Evaluating ACO Performance, 2016 Edition

(A review of 2015 ACO results appeared on The Health Care Blog on October 1, 2016.)

The Medicare Shared Savings Program (MSSP), or Accountable Care Organizations (ACOs), continue to be CMS’ flagship pay for performance (P4P) model delivering care via 432 MSSP ACOs located in every state to over nine million, or 16%, of Medicare beneficiaries. This year the agency did not announce 2016 performance year results. Instead, CMS posted without notice in late October a Public Use File (PUF) or spread sheet summarizing 2016 performance. What analysis CMS did provide was by CMS’ vendor, the Research Triangle Institute (RTI), several weeks ago to ACO participants via webinar. RTI’s slides are not made publicly available.

Like performance year one (2013), two (2014), and three (2015), performance year four (2016) once again produced limited positive results. As stated last year, CMS does not evaluate the ACO program, therefore, ACO participants and Medicare policy analysts are left to decipher how success was achieved, what performance results mean for the MSSP program and in context of the agency’s overall efforts to reduce Medicare spending growth.

2016 ACO Financial Performance Results

Here is a bulleted summary of 2016 financial performance based on the PUF and RTI’s slides.

  • In 2016 there were 432 ACOs that had their performance year results reconciled.
  • Of these, 410 were Track 1, six were Track 2 and 16 were Track 3.
  • Of the 432, 134 earned shared savings or 119 out of 410 Track 1s, six out of six Track 2s earned shared savings and nine out of 16 Track 3s earned shared savings. Four Track 3 ACOs owed $9.33 million in shared losses. Only 129 actually received shared savings checks because five of the 134 owed CMS for advanced ACO payments.
  • Physician only ACOs once again were more successful than ACOs that included a hospital, or 41% versus 23% respectively.
  • Also again longer tenured ACOs were more successful. Among the 2012-2013 ACO class 42% were successful compared to 18% of the 2016 starters.
  • The 134 2016 ACOs earned in sum slightly more than $700 million in shared savings. Actual savings paid out was close to $650 million because imperfect quality caused ACOs to leave money on the table and because of Medicare reimbursement or sequestration cuts required the 2011 Budget Control Act.
  • For 2016 30% of participation MSSP ACOs will receive a shared savings check compared to 29% in 2016, 26% in 2015 and 27% in 2014.
  • Earned shared savings were again highly concentrated. The 15 highest performing ACOs received $265 million total in shared savings as compared to the 15 lowest performing shared savings ACOs that received $20 million in total.   An August DHHS Office of Inspector General (OIG) report made note of this dynamic, i.e., about half of the spending reductions during the first three years of the program, or $1.7 billion, were generated by 36 ACOs and three ACOs in that group generated a quarter of the amount.
  • Of the remaining 294 2016 ACOs, 107 fell within their positive Minimum Loss Ratio (MLR) corridor, 105 fell within their negative MLR corridor and 82 fell outside their negative MLR corridor. This last group, the worst performing ACOs, was 19% of all 2016 ACOs, significantly less than the 24% of the worse performing 2015 ACOs.
  • Again, success was largely determined by an ACO’s financial benchmark. ACOs that earned shared savings in 2016 had a reconciled benchmark 10% higher than all other ACOs, or respectively $11,614 per beneficiary versus $10,563 per beneficiary, or a benchmark 7% higher than those within their positive MLR corridor and 12% higher than those that fell below their negative MLR. The OIG report reached the same conclusion. During the first three years of the program, ACOs that received shared savings had a $11,748 per beneficiary benchmark compared to a $10,284 per beneficiary for ACOs that did not receive shared savings, a 12% difference. As noted last year, because of this successful ACOs only had to comparatively spend a trivial amount less than their financial benchmark to be successful.

2016 Quality Performance 

  • Year-over-year, or 2015 compared to 2016, quality performance improved on 27 of the 34 ACO quality measures.
  • Five of the seven declining scores were in the care coordination/patient safety domain.
  • Mean quality score for the 330 ACOs subject to pay for performance scoring had an average quality score of 94.9%. This score is inflated since first year ACOs, or 98 ACOs in 2016, received a 1.0 score because first year participation is simply pay for reporting.
  • As in previous years, ACOs earning shared savings did not measurably outperform all other ACOs, All other 2016 ACOs had a mean quality score of 94.6%. Interestingly, those ACOs that fell below of their negative MLR had a mean quality score of 98.6%.

Did the 2016 ACOs Save Money?

In an October 31 Modern Healthcare article titled, “CMS Loses Money As Medicare ACOs Remain Risk Adverse,” Maria Castellucci argued the Medicare Shared Savings or ACO program “paid out more in bonuses to accountable care organizations than the savings those participated generated.”   Since CMS pays Track 1s, for example, only 50% of total savings in shared savings, and because savings generated within the positive MLR are not paid out in shared savings, the only way the program can pay out more in bonuses than spending saved is if total spending below the negative MLR is deducted from total savings. For example, if total savings was $100, shared savings paid out was $50 and spending below the negative MLR was $60, the program in sum would net a negative $10 – though this net amount ignores savings generated within the positive MLR. This method of calculating ACO performance is not uncommon.

This logic however is problematic for at least four reasons. First, Medicare is not budgeted spending. Therefore it is false to assume spending can be defined as excessive or beyond budget. Second, Track 1 ACOs by design are not at financial risk therefore it’s false to assume those that fall below their negative MSR over spent. Third, there is no counter-factual. No one knows what spending would have been for providers had they not participated in the ACO program. Similarly, there is no control group to compare ACO spending against. Lastly, as Michael Chernew recently noted, the program does not financially account for spillover effects, the impact ACO savings have on reducing Medicare Advantage benchmarks and savings derived from improvements in care quality.

How Was Success Achieved

Again, while CMS details financial and quality performance results the agency does not explain in any detail how results, favorable or unfavorable, were achieved. Program participants are expected to translate the ACO model into practice or find solutions that satisfy the new financial and quality performance standards. RTI’s 2016 “trends in total expenditures” slides provide some insight.


RTI’s analysis, confirmed by the OIG’s August report, shows once again ACOs that earned shared savings had markedly different utilization in certain categories compared to all other ACOs, particularly the worst performers or those that fell below their negative MLR. Short and longer term inpatient hospital expenditures along with ED visits that led to a hospitalization were down -4.4%, -6.5 % and -8.8% respectively. For the worst ACO performers these were up 8.8%, 44% and 2.3% respectively. Rehabilitation hospital use also varied considerably from 0.6% for shared savings ACOs to 12.7% for those falling below their negative MLR. Skilled nursing, home health and hospice utilization were all comparatively better for shared savings ACOs or: -18.3% v. 16.3%; -9.7% v. 9.9%; and, -6.1% v. 12.4%. Skilled nursing utilization days differed from -21.5% to 10.3% respectively.   Interestingly, Part B drug spending was up substantially for all four performance categories, or between 9.1% and 16.7%. Again, as in last year, regardless of performance, all ACOs made far greater use of primary care visits using a nurse practitioner, physician assistant and/or certified nurse specialist, these were up between 20.1% and 39.9%. Clearly, ACOs are working to reduce higher intensity care, over-all utilization including imaging and working to increase mid-level clinician utilization.

As I wrote last year, ACO performance remains at best mediocre or marginal when you consider moreover total ACO participant spending in 2016 was over $81 billion. 2016 savings were a fraction of 1%. CMS would still benefit from formally evaluating the program and do so in a transparent manner. CMS would benefit by working to correlate quality performance with financial results such that providers are better able to succeed and as Farzad Mostashari recently noted financial performance would likely improve if CMS did not effectively cap risk adjustment. The agency should also work toward knitting together, or finding synergy between, the MSSP and the agency’s other pay for performance initiatives particularly its bundled payment demonstrations. Per MedPAC’s recommendations CMS should either find a way to level the playing field between the MSSP and Medicare Advantage (MA) or work with equal fervor to measurably improve MA quality and value beyond its current MA Value-Based Insurance Design (MA-VBID) demonstration.