HHS counted value-based payment in dollars, and providers paid for it in data
HHS’s January 2015 Medicare targets measured the share of spending inside alternative payment models. The money that varied with results was small next to what providers had to buy, build and staff to earn it.
Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.
Two-minute summaryThe dek and the first sentence of each section, quoted from the article
HHS’s January 2015 Medicare targets measured the share of spending inside alternative payment models. The money that varied with results was small next to what providers had to buy, build and staff to earn it.
- The week On Monday, January 26, 2015, Health and Human Services Secretary Sylvia Burwell set the first explicit numeric goals in Medicare’s history for moving payment away from pure fee-for-service.
- What the percentage counted The 30 percent target measured Medicare payments tied to an alternative payment model.
- The revenue side: how an alternative payment model paid, and when An organization deciding whether to join a model had to forecast what it would earn.
- The cost side: what organizations had to buy and build Whatever the revenue forecast showed, taking part required spending that began before any payment.
- The contract side: how the risk was divided HHS set the targets in a public announcement, but the economics were settled in contracts.
- How the math differed by type of organization The same target meant very different things depending on the organization’s structure, payer mix and size.
- The objection: a dollar target was the right tool The strongest objection to this argument is that HHS did exactly what it needed to do.
- With hindsight HHS reached the first target early.
The week
On Monday, January 26, 2015, Health and Human Services Secretary Sylvia Burwell set the first explicit numeric goals in Medicare’s history for moving payment away from pure fee-for-service. She announced them at a meeting with nearly two dozen leaders from consumer groups, insurers, providers and employers (AHA News, January 26, 2015). She also set out the reasoning the same day in an article in the New England Journal of Medicine, which the Peterson-Kaiser Health System Tracker summarized the same day (KFF).
The announcement had two pairs of targets, both for traditional Medicare fee-for-service, not Medicare Advantage. By the end of 2016, 30 percent of Medicare fee-for-service payments were to be tied to quality or value through alternative payment models, such as accountable care organizations and bundled payments. By the end of 2018 that share was to reach 50 percent. A broader goal covered payments linked to quality in any way: 85 percent by 2016 and 90 percent by 2018 (CMS fact sheet, January 26, 2015).
To make the goals measurable, CMS published a four-category framework. Category 1 was fee-for-service with no link to quality. Category 2 was fee-for-service with a link to quality, and CMS named the Hospital Value-Based Purchasing Program, the Physician Value-Based Modifier and the readmissions and hospital-acquired condition reduction programs as examples. Category 3 covered alternative payment models built on fee-for-service architecture: accountable care organizations, medical homes, bundled payments, the Comprehensive Primary Care Initiative and the Comprehensive ESRD Care model. Category 4 was population-based payment. The 30 and 50 percent targets counted only categories 3 and 4. The 85 and 90 percent targets counted categories 2 through 4 (CMS).
CMS also gave a baseline. It estimated that at the end of 2014 roughly 20 percent of Medicare fee-for-service payments to providers already ran through category 3 and 4 arrangements, up from very little a few years earlier. It credited four programs for that growth: the Medicare Shared Savings Program, the Pioneer ACO model, the Bundled Payments for Care Improvement initiative and the Comprehensive Primary Care Initiative. Most were created under the Affordable Care Act (CMS). So the 2016 goal asked for another ten points in two years, and the 2018 goal asked for thirty.
The third piece of the announcement was the Health Care Payment Learning and Action Network. HHS described it as a forum of private payers, large employers, providers, consumers, states and federal partners. Its stated jobs were to find agreement on how alternative payment models should work and be reported, and to develop common approaches to attribution, financial models, benchmarking and risk adjustment. It was to start meeting in March (AHA News). The fact sheet gave the reason plainly: a provider has weak reasons to change how it delivers care when only some of its payers change how they pay (CMS).
Reaction was supportive in direction and cautious in detail. The American Hospital Association said hospitals backed payment that rewarded value and coordination, and asked HHS to evaluate the reforms already running, learn from the demonstrations and phase in change to suit different communities. It also asked for the details and metrics behind the program (AHA News).
The setting matters. In January 2015 Medicare physician payment was still governed by the sustainable growth rate formula, and Congress was again facing a deadline to block a scheduled cut. The law that would replace it, the Medicare Access and CHIP Reauthorization Act, did not yet exist. The accountable care programs that the targets relied on were voluntary. Most participating ACOs were in one-sided arrangements that shared savings but not losses. The goals therefore rested on providers choosing to join, and on HHS continuing to offer models they wanted to join.
The announcement was a statement of intent. It did not change a payment rate or a regulation that week. What it changed was the yardstick. From then on, the progress of payment reform was reported as a share of Medicare dollars, and the organizations that would have to deliver it began to plan against that share.
The January 2015 targets are usually remembered as a policy milestone. We read them as a procurement event. HHS described the shift in terms of the share of Medicare dollars flowing through alternative payment models. For the hospitals, physician groups and networks that had to deliver the shift, the money at stake was modest next to the cost of the systems, data and staff needed to earn it. The targets moved the economics of payment reform into health systems’ capital and operating budgets, and into contracts with vendors who had little stake in the outcome.
The argument has four parts, and they are the first four sections below. First, what the percentage counted, which explains why a dollar share overstated how much money was really at risk. Second, the revenue side: how an alternative payment model paid, and when. Third, the cost side: what organizations had to buy and build to take part. Fourth, the contract side: how the risk was divided among Medicare, providers and vendors. Two shorter sections follow. One looks at how the math differed for different kinds of organizations. The last tests the argument against the strongest objection to it.
What the percentage counted
The 30 percent target measured Medicare payments tied to an alternative payment model. The natural reading is that 30 percent of Medicare’s money would be paid on a new basis. That reading is wrong in an important way, and the difference drove a decade of budget decisions.
Payments tied to a model, not payments at risk
Most category 3 models were built on fee-for-service architecture, as CMS said itself (CMS). In a Medicare Shared Savings Program ACO, Medicare kept paying every claim at the normal fee schedule. At the end of the year, CMS compared the total spending for the ACO’s assigned beneficiaries with a benchmark. If spending came in far enough below the benchmark, and quality standards were met, the ACO received a share of the difference. In a two-sided track it could also owe a share of any excess.
The measure counted the spending that flowed under that arrangement. The scale shows it. Thirty percent of traditional Medicare spending was a sum many times larger than any bonus pool. CMS later reported that Medicare ACOs saved the program a little over 411 million dollars in 2014, and ACO bonuses were a share of savings of that order (CMS, 2016). The target had to be counting ordinary claims for attributed patients. The money that actually depended on performance was the shared-savings payment or the shared loss, a far smaller amount. The headline share therefore described how much spending sat inside a model. It did not describe how much revenue moved with results.
Why the distinction mattered to a budget
For a health system chief financial officer, the gap between those two numbers was the business case. If a fifth of Medicare revenue were truly at risk, investing heavily in population health would be an obvious decision. If what was at risk was a small bonus on top of unchanged claims, the investment case was much thinner. It depended on the size of the possible savings, the share the ACO kept and the chance of earning anything at all.
The target also counted category 2 programs toward the broader 85 percent goal. Hospital Value-Based Purchasing, the readmissions penalty and the hospital-acquired condition penalty adjusted payment by a small percentage at the margin. A hospital already subject to them counted as tied to quality. That broad goal was far easier to reach, because the category 2 programs CMS named already applied to most hospitals’ Medicare payments (CMS).
Why a dollar share was still a reasonable choice
There was a case for counting this way, and HHS had reasons. A share of spending is easy to explain, and it measures reach: how many patients were cared for by providers who had agreed to be accountable for total cost. Accountability for a population is the behavior the models were meant to change, even if the bonus attached to it was small. A narrower measure of dollars at risk would have shown a very small number and might have discouraged the effort.
The choice still had consequences. Providers, consultants and vendors began to describe value-based care as a large share of revenue. Boards approved investments on that framing. The organizations that read the methodology closely saw that the near-term money was in avoiding penalties and earning modest bonuses, while the larger prize depended on downside risk that few had yet accepted.
What the measure left out
The targets covered traditional Medicare only. Medicare Advantage, Medicaid and commercial insurance were outside them, though HHS said it would work with those payers toward similar goals (CMS). A health system’s payer mix decided how much the Medicare target mattered to it. One whose revenue came mostly from commercial plans could treat Medicare ACO participation as a pilot. One heavily dependent on traditional Medicare had to treat it as strategy.
The measure also said nothing about cost to the provider. A model could count toward the target while costing its participants more to run than it paid them. That was the experience of many early ACOs, and it is the part of the story the percentage could not show.
The revenue side: how an alternative payment model paid, and when
An organization deciding whether to join a model had to forecast what it would earn. That forecast depended on mechanics that rarely appeared in board presentations but determined the result.
Attribution decided whose costs you carried
In the Shared Savings Program, beneficiaries were assigned to an ACO based on where they received the largest share of their primary care. They were not enrolled and could see any Medicare provider. An ACO could therefore be accountable for the total cost of patients it did not control. A patient assigned to an ACO might receive most specialty care, imaging and hospital stays elsewhere. Those costs counted against the ACO’s benchmark.
Attribution also changed from year to year. Patients moved, changed doctors or died. An ACO planning care management for a list of patients found the list shifting under it. The financial result depended on a population the organization could describe only after the fact.
The benchmark decided whether savings were possible
An ACO’s benchmark was built largely from its own historical spending for its assigned population. That design rewarded organizations that had been expensive, because they had more room to cut, and penalized those that were already efficient. Low-cost systems in low-cost regions often found it hard to beat a benchmark built on their own lean history.
The benchmark also rebased over time. An ACO that succeeded in lowering spending could see its next benchmark reflect that lower spending. Sustained savings then became harder to show. Finance teams learned to model several agreement periods, not one year.
Thresholds and quality gates reduced the expected payout
Savings had to exceed a minimum savings rate before any were shared, so that random variation was not rewarded. Quality performance then scaled the share the ACO kept. Both features made sense as policy. Both lowered the probability-weighted value of joining. A forecast that assumed a share of all savings below the benchmark would overstate the likely payment.
The cash arrived long after the work
Shared savings were calculated after the performance year closed and claims had run out. The payment came well after the year it rewarded. Care managers hired in January were paid from operating budgets for many months before any savings check could arrive, if it arrived at all.
That timing turned a payment model into a working capital question. Large systems could absorb the lag. Small physician-led ACOs often could not. Many relied on outside partners who advanced the start-up costs in exchange for a share of any savings, which became part of the contract story below.
Bundles worked differently
Bundled payments moved risk for an episode of care, such as a joint replacement, rather than for a population. Medicare set a target price for the episode, including care for a period after discharge. Providers that kept total episode spending below the target could share the difference. The levers were concrete: choice of implant, length of stay and, above all, where patients went after discharge.
The revenue logic of a bundle was easier to forecast than an ACO’s, because the episode and the patient were known. The exposure was narrower but sharper. A hospital whose patients routinely went to skilled nursing facilities could see much of its margin on the episode depend on post-acute decisions made by others.
Savings could cut a provider’s own revenue
The most awkward feature of the revenue side was that success often reduced fee-for-service income. An ACO that kept its patients out of the hospital lowered admissions at its own hospital. A system that shortened post-acute stays reduced volume at its own post-acute units. The shared-savings payment had to exceed the lost fee-for-service margin before the organization came out ahead.
For a health system with high fixed costs, that calculation was often unfavorable unless the empty capacity could be filled with other patients. This is why physician-led groups without hospitals tended to find the models more attractive. Their savings came from someone else’s revenue.
Risk adjustment made documentation a revenue line
Benchmarks were adjusted for the health status of the assigned population, using diagnosis-based risk scores drawn from claims. A population that looked sicker on paper received a higher expected cost. That gave diagnosis documentation a direct financial value it had rarely had under fee-for-service, where a hospital stay or office visit paid much the same whatever secondary conditions were recorded.
Organizations responded by investing in clinical documentation programs, coding review and prompts inside the electronic health record. Those investments were legitimate when they captured conditions patients really had. They also carried compliance risk if they drifted toward recording conditions that were not supported by the chart. A compliance officer reviewing a population health budget in 2015 needed to see documentation improvement as both a revenue line and an audit exposure.
Penalties were the near-term money
For many hospitals the most certain financial effect of the value agenda in 2015 came from category 2 programs, not from shared savings. The readmissions and hospital-acquired condition programs reduced Medicare payments to hospitals that performed poorly. Value-based purchasing redistributed a slice of payment according to quality and efficiency scores. These adjustments applied whether or not the hospital joined any model.
That meant the same capabilities a hospital would build for an ACO, such as discharge follow-up, medication reconciliation and tracking patients after they left, also protected fee-for-service revenue. The business case for the infrastructure was often stronger on penalty avoidance than on shared savings. Finance teams that combined the two in one case made a more honest argument than those that relied on bonuses alone.
The cost side: what organizations had to buy and build
Whatever the revenue forecast showed, taking part required spending that began before any payment. It fell into four categories: data, analytics, people and reporting. Each was a procurement decision with long consequences.
Data that a provider had never held
A hospital knows what happens inside its own walls. To manage total cost of care, it needed to know what happened everywhere else: other hospitals, specialists, labs, pharmacies and post-acute providers. CMS supplied Shared Savings Program ACOs with claims data for their assigned beneficiaries, delivered in recurring files. Those files were the only complete picture of what attributed patients cost.
Claims files were not ready to use. They arrived in fixed formats, lagged behind the care they described and had to be matched to the provider’s own patients and clinicians. An organization had to receive them securely, load them, keep their history and join them to records from its electronic health record. That work required a data warehouse, integration staff and identity matching, and most ambulatory groups had none of these.
Analytics that turned data into lists
The data was useful only if it produced decisions. ACOs needed to know which patients were driving cost, which were likely to be admitted, which specialists and facilities were expensive and where leakage occurred. That meant risk stratification, utilization reporting and network analysis. Vendors sold population health platforms that promised all of this, often priced per attributed life per month.
The choice between buying a platform, extending the electronic health record vendor’s modules and building in-house was one of the most consequential technology decisions of the period. A separate platform could ingest data from many sources but created another system clinicians had to open. The record vendor’s module sat inside clinical workflow but was often weaker at handling outside claims. Building gave control but demanded scarce data engineers and analysts.
People who did the actual work
Savings came from changing care, and that required staff. Care managers and nurse navigators called high-risk patients, arranged follow-up after discharge and coordinated with post-acute providers. Pharmacists reviewed medications. Social workers addressed housing or transport problems that drove readmissions. These were recurring salaries, not one-time capital.
The staffing cost often exceeded the technology cost. It was also harder to cut if the model failed, because the staff had relationships with patients. Finance teams that modeled the software and ignored the care team understated the commitment.
Reporting that the model required
Each model had its own quality measures, submission methods and deadlines. ACOs reported clinical quality measures for samples of patients, and missing data could cost them part of their shared savings. Bundled payment participants had their own reconciliation and quality requirements. Category 2 programs had their own measures and penalties.
Each set of requirements meant data extraction, chart abstraction or registry work. Organizations in several models at once found themselves running parallel reporting operations. The work grew faster than the payment it protected.
Why these costs were hard to recover
Three features made the investment risky. It was front-loaded, while any return came late. It was partly specific to one payer’s model, so a change in that model could strand it. And it was hard to measure, because savings were calculated against a counterfactual benchmark rather than observed directly.
Compliance and security costs also rose. Claims data for tens of thousands of attributed beneficiaries was protected health information under HIPAA. The agreements under which CMS released it imposed their own conditions on use and disclosure. Every new warehouse, vendor and data feed added to the organization’s exposure and to its list of business associates. A chief information security officer at an ACO in 2015 was protecting a data asset that had not existed a few years earlier.
What a sensible budget looked like
The organizations that handled this best treated the investment as infrastructure for several payers, not a cost of one Medicare contract. A claims ingestion pipeline built for Medicare could take commercial and Medicaid files later. An identity-matching service could serve clinical exchange as well as analytics. Care management staff could work across payers. That approach required more upfront design but made the asset durable if any single model changed.
The organizations that handled it worst bought a point solution for one program, priced per life for that program, and found themselves with several overlapping tools when other payers launched their own arrangements.
How a sound selection process ran
The selection itself deserved more care than many organizations gave it. A sound process started with the use cases the organization would run in its first two years. They included risk lists for care managers, leakage reports for network managers, quality gap lists for clinics and financial tracking against the benchmark. It then tested vendors against the organization’s own data, including a sample of real Medicare claims files and records from its electronic health record, rather than against a demonstration data set.
Several evaluation criteria mattered. One was how the platform matched patients across sources. Others were how it handled late and corrected claims, and whether its attribution logic could be inspected and reproduced, and how quickly it could add a new payer’s file format. Pricing per attributed life looked simple but scaled with the organization’s success at growing its population, so procurement teams compared it against fixed or tiered fees. Security review belonged in the same process, because the platform would hold protected health information from outside sources. A business associate agreement, audit rights and breach notice terms were part of the selection, not paperwork after it.
The contract side: how the risk was divided
HHS set the targets in a public announcement, but the economics were settled in contracts. Three sets of agreements decided who would bear the cost if the models underperformed.
Between Medicare and the provider
The Medicare agreement set the track, the benchmark method, the sharing rate and any loss exposure. Most ACOs in early 2015 were in one-sided tracks, which meant Medicare bore the downside and the ACO risked only its own investment. That design attracted participants and helped the dollar share grow. It also meant the organizations doing the most work could lose money without ever owing Medicare anything, because their losses were internal costs that no agreement recognized.
Among the providers inside an ACO
An ACO was usually a legal entity owned or controlled by its participants: hospitals, physician groups, sometimes independent practices. Its internal agreements decided how any shared savings were distributed and who paid for shared infrastructure. Those decisions were contentious. A hospital that funded the data platform might expect most of the savings. Physicians who changed referral patterns might expect the same. Distribution formulas weighed effort, attributed patients, quality scores and capital contributed.
The internal contracts also had to respect federal fraud and abuse law. Paying physicians to change referrals or utilization can raise questions under the physician self-referral law and the anti-kickback statute. CMS and the HHS Office of Inspector General issued waivers for ACO arrangements, and organizations relied on them with counsel’s guidance. Compliance officers became part of every distribution decision.
The data agreements underneath
Beneath the commercial agreements sat data agreements that limited what could be done with the most valuable asset. Medicare claims data for assigned beneficiaries came under terms set by CMS, which restricted use to the purposes of the program and required safeguards. Beneficiaries could ask that their data not be shared. Every vendor that touched the data needed a business associate agreement that passed those limits down.
These terms affected value. A vendor that hoped to pool data across clients to build benchmarks or predictive models needed rights the provider might not be able to grant. A provider that wanted to reuse the data for research or for a commercial contract needed to check whether the original terms allowed it. Organizations should check with counsel before assuming that data received for one program can serve another. The contracts that worked best stated the permitted uses at the start, rather than leaving them to be argued later.
Between the provider and its vendors
Vendor contracts were where the risk transfer was weakest. A population health vendor paid per attributed life per month earned the same fee whether the ACO saved money or not. Its incentive was to keep the contract and expand the covered population. Some vendors and management partners offered gain-sharing instead: they funded start-up costs and took a share of any savings. That aligned incentives but gave the partner a claim on future payments and sometimes on governance.
Procurement teams had few precedents for these contracts. Key terms included who owned derived data and models, what happened to data at termination, how savings attributable to the vendor were measured and what service levels applied to data feeds. A contract that let the vendor keep de-identified data, or that made the provider dependent on the vendor’s proprietary attribution logic, could constrain the organization for years.
Where the Learning and Action Network fit
HHS created the Learning and Action Network partly to fix a contracting problem. Each payer designing its own model, with its own attribution rules, measures and reporting, multiplied provider costs. Common approaches to attribution, benchmarking and risk adjustment would let a provider build one capability and use it across contracts (CMS).
That aim was sound, and it targeted the right cost. Alignment depended on voluntary agreement among payers with different business interests, though, and commercial insurers were competing with each other on network and product design. For a provider in 2015, the practical consequence was that every contract still had to be negotiated separately, and every new arrangement brought its own data and reporting.
The term that mattered most
Across all three sets of agreements, the most important term was duration. A provider that invested in care management and data needed several years of predictable rules to recover the cost. Medicare agreements ran for multi-year periods, but the rules for later periods could change. Vendor contracts often ran for shorter terms with automatic renewal. Internal ACO agreements depended on continued participation. A mismatch among those durations left the provider holding long commitments against short or uncertain revenue.
Commercial and Medicare Advantage contracts followed the same shape
The Medicare announcement also changed negotiations with other payers. Commercial insurers and Medicare Advantage plans began offering their own shared-savings and risk arrangements, often modeled loosely on Medicare’s. Each came with its own attribution method, quality measures and data files. Some plans supplied rich claims data. Others supplied summary reports that made independent verification of results difficult.
For a provider, the negotiating priorities were consistent across payers. It needed access to claims-level data for attributed members, a benchmark method it could reproduce, quality measures that overlapped with those it already reported, and a reconciliation schedule it could plan cash around. A provider that negotiated those terms early could run one operation across many contracts. One that accepted each payer’s defaults built a separate operation for each, and paid for the duplication in staff time and software fees.
How the math differed by type of organization
The same target meant very different things depending on the organization’s structure, payer mix and size. Five types show the range.
Integrated health systems with hospitals
Systems that owned hospitals and employed physicians had the capital and data staff to build infrastructure. They also faced the strongest conflict between savings and fee-for-service revenue. Their best case for joining was strategic: preparing for a future where more payment would be at risk, and securing relationships with physicians and payers. Many joined models while managing exposure carefully, using the infrastructure more for commercial contracts and quality penalties than for Medicare shared savings.
Physician-led groups and ambulatory networks
Independent physician groups had little hospital revenue to lose, so savings from fewer admissions came mostly at others’ expense. They had better incentives but less capital. They were the main market for management partners that funded start-up costs in exchange for a share of savings. Their procurement risk was concentration: dependence on a single partner for data, analytics and sometimes governance.
Academic medical centers
Academic centers had high costs, complex referral patterns and many patients who came from outside their region. Attribution was hard to predict, and their benchmarks reflected expensive tertiary care. They often participated in bundles for specific procedures more readily than in population models. Their data capabilities were strong, but internal governance across faculty practice plans and hospitals slowed decisions.
Rural and safety-net providers
Rural hospitals and safety-net systems had small attributed populations, which made results statistically noisy. A few high-cost patients could erase a year’s savings. Their margins were thin and their technology budgets small. The fixed costs of data and reporting weighed more heavily per patient. For them, the target was a signal about Medicare’s direction more than an opportunity they could easily take up.
Health plans and state Medicaid agencies
The targets did not apply to Medicare Advantage plans or Medicaid, but both were watching. Plans that ran their own provider risk arrangements saw HHS endorsing a framework they could adopt. State Medicaid agencies that were already testing their own payment reforms gained a common vocabulary through the four categories. For a state agency, the procurement question was what to require of managed care organizations in their contracts. That included alternative payment targets, reporting on them and data-sharing obligations with providers.
What the comparison shows
The organizations best placed to act on the target were those with capital, data and limited fee-for-service revenue at stake. Few organizations had all three. The target set one national number. The ability to respond to it was distributed very unevenly, and the costs of responding fell hardest on those with the least room to absorb them.
The objection: a dollar target was the right tool
The strongest objection to this argument is that HHS did exactly what it needed to do. A headline number focused attention, gave the industry a direction and helped push participation up. Counting dollars tied to models rather than dollars at risk was a deliberate choice to measure reach. On this view, the cost providers bore was the price of change that had to happen.
Where the objection is right
The target did focus attention. HHS later reported that it reached the first goal early (CMS fact sheet, March 3, 2016). Organizations that would otherwise have waited began building data and care management capabilities. Commercial payers and states adopted the same categories, which gave the industry a shared vocabulary. Measuring reach was defensible, because accountability for a population is a precondition for anything else.
Where it falls short
The objection treats the cost of participation as a fixed price of progress. It was not fixed. It depended on choices HHS and providers made, especially the proliferation of separate models and reporting rules, and the timing of payment. A target that counted spending tied to models encouraged growth in participation without signaling how much would be at stake, or for how long. Providers that read the number as a revenue forecast overbuilt for one payer’s model. Those that read it as a signal, and built infrastructure that would serve any payer, fared better.
The objection also misses who bore the cost. Medicare bore little downside in one-sided models. Vendors were paid regardless of results. The provider carried the investment and most of the uncertainty. That allocation was not inevitable, and later policy moved deliberately to change it.
A second objection: the costs were modest
A different objection holds that participation costs were small relative to a health system’s total Medicare revenue, so the procurement framing exaggerates. For a large system, a population health platform and a care management team were indeed a small share of operating expense. The argument has more force for small organizations, and for the cumulative effect across many arrangements. A physician group of modest size could spend a meaningful share of its margin on data, analytics and reporting for one ACO. A system in several models, each with its own measures and files, multiplied the overhead. The cost that mattered was not one platform. It was the duplication that separate models and payers created, which is precisely what the Learning and Action Network was meant to reduce and could reduce only slowly.
What the comparison means for reading targets today
Policy targets stated as a share of spending will keep appearing, from Medicare and from states and payers. The useful reading separates three numbers: the spending inside an arrangement, the revenue that varies with performance, and the cost of qualifying to take part. The first measures the policy’s reach. The second and third decide whether a particular organization should join. In January 2015 only the first was published, and many plans were made as if it were the second.
Questions a finance and technology team could ask now
- For each value-based arrangement, what share of revenue actually varies with performance, as distinct from the spending that flows under it?
- Which of our data and analytics assets serve every payer, and which are tied to one program’s rules?
- When does cash from each arrangement arrive relative to the costs we incur to earn it?
- Do our vendor contracts give us ownership of derived data and attribution logic, and a clean exit?
- Which arrangements reduce our own fee-for-service revenue, and by how much, before any shared savings?
With hindsight
HHS reached the first target early. In March 2016 CMS estimated that 30 percent of traditional Medicare payments were tied to alternative payment models as of January 2016. That was about eleven months ahead of the end-2016 date set in January 2015. It credited new ACOs and wider participation in other models, and named the next goal as 50 percent by the end of 2018 (CMS fact sheet, March 3, 2016). The same fact sheet said the Learning and Action Network had been created in March 2015 and was preparing implementation guides for payers and purchasers.
The bigger change came from Congress. In April 2015 the Medicare Access and CHIP Reauthorization Act repealed the sustainable growth rate formula and created the Quality Payment Program, which began in 2017. Clinicians could report through the Merit-based Incentive Payment System, in effect a category 2 arrangement with payment adjustments, or qualify through participation in Advanced Alternative Payment Models, which required meaningful downside risk. That law added a second definition of value-based payment, built around risk rather than reach.
We have not found an HHS announcement that the 50 percent goal for 2018 was met on schedule, in the way the 2016 goal was announced. The administration changed in 2017, and the approach to models changed with it. CMS scaled back some planned mandatory episode models, and in late 2018 it finalized changes to the Shared Savings Program, known as Pathways to Success, that pushed ACOs toward downside risk sooner. The Learning and Action Network continued to measure payment across commercial plans, Medicaid, Medicare Advantage and traditional Medicare, and over time its goals shifted from any alternative payment toward arrangements with two-sided risk.
In 2021 the CMS Innovation Center reviewed its first decade and more than fifty models. Its refreshed strategy set a 2030 goal for every traditional Medicare beneficiary to be in a care relationship with accountability for quality and total cost of care. It set the same goal for the vast majority of Medicaid beneficiaries (CMS Innovation Center, 2021 strategy). The language had moved from a share of dollars to a share of people in accountable relationships.
That goal did not survive another change of administration. The strategy the Innovation Center posted in May 2025 is organized around prevention, patient empowerment, and choice and competition. It says the Center will move beneficiaries into arrangements with global downside risk, and require downside financial risk across its models, and it does not repeat the 2030 target (CMS Innovation Center strategy, 2025). Trade press reported in 2025 that the Center’s leadership had dropped the goal of putting all traditional Medicare beneficiaries in accountable care by 2030. CMS also finalized a mandatory episode model for surgical care, the Transforming Episode Accountability Model, which began in selected hospitals in January 2026.
What held up from January 2015 was the direction and the vocabulary. The four categories became the common language of payment reform across payers. The data infrastructure that ACOs built outlasted the specific models it was first built for, and became the base for later risk contracts with commercial and Medicare Advantage plans.
What did not hold up was the idea that the share of dollars tied to a model described the economic change. Policy from MACRA onward measured something closer to dollars at risk, and by 2025 CMS was requiring downside risk as a condition of new models. The organizations that read the 2015 targets as a forecast of revenue under new rules learned, sometimes expensively, that revenue followed risk, and that risk had to be accepted deliberately.
By late 2026 the lesson for budgets is settled. The infrastructure for managing total cost of care is permanent. The specific models that justify it are not, and each change of administration has changed them. Organizations that built for many payers and short model lives recovered their investment more reliably than those that built around a single program.
Spotted an error? Tell us and we will check it and correct the article where needed.