January 21, 2025 • 34 mins
Chewing It Over Is Back!
In this episode of "Chewing It Over," Jack Chew delves into the integration of artificial intelligence (AI) in musculoskeletal (MSK) healthcare. He explores how AI technologies are transforming diagnostics, treatment planning, and patient outcomes in the MSK field. He goes on to think about current applications, potential benefits, and the challenges faced in adopting AI within MSK practice.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Are you fed up with printing outexercise programs or horror
drawing them? Solve every exercise
prescription issue you can thinkof using Rehab My Patient.
Thousands of pictures, videos ofevery movement you can dream up.
Send by e-mail or WhatsApp. Translate into different
languages at the click of a button.
But don't take our word for it. Sign up for a three month free
trial now. Just go to
(00:21):
rehabmypatient.com/physio Matters.
OK, hello folks. How are you?
(00:43):
Hope this is working. If not I'll learn from it.
It's been a while. We've not gone live for a bit.
Obviously never stopped podcasting, could never stop
doing that. However, I think if I know where
the buttons are, then I've I've got things working.
There's a few new new little features that I can play around
with and so I'm just trying to work that out.
(01:04):
So if you are tuning in live andif this is working and if you
can hear me, then please post inthe comments.
It'd be really nice to hear fromyou because we're deciding to
bring Chewing it over back as a live stream and I'll explain in
a moment why that even stopped. We've obviously still been
podcasting. We've been doing it as a pre
produced and slightly edited Let's go.
Where's the captions? Had that little ticker thing for
(01:28):
a second. Good.
I think I'm finding my way around this place again.
Unfortunately, there's a retrobeat called Jack March
that's clearly tuned in. He said hello to me on YouTube
of all places. Now, if any any anyone more
important than him is tuning in and can hear me OK on the
various different locations of which I'm streaming this, then
(01:50):
please let me know because it'd be useful for me to find that
out and to make sure it's working because we've turned all
the buttons back on and it's been a little while since we've
we've gone live. I can see in the background tidy
on the plinth there, isn't it? But yeah, bit of teething
problems. But yeah, lovely to be back with
you. And I'll explain a little bit
about what we're going to do with this show for 2025 each
(02:11):
Wednesday, as you can see, 12:30to 1 O clock ish.
So lunchtime, Wednesday lunchtime, I'm going to stream.
And we stopped doing tune it over as a live show reluctantly
to some degree, just because of the stability of the Internet
and the product that we were using needed a stable Internet
that we just couldn't always guarantee.
And that was really frustrating.And so we've moved it back to a
(02:33):
weekly show on Zoom that we thenproduced and put through the
feeds. There were some people that
found that frustrating that it wasn't on their preferred
location. There were some people that only
use LinkedIn, for example, and found that to be where they
found me. And so Aston hoped that we'd
bring that back or post it on there.
Once you had failed to stream itto all those different
(02:55):
locations, manually uploading itto each individual one was, was
a bit of a pain. So we ended up just being on the
podcast feed and on on YouTube. And so now hopefully we can cast
this then again to Facebook and Twitter and, and, and YouTube
and LinkedIn. And hopefully that's useful for
you. And so you can then engage with
it as live or after the fact. We'll also post that especially
(03:16):
successful ones that aren't glitchy and stuff.
We'll make sure it's stable. We'll post those onto onto
Spotify and Apple Podcasts and all that sort of gobbins as
well. But generally we're just going
to test the water and do a few Wednesdays here in December,
sorry in January. And we can do it that way.
But it really is only the Internet that made the sort of
stability of that, that made us stop doing it live.
(03:37):
And so we prefer that prefer that format.
We prefer this suite of products.
And so hopefully as I found my way back round it and we can get
back going again with that. So welcome back for those that
are tuning in, be that live or after the fact.
What was always interesting is that some people TuneIn live in
their lunch breaks or whatever. We're back when we used to go
daily for tuning it over back just after within the sort of
(03:58):
what would still be considered within the pandemic.
So you get a few dozen TuneIn live, you'd get then a few 100
over the course of the week thatmight tune into that show and
then a few thousand once it had gone out onto the podcast feeds.
And so whilst, you know, it was interesting to see, it doesn't
matter as much about your engagement during the during the
sessions. It's just a way of us then
(04:18):
streaming that out and and also editing it on the fly, which
means that when we cover a contemporary topic or we have a
guest on that's really pertinentat the time, we don't have any
delay to it going out that following week or whatever.
So it goes out live and then we can re reproduce it across those
of the streams. So hopefully that suits you.
Now there's a few new features around here, some old ones as
(04:39):
well. What if I got here, tune it over
live stream is back for 2025. And so that can that can exist
there. But I can give you a little clue
as to what we're going to be talking about today, AI and MSK.
I'm going to keep this one briefin part because I just wanted to
make sure I tested the water, but also give you a bit of a
(04:59):
clue because I'm not going to guest on today because I wanted
to test it without one. But we will be getting guests on
all the people from our network that have got some interesting
to say. Oh, come and join me now and
again on a Wednesday. But as you guys might know
already, I'm happy to shoot the breeze on on topics, especially
those related and adjacent to MSK healthcare.
And one of the things that kind of crops up a lot and I get
asked about a lot on social media and through our Physio
(05:22):
Matters channels is regarding AIand MSK.
And that can be applying techniques and tools and
software and things like that across different parts of
clinical practice, education, how people are using it and
finding out as well how people are using it and innovating with
the large language models like ChatGPT, Grok, Meta, AI, that
sort of stuff, which we'll talk about.
But there's also then some fear and concern over MAIM and Ms.
(05:46):
care services, AI physiotherapy,whereby, and I'll use some
examples in a second as to how that's sometimes a risk of it
being a therapist replacement and how we might need to get
those risks if indeed they exist.
So that's what I wanted to just quickly shoot the breeze with
you on today. And obviously if you've got any
opinions, then P pop them in thechat wherever it is you're
viewing this and they will pop up and I will get to them if I
(06:07):
can. But if you get into this like
most people do after the fact, then obviously feel free to
still post on the channels. I just can't obviously then act
on them on the fly. Or you can always e-mail us
info@physiohyphenmatters.com is where to where we aggregate our
feedback. And so if you've got any ideas
and thoughts on that, then then please do let me know.
One thing that I wanted to sort of highlight as a new feature on
(06:29):
here is that I can pull up AQR code there.
If you look up that top left of my screen is AQR code.
It says Joe T on AI. I'm going to go into the detail
on this in a second, but one of you know, my colleague Joe
Turner wrote a brilliant piece for MSK Mag recently.
(06:50):
Well relatively recently, a few months back when there was news
of of AI encroaching on certain services and AI physiotherapy
services. And that piece is here.
ChatGPT. Can you tell me if you want my
job? If you would like to have a have
a read of that, then please do check out the the QR code in
that top left hand corner or just search Joe Turner ChatGPT
(07:12):
and you can find it just as I did.
And that was a really nice piecethat that I will be reflecting
on a little bit. And to be honest, you'll be
getting a sort of butchered version of of Joe's arguments to
some extent or her reflections from that piece.
And it was a really good one. And that you should absolutely
have a look at. So I'm just going to leave that
in the in the top corner just because I can now.
And apparently that's how these QR codes work on this new, new
(07:33):
streaming software. And so please someone test that
out for me. It'd be useful for me to know if
that works because I can use those again.
But as the caption suggests, I'mgoing to be talking a bit about
use cases. I'm going to talk about the
compatibility then with the AI and MSK as I see it, as well as
them reflecting on this therapist replacement concern
that people have. OK, so that's going to be what
today's shows about, if that's all right with you lot and I'll
(07:55):
get stuck in. So one of the the first things I
want to say is that it's really interesting and it's an emerging
picture as to how AI is interfering or I suppose being
introduced across our our sector.
OK, so MSK being what I considerthe sort of pain and injury
sector of which we work in. I'm a physiotherapist by
background, but we work alongside our osteopathic,
(08:18):
chiropractic, sports therapy, soft tissue therapy, sports
medical and orthopaedic colleagues.
You know, it's kind of we're raising standards in pain and
injury and reflecting on how that might affect the care of
pain and injury. Rehabilitation is central to
that. And that's something that
admittedly is a deep bias of mine.
But when it comes to the way in which AI is affecting the world,
(08:38):
affecting technology, its use cases across different sectors
are plentiful. And so we're all learning about
ways in which it can can be used, good cases, bad cases.
Sometimes it's being used in a bit of a gimmicky way for my
taste, and other times it's a really interesting way in which
it can be used to good effect. So just to get just a reel off a
(08:59):
few examples, and I'm always interested in yours as well, of
course, but I've been using AI for 12 months now for recording
and then interpreting my initialassessments.
And so for example, I'll have anAI software that will be
listening into my consultation with a patient.
(09:20):
So my start on my consultation says, if you don't mind, I use
an AI software that listens to our conversation, securely
stores it, but then basically itgives me a first draft of my
documentation. So it means I can focus
perfectly on the conversation we're having without having to
make as many little notes and things like that as we go along.
Is that OK with you percent and then press go and there's a
microphone that essentially bugsthe room records that records it
(09:43):
and stores it safely and it gives me a first draft of my
subjective assessment and especially useful for initial
assessments. You can use it for beyond and
there's all sorts of other ways in which I'm sure that that that
particular model I could use it more and I could use it
differently. But at the moment I've been
using that. I've been using that for the
part at least 12 months I'm pretty sure now.
And so that is one way in which AI can really help that I can
(10:04):
then ask that AI, once I've refined that documentation, I
can then say please send a referral letter in keeping with
that conversation and it will give me again a draft.
Nothing ever gets sent straight away.
You're always able to refine it and adjust it and you can train
that system to become better. If you don't like how the how
it's decided to word it or there's ways in which you can
make it, you know, make it elaborate more or, or, or
(10:26):
sharpen what it's said. And so that's just one way, one
use case of which I've been finding it to be interesting.
You notice in these models, So patient notes AI is the one that
I use at the moment, Heidi, I'vebeen playing around with as
well. And then there are other other
tools and no doubt that I don't know of that do similar things
or, or maybe even better. So let me know if you've, if
you've been playing around with that.
(10:47):
There's also some interesting chatbot tools such as Go triage
that have come out on the marketrecently, whereby essentially
ahead of an initial assessment, they get a link to, to then
interact with a chat bot which asks them subjective assessment
questions. And so it gives a bit of a pre
assessment. So a therapist can then we use
that in our team. Recently we've been trying out
especially for our junior members of staff to be able to
(11:09):
then get access to an AI interaction with the patient
ahead of time. Basically screen, you know,
things like screening red flags and and giving medical history
and the things that can sometimes take a bit of time up
is you've got that plus then theAI can give indications as to
what it feels the differential diagnosis to be aware of further
questions that you might want toask highlights within that
(11:31):
subjective assessment that the the robot's done just to mean
that you can get a bit of get ahead.
We all know that sometimes if the the notes that patients
might give you in their initially when they, when they
book in or whatever, that can bereally useful insight.
But also that a referral that might come from a doctor or from
a colleague might well have be laced with sort of biases or a
(11:52):
diagnosis that's that's too narrow for what you think it
might be. Whereas this is kind of a useful
way in which the AI can interpret that and give you an
idea and scope as to where you might focus your attention from
what they've said, as well as then giving a few clues as to
where you might take things next.
So, you know, having a, a juniortherapist making sure it says
to, to rule out cervical problems from a, from a shoulder
(12:16):
pain. Now you'd hope that a quality
MSK clinician is not going to then fail to check for referred
pain from the neck. But the way in which it might
indicate that because of some ofthe answers given by in the
subjective chat bot assessment, that's why you want to be more
thoughtful in that direction canjust give that extra bit of help
to, to a patient, sorry, to a therapist who's working with
(12:37):
that patient. So another interesting use case
of AI in practice that we've been playing around with with Go
triage and a really interesting innovation from them in terms of
software and technologies. I've heard of other ways in
which AI is is developing in theexercise prescription model, the
ways in which you can feed an AIwith, with motion capture, with
(12:58):
gait analysis. I've heard of ways in which you
can then just whatever information that you might give
it, how the model's then been trained to interpret.
I mean, from my taste, if I'm honest, that's where I start to
get annoyed and I feel like it gets a bit gimmicky is that if
that AI is being trained on a model in which we feel is a
little bit dated for. So for example, I've seen
(13:19):
whereby motion capture from, say, an iPhone camera feeding
into an AI model from a gate analysis, it's then stripping it
down and stripping it back and saying this is a crossover gate
on the left side. This is a heavy landing.
This is how they're running badly or wrong.
That starts to concern me in that we've kind of hopefully
(13:40):
moved past a model of biomechanical reasoning whereby
it's that those things that it'sobserving that are different to
the textbook gate would be then considered faulty.
And I've seen AI and obviously I'm not sure how much of this is
actually on the market and how much of this is sort of
prototype and stuff like that. So it's a bit harsh.
Maybe I'm assessing something that's a young product.
(14:02):
But generally speaking, obviously we get stuff sent to
us to, to check over on what do we think of this.
And I think that that's where I think AI can be used for a bit
gimmicky at best, but at worst, it can sort of hark us back to
an old model of practice and care because it's something that
the machine is needing to be taught rules and hard and fast
rules are something that we've kind of moved away from.
(14:24):
And the the grey area of MSK practice is obviously the
majority. And that the reasoning of of the
who's and ours of understanding the contributors to to people's
paying from bio mechanics is something that the robots maybe
won't necessarily always respectthat nuance.
Now, fortunately, these like clever large language models can
understand and will start to then factor in trade-offs and
nuance. And so maybe it's that if
(14:45):
trained appropriately to take with a pinch of salt what
they're finding, then maybe thatwill work.
We've seen some really useful innovations in medicine with
that with regards to how AI is continually being trained to
interpret imaging better. So CT heads, for example, was
something that was very early inthat From my understanding, AI
reads of CT heads for things like stroke and tumour are
(15:09):
superior inaccuracy to radiologists, human
radiologists. Now if we think about what that
then needs to be trained on is making sure it doesn't over
represent certain findings that might be incidental.
And so if we think about how that model might be trained on
reading spinal Mris, the knee Mris, that sort of stuff, which
I'm sure is coming, if not already existent and might be
already from, from my knowledge might be superior to to human
(15:33):
radiology interpretations. Those models need to make sure
they're not over representing incidental findings.
And then if that AI then he's also writing a report, for
example, that that might get to the eyes of a patient or to a
more naive, a young therapist orjunior therapist, how young they
are is relevant. Then it might well be that we
need to be stopping ourselves from succumbing to really
(15:56):
literal structural reasoning. Again, you know, if it's not
necessarily giving the, the context of which these might be
age-related findings, these might be incidental findings,
these aren't necessarily causal to symptoms, then we need to
make sure that then we as human interfaces with the patient
explain those things thoroughly.And so I think that that's
something we need to use as a mitigation.
(16:17):
It means that then for me, it's not that we need to, you know,
keep the robots out of the room and stop and not use them
because their accuracy can be brilliant.
And that was a really efficient way of using technology.
But if we don't also then work, make sure we we train them in
that nuance as well as them. We add that nuance when we, when
we're interfacing with the patient.
That's really important too. And so I'm, I'm interested with
(16:39):
that, where that goes. But just back to my gait
analysis example is that if thenthe structural specificity that
I'm mentioning with imaging interpretation becomes the same
sort of problem with biomechanical interpretation
from say, gait analysis. If you can feed an AI with a lot
of photographs and videos of your patient moving, doing a
squat, doing a, you're doing some running on a treadmill, and
you're then feeding it and all it can interpret it through is a
(17:00):
lens of which it's been trained,which is spotting movement
faults. This is a way in which I think
technology can really proliferate some of the worst
parts of MSK practice for my test and strips it back and sort
of ends up making us move backwards in time rather than
forwards in time. But the allure of it being new
and trendy would mean that it would feel like it was somehow
more accurate. And that accuracy can actually
(17:23):
be unhelpful. So the accuracy between
someone's specific postural analysis based on a Plumb line,
we kind of have hopefully grown out of narrowly saying that
you're moving differently to my perfect sort of postural Plumb
line. And therefore that's a
dysfunction. The robots, unless they're
trained to take that with a pinch of salt, might be more
literal. And therefore the patient may
(17:44):
well think of, you know, it's the the accuracy of that
technology is going to be beyondsome person's pair of eyes.
And therefore they're going to be maybe trusting the accuracy.
And that accuracy is something that can feel alluring and
tempting. It can feel almost person
centered. It can feel like it's really
specific to them. If it's not, then put into its
appropriate context to understand what to take with a
pinch of salt. That might be a concern.
(18:06):
So I'm concerned about that sortof the gimmicks of it.
But it doesn't necessarily mean that therefore it needs to be
thrown away or ignored. It just maybe the AI needs to be
trained or that we need to make sure that we are using these as
as tools and we remain that persons that the, the interface
that then personalizes that within the relevance of the case
in front of us. And we're then interpreting that
(18:27):
information appropriately. And so that's one of the things
that I think we we need to really bear in mind.
And this just speaks to what I consider to be then the
compatibility problem. So if I look at that bottom
there where it says use cases compatibility and then therapist
replacement. So to move me on to my sort of
second of those of those 3 is the compatibility.
I feel really strongly that the integration of AI into various
(18:50):
different ways of our practice, namely some of the ones I've
mentioned, although it's not an exhaustive list by any stretch,
I think there is a massive compatibility for us to use this
technology. It's not dissimilar to what I've
said before about measurement tech whereby using dynamometers
or force plates or even motion capture type software.
I think that these things do notend up being things that we need
(19:11):
to be threatened by, but actually ways in which we can
integrate them into our practice, create a massive
amount of time efficiency for ourselves, and then move them
forward. And in, in, in further in our
ways in which we can advance Ms.care practice as a conduit
between us and us and us and robots and technology.
(19:31):
It can really supplement our care rather than replace it.
Eventually I'll spat it out. So I do think it's compatible.
I think that where where it's not is based on what I've just
kind of described, whereby it can then be we're sometimes
fighting against it. If it's trying to be, it's
wanting things to be black and white rather than Shades of
Grey. If it isn't being trained on the
(19:52):
nuance. If it isn't, if it's something
that is almost directly interfacing with the patient.
So for example, I've had patients that before they've
come to see me, they've put their scan results through
ChatGPT. And sometimes it's failing to
take the edge off the the threatening behaviour that can
exist around the low grade spondylolisthesis.
When a patient puts that throughChatGPT and finds out that that
(20:16):
is a, he says, please explain this in later terms.
Then a slippage based on a fractspinal fracture is pretty
threatening. And so the ability for me to
then put that into an appropriate context with their
symptoms, with their history, how recent and new we feel that
is, means that then that was a really reassuring intervention
for me that ChatGPT was unlikelyto that the lack of literal
(20:41):
empathy that it's going to be able to share and understanding
how scary that might be to someone.
It did its job. It put an, A spinal image report
into layer language, but then happened to be that then the,
the literal interpretation of that layer language of a
slippage based on spinal fracture from, from an MRI is
scary stuff. And so helping that person to
(21:02):
understand it more was my job. And it came in the wrong order
for my test. And so it's not that that was
necessarily a bad thing, but it's just that we probably need
to interpret that with them if possible.
And trying to spot whether or not we can get upstream of the
robots or alongside the robots, I think is an important one
there for compatibility. Because I do think that
fundamentally that's where my, my head is at with this, is that
(21:23):
I do think it is compatible. So on to then this therapist
replacement thing. What is interesting about the AI
use cases at the moment is that there's all sorts of things that
are kind of alongside us or tools that we might use or ways
in which content creators are using ChatGPT to enhance and
assist their writing. We've got all sorts of ways in
(21:44):
which we can then use it to ask you questions and using it as a
search engine. All those sorts of things that
we can use in MSK, like we can use across any different area
that we might want to learn moreand educate ourselves on.
What is interesting though is that we have got examples now of
what are being considered AI physiotherapy services.
So instead of not as well as, but instead of a physiotherapist
(22:07):
in the NHS of which then is a team of physios that are working
in an MSK department, seeing patients on a regular basis, be
that initially through Zoom and and phone calls followed by them
in person assessments and treatment and rehabilitation.
Is that AI through what I understand to be sort of chat
bots and, and, and even sort of recording conversations and
(22:27):
things like that and then interacting with patients to do
then those initial subjective assessments, somewhat versions
of objective assessments throughvideo motion capture and things
like that. To then make exercise
prescription based on that and giving advice based on what
they've told them. And that, that MSK process of,
of treating and managing that pain and injury is being used
(22:50):
instead of physiotherapist because it can obviously
hundreds of people could be doing that at the same time.
You've not got the, the, the, aneasy way to jump through a
waiting list, for example. The, the, the interesting thing
for me is that I can so understand how that might have
some utility. I think it's got its limitations
I'm going to come to in a second, but the big one is can
(23:11):
that legitimately and legally becalled physiotherapy?
But that that's fascinating and I can so understand why people
are really frustrated about that.
So if we part for a second how threatening that is to say, the
professional physiotherapy and the MSK industry writ large,
like how good for the industry is, I'm going to part that for a
second, Not because I don't wantto go there, but because I just
don't think it's as important asthis next point for how people
(23:32):
are most annoyed. And what I'll admit I'm more
annoyed about is the fact that it seems pretty out there that
that is legally allowed to call itself physiotherapy.
So physiotherapy is not a treatment technique.
It is a regulated profession theworld over in the most part.
(23:53):
But certainly in the UK it's regulated and it's, it's a a
well defined scope of practice that is appropriately insured
under certain things that it does and things that it doesn't
do. It has a career structure.
Granted, we can criticize that and I do, but generally
speaking, it's a long time autonomous regulator profession,
especially in the UK and most ofthe of certainly the developed
(24:15):
world. And so therefore an AI that is
mimicking some of the things that we do, namely taking
subjective assessment questions and objectively testing
someone's movement and then creating a set of advice and
rehabilitation prescription fromthat.
Let's just say that's what that model is doing.
And a lot of the ones that I've been seeing recently and the
(24:36):
ones that have been commissionedwithin the NHS that of course
some controversy, that's fundamentally what they're
doing. Is that physiotherapy?
What? Why is that allowed to be called
physiotherapy? If they were to be doing the
same thing where they're making advice and then let's say that
they were someone was so bold asto allow for that to then
prescribe medication, would thatbe been called AI pharmacy
(25:00):
without the appropriate checks and balances?
Would that be that? Then if someone was to be doing
that, would that be called an AIGP?
And the risks that come from that are quite significant and
the reputational baggage that comes from that is significant.
Are they replicating physiotherapy?
They might not need to replicatethe entire of physiotherapy for
(25:23):
it to be legitimately called such.
But I just don't find it really interesting that we're that they
would be legally allowed to do that, right.
So if someone that isn't a regulated physiotherapist on the
HCPC register in the UK that hadappropriately been deemed
competent through that that process and that had the
appropriate qualifications. Anyone that's doing that would
declare themselves a physiotherapist.
(25:44):
That would be illegal. It's fraud.
And so the fact that then a robot, right, this, this, this,
this server of which is then running these machines has, has
not got A at least bachelor's inphysiotherapy.
It is not registered and regulated through the HCPC.
And therefore why is it that these machines are allowed to
(26:05):
call themselves physiotherapy? Whether they call themselves
physiotherapists, I'm not sure. I've seen that, you know, it's
not literal robots, is it? But I'm just meaning these
services are calling themselves physiotherapy now.
I wonder, and this is what I've heard about recently and I'm
doing a bit of investigating at the moment, is that whether or
not you've got a therapist or a team of therapists that then are
basically then screening, triaging and managing some of
(26:27):
the the work. So it's certainly
physiotherapists have fed and trained this AI.
But then also it might well be that then the decision making is
then verified by a physiotherapist, right?
So you can imagine then if it wasn't clever robots, it was
just a call centre of therapy assistance, let's say a human
therapy assistance that then before anything goes out to
(26:48):
actually 2 patients, the decision making then goes
through a specialist expert physiotherapist.
In that instance. I've heard in some ways that the
checks and balances in place arehuman physiotherapists.
And that's one of the reasons why it's then legally allowed to
call itself that or that it's legally dubious and that this,
that this sort of taking a punt on, on them being challenged on
(27:10):
that. But that's one of the things
that's happening in the NHS. That's one of the things that
then is proliferating. And I imagine that there's no
reason why those sorts of services and essentially it's
just, it's just an algorithm. And so there's no reason why
that wouldn't be able to be available or that the same thing
that is doing that would be an extension to ChatGPT in the
future and available as a free version on some more paid
(27:31):
version on a subscription. And so therefore, of course, the
scale of which that can operate is something certainly cheaper
than the humans of which it might consult.
So no wonder people are sort of threatened by it.
But there's also then interesting debates as to
whether or not it's legitimatelyallowed to be called that.
So a reflection on some of that is best best done by my
(27:53):
colleague Joe Turner. She wrote in MSK Mag relatively
recently, a few months ago. Sorry, I'm still finding my way
back round the round the software and system.
But here's here's just the frontpage of of Joe's article that
she wrote in MSK Mag. And if you scan the top left
corner on the QR code there of this screen and that will take
(28:15):
you to Jo's article, she did a great job with that.
For those that aren't familiar, MSK Mag is a monthly periodical
by Physio Matters producing content and now a print copy
that can be delivered to your door.
So go to mskmag.substack.com to subscribe to that.
Or of course, if you're a PhysioMatters member, you get that for
free or free it's included in your membership.
(28:38):
But yeah, check that QR code on the top left if you want to
listen to Jo's thoughts on this because they are very well
informed and really interesting reflections from her as to why
it's, it's something that we should be thinking about, if not
totally worried about. But also that the, the, the
challenge it has to us both as physical therapists and all in
the Ms. care communities. What is it that we do that is
(28:58):
therefore unique? How can we appropriately
personalize legitimately personalized care in ways that
the robots are unlikely to beat us?
For those that are developed sort of manual therapy.
So those whose care model is based on them particularly
intervening through manual techniques or taping or brace
application or or acupuncture orthings that of course at the
(29:18):
moment the the robots are struggling behind us on.
And they feel that actually the fact that physiotherapy and MSK
has deviated too far away from those sorts of interventions is
the reason that this is then left us vulnerable.
And I understand that argument. I think unfortunately, the way
in which it's articulated means that to reduce MSK care to the
administration of essentially pain relieving techniques is as
(29:42):
reductive in lots of ways professionally as what you're
fearful of the MSK machines and AI doing to us on exercise
prescription and advice. And I think that that's what's
interesting is that I feel that there's this interesting
balance, which I disagree with in both.
I think physiotherapy or MSK practice being reduced to advice
and exercise in the in the that makes us vulnerable to the
robots doing that job because it's it's Noddy or that it's
(30:05):
just a reasoning model that can be just applied more generically
and and less personally. Then I can you know, that that's
something I'm not not into. But similarly, if we think that
MSK practice in its purest form involves a lot of doing things
to people and that the manipulation and, and massage
and taping and braces and acupuncture and all that sort of
(30:26):
stuff is the is so fundamental to the essence of what it is to
be a quality MSK therapist. That that is where you know, we
should learn more into that as ameans of distinction from the AI
think that would be a mistake. I think that would hark back to
something that would be reductive and regressive.
But I understand the point. I think there's something in
between those two things, right.I don't think that the
(30:48):
combination of the factors that the reasoning framework that
helps us to give advice the way in which we'd be able to engage
with someone very personally as a human, we've got that
advantage, right? We can empathize and they can
empathize with us is something that is going to be useful.
The fact that we might then offer a rehabilitation advice
that might not look that different to what an AI would do
and also could be informed by anan AI program as well.
(31:09):
If we were to use it as an interface.
That compatibility I was mentioning before, if we did
that, but also as a means of facilitating functional
rehabilitation, we were able to lay our hands safely and
sensibly. We will be able to tape or to
apply brace, you know, to do other techniques to and with
patients to revolve ourselves and scale that rehabilitation
carefully by having that knowledge.
(31:30):
To be able to when to stand close to someone that's doing a
step up for the first time in a while on a knee that's sore and
being able to emote and empathize.
And to to that to lay what wouldn't be considered manual
therapy, but would be reassuringtouch within a functional
rehabilitation model that I do all the time in my gym few doors
down from here. Good luck to the robots trying
to do that. Right.
(31:50):
So you've got this interesting interface there where I think we
need to reflect and not allow itto succumb to, let's speak
conspiratorial and say that's what the robots would want.
They would want us to tear ourselves out in, up in, in, in
hands on, hands off debate againand rehashing those old
arguments like let's make sure we grow up and move past that
and try and integrate these AI related challenges to that, to
(32:12):
that story. I suppose it's the, it's the,
it's the main thing. So I've said a lot there.
I've got a lot to say. You can imagine, as usual on
everything, but particularly on to AI, it covers a lot of ground
and it won't be the last time wecover it on this show or in the
future. But I'm certainly very
interested in in what you all think about this and and also
for you to give me some feedbackas to how you feel it's gone
(32:33):
with regards to bringing the show back as a as a live format.
I will love you and leave you for now and make sure I turn off
all my turn off all my buttons. And if you tune in on Wednesdays
at lunchtimes, you will see moreof me in the in the coming weeks
and months. But I really appreciate those
that are watching live. If you have any feedback after
(32:56):
the fact, then do feel free to post on the comments.
We, we do spot those and we willbring them into other shows and
things like that as well. But if you'd rather, you can
always then give us a, an e-mailinfo@physiohyphenmatters.com for
any more further thoughts on that.
Or post comments underneath Joe's article if you do indeed
read that. Because I think that that's a
(33:18):
was a really nice reflection of,of, of, of thoughts that we
should all get behind and further the discussion rather
than having it all in isolation.We should sort of really try and
work as a community to think, right?
Where do we stand on this? What do we think is OK?
What isn't OK? What do we need to do to step
up? I think that that would be a
really progressive way of of working through this together.
But thank you so much for now. I will love you and leave you.
(33:40):
And until next time, I'll see you later.
Cheers folks. Here at physio matters, we think
Physio matters become a member today and access over 500
webinars. Get free tickets to shows and
access new content instantly. Access at homework or on the
train to make sure your CPD is on track.
Physio Hyphen matters.com More content than the all you can eat
(34:03):
buffet cart.
Are you fed up with printing outexercise programs or horror
drawing them? Solve every exercise
prescription issue you can thinkof using Rehab My Patient.
Thousands of pictures, videos ofevery movement you can dream up.
Send by e-mail or WhatsApp. Translate into different
languages at the click of a button.
But don't take our word for it. Sign up for a three month free
trial now. Just go to
(00:21):
rehabmypatient.com/physio Matters.
OK, hello folks. How are you?
(00:43):
Hope this is working. If not I'll learn from it.
It's been a while. We've not gone live for a bit.
Obviously never stopped podcasting, could never stop
doing that. However, I think if I know where
the buttons are, then I've I've got things working.
There's a few new new little features that I can play around
with and so I'm just trying to work that out.
(01:04):
So if you are tuning in live andif this is working and if you
can hear me, then please post inthe comments.
It'd be really nice to hear fromyou because we're deciding to
bring Chewing it over back as a live stream and I'll explain in
a moment why that even stopped. We've obviously still been
podcasting. We've been doing it as a pre
produced and slightly edited Let's go.
Where's the captions? Had that little ticker thing for
(01:28):
a second. Good.
I think I'm finding my way around this place again.
Unfortunately, there's a retrobeat called Jack March
that's clearly tuned in. He said hello to me on YouTube
of all places. Now, if any any anyone more
important than him is tuning in and can hear me OK on the
various different locations of which I'm streaming this, then
(01:50):
please let me know because it'd be useful for me to find that
out and to make sure it's working because we've turned all
the buttons back on and it's been a little while since we've
we've gone live. I can see in the background tidy
on the plinth there, isn't it? But yeah, bit of teething
problems. But yeah, lovely to be back with
you. And I'll explain a little bit
about what we're going to do with this show for 2025 each
(02:11):
Wednesday, as you can see, 12:30to 1 O clock ish.
So lunchtime, Wednesday lunchtime, I'm going to stream.
And we stopped doing tune it over as a live show reluctantly
to some degree, just because of the stability of the Internet
and the product that we were using needed a stable Internet
that we just couldn't always guarantee.
And that was really frustrating.And so we've moved it back to a
(02:33):
weekly show on Zoom that we thenproduced and put through the
feeds. There were some people that
found that frustrating that it wasn't on their preferred
location. There were some people that only
use LinkedIn, for example, and found that to be where they
found me. And so Aston hoped that we'd
bring that back or post it on there.
Once you had failed to stream itto all those different
(02:55):
locations, manually uploading itto each individual one was, was
a bit of a pain. So we ended up just being on the
podcast feed and on on YouTube. And so now hopefully we can cast
this then again to Facebook and Twitter and, and, and YouTube
and LinkedIn. And hopefully that's useful for
you. And so you can then engage with
it as live or after the fact. We'll also post that especially
(03:16):
successful ones that aren't glitchy and stuff.
We'll make sure it's stable. We'll post those onto onto
Spotify and Apple Podcasts and all that sort of gobbins as
well. But generally we're just going
to test the water and do a few Wednesdays here in December,
sorry in January. And we can do it that way.
But it really is only the Internet that made the sort of
stability of that, that made us stop doing it live.
(03:37):
And so we prefer that prefer that format.
We prefer this suite of products.
And so hopefully as I found my way back round it and we can get
back going again with that. So welcome back for those that
are tuning in, be that live or after the fact.
What was always interesting is that some people TuneIn live in
their lunch breaks or whatever. We're back when we used to go
daily for tuning it over back just after within the sort of
(03:58):
what would still be considered within the pandemic.
So you get a few dozen TuneIn live, you'd get then a few 100
over the course of the week thatmight tune into that show and
then a few thousand once it had gone out onto the podcast feeds.
And so whilst, you know, it was interesting to see, it doesn't
matter as much about your engagement during the during the
sessions. It's just a way of us then
(04:18):
streaming that out and and also editing it on the fly, which
means that when we cover a contemporary topic or we have a
guest on that's really pertinentat the time, we don't have any
delay to it going out that following week or whatever.
So it goes out live and then we can re reproduce it across those
of the streams. So hopefully that suits you.
Now there's a few new features around here, some old ones as
(04:39):
well. What if I got here, tune it over
live stream is back for 2025. And so that can that can exist
there. But I can give you a little clue
as to what we're going to be talking about today, AI and MSK.
I'm going to keep this one briefin part because I just wanted to
make sure I tested the water, but also give you a bit of a
(04:59):
clue because I'm not going to guest on today because I wanted
to test it without one. But we will be getting guests on
all the people from our network that have got some interesting
to say. Oh, come and join me now and
again on a Wednesday. But as you guys might know
already, I'm happy to shoot the breeze on on topics, especially
those related and adjacent to MSK healthcare.
And one of the things that kind of crops up a lot and I get
asked about a lot on social media and through our Physio
(05:22):
Matters channels is regarding AIand MSK.
And that can be applying techniques and tools and
software and things like that across different parts of
clinical practice, education, how people are using it and
finding out as well how people are using it and innovating with
the large language models like ChatGPT, Grok, Meta, AI, that
sort of stuff, which we'll talk about.
But there's also then some fear and concern over MAIM and Ms.
(05:46):
care services, AI physiotherapy,whereby, and I'll use some
examples in a second as to how that's sometimes a risk of it
being a therapist replacement and how we might need to get
those risks if indeed they exist.
So that's what I wanted to just quickly shoot the breeze with
you on today. And obviously if you've got any
opinions, then P pop them in thechat wherever it is you're
viewing this and they will pop up and I will get to them if I
(06:07):
can. But if you get into this like
most people do after the fact, then obviously feel free to
still post on the channels. I just can't obviously then act
on them on the fly. Or you can always e-mail us
info@physiohyphenmatters.com is where to where we aggregate our
feedback. And so if you've got any ideas
and thoughts on that, then then please do let me know.
One thing that I wanted to sort of highlight as a new feature on
(06:29):
here is that I can pull up AQR code there.
If you look up that top left of my screen is AQR code.
It says Joe T on AI. I'm going to go into the detail
on this in a second, but one of you know, my colleague Joe
Turner wrote a brilliant piece for MSK Mag recently.
(06:50):
Well relatively recently, a few months back when there was news
of of AI encroaching on certain services and AI physiotherapy
services. And that piece is here.
ChatGPT. Can you tell me if you want my
job? If you would like to have a have
a read of that, then please do check out the the QR code in
that top left hand corner or just search Joe Turner ChatGPT
(07:12):
and you can find it just as I did.
And that was a really nice piecethat that I will be reflecting
on a little bit. And to be honest, you'll be
getting a sort of butchered version of of Joe's arguments to
some extent or her reflections from that piece.
And it was a really good one. And that you should absolutely
have a look at. So I'm just going to leave that
in the in the top corner just because I can now.
And apparently that's how these QR codes work on this new, new
(07:33):
streaming software. And so please someone test that
out for me. It'd be useful for me to know if
that works because I can use those again.
But as the caption suggests, I'mgoing to be talking a bit about
use cases. I'm going to talk about the
compatibility then with the AI and MSK as I see it, as well as
them reflecting on this therapist replacement concern
that people have. OK, so that's going to be what
today's shows about, if that's all right with you lot and I'll
(07:55):
get stuck in. So one of the the first things I
want to say is that it's really interesting and it's an emerging
picture as to how AI is interfering or I suppose being
introduced across our our sector.
OK, so MSK being what I considerthe sort of pain and injury
sector of which we work in. I'm a physiotherapist by
background, but we work alongside our osteopathic,
(08:18):
chiropractic, sports therapy, soft tissue therapy, sports
medical and orthopaedic colleagues.
You know, it's kind of we're raising standards in pain and
injury and reflecting on how that might affect the care of
pain and injury. Rehabilitation is central to
that. And that's something that
admittedly is a deep bias of mine.
But when it comes to the way in which AI is affecting the world,
(08:38):
affecting technology, its use cases across different sectors
are plentiful. And so we're all learning about
ways in which it can can be used, good cases, bad cases.
Sometimes it's being used in a bit of a gimmicky way for my
taste, and other times it's a really interesting way in which
it can be used to good effect. So just to get just a reel off a
(08:59):
few examples, and I'm always interested in yours as well, of
course, but I've been using AI for 12 months now for recording
and then interpreting my initialassessments.
And so for example, I'll have anAI software that will be
listening into my consultation with a patient.
(09:20):
So my start on my consultation says, if you don't mind, I use
an AI software that listens to our conversation, securely
stores it, but then basically itgives me a first draft of my
documentation. So it means I can focus
perfectly on the conversation we're having without having to
make as many little notes and things like that as we go along.
Is that OK with you percent and then press go and there's a
microphone that essentially bugsthe room records that records it
(09:43):
and stores it safely and it gives me a first draft of my
subjective assessment and especially useful for initial
assessments. You can use it for beyond and
there's all sorts of other ways in which I'm sure that that that
particular model I could use it more and I could use it
differently. But at the moment I've been
using that. I've been using that for the
part at least 12 months I'm pretty sure now.
And so that is one way in which AI can really help that I can
(10:04):
then ask that AI, once I've refined that documentation, I
can then say please send a referral letter in keeping with
that conversation and it will give me again a draft.
Nothing ever gets sent straight away.
You're always able to refine it and adjust it and you can train
that system to become better. If you don't like how the how
it's decided to word it or there's ways in which you can
make it, you know, make it elaborate more or, or, or
(10:26):
sharpen what it's said. And so that's just one way, one
use case of which I've been finding it to be interesting.
You notice in these models, So patient notes AI is the one that
I use at the moment, Heidi, I'vebeen playing around with as
well. And then there are other other
tools and no doubt that I don't know of that do similar things
or, or maybe even better. So let me know if you've, if
you've been playing around with that.
(10:47):
There's also some interesting chatbot tools such as Go triage
that have come out on the marketrecently, whereby essentially
ahead of an initial assessment, they get a link to, to then
interact with a chat bot which asks them subjective assessment
questions. And so it gives a bit of a pre
assessment. So a therapist can then we use
that in our team. Recently we've been trying out
especially for our junior members of staff to be able to
(11:09):
then get access to an AI interaction with the patient
ahead of time. Basically screen, you know,
things like screening red flags and and giving medical history
and the things that can sometimes take a bit of time up
is you've got that plus then theAI can give indications as to
what it feels the differential diagnosis to be aware of further
questions that you might want toask highlights within that
(11:31):
subjective assessment that the the robot's done just to mean
that you can get a bit of get ahead.
We all know that sometimes if the the notes that patients
might give you in their initially when they, when they
book in or whatever, that can bereally useful insight.
But also that a referral that might come from a doctor or from
a colleague might well have be laced with sort of biases or a
(11:52):
diagnosis that's that's too narrow for what you think it
might be. Whereas this is kind of a useful
way in which the AI can interpret that and give you an
idea and scope as to where you might focus your attention from
what they've said, as well as then giving a few clues as to
where you might take things next.
So, you know, having a, a juniortherapist making sure it says
to, to rule out cervical problems from a, from a shoulder
(12:16):
pain. Now you'd hope that a quality
MSK clinician is not going to then fail to check for referred
pain from the neck. But the way in which it might
indicate that because of some ofthe answers given by in the
subjective chat bot assessment, that's why you want to be more
thoughtful in that direction canjust give that extra bit of help
to, to a patient, sorry, to a therapist who's working with
(12:37):
that patient. So another interesting use case
of AI in practice that we've been playing around with with Go
triage and a really interesting innovation from them in terms of
software and technologies. I've heard of other ways in
which AI is is developing in theexercise prescription model, the
ways in which you can feed an AIwith, with motion capture, with
(12:58):
gait analysis. I've heard of ways in which you
can then just whatever information that you might give
it, how the model's then been trained to interpret.
I mean, from my taste, if I'm honest, that's where I start to
get annoyed and I feel like it gets a bit gimmicky is that if
that AI is being trained on a model in which we feel is a
little bit dated for. So for example, I've seen
(13:19):
whereby motion capture from, say, an iPhone camera feeding
into an AI model from a gate analysis, it's then stripping it
down and stripping it back and saying this is a crossover gate
on the left side. This is a heavy landing.
This is how they're running badly or wrong.
That starts to concern me in that we've kind of hopefully
(13:40):
moved past a model of biomechanical reasoning whereby
it's that those things that it'sobserving that are different to
the textbook gate would be then considered faulty.
And I've seen AI and obviously I'm not sure how much of this is
actually on the market and how much of this is sort of
prototype and stuff like that. So it's a bit harsh.
Maybe I'm assessing something that's a young product.
(14:02):
But generally speaking, obviously we get stuff sent to
us to, to check over on what do we think of this.
And I think that that's where I think AI can be used for a bit
gimmicky at best, but at worst, it can sort of hark us back to
an old model of practice and care because it's something that
the machine is needing to be taught rules and hard and fast
rules are something that we've kind of moved away from.
(14:24):
And the the grey area of MSK practice is obviously the
majority. And that the reasoning of of the
who's and ours of understanding the contributors to to people's
paying from bio mechanics is something that the robots maybe
won't necessarily always respectthat nuance.
Now, fortunately, these like clever large language models can
understand and will start to then factor in trade-offs and
nuance. And so maybe it's that if
(14:45):
trained appropriately to take with a pinch of salt what
they're finding, then maybe thatwill work.
We've seen some really useful innovations in medicine with
that with regards to how AI is continually being trained to
interpret imaging better. So CT heads, for example, was
something that was very early inthat From my understanding, AI
reads of CT heads for things like stroke and tumour are
(15:09):
superior inaccuracy to radiologists, human
radiologists. Now if we think about what that
then needs to be trained on is making sure it doesn't over
represent certain findings that might be incidental.
And so if we think about how that model might be trained on
reading spinal Mris, the knee Mris, that sort of stuff, which
I'm sure is coming, if not already existent and might be
already from, from my knowledge might be superior to to human
(15:33):
radiology interpretations. Those models need to make sure
they're not over representing incidental findings.
And then if that AI then he's also writing a report, for
example, that that might get to the eyes of a patient or to a
more naive, a young therapist orjunior therapist, how young they
are is relevant. Then it might well be that we
need to be stopping ourselves from succumbing to really
(15:56):
literal structural reasoning. Again, you know, if it's not
necessarily giving the, the context of which these might be
age-related findings, these might be incidental findings,
these aren't necessarily causal to symptoms, then we need to
make sure that then we as human interfaces with the patient
explain those things thoroughly.And so I think that that's
something we need to use as a mitigation.
(16:17):
It means that then for me, it's not that we need to, you know,
keep the robots out of the room and stop and not use them
because their accuracy can be brilliant.
And that was a really efficient way of using technology.
But if we don't also then work, make sure we we train them in
that nuance as well as them. We add that nuance when we, when
we're interfacing with the patient.
That's really important too. And so I'm, I'm interested with
(16:39):
that, where that goes. But just back to my gait
analysis example is that if thenthe structural specificity that
I'm mentioning with imaging interpretation becomes the same
sort of problem with biomechanical interpretation
from say, gait analysis. If you can feed an AI with a lot
of photographs and videos of your patient moving, doing a
squat, doing a, you're doing some running on a treadmill, and
you're then feeding it and all it can interpret it through is a
(17:00):
lens of which it's been trained,which is spotting movement
faults. This is a way in which I think
technology can really proliferate some of the worst
parts of MSK practice for my test and strips it back and sort
of ends up making us move backwards in time rather than
forwards in time. But the allure of it being new
and trendy would mean that it would feel like it was somehow
more accurate. And that accuracy can actually
(17:23):
be unhelpful. So the accuracy between
someone's specific postural analysis based on a Plumb line,
we kind of have hopefully grown out of narrowly saying that
you're moving differently to my perfect sort of postural Plumb
line. And therefore that's a
dysfunction. The robots, unless they're
trained to take that with a pinch of salt, might be more
literal. And therefore the patient may
(17:44):
well think of, you know, it's the the accuracy of that
technology is going to be beyondsome person's pair of eyes.
And therefore they're going to be maybe trusting the accuracy.
And that accuracy is something that can feel alluring and
tempting. It can feel almost person
centered. It can feel like it's really
specific to them. If it's not, then put into its
appropriate context to understand what to take with a
pinch of salt. That might be a concern.
(18:06):
So I'm concerned about that sortof the gimmicks of it.
But it doesn't necessarily mean that therefore it needs to be
thrown away or ignored. It just maybe the AI needs to be
trained or that we need to make sure that we are using these as
as tools and we remain that persons that the, the interface
that then personalizes that within the relevance of the case
in front of us. And we're then interpreting that
(18:27):
information appropriately. And so that's one of the things
that I think we we need to really bear in mind.
And this just speaks to what I consider to be then the
compatibility problem. So if I look at that bottom
there where it says use cases compatibility and then therapist
replacement. So to move me on to my sort of
second of those of those 3 is the compatibility.
I feel really strongly that the integration of AI into various
(18:50):
different ways of our practice, namely some of the ones I've
mentioned, although it's not an exhaustive list by any stretch,
I think there is a massive compatibility for us to use this
technology. It's not dissimilar to what I've
said before about measurement tech whereby using dynamometers
or force plates or even motion capture type software.
I think that these things do notend up being things that we need
(19:11):
to be threatened by, but actually ways in which we can
integrate them into our practice, create a massive
amount of time efficiency for ourselves, and then move them
forward. And in, in, in further in our
ways in which we can advance Ms.care practice as a conduit
between us and us and us and robots and technology.
(19:31):
It can really supplement our care rather than replace it.
Eventually I'll spat it out. So I do think it's compatible.
I think that where where it's not is based on what I've just
kind of described, whereby it can then be we're sometimes
fighting against it. If it's trying to be, it's
wanting things to be black and white rather than Shades of
Grey. If it isn't being trained on the
(19:52):
nuance. If it isn't, if it's something
that is almost directly interfacing with the patient.
So for example, I've had patients that before they've
come to see me, they've put their scan results through
ChatGPT. And sometimes it's failing to
take the edge off the the threatening behaviour that can
exist around the low grade spondylolisthesis.
When a patient puts that throughChatGPT and finds out that that
(20:16):
is a, he says, please explain this in later terms.
Then a slippage based on a fractspinal fracture is pretty
threatening. And so the ability for me to
then put that into an appropriate context with their
symptoms, with their history, how recent and new we feel that
is, means that then that was a really reassuring intervention
for me that ChatGPT was unlikelyto that the lack of literal
(20:41):
empathy that it's going to be able to share and understanding
how scary that might be to someone.
It did its job. It put an, A spinal image report
into layer language, but then happened to be that then the,
the literal interpretation of that layer language of a
slippage based on spinal fracture from, from an MRI is
scary stuff. And so helping that person to
(21:02):
understand it more was my job. And it came in the wrong order
for my test. And so it's not that that was
necessarily a bad thing, but it's just that we probably need
to interpret that with them if possible.
And trying to spot whether or not we can get upstream of the
robots or alongside the robots, I think is an important one
there for compatibility. Because I do think that
fundamentally that's where my, my head is at with this, is that
(21:23):
I do think it is compatible. So on to then this therapist
replacement thing. What is interesting about the AI
use cases at the moment is that there's all sorts of things that
are kind of alongside us or tools that we might use or ways
in which content creators are using ChatGPT to enhance and
assist their writing. We've got all sorts of ways in
(21:44):
which we can then use it to ask you questions and using it as a
search engine. All those sorts of things that
we can use in MSK, like we can use across any different area
that we might want to learn moreand educate ourselves on.
What is interesting though is that we have got examples now of
what are being considered AI physiotherapy services.
So instead of not as well as, but instead of a physiotherapist
(22:07):
in the NHS of which then is a team of physios that are working
in an MSK department, seeing patients on a regular basis, be
that initially through Zoom and and phone calls followed by them
in person assessments and treatment and rehabilitation.
Is that AI through what I understand to be sort of chat
bots and, and, and even sort of recording conversations and
(22:27):
things like that and then interacting with patients to do
then those initial subjective assessments, somewhat versions
of objective assessments throughvideo motion capture and things
like that. To then make exercise
prescription based on that and giving advice based on what
they've told them. And that, that MSK process of,
of treating and managing that pain and injury is being used
(22:50):
instead of physiotherapist because it can obviously
hundreds of people could be doing that at the same time.
You've not got the, the, the, aneasy way to jump through a
waiting list, for example. The, the, the interesting thing
for me is that I can so understand how that might have
some utility. I think it's got its limitations
I'm going to come to in a second, but the big one is can
(23:11):
that legitimately and legally becalled physiotherapy?
But that that's fascinating and I can so understand why people
are really frustrated about that.
So if we part for a second how threatening that is to say, the
professional physiotherapy and the MSK industry writ large,
like how good for the industry is, I'm going to part that for a
second, Not because I don't wantto go there, but because I just
don't think it's as important asthis next point for how people
(23:32):
are most annoyed. And what I'll admit I'm more
annoyed about is the fact that it seems pretty out there that
that is legally allowed to call itself physiotherapy.
So physiotherapy is not a treatment technique.
It is a regulated profession theworld over in the most part.
(23:53):
But certainly in the UK it's regulated and it's, it's a a
well defined scope of practice that is appropriately insured
under certain things that it does and things that it doesn't
do. It has a career structure.
Granted, we can criticize that and I do, but generally
speaking, it's a long time autonomous regulator profession,
especially in the UK and most ofthe of certainly the developed
(24:15):
world. And so therefore an AI that is
mimicking some of the things that we do, namely taking
subjective assessment questions and objectively testing
someone's movement and then creating a set of advice and
rehabilitation prescription fromthat.
Let's just say that's what that model is doing.
And a lot of the ones that I've been seeing recently and the
(24:36):
ones that have been commissionedwithin the NHS that of course
some controversy, that's fundamentally what they're
doing. Is that physiotherapy?
What? Why is that allowed to be called
physiotherapy? If they were to be doing the
same thing where they're making advice and then let's say that
they were someone was so bold asto allow for that to then
prescribe medication, would thatbe been called AI pharmacy
(25:00):
without the appropriate checks and balances?
Would that be that? Then if someone was to be doing
that, would that be called an AIGP?
And the risks that come from that are quite significant and
the reputational baggage that comes from that is significant.
Are they replicating physiotherapy?
They might not need to replicatethe entire of physiotherapy for
(25:23):
it to be legitimately called such.
But I just don't find it really interesting that we're that they
would be legally allowed to do that, right.
So if someone that isn't a regulated physiotherapist on the
HCPC register in the UK that hadappropriately been deemed
competent through that that process and that had the
appropriate qualifications. Anyone that's doing that would
declare themselves a physiotherapist.
(25:44):
That would be illegal. It's fraud.
And so the fact that then a robot, right, this, this, this,
this server of which is then running these machines has, has
not got A at least bachelor's inphysiotherapy.
It is not registered and regulated through the HCPC.
And therefore why is it that these machines are allowed to
(26:05):
call themselves physiotherapy? Whether they call themselves
physiotherapists, I'm not sure. I've seen that, you know, it's
not literal robots, is it? But I'm just meaning these
services are calling themselves physiotherapy now.
I wonder, and this is what I've heard about recently and I'm
doing a bit of investigating at the moment, is that whether or
not you've got a therapist or a team of therapists that then are
basically then screening, triaging and managing some of
(26:27):
the the work. So it's certainly
physiotherapists have fed and trained this AI.
But then also it might well be that then the decision making is
then verified by a physiotherapist, right?
So you can imagine then if it wasn't clever robots, it was
just a call centre of therapy assistance, let's say a human
therapy assistance that then before anything goes out to
(26:48):
actually 2 patients, the decision making then goes
through a specialist expert physiotherapist.
In that instance. I've heard in some ways that the
checks and balances in place arehuman physiotherapists.
And that's one of the reasons why it's then legally allowed to
call itself that or that it's legally dubious and that this,
that this sort of taking a punt on, on them being challenged on
(27:10):
that. But that's one of the things
that's happening in the NHS. That's one of the things that
then is proliferating. And I imagine that there's no
reason why those sorts of services and essentially it's
just, it's just an algorithm. And so there's no reason why
that wouldn't be able to be available or that the same thing
that is doing that would be an extension to ChatGPT in the
future and available as a free version on some more paid
(27:31):
version on a subscription. And so therefore, of course, the
scale of which that can operate is something certainly cheaper
than the humans of which it might consult.
So no wonder people are sort of threatened by it.
But there's also then interesting debates as to
whether or not it's legitimatelyallowed to be called that.
So a reflection on some of that is best best done by my
(27:53):
colleague Joe Turner. She wrote in MSK Mag relatively
recently, a few months ago. Sorry, I'm still finding my way
back round the round the software and system.
But here's here's just the frontpage of of Joe's article that
she wrote in MSK Mag. And if you scan the top left
corner on the QR code there of this screen and that will take
(28:15):
you to Jo's article, she did a great job with that.
For those that aren't familiar, MSK Mag is a monthly periodical
by Physio Matters producing content and now a print copy
that can be delivered to your door.
So go to mskmag.substack.com to subscribe to that.
Or of course, if you're a PhysioMatters member, you get that for
free or free it's included in your membership.
(28:38):
But yeah, check that QR code on the top left if you want to
listen to Jo's thoughts on this because they are very well
informed and really interesting reflections from her as to why
it's, it's something that we should be thinking about, if not
totally worried about. But also that the, the, the
challenge it has to us both as physical therapists and all in
the Ms. care communities. What is it that we do that is
(28:58):
therefore unique? How can we appropriately
personalize legitimately personalized care in ways that
the robots are unlikely to beat us?
For those that are developed sort of manual therapy.
So those whose care model is based on them particularly
intervening through manual techniques or taping or brace
application or or acupuncture orthings that of course at the
(29:18):
moment the the robots are struggling behind us on.
And they feel that actually the fact that physiotherapy and MSK
has deviated too far away from those sorts of interventions is
the reason that this is then left us vulnerable.
And I understand that argument. I think unfortunately, the way
in which it's articulated means that to reduce MSK care to the
administration of essentially pain relieving techniques is as
(29:42):
reductive in lots of ways professionally as what you're
fearful of the MSK machines and AI doing to us on exercise
prescription and advice. And I think that that's what's
interesting is that I feel that there's this interesting
balance, which I disagree with in both.
I think physiotherapy or MSK practice being reduced to advice
and exercise in the in the that makes us vulnerable to the
robots doing that job because it's it's Noddy or that it's
(30:05):
just a reasoning model that can be just applied more generically
and and less personally. Then I can you know, that that's
something I'm not not into. But similarly, if we think that
MSK practice in its purest form involves a lot of doing things
to people and that the manipulation and, and massage
and taping and braces and acupuncture and all that sort of
(30:26):
stuff is the is so fundamental to the essence of what it is to
be a quality MSK therapist. That that is where you know, we
should learn more into that as ameans of distinction from the AI
think that would be a mistake. I think that would hark back to
something that would be reductive and regressive.
But I understand the point. I think there's something in
between those two things, right.I don't think that the
(30:48):
combination of the factors that the reasoning framework that
helps us to give advice the way in which we'd be able to engage
with someone very personally as a human, we've got that
advantage, right? We can empathize and they can
empathize with us is something that is going to be useful.
The fact that we might then offer a rehabilitation advice
that might not look that different to what an AI would do
and also could be informed by anan AI program as well.
(31:09):
If we were to use it as an interface.
That compatibility I was mentioning before, if we did
that, but also as a means of facilitating functional
rehabilitation, we were able to lay our hands safely and
sensibly. We will be able to tape or to
apply brace, you know, to do other techniques to and with
patients to revolve ourselves and scale that rehabilitation
carefully by having that knowledge.
(31:30):
To be able to when to stand close to someone that's doing a
step up for the first time in a while on a knee that's sore and
being able to emote and empathize.
And to to that to lay what wouldn't be considered manual
therapy, but would be reassuringtouch within a functional
rehabilitation model that I do all the time in my gym few doors
down from here. Good luck to the robots trying
to do that. Right.
(31:50):
So you've got this interesting interface there where I think we
need to reflect and not allow itto succumb to, let's speak
conspiratorial and say that's what the robots would want.
They would want us to tear ourselves out in, up in, in, in
hands on, hands off debate againand rehashing those old
arguments like let's make sure we grow up and move past that
and try and integrate these AI related challenges to that, to
(32:12):
that story. I suppose it's the, it's the,
it's the main thing. So I've said a lot there.
I've got a lot to say. You can imagine, as usual on
everything, but particularly on to AI, it covers a lot of ground
and it won't be the last time wecover it on this show or in the
future. But I'm certainly very
interested in in what you all think about this and and also
for you to give me some feedbackas to how you feel it's gone
(32:33):
with regards to bringing the show back as a as a live format.
I will love you and leave you for now and make sure I turn off
all my turn off all my buttons. And if you tune in on Wednesdays
at lunchtimes, you will see moreof me in the in the coming weeks
and months. But I really appreciate those
that are watching live. If you have any feedback after
(32:56):
the fact, then do feel free to post on the comments.
We, we do spot those and we willbring them into other shows and
things like that as well. But if you'd rather, you can
always then give us a, an e-mailinfo@physiohyphenmatters.com for
any more further thoughts on that.
Or post comments underneath Joe's article if you do indeed
read that. Because I think that that's a
(33:18):
was a really nice reflection of,of, of, of thoughts that we
should all get behind and further the discussion rather
than having it all in isolation.We should sort of really try and
work as a community to think, right?
Where do we stand on this? What do we think is OK?
What isn't OK? What do we need to do to step
up? I think that that would be a
really progressive way of of working through this together.
But thank you so much for now. I will love you and leave you.
(33:40):
And until next time, I'll see you later.
Cheers folks. Here at physio matters, we think
Physio matters become a member today and access over 500
webinars. Get free tickets to shows and
access new content instantly. Access at homework or on the
train to make sure your CPD is on track.
Physio Hyphen matters.com More content than the all you can eat
(34:03):
buffet cart.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy, Jess Hilarious, And Charlamagne Tha God!