SERIOUS ILLNESS · HEALTH SYSTEM STRATEGY
TIMING IS EVERYTHING
The most important decision in serious illness is made too late — and fixing it isn’t a program you run once, but a change in how your system practices medicine, and keeps practicing it.
There’s a conversation medicine keeps missing: what matters most to this patient, and what kind of care would honor their values and priorities. It’s the most decision-relevant exchange in all of medicine, and the one we’re least equipped to have. It’s uncomfortable. It isn’t reimbursed like a procedure. And there’s rarely an obvious moment to begin. So it gets pushed to the crisis — the ICU at 2 a.m., when it’s far too late to be a real choice.
We started Empower Hope to close that gap, and we’ve come to believe something that reframes the whole challenge for anyone running a health system or risk-bearing plan. This was never a failure of compassion or will. It’s a failure of timing — and timing isn’t something you fix once; it’s a discipline you build into how an institution practices, and sustain. Not a tool bolted onto a broken moment, but a re-imagining of how a system knows and honors what its patients want — early, continuously, for everyone. Not one patient rescued, but a standard of care that stays on.
The window is real — and invisible
Ask any clinician when the serious-illness conversation should happen and you hear the same tension. Too early feels alarming — the patient “isn’t sick enough yet.” Too late means the ICU, if it happens at all. One physician put it more sharply than we ever could:
“I always know the conversation was overdue — right after the patient can no longer have it.”
And the right window is rarely a single moment. As one doctor put it, “Advance care planning is a process… and multiple conversations are best.” Ideally they begin early, while a patient is well enough to explore what matters, and continue as health changes. The window is also invisible in the moment, buried in a trajectory no single visit reveals. By the time the need is undeniable, the patient is often too sick or too frightened to weigh in, and an exhausted family is left to guess.
That’s why “try harder” has never worked. More training, more reminders help at the margins but never change the math. We’re asking overloaded clinicians to add an emotionally heavy, time-intensive task, unprompted, with no trigger telling them the moment has arrived. When the right thing takes extra work at the worst possible time, it doesn’t get done. Blame lands on the clinicians, but this was never a people problem. It’s a design problem, and it’s solved by redesigning the system that sets them up to fail.
Why timing lands on your P&L
For a fee-for-service hospital, bad timing is mostly human and clinical. For a risk-bearing organization — an ACO, a Medicare Advantage plan, a capitated system — it’s also financial, and enormous. Roughly a quarter of Medicare spending for people over 65 goes to their final year of life, and more than half of that is inpatient. Some is essential and humane. But a meaningful share pays for aggressive care that patients, asked clearly and early, would have declined. We deploy our most intensive resources at the exact moment our tools for knowing what the patient wants are weakest.
Meanwhile, only about one in three American adults has documented what they’d want. Those preferences usually exist — we just don’t capture them until it’s too late to matter. If you carry downside risk on a serious-illness population, that’s your largest, most concentrated pool of spending, flowing through the least-managed door in your system: the ER. One plan medical director asked us the sharpest question we’ve heard:
“How do I reach these members before the system reaches them?”
This is not a cost-cutting exercise, and should never be sold internally as one. The savings from goal-concordant care don’t come from denying wanted treatment — they come from not delivering unwanted treatment: the ICU stay the patient would have refused, the fourth-line therapy with real suffering and vanishing benefit. Start from “what does this person actually want?” and the humane answer and the economical answer point the same way far more often than we admit. As one health-system leader told us: “We weren’t spending too much. We were spending it on the wrong things at the wrong time.” Once they could see which patients were entering the window and capture their goals, the spending didn’t just drop — it moved, toward home, toward comfort, toward what people wanted.
The rare win-win
Almost every cost initiative in healthcare trades off against experience — narrow networks, prior authorization, utilization management all create friction somewhere. Goal-concordant serious-illness care is the exception. Help patients get the care they actually want near the end of life, and they and their families report better experiences, clinicians deliver care they believe in, and unwanted intensive utilization falls. Experience up. Morale up. Cost down. Name another initiative that moves all three the same direction.
That includes a driver most retention strategies ignore. Much of clinician burnout isn’t overwork — it’s moral distress, the specific exhaustion of delivering care you believe is wrong. A veteran ICU nurse told us her hardest shifts aren’t the busy ones; they’re the ones where she’s quietly certain that if anyone had asked in time, none of it would be happening. Every timely serious-illness conversation is also, quietly, a retention intervention.
Make the moment visible — carefully
This is where technology has a role, and it’s worth being precise about it, because the fear is legitimate. No one should want an algorithm breaking hard news, prognosticating to a patient, or deciding what someone’s values mean for their care. That is sacred, human work, and it must stay human. What technology should do is humbler and far more useful: notice the window. A model that surfaces when a patient is likely entering their final year doesn’t replace clinical judgment — it gives a clinician a reason to walk into the room today instead of assuming there’s still time.
“Machine finds the moment; human owns the meaning.”
That division of labor is the design principle that makes it trustworthy — and it demands discipline. Any model that influences who gets a serious-illness conversation has to be interrogated for bias from day one — representative data, subgroup testing, clinician oversight on every output — and built to live inside the workflow, augmenting the clinician rather than issuing verdicts. A model that flags some populations later than others deepens the very inequities we should be closing. Skepticism here isn’t an obstacle to the work; it’s a prerequisite for doing it responsibly — and for building something a system can trust enough to keep.
Measure the thing that actually matters
If timing is the game, measurement is how you know you’re winning — and most organizations measure the wrong thing. ICU death rate, length of stay, readmissions, hospice use: useful proxies, but they dance around the real question. Did the care this person received match the care they said they wanted? A patient who wanted aggressive treatment and got it is a success, even if they died in the hospital. A patient who wanted to be home and died in the ICU is a failure, even if the numbers look average. The truth-telling metric is concordance — not satisfaction, since families in crisis can be “satisfied” with care that didn’t match the patient’s goals simply because the team was kind.
Concordance has stayed unmeasured for one reason: the input, the patient’s actual goals, usually isn’t captured as structured, retrievable data. It lives in free-text notes, if it’s documented at all — invisible at the moment of decision. That’s a data-architecture choice, and it’s fixable. Capture values as structured data, early, and three things become possible at once: clinicians see them at the point of care, quality teams can measure concordance, and the system can finally learn. That’s what turns a tool into infrastructure.
Infrastructure, not an intervention
Here’s the distinction that decides whether this works. An intervention helps the patients in front of it and ends; infrastructure changes what’s normal for everyone who comes after, and doesn’t stop. And this isn’t theory. Across ten of the nation’s leading cancer centers, a multi-year collaborative — the Improving Goal Concordant Care initiative — built structured goals-of-care documentation into the EHR, and documentation roughly doubled (18% to 37% of all decedents, 25% to 52% among inpatients); a parallel communication-skills program lifted patients’ own “heard and understood” scores from 79.5 to 83.8. Tellingly, the collaborative’s 70% goal wasn’t reached — proof this isn’t a switch you flip but a capability you stand up and keep feeding, which is exactly why it takes durable infrastructure, not a pilot, to finish the job.
Scaling the humane thing has always hit the same rock. Palliative care teams do extraordinary work — but as one palliative physician has argued, “scaling palliative care alone is not a viable strategy to support the entire seriously ill population.” The only way to reach everyone is to change the operating system underneath the care: embed the signal, the conversation, the structured record, and the feedback loop into everyday workflow, so clinicians “work smarter and more efficiently instead of harder.” That’s the cultural shift we mean. Culture is what a system does by default, and today the default is silence until crisis. We’re out to make the opposite true — that knowing and honoring what a patient wants is simply how serious illness is practiced here, for every patient, improving with each one.
What we’re asking of leaders
Picture it working as it should. A patient enters the window where illness is becoming life-limiting. Before any crisis, the team gets a gentle, accurate signal: now is the time. A prepared clinician has an unhurried conversation — the first in a series — about how this person wants to live with the time they have, within the bounds of medically appropriate care. Those goals are captured as structured data and revisited as circumstances change. Months later, at 2 a.m. in the ER, the on-call team pulls up not just labs but the patient’s own words — and the care that follows matches the person. Every piece of that exists today; what’s been missing is stitching it into one reliable flow.
For a leader carrying risk on a serious-illness population, the question is no longer whether to care. It’s whether you can see these patients in time to help — and whether you’ll put concordance on the scorecard beside the metrics you already watch. The conversation medicine keeps missing is the one your most expensive, most human decisions turn on. The tools to make its timing visible finally exist, and the evidence that a system can be built around them is on the record. The only question left is this: are you OK with the status quo, or do you want to partner with us to make goal-concordant care a reality?
■
We’re ready when you are
Empower Hope is selecting a small number of forward-thinking health systems and plans to prove this out; a focused, well-measured partnership built to move concordance, patient experience, and cost together, with the metrics to show it. The organizations that move first will set the standard for what serious-illness care can look like. If that’s your world, let’s talk.
EMPOWER HOPE Start the conversation: empowerhope.ai info@empowerhope.ai
Bringing humanity back to care.