The resident had been in the program for five months when the case appeared on a Monday morning schedule. Forty-one years old, construction work, low back pain radiating into the left leg, six weeks of symptoms, no red flags on the intake form. Textbook. The resident had seen this presentation a dozen times since starting. The evaluation took twenty-four minutes. The hypothesis — lumbar disc irritation with radiculopathy — was generated within ten. The treatment plan was sound: directional preference testing, neural mobilization, and graded loading. Every clinical decision was defensible.
The mentor, watching through the one-way mirror, checked a box on an assessment form: *Can perform this activity independently at this complexity level.*
Two days later, a different case. Fifty-seven years old, eighteen months of diffuse low back pain, three prior episodes of physical therapy documented in the medical record, each ending in partial improvement that did not hold. The resident’s evaluation took forty-eight minutes. Hypothesis generation was slower—not because the resident lacked knowledge, but because the clinical picture did not yield a clear pattern. There was bilateral pain without a clear directional preference. Sleep was disrupted. The patient mentioned, almost in passing, a recent change in employment status.
The mentor checked a different box: *Requires guided supervision for this activity at this complexity level.*
Same resident. Same week. Same profession. Two entirely different clinical realities, separated not by knowledge but by complexity — and an assessment system honest enough to say so.
The way most professions have answered the competency question for decades amounts to a binary: pass or fail. A clinician either meets the standard or does not. A therapist is either competent or not yet competent. An examination is either passed or failed. The entire apparatus of credentialing, from licensure to board certification, is built on this architecture. And for a long time, the architecture seemed adequate.
It was never adequate. It was simply the best the system could manage.
The problem with binary competency is not that it is wrong — it is that it is incomplete. A clinician who passes a board examination has demonstrated knowledge. But knowledge is not the same as clinical reasoning, and clinical reasoning is not uniform across complexity. The forty-one-year-old with textbook radiculopathy and the fifty-seven-year-old with an eighteen-month pain history that has defeated three prior clinicians are not the same clinical challenge. They require different depths of reasoning, different tolerances for ambiguity, and different capacities for integrating information that does not fit the expected pattern.
Binary assessment cannot see any of this. It sees a clinician who has met a threshold. It cannot see where, within the landscape of clinical work, that clinician can be trusted to operate independently — and where they still need a more experienced set of eyes in the room.
The entrustment framework was designed to see exactly this. Introduced by ten Cate in 2005 and subsequently adopted worldwide in medical education, entrustable professional activities — EPAs — reframe assessment from a question about the person to a question about the match between the person and the task (ten Cate, 2005). The question is no longer Is this clinician competent? The question becomes: Would you trust this clinician to perform this activity independently, without you in the room, for the next patient at this complexity level?
That shift changes everything.
In the model used by one ABPTRFE-accredited residency program, clinical competence is organized into seven domains, 35 competencies, and 10 EPAs. Twenty-five of those residency competencies are direct developmental precursors to fellowship competencies — the architecture honors the developmental logic rather than pretending the residency can produce the same clinician the fellowship produces. The residency builds routine expertise. The fellowship builds something else. The entrustment framework makes the boundary between them visible rather than leaving it to chance.
The practical effect is that a resident can be trusted to independently manage a straightforward mechanical low back presentation, while still requiring guided pattern recognition for complex chronic pain with central sensitization. This is not a failure. This is an honest developmental picture — one that the old binary model could never produce.
And here is the part that matters most for the profession: when a patient does not respond to competent treatment, the entrustment framework changes the interpretation. In the old model, non-response is ambiguous. Perhaps the treatment was wrong. Perhaps the clinician was inadequate. Perhaps the patient was non-compliant. The system has no mechanism for a more precise answer.
In the entrustment model, non-response to competent treatment becomes clinical data. The case exceeded the clinician’s current entrustment level. The system’s response is not a fault —it is a match. The patient needs a clinician entrusted to a higher complexity tier, or the treating clinician needs guided supervision to navigate the unfamiliar clinical territory. Either way, the response is structural, not personal.
This reframe is not theoretical. The APTA’s 2025 competency-based education framework defines 19 EPAs across 8 domains of competence for physical therapist ’ entrance into practice—the most comprehensive application of entrustment logic to the profession to date (APTA, 2025). Frank and colleagues established the theoretical foundation two decades ago: competence is multidimensional, dynamic, developmental, and contextual—not a fixed endpoint but an evolving capacity (Frank et al., 2010). The architecture exists. The question is whether the profession will build systems that actually use it.
The current system does not. Most clinical settings lack a mechanism for matching patient complexity to clinician entrustment levels. The schedule is the schedule. The next patient on the list goes to the next available therapist. When that patient’s presentation exceeds the therapist’s current capacity, the system has no structural response — only individual coping. The therapist works harder, stays later, documents more carefully, and absorbs the cognitive and emotional cost of operating beyond the boundary of what they have been prepared to do.
Medicare reimbursement, adjusted for inflation, is roughly half of what it was thirty years ago (MedPAC, 2023). Volume pressure is the inevitable consequence. In that environment, the idea of matching complexity to entrustment level sounds like a luxury the system cannot afford. But the data suggest it is a cost the system cannot afford to avoid. Rodeghero and colleagues found that fellowship training — the level of preparation designed to handle the highest clinical complexity — improved both functional outcomes and treatment efficiency in a retrospective cohort of over twenty-five thousand patients (Rodeghero et al., 2015). The additional mentored practice of a fellowship creates a measurable clinical difference. Residency alone did not produce the same outcome gains, but it built the foundation the fellowship required.
The entrustment framework does not solve the reimbursement crisis. It does not add hours to the day or reduce the caseload. What it does is replace a system that cannot tell you where a clinician’s reasoning breaks down with one that can. It replaces blame with matching. It replaces ambiguity with a graduated picture of trust that updates as the clinician develops.
The resident finished the week on Monday morning, having seen eleven patients. Four of them were straightforward — the kind of clinical work where five months of residency training had produced rapid, efficient, independent reasoning. Three were moderately complex — cases where the resident’s reasoning was sound but slower, where the mentor’s presence in the debrief added something the resident could not yet generate alone. And four were genuinely difficult — presentations where the clinical picture did not resolve cleanly, where the resident needed guidance not because of insufficient knowledge but because the complexity of the case outpaced the reasoning architecture the residency had built so far.
The assessment form captured everything. Not a single number. Not pass or fail. A picture — honest, developmental, and specific enough to tell both the resident and the mentor exactly where the work still needed to happen.
The question had never been whether the resident was competent. The question was where, exactly, that competence could be trusted — and where it could not yet be trusted, not because of any failing, but because the development was still underway.
