Comparison of Muscle Load, Comfort, Usability, and Time-On-Task Across 3 Wheelchair Handle Orientations and Ground Types: A Randomized Crossover Study

Introduction

Population aging is a major social shift of the 21st Century; in Taiwan, 17.5% of citizens are over age 65 years, officially classifying the nation as an “aging society” [1]. As mobility declines among older adults, the demand for wheelchairs, especially attendant-propelled models, continues to rise, with over half of users requiring caregiver assistance [2]. Taiwan faces a severe caregiver shortage, largely due to musculoskeletal disorders (MSDs) [3]. Approximately 88% of home caregivers report experiencing MSDs, most frequently in the shoulders (67.4%), lower back (64.2%), and wrists/hands (56.6%) [3], which is more than twice the prevalence among Danish caregivers (≈30%) [4]. Furthermore, wheelchair design directly influences attendant posture, comfort, and biomechanical load [5]. Traditionally, wheelchair design focuses on maximizing patient comfort and minimizing costs [6], with little attention given to reducing the burden on caregivers.

Previous research examined how handle orientations affect force direction in push/pull tasks [7] and upper-arm muscle activation during shopping cart pushing [8]. Pushing and pulling are major contributors to musculoskeletal issues, accounting for 50% to 75% of manual material handling tasks across industries [9]. NIOSH reports that up to 20% of hand overexertion injuries stem from these activities [10], highlighting the need to address muscular load in wheelchair use. While studied have explored factors like handle posture and force application angles [11], specific research on wheelchair handle design remains limited [12]. Moreover, most studies focused on straight-path pushing [13,14], overlooking the challenges wheelchair users face, such as ramps and uneven terrain [13,15]. Parallel handles have been reported to elicit the greatest upper-limb muscle activity on ramps [13]. Turning while pushing a cart is notably more challenging [16], but there is little research on muscle loading or usability during these tasks.

Unlike self-propelling wheelchair users, attendant pushers encounter distinct biomechanical demands, applying combined horizontal and vertical forces from behind the wheelchair that generate multi-directional loads among the shoulder, trunk, and lower limbs [6,17]. These forces demand greater proximal control and postural stabilization, engaging the erector spinae, anterior deltoid, and forearm stabilizers, including the extensor carpi radialis (ECR) and flexor carpi ulnaris (FCU) [17,18]. Handle orientation further modulates these loading patterns: horizontal handles (eg, lateral handles) increase shoulder flexion and trunk extension moments due to greater downward forces, whereas vertical handles heighten wrist and forearm stabilization demands to counter anterior–posterior sway [12,19]. Examining these effects in wheelchair attendant pushers can help identify optimal handle configurations that reduce musculoskeletal load and improve propulsion efficiency. However, the combined impact of handle orientations and ground types on wheelchair attendant pusher loads remains unclear.

Evaluating handle ergonomics requires both objective and subjective indicators. This study’s primary outcomes were upper-limb muscle activity, measured as electromyographic (EMG) amplitude normalized to maximal voluntary contraction (MVC) in the ECR and FCU [2,7,14], and usability assessed using the System Usability Scale (SUS) [20,21]. Secondary outcomes include perceived discomfort (Borg CR-10 scale) [22,23] and task completion time, providing complementary insight into physiological strain, user perception, and performance efficiency while maintaining alignment with power and error control.

Therefore, this study aimed to compare muscle load, comfort, usability, and time-on-task between parallel, vertical, and lateral handle orientations on 3 ground conditions in 28 healthy adults pushing a wheelchair load of 65 kg. A randomized crossover design was used to minimize inter-individual variability. Surface EMG was used to quantify muscle load, while subjective discomfort and usability were assessed using validated scales. Time-on-task was included as a functional performance indicator. We hypothesized that parallel handles would increase upper-limb muscle activity, especially on ramps, whereas vertical handles would reduce muscle load but potentially compromise usability.

Material and Methods

PARTICIPANTS:

Twenty-eight healthy volunteers (14 males, 14 females; age 22.8±3.1 years; height 164.4±6.3 cm; body weight 50.0±7.7 kg) participated in this study. Healthy young adults were recruited to establish a controlled baseline for assessing the biomechanical and muscular effects of handle orientation without confounding influences from age-related changes or occupational fatigue. This approach allowed clearer interpretation of fundamental movement patterns before extending the findings to professional caregivers. Participants with musculoskeletal disorders, upper-limb or back injuries, or neurological impairments affecting movement control or force generation were excluded. None of the participants had prior caregiver or wheelchair-handling experience; however, all received standardized instruction and practice in wheelchair propulsion and pushing procedures before testing to ensure task familiarity and consistency. All participants provided written informed consent before participation. The study was approved by the Human Research Ethics Committee of Human Factors Engineering, Chaoyang University of Technology (Ethics ID: CYUTIRB-114-002). Participants received monetary compensation of approximately USD 40 for their time.

EQUIPMENT:

A wireless EMG system (Noraxon, Scottsdale, AZ, USA) was used to record muscle activity from 5 sites: bilateral ECR, bilateral FCU, and the right biceps brachii (BB). The sampling frequency was 1500 Hz. A single-sided BB measurement was selected due to experimental constraints, with careful electrode placement and consistent orientation used to minimize signal crosstalk. After skin preparation to reduce impedance, disposable bipolar Ag/AgCl surface electrodes (Noraxon Dual Electrode, Noraxon, Scottsdale, AZ, USA) were placed over each muscle belly with a 20-mm interelectrode distance [24]. Signals were transmitted wirelessly to a computer for visualization and analysis.

EMG signals were processed using MyoResearch v1.06 software. Following full-wave rectification of the signal, a single-gain bandpass filter with a bandwidth of 40–400 Hz was applied to extract the linear envelope. The signal was then smoothed by calculating the root mean square (RMS) over a time window of 100 ms. To find the peak value of the MVC signal, a 300-ms sliding window was used to calculate the mean amplitude of the signal. This sliding window technique is a more stable method as opposed to finding a single peak [25].

MAXIMUM VOLUNTARY CONTRACTIONS:

Raw EMG signals were normalized to a reference value provided by the MVC of each muscle. The MVC provides maximum EMG output, and, in accordance with Konrad (2005) [25], constitutes a physiologically relevant calibration unit. Participants were instructed to flex their elbows at a 90-degree angle against manual resistance and performed a maximum contraction of the BB. The participants sat in chairs with their fists placed underneath the edge of the desk surface with stable forearm support. They were then instructed to flex their wrists against manual resistance and performed a maximum contraction of their FCU muscle. The MVC of the ECR muscle was obtained by repeating the same procedure in the opposite direction (the direction of wrist extension). The isometric test included 3 MVCs of 5 s each separated by 60-s intervals. The participants were verbally motivated to ensure maximal effort.

DISCOMFORT LEVELS: Subjective measurements were essential for assessing user discomfort [23]; therefore, the Borg CR-10 scale was used to rate discomfort in the arms, hands, and lower back. After completing each task, participants rated their perceived discomfort on a scale from 0 to 10, where 0 indicated “nothing at all” (no exertion, muscle fatigue, or breathlessness) and 10 represented “extremely strong” discomfort, reflecting the most strenuous exertion experienced. This approach provided insight into the perceived physical demands associated with each handle design.

USABILITY: To evaluate the usability of wheelchair handles, this study used the Chinese version of the System Usability Scale (C-SUS) [26], replacing the term “system” with “handle.” The C-SUS consists of 10 items, each scored 0–4. For positively-worded items, the score is the scale position minus 1; for negatively-worded items, it is 5 minus the scale position. The total score is obtained by multiplying the sum by 2.5, yielding a range of 0–100. Scores above 80.3 are classified as Excellent, 68–80.3 as Good, 51–68 as Okay, and below 51 as Poor [27]. The C-SUS demonstrated high reliability (Cronbach’s α=0.93) and content validity (0.90–0.95) [26].

Additionally, task completion time (“time-on-task”) provided a quick and objective usability measure, with shorter times reflecting better user experience [28].

SETTINGS OF HANDLE ORIENTATIONS AND GROUND TYPES:

This study used a Karma SM-352 manual wheelchair (Karma Ltd, Twickenham, UK) with a handle height of 85 cm, handles 41 cm apart, and a wheelchair weight of 15 kg. The seat measured 48 cm in height, 45 cm in depth, and 42 cm in width. An adapter allowed the wheelchair handles to be positioned in 3 orientations: (1) original parallel, (2) vertical, and (3) lateral (Figure 1A–1C). Figure 2 presents a schematic of the handle adapters, showing their angles relative to the wheel/ground and grip diameters. The handle angle relative to the wheel/ground was approximately 3–5°, with grip diameters of 22.2 mm for the parallel handles and 25 mm for the vertical and lateral handles.

Experiments were conducted in a university corridor that complied with Taiwan’s accessibility standards. Test surfaces included a ramp (600 cm long, 50 cm rise, 1/12 slope), level ground (600 cm), and a twisting cone path (600 cm) with 4 triangular traffic cones spaced 1.5 m apart to increase navigation difficulty (Figure 1D–1F).

Each task began when participants initiated wheelchair movement and ended upon crossing the finish line, defining the “time-on-task.” Two familiarization tasks preceded data collection to ensure consistent handling. Standardized 3-min rest intervals minimized fatigue. A 65-kg load simulated the weight of an occupant. Participants maintained a comfortable, self-selected walking speed, monitored via motion capture, to ensure consistency among all conditions.

EXPERIMENTAL PROCEDURE:

This study used a 3×3 within-subjects design with randomized condition order, where participants completed tasks using 3 wheelchair handle orientations (parallel, vertical, lateral) on 3 different ground types (ramp, level ground, cone path). Although participant blinding was not feasible due to visible handle configurations, data analysts were blinded to condition labels during analysis to reduce potential bias.

After collecting participants’ demographic data, the skin was prepared for EMG electrode placement on the bilateral ECR, FCU, and right BB muscles. The MVC was measured 3 times for each muscle.

Each task required participants to propel the wheelchair for 6 meters. The order of tasks for the 3 handle orientations and 3 ground surfaces was randomized using a computer-generated Latin square to ensure each condition appeared equally across participants and positions. Handle and ground surface sequences were counterbalanced to mitigate potential fatigue and carryover effects. Allocation was concealed until task assignment, and participants had 3-min rest periods between tasks to reduce cumulative muscle strain. Overall, participants completed 9 tasks. After each task, they immediately rated their perceived discomfort using the Borg CR-10 scale. Upon completing tasks on all 3 ground types with a given handle orientation, they then evaluated the handle’s usability using the SUS. A subsample of 20 participants completed the SUS again after 2 weeks to evaluate its test–retest reliability.

STATISTICAL ANALYSIS:

A power analysis for a repeated-measures ANOVA with 2 within-subject factors (3 ground types and 3 handles) indicated that a sample size of 28 participants was adequate. The primary endpoint was the change in muscle activation levels (ECR and FCU EMG amplitude) among conditions. The sample size was estimated using G*Power for a repeated-measures ANOVA (f=0.25, α=0.05, power=0.80) [29]. Other outcome measures were exploratory, as family-wise error correction was not applied.

Statistical analyses were performed using SPSS 20.0 for Windows (SPSS, Chicago, IL, USA). Normality was assessed with the Kolmogorov-Smirnov test. For normally distributed variables, repeated-measures ANOVA was conducted for handle orientation and ground type effects. For each key effect, degrees of freedom (df), F-values, P values, and partial η2 with 95% confidence intervals (CI) were reported. Sphericity was assessed using Mauchly’s test, and Greenhouse-Geisser or Huynh-Feldt corrections were applied when the assumption was violated. Pairwise comparisons used Bonferroni adjustments. Effect sizes were reported as partial η2: small (0.01–0.059), medium (0.06–0.139), and large (>0.14) [30]. For non-normal variables, the Friedman test assessed differences among conditions, with Kendall’s W and 95% CI as effect sizes, followed by Wilcoxon signed-rank tests for pairwise comparisons. Test–retest reliability of the SUS was evaluated using the intraclass correlation coefficient (ICC), and Pearson’s correlation assessed associations between SUS scores and time-on-task. Statistical significance was set at P<0.05.

Results

UPPER-LIMB EMG ACTIVITY:

Table 1 presents upper-limb EMG comparisons among handle orientations and ground types with 95% confidence intervals. Table 2 presents the repeated-measures ANOVA results for handle orientation and ground type effects (df, F, p, partial η2). Significant main effects of handle orientation and ground type were observed in the right and left ECR muscles (right: F=67.760, P<0.001; F=26.325, P<0.001; left: F=52.372, P<0.001; F=36.142, P<0.001), FCU muscles (right: F=59.601, P<0.001; F=36.649, P<0.001; left: F=64.295, P<0.001; F=38.028, P<0.001), and the right BB muscle (F=8.346, P=0.001; F=21.369, P<0.001). Significant interaction effects were also observed in the ECR (right: F=14.502, P<0.001; left: F=15.905, P<0.001) and FCU (right: F=8.811, P<0.001; left: F=17.202, P<0.001) muscles.

Bonferroni comparisons showed that amplitudes of the right and left ECR and FCU, and the right BB, were significantly higher with parallel handles than with vertical or lateral handles (P<0.05). For ground type effects, right ECR and left FCU amplitudes were lower on level ground than on ramps and cone paths (P<0.05). The left ECR, right FCU, and BB showed the highest amplitudes on ramps, intermediate on cone paths, and lowest on level ground (Table 1).

Interaction effects (Figure 3) revealed that on ramps, ECR and FCU amplitudes were highest with parallel handles, intermediate with lateral handles, and lowest with vertical handles. On level ground, parallel handles produced significantly higher amplitudes than vertical or lateral handles. On the cone path, ECR amplitude was greater with parallel than lateral handles, while vertical handles showed no difference. FCU amplitude did not vary significantly among handle orientations on the cone path.

DISCOMFORT LEVELS:

Table 3 presents discomfort levels (Borg CR-10 scale) analyzed using non-parametric tests with effect sizes (Kendall’s W, 95% CI). The Friedman test revealed significant differences among handle orientations for arms and hands on level ground (P=0.002) and for the lower back on ramps (P<0.001) and level ground (P=0.015). Ground type effects were also significant for arms and hands with parallel (P=0.001), vertical (P=0.015), and lateral handles (P=0.009), and for the lower back with parallel (P<0.001) and lateral handles (P=0.005).

Wilcoxon signed-rank tests showed that on ramps, lower-back discomfort was greater with parallel handles than with vertical or lateral handles (P<0.05). On level ground, parallel handles caused greater discomfort in arms and hands compared with other orientations, and in the lower back compared with vertical handles (P<0.05). For parallel handles, discomfort was greater on ramps than on level ground or the cone path. For lateral and vertical handles, discomfort was greater on ramps than level ground (P<0.05) (Table 3).

USABILITY:

Figure 4 presents box plots (mean, minimum, maximum) and scatter plots of SUS scores (A) and time-on-task (B) for the 3 handle orientations. Usability differed significantly among handles (SUS: F=18.380, p<0.001; time-on-task: F=8.141, P=0.001). Bonferroni tests showed lower SUS scores for vertical handles (95% CI: 49.7–60.5) than for parallel (95% CI: 66.5–74.9) and lateral handles (95% CI: 65.7–78.9) (P<0.05) (Figure 4A). Time-on-task on the cone path was longer with vertical handles (95% CI: 17.3–19.9) than lateral handles (95% CI: 15.3–18.2) (P<0.05), with no difference from parallel handles (95% CI: 16.8–19.3) (Figure 4B).

SUS reliability was excellent (2-way mixed-effects, single-measure, absolute agreement ICC=0.878, 95% CI: 0.768–0.941, P<0.001) (Figure 5A), and SUS scores correlated moderately and negatively with time-on-task (r=−0.453, P<0.001) (Figure 5B).

Discussion

UPPER-LIMB EMG ACTIVITY:

EMG amplitudes of the ECR, FCU, and BB muscles were significantly higher with parallel handles than with vertical or lateral handles, whereas no differences were observed between vertical and lateral handles. These results indicated that handle design substantially influenced upper-limb exertion, consistent with previous findings [7,11,17]. The activity of wrist extensors and flexors was critical for forearm exertion [18,31], and this activity increased when using parallel handles. Parallel handles increased flexor activation because they align with the direction of motion. Vertical handles form a 90° angle, which reduced maximum push force [11]. Extended elbows further increased flexor force generation and strengthened biceps activation [32]. In contrast, vertical handles reduced muscle activity because the neutral wrist stabilized the elbow and reduced biceps engagement, which may have facilitated easier pushing. Parallel handles required greater wrist strength due to ulnar deviation, which increased the load on the wrist and trunk muscles [33]. Lateral grips can reduce exertion through wrist pronation [14]. These findings will be important for designing push–pull handles that minimize load and improve ergonomic efficiency.

Ground type significantly affected muscle activation during wheelchair propulsion. ECR and BB amplitudes were lower on level ground than on ramps and cone paths, and FCU activity was also reduced on level ground compared with ramps. Parallel handles elicited the greatest upper-limb muscle activity on ramps, consistent with previous findings [13]. On level ground, ECR amplitudes with parallel handles were 34.1% and 34.8% higher than those with vertical and lateral handles, respectively, while FCU amplitudes were 46.5% and 51.8% higher. On ramps, ECR amplitudes increased by 85.5% and 52.6%, and FCU amplitudes increased by 78.2% and 52.9% with parallel handles relative to vertical and lateral handles. The FCU, which is essential for pushing, pulling, and steering [34], showed greater activity with parallel handles on ramps and across all handle types on cone paths. Wrist posture shifted from ulnar deviation, extension, and pronation on level ground to radial deviation, extension, and supination on slopes [35], thereby increasing BB engagement via elbow flexors and supinators. Parallel handles increased biceps brachii activation, whereas lateral and vertical handles preferentially activated the triceps during propulsion [8].

Most previous wheelchair research has focused on seated users, with limited attention to wheelchair pushers. Although pushing is not typically considered highly demanding for the hands, tasks requiring over 30% MVC are classified as high-force activities [36]. In the present study, the ECR and FCU muscles exceeded 30% MVC on ramps and cone paths when using parallel handles, placing them in the high-force category. However, all other muscle groups examined in our study remained suitable for prolonged loading, indicating their ability to manage sustained effort.

DISCOMFORT LEVELS:

Our findings indicated that parallel handles caused greater discomfort in the upper limbs and lower back compared with vertical handles, whereas no significant differences were observed between vertical and lateral handles. Previous studies suggested that parallel handles increased wrist strength demands, leading to greater trunk muscle activity [14], whereas lateral handles increased shoulder flexion and trunk extension moments due to greater downward forces, which in turn reduced trunk strain [12]. Vertical handles required increased wrist and forearm stabilization demands to counter anterior–posterior sway, and this stability helped reduce trunk and shoulder load [19]. A Borg CR-10 score of 3.5 had been identified as a threshold for predicting future neck and lower-back pain [23], and pushing tasks on inclined surfaces increased lumbar spine pressure [37]. During parallel handle use on ramps, Borg CR-10 scores exceeded 3.5, reflecting greater lower-back discomfort relative to vertical handles, consistent with prior findings [14]. However, Borg CR-10 scores may overestimate perceived exertion compared with EMG%MVC [38], highlighting the importance of combining subjective and objective measures when assessing physical workload during wheelchair propulsion.

USABILITY:

Usability ratings differed significantly among handle orientations, with vertical handles receiving the lowest SUS scores compared with parallel and lateral handles. Mean SUS scores were 55.1 for vertical handles, 70.7 for parallel handles, and 72.3 for lateral handles. The vertical handle score fell below the commonly accepted usability benchmark of 68, indicating relatively poor perceived usability [37]. In addition, time-on-task during the twisting cone path test was longer when using vertical handles compared with lateral handles. Turning is frequent when pushing a wheelchair, and different handle designs appeared to influence performance during directional changes. Although both vertical and lateral handles were generally considered more ergonomic than parallel handles [8], lateral handles provided superior leverage and control during turning. When positioned horizontally, lateral handles enabled the handcart pusher to apply force more efficiently, particularly in tight spaces or while navigating obstacles [39]. Although vertical handles may offer comfort for some users, particularly if adjustable, they lack the same level of control and leverage during turns, making lateral handles more suitable for tasks requiring precision and force application.

The SUS demonstrated strong test–retest reliability in this study, with an intraclass correlation coefficient (ICC) of 0.878, indicating good consistency across repeated measurements. ICC values exceeding 0.75 are generally considered to reflect good reliability [40], supporting the SUS as a robust tool for assessing perceived usability in handle design research. Establishing reliability was particularly important for design evaluation, as it ensured that observed differences were attributable to design features rather than measurement variability.

In the present study, SUS scores were moderately and negatively correlated with time-on-task (r=−0.453, P<0.001). This meant that longer task completion times were associated with lower perceived usability. Previous work had reported smaller but similar correlations, around r=−0.24 [27], and in some cases stronger negative correlations, such as r=−0.50 [41]. These results supported the general idea that SUS scores were related to task performance, but they were not a direct measure of performance. Therefore, SUS should have been interpreted together with task-based metrics, such as completion time, to obtain a more complete picture of usability. SUS was not designed to diagnose specific usability problems, and its predictive power for task performance was limited.

TRADE-OFF BETWEEN MUSCLE LOAD AND USABILITY:

Under controlled laboratory conditions with a standardized load, environment, and participant characteristics, handle orientation influenced both muscle activity and perceived usability, revealing a trade-off between caregiver physical demand and functional performance. Parallel handles were associated with higher upper-limb muscle activation and discomfort but were sometimes perceived as advantageous for maneuverability. Vertical handles reduced muscular load but received lower usability ratings, suggesting reduced perceived control or familiarity. In contrast, lateral handles demonstrated a more balanced profile, combining moderate muscle activation with better usability. These findings suggest that optimal handle design should balance biomechanical efficiency and perceptual control rather than prioritizing one outcome alone.

DESIGN RECOMMENDATIONS AND TRANSLATIONAL IMPLICATIONS:

From a translational perspective, lateral handles appeared advantageous for improving comfort and control, particularly during turning. Adjustable handle orientation designs may further accommodate different ground types and task demands. Leeuwis et al (2024) revealed that modifying the rake angle influenced the direction of wheelchair movement where the lateral force was maximal [42]. This adjustment allowed caregivers to push the wheelchair using the push bar with reduced effort, even on ramps. The evidence indicated that both handle orientation and rake angle could significantly affect biomechanical load and perceived usability. Furthermore, a recent systematic review of manual wheelchair usability assessments emphasized that integrating objective biomechanical metrics with subjective user satisfaction measures was crucial for advancing evidence-based wheelchair design [43], and found it was essential to validate usability outcomes of handles among attendants or caregivers based on the findings.

LIMITATIONS:

Several limitations should be noted. First, the sample size was relatively small, which may limit the generalizability of our findings. Second, the participants were healthy young adults rather than professional caregivers, which may limit the generalizability of the findings to individuals with different strength levels, experience, or older caregivers. Finally, EMG did not measure shoulder or lower-back muscles, which limited our ability to assess overall neuromuscular demands and compensatory strategies during the task. However, despite these limitations, this study provides valuable insights into the effects of wheelchair handle orientations on muscle load and usability.

Conclusions

We found that handle orientation and ground type jointly affected the physical demands on attendants during wheelchair propulsion. The parallel handles consistently caused higher wrist and forearm muscle activation levels and greater discomfort, especially on ramps, whereas vertical handles generally resulted in lower muscle activation but poorer usability and longer time-on-task on twisting paths. Lateral handles demonstrated moderate muscle activation with relatively good usability. Overall, handle design affected upper-limb muscle demands. Vertical handles minimized ECR and FCU activation, while lateral handles reduced exertion and enhanced efficiency. Given that about 70% of caregivers are women [44], future studies should examine sex differences.