The goal of falls prevention technology is not to restrict patient movement, but to avoid the negative consequences of a fall. Can Smart Care Technology actually predict falls, or only detect falls?
In the field of cardiology, the pacemaker has revolutionised the treatment of arrhythmias over the past 50 years , by providing a means of maintaining the heart in a normal sinus rhythm. Is there a simple robust solution for the problem of falls?
Any products used to prevent falls must be considered within the context of the individual patient’s needs, and the technology must be easy for staff to use. If technology is cumbersome to use, its effectiveness is reduced.
The technologies might be easy-to-use, low cost and meet electrical & cyber safety guidelines, but in terms of functionality the technology readiness levels (TRL), based on a scale from 1 to 9, are beside location systems still far removed from a 9 (being the most mature technology).
It must also be the right technology, and appropriate selection requires understanding the various settings, the best fall predictors and the current technology options.
Let’s have a closer look at the effect of technology settings, the top 5 predictors of falls and last but not least at the different options of fall-related technologies including some pros and cons.
Precision vs. sensitivity setting
In a study by Hoke and Zekany1, of 67 falls that occurred in a progressive cardiac care unit over 2 years, only one patient who fell had an activated bed alarm.2 Because that alarm was set at its lowest setting, it would be activated only when the patient was completely out of bed. The authors of the study speculate that a more sensitive setting may have prevented the fall. But is that true?
Effective use of technology is a balance between sensitivity and precision settings.
Sensitivity is a measure of how well it can identify true positives and precision is a measure of how well it can identify true negatives. There is usually a trade-off between sensitivity and precision, such that higher sensitivities will mean lower precision and vice versa. In this setting:
True positive: Fallen people correctly identified as fallen
False positive: Not fallen people incorrectly identified as fallen
True negative: Fallen people correctly identified as not fallen
False negative: Fallen people incorrectly identified as not fallen
The challenge of fall detection with sensors is finding the right balance between sensitivity and precision (or Specificity), and getting close to 100% precision to avoid “cry wolf” phenomena.
Every 24 hours, the system classifies 17.280 five second periods for fall versus no fall
Precision of 99.994% (which is pretty high) in fall detection still leads to 1 false positive/day!
To threshold the precision to 100% (to avoid false positives) sensitivity will drop to 82% (so not very sensitive), and a high number of false negatives
Is it better to have high sensitivity setting with false positives and hence a “better be safe than sorry” approach but with the risk of alert-fatigue, OR better to have a high precision setting with false negatives and miss 1 out of 5 falls…?
Top 5 Predictors of Falls
The Top 5 predictors of falls, of which the first 2 dominate (by far) the risk prediction 3:
Fall history
Unsteady gait
Antipsychotic
Antidepressant
ADL impairment
Fall history
Monitoring fall detection of residents with a fall history could increase the reaction speed of carers after an actual fall has occurred. Automated tracking data about the location, event and/or time registration of the fall could provide valuable information to work on a resident’s personalised risk avoidance plan. Fall history is an obvious indicator but It's easy to be wise after the event.
Gait
Slowing of gait (a person's manner of walking) by 5cm/s over one week is associated with an 86% probability of falling during the next 3 weeks, 4x more than someone who's walking speed hadn’t changed. A decrease in stride length of 7.6cm was associated with 50% probability for falls4.
This means gait reduction is one of the best indicators of falls!
Incorrect weight shifting as main root cause for imbalance
In your effort to predict falls you need to include the root cause of all falls: imbalance; and the activities performed while falling.
Remarkably, 47% of all falls, ~ 1 out of 2 falls, have as a root cause of imbalance incorrect weight shifting (vs eg. 14% of all falls caused by trip/stumble)5.
To furthermore understand the top 3 activities when falling (while incorrectly shifting their weight) people were:
Walking (37%)
Standing (24%)
Transferring from standing (18%)
Knowing that the majority of all people falling are on the move, fall while incorrectly shifting their weight, and that the best leading predictor of falls is a reduction of gait/motion, alerting against a reduction of motion over time could be one of the best solutions for fall prediction.
The types of Fall-related Technologies
A common method of fall detection technology is the use of sensors. There are 7 types of sensors in 2 main categories:
Wearable sensors
Inertial/acceleration sensors
Location systems
Smart ambient sensors
Pressure sensors
Sound sensors
Infrared/motion sensors
Vision sensors
Radar sensors
Wearables
Inertial sensors work based on an accelerometer which is quite enough for human fall detection as sufficient information could be extracted from its measurements.
Locating systems (RTLS) are used to automatically identify and track the location of people (or objects) in real time, usually within a building. Wireless tags are worn by people, and communicate wirelessly with devices in the ceiling from to determine their location and movements.
Ambient
Pressure sensors may be integrated into pads for chairs or toilets. When patients’ movements indicate they are getting near the edge, an alarm sounds so staff can respond and help patients exit safely.
Pressure and sound sensors also can be embedded into devices. For example some bed mat or floor sensors are still used in some facilities but most beds now have sensors incorporated into them. Alternatively, innovative bed sensors can be placed under the leg/wheel of a bed and provide a range of useful information.
Some devices contain an infrared beam to detect a visit to a room or a single movement.2 In combination with a rule base engine logical alerts can be raised, like eg. no movements in the bedroom, or a long time spent in the restroom.
The latest sensor technology is radar; passive, touchless monitoring enabled by 4D imaging. This leading-edge technology offers a balance between the visibility care providers need and the privacy that care users demand. It detects falls in any lighting conditions, dense steam, and even through materials such as shower curtains, making it ideal for use in bathrooms.
Barriers:
Challenge of accurate fall detection
Wearable sensors: form factor, battery life, attitude towards use/wearing them
Ambient sensors: need of fall to be in view of sensor, multiple sensors per room
Resident motivational factors to use fall prevention technology:
Control
Independence
Privacy
Self-efficacy
Perceived need for safety
Usability
Cost
Wearable vs. Ambient sensors
In a contact-based sensor system (wearable), residents must wear a tag/device. The main advantage of a wearable is their suitability for activities and being on the move. However, wearing sensors might be impractical, the position of a wearable sensor affects the accuracy, and the battery life of wearable devices is a challenge.
Instead, a non-contact system (based on an ambient sensor), resolves some wearable issues, but is subject to a lower acceptability, especially by the elderly, who see their privacy violated, and it is limited in range to the room it is located in.
Wired or wireless
Devices that use sensors can be wired or wireless. Wireless in terms of detecting and signaling the data back to base. Wireless systems provide two significant benefits: (1) reduced tripping hazards for residents and staff and (2) greater ease and efficiency in moving residents because the alarm and the sensor do not have to be in close physical proximity.
Mains powered vs. Battery powered
Devices are only truly wireless when they are solely battery powered. The benefit is that the device still works when mains power is not available. However, batteries need to be (1) constantly monitored to avoid running low/empty, (2) require labor to physically replace them, and (3) are harmful to the environment when improperly disposed of. Alternatives are “wireless devices” that are powered from mains (although now also “wired”), or a combination thereof, with the battery functioning as backup for power outages.
To reduce the risk of missing actual falls (aka true negatives) a combination of fall-related technologies should be considered.
Best fall predictor
Even though it is the 2nd best fall prediction method (besides fall history), it is not easy to continuously monitor gait. Instead, it would help to collect data of residents’ movements (or alternatively their step count) as a leading indicator for falls!
Knowing that the majority of people falling are on the move, and the best leading predictor of falls is a reduction of gait/motion, therefore alerting against a reduction of motion over time is one of the best solutions for fall prediction, using the currently available fall related technologies.
For more information
Contact us to assist you in evaluating your fall prevention requirements to ensure that you have the appropriate systems in place to avoid falls and major injuries, improve your quality measures and hence your star rating, and more!
Please feel free to leave a comment below...
Sources:
1) Hoke LM, Zekany RT. Two sides to every fall: Patient and nurse perspectives. Crit Care Nurs. 2020;40(6):33-41. doi:10.4037/ccn2020289
2) Oh-Park M, Doan T, Dohle C, Vermiglio-Kohn V, Abdou A. Technology utilization in fall prevention. Am J Phys Med Rehabil. 2021;100:92-99. doi:10.1097/ PHM.0000000000001554
3) Stephen N. Robinovitch, Ph.D., Natalie Shishov, PT, M.ScKaram Elabd B.App.Sc. Preventing and predicting falls and injuries in older adults with technology. 2019.
4) Philips et al., Western J Nursies Res, 2017
5) Robinovoitchet al., Lancet , 2013: Yang et al., JAMDA , 2017