300 Section - Using Auditory Dual-Task walking tests to detect dual-task performance deficits in people with mild cognitive impairment

Objectives : Mild cognitive impairment (MCI) often presents with subtle deficits in executive control and gait that become more apparent under dual-task conditions. Auditory dual-task walking paradigms simultaneously challenge motor and cognitive resources and may enhance sensitivity to early cognitive decline. This study aimed to investigate whether combining cognitive performance and gait parameters in auditory dual-task walking tests can effectively distinguish older adults with MCI from healthy controls.

Design: A total of 113 participants (66 MCI, 47 controls; mean age = 70.88 ± 4.66, 56% female) completed three tasks: (1) simple walking, (2) auditory Stroop test while walking, and (3) auditory clock test while walking. Gait parameters were recorded using APDM Wearable Technologies Mobility Lab sensors, and cognitive performance was measured using PsychoPy software. Two-way analysis of covariance (ANCOVA) was used to detect differences between the two groups. Feature selection and modeling were performed using the Least Absolute Shrinkage and Selection Operator (LASSO) regression, and receiver operating characteristic (ROC) curve analysis was applied to assess the discriminatory ability of the models.

Results: Twenty-nine gait features and three cognitive features (accuracy, reaction time, omission rate) were extracted. In the dual-task Stroop test, 17 gait features and 2 cognitive features differed significantly between the MCI and control groups, while in the dual-task clock test, 15 gait features and 2 cognitive features differed (two-way ANCOVA). LASSO regression models showed that the Stroop dual-task model achieved an area under the ROC curve (AUC) of 0.81, and the clock dual-task model achieved an AUC of 0.86 for distinguishing MCI from controls.

Conclusion: Auditory dual-task walking tests that integrate gait and cognitive performance effectively differentiate MCI from healthy controls, offering a promising tool for early detection. Future work should validate generalizability, optimize feature sets, and assess longitudinal predictive value for conversion risk.