Risk Adjustment and the
Committee on Commerce,
Subcommittee on Health and the Environment
U.S. House of Representatives
Gail R. Wilensky, Ph.D.
Medicare Payment Advisory Commission
February 25, 1999
Good morning Chairman Bilirakis and members of the Subcommittee. I am Gail Wilensky, Chair of the Medicare Payment Advisory Commission (MedPAC). I am pleased to be here this morning to discuss the issue of risk adjustment and the Medicare+Choice program.
The system used to adjust payments to Medicare’s risk-contracting plans and now Medicare+Choice plans has been widely acknowledged to be inadequate because it does not accurately reflect predictable differences in enrollees’ health spending. As a result, Medicare has overpaid plans to care for relatively healthy enrollees and underpaid plans to care for those in poorer health. Overall, payments have exceeded plans’ costs of providing the basic Medicare benefit package.
A better risk adjustment system would improve payment equity across plans and reduce Medicare’s overpayments to plans. The interim risk adjustment system proposed by the Health Care Financing Administration (HCFA)—which relies on principal diagnoses from inpatient hospital stays—is imperfect, but it represents a step in the right direction by making payments correspond more closely to enrollees’ health needs. Moreover, many of the limitations of the proposed interim system could be mitigated by moving to a system based on diagnosis data from all sites of care. MedPAC supports HCFA’s efforts to do this effective for payments in 2004.
Adopting any new system of risk adjustment would introduce swings in payments
to plans. Accordingly, MedPAC supports the phase-in proposed by HCFA that back-loads
RISK ADJUSTMENT AND WHY IT IS NEEDED
Risk adjustment is a term used to describe incorporating predictable differences in health status and service needs into the capitation payments made to health plans. When payments are risk adjusted, plans receive larger payments for their relatively sick enrollees and smaller payments for their healthier ones.
In Medicare, risk adjustment is intended both to make payments equitable across Medicare+Choice plans and to account for differences in the mix of enrollees between the traditional fee-for-service program and the Medicare+Choice program. Put another way, risk adjustment may be viewed as a means of encouraging health plans to serve beneficiaries with severe or chronic illnesses by paying plans more to care for them.
Medicare beneficiaries’ needs for health services—inpatient care, physician visits, and so on—vary, and this variation has both a random component and a systematic component. The random component reflects service needs that are, by definition, unpredictable, so that if there were no other differences among beneficiaries, risk adjustment would not be necessary. In such a situation, unexpectedly high costs for some enrollees in a plan would be offset by unexpectedly low costs for other enrollees. Given sufficient numbers of enrollees, payments to plans would be correct on average.
In fact, there are differences among beneficiaries that lead to systematic and predictable differences in their needs for health services. For example, older people use more services than younger people, and people with severe or chronic illnesses use more services than others. These predictable differences in the use of services—whether they are predictable either by health plans or by enrollees themselves—introduce the potential for risk selection. If no adjustments are made to account for these differences, plans will be overpaid for healthy enrollees and underpaid for sick enrollees. Accordingly, they will have an incentive to enroll beneficiaries whose expected costs are below average, because they will still receive the average payment. If plans act on this incentive and successfully attract relatively health beneficiaries, aggregate payments will be too high.
MEDICARE’S CURRENT SYSTEM OF RISK ADJUSTMENT
Currently, Medicare adjusts payments to private health plans to reflect only differences among enrollees in their demographic characteristics (age and sex), employment status, institutional status, and eligibility for Medicaid. This risk adjustment system accounts for the relatively greater use of health services of older beneficiaries and those who are institutionalized, and the relatively lower expected costs associated with working enrollees who have primary coverage through their employers, but it does not account for variation due to differences in health status. Until 1998, the base payment method paid 95 percent of expected fee-for-service spending for beneficiaries with similar characteristics, which was intended to account for health plans’ ability to deliver care more efficiently. Now, payments are based on updated 1997 rates.
Payment inequity and overpayment under the current system
A common complaint about the current system is that plans have experienced significant favorable risk selection—enrollment of relatively healthy beneficiaries—that is not reflected in their payments. Because it does not take health status into account, the current system rewards organizations that attract healthier enrollees because it does a very poor job of accounting for predictable differences in health spending. Plans are thus paid the same amount for two beneficiaries with identical demographic characteristics, even though differences in their health status would suggest that one will be much more costly than the other.
Empirical research supports the assertion that plans have experienced favorable selection while their payments have been based on average risks within demographic groups. For example, Riley and colleagues (1996) found that in 1994 the predicted costs of Medicare risk plan enrollees were 12 percent lower, on average, than the predicted costs of fee-for-service enrollees with the same demographic characteristics. Because payments currently are adjusted only for demographic differences, even setting rates at 95 percent of the amount Medicare expected to spend for a beneficiary in the fee-for-service program resulted in overpayments of as much as 7 percent (Riley et al. 1996, Hill et al. 1992). Those overpayments are in part why Medicare risk plans have been able to offer expanded coverage to enrollees.
Some favorable risk selection may be inevitable because the methods organizations use to recruit enrollees might not reach people with poor health status, such as the institutionalized, or because healthy people may be less particular about being able to see a specific physician. Moreover, even if selection to plans has been favorable in the aggregate, that does not mean that all individual plans have experienced favorable selection. For example, one study shows that mortality and hospitalization rates rise as length of managed care enrollment increases (PPRC 1996). This “regression towards the mean” means that in terms of their use of health services, managed care enrollees become more like fee-for-service beneficiaries over time. Plans that have participated in Medicare the longest and have long-tenured enrollees may thus experience less favorable selection.
Risk adjustment requirements in the Balanced Budget Act
In response to concerns about the current system, the Balanced Budget Act of 1997 (BBA) directed HCFA to develop a new risk adjustment system. The rationale of the Congress for mandating the new system was to make Medicare’s payments to Medicare+Choice organizations more accurately reflect predictable differences in health spending by enrollees. This new system should improve Medicare+Choice by making payments more equitable across plans and making them reflect the generally better health of Medicare+Choice enrollees as compared with fee-for-service beneficiaries.
The BBA required the new risk adjustment system to use enrollees’ health status and demographic characteristics to account for variations in their expected spending. It laid out a very tight time schedule, requiring HCFA to implement the system by January 1, 2000. To meet that schedule, the agency must:
publish a preliminary notice by January 15, 1999, describing the changes in methods and assumptions it will use to determine payment rates for 2000, compared with those for 1999 (HCFA 1999);
publish a final notice by March 1, 1999, on the payment rates for 2000 and the risk and other factors it will use to adjust those payment rates; and
submit a report to the Congress that describes the risk adjustment method it will implement with the new payment rates, also by March 1, 1999.
While HCFA has supported research to develop improved risk adjustment methods for more than a decade, implementing the new system has required HCFA to collect and analyze a substantial amount of new data in a short period of time. The agency must measure not only the health status of beneficiaries enrolled in Medicare+Choice plans, but health status and subsequent spending for beneficiaries in the traditional fee-for-service program.
HCFA must collect data from Medicare+Choice plans both to determine the monthly payments to organizations for each enrollee starting in 2000 and to inform Medicare+Choice organizations about the anticipated effects of the new risk adjustment system.
HCFA must measure health status and spending for fee-for-service beneficiaries for two reasons. First, the agency must estimate risk scores that measure relative levels of expected spending for beneficiaries with different combinations of health conditions and demographic characteristics. These scores require beneficiary-specific data on health conditions, demographic characteristics, and annual Medicare spending for covered services that are currently available only for beneficiaries in the traditional fee-for-service program. Second, once the new risk scores are developed, HCFA must adjust the per capita monthly payment rate for each county—the county rate book—to reflect the county’s expected level of per capita spending for a beneficiary with national average health and demographic characteristics.
To facilitate these tasks, the BBA permitted HCFA to collect encounter data—which provide information similar to claims data—on hospital inpatient stays from Medicare+Choice organizations, but not before January 1, 1998. Starting July 1, 1998, HCFA could collect encounter data from other providers of care such as physician offices, hospital outpatient departments, skilled nursing facilities, and home health agencies. HCFA will be able to use the diagnoses reported in the encounter data to develop indicators of beneficiary health status.
HCFA has indicated it has been meeting the time requirements of the BBA and has collected almost complete hospital inpatient encounter data records from nearly all organizations. A small number of organizations have supplied incomplete data, and HCFA is working with them to get complete data. Some organizations are less confident and believe the data generally are not complete due to systems problems. However, the actual risk scores will be based on the next round of data collection, which should afford an opportunity to work out existing problems.
HCFA’S PROPOSED INTERIM SYSTEM
The schedule outlined in the BBA restricted HCFA to adopt, at least initially, an interim system in which health status will be measured using only hospital inpatient diagnoses. Before the Congress passed the BBA, HCFA argued that it needed data as soon as possible to implement an improved risk adjustment system. However, HCFA and the Congress recognized that Medicare+Choice organizations could not establish systems for reporting data from sites of care other than hospital inpatient departments in time for implementation by January 1, 2000. Therefore, HCFA indicated to the Congress it needed inpatient data by a particular date and left the Congress to determine the remaining time frame.
Description of the proposed interim system
In the interim system, HCFA will determine payments to Medicare+Choice organizations according to the following process. First, HCFA will characterize beneficiaries based on:
their age and sex;
principal diagnoses associated with any inpatient hospital stays they had during the previous year;
whether they were eligible for Medicaid benefits at any time in the previous year; and
for aged beneficiaries, whether they were previously eligible for Medicare on the basis of a disability.
Based on this classification, HCFA will determine prospective risk scores for Medicare+Choice enrollees (see the Appendix for more detail). Risk scores are intended to measure enrollees’ expected spending in the forthcoming payment year relative to that of the average beneficiary in the traditional fee-for-service program. As in the current risk adjustment system, spending patterns in the traditional fee-for-service program will be treated as a baseline, so the risk score associated with each combination of demographic and health status factors will be estimated using fee-for-service data.
In the last step, HCFA will calculate payments for enrollees as the product of three factors:
the payment amount for 2000 for enrollees’ county of residence from the county rate book;
a factor that will adjust the county payment rate to reflect the change in risk measurement methods; and
the enrollees’ risk scores based on the interim system.
The county adjustment factors are needed to change the county payment amounts so they are consistent with the new system. Under the current system, each county payment rate is based on the updated 1997 payment rate, which reflects the current expected fee-for-service spending per capita in the county for a beneficiary with the national average demographic profile. Because the new risk adjustment system captures risk differences among beneficiaries more precisely than does the current system, HCFA needs to recalibrate the county amounts using the new adjusters. This method will ensure that the county payment rates reflect the 1997 expected fee-for-service spending per capita in the county for a national average beneficiary, as measured by the new system.
The interim system intended to improve payment equity
The interim risk adjustment system should be an improvement over the current system because payments to organizations will more accurately reflect the predictable differences in health spending by their enrollees. If it works as intended, the system will encourage organizations to compete on the basis of how effectively they manage care and not reward plans for attracting favorable risks.
The interim system is consistent with the BBA’s objectives for risk adjustment. First, it will encourage organizations to compete on factors other than risk selection because the profits from favorable selection will be lower. Second, organizations may have more resources for developing specialized care management programs for enrollees with serious conditions, which may lead to improvements in efficiency and in the quality of care enrollees receive. Finally, aggregate overpayments to Medicare+Choice organizations that result from enrolling healthier Medicare beneficiaries may be reduced.
Potential concerns with the interim system
Despite these improvements over the current system, the interim system’s dependence on hospital inpatient diagnoses raises several potential concerns that policymakers should monitor closely.
Incentives to hospitalize inappropriately. Because organizations will receive higher payments only for enrollees who have been hospitalized, the proposed system may create incentives for Medicare+Choice organizations to hospitalize enrollees inappropriately. However, the impact of such incentives is likely to be mitigated by a number of factors.
First, payments for enrollees’ hospital stays are based on their expected spending in the year following the stay, so the incremental payment may be lower in many cases than the hospitalization cost the organization incurred.
Second, organizations will not receive an increased payment until the calendar year after a hospitalization, and then only if the hospitalized beneficiary remains enrolled in the same organization.
Finally, organizations would have to influence physicians to hospitalize more patients and to overcome resistance to being hospitalized on the part of enrollees.
To further counteract any incentive to hospitalize, HCFA will treat enrollees with one-day inpatient stays and those with diagnoses for which hospitalization is discretionary the same as enrollees who were not hospitalized. HCFA considers a hospitalization to be discretionary if the principal diagnosis represents only a minor or transitory disease or disorder, is rarely the main cause of an inpatient stay, or is vague or ambiguous.
Adjustments based on fee-for-service patterns. A second potential problem is that risk scores based on fee-for-service hospitalization patterns may understate the riskiness of certain Medicare+Choice enrollees. This understatement will occur if Medicare+Choice organizations substitute other sites of care in place of hospitalizations more frequently than do providers in traditional fee-for-service Medicare. If this were true, Medicare+Choice enrollees with serious conditions would be hospitalized less often and would receive lower risk scores, on average, than fee-for-service beneficiaries with comparable conditions and demographic characteristics.
How serious this problem could be is unclear. Hill and colleagues (1992) found
that Medicare managed care organizations did not reduce the hospitalization
rate relative to fee-for-service Medicare. But Medicare+Choice organizations
have also argued that they hospitalize comparable patients for shorter stays
than do fee-for-service providers in traditional Medicare, and results from
Hill and others support this argument. To the extent organizations shorten hospital
stays to one day, HCFA’s proposal to treat enrollees with one-day stays
the same as enrollees without inpatient stays will compound any understatement
caused by calibrating risk scores based on fee-for-service data.
Potential for large changes in payments. A third issue is that implementing any improvements in risk adjustment will often lead to changes in payments to some individual plans that are much larger than the change in aggregate Medicare+Choice payments. Under the interim system, these changes could affect some Medicare+Choice organizations’ decisions to participate in some or all of the market areas they serve for Medicare and disrupt Medicare+Choice coverage for some beneficiaries.
Medicare+Choice organizations are understandably concerned about the effects of HCFA’s new risk adjustment system on their future payments. Other things being equal, adoption of this new system on January 1, 2000, will change payments for individual organizations and reduce overall Medicare+Choice payments. However, the full effects of the new system are somewhat uncertain because the data that HCFA will use to determine payments to organizations in 2000 will not be available until late in 1999 when enrollment data are available.
MedPAC has not yet made a comprehensive assessment of the impact of the new system on specific plans. But the amounts involved will be significant. Based on preliminary data, HCFA estimates that if the new system were implemented immediately and if there were no changes in the composition of enrollment:
variation in payments for individuals would range by a factor of about 25, compared with the current variation of about 6;
additional payments would be made for about 12 percent of enrollees, and about 20 percent of total payments would be redistributed;
aggregate plan payments would fall by 7.6 percent; and
payments to some plans could fall by 15 percent, whereas payments to others could increase by 5 percent.
Inpatient data inadequate. Finally, some analysts have expressed concerns that payments to Medicare+Choice organizations under the interim system will not fully account for measurable and predictable differences in spending among their enrollees because there is diagnosis and health status information that is not reflected in the demographic and hospital diagnosis data used. As a result, organizations that attract seriously ill enrollees still will be underpaid, while those that attract healthy ones will continue to be overpaid. This concern is valid, but the new system nonetheless represents a substantial improvement over the existing system.
In MedPAC’s Report to the Congress: Medicare Payment Policy that will be released next week, the Commission has two recommendations that could mitigate many of the concerns associated with implementing a new risk adjustment system for Medicare+Choice.
Recommendation to use diagnosis data from all sites of care
Many of the problems cited for the proposed interim system could be mitigated by replacing it with a permanent one in which health status is based on diagnoses assigned during both inpatient hospital and other types of health care encounters. Thus, MedPAC recommends that:
As quickly as feasible, the Secretary should develop the capability to use diagnosis data from all sites of care for risk adjustment as quickly as feasible.
In its January 15, 1999, 45-day risk adjustment notice, HCFA indicated it intends to replace the interim system on January 1, 2004, with a comprehensive system based on diagnoses from beneficiaries’ encounters with all major types of providers. To make that possible, HCFA will require organizations to augment their hospital inpatient data with information from enrollees’ encounters in physicians’ offices, hospital outpatient departments, skilled nursing facilities, and home health agencies. However, this requirement will not be implemented before October 1, 1999.
Recommendation to phase in the interim system to cushion its effects on payments
MedPAC agrees with the Secretary’s plan to phase in the interim risk adjustment system:
The Secretary’s plan to phase in the interim risk adjustment system—with a method that uses a weighted blend of the payment amounts that would apply under the interim system and those that would apply under the current system—is sound. The weight on the interim payment amounts should be back-end loaded. That is, the weights should be relatively low in the first years so that most organizations will not experience extreme changes in their total payments.
The phase-in should reduce the number of organizations that withdraw from the Medicare+Choice program, but it also will slow the benefits of adopting the interim risk adjustment system. In addition, the phase-in will raise Medicare spending because the reduction in payments that otherwise might occur under the interim system will not be fully realized.
Blended payments will be made during 2000 through 2003. In 2000, payments will be calculated using 90 percent of the existing system and 10 percent on the interim system. Progressively lower weights will be assigned to the existing system in 2001 through 2003. In 2004, payments will be based on full implementation of a comprehensive risk adjustment system that uses data from all sites of care.
Changes in Medicare’s rules for private health plans participating in the program have had both intended and unintended consequences. These changes, introduced in conjunction with the new Medicare+Choice program, were designed to improve Medicare’s risk contracting program by increasing the fairness of the distribution of payments to health plans, by creating incentives to improve quality of care, and by helping beneficiaries to make more informed choices. But taken together with lower base payment updates attributable to the BBA and to unexpected slowing in the growth of fee-for-service Medicare spending, the new rules they may have made participation in Medicare less attractive from the plans’ perspective. Plans have expressed particular concerns about the combined impact of lower base payment updates under the BBA and possible decreases from that base as risk adjustment is implemented.
Improving Medicare’s current risk adjustment system is essential. Risk adjustment is about getting relative payment rates right, so that payments for enrollees in the Medicare+Choice program more closely match their expected costs. It is appropriate that the new system of risk adjustment be phased in, both to avoid any instability that sudden swings in payments to health plans could have for their enrollees and to allow time for policymakers to assess how it is working, but the benefits of better risk adjustment should not be delayed more than necessary. While the Commission recognizes that the transition to the new Medicare+Choice program has been less smooth than many had hoped, we believe the issues raised during the transition should be considered separately from the issue of improving Medicare’s system of risk adjustment.