The (Unfortunate) Irrelevance of ACOs

Print Friendly, PDF & Email

Richard Evans / Scott Hinds

203.901.1631 /.1632

richard@ /

April 27, 2011

The (Unfortunate) Irrelevance of ACOs

  • Accountable Care Organizations (ACOs) are CMS-designed and sanctioned groups of Medicare providers that cooperate to produce greater quality of care, and savings in which member providers can share. Under the proposed rule there are few (risk-adjusted) savings available, and few if any reasons to cooperate. As proposed, ACOs have little investment relevance
  • Costs of entry are considerable (clinical systems primarily; legal and administrative infrastructure secondarily), as are risks (upside potential for shared savings is, at least after 2 years of operation, matched 1:1 by the risk of having to refund cost-over runs)
  • With a few notable exceptions achievable gains generally are too modest to warrant these costs and risks for primary care physician groups that might want to form an ACO – thus we expect very few ACOs will be formed and driven by primary care physicians
  • Hospitals are a different story; with deeper admin / systems / capital capacities hospitals are better able to bear costs of entry (and risks of over-runs). And, because physicians likely will tend to admit their commercial patients to the same hospital as their Medicare patients (i.e. to the hospital in their ACO), hospitals see spill-over benefits from forming ACOs
  • Thus ACOs form, and hospitals tend to control them – which means care patterns and associated costs are less likely to change. Because the marginal profit from performing a Medicare procedure is greater than the marginal shared savings from not doing a procedure, ACO providers will not try to meet savings benchmarks by reducing their own Medicare charges, but will instead seek to meet benchmarks by reducing other providers’ (including ACO members’) charges
  • Hospitals will not reduce their own Medicare charges – especially if they control the ACOs – but cannot effectively reduce other providers’ charges. As a result, very little has changed, other than the need for hospitals to form (and fund) ACOs in order to secure their referral networks
  • There is a bright spot: to participate in shared savings ACOs must meet quality targets, and these quality targets practically require ACOs to adopt a common electronic medical record (EMR) across all participants; and, the quality targets in no small way may define characteristics of (i.e. create standards for) the EMR. Thus to the extent ACOs form – even if they have little or no effect on care patterns or costs – the standardization and adoption of EMRs almost certainly will accelerate

Table of Contents

I. What is an ACO? 3

A. ACOs must have a legal framework and administrative / clinical systems, but shared / common ownership is not required 3

B. To participate in shared savings, ACOs must meet quality benchmarks 4

C. Benchmarks are limited to Medicare A/B FFS expenditures; ACO providers can share in total spending reductions (i.e. not only reductions in ACO providers’ charges, but charges made by non-ACO providers as well) 4

D. There is a downside … 5

E. …But do they actually work? Experience is limited, but underwhelming 6

II. The Investment Relevance of Accountable Care Organizations 9

A. Why form (or join) an ACO? 9

B. What’s the penalty for not participating? 11

C. Are there spillover (e.g. commercial business line) effects of ACO participation? 12

D. Assuming ACOs do form, how will they change patterns of care and associated earnings potentials? 13

E. Physicians on offense, hospitals on defense! 14

1. The physicians’ playbook … 14

2. The hospitals’ playbook … 16

F. Ready, set … stalemate? 18

III. Investment Conclusions 19

What is an ACO?

The Affordable Care Act (ACA) requires the Secretary of HHS to establish by no later than January 1, 2012 the Medicare Shared Savings Program, which is intended to encourage the formation of Accountable Care Organizations (ACOs) in Medicare. On March 31, HHS and CMS took a major step towards the creation of ACOs with the release of the long-awaited detailed proposed rule on how these entities should be designed and implemented. Although CMS is collecting feedback on the rule until June 6 and the final rule will be issued later this year, the recent communication provides a clear picture of how CMS recommends these entities take shape and the first opportunity to reasonably assess their investment relevance

ACOs must have a legal framework and administrative / clinical systems, but shared / common ownership is not required

The rule defines an ACO as “a legal entity that is recognized and authorized under applicable State law… and comprised of an eligible group… of ACO participants that work together to manage and coordinate care for Medicare FFS beneficiaries and have established a mechanism for shared governance that provides all ACO participants with an appropriate proportionate control over the ACO’s decision making process.” Through voluntary formation of an ACO, a set of providers (participants) cooperates to improve quality and control costs of their beneficiaries’ fee-for-service (FFS) care furnished under Medicare Parts A and B (Parts C and D are excluded). ACO participants continue to receive FFS payments, but if quality meets certain standards and costs are below target, the participants also are eligible to share a portion of the generated savings with CMS. Conversely, ACOs with costs above target will eventually be required to share in these “losses” through penalty payments to CMS. This approach to delivery and payment reform has a “triple aim” of better care for individuals, better health for populations, and lower growth in health care expenditures. But change is expected to come slowly: the rule projects roughly 1.5 – 4 million beneficiaries (of 49 million 2011 Medicare enrollees) in 75 – 150 ACOs, generating net savings of $510 million (against Parts A and B expenditures totaling over $350 billion – an estimated 3% of GDP in 2011) for 2012 through 2014[1]

Under the proposed rule eligible participants include ACO professionals in group practices, networks of individual practices, partnerships or joint ventures between hospitals and ACO professionals, hospitals employing ACO professionals, and other groups of provider and suppliers as the Secretary of HHS deems appropriate. ACOs must commit to at least 3 years of participation and assume responsibility for at least 5,000 beneficiaries in order to provide statistically significant grounds for evaluating real effects of ACO performance versus normal variation. Crucial to ACO design is the proposed method of accounting for beneficiaries: after each performance year (PY) CMS will retrospectively assign beneficiaries to an ACO on the basis of a patient incurring the plurality of their allowed charges for primary care services under an ACO’s providers. As such, the rule strives to preserve patient freedom of choice in the providers from whom they receive care. At the start of each PY, patients are to be informed of their possible assignment to an ACO and given the ability to opt-out of sharing their claims data, which would effectively remove them from the ACO’s beneficiary pool

To participate in shared savings, ACOs must meet quality benchmarks

Progress in enhancing beneficiary quality of care is a prerequisite of ACO eligibility for shared savings. This progress is gauged each PY as ACOs report, and CMS tracks and scores, 65 performance measures across five domains denoted as Patient / Caregiver Experience; Care Coordination; Patient Safety; Preventive Health; and At-Risk Population / Frail Elderly Health. The final domain accounts for nearly half of the total scoring weight and encompasses care in chronic diseases such as diabetes, heart failure, coronary artery disease, hypertension, and COPD. The majority of performance measures focus on process (e.g. percentage of patients aged 18 years and older with a diagnosis of CAD who were prescribed oral anti-platelet therapy), with the remainder assessing outcomes and patient experience. ACOs submit performance data via a combination of patient surveys, claims data, and the Group Practice Reporting Option (GPRO) data collection tool, which would be a new platform based on the current Physician Quality Reporting System

The emphasis on improvement in quality is not to be taken lightly: in order to be eligible for shared savings in PY1, ACOs must fully and accurately report on all 65 performance measures to CMS. In PY2 and beyond additional measures may be added, and ACO quality scores will be determined using a sliding scale of performance versus national FFS or Medicare Advantage (MA) rates (and eventually, national ACO rates) for many measures. E.g. if an ACO’s performance on a certain measure is in the 90th+ percentile of national FFS / MA rates or >90% of a national FFS / MA benchmark, it would receive 2 quality points; performance in the 40th+ percentile or 40% would merit only 1.25 quality points; and below 30th percentile or 30% would yield no points. The ACO’s total score across measures in turn directly influences its shared savings rate (discussed below). ACOs failing to meet the minimum standard on any performance measure are issued a warning and are ineligible for savings that PY, regardless of how successful they were in reducing costs. If a particular measure does not met minimum standards for two consecutive years, the ACO’s agreement is terminated

Benchmarks are limited to Medicare A/B FFS expenditures; ACO providers can share in total spending reductions (i.e. not only reductions in ACO providers’ charges, but charges made by non-ACO providers as well)

While demonstrated quality is required for shared savings eligibility, an ACO obviously also must contain costs in order to generate savings to share in the first place. Savings – and losses – are defined as the difference between actual expenditures for an ACO’s beneficiaries and a benchmark in each PY, with the benchmark intended to approximate what an ACO’s beneficiaries’ expenditures would be in the absence of the program. Thus the proposed methodologies for calculating an ACO’s budget benchmark and actual expenditures each PY are central to its function. CMS first sets an initial budget benchmark for each ACO, which starts with identifying average per-beneficiary total FFS Parts A and B expenditures for an ACO’s beneficiaries (as defined by the plurality of primary care rule) in each of the three most recent years for which data are available (hereafter referred to as benchmark years, or BYs) prior to the ACO’s first PY. These per-capita figures are adjusted for beneficiaries’ health risk characteristics and other factors and trended forward to PY1 using the national growth rate in per-capita Parts A and B expenditures. The average of these three trended BY figures, weighted towards the most recent year (by factors of 60/30/10), is the ACO’s initial budget benchmark, expressed in terms of average per-beneficiary total FFS Parts A and B dollar expenditures. Thereafter, the benchmark is updated every PY by the flat dollar amount equivalent to the estimated absolute amount of growth in national per capita FFS expenditures for Parts A and B. The ACO’s actual observed expenditures over each PY then represent average per-beneficiary total FFS Parts A and B expenditures, adjusted for beneficiary characteristics. Critically, the benchmark incorporates all projected Medicare A/B spending for each beneficiary, whether the care is provided / billed by an ACO provider, or by a provider / supplier outside the ACO

There is a downside …

ACOs must select one of two options for risk-sharing and calculating savings / losses, known as the one-sided and two-sided models. Under the one-sided model ACOs face no downside risk of sharing in losses if their actual expenditures exceed benchmark during the first two years of a three year agreement, but these ACOs’ rate of participation in savings is lower than under a two-sided agreement. Under the two-sided model ACOs are eligible / liable for savings / over-runs in all three years of an agreement. In the one-sided model, if actual expenditures are below benchmark and the ACO has satisfactorily met all performance measures, there are a number of moving parts that determine the amount shared by the ACO. First, in order to account for normal variation in ACO expenditures, actual spend must be a certain percentage below benchmark, called the minimum savings rate (MSR), for shared savings to be attained. MSR is contingent on ACO beneficiary population size, ranging from 3.9% for ACOs with 5,000 beneficiaries to 2.0% for those with ≥60,000 beneficiaries. If actual spend is below benchmark by the MSR, a percentage of the difference between actual spend and benchmark, net of a 2% threshold, is shared by the ACO. The percentage represents a shared savings rate and is equal to 50% in PY1 if all measures are fully and accurately reported; in PY2 and beyond, it is equal to the proportion of total possible quality points achieved by the ACO multiplied by 50%. Additional incentives for participation in Federally Qualified Health Centers (FQHC) and Rural Health Centers (RHC) of up to 2.5 percentage points allow a one-sided ACO to achieve a maximum shared savings rate of 52.5%. Lastly, shared savings payments to an ACO are capped at 7.5% of the benchmark. In sum, ACOs using a one-sided model share in savings once MSR is exceeded, net of a 2% threshold, by a shared savings rate of up to 52.5%, up to a maximum 7.5% of the benchmark

The one-sided model is intended to allow ACOs with limited initial resources for managing costs to still participate in the program, however all one-sided ACOs automatically transition to a two-sided model after PY2. The two-sided model, designed for ACOs with greater experience in managing financial risk, presents downside risk if actual expenditures exceed benchmark, but increased shared savings potential as well. In contrast to the one-sided model, the two-sided model uses a flat MSR of 2.0%, and once exceeded there is no 2% threshold netted out of savings. The two-sided shared savings rate is 60% instead of 50% (although the proportional quality score adjustment still applies); with FQHC / RHC participation incentives of up to 5 percentage points, the maximum sharing rate is 65%. Two-sided model shared savings are capped at 10% instead of 7.5% of benchmark. However, two-sided ACOs also share in losses if actual expenditures exceed benchmark by the minimum loss rate (MLR) of 2.0%. As with two-sided shared savings, there is no threshold percentage netted out of shared losses, hence they are said to be shared on a first dollar basis. An ACO’s shared losses rate is defined as 1 minus its final shared savings rate, which effectively mitigates shared losses for ACOs who nonetheless achieved high quality scores and/or participated in FQHC / RHC incentives. Shared losses are capped at 5% of benchmark in PY1, 7.5% in PY2, and 10% in PY3 (one-sided ACOs transitioning to a two-sided model after PY2 will receive the 5% cap in their first year as a two-sided ACO). Last but not least, the proposed rule recommends a 25% withhold rate be applied to all shared savings payments distributed under both one- and two-sided models in PY1 and PY2, as a mechanism for ensuring ACOs are able to repay CMS for any shared losses incurred under the two-sided model. Withheld payments are available to ACOs at no earlier than the conclusion of the mandatory three year agreement period

It is worth noting the implied autonomy that the proposed rule affords ACOs in devising a method for distributing shared savings / losses payments among their member participants. ACA requires an ACO to have a formal legal structure that “would allow the organization to receive and distribute payments for shared savings to participating providers of services and suppliers”; the proposed rule goes on to state that ACOs must demonstrate a mechanism of shared governance that provides all ACO participants with an “appropriate proportionate control” over the ACO’s decision-making process. However, HSS and CMS also concede in the rule that they lack the authority to specify how savings are to be distributed, instead proposing simply to “require ACOs to provide a description in their application of the criteria they plan to employ for distributing shared savings among ACO participants and ACO providers / suppliers, and how any shared savings will be used to align with the aims of better care for individuals, better health for populations, and lower growth in expenditures”

For a summary of key ACO design elements, please see the Appendix. Examples of how % changes in ACO actual expenditures versus benchmark translate to net shared savings / losses payments under one- and two-sided models are presented later in this call

…But do they actually work? Experience is limited, but underwhelming

Assessing the track record of ACOs is difficult due to the scarcity of truly parallel historical models as well as a persistent lack of agreement on definition. Under a broad interpretation, some have pointed to various alternative payment and delivery models that have achieved varying degrees of success in enhancing provider cooperation and accountability for quality and reigning in costs. Prominent examples include the use of global fees by, e.g. Kaiser Permanente, Intermountain Healthcare in UT, and HealthPartners in MN; bundled acute case rates by, e.g. Geisinger Health System in PA; and primary care medical home fees by, e.g. BCBS of MI. A number of states, perhaps most notably CA, host provider-sponsored organizations which resemble ACOs in that they accept varying degrees of risk for ensuring that a population of patients receives necessary care. However, none of these organizations employ the particular model of two-sided shared savings layered on top of FFS payments that characterizes the proposed rule’s vision for ACOs. There likewise has been a recent upswing in investment in pilot programs labeled as ACOs, such as the efforts undertaken by Anthem BCBS of NH with Dartmouth-Hitchcock Medical Center; Premier health care alliance; and Aetna with Carilion Clinic in VA. Yet details on the proposed workings of these pilots remain scarce. The proposed rule reiterates that alternative payment models are being tested by the Medicare Innovation Center for potential future implementation, but it remains wholly unclear whether any of these will be eventually amended to the ACO’s design as currently drafted

The CMS-sponsored Medicare Physician Group Practice (PGP) demonstration, which served as the basis for ACA’s ACO provisions, provides the most suitable barometer for how we might reasonably expect ACOs to fare in coming years. Initiated in 2005, the demonstration was designed to test the viability of incenting improved process and outcomes along specific measures through Medicare savings-based performance payments in the context of the existing FFS structure of Parts A and B. The study included ten participants representing physician networks, integrated delivery systems, freestanding group practices, and academic medical centers; operating in geographically diverse settings; ranging in size from 232 to 1,291 physicians; and covering 9,700 to 38,700 beneficiaries assigned on the basis of receiving the plurality of their evaluation and management (E&M) services at that site. The participants operated under a one-sided risk-sharing model and, after surpassing a 2% MSR, were eligible for a maximum 80% shared savings rate, with the rate adjusted by the proportion of quality targets that were met. 32 quality targets were established from MA plan benchmarks or subjective improvement objectives and included primarily process measures related to diabetes, coronary artery disease, heart failure, hypertension, and preventive care. Per-beneficiary FFS expenditures were benchmarked against those of comparison groups that drew their beneficiaries from the same geographic areas as the participants

HHS reported summary results from the project’s fourth year in December 2010, revealing that all ten participants met at least 29 of the 32 quality goals, and that three met all 32, in that year (PY4). To further contextualize the performance payments in PY4: although HHS did not disclose target or actual expenditure levels for that year, the mean rate of shared savings versus target for sites that earned performance payments in PY2 and PY3 was roughly 3.5%; assuming a similar rate in PY4, the $31.7 million in performance payments would therefore represent a 3.0% increase over the actual FFS payments to earners alone.. The success of the participants in meeting quality goals in the fourth year marked a continuation of the demonstration’s earlier years, as the sites achieved an average of 90% of their quality targets in just the first year of the project. However, the trend in shared savings tells a different story: only two sites shared a total of $7.3 million in performance payments in PY1, and four sites shared $13.8 million in PY2 (none of which were the five integrated delivery system sites). Only two sites achieved performance payments in all four years: University of Michigan faculty group practice, an academic medical center; and Marshfield Clinic, a rural freestanding group practice. Participants attributed savings to factors ranging from organizational structure to redesigned care processes to changes in market conditions

The PGP demonstration yields a number of clues and concerns about the current proposed rule. The success of the PGP sites in achieving quality targets is obviously encouraging; although the proposed rule doubles the number of performance measures, engineers them to cover a broader array of outcomes and patient experience, demands more complex reporting requirements, raises the bar to 100%, and eventually benchmarks many of them against real FFS and MA rates versus some of the comparatively subjective targets of the PGP. The demonstration also raised serious questions around the ability to distinguish real ACO performance from favorable pre-demonstration cost trends as significant determinants of performance payments. Ideally this is tempered by the proposed rule’s use of a more robust initial benchmark calculation methodology as well as higher MSRs

On the other hand, higher MSRs and lower shared savings rates in the proposed rule only amplify what the PGP demonstration suggests is already a rather steep uphill battle to achieving any, let alone economically meaningful, shared savings. Direct comparisons are certainly discounted due to the demonstration’s use of comparison groups to define and grow budget benchmarks; but the reality that nearly all PGP savings were generated by outpatient, not inpatient, services clearly suggests an operational disadvantage to ACOs that include a hospital in their group. Arguably the most glaring difference between the proposed rule and the PGP is the inclusion and eventual enforcement of a two-sided versus one-sided risk-sharing model, which almost certainly has pivotal implications for provider willingness to join an ACO and the sustainability of an ACO for those that do join, yet fail to generate early savings (particularly with the presence of the 25% withhold rate). Lastly, the PGP demonstration and separate studies point to the joint hazards of assigning beneficiaries based on the plurality of primary care rule and holding ACOs responsible for beneficiaries’ total FFS expenditures. In the demonstration’s first two years, beneficiaries incurred, on average, 85.0% of their allowed E&M charges under PGP-based providers. More alarming, in a 2007 study of Medicare beneficiary patterns of care that offered early evidence to support ACOs’ creation, Fisher et al. found that 73% of beneficiaries’ visits for E&M services took place within a primary hospital or involved its extended multispecialty medical staff; and, on average, 64% of admissions were to the primary hospital.[2] In other words, it is not unreasonable to estimate that one quarter of the average beneficiary’s E&M visits and over one third of hospital admissions are routed through non-ACO providers. The fact that substantial portions of the average beneficiary’s care and incurred charges are controlled imperfectly, at best, by an ACO’s providers further calls into question an ACO’s ability to generate shared savings

The Investment Relevance of Accountable Care Organizations

Whether ACOs have investment relevance is a two-layered question: 1) whether ACOs in fact form to any meaningful extent; and 2) whether ACOs substantially change either total Medicare payments, or the share of payments that flow to various products and services

Regarding the first question, the incentive to form ACOs appears quite weak, as shared savings payments net of costs and risks seem quite small. This tends to tamp down the relevance of the second question, i.e. if incentives are weak, so are the forces that might reduce or re-distribute Medicare payments. That said, because the strength (and direction) of incentives varies by provider, there appear to be at least isolated pockets of ACO influence that may affect healthcare portfolios

Why form (or join) an ACO?

Being able to share in reduced Medicare costs v. trend has strong conceptual appeal; yet under the proposed rule, the magnitude of available savings (which fall with time as care becomes more efficient, since ACO’s are benchmarked v. their own history), once netted against the costs (e.g. IT) and risks (ranging from less than expected savings to actual payment of penalties) of being an ACO, strike us as unappealing to the average Medicare provider

In accordance with the proposed rule, Exhibits 1 and 2 show potential net savings to an ACO (y-axis) as a function of total reduction in Medicare A/B charges for care provided (x-axis) to the ACO’s beneficiaries (Exhibit 1 is the one-sided model; Exhibit 2 is the two-sided model). Note that both exhibits likewise show the range and median actual percent reduction in Medicare A/B charges achieved in PY3 of the PGP demonstration project: if we make the reasonable assumption that real-world savings will be in this range, we get some idea of the (rather limited) potential for ACOs to participate in shared savings. Throughout this call, it’s important to bear in mind that achieving these savings comes at considerable marginal cost in terms of ACO formation (particularly with respect to underlying clinical systems), that ACOs are benchmarked against their own historic performance over time (meaning they get paid for being more efficient in one agreement period, but must beat that level of efficiency to get paid in the next agreement period); and, that ACOs face the risk of having to pay back a portion of cost over-runs (even one-sided ACOs, which must ‘go’ two-sided in the 3rd year of their first contract cycle, and remain such for all subsequent years or contract cycles). Also bear in mind that because Medicare payment rates to providers are fixed, that ACOs cannot reduce Medicare charges by reducing these payment rates, and so must work to reduce charges by managing volume and mix alone. This is a considerable constraint; as recently as 2009, volume and mix were less than 60 percent of total Medicare cost growth (Exhibit 3)

Confining our analysis to Medicare payments, i.e. ignoring any spill-over effects into commercial business, the (highly stylized) best candidate to form an ACO is a large group practice: 1) whose potential ACO assignees over-consume poorly coordinated care; 2) who admits its Medicare patients to teaching institutions more often than not; and 3) who has decided for reasons not entirely ACO-related to reduce duplication and inefficiency in the care its patients receive beyond the four walls of its own group practice – a rare animal indeed. #1 assures that the trailing three year period from which the spending benchmark is calculated is inflated relative to what can be achieved through low-hanging-fruit changes to the practice. #2 is simple exploitation of the rules; Medicare patients admitted to teaching hospitals tend to have higher payments for several reasons, not least of which being the inclusion of teaching subsidies in these institutions’ DRG calculations (Exhibit 4); thus by admitting to non-teaching institutions, a practice can realize considerable savings in total Medicare payments made on behalf of its beneficiaries. #3 – having decided for reasons beyond the ACO to invest in better care protocols and the associated systems – is necessary simply because of the very high risk that the long-term financial gains from being an ACO, even in this stylized instance, will not offset the costs of the systems necessary to achieve these gains

What’s the penalty for not participating?

Alternatively, and still ignoring any commercial spill-over effects, we might suspect that there is a Prisoner’s Dilemma: i.e. that groups choosing not to form an ACO will be dis-advantaged if others in the area do form ACOs. At least with respect to payment rates this appears not to be the case; Medicare payment rates to a non-ACO provider in a community where others form ACOs are unlikely to meaningfully decline (or even grow more slowly) simply because ACOs emerge in that community. I.e. the non-ACO provider should have little fear that ACO’s will create efficiency gains and drive down local re-imbursement[3] rates to a point that the non-ACO provider, not having invested in gaining efficiency, can no longer make a profit. Each ACO’s benchmarks are set using its trailing three year per-beneficiary Medicare expenditures and subsequently updated annually by expected rates of growth in national per-beneficiary spending over the three year agreement period. Thus only if so many ACOs form nationally that Medicare begins projecting lower national FFS growth is the non-ACO provider’s re-imbursement meaningfully affected; even then, the non-ACO provider (particularly if it’s a primary care physician practice) is unlikely to have a different cost structure than ACO providers (we’ll explain why shortly), and so is not at a relative operating economic dis-advantage versus its ACO peers

The story is at least somewhat different for hospitals, as referral patterns for both Medicare and commercial patients are likely to be influenced by ACO membership, to the extent ACOs form. Thus if a competing hospital forms an ACO and another does not, the non-ACO hospital faces some non-trivial risk of losing Medicare (primarily) and commercial (secondarily) admissions. This risk does not apply to non-ACO physicians, as it’s highly unlikely that ACO-affiliated hospitals will deny admitting privileges to competent non-ACO physicians

Are there spillover (e.g. commercial business line) effects of ACO participation?

Being unable, within the confines of Medicare, to find a compelling and broadly applicable reason for ACOs to form, we’re forced to look outside Medicare – to the impact of ACO formation on commercial lines of business

The most obvious – though illegal – route to realizing gains in ex-Medicare lines of business would be to cartel with the (otherwise competing) providers organized within an ACO to extract price concessions from commercial payors

The proposed ACO antitrust enforcement policy released by the Federal Trade Commission (FTC) and the Department of Justice (DoJ) highlights this concern, and suggests there will be little tolerance from regulators for collusive behavior outside of activities directly within the scope of the ACO. Specifically, provider actions that would raise competitive concerns include:

  • “anti-steering”, “guaranteed inclusion”, “product participation”, “price parity” clauses that discourage commercial payors from directing patients to providers outside the ACO
  • “tying sales of the ACO’s services to the commercial payors’ purchase of services from providers outside the ACO (e.g. requiring a purchaser to contract with all the hospitals in the same network as the hospital that belongs to the ACO)”
  • “with an exception for primary care physicians, contracting with other ACO providers on an exclusive basis…discouraging them from contracting outside the ACO”
  • “restricting commercial payors’ ability to make available to enrollees cost, quality, etc. information”
  • “sharing competitively sensitive pricing data that could be used to set prices for services provided outside the ACO”

Further, ACOs that are subjected to an antitrust review will be expected to demonstrate the restrictions in place to “prevent participants from obtaining information regarding prices that other participants charge commercial payors that do not contract through the ACO”

Given the prominence of the concerns about commercial collusion and the specificity with which these are addressed by the proposed enforcement policy, we are generally skeptical of ACO participants’ ability to broaden or strengthen their anti-trust protections simply by becoming an ACO. This is especially true considering the “make-or-break” power held by FTC / DoJ: if, upon review, antitrust regulators decide they are likely to challenge or recommend challenging the proposed ACO, CMS cannot (per its own regulations) approve the ACO for the Shared Savings Program. Despite the fact that ACO formation does not appear to improve anti-trust protections, we nevertheless see a likelihood that the impact of ACO formation on Medicare referral patterns is likely to spill-over into commercial referral patterns – e.g., primary care practices within an ACO presumably will admit their Medicare patients preferentially to hospitals in that same ACO, and we suspect these same practices will tend to admit their commercial beneficiaries to the same (ACO-participating) hospital

Assuming ACOs do form, how will they change patterns of care and associated earnings potentials?

Recall that ACOs share in savings if total Medicare per-ACO-beneficiary spending falls below the benchmark – i.e. ACOs can produce savings by either reducing their own Medicare billings, by reducing the Medicare billings that other providers / suppliers render on behalf of the ACO’s beneficiaries, or some combination of the two. We believe that under the current rule ACO providers will prefer to reduce others’ charges rather than their own; this fosters competition where cooperation is needed, and reduces potential savings

It’s not me, it’s you

Starting with the question of whether an ACO provider might benefit by reducing their own Medicare billings, we compared the marginal profitability of a $100 procedure with the maximum potential savings to the ACO provider of not doing that procedure. Physician practices generally run a net margin in the 12 percent range, or about 18 percent pre-tax, meaning 82 percent of revenues go to cover costs. Fixed / variable cost ratios are about 66/34, which suggests that about 28 percent of revenues go to cover variable costs. Since we’re doing the analysis on the margin, we assume the fixed costs are irrelevant, meaning the marginal pre-tax profit from the $100 procedure is $72, or after marginal (35%) taxes, about $47

The alternative to performing the $100 procedure and netting $47 now is to forego the procedure, and take the resulting shared savings. Assuming a perfect shared savings scenario – the savings from the procedure are themselves marginal, i.e. they are beyond the 2 percent minimum savings rate; the ACO provider has a perfect quality score; the ACO provider is in a two-sided risk model and thus shares 65 percent of the shared savings; and, the ACO provider is in fact the whole ACO – her organization has no one else to share the savings with — then her economic gain from not doing the procedure is $59 (Exhibit 5). The $59 gained from not doing the (marginal) procedure is greater than the $47 from doing the procedure; however no ACO provider can realistically expect to keep 100% of the shared savings, as these have to be spread in some proportion across all of the ACO’s participants. And, the $59 gain makes no provision for risk, i.e. that the savings achieved may be lower, that the savings rate may be lower, or that the provider may in fact have to pay a penalty. After reducing the $59 by the primary care physician’s participation rate in the ACO’s shared savings, and adjusting the dollar figure for risk, we see no way that the gain from not doing a procedure can be greater than the gain from doing the procedure

This leads to a simple rule: ACO providers, under the proposed rule, have no incentive to reduce their own billed services. Instead, a rational, self-maximizing ACO provider will work to lower costs by reducing Medicare payments to providers / suppliers other than the ACO-provider[4] itself

Physicians on offense, hospitals on defense!

The smaller your own charges are as a percent of total Medicare spending (physicians), and the larger your influence is over charges other than your own (especially primary care physicians), the more ideally situated you are as an ACO provider. Conversely, the larger your own charges are as a percent of total Medicare spending (hospitals), and the weaker your influence over charges other than your own (again, hospitals), the less suited you are as an ACO provider – at least with respect to your ability to produce savings

The physicians’ playbook …

Primary care physicians are roughly 8% of Medicare payments (see Exhibit 6 for total share of Medicare payments to physicians; we assume primary care is roughly 40% of physician payments, or 8% of total Medicare A/B payments), but their influence on Medicare charges made by other providers is far broader – thus primary care physicians presumably should be quite motivated to form ACOs, if they believe they can deliver care at costs below the benchmark by reducing the volume and mix of care provided (by others) to their beneficiaries[5]. And generally speaking, the more control primary care physicians have over a given cost, the more likely that cost is to fall (or at least grow more slowly). For example – we would not expect to see a reduction in office visits to primary care, since we don’t expect ACO providers to reduce their own services. Nor would we expect to see a reduction in retail prescriptions, as Part D expenditures are not included in the savings / cost analysis. We would expect to see some reduction in laboratory and radiologic charges for services ordered by primary care physician groups from third-party providers. However this effect is limited by the relationship between primary care practice size and ownership of laboratory and radiologic equipment: the larger a primary care practice, the more capable it is of forming / participating in an ACO, and the more likely it is to do its diagnostic work in-house (Exhibit 7)

As we move beyond the products and services immediately and directly controlled by the physician, we come to the next less immediate / less direct layer of primary care physician cost influence, namely the decision to refer a patient to a specialist, or admit a patient to hospital. In the former case, presumably ACO primary care physicians will be slower to make specialist referrals, and will preferentially refer to specialists who deliver a more modest mix of products and services (e.g. physical therapy instead of a spinal implant, palliative care instead of Provenge). Obviously primary care physicians’ power here is limited for two simple reasons: it’s going to be quite difficult to find specialists who reliably offer an adequate but less expensive mix of care, and even if such specialists can be found an ACO’s patients maintain the right to see whichever specialist they like, whenever they like

In the case of hospital admissions, primary care physicians have somewhat greater influence. As we mentioned before, under the proposed rule[6] practices that currently admit to teaching institutions can reduce hospital charges simply by shifting their admits to non-teaching institutions, as Medicare re-imbursements are lower for non-teaching hospitals. More importantly, where a specialist referral generally means a hand-over of patient control from the primary care physician to the specialist, a hospital admission may not necessarily be such a complete transfer of control. We believe primary care physicians in ACOs can and should place hospitalists in the hospitals they use, for the obvious purpose of controlling the care of (and charges related to) the ACO’s beneficiaries. As we’ve shown, the hospital itself has no incentive to reduce its own charges, thus the need for primary care practices to have their own physicians directing in-patient care wherever and whenever possible

Thus beyond their own charges (which no one wants to reduce) crudely speaking physicians have two ‘layers’ of cost influence: layer 1 referring to direct / immediate control over products and services ordered on behalf of beneficiaries (e.g. clinical laboratory testing and radiology); and, the less direct / immediate ‘layer 2’ of deciding when to refer (to specialists) or admit (to hospital), deciding which specialists or hospitals to use, and working (e.g. via hospitalists) to control hospital charges

The hospitals’ playbook …

Recall that ACO’s ‘form’ around primary care practices; beneficiaries are assigned to ACOs according to the ACO affiliation of the primary care practice from which the beneficiary receives the plurality of its primary care services. Thus an independent hospital that owns no primary care practices cannot participate in an ACO unless it is invited to do so by a primary care practice. The independent hospital might encourage its participation in a given ACO by using its relatively greater administrative / systems / capital capacities to provide the primary care practice(s) with the resources needed to form the ACO in the first place; i.e. in many geographies we suspect primary care physicians may not be able to form ACOs without the help or even initiative of local hospitals

On the other hand, hospitals that own or are otherwise directly affiliated with primary care practices can form ACOs, since beneficiaries can be assigned to these owned practices – provided the owned practices have enough assigned Medicare beneficiaries to qualify (5,000) and operate efficiently (perhaps more like 10,000) as an ACO

Also, bear in mind that hospitals can ‘join’ multiple ACOs, where primary care providers can ‘form’ only one. Thus an independent hospital that owns no practices might help providers in its community form a series of ACOs by offering its admin / systems / capital resources to those practices, and the hospital could participate in all of the resulting ACOs. Or, a hospital that owns practices might form one ACO with its owned practices; and, like the independent hospital, reach out into the community and form one or more additional ACOs by helping to organize independent primary care physician groups

Under the proposed rule, we view hospitals’ motive in forming (or participating in) ACOs as having more to do with gaining (or defending) market share than with affecting care patterns (and related costs) for Medicare beneficiaries. ACO participation cannot contractually influence the hospital’s share of non-Medicare admissions, though as a practical matter ACO participation likely will affect commercial admissions; ACO rules variably require or encourage physicians and hospitals in a given ACO to work under somewhat common protocols / procedures / systems, and this commonality is likely to influence commercial as well as Medicare care pathways, meaning an ACO-participating primary care practice’s commercial admission patterns are at least somewhat likely to mirror its Medicare admission patterns

But once a hospital has formed (and/or joined) an ACO (or ACOs), its motives for ACO participation have largely been met. The hospital has secured or expanded its referral base, but has no further motive to meaningfully change its approach to care – in particular, per our “First Rule of ACOs”, the hospital, as the primary cost center in Medicare beneficiaries’ care, has no reason to moderate the volume or mix of services it provides. The only way hospital-based volume and mix are likely to moderate is if primary care physicians have sufficient control over hospital care patterns, which raises the issue of balance of power within the ACO. Where ACOs are formed around hospital-owned physician groups, we see primary care physicians being less able to direct patterns of care in the hospital. Similarly, ACOs that are formed between independent practice groups and hospitals, but largely as a consequence of hospitals initiating formation by providing admin / systems / and capital resources, also are likely to have an internal balance of power that tilts in the direction of the hospital instead of the primary care physician; so here too we see limited odds of hospital care patterns being made more efficient. Only in the (potentially somewhat rare) circumstance of primary care physician groups that form ACOs independently, i.e. without relying on hospitals’ resources, would we expect primary care physicians to have meaningful influence over the mix and volume of hospital services provided to the ACO’s beneficiaries

On net, we see hospital-formed ACOs as being more defensive than offensive. Under the proposed rule, hospital-formed ACOs presumably will work to expand / defend share of admissions in the local market, and once established will tend to defend the volume and mix of care provided by the hospital to Medicare beneficiaries. Given the large percentage of total Medicare A/B charges that are hospital based, we believe that hospital-formed ACOs are inherently unlikely to create meaningful savings versus a benchmark

Ready, set … stalemate?

Under the proposed rule, where cooperation is needed, competition prevails. What primary care wants hospitals and specialists do not want; what hospitals want, primary care does not want. Stalemate is a likely result

If physician groups faced no or limited entry costs and risks associated with ACO formation, then the prospect of sharing in the savings achieved by reducing other providers’ billings would be quite attractive, since other providers’ billings are 90% or more of Medicare charges. The problem is that primary care physician groups do face considerable entry costs and risks, which when netted against achievable shared savings, may make ACO formation less or even un-attractive. Primary care initiated ACOs (‘PCP/ACOs’) require capital for legal frameworks, clinical guidelines and associated systems, and reserves to demonstrate an ability to cover any responsibility for cost over-runs. Even large and well-funded PCP/ACOs cannot expect to meet quality standards without at least some level of cooperation from hospitals, and smaller and less-well funded PCP/ACOs may not be able even to form without relying on local hospitals’ administrative / systems / and capital resources. Thus it’s unlikely PCP/ACOs can form without including hospitals in some fashion; at a minimum this means sharing of (i.e. diluting) any potential net savings with these member hospitals, and more often than not likely means ceding some level of hospital mix / volume control to the hospital. Since PCP/ACOs cannot expect hospitals to reduce Medicare volume and mix on their own initiative, the more power hospitals have in ACOs, the less likely anything is to change – i.e. stalemate

Even in the idealized circumstance of a PCP/ACO that might beat its first three year contract benchmark by repairing its history of excessive charges and by shifting admits from teaching to non-teaching hospitals (or even by postponing expensive care in the third year), we have very real concerns that the PCP/ACO would not be motivated to sign-up for a second three year agreement. Presumably it would have been ‘paid’ for resolving its legacy inefficiencies in the first three year agreement, the teaching to non-teaching arbitrage can only be played once, and any care postponed from the 3rd year of the first agreement cannot be postponed for the duration of the second three year agreement. Any mix-volume reductions the PCP/ACO achieved in its first three years logically will have come from those decisions it most directly / immediately controls, i.e. reduced diagnostic (clinical laboratories and radiologic) charges, followed by reduced rates of specialty referrals and hospital admissions, complimented if possible by some level of control over volume and mix of hospital services on a case-by-case basis. Beyond this scope of influence the PCP/ACO has little or no control, but having presumably exhausted the savings in its scope of influence in the first contract cycle, even this idealized PCP/ACO would be at a stalemate with specialists and hospitals by the second cycle

Hospital-controlled ACOs would presumably reach a stalemate more quickly, within the first contract cycle. Once the ACO is formed and referral networks thus established / defended, the hospital is likely to defend its ‘internal’ volume and mix trends, and thus can only profitably[7] achieve savings v. benchmarks by reducing volume and mix beyond its four walls. Hospitals have little influence on care patterns beyond their four walls, so if hospitals defend care patterns within their four walls, there’s precious little room for savings – again, a stalemate

Investment Conclusions

Importantly, the proposed rule may change, thus our conclusions assume the rule remains as-written, and are provisional pending the final rule

Broadly speaking, we believe that ACO formation will be limited by what appear to be relatively weak returns net of the costs and risks of forming and operating an ACO. Those ACOs that do form are more likely to be initiated by hospitals than by primary care practices, for the simple reason that hospitals are more likely to have the requisite administrative, systems, and financial capacities, and are more likely to enjoy positive commercial spillover effects. Where ACOs are formed by primary care groups, we see these being more often than not opportunistic and short-lived; i.e. we expect primary care formation of ACOs will be largely limited to those practices whose benchmarks ‘benefit’ from relatively inefficient legacy patterns of care that offer easy ‘low-hanging-fruit’ efficiency gains. Opportunities to shift admits from teaching to non-teaching hospitals, and of greater concern third year delays in necessary but costly care may add to potential savings. Most, perhaps even all of the more potent opportunities for primary care groups to earn shared savings are one-time, thus such ACOs are relatively un-likely to hang around for a second three year cycle, particularly given the greater risk in the second contract cycle of having to refund cost over-runs (easy savings are less available, and ACOs must be two-sided in their second three year contract cycle). To the limited extent that primary care driven ACOs do in fact form, we would expect to see reduced utilization of clinical laboratory services (e.g. DGX, LH), reduced use of third-party radiographic services, reduced rates of specialist referral, reduced rates of hospital admissions, reduced re-admission rates, fewer Medicare charges per admit, and a greater tendency to discharge from in-patient to home care as opposed to more expensive sub-acute settings – in other words, all the things we expect ACO designers might hope for. Presumably less desirable, and as mentioned several times, we would expect to see a shift in admissions from teaching to non-teaching hospitals. Despite these rather dramatic potential shifts in care patterns associated with physician-directed ACOs, we emphasize that such changes are unlikely, again because we have little expectation that physicians will form a substantial number of lasting ACOs under the proposed rule

Instead we see hospitals as the most common organizers behind ACO formation, the primary motive being to secure and defend (Medicare and commercial) referral networks. Once formed, hospital-directed ACOs are unlikely to meaningfully change patterns of care; under the present shared-savings calculus ACO providers generally and hospitals specifically have no motive to reduce their own charges, and hospitals’ influence on care patterns beyond their own four walls is quite limited. Thus savings are very limited, and existing care patterns remain largely unbothered

To our minds the single clear investment opportunity across the entirety of potential ACO effects lies with the electronic medical record (EMR). Despite our belief that ACOs will tend to be hospital-dominated entities with little real effect on care patterns, to be an ACO these organizations nevertheless must meet basic organizational structure and quality requirements. ACO’s are required to have some type of clinical decision making process that is shared across all of the ACO’s member providers; this requirement plainly forces the adoption of EMR standards within an ACO, will most likely catalyze EMR standards within metropolitan centers, and may ease the pathway to a national EMR standard. ACO quality requirements that emerge in the near future, even in the second and third years of the first-generation ACO agreements, are likely to explicitly assume that a common EMR exists across the ACO, and thus are in effect de facto requirements for ACOs to adopt EMRs having certain (ACO quality standard related) characteristics. Thus to the extent that ACOs form, EMR standardization and rates of adoption both are likely to accelerate

  1. 2010 Medicare Trustees Report.
  2. Fisher ES et al. “Creating accountable care organizations: the extended hospital medical staff.” Health Aff (Millwood) 2007;26(1):w44-w57
  3. In fact, ironically, because of how geography-specific inputs to DRG payment rate calculations work, payments to all practices in the area likely go up if ACO’s reduce the intensity of care in the area
  4. N.B. we’re careful here to say ACO provider rather than ACO. If I’m a primary care physician in an ACO, and you’re a hospital in my same ACO, I have every reason to reduce your Medicare charges, and no reason to reduce my own. In fact, at least with respect to my marginal economic motive to reduce charges, I’m equally willing to reduce your (‘intra-ACO’) charges as I am to reduce charges of any non-ACO provider that might bill Medicare for products and services provided to the ACO’s beneficiaries
  5. Prices, of course, are centrally determined
  6. Inevitably this arbitrage will bring comments from teaching hospitals, thus the final rule may not offer ACOs the opportunity to save by shifting from teaching to non-teaching hospitals
  7. I.e. net of opportunity costs
Print Friendly, PDF & Email