Detection of Real-World Trips in At-Fall Risk Community Dwelling Older Adults Using Wearable Sensors

Abstract

Background: Near-falls such as a trip, slip, stumble, or misstep involve a loss of balance (LOB) that does not result in a fall, occur more frequently than actual falls, and are associated with an increased fall risk. To date, studies have largely involved detection of simulated laboratory LOBs using wearable devices in young adults. Data on the detection of and kinematics of naturally occurring LOBs in people at high risk of falling are lacking. This may provide a new way to identify older adults at high risk for falls. We aimed to explore key body kinematics underlying real-world trips in at-fall risk community dwelling older adults wearing inertial measurement units (IMU). Methods: Five community-dwelling older adults with a history of falls who reported trips during the study period participated. They wore a voice recorder and 4 IMUs mounted on feet, lower back and wrist for two consecutive weeks to provide a record of the context and timing of LOB events. Sensor data prior to time-stamped voice recording of a trip were processed in order to visually identify unusual foot trajectories and lower back and arm orientations. Then, data of feet, lower back and wrist position and orientation were combined to create a three-dimensional animation representing the estimated body motion during the noted time segments in order to corroborate the occurrence of a trip. Events reported as a trip by the participant and identified as a trip by a researcher, blinded to voice recordings description, were included in the final analysis. Results: A total of 18 trips obtained from five participants were analyzed. Twelve trips occurred at home, three outside and for three the location was not reported. Trips were identified in the sensor data by observing (1) additional peaks to the typical foot velocity signal during swing phase; (2) increased velocity of the contralateral foot and (3) sharp changes in lower back pitch angles. Conclusions: Our approach demonstrates the feasibility of identifying and studying the mechanisms and context underlying trip-related LOBs in at-fall risk older adults during real world activities.

Keywords: accelerometers; balance; inertial measurement units; monitoring; near-falls; walking.

Figure 1

IMUs mounting location at feet, lower back and one wrist (dashed yellow line) and a voice recorder worn on the other wrist (dashed red line).

Figure 2

Data obtained from 3 participants, showing: (top) magnitude of foot velocity corresponding to right and left foot and lower back (middle) and wrist (bottom) tilt angles. The dots indicate instances during the stance phase where the foot was stationary. The solid arrows indicate the trip location identified by a sharp increase in velocity or several “peaks” indicating changes in foot velocity. The dashed arrows show the recovery response as a change in velocity of the contralateral foot (B) or variations in lower back pitch angles (A–C). We could not identify trips based on data obtained from the wrist sensor. Participant's self-report: (A) “Caught left foot under carpet, stumbled” (B) “climbing steep hill, tripped over speed bump” (C) “Tripped on a piece of wood”.

Figure 3

Data demonstrating trips occurring at different times during the swing phase. Magnitude of foot velocity corresponding to right and left foot. The dots indicate instances during the stance phase where the foot was stationary. (A) A sudden change in right foot's velocity during mid swing (immediately after peak velocity) while climbing stairs (stereotypical velocity profile of climbing stairs contains short periods of nearly constant velocity). (B) A sudden change in velocity of the left foot during mid swing (i.e., the foot almost reached maximal velocity), then the foot landed on the ground (foot velocity equal to zero) and a fast, compensatory step with the same foot was performed. (C) A sudden change in maximal velocity of the left foot during terminal swing. An increased velocity of the contralateral (right) foot afterwards. Arrow indicates the trip location. Participant's self-report: (A) “Stumbled going upstairs, tripped” (B) “scuffed foot on floor, toe wouldn't slide” (C) “stepped in hole, tripped”.

Figure 4

Data showing a trip that resulted in a fall. (A) (top) Magnitude of foot velocity corresponding to right and left foot and (middle) lower back and wrist (bottom) tilt angles. Numbers indicating: (1) the trip location (2) lower back moved backwards (increasing pitch angle) and rightwards (increasing roll angle) while the participant fell onto the bed, (3) participant immediately rolled back to mid position (decreasing roll angle) and remained backwards (approximately constant pitch angle). (B) (top) right and (bottom) left foot tilt angles. The participant recorded: “Walking to back door to let dog in, stepped on cane. Fell on a bed”.