Hannah Neprash is an assistant professor in the Division of Health Policy and Management at the University of Minnesota’s School of Public Health. Her research focuses on how clinician decision-making responds to incentives and affects utilization, spending, access and quality of care. 

Risk adjustment is one of those health policy ideas that’s super technical but also incredibly important. Whenever an insurer or provider is on the hook for patients’ health care spending — like in a Medicaid managed care plan, Medicare Advantage plan or accountable care organization — it is in their interest to attract relatively healthy folks who won’t spend much and wreck the budget.

Enter risk adjustment: a tool to make payments higher for plans with lots of sick people and lower for plans with healthier enrollees, on average. Done right, this also allows us to compare plans and providers “apples to apples” without wondering whether performance differences are because of the people covered. But are we doing it right?

A recent study in Annals of Internal Medicine by Jacob Wallace, J. Michael McWilliams, Anthony Lollo, Janet Eaton and Chima Ndumele sheds new light on this perennial question. The authors compared annual total spending between patients who chose a particular Medicaid managed care plan in Louisiana versus those who were randomized into the same plan. We’d expect unadjusted spending to vary widely among the “choosers” who are likely to pick a plan that matches their health risk (i.e. sicker people choose more generous plans, which will have higher spending). But if risk adjustment works perfectly, spending should end up looking identical for both groups. But it didn’t. There was still a nearly 4% difference, on average, in spending between the groups — a meaningful amount in an industry with average profit margins around 3%.

Of course, the authors were limited by only having data from one state’s Medicaid program and only evaluating one common risk adjustment algorithm. But even acknowledging limitations, this finding has major implications since risk adjustment shows up in most Medicaid managed care, Medicare Advantage and Obamacare plans, plus many value-based provider contracts.

First, this finding suggests that plans still have an incentive to attract healthy enrollees if risk adjustment doesn’t fully correct for the characteristics of enrolled patients. And second, risk-adjusted comparisons across plans — which are used by consumers to pick plans and some state Medicaid agencies to measure performance and decide which plans to contract with — still reflect differences in how sick their beneficiaries are instead of how good the plan is at controlling spending or incentivizing the use of high-value care.

We’ve known for a while that risk adjustment is a tricky business, but this study offers new insights into just how flawed it is. Fortunately, the authors also highlight other tools policymakers could use such as reinsurance, pay-for-improvement rather than pay-for-performance, and targeted subsidies to make sure plans and providers have the right incentives to cover and improve the well-being of everyone.