ACO shared-savings math can look precise and still pay the wrong practice
Shared-savings statements often arrive dressed in authority: clean tables, exact dollar figures, and risk scores carried to three decimal places. Precision does not guarantee accuracy. In ACO contracts, every savings number sits at the end of a chain of assumptions about attribution, risk, benchmark construction, trend, quality, and distribution, and any one of those assumptions can change who appears to have earned the money.
The practical warning for physicians is blunt. A distribution from an actuary, finance committee, or payer report should not be accepted at face value simply because it looks technical. Over the past five years, material distribution errors have redirected hundreds of thousands of dollars to clinicians who had actually generated the savings. The central question is not whether the ACO reported savings. The real question is whether the organization can show how those savings were calculated, validate the result, and explain why a practice was paid what it was paid.
Precision on paper can conceal a bad denominator.
Why the calculation chain invites error
A shared-savings calculation is not a direct observation. It is an estimate assembled inside a contractual and actuarial framework. The process requires judgments about which patients were attributed to which practices, how long they were actually attributed, how illness burden was measured, what baseline spending was used, how spending was expected to trend, which claims were included or excluded, whether quality thresholds changed the payment, and how earned dollars were distributed among practices.
A technically correct formula can still produce a wrong result if it is built on the wrong population, the wrong denominator, the wrong baseline, or the wrong quality definition. The spreadsheet may be mathematically elegant while the conclusion is economically false. In value-based care, that kind of error matters because it affects not only revenue but also physician behavior. If clinicians believe the contract is unrewarding or unreliable, they are less likely to keep investing in care management, access improvement, and chronic disease coordination.
Most of these failures are quiet. They are not dramatic misconduct. They are more often static member counts substituted for member-months, stale quality thresholds, benchmarks built from populations that no longer resemble the current panel, or distribution formulas that have never been revisited. Those errors can suppress legitimate savings or fabricate them.
Member-months versus a snapshot head count
The clearest example is the use of member-months. A member count is a snapshot; member-months measure actual exposure over time. That difference changes the math in a material way.
Consider a practice that begins the year with 1,000 attributed patients. In July, an employer group representing 300 patients leaves the plan, and the practice ends the year with 700 patients. The real exposure is 10,200 member-months, not 12,000 and not 8,400. If allowed cost is $5.1 million, the correct per member per month cost is about $500. If the benchmark is $520 PMPM, expected spending is $5.304 million, which produces $204,000 in savings.
If the ACO uses the opening population of 1,000 members for the full year, the PMPM becomes $425. If it instead uses the year-end census of 700 members as though that were the full-year population, the same $5.1 million in actual spending appears to be a $732,000 loss. The practice’s care did not change. The denominator did.
The same problem works in reverse when membership grows during the year. A practice can improve care and lower avoidable spending, yet still look underperforming if the denominator understates actual exposure. The physician question is straightforward: did the ACO calculate PMPM using actual member-month eligibility, and does the denominator match the same period reflected in the cost numerator?
Benchmarks can erase real savings or invent them
A benchmark is supposed to represent what spending would have been without the practice’s intervention. It is not an observable fact. It is a modeled counterfactual, and it can be distorted by incorrect baseline years, membership turnover, benefit-design changes, network changes or leakage patterns, regional trend factors that do not reflect the local market, comparison populations with different geography or illness burden, and the inclusion or exclusion of high-cost categories without a clear contractual basis.
If the benchmark does not reflect the practice’s actual population and market, the savings figure loses credibility. A flawed benchmark can create savings that never occurred. It can also erase genuine savings that a practice produced through lower emergency utilization, better access, or stronger management of chronic disease. The payment outcome then reflects model design more than clinical performance.
Physicians should ask how the benchmark was built and whether it tracks the same population the practice actually cared for during the performance period. If that answer is vague, the savings calculation is weak.
Risk adjustment can change the meaning of good care
Risk adjustment sets expected cost, and expected cost determines whether actual spending looks favorable. In other words, risk scoring helps decide whether shared savings exist at all. Commercial payers increasingly use proprietary risk models that may differ materially from CMS-HCC or HHS-HCC methods. The same patient can receive different risk scores across contracts because models can vary in diagnosis hierarchies, chronic-condition weights, diagnosis-frequency requirements, treatment of behavioral health conditions, treatment of social risk, coding-intensity adjustments, and recalibration methods.
When a model understates the burden of a practice’s patients, the benchmark is too low and clinically appropriate spending can be mislabeled as poor performance. When a model overstates risk, it can manufacture savings that are really the product of model design. A physician should be skeptical when panel complexity clearly rises but the ACO reports flat or declining risk scores. The numbers may be telling a story the clinic experience does not support.
What physicians should demand before accepting the distribution
The most useful question is not whether savings were reported. It is whether the ACO can reconstruct the calculation from the ground up. Physicians need to know which patients were attributed, for how long they were attributed, what period was used for the numerator, whether member-months were used for the denominator, what benchmark year or trend factor was applied, how quality thresholds affected eligibility, and how the final distribution formula treated each participating practice.
A practice that generated real savings should be able to trace its contribution through the mechanics of the contract. If the ACO cannot explain the denominator, the benchmark, and the risk logic in plain terms, the distribution should be treated as provisional rather than authoritative. The numbers may still be clean. They may still be precise. They may still be wrong.
For physicians in value-based care, the danger is not merely underpayment. It is the slow erosion of trust in a contract that depends on trust to function. When the math is opaque, care investment becomes harder to justify and harder to sustain.