How to divide shared savings without dividing the practice

The Medicare Shared Savings Program check arrives in the fall. It is larger than anyone expected, and within a week, the collegiality of your partnership is under more strain than it has seen in years. Everyone agrees the practice earned it. No one agrees on who earned it.

Most practices default to one of two answers, and both are wrong.

The first is to treat shared savings as profit and distribute it by ownership percentage. That is clean, it is what the operating agreement already says, and it quietly tells every nonowner physician that the work they did to produce the savings belongs to someone else.

The second is to distribute by production — work relative value units, collections, revenue per patient — because those are the metrics already built into the compensation formula. That one is worse, because it rewards precisely the behavior that shared savings exist to discourage.

Savings do not live on the revenue line

Savings arise from reduced total cost of care: avoided admissions, fewer visits to the emergency room, lower post-acute utilization, tighter referral patterns, better medication management, and timely follow-up after discharge. These savings are earned across the entire horizontal spectrum of care, and most of that work never surfaces as revenue per patient; the phone calls, portal messages and care-management outreach that prevent the next admission generate no billable encounter at all.

Revenue per patient is not a reliable measure of a physician’s individual contribution. A physician who spends more time with each patient may generate more revenue per visit but see fewer patients per day, leaving fewer encounters to cover the same fixed overhead. Revenue from ordered services can also reflect the panel’s clinical needs more than the physician’s effort. And neither revenue stream is produced by the physician alone: Both depend on staff, facilities, technology, and workflows funded by the practice. Rewarding physicians solely on revenue per patient can therefore overstate individual contribution by crediting them for patient complexity and shared resources without accounting for visit volume, costs, and staff time.

This is the core design problem. The dollars come from a population-level, longitudinal, team-produced result and most practices try to distribute them using an encounter-level, transactional, individually attributed formula. The mismatch is not a rounding error. It is the whole thing.

The seniority question, asked honestly

Seniority is usually treated as either an entitlement or an embarrassment. It is neither. There are two legitimate claims hiding inside it, and a fair model separates them.

The first is return on capital and risk. Senior partners typically funded the care-management staff, the population health platform, the analytics, the after-hours coverage and the working capital that carried the practice through the years when the accountable care organization (ACO) produced nothing. That is a real investment deserving a real return. But it is a return on capital, not a return on clinical performance, and it should be paid as such, from a separate and explicitly labeled pool.

The second is the panel itself. A physician who is 30 years into practice generally carries a larger, older, more complex, more loyal panel. That panel is the substrate on which savings are generated, and it took decades to build. This is a genuine contribution — but it is already captured by attribution-weighted allocation. Paying for it twice, once as panel weight and once as a seniority multiplier, is double-counting.

What seniority does not justify is a performance premium. If a senior partner's attributed patients have higher risk-adjusted per-member-per-month costs than a second-year associate's, no length-of-service argument fixes that. The moment the distribution formula becomes a mechanism for protecting incumbent income against measured performance, the practice has stopped being a value-based organization and has simply found a new way to fund the old one.

A five-bucket model

The most useful structural insight is that allocation is not one decision but a sequence of decisions. Three questions get to the three tiers of decision-making, as follows:

• How much is retained for infrastructure versus distributed?
• How is the distributed portion split among provider groups?
• How is it allocated within a group?

Practices that argue about the third question without settling the first two never reach an agreement.

Here is a workable structure for an independent primary care group:

Bucket 1 — Reinvestment, off the top

Care coordinators, data analytics, after-hours access, the transitional care nurse who calls every discharged patient within 48 hours. This is not overhead skimming; it is the machinery that produces next year's savings. Fund it first and say so publicly, because the fastest way to lose a shared savings program is to distribute 100% of year one shared savings and have nothing left to generate year two. Administrative costs and any debt service come out of this bucket as well.

After deducting the reinvestment of savings, a reasonable distribution of remaining savings goes to:

Bucket 2 — Panel size

Distribute a portion of shared savings based on each physician’s number of attributed beneficiaries, adjusted for patient risk. This creates a predictable base payment for accepting responsibility for the population and is one of the most common approaches used by ACOs.

Risk adjustment matters because a physician caring for a smaller number of patients with medically complex cases may be managing as much — or more — clinical and financial responsibility as a physician with a larger, healthier panel. For example, a physician with 500 attributed beneficiaries whose panel has substantially higher average risk may receive a comparable or greater panel-based allocation than a physician with 650 relatively healthy beneficiaries.

A risk-weighted panel base also recognizes the larger patient panels often maintained by more established physicians without creating a separate seniority payment category.

Bucket 3 — Financial performance

This should be the largest bucket because savings ultimately depend on managing the total cost of care below the benchmark. Allocate a meaningful share based on financial performance but ensure the cost measure is appropriately risk-adjusted and excludes catastrophic outliers. Otherwise, the formula may unintentionally reward practices for avoiding the sickest patients and the patients with the most complex cases, which is the single most destructive outcome an allocation model can produce.

Bucket 4 — Quality performance

Quality should carry equally clear importance, but it should function primarily as an eligibility gate rather than a sliding-scale bonus. A practice that misses required quality measures should not receive shared savings merely because it came in under its cost target. Strong quality performance protects patients, supports program eligibility and ensures that financial rewards reflect better care, not simply lower spending.

Bucket 5 — Citizenship

This category recognizes work that supports the practice and ACO but is not fully reflected in panel size, cost, quality, or visit volume. It can include care-management huddles, high-risk patient reviews, gap-closure outreach, timely documentation, quality-improvement projects and participation in governance.

It also provides a fair way to recognize physicians who contribute time to mentoring, staff training, committee leadership, contract negotiations, recruitment, strategic planning and representing the practice in ACO or payer meetings. Expectations should be set in advance and contributions documented to ensure the category rewards meaningful work rather than visibility or favoritism.

Then, and only then, pay a separate, explicitly disclosed return on invested capital from the reinvestment residual or from a defined percentage of total savings, allocated by ownership. Naming it honestly does two things: It gives senior partners a defensible claim, and it stops that claim from contaminating the clinical performance pools.

Guardrails

Risk-adjust or don't measure cost at all. An unadjusted per-member-per-month comparison is a referendum on panel complexity, not on physician behavior.

Consider rewarding improvement as well as absolute performance. A physician or practice that makes meaningful progress should receive recognition, not only those already performing at a high level. At the same time, the model should balance improvement with sustained achievement: Placing too much emphasis on year-over-year gains can unfairly disadvantage consistently high performers who have less room to improve.

Keep shared-savings distribution at the practice or group level when patient attribution to an individual physician is not reliable. If incentives are tied too closely to one clinician’s assigned patients, clinicians may feel financially penalized for caring for patients with complex illnesses, frequent hospitalizations or high costs. That can lead to “not my patient” behavior — trying to shift responsibility for difficult patients to someone else rather than working together to improve their care. A group-level approach better reflects how primary care is actually delivered: through teams, shared coverage, care managers, specialists and practice-wide systems. It encourages physicians to help the sickest patients rather than avoid them.

Get counsel involved early. Distributions tied to referral patterns or downstream utilization implicate the Stark Law and the Anti-Kickback Statute. Structure matters, and it is far cheaper to design correctly than to unwind.

Test the proposed shared-savings formula using prior years’ actual performance data before adopting it. A model that appears fair in theory can create large and unintended differences in payments among physicians or practices with similar contributions. Running historical scenarios in advance helps identify those problems, refine the methodology, and build confidence that the final approach is equitable and understandable before anyone is asked to approve it.

Write it down before the money arrives. Allocation designed in the presence of a specific check is negotiation. Allocation designed in advance is policy.

The real test

A good distribution formula is one that a physician would accept before knowing whether it favors them. The most successful practices treat their shared-savings formula as a practical management tool, not a permanent settlement. They begin with a model that physicians can understand, test it against actual results and revisit it each year as their data, care-management capabilities, and value-based experience mature.

The goal is not to create the most complicated formula. It is to build one that is fair, transparent, and aligned with the behaviors the practice needs to succeed: caring for patients with complex cases, improving quality, managing total cost of care and contributing to the work of the organization.

Finally, protect a meaningful reinvestment pool. Shared savings should strengthen the infrastructure that produced them — care management, data analytics, workflow support and physician leadership. Without continued investment in those capabilities, today’s distribution may come at the expense of tomorrow’s performance.

Robert Resnik, M.D., MBA, is a board-certified internal medicine physician practicing in Cary, North Carolina. He earned his medical degree from Eastern Virginia Medical School and completed his residency at East Carolina University. He also holds an MBA from Duke University.