Counter to popular opinion among therapists, AI isn’t going to “taking your job,” at any time soon, although, sentient AI is likely coming down the pipeline in the next 30-50 years, maybe less, maybe more.
Predictions of these sorts of things are always to have a huge confidence interval.
In this article, I’m going to be sharing how various AI technologies will be augmenting and replacing portions of the rehab plan of care, treatment plans, and conventional physical therapy. Will artificial intelligence in physical therapy replace therapists? Let’s find out.
What will the future of rehab look like with AI-augmented therapy?
Imagine a future where an algorithm can take a thorough history for patients, put them through a detailed examination and develop a plan of care with the same level of effectiveness as a human therapist or, even better. Better patient outcomes for a lower cost would be great, right?
You might think that’s far off, and you’re probably right, however, it’ll be here sooner than you think. Deep learning (a subset of machine learning) has made huge strides over the past decade and it’s only to continue it’s forward progression. (unless of course we have nuclear war or something).
AI is already better at reading x-rays and other diagnostic imaging than radiologists and various types of surgeons might soon be replaced by fully autonomous robotic surgery units that make less errors and never fatigue.
I think it’s completely reasonable that therapists could create a decision matrix that AI is able to take a patient through to figure out what the best course of action would be for a given diagnosis, whether it be lower back pain, shoulder pain, knee pain, or something else.
The algorithm could give the patient exercises to work on and check in routinely with the patient to ensure that symptoms are not getting worse and that the patient is progressing as expected. If the patient is not progressing as expected, the AI in more basic stages could then triage to a human therapist who could further problem solve.
With more advanced AI technology, I predict that the AI, with the feedback of the patient will be able to automatically adjust the programming if the patient is not making the expected progress, without input from a human therapist.
Of course, there are simple presentations of musculoskeletal pain and injury and then there are more complex. First, the simple presentations like non-specific low back pain will be addressed by the AI and then as it gets better it will be able to help people with even the most complex of chronic back issues, with or without the assistance of a human.
A common rebuttal to this that I hear from therapists is that computers and AI will never be able to replicate our human emotions and the connection that we build between each other.
I disagree with this sentiment. If you look at some of the advancements in facial recognition and mirroring that robotics and AI is now capable of, you’d be shocked. There are engineers working on making facial expressions and voices more life-like with AI on robots and depending on who you’re talking to, this is terrifying or exciting.
I for one welcome our robot overlords…
Eventually, we will have sentient AI brains that are placed inside of robots that will have all the same facial expressions as us, and, it will likely be difficult to distinguish the difference between them and us, other than the massive intelligence gap.
No one can know for certain how far off this is, given we don’t understand what creates our own consciousness. Some will say that since we don’t understand it we can’t build it, but, like with most things, it will probably be a gradual process over time rather than an all-at-once discovery.
This means that our understanding of “artificial or man-made consciousness,” will likely evolve in tandem as it’s happening and after thousands and tens of thousands of iteration cycles.
What are the challenges that face the rehab community?
Cost/affordability. The cost of care for many people, particularly those who either don’t have insurance or are low-income, is not affordable.
Quality. The quality of care is not equal across the board and more often than not, people who have less money are relegated to working with clinicians and practitioners who see several patients per hour and are not able to offer the highest quality of care. (some diagnoses and people are appropriate to work in a group therapy setting, not everyone needs to be seen 1:1).
Access. Some of this is related to cost, but it’s also related to the fact, at least in the United States, that to see providers who do accept the insurance of low-income Americans, namely medcaid, there just aren’t that many providers. Many people who would benefit from physical therapy could just as easily be served by an algorithm that can guide and triage their care, overseen by human therapists.
Burnout and clinician quality of life. This is one that’s not addressed often enough when it comes to making decisions about how to improve healthcare. If we want patients to be healthy and have better outcomes it’s also equally important that we make sure, collectively, that our physical therapists and other rehab professionals are feeling a strong sense of fulfillment in their work.
Too many therapists are burnt out (as well as many physicians, nurses, occupational therapists, etc.) due to long hours, repetitive tasks (documentation), and unrealistic productivity expectations.
AI would help to be able to treat the easier cases so that our focus could be placed more-so on the complex cases that are more mentally stimulating and challenging, while reducing the overall burden on the system.
I think it’s a win-win all around.
The above issues are prevalent in all healthcare systems and are not unique to the United States. Some countries do better in some categories and worse in others.
An AI-augmented rehab program could likely address all three of the above problems by allowing all people to have access to high-quality diagnostics and plans of care while also being appropriately triaged to therapists and other practitioners as needed.
Human connection is important in the rehab process but I believe this problem can be solved, too. The social connection problem that arises from mainly AI augmented rehab is likely a problem that will take much longer to solve.
What companies are already doing this?
“Freeing the world from pain.” That’s a pretty good tagline, i’ve gotta say.
Upon reviewing SWORDs website the way they have their business setup is that you work with a virtual therapist to go through various exercises for the pain that you’re having. There are also educational materials that are provided at various points throughout the experience.
I’ve listened to the CEO speak about what exactly is included but it was more of a brief overview. I am quite interested to see what type of artificial intelligence is utilized.
It appears that there is a wearable device that also shows range of motion changes while one is doing the exercises. I’m really not sure how helpful this is to make improvements. It seems mostly gimmicky however they could also find that it helps with patient compliance of exercise.
There actually doesn’t seem to be very much use of AI (from what I can see on the website), so it will be interesting to see if they choose to implement this down the road. I’ll be reaching out to the company and updating this article as I gather more information.
According to Crunchbase, they’ve already raised $323 million.
Out of all the companies working on solving the persistent and chronic pain problem, my money is on SWORD. Their team seems to be well diversified in terms of actual clinical employees and business development. I believe that this is important when it comes to growing a larger healthcare organization.
Kaia + UnitedHealthcare
“Proven MSK care that combines human care with superior technology to achieve better outcomes.”
I like the democratizing healthcare bit, but the subheading is a bit underwhelming. SWORDs got them beat, here.
Kaia is similar to SWORD in certain aspects. Let’s start with how they’re similar. They both track how the client is moving via technology. SWORD uses wearable technology and Kaia uses your phone camera.
Both companies utilized therapists to guide the client and update exercise programs to make sure that appropriate progress is being made.
Both are targeted to the B2b (business to business) market, meaning you can only use it if your employer or health plan covers it.
Here’s how they are different. SWORD only uses physical therapists (to my knowledge), while Kaia leverages PTs and health coaches to help clients make progress.
Currently, Kaia is only available to people who have United Healthcare whereas SWORD is available across various employee health plans at larger companies.
“Exercise Therapy Delivered Digitally”
“The #1 back and joint pain clinic.” “We go beyond digital PT to combine physicians, PTs, health coaches, and motion sensor technology to deliver the right care at the right time.”
Hinge health is currently the largest and earliest player in the digital health and physical therapy space. They are currently only B2b and work with, according to their website, 750+ companies to deliver this service.
The Hinge Health system has a prevention, acute, and chronic program that utilizes an app based framework to help people reduce their musculoskeletal pain.
This program uses physical therapists, physicians, health coaches, and motion sensor technology, which is again very similar to Kaia and SWORD health.
Each of these businesses appears to be more or less the same company in terms of their offering and they are approaching the same problem.
SWORD is the only company claiming to be using AI, however i’m still unclear as to how that is being implemented. As I said before, I’ll update this article when I have a conversation with some of the employees from each company.
What is the problem with most companies that are implementing this technology?
The greatest error that I see with companies that have developed initial versions of this technology is that they focus too much on “how” exercises are being performed and are not tracking, to what I’ve seen, how symptoms are progressing one hour, and 24 hours after an exercise session.
Instead of focusing on form, these companies should be developing technology that instead focuses on progressing clients and helping them navigate exercise as their symptoms permit. This is a more complicated notion (seemingly) because this requires appropriate input from the clients.
I don’t understand machine learning enough at this point to know the most efficient way to make this happen. (if you have some resources in mind for me to go through, i’m all ears).
Another problem is that there are many people who are not tech savvy, and would likely not benefit from this sort of intervention. In this case, I think the traditional model of rehab services needs to continue to exist.
Generally, as we get older, we are less likely to adopt new technology. This may change with millenials, Gen Z, and whatever generation is after Gen Z as we grew up with tech.
How these companies could be improved to serve clients more effectively
Here’s a list of what I think is missing from these companies:
- Triaging of patients to in-person therapists who have been vetted for the clinical competence
- Focusing on symptom changes and fitness improvements vs. wearable devices or “how” an exercise is being performed.
- Expanding to the B2c and small business market
What do you think?
Do you think AI will eventually replace or significantly augment the musculoskeletal rehab profession? Why or why not?