In the two current Medicare bundled payment programs, three different methodologies have been used to establish payment targets. In the BPCI program targets are based on each participant’s historical episode cost, with “cost” defined as all payments made by Medicare or beneficiaries for services provided to the beneficiary during the episode. In this program participants will achieve financial success by reducing episode costs below their historical levels. In the CJR program target in the later performance periods are computed based on the episode costs of other hospitals in the respective region of each hospital. In this program participants achieve financial success by having costs that are lower than those of their neighbors. Finally, in the Oncology Care Model (OCM) program targets are computed from a complex regression equation based on many patient characteristics, and the financial results are a consequence of patient mix and services, with no clear path to success.
When BPCI Advanced was announced as a voluntary program, we expected the targets to be based on historical cost since no rational provider would voluntarily participate in a program in which its own cost already exceeded the targets, which would occur if regional targets were used, and in the initial materials released by CMS it appeared that this method would be used. However, as additional materials are released by CMS this conclusion becomes less certain. In the CMS webinar reviewing the model, CMS described the target setting process as follows: To determine the hospital's benchmark price, CMS will use risk-adjustment models to account for the following contributors to variation in the episode standard spending amounts. One, patient case mix. Two, patterns of spending relative to the hospital's peer group over time. And three, historic Medicare fee-for-service expenditures efficiency in resource use specific to the hospital's baseline period.
Unfortunately most of these terms are not defined in these or previous documents. “Episode standard spending amounts” has a particular meaning in the CJR program, but nothing that relates to the basis for targets. Likewise, no information is available about the “risk-adjustment model” or how it will incorporate patient mix, costs of the peer group (or any definition of that group), and how “expenditures efficiency” will be measured from the hospital’s baseline period. CMS has indicated that the target methodology will be released in the next few weeks, which does little to alleviate our concern at this point.
Why Risk Adjustment Matters
Experts on our team differ in their expectations of how the “risk adjustment” will work. Some believe that the primary factor in target-setting will be the hospital’s baseline cost, with the risk adjustment making minor changes to that amount. That model would be somewhat comforting in its familiarity and the knowledge that improvement on past performance would generally result in financial success. Others are less optimistic and believe that the primary driver of targets will be the regression equations that CMS has indicated will be part of the target-setting process. In this case the path to success is less clear. In the OCM program there’s little correlation between individual episode costs and targets, and some cancer types are invariably financially successful while others almost always sustain financial losses. Thus, if the regression models are the primary drivers of targets we’ll be running detailed analyses to identify areas in which the regression factors (and not episode costs) are the primary determinants of success or failure; in other words, whether there are biases towards certain types of patients, participants and episode characteristics. Such biases would significantly alter the episode selection process, moving from favoring episodes in which clinical change opportunities exist to those episodes that are “built-in winners” by virtue of the risk adjustment process.
These details will be critical to establishing confidence in the BPCI Advanced program, which will be necessary to induce participation. In the BPCI and CJR programs the path to financial success were relatively straightforward and determinable from factors that were visible and easily understandable. In the OCM program the target development process is significantly more cloudy, and participants who believe that their cost reductions will lead to financial success may find that their cost targets have decreased due to case mix changes and that the expected success will not materialize. Participants need to understand the direct relationship between their actions and success, which can be obfuscated by the intervention of complex risk adjustment processes.