U.S. Department of Health and Human Services

Risk Selection Among SSI Enrollees in TennCare

Executive Summary

Steven C. Hill, Christopher Trenholm, Craig Thornton and Judith Wooldridge

Mathematica Policy Research, Inc.

February 1999

This report was prepared under contract #500-94-0047 between the U.S. Department of Health and Human Services (HHS), Office of Disability, Aging and Long-Term Care Policy (DALTCP) and and Mathematica Policy Research, Inc. Additional funding were provided by the HHS Health Care Financing Administration. For additional information about this subject, you can visit the DALTCP home page at http://aspe.hhs.gov/_/office_specific/daltcp.cfm or contact the office at HHS/ASPE/DALTCP, Room 424E, H.H. Humphrey Building, 200 Independence Avenue, S.W., Washington, D.C. 20201. The e-mail address is: webmaster.DALTCP@hhs.gov. The Project Officer was Hunter McKay.

The opinions and views expressed in this report are those of the authors. They do not necessarily reflect the views of the Department of Health and Human Services, the contractor or any other funding organization.

Risk selection greatly complicates the administration of Medicaid managed care. It occurs when the health care needs of beneficiaries enrolled in a specific plan differ systematically from the needs of the overall beneficiary population. When risk selection occurs, state administrators should adjust payments to the managed care plans to ensure that each plan's payment accurately reflects the needs of its enrollees. Without such a payment system, problems for plans and beneficiaries are likely to arise. Plans with adverse selection (that is, a disproportionately large number of high-need beneficiaries) are likely to lack the resources required to deliver adequate care to their enrollees. Such plans face pressures to cut back on needed care and may ultimately drop out of the state's Medicaid managed care system. Plans with favorable selection (a disproportionately large number of low-need beneficiaries) will be paid more than is necessary to provide care.

The issue of risk selection is particularly important for states that enroll blind and disabled Supplemental Security Income (SSI) beneficiaries in managed care. SSI beneficiaries often need atypical and complex services. Many will require ongoing management by specialists, especially those beneficiaries with relatively rare conditions or with mental conditions. Furthermore, many SSI beneficiaries require ongoing social support services to address chronic limitations in functioning, while others have conditions that make communicating with providers difficult. The seriousness of these limitations and the underlying medical conditions mean that individuals can experience severe declines in their health and independence if proper services are not delivered. Managed care plans, which often began by serving healthier, employed populations, will be challenged to arrange for this needed care even if capitation payments accurately reflect the underlying health care needs of their members. In the face of adverse selection, those challenges can be magnified substantially.

We examine risk selection among SSI beneficiaries in Tennessee's Medicaid managed care program, TennCare. We focus on blind and disabled SSI beneficiaries younger than age 65, who are not simultaneously in Medicare and who enrolled in TennCare at its start in January 1994. Because beneficiaries could change plans in subsequent years (and thereby mitigate or exacerbate the initial risk selection), we track risk selection through TennCare's first two years. We used 1993 fee-for-service claims data and TennCare plan enrollment data for our analysis.

The level of risk among blind and disabled SSI beneficiaries is measured using information about pre-TennCare expenditure and use patterns. Our basic risk measure was each SSI beneficiary's 1993 expenditures under fee-for-service Medicaid. Previous fee-for-service expenditures have been used in numerous studies of risk selection, and expenditures are highly correlated across years for disabled SSI beneficiaries. We use these expenditure levels to estimate a "risk index" for each managed care plan that equals the average 1993 Medicaid expenditures for a plan's SSI enrollees, divided by the average 1993 expenditures for all SSI beneficiaries in TennCare. For a given plan, the higher the risk index is above a value of 1, the stronger the evidence that the plan has experienced adverse selection.

There was substantial risk selection among the managed care plans serving SSI beneficiaries in TennCare. In particular, the initial TennCare enrollment process resulted in substantial adverse selection for plans affiliated with medical schools and for the largest plan, which did not require primary care gatekeepers. The estimated risk indexes of the SSI beneficiaries enrolled in these plans were 12 to 40 percent above the average for all SSI beneficiaries. At the same time, a few plans experienced favorable selection among this population. In those plans, we estimate that the enrolled SSI beneficiaries had health care needs as much as 50 percent below the average for all SSI beneficiaries. The pattern of risk selection remained largely unchanged throughout the first three annual opportunities to change plans, and it is consistent across several alternative risk indexes.

Risk selection may have consequences for plans and beneficiaries. Some of the plans with adverse selection among the SSI beneficiaries also had financial problems, and one plan that experienced favorable selection reported substantial profits in the early years of TennCare. However, these apparent financial impacts on plans should be interpreted cautiously, because SSI beneficiaries are only 13 percent of the TennCare population, and many other factors, including the actuarial soundness of the rates overall, affect profitability. The fee-for-service claims data used for this paper could not be employed to assess actual profits from SSI beneficiaries or any impacts on access and quality of care.

Risk selection is difficult to prevent or reduce. States will find it difficult to reduce risk selection because it is hard to address the three factors causing it: beneficiary choice; differences among plans; and, in Tennessee, the assignment process for people who did not select a plan. Risk selection cannot easily be reduced by eliminating choice, because choice in all areas of life is important to people with disabilities, and federal government regulations require that beneficiaries have choices among plans. To some extent, states can control differences among plans. However, it seems likely that differences in plans' networks, reputations, provider payment systems, and other factors will persist, thus leading consumers to choose, and their providers to recommend, specific plans. The assignment process would be the easiest factor for states to address because it is under the state's direct control. Nevertheless, states may want to use the assignment feature most responsible for risk selection, assigning beneficiaries to plans whose networks contain the providers who have served those beneficiaries, because that feature also promotes continuity of care.

States have only limited options for addressing the consequences of risk selection. States can implement payment systems that match plan compensation to the expected needs of plan enrollees. Improving the match between payments and needs will likely reduce plans' incentives to market to healthier consumers, enhance the financial stability of plans that attract members with greater needs, and reduce the extent to which states pay plans with healthier members more than is necessary to provide care. States have two basic options:

• Partial Capitation. By providing at least some compensation based on capitation, this method gives plans an incentive to provide cost-effective treatment. At the same time, providing some compensation based on costs or charges can help ensure that a plan has enough resources to arrange for the care required by high-need individuals.

• Risk Adjusted Capitation. This method creates a set of group-specific capitation rates that reflect the expected costs of care for defined groups. It creates a stronger incentive to provide cost-effective treatment. The groups must be defined by diagnosis, to effectively predict expected costs of people with disabilities.

States have implemented each of these options, and, to date, there is little information on which specific method is most effective in reducing risk selection or the effects of risk selection on plans or SSI beneficiaries.

Implementing risk-adjusted capitation or partial capitation requires good data. When managed care is initially implemented, states can use the diagnostic information included in claims data from fee-for-service Medicaid to assign beneficiaries to rate cells for risk adjustment. But, as the managed care program matures over time, states need more recent data, including diagnostic information for people newly enrolled and for those with new or worsening health conditions. To implement reinsurance or risk corridors, states need reliable data on expenditures for care from each plan. Encounter data from managed care plans is the obvious source for diagnostic and expenditure information, but states have had difficulty acquiring accurate encounter data from the plans in the early years of managed care, and states have required substantial resources and time to implement full review systems and provide the plans with the feedback necessary to improve plan data accuracy. However, the benefits of these investments may be great, because the data are useful not only for rate setting, but also for numerous other purposes, including monitoring access and quality.

Risk selection will happen, states have limited options for reducing the consequences for plans and beneficiaries, and those efforts require encounter data. We found considerable risk selection among the SSI population in TennCare, and it is fairly certain to occur in other states, too. If states do nothing to adjust for risk selection in their payments to plans, some plans with adverse selection may eventually close and beneficiaries may be affected. States can implement payment systems to reduce the impacts of risk selection, and these payment systems require states and plans to invest substantial resources in their encounter data systems.