Telecare Soapbox: Philips Lifeline AutoAlert – where’s the evidence?

Editor Steve frets about the lack of information on the accuracy of the Philips Lifeline AutoAlert and says it’s about time they published some data.

Back in March 2010, I welcomed the introduction of the Philips Lifeline with AutoAlert (for falls when the person cannot press the pendant button), particularly as a long-needed acknowledgement in the industry that traditional pendant alarms have considerable shortcomings. Philips puts nail in pendant alarm coffin. I also noted that, considering the notorious unreliability of fall detectors, Philips must be anticipating a large number of false positive alerts because there was a 30 second delay built in to enable users to cancel the autoalert.

In January this year Philips…

linked up in a big deal with WellAWARE Systems to monitor people in ‘senior living communities’ and ‘home health’ settings. Philips Lifeline, WellAWARE to co-market.

Now I see the following ringing endorsement by The Franciscan Companies, a member of the St. Joseph’s Hospital Health Network, complete with an oh-so-confident customer testimonial, on video, in a news item. Auto Alert system detects a fall without the press of a button. It’s convincing stuff.

Until you ask where the evidence is.

And is the equally robust statement on Philips’s design news blog to be taken to mean that there are never any false positives or false negatives? “It contains a series of sensors and an advanced algorithm that can accurately depict when the wearer has fallen, sending notification to the Lifeline service.”

If I were a customer who had a fall that the system failed to report I’d be using that statement in court as a stick to beat them with.

Let’s be generous and assume that Philips (or now even WellAWARE) has some research data on actual levels of reliability that it has shared with St. Joseph’s and other customers. But if it is so good why are they not sharing it with the public at large and the telecare/telehealth community in particular? I can’t find that data on any of their sites, and I’ve not heard of it being presented at any conferences…

Alternatively, if the data is not good, then the system and publicity falls into the realm of expensive but false reassurance.

Come on, Philips! If you want the industry to take the AutoAlert seriously you have to do better than just making bold assertions!

Categories: Soapbox.

Comments

  1. I am surprised that a company of Philips’ size and prestige has not garnered some objective outside research studies, from the equivalent of an ‘academic research council’ composed of major/name researchers, on Auto Alert. These typically have a relationship with the company to ensure access but at arms length. But to think of it, we see very little of this in eHealth, don’t we?

  2. Ethan Templeton

    I have severe doubts about fall detectors as you do. I would like to make a few statements on this particular fall detector. One, I personally spoke to a Philips Lifeline healthcare channel partner who told me about on of their first customers who fell using auto alert and it did not activate. Lifeline employees told them it was due to the women’s large breasts of all things and the way she fell. Secondly, They tell you to “press the button” that is on the fall detector when you fall just in case. Lastly and most shocking, I was told from a former wellcore employee that Philips licensed the core algorithms for their fall detector from wellcore, but Philips so tweaked the software that it would barely activate so the call center wouldn’t be flooded with false alarms. At the end of the day this product/service is a marketing tool for them to drive almost 30% more revenue per subscriber. I am glad you took time to call them out on this. I would have done so earlier, but I did not have a platform to do it.

  3. Kevin Doughty

    My students and I designed and trialled a number of different automatic fall detectors between 15 and 16 years ago. It took years of simulated falls, using a range of technologies, algorithms and device form factors onto surfaces from stairs through to thick pile carpets to enable us to come to the conclusion that no single approach would work reliably in all situations and for all potential wearers.

    Attempts to make a reliable and commercial device were subsequently aimed at getting the right balance between unregistered falls and false alarms. Our conclusion was that it was necessary to “fix” as many of the variables as possible. The easiest was the location of the device on the torso somewhere above the hips (waist through to chest) where the degrees of freedom would be small in number. The second was the type of fall – forward, backward, sideways, slip, trip or collapse. Reliability can be 99% or better for a single type of fall for a detector located at a specific point. False alarms are better than 5% under the same circumstances.

    Unfortunately, assessments are not yet sophisticated enough to separately calculate the risk of falling for each potential mode of fall. Consequently, manufacturers have had to compromise their designs to try to cover as many types of fall as possible, leading to inevitable detection failures and false alarms.

    This is the problem faced by Philips. Perhaps the solution is to have fall detectors with multiple detection hardware and software elements whose outputs combine with suitable logic to detect genuine falls with less uncertainty. The price is hardware redundancy, increased device size and greatly reduced battery life.
    We mustn’t forget that the greatest risks of a “long lie” are associated with falls that occur during the night, because the chances of being found and helped by a regular visitor or domiciliary carer are much less. It might therefore be reasonable to focus efforts on indirect fall detection methods including the use of bed exiting alarms with appropriate delays of 5 to 10 minutes to enable the user/patient to return to bed perhaps after using a commode or visiting the bathroom. There are over a dozen sensor arrangements of this type currently on the market and used by telecare and AT services. Their success depends on an assessment of both the individual’s needs, habits and preferences, the type of bed, and the availability of a carer or telecarer to respond.

    These issues are discussed fully in the next Telecare EPG Group Comparison Report which will be published at the end of August by T-Cubed and CUHTec at the University of York. Worn fall detectors will be reviewed in the same way in a few months.

  4. I have been struggling with the notion of the auto-alert efficacy. Reading the above well-articulated points, I wonder – what will make everyone happy? A false positive is better than no alert. With a traditional pendant, there is no alert if the wearer is disoriented or unconscious and unable to press the button. But at least with the current fall sensor technology, there’s a reasonable chance (I’ve heard 80%) that it will activate if a “hard fall” occurs. As long as wearers understand that the technology is not fool-proof, I say it’s better than a standard pendant.