You Can't Move Care Upstream by Dropping People Back Above the Rapids

The healthcare industry has been talking about "moving care upstream" for as long as anyone in value-based care can remember. It is the metaphor of choice in every keynote, every investor deck, every CMMI model rollout—deployed by executives explaining why their latest initiative will finally do the thing the industry has been promising for decades. They likely believe it. The buzzword is a comfortable fit for almost any work an organization wants to describe.

And yet—measured by what actually matters, the rates of preventable chronic disease, lifestyle-related morbidity, behavioral health crises—almost nothing about the trajectory of population health has changed.

The metaphor itself explains the problem if you span out to see the whole picture. Most medical intervention happens after the patient has already experienced an adverse physical event—the heart attack, the stroke, the diabetic complication, the depressive episode. Think of it as the pool at the bottom of a waterfall. There is the violent drop-off, then the water calms. The pool is where most healthcare intervention is happening today: stabilize the patient, manage the chronic condition, prevent the next acute episode.

The industry's current "upstream" initiatives pull patients out of the pool and place them back into the rapids above the waterfall. The patient is alive, the fall is over, they get a little more navigable water before the next adverse event. But the rapids are still there. The waterfall is still ahead.

Genuine upstream care means going further up the river—all the way to the calmer waters above the rapids. There, we can build a dam and create a navigable route around the disruption entirely. Fewer patients ever reach the rapids in the first place.

What the industry currently calls "upstream"

Three patterns dominate the industry lingo around "upstream" care. Each is real work. None of them is upstream care in the sense the word implies.

Top-of-license and expanded primary care. Primary care physicians are taking on more chronic disease management, more care coordination, more behavioral health screening. Medical assistants and nurses are absorbing work that physicians used to do. Nurse practitioners and physician assistants are running independent panels. This is genuinely valuable work, and it improves access in a system that desperately needs it. But the patient already has the chronic condition. The disease is already present. The intervention is downstream of the etiology—the chronic kidney disease, the type 2 diabetes, the heart failure, the depression—has already developed by the time the expanded primary care team gets to work.

Site-of-service shifts. Procedures move from inpatient to outpatient. Outpatient procedures move to ambulatory surgical centers. ASC procedures move to physician offices. Total knee replacements that required a five-day hospital stay in 2005 now happen as same-day discharges. There are real cost savings, real patient experience improvements, real quality gains in these shifts. But the surgery, the infusion, the procedure was already predetermined at the point of intervention. The intervention point has not moved upstream—only the location has changed.

Earlier disease detection. Risk stratification tools identify chronic conditions in earlier stages. Annual wellness visits surface conditions that would have gone undiagnosed for years. Panel management catches the patient with prediabetes before they progress to type 2. This is closer to genuine upstream than the other two, but it is still operating on a disease that has already developed. The metabolic dysfunction has already started its decade-long drift. The depression has already taken hold. The hypertension is already damaging the cardiovascular system. Catching disease earlier is better than catching it later. But disease prevention is the standard upstream framing—catch the metabolic markers before diabetes develops. Health promotion is the framing past that—build the lifestyle, environment, and behavioral conditions where the metabolic markers don't drift in the first place. Disease prevention is great. Health promotion is better.

Each of these is real work and worth doing. None of them is upstream care in the sense the word implies. They are reactive care made more efficient.

What upstream actually means

Upstream isn't a single intervention. It's a time horizon.

Reactive care operates on a timeline of weeks to months—the acute episode, the chronic exacerbation, the hospitalization, the recovery. Industry-upstream care operates on months to years—the screening cycle, the chronic disease management plan, the care coordination across the trailing edge of disease. Genuine upstream care operates on years to decades. It intervenes on the conditions that, left alone, will eventually drive someone into the reactive healthcare system.

That framing matters because it clarifies what upstream care is and isn't. Upstream care isn't anti-healthcare-system. People should be engaged with the system—they should be in regular contact with primary care, they should have someone who knows them medically, and they should have a relationship with their care team. What needs to change is the mode of that engagement: from acute and reactive, which is what consumes nearly all of healthcare's current capacity, to proactive, promotional, and preventive.

Concretely, genuine upstream care looks like this:

• Identifying metabolic function and dysfunction—insulin resistance markers, inflammatory signatures, sleep patterns—years before someone is diagnosed with type 2 diabetes (a use case AI is well-positioned to support, given the right longitudinal data substrate)
• Detecting cognitive decline trajectories years before Alzheimer's symptoms become clinically apparent
• Recognizing the behavioral and environmental patterns that predict cardiovascular events in people who currently have no clinical heart disease
• Addressing the social determinants—food access, housing stability, economic security, environmental exposures—that the National Academy of Medicine estimates account for 80 to 90 percent of the modifiable contributors to health outcomes, while medical care accounts for only 10 to 20 percent.¹

The healthcare system isn't structurally set up to do any of this at scale. That's the problem.

Barriers to building the dam

Four interlocking problems prevent the healthcare system from doing genuine upstream care. None of them is solvable in isolation. Each is a real constraint, and each reinforces the others.

Economics

The economics of upstream care are at odds with the economics of the healthcare system in four distinct ways.

First, fee-for-service reimbursement structurally doesn't reward care that prevents the need for care. The model pays for events—visits, procedures, admissions. Upstream care produces non-events—the diabetes that doesn't develop, the heart attack that doesn't happen, the hospitalization that's avoided. FFS has no mechanism for paying for things that didn't occur. Value-based care arrangements try to fix this, with varying degrees of success.

Second, physician compensation is inverted. Specialists earn two to three times what primary care physicians earn. The talent flows downstream because that's where the money is. Until the compensation scale inverts—until the doctors doing the upstream work are paid as well as the doctors doing the downstream work—the workforce that would do upstream care at scale doesn't get built.

Third, supply and demand are mismatched at the structural level. The primary care physician shortage is well documented. Residency match rates for primary care specialties have been declining for years. Burnout is driving experienced PCPs out of the field. The supply side is contracting while demand for upstream-capable primary care is growing. Even organizations that want to do upstream care can't reliably staff it.

Fourth—and this is the sharpest point—patient mobility across insurance products breaks the longitudinal economics of upstream care. A patient cycles through commercial insurance during their working years, possibly through the ACA marketplace during a transition, possibly through Medicaid during a period of low income, and eventually onto Medicare. Within each phase, they may change insurers multiple times. The problem isn't that risk contracts need to follow patients across insurers—patients don't change primary care providers as often as they change insurance. The problem is that providers don't have value-based contracts with every insurer their patients move through. Outside Medicare Advantage and select Medicaid markets, commercial insurance's value-based care ecosystem is underdeveloped. That means a primary care organization committed to upstream care still loses the financial benefit of those investments every time a patient transitions to a payer the organization doesn't have a risk contract with. Until value-based contracting penetrates commercial insurance the way it has penetrated Medicare Advantage, the longitudinal economics of upstream care don't close.

Data — the chicken-and-egg problem

You can't deliver the right upstream care without the right data. Claims and EHR snapshots don't capture the years of metabolic drift, behavioral patterns, dietary choices, sleep quality, social connection, financial stress, and environmental exposure that predict disease. The signals that matter for genuine upstream care exist almost entirely outside today's clinical data infrastructure.

But you can't capture that data without sustained patient engagement. And you can't get sustained engagement without already doing the upstream care that builds the relationship. The loop is the problem.

Returning to the river analogy: think of the dam as the data infrastructure. You can't build it without continuous engagement, and you can't get continuous engagement without partial dam-building already underway. The way out isn't a single intervention. It's sequencing—building the infrastructure in stages, with each stage enabling the next stage of engagement, which enables the next stage of infrastructure. This is what we mean when we say intelligence first is the precondition for upstream care. The data substrate has to be built before the rest of it works.

Culture

The healthcare system has lost ground in the upstream conversation, and physicians know it. The common refrain from medical organizations is some version of this:

That framing isn't always fair—much of the wellness and functional medicine ecosystem is doing serious work on diet, sleep, stress, and longitudinal monitoring that primary care doesn't have the time, training, or reimbursement to do. But the underlying observation is right: the upstream conversation patients want to be having about their long-term health is happening almost anywhere except in their fifteen-minute clinical encounter.

"Sick care" is another tired phrase—tired because it's right. The healthcare system is structurally oriented around the clinical encounter and procedural volume. Genuine upstream care happens between encounters, often without one. The current operating model has no good place to put the work, no good way to compensate the people doing it, and no good way to measure the outcomes over the time horizons that matter. The orientation has to invert.

Patient engagement

Roughly a third of Americans don't have a primary care provider.² Genuine upstream care requires sustained engagement with the population most disconnected from the system. The disengagement problem isn't peripheral to upstream care—it's central. You cannot do upstream care for people who don't have a relationship with the system, and the people most likely to benefit from upstream care are disproportionately the people without that relationship.

But—and this is where the constructive turn begins—some primary care models have figured out how to solve this.

What upstream-ready actually looks like

A handful of primary care organizations have already figured out what the system as a whole hasn't. They're navigating their patients' diet, exercise, sleep, social connection, and financial stress as the central work of primary care—not as adjacent services. The check-points along their patients' route are visible in the model itself:

• Capitated economics that align incentives with health outcomes rather than visit volume
• Care teams that go beyond the physician-MA dyad—health coaches, behavioral health specialists, community health workers, social workers, pharmacists, and care navigators
• Community programming that builds engagement outside the clinical encounter—tai chi classes, yoga, group medical visits, social gatherings, diet education, fitness programs, and healthy incentive programs
• Continuous access models—same-day appointments, longer visits, twenty-four-hour patient support lines, asynchronous communication, and connected wearable devices
• Population-level investments rather than encounter-level economics—investing in the longitudinal conditions and relationships that produce health, on a time horizon that doesn't show up in any single quarter's P&L

The named examples are familiar. ChenMed reports that its patients have 33 percent fewer hospitalizations and ER visits than comparable patients.³ Oak Street Health reports a 44 percent reduction in hospital admissions compared to Medicare benchmarks.⁴ CenterWell, Humana's senior-focused primary care subsidiary, publishes similar outcomes data through its parent company's value-based care reports. The direct primary care movement is demonstrating that engagement at scale is possible outside traditional reimbursement structures.

The model works. The barriers to scaling it aren't clinical—they are the four interlocking economic, data, cultural, and engagement problems described above. The clinical model is established. The infrastructure is not.

The work ahead

There's a trilogy of arguments emerging across our recent writing. Intelligence first—the data substrate that everything else depends on. Upstream care—the goal that data substrate enables. And, coming next, evidence-based and value-based medicine working in tandem as the operational model that makes both possible and enables a Kaizen-style continuous improvement loop driven by data.

We keep pulling people out of the pool below the waterfall and dropping them back above the rapids. It is real work, and it is not the same thing as moving care upstream.

Building the dam—the data infrastructure, the longitudinal economics, the cultural shift, the engagement models—is the actual work. None of it is happening at scale yet. Most of it is being built piece by piece by a small number of organizations who have the patience and the capital to wait for results that show up over years instead of quarters.

The MSSP and ACO REACH Explorers exist to help organizations understand where they sit in that progression. If you want to talk about where your organization fits in the upstream conversation, or where your data infrastructure is and isn't ready for the work ahead, we'd like to talk.