The wrong default
When an ACO decides to take palliative care seriously, the conventional first move is to contract with an external palliative care provider. The pitch sounds clean: a specialized team handles a clinically demanding population the primary care practice can't service inside a thirty-minute visit, and the ACO pays a per-member-per-month fee or a capitated rate to outsource the problem.
In our work at Pearl Health, we examined this approach across a number of vendor partnerships and concluded that "contract with a vendor" isn't actually a strategy. It's one of three patterns, each appropriate in a different setting. Defaulting to the vendor contract regardless of what the practice already has is what produces the disappointing outcomes most ACOs experience with palliative care: over-medication, misaligned incentives, poor patient selection, and a cost differential between inappropriate external care and proper internal management that exceeded $72,000 per case in some of the situations we reviewed.
Palliative care isn't one thing. It's a population-health workflow that needs to be matched to the practice executing it. Three patterns work, in three different contexts.
Why this matters for the patient
Before the operating models: one thing is true regardless of which pattern fits a given practice. Palliative care, properly delivered, exists to align medical care with what the patient actually wants. A trusted clinician who knows the patient (who has built the relationship over time) is dramatically better positioned to help that patient navigate the trade-offs between aggressive treatment, comfort, time at home, and time with family. Without that relationship, the system defaults to maximum intervention. With it, patients consistently choose differently than the default, and live the time they have left more in line with what matters to them.
Think of palliative care the way you think of any other specialty. You have an orthopedist on your team for the moments when your shoulder flares up; you don't see them when your shoulder is fine. Palliative is the same. When pain, symptoms, complex decisions, or transitions flare up, the patient should already have a relationship with the clinician who's good at managing them. The earlier that relationship is established, the more it can do, for the patient and for the math.
That relationship is the substrate. Everything else in the operating model is about how to make it real.
Pattern A: Vendor partnership (where capacity doesn't exist)
Some practices have no realistic path to building internal palliative capacity. A multi-site primary care group in a rural geography. A small independent practice without the scale to hire a part-time palliative provider. An MSO that runs lean on clinical operations. For these practices, contracting with an external palliative care vendor is the right answer. The work doesn't exist otherwise.
The vendor partnership pattern has known failure modes, they're real, and worth being explicit about:
Over-medication approaches. External palliative providers paid on a fee-for-service or per-visit basis are economically incentivized to do more, not less. In a population already heavily medicated, "more" usually means additional symptom-management drugs, more frequent visits, and a clinical posture that prioritizes intervention intensity over goal-concordant restraint.
Misaligned financial incentives. A capitated palliative care vendor with a fixed PMPM has the opposite problem: every dollar spent on the patient is a dollar of margin lost. The economic pressure is to under-engage and let the most expensive cases drift back into the medical system as the ACO's problem.
Poor patient selection. External vendors run their own targeting algorithms, optimized for their economics rather than the ACO's. The patients who get enrolled aren't always the ones where palliative wraparound has the highest avoided-cost potential. They're the ones where the vendor's enrollment funnel has the lowest friction.
These failure modes don't make the vendor pattern wrong. They make it conditional. The vendor pattern works when (a) the practice genuinely has no capacity, (b) the contract aligns vendor economics to outcomes the ACO actually cares about, and (c) patient selection is jointly governed rather than left entirely to the vendor's funnel. Get those three things right and the model fills a real gap.
Pattern B: Pay the practice a PEPM (where capacity could exist but doesn't pencil)
The most underused pattern is the second one. Some practices already have the clinical sophistication and the people to do palliative care well. What they don't have is the economic model to make it work.
A practice we worked with at Pearl is a good example. The medical leadership is, in their own words, four-hundred-level end-of-life thinkers. The practice has been doing serious-illness conversations for years. They want to hire a dedicated palliative-trained provider (an MD or a mid-level under the direction of an experienced clinician) to staff a panel of the sickest two hundred patients in their attribution. They believe it's the right thing for the patient. They also believe, correctly, that it's a money-losing proposition under fee-for-service economics, because palliative time isn't reimbursed at anywhere near the rate of acute care visits.
For practices like this one, the right answer isn't a vendor. It's to pay the practice itself a PEPM (a per-engaged-patient-per-month payment) to staff and run their own palliative program internally. The economics of the PEPM are designed to cover the cost of the dedicated clinician across a defined panel of the sickest patients, where the avoided-cost math justifies the investment many times over.
The advantages of this pattern, when it fits, are significant. Trust is already established between the patient and the practice. The relationship that makes palliative care work doesn't have to be rebuilt with a stranger. Workflow integration is automatic, because the palliative work happens inside the practice's existing operations. Continuity is preserved across the patient's other care. And the practice's clinical team, not an external vendor's algorithm, decides who gets engaged and when.
The contracting work is more nuanced than a vendor deal. Payments tie to operational adherence: engaging a defined percentage of the target panel through monthly visits, completing POLST or advance directive forms, hitting documented quality metrics, so the PEPM isn't just an open-ended subsidy. Done right, the PEPM functions as an advance on shared savings the practice would otherwise generate, structured in a way that protects the ACO if execution falls short.
Pattern C: Signals and workflow integration (where capacity already exists)
A small number of practices already have everything they need. Internal palliative team in place. Workflow that can absorb high-acuity patients. Clinical team aligned on the value of early palliative engagement. For these practices, the highest-impact thing the ACO can do is feed them risk signals and let them act.
This is the lowest-cost pattern in the program. The ACO's claims-based algorithm identifies the seriously ill cohort: patients with severe single-organ disease, metastatic cancer, advanced dementia, or three-plus chronic conditions, plus a recent hospitalization, plus a functional limitation marker (home health, SNF, DME). The signal lands inside the practice's existing care-team workflow, automatically triggering a palliative consult. No vendor contract. No PEPM. No new infrastructure. Pearl's modeling of this pattern alone captured 10–20 percent of the program's total opportunity at minimal incremental cost.
Pattern C is rare because the practices that qualify for it are rare. But when they exist, they're the highest-yield, lowest-effort engagement in the entire palliative care program.
What's the same across all three
What changes across the patterns is the operating model. What stays constant is everything that makes palliative care work clinically.
Patient identification: the same algorithm runs underneath all three patterns. A claims-based serious-illness definition combining diagnosis criteria, recent inpatient utilization, and claims-based functional-limitation proxies. The cohort it identifies has roughly 28% one-year mortality. Whether that cohort gets engaged by an external vendor, an internally funded palliative provider, or an existing palliative team responding to a signal, identifying them precisely is the prerequisite.
Conversation skills training: primary care providers across all three patterns benefit from evidence-based conversation frameworks (Ariadne Labs Serious Illness Conversation Guide, the NURSE framework, VitalTalk), because much of the early values-elicitation and POLST completion work happens upstream of any specialist palliative engagement. Trained clinicians find these conversations significantly more rewarding and feel more confident having them. The training respects existing clinical expertise. These are scripts and decision frames, not remedial training.
Goal-concordant care as the actual outcome: the financial savings in palliative care follow the alignment of medical care with what the patient actually wants. They don't precede it. Programs that chase savings without doing the relational work of understanding patient values produce neither.
What this generalizes to
The meta-lesson isn't about palliative care specifically. It's about how clinical strategy in VBC actually works. The right operating model depends on what the practice already has, what the population needs, and what the economics make possible. Strategies that pretend to be universal usually aren't. The work is in matching the model to the practice, and being willing to run multiple models simultaneously, across the same network, when that's what the situation calls for.
This is one of three posts on Pearl Health's end-of-life program. The companion pieces examine why POLST is the right wedge for primary care, and how to curate a hospice network, which together form the three layers of a serious end-of-life program.