What constitutes a ´near-miss´ among frequent fallers with Parkinson´s: can sensors detect the indications of instability that clinicians note?

Stack, E., Agarwal, V. and Ashburn, A. (2016) What constitutes a ´near-miss´ among frequent fallers with Parkinson´s: can sensors detect the indications of instability that clinicians note? Physiotherapy, 102 (1), e235-e236. (doi:10.1016/j.physio.2016.10.292).

Abstract

Relevance: Sensors in the home afford the opportunity to monitor a patient's risk of falling, allowing timely intervention. Sensors can detect falls (and raise the alarm). The challenge is to develop technology capable of measuring the frequency of near-misses, so that help may be provided before a significant fall.

Purpose:(1)
To identify the features of ‘near-misses’ that observers note in people at high risk of falling moving freely around their homes.
(2)
To design a protocol for establishing whether sensor can identify the same features and thus determine whether an individual is at imminent risk of falling.


Methods/analysis: Via support groups, with Ethics Approval (from the University of Southampton), we recruited five people with Parkinson's who could (a) describe frequent falls or near-misses and (b) mobilise independently indoors. We observed and video-recorded each participant at home (a maximum four times) performing their usual activities. An independent reviewer noted and counted instances when participants appeared to ‘lose their balance’ and nearly fall. We then designed a protocol for testing the ability of depth cameras and wearable amonitors to detect indications of instability during frequently challenging activities.

Results: Participants had been diagnosed between 5 and 11 years; they all had moderate or severe Parkinson's and were largely sedentary. Reviewing 246 minutes of video, participants appeared at imminent risk of falling 227 times, most frequently when turning, on steps, if they walked immediately on rising or if they did not use support when transferring. Balance was often lost backwards unless walking (when participants tended to stumble forwards or sideways if their feet did not clear the floor or crossed, or they tripped or froze). Unsteady transfers were characterised by swaying backwards on rising (toes lifting off floor) or by falling into the chair, either on rising or during sitting (feet lifting off floor). We designed a protocol to capture these features plus ‘pausing and repositioning’, ‘aborted attempts’, ‘increased sway’ and ‘loss of control’ during the following activities: chair transfers, walking, turning and standing.

Discussion and conclusions: ‘Teaching’ machines to recognise changing postural instability is a multistage process. By observing individuals who nearly fell almost every minute they were active we have identified the features that ‘flag’ a near-miss to clinicians. Unlike the simulated falls (by healthy volunteers) on which others have based their machine learning, we are progressing towards sensor-based detection of fall risk using real events demonstrated by people with marked postural instability.

Impact and implications: Clinicians cannot observe their patients all day every day, certainly not at home. As we move through a planned series of studies, we move closer to addressing one of the most significant questions in falls research: can you identify someone at risk of falling before they have their first fall? If sensors can detect worsening postural stability (i.e. increasingly frequent near-misses), physiotherapists will be better informed and have the chance to intervene early when a patient is optimally placed to benefit.

Funding acknowledgement: This work is funded by the Engineering and Physical Sciences Research Council and is part of the SPHERE Interdisciplinary Research Collaboration.

This record has no associated files available for download.

Contributors

Author: V. Agarwal

Download statistics