Abstract Background and Objectives TechSANS (Technology for Smartphone Assessment of Neurocognitive Symptoms) is a digital phenotyping project exploring passive smartphone sensing as a complement to infrequent clinical assessments for long-term cognitive characterization. This manuscript presents a guide for passive sensing applications, including a primer on data modalities and derived measures, initial findings on feasibility and analytic considerations, and preliminary relationships with cognitive performance from an ongoing study of older adults. Research Design and Methods An analytic pipeline cleaned the raw data and extracted 145 digital phenotyping features from 6 months of multimodal passive smartphone sensing data for 21 participants (aged 75.81 ± 4.86 years, 13 cisgender women; 17 cognitively normal, 4 with mild cognitive impairment/dementia), characterizing daily behaviors and smartphone interactions. Generalized linear mixed models assessed associations between these measures and baseline cognitive performance. Results Digital measures were extracted for 141 days per participant on average (74.4% of the data inclusion period), suggesting good study adherence. Statistical analyses identified relationships between cognitive performance, smartphone typing, and gait. Specifically, poorer working, episodic, and semantic memory were associated with slower typing, more frequent typing errors, slower walking, higher walking asymmetry, and lower walking cadence. Discussion and Implications This manuscript introduced clinical scientists to the technical foundations of passive smartphone sensing. The exploratory cross-sectional analysis suggested the feasibility of this approach in older adults for scalable, long-term cognitive characterization. We also provided practical considerations to improve future research and highlighted the need for larger, more diverse cohorts to discover and validate generalizable digital biomarkers.
Shen et al. (Sat,) studied this question.