Skip to article content

The mission of the Harvey L. Neiman Health Policy Institute® is to establish foundational evidence for health policy and radiology practice that promotes the effective and efficient use of health care resources and improves patient care.

September 19, 2016

Radiology and the Shared Savings ACO

Radiology practices face a number of challenges when it comes to meeting the sweeping changes to Medicare reimbursements under  the Medicare Access and CHIP Reauthorization Act. Among these is determining whether they will participate under the Merit-Based Incentive Payment System (MIPS) or the alternative payment model (APM) provisions. While I believe the majority of radiology practices are best served in the near term participating through the MIPS framework, researchers at the Harvey L. Neiman Health Policy Institute have taken the time to think carefully about radiology’s role in APMs, having recently proposed a bundled payment model for breast cancer screening and developing the Inpatient Cost Evaluation Tool (ICE-T) to provide national benchmarks for inpatient bundles.

A new Neiman Institute paper, published in collaboration with industrial engineers from Virginia Tech, investigates how the incentives in another APM – the Accountable Care Organization Share Savings Program (ACO SSP) – might influence cost, quality, utilization, and technological investment. The full details of the model can be found in our paper – which can be a little intimidating for those shy of math.  However, the assumptions and implications of this model are relatively straightforward to explain.

The basic set-up:

  • Patients care about getting well and will follow physicians’ diagnostic and treatment recommendations.
  • All of the patients are covered by a public payer that wishes to maximize patient health at the lowest possible cost.
  • Hospitals, primary care physicians (PCPs), and radiologists care about both their patients’ health and the monetary benefits received from providing care.
  • We also assume that PCPs and radiologists like to work with the latest value-adding technology and hospitals have to decide if they are willing to purchase new technology – in this model assumed to be an improved computed tomography (CT) scanner.

On top of this basic framework, hospitals and physicians can choose to form and participate in an ACO SSP.  The hospital will manage the ACO organization and the two physician groups (i.e. PCPs and radiologists) can separately choose whether to participate.  Choices to form and/or participate in the ACO will depend on the answers to these questions:

  • What is the cost benchmark the ACO must not exceed to achieve savings?
  • How are the savings shared between the different provider groups?

Other decisions the various providers must face:

  • Will ACO participation influence whether the hospital purchases the new CT scanner?
  • How aggressively will providers order images?

Because this is a numerical model, we don’t have to answer these questions directly. Instead, we simply recalibrate the model hundreds of times with different parameter values to explore entire ranges of outcomes and look for the breakpoints where different decisions are made.  The answers to these questions will alter model solutions for patient health, image utilization, and overall cost.

Here’s what we learned:

Provider participation in ACOs occurs at very different cost benchmarks.

As seen from Dartmouth’s recent exit from the ACO program, if the cost benchmark is too aggressive hospitals simply can’t participate – especially low cost providers that substantially reduced waste prior to ACO participation.  Further, our model indicates that even when hospitals are willing to form an ACO, it requires an even lower threshold for accruing savings to induce PCP participation – and lower still for radiology practice participation – because of their limited share of any potential savings.

Physician provider shares vary according to the cost benchmarks set between payers and ACOs.

At very aggressive cost benchmarks, ACOs are willing to pass all of the savings to physicians to incentivize lower cost through less imaging utilization. At easy to achieve cost thresholds, hospitals retain all of the potential savings and simply engage with the other providers through fee-for-service.  The latter case is consistent with recent evidence on radiology practice participation¹.

The dynamic relationship between provider shares and the cost threshold have important implications on utilization and patient health.

At aggressive savings targets, PCPs ordered fewer images – often beneath the threshold required to maintain patient health. However, with easier to achieve benchmarks, physicians actually ordered more imaging under the ACO program as hospitals retained all of the savings – raising the dangerous possibility of excessive radiation exposure.

ACO participation reduces the likelihood hospitals will invest in new technology.

This doesn’t mean hospitals won’t invest at all. However, new technologies must produce more than marginal increases in clinical effectiveness to justify the expense and still meet hospitals’ savings targets.

The big takeaway:

I’m often asked whether I can provide a set of decision rules for when and under what conditions a radiology practice should participate in an ACO.  The short answer is, “No.”  The many moving parts indicate that there is no magic bullet for radiology practice participation and that the likeliest outcome of participation is either no savings accruing to radiologists, reduced imaging volumes, or both.  Of course, this doesn’t mean that no ACO participating radiologists have seen incentive payments, just that these will likely occur in short-lived ACOs as this is most likely to happen in ACOs with extremely aggressive savings thresholds in our model.

Of deep concern were the potentially health compromising decreases and increases in utilization we observed at relatively modest parameters away from their non-ACO status quo values. These indicate that the Centers for Medicare and Medicaid Services (CMS) must be careful to negotiate savings benchmarks with ACOs that are neither too stringent nor too lenient. Other solutions could incorporate the use of clinical decision support systems to ensure appropriate imaging utilization and protect the patient’s quality of care.

Although our study only considers the CMS’ ACO Shared Savings Program, it is likely many of these results are applicable to radiology practices under other APM arrangements such as the Next Generation ACO Model or the BPCI program.  Perhaps,  as others have said, the unintended consequences of these programs are related to the volume of APM programs available – causing providers to question these programs’ permanence and not really change their behaviors. However, it’s always possible, as we learn from ACO SSPs, that at least some of these consequences may be built into the program’s incentive structures themselves.


¹P. Kassing. P. Hashmian. “Still Stuck in the Fee-For-Service Model? Don’t Worry…You Are Not Alone!” RBMA Bulletin. July-August 2015.


Danny Hughes, PhD

Danny Hughes, PhD

Danny R. Hughes, PhD is Affiliate Senior Research Fellow, Harvey L. Neiman Health Policy Institute, and Professor in the School of Economics and Director of Georgia Tech’s Health Economics and Analytics Lab (HEAL). Follow him on Twitter @economeer.