendix A: Synchronization of Data Sources This appendix describes the method for synchronizing kinematic and kinetic data sources. It was published in 2017 in the journal Gait and Posture, with the title A novel method for synchronizing motion capture with other data sources for millisecond-level precision. The authors are myself, Dr Alison Novak, and Dr Bruce Haycock. My role in this paper was in Part B, in which we verified the synchronizing system in the context of balance recovery data from nine healthy, young participants. This included demonstrating the need for a method that enables synchronization of data sources for every trial due to the high variability in temporal offsets between trials. Dr Haycock developed the content for Part A, which describes the synchronization system and how it is implemented in the Challenging Environments Assessment Laboratory at Toronto Rehab-UHN. Dr Novak did the majority of the writing for the introduction, discussion and conclusion. Abstract Synchronization of multiple data collection systems is necessary for accurate temporal alignment of data, and is particularly important when considering rapid movements which occur in less than one second. This paper describes a novel method for synchronizing multiple data collection instruments including load cells and a motion capture system, using a common analog signal. An application of the synchronization method is demonstrated using biomechanical data collected during a rapid reach-to-grasp reaction, where data from motion capture and load cells are collected. Results are provided to validate and demonstrate the accuracy of the synchronization of motion capture with other data collection systems. During the reach-to-grasp trials, delays between the data collection systems ranged from 4ms to 235ms. The large range and variability in delay times between trials highlights the need for synchronization on a continual basis, rather than application of an average or constant value to correct for time delays between systems.

156 142 A-1 Introduction Synchronization of multiple data collection systems, such as kinematic motion capture and kinetic force plate data, is a typical requirement tasked to laboratories to produce accurate biomechanical analyses of human movement. Each laboratory will likely employ an internal procedure to synchronize the data systems, although this procedure is usually not detailed in publications. In many cases, an approximate temporal alignment of the systems is obtained by simply having the primary data collection software start and stop the motion capture recording process. Alternatively, analog signals can be synced with motion capture by collecting them directly within the proprietary motion capture software. Both approaches provide a reasonable means of synchronizing multiple signals to analyze many types of movements and tasks. However, due to differences in system response times or built-in system latencies, this can result in a misalignment of data which could impact interpretation of data when millisecond responses are being evaluated, such as during balance recovery reactions that involve rapid compensatory stepping or grasping. There are also numerous scenarios in which these approaches may not be feasible, such as when collecting signals that are not simple analog inputs, wireless Bluetooth sensors or USB connected equipment for example, or when the data acquisition is part of a larger program controlling hardware. Following large, unpredictable balance perturbations, young and older adults execute balance recovery reactions very rapidly, including reaching to grasp nearby handholds when they are present [74, 16, 185, 3, 77]. In response to perturbations on level ground, mean reaction time of the upper arm can be as quick as 143 ms [77]. Following a simulated stair fall, it has been shown that the time to handrail contact can range from 324 ms to 869 ms, dependent on perturbation magnitude and individual [3]. Given the rapid nature of the balance recovery responses, accurate evaluation of temporal aspects of movement may depend on precise synchronization between motion capture, force plates and other biomechanical measurement instruments when data collected from these systems are integrated. The purpose of this paper is to describe a novel method for synchronizing multiple data collection instruments including a motion capture system (Part A). Data collected from a study evaluating rapid balance recovery reaction will be used to illustrate the effect of the synchronizing method on biomechanical movement analyses (Part B).

157 143 A-2 Part A: Development of Synchronization Method A-2.1 System Implementation To synchronize the motion capture data with other biomechanical measurement instruments, an analog output signal is created from the main program responsible for data collection, and is subsequently sampled by the motion capture system. For our particular data collection setup, the main/primary data collection system uses QUARC (Quanser Inc., Canada) a real-time control software which integrates with and builds on MATLAB Simulink (The MathWorks, Inc., USA) to record data in a MATLAB file format. The motion capture software is Cortex (Motion Analysis Corp, California). In this particular instance QUARC is used to collect signals such as force plates, load cells, and encoders used to track the movement of our laboratory s perturbation platform, but the same methods presented in this paper would hold for MATLAB, LabVIEW, etc. While any number of continually-varying analog output signals could potentially be employed, for ease of use in this application a sawtooth wave is created via program run time modulus 1. A bias and scale are included to make use of a reasonable range of the analog data acquisition systems, mapping the [0,1] second time range to [- 9,9] Volts in this case. The analog signal is shown in Figure A-1. Figure A-1: Time reference signal: a sawtooth wave with added bias and scaling to make use of a reasonable range of the analog data acquisition systems. To yield an initial approximate temporal alignment of the systems, the primary data collection software starts and stops the motion capture recording process. This can be accomplished, for

158 144 example, by using the Software Development Kit (SDK) for Cortex (as was specific to our set-up), or UDP Ethernet messages for others. The analog output signal from the main collection system described above is connected to an analog input on the motion capture system. This requires a separate analog input device dedicated to the motion capture software, which can be entirely housed within the same computer as the main data collection software, or on a physically-separate machine. In this particular application, two computers were used for performance considerations; a Quanser QPID card was used for the primary PC analog output, with a 16-bit resolution, +/-10 V range, and1 μs response time. The signal was sampled by a National Instruments USB-6225 data acquisition system on the motion capture PC with a similar 16-bit resolution, +/-10 V range, and quoted 3 mv accuracy. A block diagram showing this arrangement is provided in Figure A-2. The analog input is sampled by the motion capture software along with a corresponding set of camera images that will form one frame of motion capture data. This analog signal gives the precise clock time (in fractions of a second) of the main collection program when the motion capture frame was collected. With the main collection program running at an update rate of 1 khz, the signal provides millisecond accuracy. With the previously mentioned scaling, the 1 ms time step corresponds to a voltage step of 18 mv, well above both the quoted accuracy of the analog input and the observed noise. This time in milliseconds (0-999) is combined with the above initial start time approximation in whole seconds to provide the actual timestamp for the motion capture system. Figure A-2: Schematic diagram of system architecture. A continuously-varying analog signal is created and output from the main data collection software, which collects analog and other data sources as well as triggering motion capture recording. This analog signal is sampled by the motion capture software, and allows data from the respective systems to be temporally-aligned.

159 145 A-2.2 Verification of synchronization method For testing purposes and real-time monitoring of time delays between motion capture data, the analog input collected through the motion capture system can be sent back to the primary data collection system through the SDK along with other frame information. Figure A-3a shows a fivesecond recording of this timing reference signal with the applied scaling and offset from above removed. A loopback reference from an analog input of the main collection system connected directly to the original outgoing signal is included for comparison, which was also used to confirm that the analog output was as expected with no time delays. The time on the horizontal axis has had the whole second value subtracted (i.e. modulus 1), which folds all data back into the range of 0 to 1 second. Since the motion capture system sampling frequency is reduced compared to the primary data collection system (250 Hz versus 1 khz in this particular example), the motion capture analog values remain constant for four frames on average when read back in the primary data collection system. This reduced update rate as well as the time delay present in the motion capture system can clearly be seen in Figure A-3. Figure A-3: Timing reference signal recording compared to a loopback reference from an analog input on the main collection system connected directly to the original outgoing signal, with the added scaling and bias removed. Time on the horizontal axis has had the whole second value subtracted (i.e. modulus 1), folding all data back into the range of 0 to 1 second. Figure 3a shows the original signal while Figure 3b shows the result after processing to remove any bias and repeated data points. While these results look quite promising, there are several minor issues common to analog signals, with some noise visible as well as the possibility of a voltage bias. When looking at a small section of analog signal data as illustrated in the inset of Figure A-3a, any voltage bias if present (which would shift points up or down) would be indistinguishable from the time delay to be identified (which

160 146 would shift points to the right). However, the known nature of the signal (i.e ms repeating in the presented example) provides additional information, allowing the average over each period (which must be ms here) to be used to identify any bias in the analog signal. Further, the noise can be effectively eliminated simply by rounding to the nearest time step as the variations due to analog noise and resolution are well below the output step size. Finally, the repeated frames of motion capture data due to the lower sample rate are dropped, using only the first instance of each frame, which is the time that frame of data was actually recorded. After this processing, the timing signal appears as shown in Figure A-3b, with an average time delay of approximately 7.5 milliseconds across this example trial. There is one significant assumption in the above technique: that the motion capture software is in fact collecting the analog time reference signal at the same instant in time that the motion capture images are taken. In order to validate this, an accurate independent measure of the motion capture system time delay is required. To accomplish this timing measurement, motion capture markers were mounted on a rotating bar. The bar was rotated using a servomotor with a built-in position encoder, driven and sampled using the Quanser QPID data acquisition card, to directly determine the angular position of the bar with essentially zero time delay. The current position as observed in motion capture could also be determined using the line segment connecting two markers on the bar. Since our motion capture system is a passive optical system with reflective markers, three markers positioned at uneven intervals were employed to minimize the likelihood of markers being swapped mid-trial. The lag between the actual position as measured with the encoder and the position as measured through motion capture gives a direct measure of the time delay. With the servomotor position driven in a sinusoid, the phase lag between the two angular position signals can be used to obtain the time delay averaged over a large number of data points. This time delay is then compared against the value obtained using the analog clock reference signal described above. This technique was repeated a number of times using a variety of sine waves. Two sample trials are illustrated in Figure A-4. In both cases the horizontal axis is wrapped back after each period, to fold all data into one full oscillation.

161 147 Figure A-4: Two sample trials of the time delay comparison between the position command angle, the angle as derived through motion capture and the angle measured by the encoder, consisting of (a) a 10 second recording of a 90 degrees at 0.5 Hz motion and (b) 4 second recording of 45 degrees at 1 Hz motion. The horizontal axis has been wrapped back after each period, to fold all data into one full oscillation. The insets present a detailed look at one short slice of time, to better show the nature of the various signals. The results from the full set of eight trials used are presented in Table A.1. The first item of note is that the delays between the primary data collection and motion capture systems are not constant. Taking place only a few minutes apart with an identical setup, the delay differs even across these trials and from the 7.5 ms value noted in the Figure A-3 example. While these values are all reasonably small, we have observed much larger fluctuations and delays (demonstrated in Part B), particularly when using larger numbers of cameras and markers, presumably due to the increased processing time required. Table A-1: Comparison of time delays derived using the analog reference signal or through comparing position signals Trial Time Delay (ms) Analog Reference Position Comparison Difference Average SD

162 148 Second, the time delay as determined directly through comparing position signals does not precisely match that obtained with the analog time reference. However, the difference between these two methods is relatively consistent, with a discrepancy of approximately 3.5 ms, meaning that the position signal obtained through motion capture is 3.5 ms older than the analog time reference would imply. As previously noted, the motion capture system in this case was operating at 250 Hz, with a 4 ms frame time. This implies that the analog inputs are indeed sampled during the same time period as the images from each camera used to create one frame of motion capture data. However the camera images would appear to be taken closer to the start of the frame while the analog inputs are sampled at the end of the frame. To account for this and obtain a better temporal match in general use, an additional half frame of delay is added to that indicated through the analog time reference. A-3 Part B: Application of synchronization method during balance recovery reactions To demonstrate the application of the synchronization method during a balance recovery experimental protocol and understand the trial-to-trial variability in delays between optical motion capture and other data recordings collected in parallel, we conducted a pilot test involving a total of 612 balance disturbance trials across nine young adults. A-3.1 A Methods Experimental environment All data were collected in the Challenging Environments Assessment Laboratory (CEAL) at Toronto Rehabilitation Institute University Health Network: a 5m x 5m laboratory secured to a six-degree-of-freedom robotic platform (Figure A-5). The robotic platform can deliver walkingsurface perturbations during level- and sloped-ground walking. The walking surface consisted of five large (120cm x 120cm) force plates (AMTI, Watertown, MA) covered with fiberboard panels. A handrail positioned along the walkway was instrumented with tri-axial load cells (MC3A-1000, AMTI, Watertown, MA) on the top of the two vertical posts to quantify the forces and moments that participants applied to the handrail during balance recovery. All participants wore a safety harness secured to the laboratory ceiling, which helped to protect them from actual falls, but interfered minimally with natural gait. Each participant provided informed consent prior to participation in the study.

163 149 Figure A-5: The Challenging Environment Assessments Laboratory. Left: Exterior view of the laboratory environment a 5x5m laboratory is mounted on a robotic motion base. Right: Interior view of the laboratory space. A Protocol Participants walked back and forth beside a nearby handrail with their arms at their sides on a level and an 8-degree inclined surface. The pilot study protocol has been previously described in detail [185]. During randomly-selected walks, participants were destabilized with sudden, posteriordirected platform translations (perturbations), which were initiated when participants stepped on a trigger force plate. Participants were then instructed to reach and grasp the handrail as quickly as possible following the perturbation (platform acceleration > 0.1 m/s 2 ). A trial was repeated if the participant forgot to reach for the handrail after the perturbation. A range of 67 to 73 complete trials were collected for each participant. The demands on the data collection systems were substantial: twelve motion capture cameras tracked a total of 76 passive motion capture markers at any given time (51 markers on the participant, as well as 25 markers on the handrail posts and laboratory floor to characterize the environment and allow for continuous dynamic calibration of the motion capture system), along with five force plates on the floor and two load cells on the handrail itself. We expected that this would comprise a relatively demanding environment for collecting data from multiple systems in

164 150 order to understand the effect of synchronization delays on interpretation of balance recovery reactions. To illustrate the effect of synchronization of the multiple systems on data analysis, handrail contact time defined by using a motion capture signal was compared to definition using an analog force signal from the load cells on the handrail posts. Vertical position of the hand, determined by a motion capture marker on the base of the third metacarpal of the participant s reaching hand, permits identification of handrail contact time using the kinematic signal. In this case, handrail contact time was defined as the time at which position of the marker stabilizes in the vertical direction following the perturbation, indicative of the hand terminating vertical movement following sudden contact with a rigid handrail. The signal derived from the handrail forces was used as a kinetic measure of handrail contact time, defined as the time at which the force magnitude in the lateral direction increased suddenly from baseline (extracted via visual inspection). Note that the handrail force signal considered in this application was based on the sum of both load cells (one on each handrail post) to quantify the total lateral force that a participant applied to the handrail at any given time. A-3.2 Results A High delays present between data from primary data collection versus motion capture systems We define delay as the extent to which data collection initiation of the motion capture system lags behind the primary data collection system (as described in Part A). This is the difference between applying the described synchronization technique to accurately align the two recordings in time versus assuming a common start time obtained through remotely triggering the motion capture recording process. As demonstrated by our results, the range and variability in delays between these two systems were surprisingly high: delays ranged from as low as 4 ms to as high as 235 ms (mean delay 38 ms; median delay 36 ms; inter-quartile range ms; standard deviation 27 ms). High delays (> 100 ms) were observed in ten of the 612 trials in this evaluation. Figure A-6 presents sample motion capture with the synchronization method applied versus unsynchronized data and handrail load cell data. This demonstrates kinematic and kinetic features of a reach-to-grasp reaction following a perturbation and the importance of temporal alignment in these datasets. As illustrated in Figure A-6, handrail contact time as defined using handrail forces

165 151 aligns temporally when synchronizing is applied to the motion capture data. Without synchronization of the systems, an apparent delay of nearly 200 ms occurs between when the hand appears to contact the handrail in motion capture and when the signal on the load cell following reach-to-grasp is observed to change substantially. Figure A-6: Sample vertical hand position (a), with and without synchronization technique applied, and lateral handrail forces (b) during a single trial in which the participant reaches to grasp a handrail following a balance perturbation. The time scales are with respect to platform perturbation onset, indicated by the dotted black line. The vertical dashed line indicates the properly aligned handrail contact time on both signals, while the dash-dot line indicates the apparent time delay between identified contact times without synchronization applied. A High trial-to-trial variability in delays between the primary data collection system and the motion capture system We also evaluated the trial-to-trial variability in delays between data from the primary collection system versus motion capture system, with results demonstrating very high trial-to-trial variability. A histogram showing the distribution of delays across the 612 trials is presented in Figure A-7.

166 152 Figure A-7: Histogram of motion capture recording time delays across 612 trials (bin size is 5 ms). A-4 Discussion This study presented a new method for synchronization between motion capture and kinetic data, and the application of this high-precision synchronization when evaluating rapid reach-to-grasp balance recovery reactions. Accurate integration of kinetic and kinematic data is necessary for biomechanical evaluation of human movement. High-precision synchronization is particularly important when considering rapid movements. Reach-to-grasp balance recovery reactions occur over very brief time scales, with previous studies reporting approximately 143 to 260 ms between an initial balance disturbance and reaction onset of key arm muscles (including anterior deltoid, middle deltoid and biceps [74]). Healthy adults have also demonstrated handrail contact times of approximately half a second while walking [16] or standing [3]. As shown by our current results, synchronization errors can result in