Can Gait Characteristics Be Represented by Physical Activity Measured with Wrist-Worn Accelerometers?

Area of Science:

• Biomedical Engineering
• Clinical Trial Technology
• Wearable Sensor Technology

Background:

• Wearable accelerometers enable continuous monitoring in naturalistic settings.
• Lumbar devices track gait, while wrist devices measure physical activity (PA).
• Using multiple devices increases patient burden and trial complexity.

Purpose of the Study:

• To assess if wrist-worn device PA data can represent gait features from lumbar devices.
• To explore reducing sensor numbers in clinical trials for improved compliance and data quality.

Main Methods:

• Utilized 7-day continuous at-home data from wrist- and lumbar-worn GENEActiv devices from healthy participants.
• Employed statistical methods: penalized regression, principal component regression, partial least square regression, and JIVE.
• Incorporated multilevel models to analyze both between- and within-subject effects.

Main Results:

• Selected gait features from lumbar devices were adequately represented by PA features from wrist-worn devices.
• Demonstrated preliminary evidence for reducing the number of wearable sensors in clinical trials.
• Provided an analytical framework for comparing repeated measures from multiple data modalities.

Conclusions:

• Wrist-worn accelerometers can potentially replace lumbar devices for gait analysis in certain contexts.
• Reducing sensor burden can enhance patient comfort and compliance in clinical trials.
• The statistical framework is applicable for multimodal wearable sensor data analysis.