Turning versus straight walking under dual-task conditions: a preliminary study for identifying mild cognitive impairment and Alzheimer's disease.

BACKGROUND: This preliminary study is the first to systematically compare dual-task turning versus straight walking metrics for detecting cognitive impairment in older adults. METHODS: Seventy-five participants, 26 neurotypical (NT), 25 amnestic mild cognitive impairment (aMCI), and 24 mild Alzheimer’s disease (AD), performed a dual-task walk and turn test. Gait and turn features were extracted and compared through two distinct models: walk-phase and turn-phase. Two models were evaluated (walk-phase vs. turn-phase) using a repeated nested cross-validation pipeline. RESULTS: The turn-phase model achieved multiclass balanced-accuracy (macro-averaged recall across NT/aMCI/AD) 68.8% vs. 58.8% for the walk-phase (Wilcoxon p < 0.001, Cohen’s d = 0.67), a statistically significant yet clinically modest ~ 10-point gain. Class AUCs favored the turn-phase for AD (0.92 vs. 0.84) and NT (0.88 vs. 0.82), whereas aMCI discrimination was modest for both (≈ 0.63–0.65). Feature interpretation identified gait speed (walk-phase) and peak head speed and peak segmental angle (turn-phase) as the strongest contributors, with medium-to-large Hedges’ g for key turn features. CONCLUSIONS: Findings provide proof-of-concept that turn-phase metrics add discriminative information beyond straight walking. Given the modest absolute accuracy and limited aMCI performance, results should be considered preliminary and warrant external validation in larger and more diverse cohorts.