What the studies show, where each one stops, and what we do not claim.
Nine papers carry most of the weight on this site. Each is summarized with its actual numbers, its design, and the limit that keeps it from proving more than it does. The paper is linked so you can check us.
Home-exercise adherence is a real and large problem, with estimates of nonadherence as high as half of patients.
Monitoring changes what a clinician can see between visits, and the best available evidence associates monitored in-person PT with better functional outcomes than in-person PT alone.
A named clinician on record, paid only when the monthly interaction happens, is a workable accountability model, and the practice's own employed PT can be that clinician.
Physical therapy that is actually done, aligned with best practice and given time, works and is cost-effective; there is no magic bullet, and the variable is effort, adherence and time.
When and in what order therapy starts matters too: delay between diagnosis and therapy is associated with more opioid use in graded steps (67,245 knee OA patients), and among 87,305 Medicare Advantage older adults who received care for new low back pain within 90 days, episodes that started with PT had fewer spinal injections in that window (5.5% versus 21.8%; observational, not causal).
We do not claim
That JointCoach itself improves outcomes. No trial of JointCoach exists yet.
That any JointCoach measure is risk-adjusted, or that any per-clinician comparison is valid below about 40 episodes.
That digital tools change behavior on their own. The meta-analysis below says they mostly do not.
That the camera replaces in-person goniometry or examination.
That a practice is billing RTM on JointCoach today. As of September 2026 none is. The pilot is where the first real claims come from.
The nine studies
Each one, with its limit stated
01 · Monitoring plus in-person PT
In-person PT with remote therapeutic monitoring was associated with better functional outcomes than in-person PT alone.
DesignRetrospective case-control study of patients with musculoskeletal conditions in outpatient PT, with and without RTM.Finding72% of PT-plus-RTM patients reached the functional-status benchmark at discharge versus 63% of controls (P = .004). More RTM patients attended more than two visits a week (36% versus 24%). Controlling for other variables, RTM participation predicted reaching the benchmark (adjusted odds ratio 1.53, 95% CI 1.04 to 2.22).LimitRetrospective and observational: patients who opt into monitoring may differ from those who do not, and the study cannot separate the monitoring from the engagement it selects for. One provider network; not joint-replacement specific.What it supportsThe claim that monitoring alongside in-person care is associated with better outcomes, not that any particular product causes them.
Marshall T, Goldman A, Lyles R, Grundstein MJ. Retrospective case-control study on the effect of in-person physical therapy with remote therapeutic monitoring on functional outcomes and plan of care adherence amongst individuals with musculoskeletal conditions. Archives of Rehabilitation Research and Clinical Translation, 2025. PubMed 40980535
02 · Real-world RTM in outpatient PT
The first published implementation study of remote therapeutic monitoring in physical therapy: monitoring was implementable, and activity rose over six months.
DesignMixed-methods implementation study in outpatient physical therapy: 42 patients monitored for steps and exercise, with clinician goal-setting and interviews.Finding81% of participants completed six months of monitoring. Daily steps rose from about 4,900 to about 6,000 and weekly exercise from 44 to 83 minutes, with improvements that decelerated over time. Motivation and accountability drove engagement; health and technology problems were the barriers.LimitSmall, single program, no control group, and the monitored measures were steps and exercise minutes rather than joint range of motion. Published in a neurologic physical therapy journal, so the population is not orthopedic post-operative patients.What it supportsThat therapist-supported monitoring can be run in real outpatient practice and sustain engagement for months. The full paper also reports billing fidelity: the 20-minute management code was billed in roughly 53% of monitored months even for tracked patients (this figure is in the full text, not the abstract). That is where our calculator's 55% default comes from; the pilot replaces it with your own number.
Hohl et al. Remote monitoring enhanced outpatient physical therapy: a mixed-methods implementation study. Journal of Neurologic Physical Therapy, 2026. PubMed 42489634
03 · Digital tools and self-care
Digital health tools improve self-care monitoring by a moderate amount and barely move behavior or medication adherence.
DesignMeta-analysis of 55 randomized trials, 5,889 patients with chronic diseases.FindingBehavior and medication adherence moved very little (standardized mean difference 0.06 across 17 trials). The one clear monitoring signal is narrower than it looks: standardized mean difference 0.49 is heart failure only, from 5 studies and 364 participants, at low certainty. In diabetes the same review found little to no clear improvement across every self-care domain it pooled, at low to very low certainty.LimitChronic-disease populations, not orthopedic recovery, and heterogeneous interventions pooled together.What it supportsWhy JointCoach sells the monitoring half and not the behavior half: the instrument makes the clinician's view better; it does not by itself make patients do their exercises. Read strictly, this paper is stronger on the second half than the first — the near-null adherence result is pooled across 17 trials, while the positive monitoring result is one low-certainty subgroup. It is good evidence for what we do not claim, and weak evidence for what we do.
Longhini J, et al. Effectiveness of digital health interventions to improve self-care in patients with chronic diseases: a systematic review and meta-analysis. Journal of Medical Internet Research, 2026. PubMed 42263266
04 · Standardized measurement
Physical therapists use standardized outcome measures almost universally, but fewer than half believe they are administered in a standardized way across the profession.
DesignCross-sectional survey of 514 U.S. physical therapists.FindingNearly all used outcome measures, with performance-based tests used more than self-report surveys (97% versus 83%). 89% said performance-based tests help them make clinical decisions, against 61% for self-report surveys. Fewer than half agreed outcome measures are administered in a standardized way throughout the profession.LimitSelf-reported survey; respondents may not represent all clinicians.What it supportsThe comparability gap, and the reason to compute a joint angle the same way on every phone and every session. It says nothing about whether JointCoach's angle is more accurate, or more consistent, than a goniometer; that has not been measured.
Kristal A, Gaunaurd IA, Morgan SJ, et al. Use of standardized outcome measures among physical therapists in the United States: a cross-sectional survey. PLoS One, 2025. PubMed 40834029
05 · The employed PT as owner
In a randomized trial, physical therapists' diagnoses agreed with senior physicians' at high levels, with no adverse events.
DesignRandomized controlled trial, 102 patients, single site.FindingAgreement between PT and senior-physician diagnoses ranged from kappa 0.72 to 0.96, with no adverse events reported.LimitOne site, diagnostic agreement rather than treatment outcomes, and a setting that may not resemble a U.S. orthopedic practice.What it supportsThat putting the practice's own PT on record as the owner of a monitored episode is reasonable, which is exactly what the CY2027 proposed rule requires. It does not show that PT-owned monitoring improves recovery.
Nonadherence to home exercise in rehabilitation is estimated to reach as high as 50%, and connected technology is proposed as a way to measure and support it.
DesignViewpoint article summarizing the adherence literature and the case for connected-health measurement.FindingEstimates of nonadherence as high as 50%, with a potentially detrimental effect on outcomes; the authors argue adherence cannot be judged, or improved, without a robust measurement tool.LimitA viewpoint, not a trial; the 50% figure is an estimate drawn from earlier studies. A companion systematic review of adherence interventions found insufficient evidence to endorse any one of them in practice.What it supportsThe problem statement, and the measurement half of the answer. Not that any tool fixes adherence.
Argent R, Daly A, Caulfield B. Patient involvement with home-based exercise programs: can connected health interventions influence adherence? JMIR mHealth and uHealth, 2018. PubMed 29496655 · Peek K, et al. Interventions to aid patient adherence to physiotherapist prescribed self-management strategies: a systematic review. Physiotherapy, 2016. PubMed 26821954
07 · No magic bullet: physical therapy versus injection
A course of physical therapy was cost-effective compared with glucocorticoid injection for knee osteoarthritis, with more quality-adjusted life and no surgeries in the therapy group.
DesignSecondary economic analysis of a randomized trial: 156 adults with knee osteoarthritis at two U.S. military hospitals, randomized 1:1 to physical therapy (8 visits over 4 weeks) or glucocorticoid injection, followed one year; health economists blinded to outcomes.FindingPhysical therapy gained 0.076 more quality-adjusted life-years at one year (95% CI 0.02 to 0.126). Unadjusted knee-related costs were nearly identical between the groups. The incremental cost-effectiveness ratio, for knee-related costs and for all medical costs, fell far below the conventional per-QALY threshold, with a 99.2% probability of cost-effectiveness for knee-related costs. Most injection patients needed more than one injection (mean 2.6); 14 later went to physical therapy; four had knee surgery, all in the injection group and none in the therapy group.LimitA closed single-payer military system with no copays, so out-of-pocket costs were not captured; an unplanned secondary analysis; a one-year horizon; quality of life mapped from WOMAC scores rather than measured directly; total medical costs for any reason were higher with physical therapy.What it supportsThat physical therapy which is actually done, aligned with best practice and given time, works and pays for itself, and that the "one quick shot" rarely stays one. There is no magic bullet in musculoskeletal care; the variable is effort, adherence and time, which is the variable monitoring makes visible. It says nothing about JointCoach itself.
Rhon DI, Kim M, Asche CV, Allison SC, Allen CS, Deyle GD. Cost-effectiveness of physical therapy vs intra-articular glucocorticoid injection for knee osteoarthritis: a secondary analysis from a randomized clinical trial. JAMA Network Open, 2022. PubMed 35072722 · Parent trial: Deyle GD, et al. Physical therapy versus glucocorticoid injection for osteoarthritis of the knee. New England Journal of Medicine, 2020. PubMed 32268027
08 · When therapy starts
The longer physical therapy was delayed after a knee osteoarthritis diagnosis, the higher the risk of opioid use, and the relationship was graded.
DesignReal-world cohort of 67,245 American adults with incident knee osteoarthritis who were referred to physical therapy within a year, using commercial and Medicare Advantage claims from 1999 to 2018, split into opioid-naive (35,899) and opioid-experienced (31,346).FindingCompared with starting therapy within one month, each further delay carried more risk, in steps. Among people who had never used opioids, the adjusted relative risk of any opioid use rose 1.18, 1.49, 1.73 and 1.93 across delays of 1–3, 3–6, 6–9 and 9–12 months; for chronic opioid use it rose 1.25, 1.83, 2.29 and 2.50. Among people with prior opioid experience the same pattern held but more gently, reaching 1.54 and 1.65. Active, exercise-based therapy was associated with lower chronic opioid use in the opioid-naive group, but only under the stricter definition: the primary estimate was 0.86 with a confidence interval of 0.72 to 1.02, which crosses 1; the 0.77 (0.62 to 0.95) comes from a sensitivity analysis that required every visit to carry an active code. On dose, 6 to 12 visits was associated with lower chronic opioid use than 1 to 5; more than 12 visits was not, which the authors read as a marker of worse pain rather than a harm of therapy.LimitClaims data and observational: the diagnosis could not be validated against clinical criteria, people who wait for therapy differ from people who do not, and confounding by pain severity runs in exactly the direction of the finding. The authors adjusted for prior use and healthcare utilisation and report E-values for unmeasured confounding, but this is an association, not a demonstrated cause.What it supportsThe timing half of the same claim the trial above makes about the choice: physical therapy that is actually done, early, is where the value is. The graded step is the important part — 1.18 to 1.93 for any opioid use, 1.25 to 2.50 for chronic — a dose-response relationship of this shape is one of the classical reasons to take an observational finding seriously. It says nothing about JointCoach, which starts after the referral has already happened.
Kumar D, Neogi T, Peloquin C, Marinko L, Camarinos J, Aoyagi K, Felson DT, Dubreuil M. Delayed timing of physical therapy initiation increases the risk of future opioid use in individuals with knee osteoarthritis: a real-world cohort study. British Journal of Sports Medicine 2023;57:958–964. PubMed 36822841
09 · Care sequence in Medicare Advantage (N = 181,980)
Among 87,305 Medicare Advantage beneficiaries who received follow-up care within 90 days of new-onset low back pain, 78.4% received low-value care first (imaging or injection) and 21.6% received PT-first high-value care. Spinal injections were less common in the PT-first group (5.5% versus 21.8%).
DesignRetrospective cohort study of a 20% national sample of Medicare Advantage encounter data from 2021 (181,980 continuously enrolled adults aged 66 and older with new-onset low back pain after a 90-day washout, nested across 1,884 U.S. counties in a hierarchical two-level logistic regression). Services occurring within seven days of one another were grouped as the same week to account for scheduling logistics.FindingOver 90 days, 52.0% (94,675) received no tracked physical therapy, thoracolumbar imaging, or epidural steroid injection, while 48.0% (87,305) entered active procedural or rehabilitative care. Among the 87,305 receiving active care, 78.4% (68,408) received low-value care (48.3% isolated imaging without PT or injection; 41.5% of all 32,278 PT recipients underwent imaging or injection more than seven days before their first PT visit), and only 21.6% (18,897) received guideline-concordant, PT-first high-value care. When PT came first, 69.1% (13,065) needed PT alone and only 5.5% (1,039) ever received an epidural steroid injection within 90 days, compared with 21.8% (14,895) in the low-value group. Compared with primary care index visits (58.1% of the cohort), an index visit with a surgeon was associated with 33% lower adjusted odds of high-value care (aOR 0.67, 95% CI 0.63–0.70; 83.5% of the 12,077 active-care surgical index patients entered low-value sequences), pain specialists with 19% lower odds (aOR 0.81, 95% CI 0.77–0.85), and integrated medicine providers with 12% lower odds (aOR 0.88, 95% CI 0.81–0.95). County Medicare Advantage penetration of 46% or higher had no association with high-value care (aOR 1.00, 95% CI 0.95–1.06), whereas enrollment in MA plans rated below 4.0 stars (aOR 0.94), rural residence (aOR 0.86), high neighborhood deprivation with Area Deprivation Index of 50 or above (aOR 0.67), and Hispanic ethnicity (aOR 0.81) were each associated with lower odds of receiving PT-first care.LimitObservational 2021 administrative encounter claims for low back pain, not post-operative joint recovery. The dataset had no medication or referral data, so the 52.0% receiving no additional tracked care mixes mild self-resolving pain with patients who faced access barriers (dual-eligible 22.2% versus 16.3%, standardized difference 0.151; Hispanic 15.1% versus 11.5%) or received untracked analgesics.What it supportsThat sequence and entry point matter as much as whether physical therapy ever happens: when imaging comes before conservative care, 41.5% of PT episodes are already downstream of low-value testing, whereas PT-first pathways resolve 69.1% of episodes with PT alone and had fewer 90-day injections (5.5% versus 21.8%), an association rather than a proven effect. County Medicare Advantage penetration was not associated with high-value care (aOR 1.00), while rural and high-deprivation counties were (aOR 0.86 and 0.67). It says nothing about JointCoach itself.
Kumar A, Deutscher D, Chou LN, Fritz JM. Variation in High Value Care for Low Back Pain Among Older Adults Enrolled in Medicare Advantage Plans. Journal of the American Geriatrics Society 2026 (published online 2026-10-09). doi:10.1111/jgs.70779 · PubMed 42852584. The 5.5% and 21.8% injection shares and the 41.5% figure are computed from the paper’s Table 1 and Table 2 counts.
Also cited on this site: video-based motion analysis is already used by roughly half of physical therapists, with established methodological standards (International Journal of Osteopathic Medicine, 2023); the CMS rules and rates are on the FAQ with their primary sources.
How people read their pain
Two trials on pain beliefs, each with its limit stated
Neither is about joint replacement, and neither tests JointCoach. They are here because fear of movement and beliefs about pain come up in every rehab program, and the headline numbers travel further than their limits do.
A · A physical therapist works on the fear
When physical therapists worked on how people understood and feared their back pain, alongside graded movement, disability fell more than with usual care, and most of the gain was still there three years later.
DesignRandomized trial of 492 adults with chronic, disabling low back pain in 20 Australian physiotherapy clinics: cognitive functional therapy, with or without movement-sensor biofeedback, against usual care. 312 people were followed to three years.FindingDisability on the Roland-Morris scale (0 to 24) improved 4.6 points more than with usual care at 13 weeks, with similar results at 52 weeks and lower societal cost. At three years disability was still 3.5 to 4.1 points better, while pain was only 1.0 to 1.5 points better on a 0 to 10 scale.LimitNot blinded, and the method’s developers co-authored the trial. Chronic low back pain, not joint replacement or knee arthritis. Pain moved much less than function. In a secondary analysis of 301 participants, the therapy did not reduce opioid dispensing.What it supportsThat what a person believes about their pain, and whether they are afraid to move, are things a physical therapist can work on in the room, with gains in function that last. It says nothing about JointCoach, which does not deliver this therapy.
Kent P, et al. Cognitive functional therapy with or without movement sensor biofeedback versus usual care for chronic, disabling low back pain (RESTORE): a randomised, controlled, three-arm, parallel group, phase 3, clinical trial. The Lancet, 2023. PubMed 37146623 · Three-year follow-up: Hancock M, et al. Lancet Rheumatology, 2025. PubMed 40780241 · Opioid dispensing: Thiveos L, et al. European Journal of Pain, 2026. PubMed 41979226
B · Reading pain as a signal, not damage
In one Boulder trial, two-thirds of people with chronic back pain were pain-free or nearly so after a therapy that teaches them to read their pain as a real but non-dangerous brain signal. No independent team has repeated it yet.
DesignRandomized trial of 151 adults with chronic back pain, mostly low to moderate (4.1 of 10 on average): pain reprocessing therapy (a physician visit plus eight hours of therapy), an open-label placebo injection, or usual care.FindingPain-free or nearly pain-free (0 or 1 of 10) after treatment: 33 of 50 (66%) with the therapy, 10 of 51 (20%) with placebo, 5 of 50 (10%) with usual care. At five years 113 of the 151 answered: 21 of 38 (55%) in the therapy group were pain-free or nearly, against 10 of 39 (26%) and 13 of 36 (36%). Counting everyone randomized to the therapy, 21 of 50 (42%) is the floor. Mind-or-brain explanations rose from 10% of the reasons people gave for their pain to 51% with the therapy, against 8% in the other arms; the shift tracked the drop in pain but explained only about 9% of its variance.LimitOne trial, not yet repeated by an independent team. The therapy was delivered by its developer or one other expert therapist, and several authors disclosed financial ties to pain-treatment businesses. The placebo arm got one injection while the therapy arm got hours of attention, and therapy cannot be blinded. Participants were mostly college-educated, active volunteers from one city with low-to-moderate pain. A quarter were missing at five years, and nobody measured the years in between.What it supportsThat how a person explains chronic back pain may be one route to relief, and that the idea deserves an independent trial. It does not support telling someone recovering from joint surgery that their pain is only a brain signal: new or worsening pain after surgery gets checked first. JointCoach does not deliver this therapy.
Ashar YK, et al. Effect of Pain Reprocessing Therapy vs Placebo and Usual Care for Patients With Chronic Back Pain: A Randomized Clinical Trial. JAMA Psychiatry, 2022. PubMed 34586357 · Five-year follow-up: JAMA Psychiatry, 2025. PubMed 40736996 · Pain attributions: JAMA Network Open, 2023. PubMed 37768666
What JointCoach does with this today is small. After a session, a pain rating of 4 or more brings up the warning signs first, then says that some pain during rehab exercise is expected and that how much is right is the physical therapist’s call. The message is shown on the phone and nothing is sent. The app does not deliver either therapy and does not ask how you explain your pain.
What would change these claims
Three things would move this page, and we will say which one did.
The pilot's own numbers: days with data, minutes reviewed, claims paid and denied, and range-of-motion trends for real patients. The regulatory consult on whether camera-based pose estimation sits inside the goniometer device class, which decides whether the device-supply codes apply. And the CY2027 final rule, which decides who may furnish the monitoring time. Until then the honest position is the one above: strong evidence for the problem, moderate evidence for monitoring alongside care, and none yet for JointCoach itself. The pilot is how that changes.