Risk-adjusted Capitation Primer

Managed care organizations (MCOs) are paid by Medicare, Medicaid and private companies to provide insurance functions (e.g., claims payment) and preventive healthcare interventions designed to improve patient health outcomes. MCOs, in turn, pay providers for the direct delivery of healthcare to members enrolled in their plans. Typically, MCOs receive per-member-per-month prospective payments for the health care management of their enrolled members, called capitation. To make a profit, a MCO must manage the health services of their population in ways that keep the actual healthcare costs below the capitation payment, on average.

While promoting cost-efficiency, capitation payment encourages MCOs to attract members whose likely costs are below the capitation rate. MCOs can behave in ways that lower the likelihood of attracting the unprofitable members - through benefit design, provider network composition, selective marketing, or other methods. These prohibited activities are called risk selection (“cherry picking”) and discriminatory disenrollment (“dumping”).

To mitigate cherry picking and dumping, payors have begun to adjust the payments made to MCOs based on patient diagnosis, so that more is paid to MCOs for the enrollment of sicker individuals than for healthier ones. Conventionally, these capitated payments can be “risk-adjusted” to account for certain factors that predict future healthcare costs, such as demographics (age, gender, geography), diagnoses, and pharmaceutical utilization.

With risk-adjusted capitation payments, more is paid to MCOs for the enrollment of sicker individuals than for healthier ones. The more accurately the risk adjustment payment model predicts the future healthcare costs of a member, the less incentive the MCO has to cherry pick or dump members. Sicker members mean higher capitation payments for the MCO.

Predictive power of risk adjustment methodologies continues to improve, but comparisons in the literature investigating Medicaid recipient populations show that the predictive power (R²) of the risk adjustment models ranges from 0.11 to 0.18 (Kronick, Gilmer, Dreyfus, & L. Lee, 2000) and from 0.15 to 0.23 (Gilmer, Kronick, Fishman, & Ganiats, 2001). While risk adjustment does not explain much of the variability in future healthcare expenditures, it is still sufficient for many government payors, such as Medicare, to pay MCOs for managing beneficiaries’ care. (It should be noted that perfect prediction of future health expenditures is not the goal. If future healthcare utilization was known, then insurance would not be necessary.)

Notwithstanding improvements risk adjustment accuracy, about 80% of the future health care costs are unexplained using risk adjustment. This remaining risk leads MCOs to engage in risk selection behaviors. While risk adjustment is a significant improvement over average cost capitation, the incentive to attract healthy enrollees and avoid the sick ones still exists. Are risk-adjusted payments better than nothing, though?