ACA risk adjustment: Special considerations for new health plans

  • Print
  • Connect
  • Email
  • Facebook
  • Twitter
  • LinkedIn
  • Google+
By Jason Siegel, David M. Liner | 02 July 2015

The Affordable Care Act (ACA) implemented a risk adjustment program to mitigate the risk of adverse selection starting in 2014. Data recently released by the Centers for Medicaid & Medicare Services (CMS) indicates that health plans new in 2014 are making net payments into the risk adjustment pool of about $142 million. This paper explores potential causes and proposes strategies to optimize risk adjustment performance.

Risk adjustment is the only permanent one of the three premium stabilization programs created by the ACA. Federal regulations describe the intended purpose of the premium stabilization programs:1

“These programs will mitigate the impact of potential adverse selection and stabilize premiums in the individual and small group markets as insurance reforms and the Affordable Insurance Exchanges (‘‘Exchanges’’) are implemented, starting in 2014.”

Federal regulations also provide a specific description of the risk adjustment program:

“The permanent State-based risk adjustment program provides payments to health insurance issuers that disproportionately attract high-risk populations (such as individuals with chronic conditions).”

The significant net payment into the risk adjustment pool was likely driven by some combination of a healthier than average enrollment mix achieved by some plans, and also a series of dynamics that will tend to disadvantage new plans. It is impossible to say to what extent each of these factors is at play in the 2014 markets. This paper explores the dynamics which, in some cases, may cause new plans to face a competitive disadvantage, and also strategies that any issuer (new or existing) can use to ensure the best and most accurate risk transfer payments.

The dynamics that we explore which in some cases may put new health plans at a competitive disadvantage are:

1. Lack of prior year data

2. New provider and administrative relationships

3. Low relative market share

How risk adjustment works

Risk adjustment starts by quantifying an enrollee’s health status based on age, sex, and diagnoses recorded during the course of the year. Diagnoses are coded by medical providers when an enrollee utilizes medical services.

An underlying issue with this mechanism is that a health plan’s population may be measured as healthier than it actually is if some diagnoses are not captured, resulting in potentially unfavorable risk adjustment transfer payments.

For example, a diabetic enrollee will not be coded for diabetes until the enrollee encounters a medical provider and the provider properly codes for the enrollee’s condition. Because the ACA risk adjustment program does not consider prescription drug data in its diagnosis assignments, a diabetic enrollee who receives insulin—but does not visit a physician during the year—will not be coded as a diabetic.

What the data show

On June 30, 2015, CMS published the first glimpse many issuers have had into what their risk adjustment transfers may look like in a report titled, “Summary Report on Transitional Reinsurance Payments and Permanent Risk Adjustment Transfers for the 2014 Benefit Year.” This report includes data for all issuers subject to the 3Rs, except that it excludes risk adjustment results for issuers in Massachusetts because that state implements its own risk adjustment transfer program. Risk adjustment results for these issuers were also released on the same day in a press release titled “Commonwealth Completes First ACA-Required Risk Adjustment Process with Carriers in State’s Merged Market.” By definition, the net payment into and out of the risk adjustment pool will be $0.

We limited these data to all Consumer Operated and Oriented Plans (CO-OPs). There may be new issuers in 2014 other than CO-OPs, but we used this class of issuers as a proxy for all new plans because they are readily identifiable as new plans. We then determined the net payments into or out of the risk adjustment pool for these issuers.

Virtually all of these plans in the small group market paid into the program, resulting in a net payment of about $73 million. In the individual market, the results are less extreme, and only about half of these plans paid into the pool. However, those paying into the pool experienced significantly higher transfers than those receiving money from the pool, resulting in a net payment of about $66 million. One of these plans is operating in a merged market (i.e., individual and small group combined) and it paid about $3 million into the pool.


We identified three potential issues that may have affected these results. It is not clear to what extent each of these issues may have affected any particular issuer in 2014.

Issue #1: Lack of prior year data

A year-over-year analysis of commercial market data suggests that many of the chronic conditions included in the Department of Health and Human Services (HHS) risk adjustment program are not coded in any given calendar year at least 20% of the time. Supplementing current year data with experience from prior years to identify potentially missed diagnoses is a common starting point for identifying coding improvement opportunities. This technique is often considered the “low hanging fruit” of risk adjustment revenue optimization. However, a new plan without prior year data will not be able to employ this method.

For example, consider an enrollee who was coded for asthma in one year but not in the next. It is likely that the enrollee was an asthmatic in both years, but was not properly coded for asthma in the second year. Existing health plans may extrapolate from prior year data to identify enrollees of this type, but new health plans do not have the data to identify enrollees missed in this fashion, creating a competitive disadvantage for these plans. Members identified using this method still must receive a valid diagnosis from a provider in order for the plan to be credited for the condition.

To some extent, this issue affects all issuers as even those that have been in existence for some time will cover some members in a given year that are new.

Issue #2: New provider & administrative relationships

Experience with risk adjustment in the Medicare Advantage market suggests that provider engagement can lead to improved diagnosis coding and thus risk adjustment revenue, whereas unmanaged diagnosis coding may lead to significant deficiencies. Medicare Advantage plans that focus on improved diagnosis coding have traditionally increased risk adjustment revenue by at least 1% to 2% annually.

An optimal risk adjustment strategy relies on a robust administrative system to report valid demographic and diagnosis information, and to carry those data through to the EDGE server. New health plans must manage new administrative system implementation and are often faced with negotiating new vendor relationships. New health plans with an evolving administrative system may have a greater proportion of invalid data submissions that do not meet risk adjustment reporting requirements. Invalid diagnosis data submissions will result in reduced risk score revenue for health plans.

In addition, provider relationships also affect a health plan’s ability to optimize diagnosis reporting. Health plans benefit from providers that report conditions in a consistent manner and at a correct level of severity, and support diagnoses with sufficient documentation. New health plans disproportionately rely on rental networks given the lack of time, resources, and economies of scale needed to build their own. This results in less control over provider practices than if using a captive network, resulting in additional challenges to ensure that providers properly code patients for optimal risk adjustment performance.

Issue #3: Low relative market share

Given a lack of brand recognition and an existing enrollment base upon which to build, new health plans often start out small relative to their competitors.

Given the mechanics of the risk adjustment program, small issuers face greater variability risk in their potential risk adjustment transfer payments. This is because small issuers can enroll a wildly different population mix than their competitors, and their risk transfer payment will be based on the full difference in morbidity between their enrollment and the average for their market.

Large issuers with a more stable block of business will experience significantly less variability in the average morbidity of members they enroll simply due to the law of large numbers. Additionally, to the extent that an issuer makes up a material portion of the enrollment in its market, its risk transfer payment will actually be dampened because it is affecting the market average to which it is being compared. An extreme example would be an issuer enrolling 100% of the members in its market; by definition, it would experience no risk transfer because it is a zero sum program.


So what can be done?

New health plans face a number of competitive disadvantages related to risk adjustment as discussed in this paper. However, these disadvantages can be mitigated with effective optimization strategies. Note that many of these strategies apply to existing health plans as well and should be explored by any issuer in the ACA markets.

There are at least three components to an impactful risk adjustment optimization strategy:

1. Robust administrative system

2. Coding accuracy initiatives

3. Provider and enrollee engagement

An optimization strategy that does not include each of these components will not optimize risk adjustment outcomes.

A robust administrative system serves as the foundation for risk adjustment optimization strategies. A sound administrative system is required for valid data submission and enables health plans to effectively pursue coding improvement initiatives. System audits are an effective technique for validating a recently implemented administrative system.

Initiatives to improve the accuracy of diagnosis coding can significantly reduce the competitive disadvantage for new health plans. For example, a lack of full year enrollment and prior year data may result in a new plan initially under-coded relative to an existing plan. However, that also means that robust algorithms for identifying missed diagnoses are likely to identify more of these opportunities for new plans, potentially closing a large part of that gap.

For example, one cutting edge technique that health plans have used to improve coding accuracy is to leverage relationships with pharmacies and pharmacy benefit managers (PBMs) in order to gain access to multiple years of prescription drug utilization data for a majority of the nationwide commercial market population (i.e., not limited to their own members during the time period in which they are enrolled). When a member initially enrolls and completes the necessary HIPAA authorizations, these databases can be linked to a health plan’s enrollment data to identify each of these member’s drug utilization history and thus many likely chronic conditions for new enrollees.

Another best practice involves developing elaborate algorithms on top of large commercially available datasets, using all possible elements from a health plan’s data to identify potentially missed diagnoses. This approach relies on identifying patterns among at least medical procedures, comorbidities, specialist office visits, and prescription drug utilization. Best in class models are set up to handle numerous interactions between these data elements and maximize the extrapolation power of these data through machine learning techniques.

Provider engagement is also a key strategic component because diagnosis coding starts with providers. Levels of engagement may range from education only to elaborate compensation schedules. Educating providers on the importance of valid diagnosis coding may improve risk adjustment outcomes. An additional level of provider engagement may be achieved by incentivizing optimal coding through reimbursement arrangements. Achieving a high level of provider engagement may require more effort in the short term than other strategies, but can also produce benefits over a longer horizon.

Enrollee engagement is another optimization strategy component. A focus on enrollee satisfaction to minimize intra-year lapses and maximize inter-year re-enrollment will reduce the negative effect of partial year enrollment on risk adjustment. Enrollee education on the quality of care benefits from improved diagnosis coding may complement a provider engagement strategy.

Thoughtful plan design features can also encourage member engagement and help ensure that care is managed properly and diagnoses are coded accurately. For example, some plans will offer free health care screenings when members first enroll. This aids in getting the right members to providers quickly and documenting relevant conditions they may have.


If you have any questions or comments on this document, please contact Jason Siegel or Dave Liner, who are actuaries with Milliman. Jason and Dave are members of the American Academy of Actuaries and meet the Qualification Standards of the American Academy of Actuaries to render the actuarial opinion contained herein.

Jason Siegel, FSA, MAAA
+1 262 796 3414

David M. Liner, FSA, CERA, MAAA
+1 860 687 0137


This communication has been prepared for the specific purpose of discussing the mechanics of the ACA risk adjustment program. This information may not be appropriate, and should not be used, for any other purpose.

In performing this analysis, we relied on information published by others. If this data or information is inaccurate or incomplete, the results of our analysis may likewise be inaccurate or incomplete.

Milliman does not intend to benefit or create a legal duty to any third-party recipient of its work. This communication must be read in its entirety. Differences between our estimates and actual amounts depend on the extent to which future experience conforms to the assumptions made in this analysis. It is certain that actual experience will not conform exactly to the assumptions used in this analysis. Actual amounts will differ from projected amounts to the extent that actual experience deviates from expected experience.

1Federal Register Retrieved December 18, 2014, from