The principles of value-based payment models in health care are elegant, intuitive, and appealing: pay clinicians for delivering high-quality care. In practice, however, we have not yet agreed on many of the important details on either cost or quality. The goal of measuring true quality remains elusive, with important unresolved issues of conceptualizing, operationalizing, and implementing quality measurement. In addition, technical and philosophical challenges remain on determining how to appropriately pay clinicians. Risk adjustment of payments and penalties raises the fundamental question of how to determine the right amount to pay for the highly varied patients that each clinician sees, and it has a profound impact on how clinicians function under value-based models. Risk adjustment can influence how organizations develop clinician networks, invest in service lines, plan locations, and treat patients. Under value-based payment models, avoidance of treating high-risk populations may be an appealing option for physician organizations, hospitals, or payers concerned that they will need to expend more resources for certain patients than they will receive to care for them. This phenomenon is known by many names, including adverse selection, cherry picking, cream skimming, and patient dumping, and has been found in a variety of contexts related to quality reporting or pay for performance. Adverse selection is a serious threat to successful value-based payment. Poorly executed risk adjustment is perhaps the biggest potential harm to high-risk patients, who may experience decreased access to high-quality clinicians as a result.