Sometimes at Think Tanks, You Think

At its inception, Medicare was a fee-for-service health system. Seniors went to doctors (and hospitals, and…); doctors provided services; doctors submitted claims to Medicare for those services; and Medicare reimbursed based on a fee schedule for all services. As long as an accurate accounting of services was provided to Medicare, physicians and other providers received an accurate payment. The problem with this approach is that the more care doctors provided to beneficiaries, the more money they made. The incentives lead to very costly health care characterized by overuse and misuse of tests, therapies, and treatments.

Over time, managed care also found a place in Medicare – most recently as Medicare Advantage (MA) in the 2003 Medicare Modernization Act. In this approach, MA plans received a single, lump-sum payment for each beneficiary. In contrast to traditional fee-for-service Medicare, the incentive is to provide as little in tests, therapies, and treatments as possible, in order to minimize costs and maximize earnings. In its purest form, a capitated payment system’s incentives are to avoid sick, costly patients (to the extent possible) and to provide the bare minimum in care otherwise. It is a much cheaper system, but not high quality.

Among the solutions to the MA incentive problem is risk adjustment – that is, calculate the lump-sum payment needed for the beneficiary with the average health and, specifically, health care costs. Then, when a Medicare beneficiary is identifiably going to be more expensive (e.g., a diabetic), provide a larger lump sum. That way, MA plans will not try to avoid the more costly patient. (The reverse is also true; for a healthier beneficiary, provide a smaller payment.)

How, exactly, to do this is worth thinking about. Since AAF is a think tank, we did.

Specifically, The Center for Medicare and Medicaid Services (CMS) uses a statistical model to implement risk adjustment in MA. Perhaps surprising to most people, the data used to calibrate this model is not from MA; it is the claims data that is collected under traditional Medicare. Since there is no automatic reason that the characteristics of the beneficiary pools or the frequency and types of treatments employed by providers would be the same in traditional Medicare and MA, it raises the question as to whether this data source is the best. An alternative would be to use so-called encounter data (records of the treatments provided by MA plans to beneficiaries) to calibrate the MA risk adjustment model, and CMS has announced its intent to do just that.

There is currently no specific proposal from CMS to recalibrate the MA risk-adjustment model, but doing so would raise policy, technical, and operational issues. To think about these, this past October AAF and the Urban Institute convened a one-day summit with 21 experts to discuss implications of calibrating the risk adjustment model using encounter data. The result was a white paper and an event yesterday that presented the results of the summit.

As it turns out, it is far from a no-brainer to use MA data to calibrate the MA risk-adjustment model. Indeed, the discussion ultimately emphasized the importance of a deliberate, stakeholder-involved process that vets the quality of the encounter data, the tradeoffs between accurate reimbursement of MA costs and the incentives for both plans and beneficiaries that lead enrollees to choose either MA or traditional Medicare, and the fact that risk-adjustment techniques must be decided in coordination with other aspects of payment in the Medicare system.