Editor’s Note: This is the first installment of a series of blog posts by Ron Klar offering suggestions on how to make the Medicare Shared Savings Program a more viable vehicle for the creation of accountable care organizations. This installment focuses on (1) the sharing model(s) for ACOs, and (2) the beneficiary issues of assignment, identification, and choice. You can also read the second installment, which focuses on the financial issues of (1) computing the benchmark for ACO payments, (2) adjusting expenditures during the performance years, and (3) determining savings (and losses). The third installment focuses on alternatives regarding (1) evaluating quality, including the measures, performance scoring, and standards, (2) determining shared savings (and losses) rates and limiting payments (and losses), and (3) sharing CMS data and information.

My Health Affairs Blog of April 7 expressed the view that the long-awaited Medicare Shared-Savings Program proposed rule released on March 31 “creates a program that is likely to have few participants”.  This view also appears in virtually every written commentary since.  There seems to be the widespread opinion that interested provider groups either could not qualify or would not want to try under the terms presented.  This skepticism extends even to “the nation’s highest-profile health care centers” generally thought ready to participate, including the Marshfield Clinic, the Cleveland Clinic, the Mayo Clinic, Intermountain Healthcare and the Geisinger Health System (as reported by “CQ Healthbeat” on May 3 and 6).

What I did not challenge, nor have other commentators, is the significant potential and prospects for groups of providers to form and function as accountable care organizations (ACOs) to achieve the “three-part aim” of better care for patients, better health for populations, and slower growth in costs.  The clear purpose of the Medicare Shared Savings Program (MSSP) is incentivizing care process enhancements by adding the possibility of supplemental performance-based payments to continued fee-for-service.  The MSSP is not another “managed care” variant; the MSSP is not “MA-lite” (referring to Medicare Advantage).

Given the tremendous expression of interest and support for this concept, while recognizing the uncertainties, it is important to capitalize on the opportunity of this public comment period (until June 6) to advocate for “a major makeover” to create a program that can more likely lead to a “four-part success”: meaningful participation by providers; widespread acceptance by beneficiaries; prerequisite achievement of quality performance; and documented earning of savings. Therefore, rather than extend any further critique of the terms of the proposed rule, below and in two follow-up posts I will instead offer specific suggested changes, with accompanying rationale, to stimulate consideration among potential participants, affected stakeholders, and policy analysts as they formulate their positions to send to the Centers for Medicare and Medicaid Services (CMS).

Part One below will address (1) the sharing model(s), and (2) the beneficiary issues of assignment, identification, and choice. Part Two will address (3) the financial issues of the benchmark and determining/allocating savings (and losses), (4) the issues of quality measurement and performance, and (5) the content and timing of CMS data and performance information. Part Three will address (6) the issues of eligibility for participation, and (7) the issues of program monitoring and administration.  To save space, I will not repeat the content of my April 7 and January 25 blogs.

Suggestion A: The MSSP Should Include a Shared Savings (SS) Model (With or Without a Shared Savings & Losses (SS&L) Model)

I firmly believe the shared savings model (without losses) is desirable and viable and, if properly structured, could be a good deal for providers, beneficiaries, the Medicare program, and the public.  So do many stakeholders.  It must be included in the final rule to assure “meaningful” participation, the prerequisite step to the four-part success.  Then why was this model not included in the proposed rule?  [Track 1 is not a true SS model, as the third-year changes to a  SS&L model and a 25 percent withhold would be applied to any year one and/or two shared savings, which would either be applied to any year three losses or forfeited if the ACO withdrew or was terminated.]

The answer revolves around two concerns: (a) “unearned” savings, i.e., having actual expenditures less than benchmark expenditures without implementing the coordinated and redesigned care processes (“winning the lottery”), and (b), “insufficient” incentive to perform, i.e., inadequate “skin in the game” and no “down-side” risk.  So let’s address these concerns directly in the SS model, rather than ignore the clear Congressional intent and eliminate this model before even giving it a chance.  [The Physician Group Practice (PGP) Demonstration was not this chance as the details of so many its fundamental features, e.g., computing and updating the benchmark, the participation requirements, and the performance measures and standards, were very different from those which will be part of the MSSP.]

The first concern is valid but easily satisfied with four safeguards (the first three of these are included in the proposed rule, although (1) and (2) are in an overly prescriptive form, which I will address in Part Three) and others I will suggest below: (1) require “evidence of capability”,  which I suggest (# A1) as a detailed description by each ACO of the specific processes they will have operational at the start and a listing and timeline of others planned to become operational, (2) require a “certification of compliance”, which I suggest (# A2) as an attestation by each ACO each year to both having operationalized their described and listed processes in (1) and satisfied the other participation requirements included in their application, (3) establish a size-related (number of assigned beneficiaries) minimum savings rate (MSR) to mitigate year-to-year random variations, and (4) establish a benchmark that accurately represents the historical expenditures of the beneficiaries actually assigned and whose actual expenditures are used in comparison, which I will suggest (in Part Two) as a modified “Option 2”.

The second concern is largely one of theoretical interest rather than empirical reality.  As many have already noted, CMS’s own impact analysis cited initial start-up and first year operating costs of $1.8 million.  However, as this figure ignores ongoing operating costs, and was determined by the GAO in 2008 for the PGP Demo sites, it is likely seriously understated for two reasons: (a) these sites were already organized and functioning entities, and (b) there are many more operational and performance requirements proposed for ACOs.  By any measure, such amounts will be substantial, and will be more than any rational business would invest, along with their substantial effort and time, just to participate.  Finding funds in such amounts in advance has itself also been raised by some as a barrier to their participation.

As importantly, I believe the nay-sayers can be satisfied by adding the incentive, and disincentive, of public transparency.  I therefore also suggest (# A3) that there be included in the MSSP an “ACO Compare Website”, with a planned posting of each ACO’s care enhancement processes, detailed performance information, and any operational infractions, e.g., “corrective action plan(s)”.  Even an organization considering participating in the MSSP as a “loss-leader” (to reap some other presumed benefits) would not want to jeopardize their reputation by being perceived as a poor performer or one under special monitoring.


Suggestion B: The SS&L Model Should Be Offered by CMS as an ACO Option

I continue to believe, as I argued in my prior blogs, that the SS&L model has fundamentally different implications to provider groups and beneficiaries than the SS model.  As such, it should only be considered as an addition to, not as a substitute for, the SS model.  An SS&L model could be accomplished within the MSSP, although it would complicate the operations by having two concurrent tracks with different features.  Furthermore, recognizing that there has been absolutely no experience with the SS&L model, I suggest (# B1) that it might be more prudent to treat this as an experiment first, utilizing the authorities of the Center for Medicare and Medicaid Innovation (CMMI), before including this model in the MSSP.

However, because of the momentum of having the SS&L model implemented somewhere, either concurrently by the CMMI or as part of the MSSP, I will proceed with this blog series including it.  [I assume that a “partial capitation” model for ACOs will also be tested by the CMMI, as they should, but I will not include any suggestions for it in these blogs so as not to further complicate the discussions.]


Suggestion C: The Timing for the SS Model Should Be Purely “Retrospective”

Probably the one issue in the proposed rule drawing the most attention, as it did in the CMS “Request for Information” last November, is the timing for beneficiary assignment.  My earlier blogs, particularly the January one, discussed “retrospective” versus “prospective” timing and my rationale for the former and against the latter.

Many have reported that the proposed rule includes “retrospective” assignment, i.e., identification, after the close (as opposed to before the start) of the performance year, of beneficiaries whose care and claim experience “will count” toward the ACO’s performance.  However, this is not entirely accurate.

A more accurate portrayal of the proposed assignment rule would be “prospective-possible plus retrospective-reconciliation”.  That is, it has two parts: (a) before the start of each performance year, a list of beneficiaries who would have been assigned during a pre-ACO period or those who were assigned during an earlier-performance year would be given to the ACO, and (b) after the end of each performance year, based on the services during that performance year, the identities of beneficiaries in (a) who are not still assigned, and those who are newly-assigned who were not in (a), would be given to the ACO.

The basic problem with “pure” prospective assignment (no reconciliation after the end of each performance year) in the MSSP, which cannot have a “lock-in” (any restriction on beneficiaries seeking any covered services from any non-ACO practitioner, provider, or supplier) is that it would (a) not give ACOs accountability for additional beneficiaries they take responsibility for during the performance year, and (b) give them accountability for beneficiaries they were no longer responsible for.  Because of the significant “time-lag” to administer assignment prospectively, this ends up being very significant.

An evaluation of the year-to-year changes in beneficiary assignment within the PGP Demo, which relied on the same prospective-plus-retrospective rule, was one-third of beneficiaries (due to both changes in choices and circumstances) were different in a succeeding year from the previous.  The rule mentions that when assignment was simulated with PGP data using the proposed criteria of primary care services by only primary care physicians, the changes were one-quarter.

The proposed rule includes a six-month claims data run-out to capture virtually-complete experience.  Therefore, any prospective assignment determination would be at least two years out-of-date; for example, the claims data used to determine a prospective assignment list for the CY2012 performance year would have to be from 2010 (or earlier).  Thus, it can be estimated that between 44 percent [1 – (0.75 x 0.75)] and 55 percent [1 – (0.67 x 0.67)] of beneficiaries assigned before the start of a performance year would not still be assigned after the end of that performance year.  This should be unacceptable to both ACOs and to CMS.

So why not have the prospective-plus-retrospective option, which CMS suggests this “is a good compromise” (page 157)?  Because a compromise is not a solution.  Most importantly, I believe having the “prospective-possible” still creates an undesirable distinction among Medicare beneficiaries.  Under any circumstances, they would be treated differently, and that is not the expectation in “traditional”, non-managed care, Medicare.

Additionally, this creates a real possibility for unacceptable intra-Medicare “cost-shifting” by an ACO to non-prospective-possible beneficiaries, especially those who become hospitalized by non-ACO physicians.  That is, ACOs could increase the amount and intensity of services to beneficiaries to whom the ACO was or is not also providing primary care.  Finally, by knowing that the “retrospective-reconciliation” will be forthcoming, ACOs might be tempted to try to “manage” their risk for these possibly-assigned beneficiaries by minimizing primary care services to their highest utilizers and maximizing these to their lower utilizers.  I find even the possibility of this incentive very disturbing.

For these reasons, and those detailed at length in my earlier blogs, I suggest purely retrospective assignment for the SS model.  I will, however, make a different suggestion for the SS&L model.

Suggestion D: There Should Be No Identification to ACOs of Specific Beneficiaries for the SS Model, i.e., It Should Be “Invisible”

Again, because of the administrative lag-time to be able to identify the retrospectively-assigned beneficiaries, and the desire of CMS (they state as “a presupposition”) and among stakeholders to preserve beneficiaries’ freedoms of choice and equality, I find no good reason to ever identify beneficiaries to a SS model ACO.  Invisibility will promote uniformity.  Of course, providers already know the identities and the conditions of the patients they are taking care of, and thus can still personalize their programs.  I suggest (# D1) that if assignment is to be both retrospective and invisible, we refer to these beneficiaries as “attributed beneficiaries” (sounding more accurately like “after the fact” and “for analysis purposes”).

I do believe, however, that ACOs are entitled to know that CMS, through its contractors, has accurately determined assignment along with all other administrative activities, including the determination of benchmark and actual expenditures and measurement of quality performance.  Therefore, I also suggest (# D2 & D3) that (a) an independent contractor be retained to scrutinize and audit these processes, and (b) ACOs have the right to appeal the attribution (and expenditure and quality information) of a “limited number” of specific beneficiaries, i.e., 5-10 percent of their total.

Suggestion E: The Timing for the SS&L Model Should be Purely “Prospective”

Most providers believe, and many have proposed, that when “downside risk” is present in an ACO program, it is essential that beneficiaries are assigned prospectively.  The usual reasoning revolves around (a) “targeting” advance planning and monitoring of these persons, and (b) engaging them in the ACO’s care processes.  I tend to agree, but only if these fee-for-service beneficiaries are given a choice; CMS must not make such a decision for them.

Therefore, I suggest (# E1 & E2) that beneficiaries be (a) notified by CMS (rather than the ACO, to avoid any issue of method, message, or influence) that they have been “tentatively aligned” with an ACO, and (b) extended the choice to “opt-out”, i.e., to continue their care with the ACO’s professionals and providers but not have their experience included in the ACO’s performance assessments.  These beneficiaries would also be encouraged to discuss this decision with their ACO physician(s) and the “ACO Relations” staff (a feature I will suggest in Part Three).

If these beneficiaries choose not to opt-out, I suggest (# E3) that we refer to them as “aligned beneficiaries” (sounding more accurately as not “locked-in”).  As such, it will be made clear that they continue to retain the freedom of fee-for-service Medicare to choose to receive services from professionals and providers not part of the ACO.  The ACO program is not a managed care plan; the absence of a “lock-in” is the difference. Of course, it will be necessary and appropriate for these SS&L ACOs to do everything they can to gain the trust and cooperation of their “aligned beneficiaries”.

With these provisions, I suggest (# E4) that it is both feasible and fair to have no “retrospective- reconciliation”, except for beneficiaries who move out of the area (and only for the proportion of time they are out of the area).  This avoids the disturbing possible incentive to “manage risk” discussed above.  Additional beneficiaries they become responsible for will subsequently become “tentatively aligned”; unless benchmark Option 2 is adopted (which I will address in Part Two) there should be no other reconciliation.  Unfortunately, pure prospective assignment increases the possibility for the intra-Medicare cost-shifting; I will also offer suggestions to deal with this in Part Three.


Suggestion F: The SS Model Should Incorporate a “Majority” Method

As discussed in the proposed rule (page 159), relying on a majority of primary care services is considered “too strict a standard” that “would likely somewhat reduce the number of beneficiaries assigned”.  However, I believe a plurality could lead to the undesirable consequence of accountability without responsibility whenever the percentage is less than the majority.  By definition, a “plurality” is simply more than any other, and the proposed rule did not recommend any minimum percentage.

Also, it must be recognized that the percentage of primary care services is not of any one ACO physician, but of the collective sum of all qualifying ACO physicians (whether primary care, or all as I suggest).  Therefore, it is not unreasonable to expect that when one compares a beneficiary’s use of primary care services of all of the ACO’s physicians (by all of their ”participant” taxpayer identification numbers (TINs)) to those of each non-ACO physician TIN, that there will be a majority.  Requiring a majority also resolves any assignment controversy in areas that may have more than one ACO, considered by most to be a very desirable development.

Furthermore, given the earlier-stated concern by some for SS model ACOs “winning the lottery”, requiring a majority would be a helpful mitigating feature.  Finally, as I am suggesting retrospective timing for the SS model, reducing (possibly) the number of assigned beneficiaries should also be satisfying to those with this concern.

Suggestion G: The SS&L Model Should Incorporate a “Minimum Plurality” Method

For all the reasons stated above for the SS model, I believe that “majority” could also make sense for the SS&L model.  Additionally, if “tentative alignment” is based on a majority rule, “opting-out” should be much less likely as the ACO is already rendering “most” of the primary care services to these beneficiaries.  Nevertheless, in the interest of increasing the number of “aligned beneficiaries”, desirable to both ACOs and CMS, and in combination with pure prospective timing, I believe that “minimum plurality” makes more sense.  I suggest (# G1) that the minimum percentage is between thirty and forty.

Suggestion H: Beneficiary Assignment for Both the SS and SS&L Models Should be Based on the “Primary Care Services” of All Physicians (Who Are Exclusive to the ACO)

The “basis” for beneficiary assignment is one of the most critical issues to the participation and operations of an ACO.  My earlier two blogs contained detailed, and what I believe are already compelling, arguments supporting this suggestion.  Nevertheless, here are some new ones.

Most importantly, this policy would stimulate organizations that form ACOs to have patient care physicians of all specialties included or affiliated (as ACO “members” or “participants”).  This would have very beneficial consequences.  For one, physicians of the full range of specialties delivering “primary care services”, as defined by specific procedure codes in the proposed rule, would then be “tied-in” to the ACO’s agreements with CMS and the ACO’s systems and care enhancement processes.  In addition, the number of assigned beneficiaries would likely increase with this basis, as some beneficiaries do not receive any care from primary care physicians.

For another, this broader assemblage of physicians would then be represented in both the governance and the opportunity for shared savings (and losses).  This would clearly promote a situation of having all physicians “work together to manage and coordinate care for Medicare fee-for-service beneficiaries” [section 1899(a)(1)(A)].  While a more limited range of physicians, along with primary care practitioners, may be the norm for the “Patient-Centered Medical Home” project, an ACO is a different and more comprehensive approach to reform (even though aspects and principles of PCMH will certainly be included in ACOs).

Another new argument is the inappropriate reliance for guidance in the proposed rule (pages 149-151 as posted) on section 5501 of the Affordable Care Act (ACA), “Expanding Access to Primary Care Services and General Surgery Services”.  First, this is a separate provision of ACA than section 3022 which created the MSSP, and there are no cross-references in either section to the other.  If the Congress intended such guidance, they would have included it.  Second, section 5501 deals specifically with services “furnished by a primary care practitioner” [italics added here and below for emphasis], defined to include others besides “physicians”, which is specifically different from section 1899(c) and (h)(1)(A).  Finally, section 1833(x), created by section 5501, separates the definition of “who” is a “primary care practitioner” [paragraph (2)(A)(i)] from “for whom primary care services accounted …” [paragraph (2)(A)(ii)]; thus, these are distinct concepts, and should be treated as such.

Still another new rationale is to be consistent with related provisions within the rule itself.  The “data collection tool” proposed for assessing many of the proposed quality measures is “GPRO”, the group practice reporting option (see pages 197-199).  This is a carefully-conceived and fully-operational methodology that was developed for the PGP Demo but since extended to other groups.  With it, Medicare patients for each of the clinical conditions are considered for the group’s reporting on the measures if the patient had received a plurality of their primary care services (billing codes) from physicians of the group without regard to their specialty.

Finally, the rule states on page 221 that CMS “considered proposing to limit incorporation of the Physician Quality Reporting System incentive under the Shared Savings Program to the ACO’s group practices that were used for beneficiary assignment”, but decided instead to include all associated groups and eligible physicians (EPs).


The MSSP for ACOs has turned out to be a very complicated and controversial endeavor, and one which deserves the best possible collaboration of thought between those inside and outside of government.  I believe that we can find a new set of features, terms, and conditions that will achieve both the “three-part aim” and a “four-part success”.  We must seize the opportunity Congress has extended to improve the quality and efficiency of care to Medicare fee-for-service beneficiaries.  Saving the shared savings program would take advantage of this opportunity and accelerate needed reforms of our overall delivery system.