How Accountable Care Organizations Try to Pay for Health, Not Volume

An Accountable Care Organization, or ACO, is a group of doctors, hospitals, and other clinicians that agrees to be measured against a spending target, called a benchmark, for a defined group of patients. If the ACO keeps total spending below that benchmark while hitting quality targets, it keeps a share of the difference. In two-sided arrangements, it also repays a share when spending runs over. The idea is to reward keeping people healthy rather than billing for more visits, tests, and procedures. Whether that idea actually changes behavior depends almost entirely on how the benchmark is built and how much financial risk the group accepts.

This article explains the mechanics and trade-offs. It is educational and not medical advice, and it takes no position on whether any model is good policy.

The problem these models are trying to solve

Traditional Medicare pays fee-for-service: each service generates a separate payment. That structure reliably pays for volume, and it does not directly reward avoiding an unnecessary scan or preventing a hospital readmission. ACOs sit on top of fee-for-service. Clinicians still bill the usual way, but at the end of the year their combined spending on an attributed population is compared to a benchmark. The gap between actual spending and the benchmark becomes shared savings or shared losses.

The reach of this approach is now large. According to the Centers for Medicare and Medicaid Services (CMS), an estimated 14.3 million people in Traditional Medicare received care through an ACO as of January 2026, up from 13.7 million a year earlier. The Medicare Shared Savings Program alone included 511 ACOs serving 12.6 million people, drawing on more than 700,000 participating providers and organizations.

Shared savings versus two-sided risk

The central design lever is how gains and losses are split.

In a one-sided or upside-only arrangement, an ACO shares in savings if it beats the benchmark but owes nothing if it exceeds it. The incentive is real but muted, because the worst case is simply earning no bonus. In a two-sided or downside arrangement, the ACO also repays a percentage of any overspending. That sharpens the incentive and, in principle, discourages a group from booking savings in good years while walking away in bad ones.

The Shared Savings Program organizes this as a ladder. Its BASIC track runs through levels A to E: the first levels are upside-only, and the group takes on progressively more downside risk as it climbs, up to a 50 percent savings share and 30 percent loss share at level E. The separate ENHANCED track carries the most risk and the highest potential reward, with a savings share up to 75 percent. CMS reports that in 2026, 82.8 percent of Shared Savings Program ACOs sat in level E or the ENHANCED track, the highest share of risk-bearing participation since the program began in 2012.

Risk-bearing is not automatically better. Downside risk pushes groups to manage spending aggressively, but it can also deter smaller, rural, or independent practices that cannot absorb a bad year, which narrows who participates.

Why benchmarking is the whole game

A benchmark sounds like a technical detail. It is closer to the rulebook.

Set the benchmark too high and an ACO earns bonuses for spending that would have been low anyway. Set it too low and even efficient groups lose money and exit. Two design choices matter most. The first is rebasing: benchmarks are periodically reset toward an ACO's own recent spending, which means a group that saves money this year can find its future target lowered, effectively being penalized for prior success. CMS has extended agreement periods to at least five years partly to reduce how often this happens. The second is regional adjustment: blending an ACO's own history with spending in its broader region rewards groups that are efficient relative to their neighbors rather than only relative to their past.

From ACO REACH to LEAD

The CMS Innovation Center, known as CMMI, tests models that push further than the standard program, and its choices show how design shapes incentives.

The current advanced model, ACO REACH, ends on December 31, 2026. It offered a Global option in which an ACO could take on full responsibility for the total cost of care, effectively 100 percent of savings and losses, through capitation arrangements that pay a set amount rather than per service. To guard against overpaying, CMS applied a benchmark discount that rose over time, reaching 4 percent under the Global option in 2026. That discount is a direct policy dial: a larger discount extracts more savings up front but leaves participants a thinner margin.

Its successor, the LEAD model (Long-term Enhanced ACO Design), begins January 1, 2027, and runs a full ten years through December 31, 2036, with a request for applications available in March 2026. Two design signals stand out. The ten-year horizon is the longest CMS has tested and is meant to reduce the disruption of frequent rebasing, giving groups a more stable target to invest against. And CMS has said the model uses revised benchmarking intended to attract a broader mix of participants, including smaller, independent, and rural practices and those serving high-need populations such as people dually eligible for Medicare and Medicaid. Read together, these design choices target known limitations of earlier models: benchmarks that lowered targets after success, and risk terms that were difficult for all but the largest organizations to bear.

What to make of it

The evidence so far is mixed but not empty. In performance year 2024, the most recently reconciled year, Shared Savings Program ACOs generated $4.1 billion in shared savings and saved Medicare $2.5 billion, by CMS accounting. Those figures depend heavily on how benchmarks were set, which is exactly why the benchmarking method, rather than the label on the model, deserves the scrutiny. A reader evaluating any ACO claim should ask three questions: how is the benchmark calculated and rebased, how much downside risk does the group actually bear, and which patients are attributed to it. The answers determine whether a model rewards genuinely better care or simply favorable math.