Time-stratified daily walking speed measurement via smartphone and its predictive utility for mild cognitive impairment

Abstract

This study evaluated the utility of a smartphone-based walking-speed application for early mild cognitive impairment (MCI) screening under every day, real-world conditions. Community-dwelling adults aged ≥ 65 years were recruited from dementia prevention activities in Yokote City, Akita Prefecture, (entry period: October 2022-January 2023; assessment period: October 2022-May 2023). A walking-speed app was installed on each participant's personal smartphone, and outdoor, GPS-enabled daily walking speed was continuously recorded. At baseline, we collected clinical and physical measures, including infrared-measured 5-meter usual walking speed (UWS), the National Center for Geriatrics and Gerontology-Functional Assessment Tool (NCGG-FAT), and the Touch Panel-type Dementia Assessment Scale (TDAS). Of 123 enrollees, evaluable walking-speed data were obtained from 115. To reduce noise in the raw dataset, segments meeting predefined criteria for walking distance, walking time, walking speed, steps per unit time, and step length were extracted and defined as unconscious walking speed (UcWS). In time-stratified analyses evaluating its potential as a proxy for UWS, UcWS showed a positive correlation with UWS (maximum r = 0.47). Compared with clinical measures, UcWS was significantly slower in MCI positive participants (median comparison p = 0.018; mean comparison p = 0.011), supporting UWS substitutability. Using walking-duration stratification, we identified a foundational UcWS analytic framework that may serve as a practical proxy for UWS using readily obtainable smartphone-based GPS and accelerometer data. Reduced UcWS was also associated with MCI screening positivity, supporting its potential utility as a scalable real-world digital biomarker for early cognitive decline research.