Young man and senior woman sit on couch discussing

Catching Seniors Before They Fall: Part II

High-tech Fall-Risk Metrics and Technologies

We often think of looking into the future like taking a peek at Christmas gifts before the big day – sometimes it’s just better not to know. When it comes to senior care, however, a crystal ball could come in handy, helping to prevent future falls. In our last blog, we extolled the virtues of high-tech functional performance testing, but did not have space to explore it in detail. In a nutshell, the existing approaches come from academic research articles that collect data using various sensors while individuals perform some functional task and compute one or more fall-risk metrics. Let’s dive in.

The Sensors

Motion capture, accelerometers/IMUs, and instrumented flooring (force plates, sensor mats) are common means of collecting biomechanical information for an individual. In a formal gait lab, motion capture involves the application of markers to key anatomical locations. These key locations are tracked over a recording and a skeletal model is fit to the data, allowing calculation of joint angles, angular accelerations, etc.

While the technology can accurately capture full-body movement, the markers make it quite obtrusive. Some marker-less motion capture systems are being used but are currently less trusted in research because of their reduced positional accuracy.

Photo Credit: https://www.razer.com/eu-en/gear-accessories/razer-x-fossil-gen-6-smartwatch/FTW4065SET

Accelerometers and IMUs (inertial measurement units) are compact wearable devices (e.g. with a wrist or waist band) that measure movement acceleration (and other quantities) in up to 3 axes. Acceleration data can be used to infer velocity and even position, relative to some starting point, and so can be used to determine parameters of postural sway and gait. Multiple accelerometers can also be worn to capture movement at key locations, if desired.

These sensors are relatively lightweight but still generally have to be worn or applied. Also, to make proper use of the 3 axes of data, they should be positioned carefully and consistently when put on. Smartphones regularly include an accelerometer, encouraging the prospect of fall risk monitoring. However, where and how the phones are carried puts some limits on the usefulness of this data.

In contrast, floor-based sensors are completely unobtrusive though not as portable. For example, force plates are typically embedded in the floor. Such devices record data from only a few sensors but at a high sampling rate and are good at providing postural sway-related information, which is based on total movement rather than fine underfoot details. For studying gait, however, force plates are not well suited.

Pressure-sensitive mats can provide a larger floor surface than the force plate and capture a video of your footprints at high resolution (as fine as 5 mm between sensors) as you walk across them. Some mats have simple arrays of binary switches (pressure vs. no pressure) and some mats actually measure the pressure. Pressure-measuring mats permit detection of high-pressure anomalies and other plantar pressure events.

Spatial and temporal parameters of gait (e.g., gait speed, cadence, stride length, stance time) can be captured by both types of mat. The pressure-measuring mats, however, can also be used for postural sway analysis because, like the force plate, they measure the vertical force/pressure of the individual over time.  A twist on this technology that has been used in fall-risk studies is the pressure-sensitive insole, which is more portable and wearable like an accelerometer. Estimating between-foot spatial gait parameters (e.g., step length), however, cannot be achieved by an insole alone. Other accompanying technology is needed to add this.

The Functional Tasks

While sensors record their movements, individuals are instructed to perform one or more tasks. Common fall-risk functional performance tasks include walking on a treadmill or on the ground, standing balance for capturing postural sway, the Timed-Up-and-Go (TUG) test, and the Five-Sit-to-Stand test (5STS). The TUG test was described in our last blog. The 5STS test is aptly named, where the time taken to complete standing 5 times from a sitting position as quickly as possible is used to evaluate fall risk.

Temporal breakdown of the gait cycle including double support and single support

The Fall-Risk Metrics

Commonly evaluated measures of fall-risk while walking are spatial and temporal parameters of gait. Of these, gait speed is of particular interest because of its high correlation with falls. Slower gait corresponds to an increase in fall risk and higher than average speed corresponds to decreased risk. This is probably a significant part of the TUG test’s success as a fall-risk tool (see below). Increased variability of certain gait parameters has also been found to correspond to increased fall risk. It seems reasonable that inconsistency or irregularity in the timing and placement of footsteps in our walking pattern will eventually lead to a fall. Other common walking-related fall risk metrics include harmonic ratio (a measure of gait smoothness/stability, an accelerometer-based metric) and gait symmetry.

Postural sway determined from standing balance indicates the level of stability of an individual. A high level of sway is associated with fall-risk. Another interesting high-risk indicator is when the degree of sway is almost as poor with eyes closed as with eyes open.

Collage of Stepscan system arranged in a T, balance assessment foot pressure maps, and a Fall Risk Assessment Summary report.

The key metric for the TUG test is the time, where individuals whose time exceeds a certain threshold is considered a high fall risk. Slow walkers exceed the threshold. In some studies, even the subtasks of the test are timed and evaluated for fall-risk. Other TUG-based metrics include spatial and temporal gait parameters and the number of footsteps made per trial.

High peak underfoot pressures and sustained footstep pressures for longer times (pressure-time integrals) have been associated with somewhat increased fall-risk as well. High underfoot pressures are associated with foot pain, and irregular stepping patterns can be expected from people who are accommodating a painful foot.

Screenshot of KPIs on Fall Risk Report Summary

Putting It Into Practice

The real question is, “What does a clinician need from a high-tech fall-risk tool in practice?”

Most of the papers try to predict whether or not an individual is expected to fall in the next year based on their data to compute the sensitivity/specificity of their approach. Is such a yes/no answer the desired indicator? Some papers include a risk ratio describing the increased/decreased risk of a certain fall-risk metric. Is that more helpful? Another approach used by some is to estimate the overall probability (i.e., 0% to 100%) of an individual falling in the next year. Such a number could be used to rate an individual before and after certain interventions and serve as a consistent measure to group those at highest risk of falling in a facility for added care.

This is the primary indicator we have implemented at Stepscan. In the upcoming 2.5 release of our software, we have introduced a Fall-risk assessment tool based on gait and on standing balance that looks at 5 different categories of fall-risk metrics (gait speed, gait variability, gait rhythm, gait pressure, and postural sway) based on a dozen specific fall-risk metrics spanning six different studies. For each category, we calculate and display a risk ratio, and these are combined to provide a single fall risk estimate.

But enough about us. What reasons would you have for choosing one high-tech fall-risk assessment tool over another? Ease of use? Cost? Integration with your electronic health records database? Are there any make-or-break features for you? Please leave your thoughts in the comments section below.

About Patrick Connor

Patrick holds a bachelor’s in computer engineering and master’s in electrical engineering from the University of New Brunswick and a Ph.D. in Computer Science from Dalhousie University. His research interests include gait biometrics, pattern recognition, computational linguistics and computational neuroscience. Prior to doctoral studies, Patrick worked as a software developer for 5 years in private industry. After his studies, Patrick joined Stepscan Technologies as a postdoctoral fellow to investigate the use of the Stepscan underfoot pressure sensing technology for security applications. He currently serves as Research and Development Lead with the company.

Leave a Comment