Reproducibility and minimal detectable change of three-dimensional kinematic analysis of reaching tasks in people with hemiparesis after stroke.

BACKGROUND AND PURPOSE Three-dimensional kinematic analysis of reaching has emerged as an evaluative measure of upper-extremity motor performance in people after stroke. However, the psychometric properties supporting the use of kinematic data for evaluating longitudinal change in motor performance have not been established. The objective of this study was to determine, in a test-retest reliability manner, the reproducibility and minimal detectable change for reaching kinematics in people after stroke. SUBJECTS AND METHODS Fourteen participants with hemiparesis after stroke performed forward reaching tasks on 2 occasions 37.3 (SD=9.8) days apart. At each session, participants performed 4 forward reaching tasks produced by the combination of 2 target heights (low and high [109 and 153 cm from the floor, respectively]) and 2 instructed movement speeds (self-selected and as fast as possible). Two analytical methods were used to calculate kinematic parameters. RESULTS Relative reliability (intraclass correlation coefficient) ranged from .04 to .99, and absolute reliability (standard error of measurement) ranged from 2.7% to 76.8%, depending on the kinematic variable, the demands of the motor task (target height and movement speed), and the analytical method. Bland-Altman analysis, a statistical method used to assess the repeatability of a method, revealed few systematic errors between sessions. The minimal detectable change ranged from 7.4% to 98.9%. DISCUSSION AND CONCLUSION Depending on the demands of the motor task and the analytical method, most kinematic outcome measures (such as peak hand velocity, endpoint error, reach extent, maximum shoulder flexion range of motion, and minimum elbow extension range of motion) are reliable measures of motor performance in people after stroke. However, because of the magnitude of within-subject measurement error, some variables (such as peak hand velocity, time to peak hand velocity, and movement time) must change considerably (>50%) to indicate a real change in individual participants. The results of our reliability analysis, which are based on our cohort of participants with hemiparesis after stroke and our specific paradigm, may not be generalizable to different subpopulations of people with hemiparesis after stroke or to the myriad movement tasks and kinematic variables used for the assessment of reaching performance in people after stroke.