Latest / MSKMag OutLoud / AI solved our notes - but can it improve care?

Transcript

AI solved our notes - but can it improve care? · MSKMag OutLoud

• 0:00
  SPEAKER_00: AI sold our notes, but can it improve care?
• 0:04
  SPEAKER_00: Written by Callum Trott.
• 0:07
  SPEAKER_00: Sponsored by Preve.
• 0:10
  SPEAKER_00: Thousands of physios use Preve to automate treatment plans,
• 0:14
  SPEAKER_00: exercises, programs, notes and letters.
• 0:17
  SPEAKER_00: And those physios see patients twice as often.
• 0:21
  SPEAKER_00: You can get AI built for physios with Preve.
• 0:25
  SPEAKER_00: Use code CHEW26 for 10% off.
• 0:31
  SPEAKER_00: A few years ago,
• 0:32
  SPEAKER_00: if you asked most clinicians what they wanted AI to do,
• 0:36
  SPEAKER_00: write my notes, would have been pretty high on the list.
• 0:39
  SPEAKER_00: Which is fair enough.
• 0:40
  SPEAKER_00: Nobody became a physiotherapist because they had a deep passion for clinical documentation.
• 0:45
  SPEAKER_00: AI scribes have become remarkably good at solving that problem.
• 0:48
  SPEAKER_00: They listen to our consultations,
• 0:50
  SPEAKER_00: turn conversations into structured notes and draft letters
• 0:53
  SPEAKER_00: before we finished our coffee.
• 0:55
  SPEAKER_00: But after more than a decade treating patients and running clinics,
• 0:58
  SPEAKER_00: I've come to think we've aimed AI at the easiest part of the problem.
• 1:02
  SPEAKER_00: Whilst notes are frustrating,
• 1:03
  SPEAKER_00: the bigger admin problem is everything required to deliver good care
• 1:07
  SPEAKER_00: between appointments.
• 1:12
  SPEAKER_00: Think about what happens after a fairly normal MSK consultation.
• 1:17
  SPEAKER_00: You might need to update the patient's exercise program,
• 1:20
  SPEAKER_00: film a new exercise,
• 1:21
  SPEAKER_00: find the right video, and write down the sets, reps, and loading parameters.
• 1:25
  SPEAKER_00: You might want to send the patient a summary of what you think Perhaps there's
• 1:33
  SPEAKER_00: an insurer's form that needs completing before another block
• 1:36
  SPEAKER_00: of treatment is approved.
• 1:38
  SPEAKER_00: Then there's the patient you haven't seen for three weeks who you were supposed
• 1:41
  SPEAKER_00: to book again in seven days.
• 1:43
  SPEAKER_00: Should you message them?
• 1:44
  SPEAKER_00: Has reception already tried?
• 1:46
  SPEAKER_00: What were their goals again?
• 1:48
  SPEAKER_00: Did they stop coming because they were better?
• 1:51
  SPEAKER_00: Because they weren't improving?
• 1:52
  SPEAKER_00: Because life got busy?
• 1:54
  SPEAKER_00: Or because they forgot why the next appointment mattered?
• 1:56
  SPEAKER_00: None of these jobs are particularly difficult,
• 1:58
  SPEAKER_00: but multiply them by 12 patients a day,
• 2:01
  SPEAKER_00: 5 days a week, and suddenly good patient care has generated a second job.
• 2:05
  SPEAKER_00: That's the opportunity in AI that interests me most.
• 2:08
  SPEAKER_00: Not simply making the documentation of care faster, but making the delivery of good care easier.
• 2:15
  SPEAKER_00: We started with clinician-centered AI.
• 2:18
  SPEAKER_00: The first wave of healthcare AI has understandably been clinician-centered.
• 2:22
  SPEAKER_00: Documentation is repetitive,
• 2:24
  SPEAKER_00: expensive, and unpopular, so we built machines that could do more of it for us.
• 2:30
  SPEAKER_00: Whilst that's a win, faster notes don't necessarily mean better care.
• 2:35
  SPEAKER_00: If an AI scribe saves me 30 minutes at the end of the day, I've improved my working life.
• 2:41
  SPEAKER_00: If AI helps my patient understand their diagnosis,
• 2:43
  SPEAKER_00: remember their exercises,
• 2:45
  SPEAKER_00: see their progress,
• 2:46
  SPEAKER_00: and stay engaged with their rehabilitation,
• 2:48
  SPEAKER_00: we may have also improved their life too.
• 2:51
  SPEAKER_00: All practitioners want to deliver optimal patient care,
• 2:54
  SPEAKER_00: but we're often limited in what's achievable by the number of hours in the day,
• 2:58
  SPEAKER_00: and this is where AI for physiotherapy starts to become much more interesting.
• 3:04
  SPEAKER_00: Exercise adherence in MSK care is hardly a solved problem.
• 3:07
  SPEAKER_00: Recent reviews continue to describe uptake and adherence
• 3:10
  SPEAKER_00: to exercise-based rehabilitation as suboptimal.
• 3:15
  SPEAKER_00: We also know that adherence isn't simply a matter of telling someone to do their exercises.
• 3:20
  SPEAKER_00: Self-efficacy, social support, goal setting, instruction and demonstration are all relevant.
• 3:27
  SPEAKER_00: Digital interventions aren't a magic wand either.
• 3:29
  SPEAKER_00: A systematic review and meta-analysis found
• 3:32
  SPEAKER_00: that digital rehabilitation improved therapeutic exercise adherence
• 3:35
  SPEAKER_00: at intermediate follow-up, but not consistently at short or long-term follow-up.
• 3:40
  SPEAKER_00: The lesson clearly isn't to throw more technology at patients,
• 3:43
  SPEAKER_00: but rather to make the care we are already trying to provide more individualized,
• 3:48
  SPEAKER_00: consistent, and easier to act upon.
• 3:51
  SPEAKER_00: The gap between knowing and doing Most clinicians already know what good care looks like.
• 3:57
  SPEAKER_00: We know patients need clarity about their condition,
• 4:00
  SPEAKER_00: we understand the importance of regular goal setting,
• 4:02
  SPEAKER_00: and we know progress should be measured and visible.
• 4:06
  SPEAKER_00: We're also clear on the benefits of inter-appointment check-ins and the importance
• 4:10
  SPEAKER_00: of following up with patients who have dropped out of care prematurely.
• 4:13
  SPEAKER_00: But there is often a gap between knowledge and execution.
• 4:17
  SPEAKER_00: A busy clinician can genuinely believe in all of those things and still finish
• 4:20
  SPEAKER_00: a Tuesday afternoon with three exercise programs to update,
• 4:23
  SPEAKER_00: two insurer forms to complete and a patient from last week they meant to follow up.
• 4:28
  SPEAKER_00: What does patient-centered AI actually look like?
• 4:33
  SPEAKER_00: Imagine finishing a consultation and before your patient has walked out of
• 4:36
  SPEAKER_00: the clinic, they receive a message with an up-to-date treatment plan.
• 4:39
  SPEAKER_00: Not a generic post-appointment email, but their plan in a living portal.
• 4:43
  SPEAKER_00: It explains the working diagnosis in understandable language,
• 4:46
  SPEAKER_00: their goals and objective measures are there,
• 4:48
  SPEAKER_00: progress since their initial assessment is visible,
• 4:51
  SPEAKER_00: their exercise program reflects what you discussed five minutes earlier,
• 4:54
  SPEAKER_00: and their recommended appointment schedule is clear.
• 4:58
  SPEAKER_00: Then you see them again next week and rather than recreating the document,
• 5:01
  SPEAKER_00: the new consultation updates it.
• 5:04
  SPEAKER_00: Their shoulder flexion has moved from 120 to 150 degrees.
• 5:08
  SPEAKER_00: Progress measure automatically updates.
• 5:10
  SPEAKER_00: They're back to swimming twice a week.
• 5:12
  SPEAKER_00: The goal updates.
• 5:14
  SPEAKER_00: You've progressed their external rotation exercise.
• 5:16
  SPEAKER_00: The rehab program automatically progresses with their newly prescribed exercise.
• 5:22
  SPEAKER_00: The treatment plan becomes a living representation of their rehabilitation rather
• 5:25
  SPEAKER_00: than a PDF that was accurate for approximately seven minutes
• 5:29
  SPEAKER_00: after their initial assessment.
• 5:32
  SPEAKER_00: Another example is Exercise Prescription.
• 5:34
  SPEAKER_00: Historically, creating a genuinely individualized program has been surprisingly fiddly.
• 5:38
  SPEAKER_00: Search a library,
• 5:39
  SPEAKER_00: find something close enough,
• 5:41
  SPEAKER_00: add instructions,
• 5:41
  SPEAKER_00: change dosage,
• 5:42
  SPEAKER_00: and maybe even film the patient on their phone if you can't find what you're looking
• 5:46
  SPEAKER_00: for with no record of the video for the practitioner.
• 5:49
  SPEAKER_00: Now imagine the exercise content being created from what
• 5:52
  SPEAKER_00: the practitioner actually says during the appointment.
• 5:56
  SPEAKER_00: A suitable video can be automatically matched.
• 5:59
  SPEAKER_00: An AI-generated image can demonstrate an unusual exercise that doesn't exist in the library.
• 6:03
  SPEAKER_00: Or better still,
• 6:04
  SPEAKER_00: you can film the patient performing their own exercise correctly and add it
• 6:07
  SPEAKER_00: to their program in real time.
• 6:10
  SPEAKER_00: The clinician still makes the clinical decisions,
• 6:12
  SPEAKER_00: but AI removes the effort required to turn that decision into something useful
• 6:16
  SPEAKER_00: and accessible for the patient.
• 6:19
  SPEAKER_00: So what about when the patient disappears?
• 6:23
  SPEAKER_00: This might be the part I'm most interested in and excited by.
• 6:26
  SPEAKER_00: We talk a lot about exercise adherence in physiotherapy,
• 6:29
  SPEAKER_00: but we talk less about treatment plan adherence.
• 6:32
  SPEAKER_00: Imagine a patient presents with a problem that we reasonably expect will require
• 6:36
  SPEAKER_00: a period of rehabilitation.
• 6:38
  SPEAKER_00: Together, we establish some goals and start treatment.
• 6:41
  SPEAKER_00: Then somewhere between feeling a bit better and actually rebuilding
• 6:44
  SPEAKER_00: the capacity required for their goal, they disappear.
• 6:48
  SPEAKER_00: Historically, our systems haven't been particularly sophisticated at handling this.
• 6:52
  SPEAKER_00: Maybe reception runs a recall list,
• 6:54
  SPEAKER_00: or the practitioner notices an empty space in the diary and remembers them.
• 6:58
  SPEAKER_00: Maybe nobody does.
• 7:00
  SPEAKER_00: What if the software already understood the plan?
• 7:04
  SPEAKER_00: It knows the patient was recommended to return in 7 days, and it knows they haven't booked.
• 7:08
  SPEAKER_00: It knows their goal was to get back to running 10km but only progressed to 3km
• 7:12
  SPEAKER_00: at their last review.
• 7:14
  SPEAKER_00: That creates the possibility of a very different automated message to patients.
• 7:18
  SPEAKER_00: Not, hi John, you're due for an appointment.
• 7:21
  SPEAKER_00: But something close to,
• 7:23
  SPEAKER_00: hi John,
• 7:24
  SPEAKER_00: at like your last appointment,
• 7:25
  SPEAKER_00: you'd built your running back to three kilometers and we're working
• 7:27
  SPEAKER_00: towards your goal of 10 kilometers.
• 7:29
  SPEAKER_00: Your next review was planned for this week so we can reassess your progress
• 7:32
  SPEAKER_00: and adjust your loading.
• 7:33
  SPEAKER_00: Tom has a free spot tomorrow at 3 p.m. Would you like to book that?
• 7:38
  SPEAKER_00: Add a one-click booking option based on their usual clinic,
• 7:40
  SPEAKER_00: practitioner and appointment preferences
• 7:42
  SPEAKER_00: and suddenly following up patients doesn't require the clinician
• 7:45
  SPEAKER_00: to spend their lunch break trawling through a recall list.
• 7:48
  SPEAKER_00: And importantly,
• 7:49
  SPEAKER_00: the purpose isn't to squeeze another appointment out of someone
• 7:52
  SPEAKER_00: who doesn't need one.
• 7:53
  SPEAKER_00: If the patient has reached their goal,
• 7:55
  SPEAKER_00: brilliant,
• 7:55
  SPEAKER_00: the purpose is to reduce the number who accidentally fall off
• 7:58
  SPEAKER_00: a plan they haven't finished.
• 8:01
  SPEAKER_00: Horizontal AI versus Vertical AI.
• 8:04
  SPEAKER_00: This is where I believe healthcare AI will increasingly diverge.
• 8:08
  SPEAKER_00: A general medical scribe has an extraordinary broad job,
• 8:12
  SPEAKER_00: such as understanding conversations across different professions,
• 8:15
  SPEAKER_00: specialties, conditions, and workflows.
• 8:17
  SPEAKER_00: But an MSK clinician doesn't just need software that understands general medicine.
• 8:22
  SPEAKER_00: We need software that specifically understands MSK care,
• 8:25
  SPEAKER_00: both in terms of content and practitioner workflow.
• 8:29
  SPEAKER_00: That's the difference between horizontal and vertical AI.
• 8:32
  SPEAKER_00: A horizontal AI system might understand that I said three sets of eight split squats
• 8:37
  SPEAKER_00: and document it correctly,
• 8:38
  SPEAKER_00: whereas AI built for the vertical of physiotherapy would automatically convert that
• 8:43
  SPEAKER_00: into rehab videos or images and deliver it to a patient's phone
• 8:47
  SPEAKER_00: and track daily progress.
• 8:49
  SPEAKER_00: It would understand that today's objective measures may represent progress
• 8:52
  SPEAKER_00: against measures taken four weeks ago and track these improvements over time in
• 8:56
  SPEAKER_00: a way that's visible for both the practitioner and the patient.
• 9:00
  SPEAKER_00: And it understands that the patient's goals,
• 9:02
  SPEAKER_00: treatment plan,
• 9:02
  SPEAKER_00: exercises,
• 9:03
  SPEAKER_00: outcome measures,
• 9:04
  SPEAKER_00: appointment recommendations,
• 9:05
  SPEAKER_00: and clinical record are a continuum or relevant context across appointments rather
• 9:11
  SPEAKER_00: than six unrelated pieces of information.
• 9:15
  SPEAKER_00: This has been a major lesson for me while building PREEV.
• 9:18
  SPEAKER_00: Full disclosure, my interest in this isn't entirely academic.
• 9:21
  SPEAKER_00: I'm a physiotherapist and clinic owner,
• 9:23
  SPEAKER_00: and these frustrations were a large part of why we built PREEV in the first place.
• 9:27
  SPEAKER_00: From day one,
• 9:28
  SPEAKER_00: it was about building a tool
• 9:29
  SPEAKER_00: that was centered around better patient care whilst also doing the heavy lifting for
• 9:33
  SPEAKER_00: the practitioner.
• 9:35
  SPEAKER_00: What happens in the real world?
• 9:38
  SPEAKER_00: We're now starting to see what happens when AI is applied to
• 9:40
  SPEAKER_00: the patient journey rather than just the clinical record.
• 9:43
  SPEAKER_00: In an internal study across hundreds of practitioners using Preve,
• 9:47
  SPEAKER_00: we've seen uplifts of more than 95% in patient visitation rates due
• 9:52
  SPEAKER_00: to those patients having an updated treatment plan.
• 9:55
  SPEAKER_00: That's obviously not the same as evidence from a randomized controlled trial,
• 9:58
  SPEAKER_00: and visitation shouldn't be treated as a clinical outcome in itself,
• 10:01
  SPEAKER_00: but it is exciting.
• 10:03
  SPEAKER_00: Particularly because adherence to physiotherapy and exercise remains challenging,
• 10:06
  SPEAKER_00: and interventions including goal setting,
• 10:08
  SPEAKER_00: written instructions,
• 10:09
  SPEAKER_00: feedback,
• 10:09
  SPEAKER_00: and communication have all been investigated as ways of improving engagement.
• 10:14
  SPEAKER_00: We've also seen clinics report around a 50% increase
• 10:17
  SPEAKER_00: in monthly five-star Google reviews after implementing AI that focuses on
• 10:21
  SPEAKER_00: a more patient-centric model of care.
• 10:24
  SPEAKER_00: Again, Google reviews aren't an outcome measure,
• 10:27
  SPEAKER_00: but they are a useful window into something we probably don't measure enough,
• 10:31
  SPEAKER_00: the patient's perception of the care surrounding the treatment.
• 10:34
  SPEAKER_00: Do they understand what's happening?
• 10:36
  SPEAKER_00: Can they see their progress?
• 10:37
  SPEAKER_00: Do they know what they're working towards?
• 10:39
  SPEAKER_00: Do they feel looked after during their other 167 hours of the week
• 10:43
  SPEAKER_00: when they're not standing in our treatment room?
• 10:45
  SPEAKER_00: And then there's the clinician.
• 10:47
  SPEAKER_00: In our clinic data,
• 10:48
  SPEAKER_00: the practitioners using this broader workflow are saving around
• 10:51
  SPEAKER_00: an additional hour per day of administration compared
• 10:54
  SPEAKER_00: with practitioners already using an AI scribe for notes and letters.
• 10:58
  SPEAKER_00: The additional saving comes from the other work, plans,
• 11:01
  SPEAKER_00: exercises, forms, communication, and follow-up, the stuff surrounding good care.
• 11:06
  SPEAKER_00: Perhaps we asked AI the wrong question.
• 11:09
  SPEAKER_00: The conversation around AI in healthcare has understandably started with,
• 11:13
  SPEAKER_00: how much time can this save the clinician?
• 11:15
  SPEAKER_00: It's a good question, because clinician burnout and administrative burden matter.
• 11:19
  SPEAKER_00: Nobody benefits when skilled healthcare professionals spend their evenings writing notes.
• 11:23
  SPEAKER_00: But perhaps there's a better question.
• 11:26
  SPEAKER_00: What could I do for every patient if time was no longer the constraint?
• 11:30
  SPEAKER_00: Could every patient leave with a clear plan?
• 11:32
  SPEAKER_00: Could every exercise program actually reflect what happened in the room?
• 11:36
  SPEAKER_00: Could progress always be visible?
• 11:38
  SPEAKER_00: Could insurer paperwork be completed before it delays care?
• 11:42
  SPEAKER_00: Could every patient who unexpectedly drops off receive
• 11:46
  SPEAKER_00: a personalized message reminding them what they're working towards?
• 11:49
  SPEAKER_00: Could we provide a level of communication
• 11:54
  SPEAKER_00: And continuity that currently requires an impossibly organized clinician with unlimited time.
• 12:01
  SPEAKER_00: That's where AI gets exciting for me.
• 12:03
  SPEAKER_00: Not because I want less clinician involvement,
• 12:05
  SPEAKER_00: but because I want the clinician's limited time spent on the things
• 12:08
  SPEAKER_00: that actually require a clinician.
• 12:10
  SPEAKER_00: Things like listening,
• 12:11
  SPEAKER_00: reasoning,
• 12:12
  SPEAKER_00: reassuring, educating, motivating, connecting, and using their physical skills.
• 12:16
  SPEAKER_00: The first generation of clinical AI gave many practitioners some hours of their evenings back.
• 12:22
  SPEAKER_00: But the next generation has a bigger opportunity.
• 12:25
  SPEAKER_00: Perhaps the best measure of AI
• 12:26
  SPEAKER_00: in physiotherapy won't be how many minutes it saves us after an appointment,
• 12:30
  SPEAKER_00: but rather how much better we become at looking after this patient before
• 12:34
  SPEAKER_00: the next one.