Longitudinal Wearable Physiological Dataset of Elderly Individuals

ElderlyPhysio-v02 is a multimodal wearable-physiology dataset designed to support research on elderly health monitoring, virtual replicas/digital twins, and AI-based decision support. The release aggregates longitudinal Fitbit-derived time series captured over 49 days at multiple temporal granularities, including seconds-level heart rate (1–3 s), minute-level heart rate, daily heart-rate variability (e.g., RMSSD), minute-level SpO₂ and estimated oxygen variation, daily respiratory rate, sleep stage epochs (sub-minute/second-level categorical labels), activity-derived live pace, daily calories, and high-frequency skin temperature streams (body and wrist, 1 s), plus derived composite sleep-session metrics. Data are provided in standard machine-readable formats (JSON/CSV) with timestamps to enable multimodal fusion, missingness studies, forecasting, and anomaly/event detection; accompanying metadata/documentation describe variable definitions, units, and file organization for reproducible reuse.