Bundled payments refers to a reimbursement method which gives some provider one payment for an entire episode of care. To be workable, the provider has to have some ability to control care across the entire continuum and the “episode” has to be discrete and definable. The Medicare program is testing bundled payments and the Lewin Group released an evaluation of the initial stages of the program. (BP Eval.) All the episodes are triggered by a hospitalization and fall into one of 48 diagnostic related groups. Four models are being tested, three of which are reported on in this evaluation. In one model, a physician or hospital is responsible for managing the episode, which can include 30, 60 or 90 days related services after the hospitalization. The various providers continue to receive fee-for-service payments, but all the costs are reconciled against a target with the managing provider receiving savings are having to repay excess costs. In another model, either a physician group or a post-acute care provider is responsible for the bundle of services, which begins within 30 days of the hospitalization and includes all the post-hospital costs. Again, a reconciliation method against a target is used. For the third model, hospitals manage the services and cover the hospitalization and all services for 30 days after it ends, unless the services are explicitly excluded. The hospital gets paid prospectively and must pay the other providers.
The researchers constructed comparison episodes from Medicare fee-for-service data to help evaluate the effects of the bundled payments. There were nine participating bundled payment schemes under the first model, and in general they were larger hospitals in more competitive markets. Participants were interviewed about how they created partnerships with the other providers offering services in the bundle and why they were accepting bundled payments. Knee and hip replacements were the most common episode, with congestive heart failure, pneumonia, and COPD also accounting for a significant number of episodes. Length of stay tended to decline slightly in the hospitalization starting the episode and to be slightly lower than for non-bundled payment episodes. A substantial decrease in the number of discharges to an institutional post-acute care setting occurred in the bundled episodes, with no change in that number for the comparison episodes. There do not appear to be cost savings associated with the bundled payment method, and quality, by measures like readmission rates, may have been slightly worse. In the second bundle model, the one for post-acute care, the bundled payment episodes showed very significant reductions both in utilization and in cost, primarily due to less use of home health care. Quality as measured by readmission rates and ER use, did not appear to be different from that in the comparison episodes. In the last model, there was only one participant, and for just one type of episode, so there was not enough information for analysis. Overall, the results would have to be called mixed, with a suggestion that for a fully-bundled episode there may not be savings, but for a post-acute care one, there could be substantial benefits. It is early in the program, and the next evaluations should be more enlightening.