Is Artificial Intelligence Ready for Musculoskeletal Rehabilitation?

Earlier this week, United Healthcare had announced that it would be partnering with Kaia Health, a virtual healthcare company, to provide virtual physical therapy utilizing artificial intelligence (AI) and online coaches. To no surprise, the news of this announcement spread like wildfire across social media amongst the physical therapist community, with most of the comments being highly negative towards the news. Many individuals stating that this was breaking state practice acts, since the company does not have a Physical Therapist (PT) on staff, some were out crying that it was a threat to the Physical Therapist profession, and others saying that the idea was just downright idiocracy. On the other hand, there were a small handful of individuals that commented the potential that this could have for patients. I currently have a neutral stance towards this issue, and only wish to side with whichever approach is going to be in our patients’ best interests. So, I figured I’d share some of my thoughts as well as what the current research says about this treatment approach regarding general musculoskeletal rehabilitation, injury prevention, non-specific spinal pain, and preoperative/postoperative rehabilitation for joint replacement surgery.

Current Capabilities of Artificial Intelligence in Musculoskeletal Rehabilitation

Let me start off by saying that we still do not know with certainty if current AI technology is able to consistently perform as well as humans, when it comes to different aspects of musculoskeletal rehabilitation, however there are promising results so far. According to Tack, AI technology was able to classify orthopedic images comparably to humans and identify successful exercise performance with 99.4% accuracy (1). One thing that’s important to point out is that the author makes it clear in his article that the AI is meant to be used to enhance physical therapist practice by automating mundane tasks for us, as opposed to removing the need for physical therapists (1). There are clear biopsychosocial factors of musculoskeletal rehabilitation that current AI technology is not anywhere near being able to process/interpret at this time, and therefore we should not expect AI to accurately diagnose or assess our patients for us anytime soon.

Injury prevention

When it comes to injury prevention, there has been conflicting evidence for decades as far as how to best monitor and predict a person’s risk of injury (2, 3). The most promising method at this time is likely the acute to chronic workload ratio, which essentially looks at an individual’s short term training programming, compares it to their long term training programming, and monitors various metrics to track their responses to individual training sessions (2). There is state of the art data that is collected when done right, however despite the vast amounts of data that can be collected, when you compare that data to a person’s self-reported responses to training, the subjective responses are actually a more accurate measurement of their training response compared to objective measures… (3) So, is artificial intelligence ready to take over the injury prevention space? No, not even close. Sorry to those of you who spent money on Whoop watches or Oura rings.

Non-specific Spinal Pain

Let’s just cut to the chase on this story of non-specific spinal pain (Warning: Hot Take). Non-specific spinal pain is by far the most common musculoskeletal condition that patients suffer from. The sad news is that the research is pretty clear that acute non-specific spinal pain will likely alleviate on its own with time, and chronic non-specific spinal pain will likely not; regardless of what we do with conservative interventions, the effects of our interventions are no better than placebo (4, 5). So, where do we go from here? This is my personal opinion. I believe if a patient is suspected to have non-specific spinal pain, the patient should still be referred to a physical therapist for an examination. A physical therapist is more likely to conduct a comprehensive evaluation of all of the biopsychosocial factors, potential red flags, expectation management, etc. that are necessary to assess for this patient population, simply due to the fact that we have the recommended time of at least 30 minutes to complete an evaluation, compared to physicians averaging 10 minutes of patient interaction time (6). From there, an artificial intelligence platform could be utilized in conjunction with the data that was collected during the physical therapist evaluation. Since patient outcomes are the same with or without supervised care, we might as well at least offer the option to utilize the lesser expensive route of the artificial intelligence platform; depending on patient circumstances and preferences of course (4, 5).

Preoperative Rehabilitation for Joint Replacement Surgery

This is also another controversial topic, as many physical therapists and orthopedic surgeons tend to butt heads on this (as I randomly found out on LinkedIn). In summary, most Orthopedic Surgeons believe they should have the authority to schedule a joint replacement surgery, when clinically indicated, without there being a preoperative rehabilitation requirement. Most Physical Therapists believe that it’s in the patients’ best interest to participate in preoperative rehab, in case there is a chance that the patient is able to recover without the need of surgery after all. If not, then at least preoperative rehab will improve patient outcomes for after the surgery, right? When we look at the evidence, it appears that preoperative rehabilitation for joint replacement surgery produces inconsistent, short term effects that are not clinically meaningful (7). There is some evidence to suggest that preoperative rehab may allow the patients to experience less pain, increased strength, and increased function compared to those who did not participate (8). However again, these effects were inconsistent, and short-lived (7-8). When we are talking about a procedure with a lifespan of ~20 years, does this suggest significant benefits to our patients? I will leave it up to you to decide. I personally believe that patients should at least be given the option to choose between completing 6 weeks of preoperative rehab with either a Physical Therapist or with an AI supervised home exercise program.

Postoperative Rehabilitation for Joint Replacement Surgery

There is also evidence suggesting that there are no differences in outcomes between independently completing a home exercise program vs. individualized, supervised outpatient physical therapy after a total knee arthroplasty (9). There is more research being conducted on post surgical rehabilitation considerations as we speak, so stay tuned for that (10). We also cannot control our patient’s physical activity levels once they complete their supervised care, therefore leaving our patients susceptible to a decline in functional abilities as time goes on (11). Given what we have seen so far in the research, I would hypothesize that completing a home exercise program with AI supervision would also produce similar results to PT supervised postoperative rehab in short, medium, and long term follow-up measurements. Only time will tell if that ends up being true though.

Final Thoughts

I just want to end this article by saying that at the end of the day, the world will continue to need physical therapists for a long time to come, even with the advancements of AI technology. At this time, we should look forward to the prospects of AI technology as a way to enhance musculoskeletal rehabilitation, as it has the potential to improve accessibility and affordability of these kinds of services. I do also want to warn my readers that this technology could prove harmful to our patients if not used correctly. Again, this technology does not have the ability to process/interpret biopsychosocial factors that are essential for clinicians to assess when providing musculoskeletal care. We must do our best to prevent the commoditization of our healthcare services as best as possible. We must also be aware of the possibility that this technology could unintentionally worsen disparities in our healthcare system with those that struggle with technology and health literacy. Meaning that there is a possibility that this technology would mainly benefit the wealthy and well-educated, which is the opposite of what this technology should be intended for, in my opinion. Only time will tell…

References:

1. Christopher Tack, Artificial intelligence and machine learning | applications in musculoskeletal physiotherapy, Musculoskeletal Science and Practice, Volume 39, 2019, Pages 164-169, ISSN 2468-7812, https://doi.org/10.1016/j.msksp.2018.11.012.

2. Saw AE, Main LC, Gastin PB. Monitoring the athlete training response: subjective self-reported measures trump commonly used objective measures: a systematic review. British Journal of Sports Medicine 2016;50:281-291.

3. Maupin, D., Schram, B., Canetti, E., & Orr, R. (2020). The Relationship Between Acute: Chronic Workload Ratios and Injury Risk in Sports: A Systematic Review. Open access journal of sports medicine, 11, 51–75. https://doi.org/10.2147/OAJSM.S231405

4. O'Connell NE, Kamper SJ, Stevens ML, Li Q. Twin Peaks? No Evidence of Bimodal Distribution of Outcomes in Clinical Trials of Nonsurgical Interventions for Spinal Pain: An Exploratory Analysis. J Pain. 2017 Aug;18(8):964-972. doi: 10.1016/j.jpain.2017.03.004. Epub 2017 Mar 25. PMID: 28347797.

5. Adding Physical Activity Coaching and an Activity Monitor Was No More Effective Than Adding an Attention Control Intervention to Group Exercise for Patients With Chronic Nonspecific Low Back Pain (PAyBACK Trial): A Randomized Trial. Crystian B. Oliveira, Diego G. D. Christofaro, Chris G. Maher, Márcia R. Franco, Anne Tiedemann, Fernanda G. Silva, Tatiana M. Damato, Michael K. Nicholas, and Rafael Z. Pinto Journal of Orthopaedic & Sports Physical Therapy 2022 52:5, 287-299. https://www.jospt.org/doi/10.2519/jospt.2022.10874.

6. Col, N., Hull, S., Springmann, V., Ngo, L., Merritt, E., Gold, S., Sprintz, M., Genova, N., Nesin, N., Tierman, B., Sanfilippo, F., Entel, R., & Pbert, L. (2020). Improving patient-provider communication about chronic pain: development and feasibility testing of a shared decision-making tool. BMC medical informatics and decision making, 20(1), 267. https://doi.org/10.1186/s12911-020-01279-8

7. Wang L, Lee M, Zhang Z, et al. Does preoperative rehabilitation for patients planning to undergo joint replacement surgery improve outcomes? A systematic review and meta-analysis of randomised controlled trials. BMJ Open 2016;6:e009857. doi: 10.1136/bmjopen-2015-009857

8. Moyer R, Ikert K, Long K, Marsh J. The Value of Preoperative Exercise and Education for Patients Undergoing Total Hip and Knee Arthroplasty: A Systematic Review and Meta-Analysis. JBJS Rev. 2017 Dec;5(12):e2. doi: 10.2106/JBJS.RVW.17.00015. PMID: 29232265.

9. Florez-García M, García-Pérez F, Curbelo R, Pérez-Porta I, Nishishinya B, Rosario Lozano MP, Carmona L. Efficacy and safety of home-based exercises versus individualized supervised outpatient physical therapy programs after total knee arthroplasty: a systematic review and meta-analysis. Knee Surg Sports Traumatol Arthrosc. 2017 Nov;25(11):3340-3353. doi: 10.1007/s00167-016-4231-x. Epub 2016 Jul 11. PMID: 27401004.

10. Mark-Christensen T, Thorborg K, Kallemose T, Bandholm T. Physical rehabilitation versus no physical rehabilitation after total hip and knee arthroplasties: Protocol for a pragmatic, randomized, controlled, superiority trial (The DRAW1 trial). F1000Res. 2021 Feb 25;10:146. doi: 10.12688/f1000research.50814.2. PMID: 34316356; PMCID: PMC8276181.

11. Stathi A, Greaves CJ, Thompson JL, Withall J, Ladlow P, Taylor G, Medina-Lara A, Snowsill T, Gray S, Green C, Johansen-Berg H, Sexton CE, Bilzon JLJ, deKoning J, Bollen JC, Moorlock SJ, Western MJ, Demnitz N, Seager P, Guralnik JM, Rejeski WJ, Hillsdon M, Fox KR. Effect of a physical activity and behaviour maintenance programme on functional mobility decline in older adults: the REACT (Retirement in Action) randomised controlled trial. Lancet Public Health. 2022 Apr;7(4):e316-e326. doi: 10.1016/S2468-2667(22)00004-4. Epub 2022 Mar 21. PMID: 35325627; PMCID: PMC8967718.