Under the Affordable Care Act (ACA), the Centers for Medicare & Medicaid Services (CMS) was tasked with implementing a risk-adjustment model for ACA-compliant business in the commercial non-group and small group markets beginning in 2014. The risk-adjustment model is one of the “premium stabilization programs” Congress wrote into the law and was intended to “reduce or eliminate premium differences between plans based solely on expectations of favorable or unfavorable risk selection.” In what follows, Kurt Giesa, head of Oliver Wyman Actuarial Consulting’s Healthcare Practice, presents four charts summarizing the impact of the model:
Taken together, the four charts below tell the story of risk adjustment in the non-group market in 2014. Each point in these charts represents an issuer’s experience in a state, and the y-axis in each of the charts shows the issuer’s risk-adjustment receipts per member per month (PMPM). A positive receipt means the issuer received funds through the risk-adjustment model, and a negative means the issuer paid into the system.
We think these charts show that, while there are certainly outliers, and there likely is room for improvement, from a nationwide perspective, the CMS-operated risk-adjustment system has contributed positively toward the intended objective of allowing issuers to compete on bases other than risk selection.
We show how risk-adjustment receipts varied with the payments the issuer received under the transitional reinsurance program, another of the “premium stabilization programs.” In 2014, the transitional reinsurance program paid issuers 100% of an enrollee’s claims between $45,000 and $250,000. The transitional reinsurance program phases out over time, and 2016 is the last year it will operate. Chart 1 shows that risk-adjustment receipts were strongly correlated with transitional reinsurance receipts (R2 = 0.43). In other words, and not surprisingly, plans with a relatively large number of high-cost claimants as measured by transitional reinsurance program receipts tended to be recipients of risk-adjustment payments as well.
We show risk-adjustment receipts by issuer size where issuer size is measured in terms of ACA-compliant, non-group member months. The least-squares line included in Chart 2 shows that risk-adjustment receipts PMPM tended to decline as issuer size increased, though the effect is small (R2 = 0.01). In other words, it does not appear as though there was a systematic bias based on member months in the risk-adjustment model in 2014.
The risk-adjustment model operates at the state-wide level. In Chart 3, we show how an issuer’s risk-adjustment receipts varied with claims relative to the state-wide average claims. Those issuers with claims that were higher than the state-wide average were also more likely to be recipients of risk-adjustment payments (R2 = 0.59). In essence, the risk-adjustment system tended to bring down the cost of claims for plans with high claims PMPM and bring up the cost of claims of plans with relatively low claims PMPM.
Finally, in Chart 4 we show how risk-adjustment receipts PMPM varied with net underwriting gain. (In calculating net underwriting gain, we included risk-adjustment and transitional reinsurance payments, but excluded reported payments under the third premium stabilization program, risk corridors, due to a lack of funding for this program.) Based on the least-squares line included in Chart 4, underwriting gains increased with risk-adjustment payments PMPM, though the correlation between underwriting gain and risk-adjustment receipts from Chart 4 is much lower than the correlation between risk-adjustment receipts and the difference between an issuer’s claims PMPM and the state-wide average claims PMPM illustrated in Chart 3. (R2 from Chart 3 = 0.59; R2 from Chart 4 = 0.01.) This would suggest that risk-adjustment has a much larger effect on normalizing claims to a state-wide level than on an issuer’s profitability, and that factors other than the payments or receipts under the risk-adjustment system are more likely to be driving issuers’ underwriting results.
While there are clearly outliers in the data, from a system-wide perspective the CMS operated risk-adjustment system appears to have worked as intended in 2014, transferring funds from plans with relatively low risk (using claims as a proxy for risk) to plans with relatively high risk.
A word on data: The results as presented make use of the data as reported by issuers in their MLR reports. We did not audit or validate this data. It is not possible given the data available to adjust the data for differences among issuers in provider contracts and reimbursement levels, care management capabilities, issuers’ efforts related to managing the risk-adjustment process, or a number of other of the characteristics of the issuers’ business that could have an impact on the results we present here. In addition, 2014 was the first year of the operation of the risk-adjustment system. We expect that issuers’ ability to submit complete and accurate data will improve over time as they become more familiar with the risk-adjustment system. For this and other reasons, the results from 2014 may not be representative of results in the future. The data used in this analysis are available here. In developing these charts, we excluded issuers in a state if they had less than 12,000 member months in that state to remove excess variability from the results. We also excluded issuers from Massachusetts (because Massachusetts operated its own risk adjustment system in 2014), West Virginia (because there was only one issuer operating in the state), and Vermont (due to its merged market).