Real-world falls collected with a wearable multi-sensor and multi-domain setup

Falls are a leading cause of injury and disability in older adults. Yet, knowledge about their real-world patterns is still limited, despite successful efforts such as recorded videos of falls in long-term care residences (Robinovitch et al, Lancet, 2013) and recorded kinematics of falls using a single IMU sensor worn on the lower back (Farseeing project and related dataset, Klenk et al, Eur Rev Aging Phys Act, 2016). Still, no data on falls recorded with more than one sensor worn on multiple positions are available in the literature.

During the DARE FALLSPREDICT-GP study (ongoing study in Bologna, Italy, targeting the general over-65 population), participants were monitored for one week using a wearable multi-sensor setup: an inertial measurement unit (accelerometer + gyroscope) on the lower back (McRoberts Dynaport 7), and a smartwatch-like sensor (Empatica Embrace Plus), featuring an accelerometer detecting non-dominant wrist movement and a photopletismograph for cardiac activity. The subjects recorded information (e.g., date, time, description, location) about any falls in a diary. This information was checked and consolidated during monthly interviews after the monitoring week.

In this contribution, to the best of our knowledge, we will present for the first time three real-world falls from three different subjects, recorded with a multi-sensor setup covering axial and distal positions and multiple domains (movement and cardiac functions). We will show the physiological signals before, during, and after the fall, providing preliminary insights into how combining information from different sources may help improve understanding of fall patterns, support better fall detection systems, and improve fall prediction and prevention. 

This research was funded by the Italian Complementary National Plan (PNC-1.1), DARE – DigitAl lifelong pREvention (PNC0000002; CUP B53C22006450001).