As ACOs and bundled payment programs get underway, we’ve had some inquiries from clients participating in one or both programs wondering how they interact financially. Specifically, the questions center around a patient who may be an attributed member of an ACO, yet who is admitted in a participating DRG in a hospital participating in the Medicare Bundled Payment for Care Improvement (BPCI) program. These interactions are somewhat counterintuitive and have some pretty strange financial effects, and therefore provide some great blog fodder for your humble analytics writer.
By themselves, the payment methods for each of these programs are pretty straightforward. Under BPCI, the participating provider (which is usually a hospital) agrees to give up a discount (2% or 3%) off of the Medicare payments that all providers who provide services to the patient from admission to 30-90 days post-discharge will receive. (The actual target-setting process is far more complex, but that’s the general gist of it.) So if the historical Medicare payments to the hospitals, physicians and post-acute providers for joint replacements over that period is $25,000, the participating hospital will forego $500 of revenue and accept a payment target of $24,500. If the respective Medicare payments to the providers turns out to be lower than that amount, the hospital keeps all of the difference. Similarly, if Medicare payments are greater than the target, the hospital must repay Medicare for the difference. Note that there is no shared savings component – Medicare gets the discount regardless of the amount of savings achieved, if any, and the hospital retains all of the savings. The discount ends up being a “sunk cost” to the hospital.
ACOs, however, operate under what’s essentially a capitated "shared savings" model. In an ACO a target expenditure amount per member per month is established, and if Medicare payments to providers are lower than the target approximately half of the savings goes to the ACO, with the remainder retained by CMS. (Note the part about “Medicare payments to providers” – we’ll return to it later.)
The financial incentives provided by these arrangements are quite different, as we've previously written. However, the interactions are even more interesting.
Let's start with the ACO, whose financial objective is to achieve shared savings by reducing utilization by Medicare providers to its attributed members somewhere across the care continuum. Since most of the care dollars are in hospital inpatient services, and readmissions are a frequent target for savings, the ACO may adopt the strategy of providing comprehensive care management services to patients with congestive heart failure in the hope of reducing their readmission rate. (And let's assume that this is the physician-owned ACO, so it is unaffected by the reduction in hospital revenue that results from elimination of these readmissions.) So the ACO sets its care management processes in place, watches the number of readmissions decrease, and starts counting up the shared savings that will result.
What the ACO didn't know, however, is that the hospital is participating in the BPCI program for CHF DRGs. Because Medicare essentially pays the hospital a bundled payment for these DRGs, its payment will not decrease as the readmissions decrease. And if Medicare payments don’t decrease, no savings results, which means there's no shared savings for the ACO. So what happens to the “savings” that are created by the reduction in readmissions? It all goes to the hospital. Since the Medicare payments for those episodes will be below the payment target, the bundled payment target will exceed the actual Medicare payments, and at reconciliation time CMS will write the hospital a check for the net amount of all the savings. Hopefully the hospital will thank the ACO physicians for all their efforts at readmission reduction, and will invite them to a nice cocktail buffet at which jumbo shrimp are served.
Does this mean that the ACO is totally out of luck with these patients? Not entirely – remember that there is a "reduction in Medicare payments" resulting from the BPCI program in the form of the discount given by the hospital as the price of participation in the program. According to CMS, that discount creates "savings", from which the ACO can derive some shared savings. So in the example given above, the "savings" to Medicare is $500 per episode, approximately $250 of which would go to the ACO. And the ACO would receive this payment for each admission regardless of whether the BPCI hospital achieved any savings, since the discount applies to every BPCI episode.
Of course, this totally changes the ACO’s financial incentives to reduce those readmissions in the first place. Instead of investing in care coordination processes in the hope of achieving a shared savings return that would never happen, they can simply refer their CHF patients to the hospital and receive a small "shared savings" payment as a portion of the hospital discount. It won’t be very much money, but it's certainly larger than the amount of shared savings that they would receive from the readmission reductions, which would be zero. And they can receive this payment without any financial or time investment.
A previous version of this article indicated that the portion of the discount given by CMS to the ACO would be recouped from the hospital. Further guidance from CMS indicates that this recoupment would only occur if the ACO was owned by the BPCI participant, not by an independent physician group.
So now the ACO can create guaranteed shared savings revenue without having to do any work, but loses its opportunity to create shared savings through all of the things that ACOs are supposed to do to create savings – at least for this group of patients.
And there ends the story of the strange interaction between ACOs and bundled payments, showing the unpredictable results of combining two different payment systems having significantly different structures and incentives. Perhaps this is one of those cases in which unpredictability is the only thing that can be predicted.