Why ACO Savings Aren’t About Location.


One of the big questions since the inception of the Medicare Shared Savings Program has been whether the model would only work in regions with extremely high baseline costs.  Farzad’s state-level analysis of earlier MSSP results suggested that ACOs in higher-cost areas were more likely to receive shared savings. It’s one of the questions that Bob Kocher and Farzad received in the wake of the op-ed on Rio Grande Valley Health Providers last week.

So we decided to dig into the data.

We’re still waiting for CMS to make baseline costs for ACOs – and the local areas they serve – public. But in the meantime, we linked each ACO to a Hospital Referral Region using the main ACO address provided by CMS – and took a look at the region’s per capita Medicare costs as a predictor of ACO success.

What we found is that baseline (2012) per capita costs were not a predictor of ACO success in achieving shared savings. Several ACOs had considerably lower costs than their risk would have projected, and they still achieved savings. ACOs with higher costs than their risk would suggest failed to achieve savings.

As we build the first Aledade ACOs, we’re focusing on finding the right doctors, identifying which of their best practices can be exported to our other providers, and finding ways we can make workflow in their offices smoother and more efficient.

That’s why this is encouraging news – the evidence confirms that the single biggest factor in ACO success are the actions of the affiliated doctors – not simply starting from a high cost baseline.

Travis Broome is a researcher with Aledade.

Spread the love

21 replies »

  1. Do you think the actions of affiliated doctors resulted in the crash of this earlier ACO model? In the link I’ve added Farzad acknowledges how many of the doctors switched to other models. What is your take on this? http://goo.gl/UQbFUW

  2. Thanks Jeff… lovin’ the exchange! Just sayin’ metrics, metrics. depends on lens….

    Bottom-line is we still live on a production driven healthcare ecosystem – ‘capitation’ (PMPM) still a fraction of total contract spend (even if you include ‘lite versions’ ie, bundled payment, DRGs, or ambulatory case rates, or OWA’s .

    Share of GDP has been and continues to disproportionately claim an obscene allocation of the U.S. (public, private) spend and growing; all while a grand COST SHIFTING CHARADE proceeds under the convenient ‘consumer directed/skin in the game’ brand play by payors/health plans/or more aptly put ‘benefits solutions providers’.

    There are no more ‘health insurers’ per se. they’ve collectively failed to manage clinical risk. PERIOD. They are ‘transaction processors’ increasingly living off of ‘fees’ and investment returns as ‘banks’, with the great hope that ‘technology plays’ (mhealth, digital health, tech-enable patient engagement), etc… can cure the beast.

    So yes, today and in the near term, clever (and well paid) manager’s are subject to production incented growth or share objectives (even amidst declining units primarily due to the slowing (and cost shift) economy and less discretionary spending for elective services).

    The handful of creative ‘comp plans’ that scaled the transformative shift from volume to value remain a fraction of total [see my piece ‘Eating Glass’ http://acowatch.me/2014/08/14/eating-glass-a-davita-healthcare-partners-hiccup-or-physician-integration-implosion/ about Craig Sammit’s abrupt departure from DaVita/Healthcare Partners ] are at least on the table given the ACO triple aim sustainability mission. If units decline, skilled managers find ways to drive UP price. Consolidations are precisely that, no?

    I remember when per diems and case rates were first introduced back in the 80s. The CFO calculus was pretty simple: budget revenue requirements divided by projected units of service and voila, you got your case-mix service tiered or global per diem for contracting. Pretty simpleton, but true.

    When shift to ambulatory from inpatient began, Outpatient surgery/procedure case rates were benchmarked to historical inpatient revenue yield. Only growth of physician owned ASC’s forced some competitive restraint to price [ and that theme remains alive today via OIG report: http://www.beckershospitalreview.com/finance/oig-says-bring-down-hopd-rates-for-surgery-to-asc-rates-cms-disagrees-11-things-to-know.html ]. That the aggregate trend UP is rather obvious, no? It has not abated from a total cost of care perspective – the only measure that really matters.

    Thank Jeff. I am not an economist, just a grunt in the c-suite who negotiated a fair amount for global (hospital, physician, ancillary and pharma) full risk downloads (from license entity to delivery system) via multiple health plans in different states.

    Things don’t seem to change much in the ‘healthcare borg.’

  3. Jeff:

    Totally agree…. the financing paradigm must shift to align and change root behavior driving the spend. And if we can use the recent ‘bundled payment’ metrics published by CPR (Coalition for Payment Reform) a 40% penetration of payor contracts reflecting some VBP (value based payment) is progress. While bundled payment remains a sidebar as the Zeigeist can only shift when the incentives align with desired outcomes. FFS is the problem. Quality assured, outcomes focused, accessible via high value networks under ‘capitated [PMPM] terms and conditions, are indicia of the likely remedy.

    Still see ACOs as kindergarden or even mezzanine boot camp offering a range of structural transformational options to tame the ‘rapacious appetite’ of our predominantly fee-for-services fueled delivery system.

    • Rapacious appetite? Inpatient utilization has fallen for the last five years.
      Outpatient services use would have flatlined if not for the observation unit reclassifications. Imaging unit volume growing maybe 1-2% a year.
      Even if ACO’s worked, the diagnosis for which they were the presumed prescription is at least five years old.

      If you look a little closer at CPR’s data, they are basically counting “fee for check the box” as value based payment. So they withhold some of your money at the front end and give it back to you if you check the right number of boxes. That’s not managed care either.

      Most provider organizations are not now and will never be successful risk bearing enterprises. It takes a unique combination of leadership, market circumstances, partners and customers (the latter is the most serious problem right now) and maybe twenty years of hard work to make the transition to a capitated model.

      Gregg, it’s not happening this time.

  4. Ah some talented peeps in this thread. Hesitate to venture in, but here goes….

    Not much to add other than budget driven targets, baselines, etc., are always a ‘devil in the details’ construct in the imperfect world of ‘data’ presentation. Milliman and others have been handsomely remunerated to vet cost & utilization data in proprietary stacks for many at risk players both then (80s -90s) and now.

    By my recall the AAPCC (adjusted area per capita cost) was -5% (the MA contractor got 95%) during the TEFRA rollout (including Medicare Choice+). Today’s ACA ACO/MSSP minimum savings targets before distribution a tad less ambitious, no?

    So regardless of the internal consistency or representational accuracy of cost, quality and outcomes (savings vs.loss scenarios) data isn’t the broader question what are the essential structural characteristics of an ACO given it’s geo-political footprint and share standing particularly in relationship to institutionally led competitors? Whether low hanging fruit is more plentiful in lower per capita cost markets, really isn’t the issue…

    The issue is ‘business as usual’ medicine is NOT on table (and will never again be placed before a governing board in our life times). The only question is whether you start with an MSSP ‘toe in the water’ (hedged to best efforts risk adjusters) or end run to MA. The underlying ‘clinical integration’ is not optional in our present iteration of ‘managed competition’, though the definition of what passes for a ‘CIN’ is wide open to local expression.

    Thanks for the piece Travis! Keep up the good work at Aledade. Too many dismiss the ACO as too little too late ‘HMO-lite’. I disagree.


    • Gregg,

      Managed care without downside risk is like vegan BBQ. It’s the risk-the global budget- that forces you to make choices. Absent the forcing function, what do you have? Fee for service running in the background, and a “wouldn’t it be nice” approach to saving money. It isn’t managed care.


      How are we doing on our new chart savings vs. aapcc?

  5. AAPCC is not a measure of total health care spending…

    from the Brookings Institution..


    The biggest issue is that the ‘energy of activation’ to possibly get benefits, and then the size of the benefit, is difficult to justify — so ,yet again, another reason that federal regulation is driving consolidation since regulatory (activation) costs are so large that unless you are massive, you cannot ‘sink’ those costs to a manageable level within an organization.

    • That is correct.
      AAPCC is a measure of per capita MEDICARE spending.
      It’s what the MSSP, a MEDICARE payment program, is trying to reduce.

  6. Stinting is something we have to be constantly vigilant about, but don’t know that we can assume that savings in a low cost area are a result of stinting. The shared savings part of an ACO definitely has a shelf life. Lots of disagreement over how long it is, but I think most agree that it is finite.

  7. But you don’t want cost savings in places where costs are low. To achieve this means you are skimping on care.

    The entire exercise of ACOs has to be temporary because one cannot keep lowering costs year after year nor can one keep improving quality year after year.

    In the big picture, the ACO effort has to be a one time event.

  8. Framing the question as “whether the model would only work in regions with extremely high baseline costs” creates a false dichotomy.

    I think Farzad got it right in the Brookings article you link to:
    “3. It’s easier to cut costs if you start high
    “While individual ACO benchmarks have not yet been released (something we strongly encourage), there is some evidence that ACOs in the highest cost states are more likely to be achieve shared savings. The states with the most expensive (risk adjusted and standardized) regions for Medicare are Florida, Louisiana, Mississippi, and Texas. ACOs in these states account for 25 of the 114 ACOs (22%) but include 10 (34%) of the 29 ACOs with shared savings (p~ 0.07). While reducing costs in high-cost areas is an important policy objective, achieving physician participation in alternative payment models nationwide may require CMS to consider modifications to the baseline calculation formulas in the next round of ACO rulemaking (expected this fall).”
    I also think your conclusion is misleading. “The evidence confirms that the single biggest factor in ACO success are the actions of affiliated doctors – not simply starting from a high baseline”. Even if grant you that the “right docs” is #1 for success, NOT having the payment formulas to optimize risk/reward will also be a deal killer for ACOs.

    Finally, I think your data is weak in supporting your conclusions. You make “savings/NOT savings” savings a dichotomous variable in your analysis. From the ACO POV, the more important factor is the perceived risk/reward ratio of developing an ACO. The feedback we are getting from the market is that this ratio is way out of whack – ACO’s see very little potential for a big upside. The fact that they can achieve a small amount of shared savings (triggering a YES in your analysis) is perceived to be out of line with the amount of risk and effort.

    Bottom line: regardless of whether you have the right docs, many ACOs will fade into the sunset UNLESS CMS adjusts the formulas to improve the risk/reward ratio from the ACO POV.

    • That was why I asked for the AMOUNT of savings, and the correlation with the market’s baseline costs. The real problem as Vince said is the negative ROI for the vast majority of MSSP/Pioneer participants on their start up investment. The MSSP is Tom Sawyer’s fence painting project. . .

      The other problem, of course, is the concentration of savings. 3% of the participants generated 50% of the savings, and 15% of participants generated 80% of the savings. It’s even more concentrated than the Physician Group Practice demo was.

    • The savings would have to cross the 2% threshold for actually getting savings. This simple chart says nothing more than some ACOs have crossed that savings threshold have been able to do so despite having lower actual costs for the HRR for the corporate HQ address than their HCC score would have indicated compared to other ACOs. This is the first snippet blog to a much deeper analysis that is ongoing and hopefully will include more ACO specific data if we can coax it out of CMS.

      • CMS published ACO-specific savings amounts. AAPCC=average adjusted per capita costs. You have half. Just graph the actual amount of the savings CMS released and you’ve got another, more illuminating, graphic.
        WIll bet you a latte that there’s a statistically significant positive correlation.

  9. I like that question. We will look into it. We haven’t integrated AAPCC into our data, but it would let CMS do a lot of the risk adjustment for us.

  10. Robert that is absolutely possible. We eagerly await ACO specific data. We plan on diving into this more and more as more specific data becomes available and looking into proper statistical testing once we get better data.

    • Travis,

      What was the correlation co-efficient between the AMOUNT of shared savings (not just the bonus) and the AAPCC of the market in which they operated?

  11. Couldn’t a high-priced system (and ACO) be located in a region with average, or lower to average, costs? I think you need system-specific cost data to make an assertion like your conclusion. But tell me if I’m wrong.

  12. suggesting that the much vaunted Dartmouth data is seriously flawed when it implies ‘excessive’ spending in certain places…

    and what broader conclusions would that suggest?

Leave a Reply

Your email address will not be published. Required fields are marked *