The gap
American medicine is extraordinary at saving lives. The tools, treatments, and technologies available in a modern hospital are genuinely remarkable. What we're not good at is recognizing when it's time to do less than the maximum, and the result is that many patients suffer unnecessarily at the end of their lives because a system built to fight keeps fighting, even when the patient would choose otherwise.
ACOs have a strong financial incentive to care about this, but most are focused on readmissions, risk coding, and quality gap closure, work that's more operationally familiar and easier to measure. That's understandable. End-of-life is clinically complex, emotionally difficult, and hard to systematize. But for the seriously ill patients who drive disproportionate spend in the final twelve months of life, the savings from getting this right dwarfs what you'll capture from readmission prevention and risk adjustment combined, and the patient outcomes are better, because care actually aligns with what people want.
The question is where to start. Our answer, built and tested at Pearl Health across a pilot now scaling to 225,000 lives: start with POLST.
Why advance directives aren't enough
Most ACOs that do address end-of-life focus on advance care planning: getting providers to document an advance directive. That's necessary. It gets patients thinking about what matters and gives families a framework for decisions.
But advance directives are conditional documents. They say things like: "If I have a terminal condition with no hope of recovery, I do not want CPR." That conditional creates a gap that swallows the document's usefulness in the moment it's supposed to work.
A 68-year-old woman has stage IV colorectal cancer, well-managed on FOLFOX chemotherapy, with a two-to-three-year prognosis. Her advance directive says: no CPR if terminal with no hope of recovery. After an infusion, she develops a chemo-induced cardiac arrhythmia (QT prolongation degenerating into ventricular fibrillation) and collapses at home. EMS arrives. She's unconscious and pulseless.
Is this terminal? The oncologist would say no: the cancer is managed, and the arrhythmia is a reversible drug side effect. Shock her heart back and she walks out in two days. But the ED team doesn't know her chemo regimen, doesn't know her QTc history, and doesn't have time to sort it out. So they do what ED teams always do when the document is ambiguous: they run the code. Full resuscitation. The advance directive is effectively ignored, not out of malice, but because it doesn't give an unambiguous order.
What makes POLST different
A POLST (Physician Orders for Life-Sustaining Treatment) is not conditional. It's a set of standing medical orders, signed by the patient's physician, in effect right now. Not "if terminal." Now.
A POLST that says "Do Not Resuscitate" means do not resuscitate, irrespective of diagnosis or circumstance. It doesn't require the ED to evaluate whether the underlying condition qualifies. It's a physician's order, not a preference document. And there's malpractice risk if a physician sees a valid POLST and disregards it. POLSTs are respected because they have to be.
The same patient with a POLST has a completely different experience. If she and her physician had decided she did not want CPR under any circumstances, the POLST is honored. If she'd decided she did want full treatment, because the arrhythmia is treatable and she has years left, the POLST says "Full Treatment" and the code runs. Either way, the care that happens is the care she chose.
Advance directives are still necessary: they name a healthcare proxy, capture values, prepare the family. But an advance directive without a POLST won't be followed when it matters most. You need both. If you have to pick where to start, start with POLST.
The math at scale
Per patient, the average TME reduction from a completed POLST is roughly $1,800. That's an expected value: for some patients, the POLST prevents an $80,000 ICU admission they didn't want; for others, it never gets activated. But the variance cuts only one way, because there is no clinical downside to completing a POLST. It's just writing down what the patient wants. The floor is zero harm. The ceiling is a prevented hospitalization, an avoided ICU death, and tens of thousands in savings. When you're running the expected-value math across the sickest 10–15 percent of an ACO's population, those averages add up fast.
We built the program at Pearl around three components. A $150 value-based payment per completed POLST, on top of the existing ACP CPT code (99497, ~$86 from Medicare), so the practice gets paid roughly $236 for the conversation. An AI-assisted intake workflow that processes scanned POLST forms, matches them to patients, and triggers payment without a per-form human in the loop. And a patient-identification feed: a risk-stratified list of seriously ill patients where the conversation has the highest expected value.
POLST scales in a way most clinical programs don't. The form is structured and machine-processable, so the operational infrastructure gets built once and reused at every new practice. The first practice is expensive. The fiftieth is nearly free. That compounding is what made the leap from a five-practice pilot to 225,000 lives viable.
The wedge, not the program
POLST gets a program off the ground. It doesn't finish the job. A serious end-of-life strategy also requires palliative care capacity matched to each practice's readiness, and a curated hospice network to prevent the benefit abuse that quietly destroys ACO economics. But POLST is the entry point that makes both tractable: practices that complete POLSTs build the muscle for harder conversations, and patients with POLSTs on file become the identifiable cohort where deeper interventions pay off most.
This is one of three posts on Pearl Health's end-of-life program. The companion pieces cover why palliative care is not a single strategy, and how to curate a hospice network, the largest per-patient savings lever in the entire program.
We're not forcing anything on anyone. We're paying the doctor to ask the patient what they want, and writing the answer down somewhere it can't be ignored.