Wearable monitoring during music-based interventions in dementia: physiological and behavioral observations from a pilot study.

Introduction: Music-based interventions (MBIs) are widely used in dementia care, but objective methods for characterizing participant responses during intervention sessions remain limited. Synchronized datasets combining wearable physiological signals and behavioral observations are particularly scarce. Methods: We conducted a pilot feasibility study involving five individuals with Alzheimer's disease and related dementias (ADRD) who participated in 13 formal MBI sessions. Physiological signals, including photoplethysmography (PPG), electrodermal activity (EDA), skin temperature (TEMP), and accelerometry (ACC), were collected using a wrist-worn wearable sensor and synchronized with intervention playlists and time-stamped behavioral observations. Exploratory analyses examined physiological responses across intervention phases, participant-specific response patterns, time-of-day effects, and music-preference effects. Results: The dataset contains 13 intervention sessions, 99 music segments, and 248 behavioral observations. PPG, ACC, TEMP, and behavioral observations were available for all sessions, while EDA quality varied because of sensor-contact challenges. Behavioral responses were highly heterogeneous across participants, with engagement and calm behaviors observed most frequently. Physiological responses also showed substantial inter-individual variability, and case studies demonstrated that physiological and behavioral responses were not always concordant. Conclusion: This study demonstrates the feasibility of collecting synchronized physiological, behavioral, and intervention-context data during MBIs in people living with dementia. The resulting publicly available multimodal dataset provides a foundation for future investigations of participant-specific responses and adaptive music-based interventions.