ACOs--promise, not panacea.

IN ITS LANDMARK 2001 REPORT, CROSSING THE QUALITY Chasm, the Institute of Medicine not only documented high rates of defect in American health care but also wrote, stunningly, “In its current form, habits, and environment, American health care is incapable of providing the public with the quality care it expects and deserves.” The term incapable is an indictment: profound, uncomfortable, and warranted. America’s forms of care are largely soloists, but patients need symphonies; its habits of care are of excess, but society needs elegance; the environment preserves the status quo, instead of encouraging change. That US health care nonetheless manages to work as well as it does is testimony to the dedication of professionals and to a workforce that somehow makes the best of a bad situation. Now, I believe and hope, the nation (not government, alone, but rather the public and private sectors together) is going to repair this. It has finally become intolerable not to do so, the price of inaction is too high for our present economy and our posterity. The transition will be messy because so many interests are so deeply invested in the “incapable” status quo, and, equally, because although a great deal is known about what better forms and habits look like, what is not known with equal certainty is what environment can best nurture them. The accountable care organization (ACO) is a guess. As described in the Affordable Care Act and elaborated on in regulations that the Centers for Medicare & Medicaid Services (CMS) issued during my time as administrator, ACOs combine incentives for cost saving; encouragement for cooperation among primary care physicians, specialists, hospitals, and other care clinicians; choices for patients; and surveillance of quality. As an experiment, ACOs make sense—at least 5% of Medicare beneficiaries (and likely more soon) are currently enrolled in ACOs. The ACO model is not a panacea. The models from the Congressional Budget Office and the CMS actuary estimate only modest financial results at best. Nor are ACOs the only hope; they are just one program—albeit a highly visible one—in a suite of new forms and environments that the nation will be testing in the next few years. The largest predecessor trial informing ACO design is the CMS Physician Group Practice Demonstration (PGPD). In this issue of JAMA, Colla and colleagues reanalyze the results of this project for the period 2005-2009. Prior PGPD analyses showed widespread gains in health care quality in the 10 demonstration sites but only inconsistent and generally small effects on costs. In the current article, the Dartmouth investigators add depth and texture to those findings. Using adjustments for case mix less vulnerable to coding games than Medicare’s hierarchical condition categories (HCC), and using a sophisticated “differences of differences” analytical model, they show a slowing of the rate of rise of costs across PGPD sites compared with controls averaging $114 per person per year, almost all attributable to more substantial savings averaging $532 per year for the 15% of beneficiaries who are dually eligible, that is, covered by both Medicare and Medicaid because of combined burdens of age, disability, and poverty. The study also offers clear evidence of coding games—lawful but games nonetheless—as PGPD sites tended to list more diagnoses at visits than control sites did, increasing the severity score for their patients, and thus making it easier to claim savings, because the HCC system adjusts expected costs upward as the number of diagnoses increases. What does this new analysis add to predictions about the promise of ACOs? First, the results for dually eligible beneficiaries are important and encouraging. Most of the 9.2 million people in that Medicare subgroup receive poor, uncoordinated care in the status quo, and they account for over $300 billion in annual costs and 40% of state Medicaid expenditures. Improvements of cost and quality for them can have big payoffs. Second, the Dartmouth group documented a small overall savings for the entire beneficiary population and, were this to be multiplied over the whole of Medicare, the total would be about $5 billion per year, that is, about 1% of the budget. Third, the substantial variation of results among PGPD sites offers hope for continual learning about best practices, and therefore, maybe, better re-