Editor’s Note:This post is part of a series of Health Affairs Blog posts examining the proposed rule implementing the Medicare Shared Savings Program, issued March 31 by the Centers for Medicare and Medicaid Services. You can read earlier posts in the series by Mark McClellan and Elliott Fisher, Douglas Hastings, and Ron Klar. Watch this blog for additional posts.

Unless fundamentally changed, the proposed regulation on ACOs and the Medicare Shared Savings Program will stifle (if not halt) creation of accountable care organizations (ACOs).  The proposed regulation imposes unfavorable economics, unrealistic requirements, high uncertainty, and significant risks for ACOs.

Creating an untenable program squanders a crucial opportunity to move care away from unaccountable, unorganized, and expensive fee-for-service.  Widespread interest in creating ACOs will ebb as interested parties assess costs and benefits under the unworkable proposed standards.

This posting highlights five key problems and concludes with several observations.

The Proposed Regulation Makes ACOs Not Economically Viable

Even highly successful ACOs are likely to need years before recouping start-up and operating expenses under the proposed regulation.  Under the CMS 2-sided shared savings model, ACOs can share in bonuses if they save money but are liable for spending in excess of budget. Under the CMS 1-sided model, ACOs can share in bonuses without liability for losses in the first two years, after which they must adopt the 2-sided model.  Barring special circumstances, financial risks greatly outweigh potential rewards from accepting 2-sided versus bonus only (1-sided) shared savings.  New entities, and especially primary care oriented physician entities, will face prohibitive obstacles in qualifying as ACOs.

Recouping Investment and First Year Operating Costs. Saving 5 percent in the first year of ACO operation would strike most observers as an outstanding success.  Yet, a simple financial analysis incorporating CMS rules and assumptions highlights that annual savings may have to approach 7 percent to simply break-even on operating and start-up costs, before paying bonuses.

Start-up costs in Medicare’s Physician Group Practice (PGP) Demonstration averaged $1.76 million, with the $0.49 million for investment and $1.27 million for first year operation, varying widely across sites.  Despite noting that ACOs are likely to incur greater start-up costs for a variety of reasons (such as lacking electronic health records, which were present in 8 of the 10 PGP sites), the CMS “Regulatory Impact Analysis” assumed ACO start-up costs equaled PGP costs.

An ACO with 10,000 assigned Medicare patients averaging $10,000 in annual cost would have a budget target of $100 million.  With the first $2 million (or 2 percent) accruing to Medicare, in the first year the ACO would appear to need $5.5 million in total savings to offset start-up cost (if credited for reporting quality data, it receives 50 percent of savings above the 2 percent threshold).  However, CMS mandates withholding 25 percent of ACO savings during the 3 year agreement, plus the added year or two it takes to “close” the books.

Saving 6.7 percent ($6.7 million) in year 1 means receiving payment in year 3 that equals ACO start-up costs, assuming timely CMS financial reconciliation and ignoring interest costs.  Without an external subsidy (such as by a hospital) and ignoring timing, saving 6.7 percent in year 1 would fund start-up costs but not any provider bonuses.

Several factors could worsen the financial outlook.  With ACO patients unidentified, entities have to institute care management and quality improvement strategies for all patients, thereby increasing start-up costs (which CMS acknowledges are understated).  After year 1, if quality scores are not perfect, ACOs might only qualify for 30 percent or 40 percent of savings.  For example, if operating costs remain the same and an ACO scores 70 percent on quality measures (capping savings at 35 percent), saving 6.8 percent would cover year 2 operating costs, with the ACO receiving payment in year 4.

Added Risk Outweighs Benefits from 2-Sided Shared Savings. Compared with the 1-sided model, the 2-sided model increases potential shared savings mainly by increasing maximum savings to 60 percent from 5 percent, and by exempting qualifying ACO savings from the 2 percent threshold. However, the potential upside gain appears inadequate to offset the potential downside loss.  ACOs will prefer 1-sided over 2-sided shared savings because the payoff (“risk premium”) is too small.

An unusual asymmetry exacerbates the downside risk.  Rather than having the same percentage determine upside bonus payments and downside losses, CMS proposes to calculate ACO losses differently.  As a result, if an ACO’s quality score falls below 83.3% (where the loss/gain exceeds 2 percent), the formula has potential losses exceeding gains.  An ACO with a quality score of 70 percent under the 2-sided model would receive 42 percent of any savings but pay 58 percent of any loss.  An ACO with a 30 percent quality score would receive 18 percent of any shared savings but pay 82 percent of any losses.  ACOs not achieving a 30% quality score would pay 100 percent of losses (up to the maximum cap) but remain ineligible for shared savings.

Two potential windfalls, as well as other idiosyncratic factors, might influence an ACO to adopt the 2-sided model.  Because the CMS proposal favors certain geographic areas, an ACO operating in a low cost region or with a low growth rate might opt for 2-sided risk.   Receiving the national growth rate could overstate ACO base year amounts (compared to actually incurred costs).  Similarly, the national (absolute dollar) per capita increase could increase a budget target significantly above expected costs.  Creating an inflated budget target could virtually assure savings, making the risk of losing money small.

New or Primary Care Entities Face Prohibitive Obstacles. As a condition for approving any ACO, CMS requires demonstrating the capacity to repay (within 30 days) at least 1 percent of annual Medicare Part A and B spending.  (ACOs also remain liable for losses, up to a maximum that varies from 5 percent to 10 percent of Medicare spending.)  A newly created physician entity is unlikely to have either a track record with managing risk or sizeable capital reserves, making obtaining a letter of credit or similar financial instrument difficult.

Other factors exacerbate the challenges associated with assuming risk by year 3.  Due to time-lags, an ACO appears unlikely to know its year 1 financial and quality results before year 3.  As a result, the mandated year 3 transition from 1-sided to 2-sided risk occurs before knowing performance in either year 1 or year 2.

CMS proposes that prospective ACOs formed by primary care physicians must assume risk for Medicare spending in excess of budget targets.  Primary care typically accounts for 6—12 percent of cost, while physician-hospital organizations might account for over 80 percent of Medicare spending.  With hospitals typically having access to capital, hospital based ACOs could more readily meet the CMS financial surety requirement than a physician organization (and especially primary care entities).

Unrealistic Quality and Performance Standards

The proposed quality standards include 65 measures grouped into 5 domains:  Patient/Caregiver Experience (7 measures); Care Coordination (16 measures); Patient Safety (2 measures); Preventive Health (9 measures); and At-Risk Population/Frail Elderly Health (31 measures).  The quality standards pose major barriers to qualifying as ACOs, create substantial uncertainty, raise serious questions about discriminating against providers serving low-income or vulnerable populations, and may inhibit focused, real-world clinical improvement.  Only 11 of the 65 proposed quality measures can be met with claims data; 54 require potentially expensive data collection from medical records or surveys.

Although the 2010 “group plan reporting option” (GPRO) involves 36 large groups reporting 26 measures, CMS would expand it to 47 measures and all ACOs.  The timeline appears to require committing substantial resources to develop Medicare ACOs without knowing these group plan reporting option specifications and related CMS reporting requirements.  CMS indicates subsequent guidance will establish the specific standards for quality measures (but current language equates percentile and percent).  Nonetheless, ACO quality scores significantly drive both financial results and continued participation in the program.

The proposed regulation is silent on validating measures for ACOs and handling the potential impact of race and socio-economic status (e.g., Hemoglobin A1c (#36; 40) or blood pressure control (#41) for diabetes).  The Brookings—Dartmouth commercial pilots all strongly recommended focusing on a limited number of high impact measures that improve quality, rather than a relatively undifferentiated “kitchen-sink” approach.  Beyond the cost of capturing medical record data, 50 percent of ACO primary care providers in year 2 must be “meaningful EHR users” (p. 200), a requirement that may disqualify many potential ACOs.

Retrospective Attribution:  Using a Rear-View Mirror

A key issue involves whether ACOs know in advance their patients and performance targets.  If patients are prospectively assigned based on where they have received their care in prior years, CMS can establish a budget target for a defined group of patients before the start of year 1.  If patients are retrospectively assigned based on where they actually received care in year 1 (or subsequent years), membership and budget targets can only be established well after the close of the performance year (e.g., year 1).

Retrospective attribution shifts accountability to after-the-fact “gotcha” scorekeeping.  CMS repeatedly asserts:

[R]etrospective assignment is to encourage the ACO to redesign its care processes for all Medicare FFS beneficiaries, not just for the subset of beneficiaries upon whom the ACO is being evaluated. (p. 114)

If economic incentives don’t matter, why bother with payment reform?  Conversely, if incentives do matter for changing care processes, why tremendously dilute the economic benefit by including patients who aren’t participating in reform?

Retrospective attribution undermines setting explicit performance targets, which require knowing in advance the patients and targets assigned to an ACO.  Retrospective attribution can also lead to selection bias.  Assigning patients after the close of the year does little to provide the tools needed to change provider behavior and culture.  In contrast, providing timely, detailed claims data on assigned patients facilitates performance improvement.

CMS should clarify its plans to provide timely, detailed claims data so ACOs can improve performance.  CMS should also simulate the effects of prospective attribution models, determining their robustness and suitability, as well as evaluating its proposed hybrid of “historically assigned patients” and retrospective attribution against prospective models.

Expensive and Burdensome Requirements

By creating systems of care and attributing patients to systems, ACOs make possible a focus on population health, measurement and accountability that simply cannot exist in fee-for-service.  However, if ACOs do not form, beneficiaries will remain in fee-for-service.

The proposed regulation adds substantial requirements and significant costs, raising barriers to entry which restrict the number and types of ACOs.  Even if CMS shifted to prospective attribution and substantially improved financial terms, the extraordinary range of requirements (including quality measures) brings into question the economic viability of ACOs and shared savings.

At best, shared savings represents relatively modest change to fee-for-service incentives, with limited potential for bonuses.  Adding requirements and costs without evaluating their potential effect on the number and range of participating ACOs is a classic case of the “perfect” endangering the “good”.

Space constraints permit citing only two of many examples.  First, CMS will apply to ACOs Medicare Advantage rules for approving “marketing” materials (broadly defined), which adds significant administrative costs.  Second, based on tax identification numbers (TINs), Medicare physician pay-for-performance (PQRS) bonuses are contingent on an ACO receiving shared savings; this linkage raises significant policy and operational issues.

Timelines and Uncertainty

Drafting, clearing, and having the final regulation take effect after the comment period closes on June 6 will require months.  Added months are required to allow ACOs to apply, and for CMS to review and approve applications.  These steps will most likely delay the start of ACO operations for months after January 2012 – perhaps to July.   The time pressures also mean that many critically important details – such as the proposed GPRO system, the specific quality criteria, the details of data sharing, etc. – will be unknown or not final.

The timelines for critical information and processes add uncertainty and unanswered questions.  CMS proposes allowing 6 months after the close of a year to receive claims.  Only after closing claims can patient attribution and measuring financial and quality performance begin.  Completing these steps may require more than 12 months, putting settlements for year 1 into year 3, and for year 3 into year 5.  As one example, CMS needs to clarify its planned non-renewal process for ACOs with net losses:  how would the policy work, if results aren’t known for 2 years beyond the 3 year agreement?

More generally, the CMS goal of transforming care processes requires time, patience, incentives, and tolerance of mistakes.  Transforming how providers function involves phenomenally hard work and changing culture.  Insisting on short-term results is a sure prescription for failure, especially with modest incentives.


An initial review suggests that the regulation may not hang together in important regards.  Examples of the range of more operational or technical concerns include problems associated with: the proposed use of tax identification numbers; treatment of patients who die in a base year; the timing and availability of data; guidance on materials that need CMS approval; and, the lack of empirical analyses to inform regulatory choices and alternatives, such as comparing alternative attribution models.

A potential “silver lining” exists because CMS generally has enough flexibility to fix these serious defects in the final regulation.   However, salvaging the shared savings program will require significant changes to the current proposed regulation.