Observations from a New Frontier: Medicare’s Experience with Risk Adjustment

Medicare’s attempt to implement a popular new payment model may have backfired. In 2004, the Center for Medicare and Medicaid Services adjusted reimbursement rules to prevent overpayments. New research suggests that Medicare Advantage plans mitigated the new rules’ effects by adapting their behavior. In the article “How Does Risk Selection Respond to Risk Adjustment? New Evidence from the Medicare Advantage Program,”Jason Brown, Mark Duggan, Ilyana Kuziemko, and William Woolston investigate system responses to risk-adjusted payments. This example of unintended consequences contains valuable lessons for US policymakers experimenting with new healthcare payment models.

Before 2004, Medicare overpaid Medicare Advantage (MA) plans. The government used local demographic features to determine per-enrollee payment amounts. MA plans are required to enroll every eligible Medicare applicant, but, in practice, MA plans enrolled people who were healthier than average. These relatively healthy people cost less than the average population. The MA plans were then able to keep the overages from their healthier-than-average population.

After 2004, Medicare paid the MA plans a risk-adjusted amount. They based the amount on a complex formula, including data on disease conditions, to determine a health risk factor. Medicare pays MA plans more for people with high risk factors and less for people with low risk factors. The government designed this policy to limit overpayments and encourage MA plans to enroll the sickest individuals. Recent research suggests the policy met neither goal.

In their study, Brown, Duggan, Kuziemko and Wollston developed a theoretical model for risk-adjusted payments. The model predicts that risk-adjusted payments encourage MA plans to enroll less healthy individuals but not the sickest individuals. The MA plans still profit from overpayments, but increased screening costs cut profits. The key component is the relationship between the health risk factor and health cost variance. As the risk factor rises, the variance in health costs associated with people at that risk level also rises. Medicare bases payments on the mean healthcare costs at a risk-factor level. The risk score better predicts health costs for healthy individuals. For sicker individuals, the risk score is more likely to widely overestimate and underestimate true costs.

The authors suggest that MA plans can identify and enroll the individuals with health cost overestimates. The formula for the risk score is not perfect. Certain observable characteristics may predict if risk score prediction will be an overestimate or an underestimate. Using the same techniques previously used to target healthy patients, MA plans could instead target and enroll these sicker but more lucrative patients.

Next, the authors look for empirical evidence to match their model’s predictions. They focus on post-2003 data for individuals who switched from fee-for-service Medicare plans to MA plans. The results support predictions of their model. The post-2003 risk factor score of an individual switching fromfee-for-service Medicare to an MA plan rises significantly. This is consistent with the theory that MA plans would enroll sicker people. The model also predicts that MA plans will target people who are relatively healthy when compared with those with similar risk factors. The authors find that holding risk score constant, the health expenditures of individuals significantly fall post risk adjustment.

Further extrapolating their results, the authors predict risk adjusted payments increased Medicare overpayments to MA plans. Without more detailed data from the MA plans, the authors are not able to make a claim about MA profits. The authors also do a preliminary search into possible welfare benefits and come up empty-handed. Their brief analysis does not suggest that patients benefited from the new payment scheme.

Risk-adjusted payments are a key component of the ACA and healthcare reform in general. The current fee-for-service model misaligns incentives and results in overspending. Brown et al. remind us that risk-adjusted payments also produce incentives and market responses. It is important for policy makers to consider these unintended consequences as they implement health care reform.

Article Source: Brown, Jason, Mark Duggan, Ilyana Kuziemko, and William Woolston. 2014. “How Does Risk Selection Respond to Risk Adjustment? New Evidence from the Medicare Advantage Program.”American Economic Review, 104(10): 3335-64.

Feature Photo: cc/(mateoutah)