Smartwatch-based automated lifelog platform: feasibility and preliminary real-world evaluation with real-world data from post-stroke patients

Abstract

Abstract Real-world monitoring may complement clinic-based assessment after stroke by capturing daily physiological and activity-related information outside scheduled visits. We developed and evaluated a wearable-to-cloud platform that integrates smartwatch-derived activity and photoplethysmography-derived heart rate variability (PPG-HRV) signals with digital questionnaire-derived clinical scores in post-stroke patients. In this three-center feasibility study, 52 stroke survivors were enrolled, and 47 participants were included in the final analytic cohort after 5 withdrawals. The mean monitoring duration was 2.65 ± 1.69 months per participant. Valid participant-level inertial measurement unit (IMU)-derived activity and PPG-HRV summaries were available for 37 participants, and the primary clinical-wearable association analysis included 36 participants with complete paired data. In the primary eight-test analysis, higher MoCA-K scores were moderately associated with higher HRV/autonomic composite values, indicating that better cognitive performance was associated with a more favorable wearable-derived autonomic/HRV profile (Spearman rho = 0.488, 95% CI 0.19–0.70; raw P  = 0.0026; Benjamini-Hochberg q = 0.021). This association was similar after age adjustment (rho = 0.503). These findings support the feasibility of multi-center smartwatch-based lifelog monitoring and identify an early cognition-autonomic physiology signal requiring validation in larger longitudinal cohorts.