Risk Adjustment Methodology: Reviewing Proposed and Current Model Improvements

On Thursday, March 24, 2016, the Centers for Medicare & Medicaid Services (CMS) released a white paper regarding the risk adjustment methodology. There has been a lot of criticism and discussion about the U.S. Department of Health & Human Services (HHS) risk adjustment program working appropriately. As indicated in the 2017 Notice of Benefit and Payment Parameters (NBPP), the white paper addresses the comments HHS has received regarding the risk adjustment methodology and serves as the basis for discussion for the Thursday, March 31, 2016, meeting. This document provides the story of the evolution of the HHS risk adjustment program, including a summary of historical information on how the HHS risk adjustment program was developed, changes made to the model thus far, and further discussions and considerations for enhancements in the future.

The risk adjustment model is intended to work with the fair rating rules under the Affordable Care Act (ACA). These rules are in place to reimburse issuers that have riskier, costlier enrollees by issuers that carry lower risk, regardless of any other factors, such as being a new issuer or narrow network plan. To enhance the existing stabilization program, modifications have been applied to recalibrate the model for the 2017 benefit year. The white paper just released highlights the areas needing further refinement, such as methodology for inclusion of prescription drugs, partial year enrollments, and future year recalibrations.

History: When the HHS risk adjustment program was being developed, the primary goal in mind was to “compensate health insurance plans for differences in enrollee health mix.” The intent was to allow differences in premium for plan design and benefits but not allow premium differences for any health conditions the prospective member may currently have. When developing how this goal would be accomplished, three specific areas were addressed:

• New Population — During development, data was not available to analyze since this was uncharted territory.  The risk adjustment program needed to include members enrolled into a plan that adhered to the ACA regulations, regardless of whether the member purchased the plan through the Marketplace or directly from the issuer. In lieu of not having specific data to utilize, HHS used a commercial dataset consisting of 2010 Truven MarketScan^® Commercial Claims and Encounter data.
• Market Factors — This introduced the different plan actuarial values in comparison to a standard benefit level.  This is an extremely complex portion of the risk adjustment methodology. HHS encountered challenges defining “how to preserve premium differences that reflect differences in generosity of plan coverage.”
• Balanced Transfers — Since issuers were no longer allowed to adjust premiums for enrollees based on health status, the risk adjustment transfer payment is the part of the risk adjustment program that will make the issuer whole from a financial perspective.

There are apparent similarities between the CMS and HHS risk adjustment programs.  One of the outstanding questions during development was considering which disease classification grouping would be utilized for the commercial market. It was decided to use the CMS-Hierarchical Condition Category (HCC) grouping utilized for Medicare risk adjustment as the basis to develop a diagnosis clinical classification for commercial.  The CMS-HCC grouping consists of 201 HCCs compared to the HHS-HCC grouping that has 264 HCCs. Only a subset of the 264 HCCs is utilized for risk adjustment. These HHCs are classified as “payment HHS-HCCs” since these are the categories that carry a factor weight to be part of the plan liability risk average score (PLRS) calculation. Many HCCs that are part of the Medicare model were split in the commercial model to better predict costs within disease groups, such as those in the metabolic, blood, psychiatric, andDepartment o injury hierarchies.

Model Improvements: There have been improvements to the risk adjustment payment model since the 2014 HHS risk adjustment process was finalized.  Each year the commercial risk adjustment program has been enacted has brought enhancements, subtle at most, to improve the accuracy of the program.

• Payment Year 2015 — Primary goal was to maintain stability. Utilized the same factors for cost-sharing reductions (CSR) plan variation in the corresponding Medicaid alternative plan variations.
• Payment Year 2016 — First year the risk adjustment model was recalibrated. In order to maintain stability and avoid using a small sample size, an approach was utilized to average coefficients using separately solved models of 2011, 2012, and 2013 MarketScan^® data for 2016 benefit year risk adjustment. This was also the year in which HHS addressed two issues surrounding the classification and reporting for infants.  The first issue surrounded infants, classified as age 0 who did not have any birth codes, an issue primarily driven by bundled mother/baby claims. The second issue addressed was for six transplant status HCC coefficients utilized in the child risk adjustment model.
• Payment Year 2017 — Second year the risk adjustment model was recalibrated. A similar blended approach was utilized in the same manor that it was for the 2016 Payment Year. The dataset was updated to include the three most recent years of MarketScan^® data; 2012, 2013, and 2014. For this year, preventive services are included in the simulation process of calculating the plan liability. The 2017 NBPP requested comments regarding further discussion topics that will be addressed at the Thursday, March 31, 2016, CMS meeting.

Proposed Model Improvements: CMS is hearing the concerns coming from the health plans and is actively engaging them for comments in regards to issues. This interaction allows for an avenue of different perspectives to enhance the risk adjustment program with solutions that will further advance the model. There is no specific avenue HHS is taking in regards to the below improvements at this time. Rather, they are gathering all of the feedback they have received thus far and opening it up for discussion to determine the next best step to take.

• Partial Year Enrollment — Health plans are concerned about the adverse effect partial year enrollments is having on their risk adjustment payment transfer outcome. The industry is seeing enrollees with 6 months or less enrollment with high Medical Loss Ratio (MLR). This type of membership “flip flop” has been a concern of health plans since the onset of the ACA. This is where an individual elects an ACA plan on a temporary basis to only handle a current condition and not to maintain health insurance coverage long-term. In turn, health plans experience high claims cost, minimal revenue, and subpar balance to the risk they are carrying for these individuals. With the commercial demographic, the majority of conditions are considered acute as opposed to chronic conditions. This statement is the exact opposite of what you would see in the Medicare space.
• Proposed Drug Modeling — The use of pharmacy claims is an important analytical element used to validate a member’s diagnosis and to proactively research for potential undocumented chronic conditions. HHS is talking about introducing a hybrid risk adjustment model that would use prescription drug utilization as risk indicators for the HHS-HCC model. The framework and operational impact to maintain a model like this is up for discussion but certainly is a step in the right direction.
• Use of a Concurrent Model — There is discussion around if utilizing a prospective model, rather than a concurrent model, will be more beneficial for the HHS risk adjustment program. A concurrent model predicts costs within the current year. This type of model tends to address more acute costs, whereas a prospective model allows greater time to review prior encounters to better predict the future costs. This allows a greater time lag for encounters to be analyzed and submitted, allowing time to capture more advanced acute and chronic conditions.

CMS clarifies the program is functioning as it was intended to, with which I would agree. The operational complexities, concurrent model, and mediocre data management practices established by health plans is leading to incomplete data submissions and, therefore, inaccurately reflecting the risk of the organization. The established model and demographic understanding will evolve over time by utilizing more accurate data to stabilize the factors used in the calculation when it becomes available. Until then, it is important for health plans to work on what they can directly impact by refining the risk adjustment processes they have in place, establish analytics and reporting practices, and ensure a thorough extraction process is conducted for the EDGE server submission inclusive of controls, pre-validation checks, and error resolution processes.

Resources

GHG can help you streamline the execution of your risk adjustment approach, and build a roadmap to ensure you’re keeping pace with CMS and/or HHS expectations in both compliance and health care outcomes. Contact us today to learn more >>

For actionable advice and best practices, join us at our annual Gorman Health Group 2016 Forum, April 19-20, at the Worthington Renaissance Fort Worth Hotel in Fort Worth, Texas. During this year’s information-packed two days, our elite team of experts, operators, clients, and partners will help you figure out what matters and what doesn’t. We will share proven tactics to cut costs, increase member satisfaction, and manage and drive sustainable growth. The hotel room block expires on April 4 so register now  to reserve your seat!

Stay connected to industry news and gain perspective on how to navigate the latest issues through GHG’s weekly newsletter. Subscribe >>