The AI giveth and the AI taketh away
The first fully autonomous AI physiotherapy app has now been approved for use by the NHS
I’ve written a lot over the years about physiotherapists’ inability to engage with social and political questions. It’s a learned anosognosia that the profession deliberately embraced early on, and parcelled into its founding belief that treating the body-as-machine would be sufficient. This is really what The End of Physiotherapy was all about.
In three weeks time, I’ll be delivering a keynote at the 50th anniversary IFOMPT conference, and a large part of that talk will be spent trying to convince the audience that the post-professional era isn’t just something that’s happening to other people.
If ever we needed more evidence that we are on the cusp of a long fuse big bang in healthcare, it came this week with a report in The Guardian that the NHS would soon have its first fully AI physiotherapy clinic.
Let’s first be clear, first, about what we’re talking about. From the limited information available, Flok Health has developed an app that can allow UK-based subscribers to receive AI-generated exercise advice for low back pain.
‘Our Services comprise the provision of multidisciplinary musculoskeletal treatment for non-specific low back pain with or without associated leg symptoms. These symptoms may be acute or chronic. Assessment and treatment is primarily delivered in the form of interactive videos, which continuously adapt to based on multiple choice answers or inputs that you provide. We also offer telehealth appointments with members of our clinical team (this can be a doctor or a physiotherapist) where we determine that a more in depth consultation may be necessary or relevant as part of your treatment’ Link.
The Flok Health App is ‘the first technology to be granted medical device clearance under MHRA regulations for fully automating the triage, assessment and treatment of back pain’ Link.1 Information on the company’s registration with the Care Quality Commission also tells little about the app.
The justification given for the development of the app in The Guardian article centred on reducing waiting times in the face of growing specialised staff shortages. People can self-refer or follow referrals from a healthcare provider. Access is instantaneous, and all of the assessment, triaging and therapy is run through the app.
But what’s perhaps most intriguing is that last year a 3-month trial of 1,000 NHS workers with self-reported low back pain across four NHS trusts showed that ‘their experience with Flok had been at least equivalent to seeing a human physiotherapist, and 57% of patients said they thought the AI experience was better’.
‘More than four in five participants reported that their symptoms had improved during treatment with the platform’ Link.
Importantly, the trial data presented by The Guardian has not been published publicly, and there have been no formal, independent clinical trials of the app yet, but it has certainly points to the changing world of physiotherapy.
Some perspective
Firstly, let’s acknowledge that the app steps into perhaps the most instrumental, systematic, rule-bound dimension of physiotherapy: exercise advice for musculoskeletal (MSK) problems.
Throughout the profession’s history, MSK physiotherapists have bathed themselves in diagnostic algorithms, standardise tests, and anatomy-drive biomechanical reductivism. They have been the standard-bearers for treating the body-as-machine.
But in recent years there’s been quite a backlash directed towards the specialty’s traditionalists. It’s become the source of a great deal of acrimony within IFOMPT and other professional bodies, especially in the traditional Australasian, European and North American heartlands of manual and manipulative therapy.
On the one side are the reformers who have embraced a largely hands-off, patient-at-the-centre, intersubjective humanism. Their work is underpinned by cognitive behaviourism and the spirit of ‘therapeutic alliance’. They reject the mechanistic, anatomo-pathological, ‘passive’ manipulation- and mobilisation-based model of ‘old’ musculoskeletal physiotherapy, whose advocates are under now siege. It’s all getting rather tense.
But to my mind, both sides are fishing in an old stagnant pond; pushing and shoving to land some pretty sickly fish.
Hollowing out the profession
Towards the end of Physiotherapy Otherwise I wrote about the need for us to ‘hollow out’ the profession. This was a thought exercise designed to cut out anything from physiotherapy that could, and will, be replaced by ‘feasible alternatives’, as the Susskinds called it.
Hollowing out involves us actively cut out anything that:
can be written down or explained to someone else in a structured, logical way
can be written out as a series of steps, a therapeutic plan, or as an instructional guide, for instance.
you do in your patient assessments, advice, or treatment programmes, that is routine or can be standardised, that you do repeatedly,
is labour-intensive, or a physical task that requires little specialised knowledge or skill
is common or shared between you and a large group of your colleagues, whether it is an established clinical approach or method of thinking or working
involves you giving advice or guidance (on things like exercise programs) because this is already, or will soon be, available to people online
The reason for doing this is that these things ‘represent the things that will almost certainly be given to someone (or something) much cheaper to train, employ, and update than us in the near future’. They are ‘the kinds of work that is now increasingly delegated to assistants’ because they are ‘habitual, generic, context independent, or conventional’ Link.
Most tellingly, I suggested that we should cut these out ‘because AI algorithms will almost certainly take these from us’. And so it now seems.
Musculoskeletal physiotherapists — like most everyone else in the PT profession — still seem to believe that more evidence-based practice research is the answer to the profession’s declining fortunes (this, this and this, for example). But this both puts too much emphasis on preserving the good name of the profession (a goal that should never take the place of the best patient care), and misunderstands the biggest threats now facing PT: late capitalist atomisation, digital disruption, and the anthropogenic climate emergency (see this, this, and the bigger existential picture).
When it comes to apps like the Flok Health app, ‘the real threat is a mix of good enough outputs from AI paired with outsourcing and downsizing’
If electric guitars gave birth to rock and roll and sampling gave birth to hip hop, it’s entirely feasible that we’re now in the throes of an AI-induced change to the way we produce and consume music Link. Could a similar thing happen in healthcare?
AI is starting to influence healthcare in all sorts of ways never before imagined; ‘sex robots, care providers, domestic servants, and the disembodied voices of our digital tools and personal assistants’ Link are all becoming more common.
How far AI penetrates into healthcare and education remains to be seen. What is pretty clear, though, is that we are only at the beginning of a revolution, and much more is yet to come.