Studies: Quiet Changes To How Data Is Analyzed Leads To Medicare Advantage Payment Cuts

by Tom Kornfield and Greg Berger

February 1, 2017

Medicare Advantage (MA) health plans provide health benefits to more than 18 million seniors and people with disabilities across the United States. As it sets payment rates for MA plans each year, the Centers for Medicare & Medicaid Services (CMS) determines how much it will pay toward each plan through a process called “risk adjustment.” The sicker the patient population within a plan, the higher the payment that plan receives from CMS. In this way, the federal government helps protect patients who need care the most.

Here’s how that process works: CMS measures the health of an MA plan’s beneficiaries by analyzing the data about their diagnoses. This results in a “risk score.” Recently, the way CMS calculated risk scores began to change. The agency had collected a limited set of data from plans to calculate risk scores.  This system CMS used was called the Risk Adjustment Processing System, or RAPS. Recently, CMS began collecting a new source of comprehensive claims data, called encounter data. These encounter data are similar to what hospitals and physicians submit to traditional Medicare. Encounter data accounted for 10 percent of the risk score in benefit year 2016, increased to 25 percent in 2017, and may increase to 50 percent in 2018.

CMS predicted that use of the encounter data should produce the same risk scores as what would have been produced from the old system (RAPS), and would therefore produce the same payments to plans. However, two new studies suggest that encounter data hasn’t proven to be complete or accurate enough to be a reliable gauge for MA payments.

Avalere and Milliman Study Results

On Jan. 26, Avalere announced it found that on average, risk scores based on encounter data were 16 percent lower in payment year 2016 (using 2015 claims data) compared to original data source (RAPS).  At a blend of 10 percent encounter data, the 16 percent average difference in raw risk scores would translate to an average payment reduction of 1.6 percent in 2016. This reduction in payments could lead plans to reduce benefits or raise premiums as a result.

Similarly, Milliman released a white paper in January estimating the difference in plan risk scores between RAPS and encounter data. Milliman found that the median percentage difference between PY2016 risk scores based on RAPS and the encounter data-based risk scores is -4.0 percent. Based on Milliman’s analyses, at a blend of 10 percent encounter data, the 4.0 percent median difference in raw risk scores would translate to a median payment reduction of 0.4 percent in 2016.

In addition, the GAO released a new report updating its July 2014 study on the steps CMS has taken to validate MA encounter data. GAO determined that CMS has yet to take needed steps to fully address encounter data accuracy, such as reviewing medical records, and the agency should “fully assess data qualify before use.”

What’s Next?

By Feb. 2, CMS is required to release its proposed payment policies for MA plans in 2018, including any changes in how encounter data is used to determine risk scores. CMS should refrain from using encounter data for MA payment until more can be done to ensure the data does not drive down risk scores incorrectly. In this way, we can ensure MA plans continue to deliver affordable, high-quality health care options for seniors and beneficiaries across the nation.

Tom Kornfield is Vice President, Public Programs Policy and Greg Berger is Executive Director, Medicare Policy at AHIP.