Reliability, construct validity, and usefulness of an assessment device for agility and motor-cognitive reactive stepping performance in community-dwelling older adults

Age-related cognitive and motor decline impairs daily functioning and increases fall risk. Although both domains interact in everyday activities, they are often assessed separately, limiting ecological validity. This study examined reliability and construct validity of an assessment device for agility and motor-cognitive reactive stepping performance in older adults. In a test-retest design, 72 older adults (48 women; 73.7 ± 7.6 years) completed agility tests (Random Star Run and a dual task with multiple object tracking) and motor-cognitive tasks on the SKILLCOURT across four sessions: familiarisation, test day 1 and day 2 (7 days apart), and test day 3 (3 months later). Agility tasks required reactive changes of direction in response to visual cues, while motor-cognitive stepping tasks assessed simple and choice reaction time, task switching, 2-back, and Stroop word-color interference. Intersession Reliability was evaluated using intraclass correlation coefficients (ICC) and coefficients of variation (CV), learning effects with linear mixed models, and construct validity via age-adjusted regression against established motor (Timed-up-and-Go; TUG; single-/dual task) and seated PC-based cognitive measures of reaction speed and executive functions corresponding to the SKILLCORT tasks. Significant learning effects were observed mainly between test days 1 and 2. Relative reliability was moderate to excellent for SKILLCOURT agility, simple reaction time, and Stroop word-color performance (ICC = 0.70-0.94), but lower for choice reaction time, task switching, and the 2-back task (ICC = 0.49-0.66). Absolute reliability was generally good (CV = 4.6-7.2%), except for the SKILLCOURT task switching and 2-back (~12%) and, in particular, the dual task agility test (32%). Random Star Run agility was predicted by TUG single-task performance, PC-based Stroop interference control, and simple reaction time (R2 = 0.77), while dual task agility was predicted by TUG dual task and PC-based 2-back performance (R2 = 0.37). All motor-cognitive outcomes were significantly associated with corresponding PC-based cognitive measures (R2 = 0.20-0.41). The SKILLCOURT demonstrated good reliability for the Random Star Run agility test and most motor-cognitive measures, although adequate familiarisation appears necessary to minimise learning effects. Construct validity was confirmed for agility measures, while reactive stepping tasks appeared to involve greater motor demands and task interference, which may weaken correlations with comparable PC-based cognitive tests.