The Taxonomy Is Lagging the Trials

Palliative care's buckets are overflowing with the arrival of new evidence.

Not long ago, Jennifer Temel's group published a stepped palliative care trial in JAMA that I'm struggling to categorize. Specialist clinicians delivered the care, but they didn't show up monthly—they showed up when the patient's quality of life dropped, when treatment changed, when the patient was hospitalized. Patients in the stepped arm averaged 2.4 visits over 24 weeks. The early-monthly arm: 4.7.

Quality of life was preserved with fewer visits by the same specialist team.

So what was that? "Specialist palliative care" at half the dose? "Primary palliative care" delivered by fellowship-trained clinicians? The answer, if you're being honest, is that it doesn't fit either box—and the boxes are the problem.

The Quill–Abernethy framework from a 2013 NEJM perspective split the palliative care world into two categories: primary (the basics, done by any clinician, and which I've always preferred to call "generalist") and specialist (the hard stuff, done by the fellowship-trained team). That binary has organized our workforce advocacy, our training programs, our research agenda, and our scorecards for 12 years. It was the right scaffold for the problem it was built to solve.

Three trials in the last two years just built something it can't describe.

What the Binary Got Right

In 2013, the field faced a novel math problem. Lupu's workforce analysis estimated the gap between HPM physician supply and demand at somewhere between 6,000 and 18,000 physicians. One hundred twenty fellowship slots a year. A patient denominator that Morrison and Kelley were about to widen from "the dying" to "the seriously ill."

The binary was the right answer to that math. If you can't put a specialist in front of every patient who needs palliative care, you need non-specialists to handle the basics. That logic produced PCHETA's two-track funding model, the CAPC Scorecard, and every workforce argument we've made to CMS and Congress for the last decade. Clean taxonomy, clean policy story, real results.

That was good work. It got us here.

But here is 2025, and the evidence doesn't fit the 2013 categories anymore. Three recent trials show why.

Three Trials That Broke the Boxes

Temel's stepped palliative care (JAMA 2024, N=507). I started with this one because the design makes the binary incoherent. Stepped care isn't "specialist" in the traditional sense—the team steps back unless triggered by a sentinel event. It isn't "primary" either—the clinicians are fellowship-trained, the initial assessment is comprehensive, the step-up threshold is a validated PRO. What it actually is: a dose-response model applied to specialist palliative care, titrated to the patient's state over time.

The binary has no category for variable-dose specialist intervention. None.

And here's the finding the field needs to sit with: hospice length of stay in the stepped arm was 19.5 days versus 34.6 in the early arm. Noninferiority was not demonstrated for that outcome. The stepped model preserves quality of life with fewer visits, but we don't yet know what it costs at the end. That trade-off is real, it's uncomfortable, and it's completely invisible inside a framework that treats all specialist contact as categorically equivalent.

Parikh's BE-a-PAL (JAMA Network Open 2025, N=562). Ravi Parikh's group embedded an EHR algorithm in a Tennessee community oncology network that auto-generated specialist palliative care referral orders as a default. Clinicians could opt out. They had to actively do so.

The consultation rate: 44% versus 8%. Opt-out rate under 10%.

Read that again. When you flip the default, oncologists allow the referral almost every time. The gap between "patients who need palliative care" and "patients who receive it" isn't primarily a workforce gap or a training gap. It's a behavioral gap. This surprises me (and I'm betting you) not at all. The binary assumes the lever is workforce supply (more specialists) and training supply (teach generalists the basics). BE-a-PAL says the lever is infrastructure: defaults, algorithms, nudges.

One more thing. Despite quintupling the consultation rate, the trial showed no improvement in patient-reported outcomes. The single initial consult wasn't enough. Referral isn't integration. The binary treats the referral as the intervention. The data say it's a gate, not a destination.

Aldridge and Morrison's TIER-PC (NCT06228209). Still enrolling, but the design is the point. TIER-PC builds a three-tier delivery model—community health worker, social worker, APN/MD—layered as needs intensify. The protocol describes "the right level of care to the right patients at the right time." The trial name says it: tiers. The Mount Sinai group built what our taxonomy needs to contemplate now.

Meanwhile, in Countries That Use More Than Two Boxes

The binary is an American invention. The European Association for Palliative Care uses a three-tier framework: palliative care approach (every clinician), general palliative care (clinicians who frequently encounter palliative patients but whose primary work is something else), and specialist palliative care. NHS England operationalizes the same tiers for commissioning.

The Dutch framework—van der Padt-Pruijsten et al., BMC Palliat Care 2024—goes further: generalist, specialist, and expert, because even the specialist tier is heterogeneous. Their survey data are revealing: clinicians with identical training self-identify as generalist, specialist, or expert depending on context. The label tracks the moment, not the credential.

We spent a decade debating whether non-specialists can "do" primary palliative care. Our colleagues across the Atlantic spent the same decade building competency tiers that describe what happens at the bedside. We could learn something.

The Reframe: Competency at the Moment

The unit of analysis should be the competency needed at the moment of care, not the category of clinician delivering it.

A hospitalist is sitting with a family at 2 a.m. The question is whether to continue vasopressors. The skill that matters right now—right here, in this room—is structured recommendation-giving calibrated to that family's decision-making style. Lai et al. (J Palliat Care 2023) showed that residents who provide recommendations in goals-of-care conversations learned the skill by watching it done on palliative care or critical care rotations. Not from didactics. Not from a training module labeled "primary palliative care." The skill transferred through proximity to the specialist tier, then lived in the hospitalist's hands at the moment it mattered.

An oncologist's patient with bone metastases has a pain score of 8. The prescription drops into a pharmacy system that won't fill it. The skill needed is prescriber confidence, state-law literacy, and the nerve to pick up the phone. Admane et al. (JCO Oncol Pract 2025) documented a five-fold decline in opioid doses prescribed by oncologists for cancer pain between 2016 and 2021. That isn't a specialist consultation failure. It's a competency failure distributed across the system, where the right skill at the right moment can live in many different hands.

Whose hands does this competency need to be in, for this patient, at this hour? Sometimes that's the fellowship-trained physician untangling delirium superimposed on opioid neurotoxicity. Sometimes it's the ICU nurse recognizing the moment for a family meeting. Sometimes it's the oncologist prescribing adequate analgesia without flinching. The precision unit isn't the referral. It's the competency at the moment.

Sedhom and Parikh's precision palliative care framework (JCO 2023) gets close to this—matching specific team members to specific needs using PRO-based screening and behavioral nudges. Good framework. But they apply precision logic within the specialist team while preserving the binary boundary around it. The move the field needs is to apply that same logic across the boundary: the entire workforce, tiered by competency, with infrastructure that routes the right skill to the right moment.

What This Means for Workforce Advocacy

I think we all know that I'm writing from inside the palliative policy house.

The binary has served the field's advocacy well. But, for what we are facing now, the advocacy becomes stronger—not weaker—under a continuum, because a continuum specifies what specialists are uniquely for. Under the binary, the specialist tier is defined by subtraction: everything that isn't "basic." Under a continuum, it's defined by expertise: refractory and complex symptom management, complex prognostic communication, team design and leadership, system architecture, training and education of the tiers below. That's a sharper argument to make to CMS. And it's a more defensible argument against the scope dilution I wrote about in The Monster We Built.

The substitution evidence—the idea that we close the gap by training non-specialists to do what we do—is weaker than the field has acknowledged. The CONNECT trial (Schenker et al., JAMA Intern Med 2021), an oncology nurse-led primary palliative care intervention in advanced cancer, found no improvement in patient-reported outcomes. Ernecoff et al.'s comparative analysis (J Palliat Med 2020) found primary palliative care interventions less comprehensive, less likely to show benefit on physical symptoms, and at higher risk of bias than specialist interventions. "Everyone does a little palliative care" has not been the scale solution it was advertised to be.

A continuum framework doesn't abandon non-specialist training. It specifies it: which competencies, for which clinicians, at which moments, with what infrastructure. That's a better model than "train everyone in primary palliative care and hope." And it's a better clinical architecture than "refer or don't."

Where I Might Be Wrong

No head-to-head trial has compared a continuum delivery model to a binary one. TIER-PC will eventually be informative; it hasn't reported yet. The international tiered frameworks operate in health systems structurally different enough from the U.S. that direct policy translation is speculative at best.

The CONNECT null result certainly doesn't generalize to all primary palliative care designs. The evidence base is thin and heterogeneous enough that confident generalizations in either direction are premature.

And the binary's policy clarity is real. "Specialist" and "generalist" translate into billing codes, fellowship accreditation, and legislative language. A continuum is harder to legislate. I don't dismiss that cost.

But the taxonomy should describe the work, not constrain it. Right now, our best trials are describing work that the taxonomy can't see.

Three trials in two years—Temel, Parikh, Aldridge and Morrison—from three respected research groups in the field. A dose-response specialist model. A behavioral infrastructure model. An explicit tiered model. None of them sorted the world into "primary" and "specialist." All of them organized around the question the field must now be asking: what does this patient need, who can deliver it, and when?

How do we get the right intervention to the right patient in the right moment delivered by the right team member?

That question doesn't have a binary answer. It has a continuum. And if "precision palliative care" is going to mean something beyond a JCO commentary—if it's going to be an operating principle—then the taxonomy has to catch up to the trials that are already building it.

Next week: where the binary came from, what it was designed to solve, and why 2026's problems need a different scaffold. If you thought this piece was about nomenclature, the follow-up is about strategy.

I am a palliative care physician, educator, and professional strategery expert. Known for turning rounds into rants and rants into teaching points. Rounds & Rants represents my views — not those of any institution or professional membership organization where I hold a role. I don't write on their behalf and they don't vet what I publish.