1875 require treatment. CMS recognizes the importance of random variation and the need for actuarially stable populations in its Medicare Shared Savings Program, in which organizations cannot participate unless they serve at least 5000 beneficiaries. The bundled-payment initiative sets no such limits, and medium-sized hospitals typically have 100 to 200 cases for their highest-volume types of Medicare episodes. At these lower volumes, hospitals may have large year-to-year shifts in average spending for particular episode types owing to random variation in illness severity. For example, our simulations showed that 25% of hospitals with 100 to 125 annual admissions for congestive heart failure would incur financial losses of at least 11% simply because their patients in the program’s first year differ from those in the period used in setting their target prices.3 Another 25% of these hospitals would achieve financial gains of 6.1% or more. Such random variation in illness severity could overwhelm the effect of performance improvements in determining a hospital’s shortterm financial outcomes — making early wins and losses artifacts of variation rather than marks of true success or failure. The Bundled Payments for Care Improvement Initiative has great potential to engage hospitals in clinical redesign and care coordination that could improve both care and efficiency. Hospitals already have incentives through Medicare’s DRG payment system to make inpatient care more efficient, but few have invested in managing care after discharge. The bundled-payment initiative provides an opportunity for hospitals to gain experience with coordinating care across a continuum of services for discrete clinical conditions. Nonetheless, program managers must be vigilant to ensure that the financial incentives don’t cause stinting on care or avoidance of high-risk patients.4 The success of the initiative will depend on whether it protects participating hospitals against losses resulting from both random and systematic variation in illness severity. Certain design features will make it much more attractive to hospitals, including risk adjustment, stop-loss protection for high-cost cases, an ability to exclude cases with highcost primary diagnoses from episode definitions, and so-called risk corridors that allow hospitals to share both gains and losses as they acclimate to the program. CMS has begun to discuss changes to the proposed financial model with applicants. If hospitals are confident that the program will financially reward successful clinical performance, many more will be willing to pursue the opportunities for care improvement that this program seeks to encourage.