New study from Toronto Rehab-KITE suggests falls can be predicted, even prevented

Dr. Andrea Iaboni, a clinician-researcher at KITE and Toronto Rehab, says researchers have developed algorithms that can predict those patients who are most at risk of falling. (Photo: UHN)

Falls are a major cause of hospitalization for elderly people across Canada.

But what if there is a way to predict who is most likely to fall next – and intervene before it happens?

That's precisely the question a team of scientists at The KITE Research Institute at Toronto Rehab set out to answer when they embarked on a project utilizing a vision-based computing system to measure the gait and balance of people living with dementia. 

"If you can detect those people who are most likely to fall imminently – for instance in the next month – then you can target resources towards those people in the form of additional supervision, rehabilitation or other assistance," said Dr. Andrea Iaboni, a geriatric psychiatrist and clinician-researcher with KITE and UHN.

"More specifically, we are interested in how gait and balance change over time. A sudden change could be an important marker for a change in health status or a marker of an imminent risk of falling."

The findings were published this month in the Journal of Gerontology in a paper titled "Vision-based assessment of gait features associated with falls in people with dementia."

In order to carry out the research work, the team installed a system called AMBIENT to Toronto Rehab's Specialized Dementia Unit, an 18-bed care inpatient unit that admits older adults with behavioural and psychological symptoms of dementia.

Researchers used a vision-based computing system to measure the gait and balance of people living with dementia. (Photo: iStock)

AMBIENT consists of a Kinect video camera mounted on the ceiling of a hallway that is activated by tiny radio frequency identification (RFID) tags that are ironed into the inside of the pants of research subjects.

"Once activated, the cameras record participants as they walk along the corridor capturing repeated bouts of natural real-life gait," said Dr. Iaboni. "From this the cameras can measure a number of features of gait, including dynamic stability, spatiotemporal measures, and gait symmetry."

Among the factors researchers studied, the most important is something called margin of stability, or how well people control their weight as they walk, and how close they get to the limits of their balance as they walk.

"The big takeaway is that we can teach a computer to observe and measure someone's pattern of walking," said Dr. Iaboni. "The information the computer gathers can be used to develop algorithms to predict those who are most at risk of falling and identify people who are at the limits of their balance."

The next steps for the research team are the deployment of AMBIENT on two floors on the UHN long-term care facility at Lakeside, and switching to the use of regular video images for measuring gait.

The long-term goal is to implement the system widely in residential care settings, by using streams from their video surveillance systems to provide information about the health status and safety of residents.

Funding for this research was provided by the Alzheimer's Association, Brain Canada, Natural Sciences and Engineering Research Council of Canada, FedDev Ontario, Department of Psychiatry, University of Toronto, the Canadian Institutes of Health Research, AGE-WELL Canada and The Walter and Maria Schroeder Institute for Brain Health and Recovery.