A-112: Characterizing Passively Collected Digital Biomarkers as Predictors of Subtle Functional Decline in Older Adults

Abstract

Objective: Smartphone-based passive sensing offers an unobtrusive way to assess real-world behavior in older adults; however, it is unclear how these markers relate to perceived changes in functional status. We examined these associations in participants from the ongoing Technology for Smartphone Assessment of Neurocognitive Symptoms (TechSANS) study. Method/Data Selection: Participants (N = 55; Mage = 75.16, 60% female) were classified as functionally unimpaired (n = 30) or functionally impaired (n = 25) using a previously established dichotomous cut-point (≥1.37) of the self-reported 12-item Everyday Cognition Scale (ECog-12). Independent samples t-tests examined group differences in 30-day aggregated smartphone-based metrics spanning gait, life-space mobility, and device usage. In exploratory analyses, logistic regression evaluated relationships after adjusting for age and education. Results/Data Synthesis: Of several potential markers, passive gait markers most consistently emerged as predictors of functional difficulties with small-to-medium effect sizes (Cohen’s |d| = 0.39 to 0.54). Adjusting for age and education, average walking speed in particular significantly improved model fit (Nagelkerke R2 = 0.148), such that each 0.1 m/s increase was associated with 28% lower odds of being classified as functionally impaired (OR = 0.72, 95% CI, [0.53, 0.99]). Conclusions: Preliminary findings indicate that passive gait markers may be promising real-world metrics that can identify subtle functional decline. Further work is needed to validate gait metrics and account for various functional confounds. Other smartphone-based metrics warrant ongoing investigation to characterize the behaviors they capture and their potential relevance to subtle functional changes in daily life.