Abstract
Inertial Measurement Units (IMUs) within everyday consumer smartwatches offer a low-cost method to monitor hospital patients' natural behavior. However, their clinical accuracy and acceptability remain unproven. This two-stage study first compared wrist-worn IMUs (Apple Watch Series 3 and 5, Xsens MTw Awinda, and Axivity AX3) against gold-standard optical motion tracking (OptiTrack). Following validation, we evaluated patient (N=44) and staff (N=15) attitudes toward smartwatch use in a clinical setting. Results showed that while Xsens was most accurate for acceleration ($R^2=0.88$), smartwatches were marginally more accurate for angular velocity ($R^2=0.85$). Surveys indicated high levels of comfort, ease of use, and clinical acceptability. Smartwatches represent a viable, pervasive tool for inpatient motion monitoring.
Introduction
Monitoring physical activity and recovery in hospital settings—particularly for stroke or neurology patients—is often manual and subjective. While optical motion capture systems like OptiTrack provide high precision, they are impractical for daily ward use. Research-grade IMUs offer portability but can be expensive and obtrusive. Consumer smartwatches present a ubiquitous alternative, yet their clinical utility is often limited by proprietary "black box" algorithms and unknown patient/staff acceptance. This study addresses these gaps by validating raw sensor data from consumer wearables and assessing their feasibility in a clinical environment.
Methods
The study employed a two-stage approach:
- Technical Validation: Twelve healthy volunteers wore a "sensor stack" consisting of Apple Watch Series 3, Apple Watch Series 5, Xsens MTw Awinda, and Axivity AX3. Participants performed standardized upper-body movements tracked by an OptiTrack system. A custom WatchOS app was developed to extract raw triaxial acceleration and angular velocity at 100Hz, bypassing standard processed metrics.
- Clinical Acceptability: Apple Watches were deployed to 44 inpatients for 1.5 to 24 hours. Post-trial surveys assessed the experiences of patients and 15 healthcare professionals regarding comfort, ease of use, and potential clinical integration.
Results and Discussion
Sensor Accuracy and Consistency
The technical validation revealed that sensor fusion is critical for accuracy. The Apple Watch and Xsens, which utilize both accelerometers and gyroscopes, significantly outperformed the Axivity AX3 (an acceleration logger). For acceleration, Xsens achieved the highest agreement with the gold standard ($R^2 = 0.88$), followed by the Apple Watch ($R^2 = 0.64$). Notably, for angular velocity, the Apple Watches were marginally more accurate than the research-grade Xsens ($R^2$ of 0.85 vs 0.82). Furthermore, the Apple Watch Series 3 and 5 produced nearly identical raw data, indicating that different hardware generations can be used interchangeably for longitudinal clinical monitoring.
Clinical Feasibility
Patients were highly receptive to the technology, with 88% strongly agreeing the devices were easy to use and 95% reporting no anxiety. The "normalcy" of smartwatches compared to medical-grade sensors reduced stigma. Healthcare staff were generally positive but raised valid concerns regarding data privacy and the potential for "false-positive" alerts in automated systems.
Conclusion
Consumer smartwatches, when "unlocked" via custom software to provide raw data, offer accuracy comparable to research-grade sensors. Their high acceptability among patients and staff suggests they are well-suited for long-term, pervasive monitoring in hospital wards, potentially enabling more objective recovery tracking and personalized rehabilitation.
Limitations and Future Work
The study's validation phase used short-term (5-minute) protocols, which may not account for long-term sensor drift. Additionally, the staff sample size was relatively small. Future work should focus on "closed-loop" systems that provide real-time feedback to clinicians and patients to assist in active rehabilitation.
Summary updated