TL;DR
This paper introduces a low-cost, batteryless system using Wireless Identification and Sensing Platform (WISP) tags to automatically detect when patients enter or exit their beds. The WISP worn on the sternum proved more accurate than a mattress-based approach, offering a promising solution for fall prevention in healthcare settings.
What Problem Does This Paper Address?
Falls are a significant concern in hospitals and residential care facilities, with a substantial proportion occurring around the bed. Existing bed-exit alarm systems have shown mixed results in reducing falls and can be costly, wired, or require frequent maintenance. The authors aim to address this by developing a more accessible, low-cost, and batteryless monitoring solution that can accurately identify posture transitions associated with bed entry and exit, potentially alerting caregivers before a fall occurs.
Methods
The study explores two primary technological approaches using WISP devices, which are batteryless RFID tags equipped with kinematic sensors.
- WISP Worn on Sternum: A WISP tag is attached to the subject's clothing over the sternum. The system uses a 20-second segmentation of acceleration data from three axes (anteroposterior, mediolateral, and dorsoventral) and the Received Signal Strength Indicator (RSSI) from the RFID reader.
- Lying to Sitting/Sitting to Lying: Detected using the anteroposterior axis acceleration. Thresholds are applied to differentiate between lying and sitting/standing states.
- Sitting to Standing/Standing to Sitting: Identified by analyzing the trunk displacement angle (derived from acceleration data), the duration of the posture transition (TD), and the RSSI pattern to infer proximity to the RFID reader. A bed exit event is defined as a lying-to-sitting transition followed by a sitting-to-standing transition. Bed entry is the reverse sequence.
- WISP Attached to Mattress: A WISP is placed on the side of the mattress. This method focuses on the derivative of the z-axis acceleration (perpendicular to the mattress) to detect changes in mattress deformation. A novel algorithm identifies static states (empty, sitting, or lying) and classifies transitions based on changes between these states. A bed exit is detected when the state transitions from sitting/lying to empty.
Datasets
The study evaluated the algorithms on data collected from ten healthy adult volunteers. These volunteers performed scripted routines involving various postural transitions, including getting into and out of bed, walking, and sitting/standing from a chair. The data was collected in a clinic trial room. No pre-existing datasets were used; the study generated its own data for evaluation.
Key Insights and Ideas
- Batteryless and Low-Cost is Achievable: The use of WISP technology demonstrates that effective posture monitoring can be achieved without the need for batteries or high costs, making it more scalable for healthcare environments.
- Sternum Placement Superiority: The WISP worn on the sternum significantly outperformed the mattress-based approach in accurately identifying bed entry and exit events. This suggests that direct body-worn sensors, even simple ones, can capture more reliable kinematic data for these specific transitions.
- RSSI as a Proximity Indicator: The integration of RSSI data proved useful in inferring the subject's position relative to the RFID reader, aiding in the classification of posture transitions.
- Simplicity of Algorithms: The algorithms, particularly for the sternum-worn WISP, rely on relatively straightforward analysis of acceleration and signal strength, making them potentially easier to implement and adapt.
Key Results
- WISP Worn on Sternum: Achieved high performance metrics:
- Getting into bed: Sensitivity of 92.8%, Specificity of 97.5%.
- Getting out of bed: Sensitivity of 90.4%, Specificity of 93.8%.
- WISP Attached to Mattress: Showed lower performance:
- Getting into bed: Sensitivity of 84.2%, Specificity of 97.4%.
- Getting out of bed: Sensitivity of 79%, Specificity of 97.4%.
- ROC Curves: The WISP worn on the sternum method demonstrated better performance with higher Area Under the Curve (AUC) values (0.931 for getting in, 0.859 for getting out) compared to the mattress-based method (0.882 for getting in, 0.855 for getting out).
Limitations and Future Work
- Healthy Adult Subjects: The study was conducted on healthy young adults. The authors acknowledge that frail older patients, who are the primary target population, may exhibit different movement patterns and speeds, requiring further investigation.
- Garment Influence: Loosely fitted hospital garments could affect the sensor's ability to closely track body movements, potentially impacting algorithm effectiveness.
- Limited Range and Blockage: WISP transmissions can be blocked by certain positions (e.g., lying prone with the sternum-worn WISP) and have a limited range (3-4m), necessitating multiple antennas for adequate coverage.
- Non-Conventional Exits: The study did not extensively investigate less conventional bed exit methods such as shuffling, using mobility aids, or exiting over bedrails.
- Future Work: The authors suggest further investigation into the system's performance with frail older patients, exploring adaptive threshold adjustments for individual patients, and addressing the limitations of signal blockage and range.
Summary updated