Association for Public Policy Analysis & Management

Purpose. CMS recently announced a goal to have 30 percent of Medicare payments outside traditional fee-for-service (FFS) systems by the end of 2016, and 50 percent of payments outside FFS by the end of 2018. One challenge to establishing non-FFS payments is that such systems may create incentives for providers or insurers to select lower-risk beneficiaries. Risk-adjusted payments may limit the potential selection effects of such systems, but create incentives for more intensive documentation of enrollees’ service use and chronic conditions. If not anticipated and accounted for, those incentives may cause risk-adjusted payments to be higher than expected. Using Medicare Advantage (MA) as a case study, this research analyzes how MA enrollment affects beneficiary risk scores using a difference-in-differences (DID) research design to address selection concerns raised in previous research.

Design. We use Medicare administrative data to compare the growth in risk scores of 2.5 million beneficiaries that switched from FFS to MA (switchers) to those of 18.2 million beneficiaries that remained in FFS (stayers) between 2006 and 2012. For simplicity, we exclude beneficiaries who switched back to FFS and those enrolled in MA plans other than private FFS plans and HMOs. We estimate both the MA switch effect and the incremental effect of additional years in MA, controlling for beneficiary demographic and enrollment characteristics. The model addresses selection bias using a FFS baseline for both groups and estimating the effects of MA enrollment on the change in risk scores.

Preliminary Findings.Our preliminary findings show that risk scores rise with MA enrollment and its duration. Between 2007 and 2012, the mean risk score for switchers increased 0.48 points (from 0.90 to 1.38) compared to 0.41 points for stayers (from 0.97 to 1.38). Our results suggest that MA enrollment can explain most of that difference—switching to MA is associated with a 0.054 increase in risk score, and each year of continued MA enrollment is associated with an additional expected increase of 0.014 (see Table).

Discussion. Our findings support the hypothesis that risk scores increase upon switching to MA. That finding is consistent with existing research, and CMS adjusts risk scores of MA enrollees for payment purposes to reflect those differences. However, our findings also show that risk scores continue to rise with duration of MA enrollment. That distinction is important when considering how to adjust risk scores for payment purposes. The findings reported here have implications for MA payments specifically but also apply to other payment mechanisms in Medicare and Medicaid. As CMS aims to move payments away from traditional FFS systems, reliance on risk-adjusted payment mechanisms is likely to increase. This research suggests that those mechanisms could increase total spending if they do not account for how insurer and provider incentives may affect beneficiary risk scores—both at initial enrollment and over time.