The Practice We Think We Deliver Isn't Always the One We Give

Five papers. One uncomfortable question running through all of them: how far is the profession from where it says it wants to be?

There is a particular kind of cognitive dissonance that physical therapists live inside every day. We are trained to think of ourselves as evidence-based practitioners. We read. We attend conferences. We complete continuing education requirements. We believe, genuinely, that what we do is grounded in science.

This week's papers suggest the gap between that self-image and reality is wider than most of us would like to admit. But the same papers also point toward something useful: if we can see the gap clearly, we can do something about it.

The five studies reviewed this week span artificial intelligence in PT education, guideline adherence in clinical practice, the neuroscience of therapeutic empathy, the evidence landscape for patellofemoral pain, and the long-term economic consequences of osteoporotic fracture. Read in sequence, they form something like a diagnostic workup on the state of the profession itself.

Monday: Teaching Students to Think With AI, Not Through It

Zhang & Rapport, 2025 — Internet Journal of Allied Health Sciences and Practice

Start with where the next generation of clinicians is being trained. At Hawai'i Pacific University, a team embedded AI tools directly into a Doctor of Physical Therapy Evidence-Based Practice course, not as a novelty, but as a structured pedagogical scaffold. Students used AI to search literature, appraise studies, draft abstracts, and generate feedback. The course followed Bloom's Taxonomy, moving students from foundational knowledge toward critical synthesis.

Across all 11 measured domains, students showed statistically significant gains in both AI literacy and research self-efficacy. The largest improvements came in measurement strategy and study design, which is arguably where clinical judgment lives.

But the more interesting finding is also the cautionary one. Some students adopted AI-recommended study designs without any independent scrutiny. The authors describe this as overreliance, and it is exactly the failure mode that makes AI-assisted education dangerous if the scaffolding is missing. The tools are only as good as the critical lens applied to their outputs.

For educators, this paper is a practical template. For clinicians already in practice, it raises a more uncomfortable question: if future students need formal training to use AI without over-deferring to it, what habits have already been built around whatever clinical decision-support tools we use today?

Tuesday: Most of Us Are Not Practicing What the Evidence Preaches

Zadro, O'Keeffe & Maher, 2019 — BMJ Open

If Monday's paper is about preparing students to think carefully, Tuesday's is about what happens when they stop doing that in practice. This systematic review synthesized 94 studies from multiple countries, using surveys, clinical audits, and billing code analyses to ask a simple question: do physical therapists actually follow the evidence when they treat musculoskeletal conditions?

The answer, delivered with the kind of bluntness that BMJ Open tends to favor, is mostly no.

Only 54% of PTs chose guideline-recommended treatments in surveys. Forty-three percent chose treatments that guidelines explicitly discourage. Eighty-one percent used treatments with no guideline recommendation at all.

Passive modalities continue to appear in clinical records despite consistent guideline recommendations against them. Core interventions like exercise, advice, and education are delivered inconsistently and, when they are delivered, often go undocumented because clinicians treat them as too routine to record.

The documentation gap is worth sitting with for a moment. If reassurance and patient education are so routine that we do not bother to write them down, we have no way of knowing whether or how we are delivering them. Invisibility in the record is functionally the same as absence in a research database.

This paper is six years old. Whether those numbers have improved since 2019 is an open question, though there is little structural reason to expect a dramatic shift without targeted intervention. The forces that produce guideline non-adherence, professional habit, patient demand for doing something, time pressure in busy clinics, are not self-correcting.

Wednesday: Empathy Is Not a Soft Skill. It Is a Clinical One.

Chapman et al., 2026 — Physical Therapy

By Wednesday, we shift from what clinicians are doing to how they are doing it. This prospective cohort study from Physical Therapy followed 31 PT-patient dyads across 99 audio-recorded visits over six weeks, coding empathic communication behaviorally rather than relying on self-report. The researchers tracked empathic opportunities, moments when patients expressed something emotional, and empathic responses, moments when PTs recognized, validated, or explored that emotion.

The overall response rate was 67%, but the range across clinicians was striking: from 27% to 84%. That is not a minor individual difference. That is a fundamentally different kind of clinical interaction.

Patients treated by PTs in the highest empathy quartile, those responding to 91% or more of emotional cues, showed nearly three times the reduction in pain intensity compared to patients treated by PTs in the lowest quartile. This survived adjustment for age, sex, and depressive symptoms, and the effect strengthened over the six-week period.

The secondary outcomes, therapeutic alliance, affect, adherence, did not show significant changes. The authors suggest this may reflect insufficient statistical power or may indicate that empathy's analgesic mechanism operates through pathways not captured by those measures. Either way, the primary finding is not easy to dismiss: how a clinician responds to emotion shapes clinical outcomes in chronic pain, independently of the physical intervention delivered.

For a profession that has invested heavily in manual therapy techniques, exercise prescription parameters, and outcome measurement tools, this paper is a quiet argument that some of the highest-value clinical behavior happens in the first 30 seconds after a patient says something vulnerable.

Thursday: We Have 307 RCTs on Patellofemoral Pain. We Are Still Missing the Point.

Hart et al., 2026 — Journal of Orthopaedic & Sports Physical Therapy

Thursday's paper is the one that will likely provoke the most discomfort among researchers and research consumers alike. This systematic evidence and gap map reviewed 307 randomized controlled trials on patellofemoral pain, the largest such analysis to date. The question was not just what the evidence shows, but what shape the evidence has taken, and where the holes are.

The shape is not what it should be.

Therapeutic exercise appeared in 82% of all studies. Psychological interventions appeared in 2%. Pharmacological interventions in 3%. Sleep was measured in fewer than 10% of studies. So was quality of life. So were psychological factors. For a condition that contemporary pain science frames as biopsychosocial, the research base remains almost exclusively biomechanical.

Diagnostic rigor compounds the problem. Only 15% of studies met all seven items on the recommended PFP diagnostic checklist. That means the majority of the evidence base may not even be studying the same population, and we are synthesizing across that inconsistency as if it does not matter.

The clinical implication is not nihilistic. Exercise matters for PFP. The evidence for that is robust enough to survive the methodological critiques. But the implication is that clinicians treating patients who are not responding to standard physical interventions, and there are many of them, are operating in a near-total evidence vacuum when it comes to alternatives. That is a problem worth naming.

Friday: The Cost of a Fracture Lasts Five Years. The Diagnosis Often Comes After.

Tran et al., 2021 — Osteoporosis International

The week closes on a different kind of evidence gap: not a gap in research, but a gap in clinical action. This retrospective cohort study analyzed over 226,000 postmenopausal women with non-traumatic osteoporotic fractures, comparing direct and indirect costs against matched controls for five years following the index fracture.

The numbers are stark. Year-one incremental direct costs for hip fracture in commercially insured patients reached nearly $60,000. That burden did not disappear at year two or three. It persisted, measurably, for the full five-year observation window.

What makes this relevant to the clinical PT is the pre-fracture data. Fewer than 12% of these women had a documented osteoporosis diagnosis before their fracture. Only 21 to 31% were using osteoporosis-related medication. The fracture was, for most of them, the first signal that anything was wrong. Which means it was also too late.

Physical therapists occupy one of the few clinical touchpoints where high-risk patients show up routinely, complaining of back pain, balance problems, or general deconditioning, before anyone has thought to ask about bone density. That is not a marginal opportunity for secondary prevention. It is a primary one.

For working-age women, the indirect cost burden, short-term disability claims and absenteeism, was also meaningful. This is not just a Medicare problem. It is a workforce productivity problem. And physical therapy, at its most proactive, is one of the few interventions positioned to intercept it.

These five papers describe a profession that is capable of more than it is currently delivering. Not because clinicians are incompetent or indifferent, but because the systems, habits, and evidence structures that shape practice have not caught up to what the science is asking for.

The clinicians still using passive modalities. The PTs missing two out of three emotional cues. The 307 PFP trials that have collectively measured sleep in fewer than one in ten studies. The women who fractured before anyone thought to screen them.

None of these are indictments. They are descriptions of a direction that is not yet finished pointing.

Full citations available for all five articles. Articles summarized represent the author's weekly selection from current MSK literature and are intended for educational discussion among licensed clinicians.