INTRODUCTION

In 2005, the Centers for Medicare and Medicaid Services (CMS) launched “accountable care organizations” (ACOs). The intent for this voluntary program was to financially incentivize provider groups to reduce fee-for-service billings while maintaining care quality. Some analysts believed ACOs to be a pivotal mechanism to control Medicare spending.1 Others questioned the theoretical basis.2 Recent ACO assessments conclude that 2019 net savings are minimal despite selection and attrition bias (greater drop-out of ACOs that failed to achieve savings)3, programs both make and lose money for Medicare4, ACOs widely vary in implementation and performance5, and savings evaporate when bias is removed6.

However, no report broadly reviews ACO financial performance. We collect and compare financial performance data from all four CMS ACO programs from 2005 to 2018, examining net CMS cost: gross savings in medical billings minus “bonus” payments to ACOs.

METHODS

We assembled all estimates of savings and losses reported by CMS and the Medicare Payment Advisory Commission (MedPAC) in June 2018 and June 2019 reports to Congress, and in academic evaluations (full citations on request). Costs include all medical care within ACOs. We standardized findings to percentage net cost (or savings) to CMS. We excluded studies that report only gross savings because they cannot reveal whether CMS saved or lost money. We compared “benchmark” analyses (which contrast ACO costs to past spending for the same individuals) and “counterfactuals” (which use concurrent non-ACO Medicare beneficiaries as controls) because of claims by some analysts that counterfactuals reveal real savings obscured with benchmark methods.

RESULTS

Findings are in Table 1. The Physician Group Practice (PGP) demonstration saved CMS 0.3% net over 5 years (2005–2010). Pioneer saved CMS several tenths of a percent net from 2012 to 2016, without the ACOs that dropped out, most of which were losing money. The Medicare Shared Savings Program (MSSP) lost CMS money for 4 years, and in year 5 (2017) yielded CMS net savings of 0.3%. Next Generation saved CMS 0.2% net in its first year and lost 0.3% over the full program.

Table 1 Accountable Care Organization Evaluations of Net Cost to CMS, 2005–2017
Full size table

There was little difference between the two analytic methods. The range for both benchmark and counterfactual approaches was CMS 0.7% savings to 0.3% added costs. For program years with both methods, results differed by 0.1%.

DISCUSSION

We found that overall, ACO programs roughly broke even from the CMS perspective. That is, when bonuses CMS paid to ACOs are subtracted from gross savings, the programs lost money or saved no more than a few tenths of a percent. We found very similar results for the “benchmark” and “counterfactual” approaches.

Most reports on ACO financial performance obscure how programs affect CMS spending. Many focus on gross savings, omitting or only tangentially mentioning net saving/cost. Some reports present only absolute dollars, without percentages, creating a misimpression of large effects. Some include only ACOs that saved money, excluding those that did not, and some reports omit ACOs that dropped out.

Our analysis has important limitations. First, we could not include three studies because they reported only gross savings. Second, we omitted ACO overhead costs, which are not reported. While they do not affect CMS net costs, they do influence provider costs. MedPAC estimates overhead for ACO programs at 2% of Medicare spending.7 If ACOs save CMS and providers at most a few percent, this implies that ACOs are actually raising systemwide costs.

We recommend that future ACO evaluations adopt a consistent reporting framework, including gross savings, offsetting bonuses, and net costs in both dollar and percent terms. Studies should report overhead, including education, training, monitoring, data collection, analysis, and reporting.

On the policy side, it is time to draw the ACO experiment to a close. We now have a decade of impressive empirical evidence demonstrating minimal if any benefit from ACOs of several designs. This model arose from laudable goals, drawing on theories of human and organizational behavior. However, reality as measured in formal evaluations failed to align with theory. ACOs have not worked for Medicare and will not work for the broader health care system. Indeed, CMS recently ended the Next Generation ACO program. However, CMS is launching another unproven program, direct contracting, to replace the open structure of traditional Medicare.

We propose instead turning to empirically proven tools used by other high-income nations to control health care costs. These include uniform benefits, simplified financing that dramatically reduces payer and provider overhead, removal of profit for primary insurance, regulation of pharmaceutical prices, and public control of capital investments. Over the past 45 years, other wealthy countries have, far more than the USA, achieved improved equity combined with slowed cost growth and extended longevity.