Digital physical therapy: what the studies show—and what remains uncertain
A plain guide to reading the research, built from five published sources.

An app promises to ease your back pain from the living room. A friend swears by it. A news item says a study "found it works." How would you know what to make of that?
You do not need a medical degree to read research more carefully. You need a few plain questions, and the habit of asking them: who paid for this, who wrote it, what was compared with what, and what was actually measured?
This article is general information about research, not medical advice. Talk with a qualified health professional about your own care. We are a new organization and our programs are in development; we have no results to report here. What we can offer is a way of reading what others have reported.
What the pooled evidence reports
In October 2026, the Journal of Medical Internet Research published a systematic review and network meta-analysis of digital approaches to chronic musculoskeletal pain. It pooled 92 randomized trials with 12,595 participants. Across app-based programs, virtual reality and telerehabilitation, the authors reported "small to moderate favorable average effects" on pain and function.
They also reported limits. The certainty of the evidence was "predominantly low or very low," and the prediction intervals crossed the null. The authors say clinical importance and comparative effects remain uncertain. In 72.8% of the trials, the risk-of-bias assessment raised "some concerns." The average intervention lasted about nine weeks. The authors declared no conflicts of interest and reported funding from Chinese national and Hunan provincial programs. Those disclosures do not establish the independence of every trial the review included.
The review compares categories of digital interventions; its pooled findings are not a verdict on any particular commercial product.

Reading a disclosure: one trial, two statements
Disclosures are where funding and ties show up, and they can be easy to miss. Here is one example. A company tie does not by itself establish that a result is wrong; study design and methods still matter.
In July 2023, the journal npj Digital Medicine published a randomized trial of 140 adults with chronic low back pain, comparing a digital care program with conventional physical therapy over eight weeks. Both groups improved; the trial did not detect a statistically significant difference in disability improvement at eight weeks. That is not the same as showing the two approaches are equivalent. Dropout was lower in the digital group (15.7%) than in the conventional group (34.3%).
Now read the disclosures. In the acknowledgements, the authors write, "No funding was obtained for this work." Their competing-interests statement nevertheless identifies the program's developer as the study sponsor and discloses employment and equity ties among the authors.
The authors note the lack of blinding and objective outcome measures, high dropout in the conventional group, and the possibility that interventions in controlled settings may not reflect real-world conditions.
What a real-world claims study can and cannot say
In November 2025, the Journal of Health Economics and Outcomes Research published a retrospective claims analysis. The study compared 2,165 digital participants with 2,165 matched patients who had a back-pain-related physical therapy visit. In the year after the index event, 14.2% of digital participants and 18.2% of matched PT patients had at least one spinal imaging visit.
The study states that it was funded by the program's developer, and that two authors are employed by that company and hold equity in it. The authors say the analysis examines an association rather than demonstrating causation, and caution that digital participants may differ from the broader back-pain population. They also note that some clinical factors are missing from the data.
Notice what was measured. Imaging visits are not pain or function, and this is not a claim about reducing medically unnecessary imaging or improving health. A matched, non-randomized comparison is not a randomized trial. It answers a different question from "does it reduce pain?"
What other assessments say
In August 2025, the Cochrane Library published a review of telerehabilitation for neck pain: 13 trials and 1,042 participants. For exercise-based telerehabilitation versus minimal intervention in chronic neck pain, the review found low-certainty evidence of reduced pain and very-low-certainty evidence of improved function at short-term follow-up. The review considered several comparator types; these particular results concern minimal intervention. Its overall conclusion was that the evidence is "inconclusive due to its very low certainty." It is about neck pain only, and it is more than a year old.
In June 2024, the Peterson Health Technology Institute published an assessment of virtual musculoskeletal solutions. Its summary says the solutions it assessed "deliver clinically-meaningful improvements in pain and function compared to usual care." It defines "usual care" as generally physician visits and pain management, not physical therapy. PHTI's FAQ states that it and the Peterson Center on Healthcare are funded solely by the Peter G. Peterson Foundation, and that PHTI accepts no other outside funding for its work. That is PHTI's own disclosure, not an independent audit. We cite no figures from the full report.
These are examples, not a representative sample. We have not reviewed every trial in this field, and we make no claim about how many are company-funded.

A reader's checklist for any digital PT study or program
- Who paid, and who wrote it? Read both the funding statement and the competing-interests statement. They can say different things.
- What was compared with what? Digital care versus in-person care, versus usual care, versus no treatment, or versus minimal intervention are different questions.
- How was it assigned? A randomized trial and a matched claims analysis answer different questions.
- How long did it last? The trial above ran eight weeks and the pooled trials averaged about nine. Ask what is known beyond that.
- What was measured? Pain and function are different from engagement, completion or insurance claims.
- Who dropped out, and how many? High dropout in either arm changes how to read the result.
- Who was in the study? Ask whether people like you were included.
- How certain are the authors? Look for words like "low certainty," "association" and "inconclusive."
- Is it the whole picture? One study, especially a sponsored one, is a piece of the evidence, not all of it.
Use these questions on any study or program, including anything we publish.
Where this leaves a curious reader
The JMIR review reports small-to-moderate average effects, with evidence certainty predominantly low or very low. Each study design answers a different question, and sponsorship disclosures are one part of assessing the evidence. The JMIR and Cochrane authors call for stronger evidence. None of that makes digital care good or bad; it tells you to read with care.
Read more research explainers on the Global Wellbeing Institute's Insights page.
Sources
- Wang X, Jing W, Li Y, et al. "Effects of Digital Health Interventions in Chronic Musculoskeletal Pain: Systematic Review and Bayesian Network Meta-Analysis of RCTs." Journal of Medical Internet Research 2026;28:e95903. Published 7 October 2026. https://www.jmir.org/2026/1/e95903
- Cui D, Janela D, Costa F, et al. "Randomized-controlled trial assessing a digital care program versus conventional physiotherapy for chronic low back pain." npj Digital Medicine 2023;6:121. Published 7 July 2023. https://www.nature.com/articles/s41746-023-00870-3
- Lu L, Yadav S, Bailey J. "Real-World Effect of a Digitally Delivered Conservative Musculoskeletal Care Program on Spinal Diagnostic Imaging Utilization in a Commercially Insured Population with Chronic Back Pain." Journal of Health Economics and Outcomes Research 2025;12(2). Published 17 November 2025. https://jheor.org/article/145231
- Fandim JV, de Oliveira LA, Yamato TP, et al. "Telerehabilitation for neck pain." Cochrane Database of Systematic Reviews 2025, Issue 8: CD014428.pub2. Published online 12 August 2025. https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD014428.pub2/full
- Peterson Health Technology Institute. "Virtual Musculoskeletal (MSK) Solutions." Assessment published 5 June 2024; linked report updated 10 October 2024. https://www.phti.org/assessment/virtual-msk-solutions/ ; funding disclosure: https://www.phti.org/faq/
Global Wellbeing Institute is a new organization and its programs are in development. We do not report program results.
