Physical human-robot interaction mediates the association of motor impairment and kinematic performance for poststroke arm rehabilitation.

BACKGROUND: Robot-assisted therapy (RAT) has demonstrated potential in revolutionizing poststroke rehabilitation settings. Beyond facilitating motor recovery, robotic systems can offer kinematic assessments to objectively measure arm performance. Physical human-robot interaction (pHRI), a core component of RAT, enables the system to adapt kinematic performance tailoring arm motor impairments. While previous studies have explored correlations between onboard kinematics and clinical outcomes, the specific role of pHRI in mediating motor impairment and kinematic performance remains under-investigated. This study aims to examine whether pHRI mediates the relationship between arm motor impairment and onboard kinematic performance, and whether higher-quality pHRI contributes to improved motor outcomes in individuals with stroke. METHODS: A mediation analysis study was conducted using data from the Exoskeleton-assisted Anthropomorphic Movement Training (EAMT) trials between December 2018 and August 2021. Data were analyzed September to December 2024. Physical human-robot interaction that provides real-time adjustments to maintain a more physiological arm posture was applied during functional tasks in RAT. Participants underwent RAT using an exoskeleton that measured kinematic metrics (movement time, smoothness, and postural error) and pHRI. Motor impairment was assessed using the Fugl-Meyer Assessment for Upper Extremity (FMA-UE). Mediation analysis was employed to examine whether pHRI mediated the association between motor impairment and kinematic performance. RESULTS: After screening 312 patients for eligibility, data of 50 participants (mean [SD] age, 49.5 [10.2] years; mean [SD] time since stroke onset, 55.2 [44.0] days) with subacute stroke were included in the study. Higher FMA-UE scores were significantly associated with improved kinematic performance, including reduced movement time (standardized r = -0.459, P < 0.001), increased smoothness (standardized r = -0.369, P = 0.005), and reduced postural error (standardized r = -0.488, P < 0.001). Moreover, pHRI significantly mediated the association between motor impairment and postural error, accounting for 43.7% of the total effect (95% CI: 12.3 to 75.1, P = 0.006). No mediation was observed for movement time or smoothness. Sensitivity analyses supported the robustness of these findings. CONCLUSIONS: pHRI accounted for 43.7% of the relationship between motor impairment and postural error during robotic training for stroke rehabilitation, indicating a substantial mediating effect. These findings highlight the potential of optimizing pHRI to improve the effectiveness of rehabilitation robots for broader motor outcomes, particularly in enhancing motor control and investigating the underlying pHRI mechanisms regarding robotic design and control strategies.