I’ve been wanting to write this article for a long time. I’ve been working in geropsychology and the long-term care (LTC) industry for the last 14+ years. I’ve worked mostly in the area of skilled nursing (what most people know of as nursing homes), as a consultant and staff member. Skilled nursing facilities are where people, typically OA (OA), live and receive medically necessary professional services from nurses and other allied health professionals to provide them assistance with activities of daily living.
Over this time, I’ve learned a few things. Skilled nursing care, at least as it’s practiced in the United States (but this is largely true across the industrialized world) is highly regulated, very labor intensive, and because of these two reasons – is very expensive. Caring for medically complex and frail, chronically ill, primarily older adults (OA) involves a lot of monitoring and supervision, along with the physically demanding tasks of caring for their bodily needs.
Back to the issue of cost – according to recent statistics, the average monthly cost of a stay in long-term care in the United States is over eight thousand dollars per month ($8,121 to be exact). While that number is obviously high, it’s sobering to recognize that the cost is 13% higher than it was just 5 years ago (citation here).
Setting the stage – Recapping the “Demographic Tsunami”
And these trends aren’t going to be slowing down anytime soon. As I’ve written here, and here, and a number of other places, the US and the rest of the industrialized world is currently poised to be swept away by a “demographic tsunami.” This refers to the fact that from the year 2000 to 2050, the world’s population who is 60 years or older will approximately double from about 12% to 22%. The number of “oldest old” (those over age 80), will quadruple.
Moreover, there’s strong data suggesting that there are, and will continue to be growing and severe shortage of healthcare professionals out there to meet the needs of OA (both in long term care and otherwise). In my own particular field, geropsychology, the shortages are already severe and are projected to continue to grow. The story is not much different for the fields of geriatric medicine or geriatric nursing – even the front-lines of LTC, the people who do the real demanding, physically laborious work of caring for elders, the nursing assistants – even this field is experiencing growing shortages.
There’s a number of reasons why this is. One, reimbursement. Geriatric medicine is a field with one of the lowest reimbursement rates of any medical specialty, so, in response, it’s a very poorly sought-out specialty for the hordes of newly-minted physicians who want nothing more than to pay off their increasingly-herculean student loans and to start providing a good living to their families (and who can fault them for that?). Geriatric nursing isn’t much different, and in the case of nursing assistants, hourly wages of 10$ per hour in very demanding conditions provides a poor retention incentive.
So why is this such a problem?
So far, it’s fair to say that the rapidly growing cost of nursing home care and elder homecare and facility care in the United States and industrialized countries is not stopping. While our current generation of 65-year-olds and older are (arguably) the healthiest they’ve been in a long time, the fact is – when people get old, they are at far greater risk of developing a whole host of problems that often require significant and at times, round-the-clock care. For example:
- Dementia – I never fail to mention this one. The #1 risk factor for developing dementia (such as Alzheimer’s disease), is advanced age – and, dementia is endemic in LTC facilities (at the VA nursing home I am employed at in my day job, around 70% of my population have dementia, which is roughly in line with US averages).
- Falls and ambulation problems – chronically ill OA are more likely to lose the ability to walk as they get older. Musculoskeletal issues (like degenerative joint disease), deconditioning and muscle wasting, dementia, and other issues can often render OA wheelchair bound or worse. Sometimes they are unable to transfer from wheelchair to bed or toilet without assistance, or require lifts to be moved. Sometimes they are even unable to turn themselves in bed. Also – closely related to the issue of ambulation is falls – older people are much more prone to injuring themselves during a fall, and of falling more frequently. This is due to issues like cognitive impairment and poor judgment (due to the aforementioned issues with dementia).
- Incontinence – another major issue is lack of control of one’s bowels or bladder. Again, medical conditions like dementia, spinal cord injury, or other neurological problems can put chronically ill OA at risk for this. Obviously, incontinence, when combined with other issues, can require care as incontinence briefs need to be regularly changed so as to prevent other issues.
- Difficult-to-heal, or nonhealing wounds – often the above issues of problems with ambulation, incontinence, or frequent falls often put chronically ill OA at risk for developing wounds that are often very difficult to heal. Often this is because they are less likely to move when they are in bed or in a chair (leading to pressure sores), or they can bang or scrape themselves during a fall. Due to thin skin and reduced ability to heal (often because of poor circulation, diabetes, etc.), their wounds take a very long time to heal, and without constant care, can at times become infected.
And this isn’t even scratching the surface. As you can see, chronically ill OA require significant amounts of monitoring and care by professionals in order to just exist, and without it, they can rapidly become acutely ill and require much more expensive care (such as in an emergency room).
So what’s the solution?
There’s been much discussion about solutions over the years, from making OA healthier, to strengthening home and family caregiving options, to training more doctors, nurses, and psychologists to help OA living in nursing homes, as well as improving existing models of care.
It’s also worth mentioning that with all the hullabaloo of the so-called “Affordable Care Act,” AKA Obamacare, there has been virtually no attention paid to reforming the broken state of LTC funding in the United States (I’ll quickly get off that soapbox!).
So what if there was some other solution? We all know, as is was said here by venture capital investor Shourjya Sanyai, that the “rapid(ly) aging demographic will directly affect social, economic and health outcomes for these growing economies. Particularly healthcare delivery pathways need to be readjusted, keeping in mind the prevalence of chronic diseases, comorbidities and polypharmacy requirements of the elderly and geriatric patients.”
Sanyai goes on: “Given the situation, healthcare providers are starting to offload certain parts of the care-pathways to artificial intelligence (AI) based automatization. AI can now be found in every step of the care-pathway, starting from intelligent tracking of biometric information to early diagnosis of diseases.”
So what is AI?
Artificial Intelligence and Machine Learning – Definitions, and the Example of Alexa
The definition of artificial intelligence, as found via Google: “the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.” Note that artificial intelligence (or AI, as it’s frequently called) is frequently combined with smart machine learning algorithms – which thereby assure a system that can do specialized cognitive tasks and also learn from experience as it does it’s work.
Moreover, one of the nice things about today is that AI and machine learning is a concept that’s rapidly seeping into the consciousness of consumers worldwide:
Figure 1. Amazon Echo (http://www.bestaiassistant.com/google-home/amazon-echo-vs-google).
Amazon probably has done more to make the concept of AI (and it’s capabilities) obvious to the masses with the introduction of the Amazon Echo device, which is the most well-known vehicle for its so-called Alexa AI technology. What does the Echo do? One of the things we like most about the Echo in our family (we have several) is that it takes rote tasks and automates them.
For example, we have most of the lightbulbs in our house now controlled by Alexa – instead of having to get up and flip a switch, we say “Echo, turn off bedroom lights,” etc. We’ve also hooked it to our music streaming service, so if we want to hear a song, we say “Echo, play (insert favorite song),” and off it goes – this is as opposed to fiddling with our phones, or a CD player, or whatnot.
There’s more – Alexa will now be doing double-duty as a burglar alarm! One of the newer “skills” that Alexa has been enabled with is something called “Guard Mode” – whereby if a user leaves their home, they can say “Alexa, I’m leaving,” and Alexa will listen for the sound of glass breaking and alert the user. So – Alexa is also a smart monitoring system – while you can still purchase an analog burglar alarm system monitored by “24/7 security personnel” (which is expensive and requires people on duty to constantly monitor your home), you don’t have to – because AI (in the form of Alexa), will do it for you.
AI and Machine Learning in Healthcare
So let’s go full-circle back to LTC. Why would be interested in it?
First, remember all the examples I listed above, regarding the kinds of problems regularly addressed by nursing staff in LTC facilities (e.g., dementia, falls, wounds, incontinence, ambulation and movement issues). What they all have in common is that in order to address them in the LTC environment, they require a significant amount of personnel to perform rote tasks relating to monitoring and rounding. I would suggest that a very significant number of these tasks currently performed by nursing staff could be offloaded to so-called “smart systems.”
Smart Device Assisted Living and Monitoring. What happens when an older adult with dementia tries to leave their nursing home (trying to “go home”)? The current standard approach to “wandering” is to attach “wander guards” to residents at risk, which, when residents cross a perimeter, will alert staff and allow them to redirect them back into the facility. The downside of this is that often these alarms don’t help to locate residents (it just alerts them that the perimeter has been breached), nor does it distinguish between those who legitimately are trying to escape, versus those who merely accidentally trip the alarm when they are merely, say, rolling outside to get a breath of fresh air.
Or how about when an older adult is bedbound, is at high risk for nonhealing wounds, but due to neurological impairment fails to turn themselves? The current standard of care is for nursing staff to regularly “turn” residents (say, on an hourly or per-shift basis). However, this requires that nursing staff, who are often busy, overworked, tired (and human) to remember to remember – moreover, it’s often possible that these residents may be still turning on their own and don’t need this extra intervention.
AI is tailor-made for the situation I sketched out above, in the case of “turning” residents. You can apply a wrist-mounted (or bed-mounted) sensor to a resident and monitor their movements. If a resident has not turned after a certain amount of time, you can have the AI system alert staff proactively, essentially telling them, “hey, Mr. Jones hasn’t turned in a while – can you go help him?” This prevents staff from having to round unnecessarily on residents who are not at risk for wounds and who are turning in bed, and also relieves nursing staff from having to “remember to remember.” One of the companies mentioned in this article are already developing an AI system to address this very issue.
Wrist-actuated actigraphy (such as what you see with Fitbits, and Apple Watches), combined with AI, is also potentially ideal for replacing the old “wanderguard” system – I am familiar with a company called Carepredict, who is essentially doing just this – they have facility residents wear their own proprietary wristbands which detect a resident’s movements within the unit (as well as level of activity). The system is designed to provide “early warnings” to staff when a resident’s behaviors deviate from their established norms – and can precisely locate a resident when they are trying to escape (this is also something AICare is also trying to do).
Not only that, Carepredict claims they can provide “early warning” for staff to let them know if a resident is becoming depressed (say, if they begin to isolate in their rooms when their previous pattern was to be out and about regularly), or if they have stopped eating.
Fall Detection and Prevention. How are falls currently addressed by nursing staff?
Figure 2. Bed / chair alarm pressure pad (courtesy of Alimed, Inc).
Currently, it’s the ol’ analog pressure-pad system. In other words, residents identified as being high fall risk are issued pressure pads placed on their beds or chairs, and if a resident gets up from their bed or chair, the alarm sets off a loud racket, and nursing staff come running. The downside of the current system are manifold – one, it has a significant number of false positives – residents who merely move in their chair or bed (something we *want* them to do, actually) set the alarms off. Second, the noise is annoying to residents, and for those with dementia, can serve to agitate them further – thereby inadvertently raising their fall risk. Third – it leads to “alarm fatigue” in staff (due to the frequent false positives) – staff sometimes don’t respond to the alarms because they know they are often wrong. Despite all of this, residents continue to fall at high rates, staff often find a resident on the floor and are left to question these frequently-memory-impaired residents and otherwise piece together what happened, and then institute fall prevention measures after the fact.
Enter Safely-You, a company I’ve been very excited about (although note they’re not the only market participants in this space). Instead of pressure pads, they offer a camera placed in a resident’s room, typically at bedside (since this is where most falls occur) and the camera then continuously monitors the resident, constantly capturing video of the resident.
However, there is never more than 10 minutes of video saved in the systems’ buffer at any one time, and video is only ever permanently saved if a fall is detected. The AI and machine learning built into the system detects falls at apparently a 94% level of accuracy, and immediately alert staff when a fall occurs. Staff are immediately able to review the video and institute fall prevention recommendations based on exactly what they see the resident do (as opposed to what they imagine happened).
Virtual Companions. This is a subject near and dear to my heart (see here, here, and here). Let’s go back to the example of Amazon’s Alexa (and it’s various competitors – like the Google Home or the Apple Homepod) – these digital assistants are useful, but they aren’t exactly companionable – more just disembodied and mildly robotic voices that do what you tell them (although Alexa can tell jokes, or sing songs for you if you ask it).
One of the other “rote” tasks in nursing (hate to put it that way) revolves around the insubstantial yet extremely important task of providing companionship to residents. The hug, the touch on the shoulder, the listening intently to the older adult as they tell a story – these are all vital to the health of OA but due to the abovementioned issue of staffing and sky-high demand for long-term care services, nursing staff are much less able to provide this service to their clients.
So, how about this?
Figure 3. Paro robot doing its thing. Courtesy of the Toronto Star.
Above is the Paro robot – a robotic companion that uses its built-in machine learning algorithms to learn the name users give it, and to respond preferentially to being stroked, and to avoid being hit or dropped. Moreover, it’s adorable – and research tends to suggest that it delivers beneficial affects to the users (which includes calming dementia patients), by stimulating oxytocin production. Oxytocin, of course is the feel-good chemical that parents get when bonding to their children or when new mothers first nurse their babies.
Other Applications for AI and Machine Learning in Facility and Home Care?
This kind of technology has applications that literally are only limited by imagination and a few smart programmers. A worthy mention is the company Winterlight Labs, which has a proprietary assessment tool which claims to assess for the presence of dementia via speech-sample assessment of patients to a degree heretofore impossible using standard, human-administered cognitive tools. This kind of innovation has the potential to put geriatric neuropsychologists out of business!! (Well – maybe not quite yet).
Also an honorable mention goes to CareAngel – they have a system whereby the digital assistant (like Alexa) calls OA and simply asks them how they are doing (they call these “care touches,”) and then has an actual conversation with them. For example, if the older adult says “terrible, I’m in a lot of pain,” then the system asks them additional questions (like what level their pain is, where it’s located, etc). and then depending on their answers, summons a live care provider.
The bottom line is that AI and machine learning are poised to revolutionize the care of OA both within and outside the LTC industry. This revolution will result in a lowering of costs, mostly in the form of less staff required for routine, rote monitoring and rounding of residents, but also – in the form of less costly trips to the emergency room or ICU due to real or even misclassified falls, as well as infections and injuries. It may even result in less need for humans to provide companionship to residents, as we might be able to offload some of that work to social robots and digital companions (as creepy, and potentially ethically questionable as that may be to some).
Nursing homes and OA care are going to see skyrocketing demand over the coming years. In order for our nation to not get completely swamped by the sheer weight of the cost and labor of caring for our most needy and vulnerable citizens, we’ll need to find ways to innovate our way out of this. The AI and machine learning revolution may in fact help us to do just that!