Risk adjustment is a key mechanism to ensuring appropriate payments for Medicare Advantage plans, Medicare Part D drug plans, and Medicaid health plans.  Since health plans vary in their mix of healthy and sick enrollees, risk adjustment modifies premium payments to better reflect the projected costs of members served and compensate plans that enroll high-cost patients.

Historically, risk adjustment was only used in Medicaid and Medicare – in effect, redistributing some revenue from health or drug plans with a relatively healthier mix of members to those plans with a more costly enrollment profile.  However, the Affordable Care Act (ACA) extends risk adjustment to the individual and small group health insurance markets starting in 2014.

A new brief from The Synthesis Project tackles the issue and makes several interesting recommendations for how to improve risk adjustment methods for the post-ACA market. Without accurate risk adjustment, health plans have a strong financial incentive to seek out only the healthiest enrollees, especially under ACA-mandated adjusted community rating.  Under adjusted community rating, health plans may not vary premiums based on health status or sex and are limited in how much they may vary premiums based on age.  Under ACA, the healthy, the young, and men subsidize the health costs of the unhealthy, the older, and women.

Risk adjustment is therefore a necessary factor in stabilizing the dramatically new post-ACA health insurance marketplace, particularly the new Health Insurance Exchanges.  Even then, the ACA is a giant game of musical chairs.  The market under ACA will be chaotic and challenging, with a mix of winners and losers once the music stops and the dust settles, which will take at least three to five years.

The Basis for Risk Adjustment:

The amount a consumer pays for insurance, in theory, should equal the expected cost of insuring that consumer, which is calculated as the cost of an event times the probability that event will happen. For example, a person’s car insurance premium for a given year would equal the costs associated with an accident times the probability the person will be in an accident that year. Health insurance is more complicated but works essentially the same way.

Measuring expected health costs is often tricky, and some patients end up costing health plans far more than what they pay in premiums. Insurers also have to tussle with adverse selection, which happens when the sickest people buy insurance and the healthy stay away – driving up health costs and destabilizing the risk pool upon which premiums and ultimately the insurance model are based.  As a result, some health plans in the past have tried to limit the chance that costs will exceed premium revenues by encouraging healthier people to enroll (“cherry pick”) and by discouraging potentially costly enrollees (“lemon drop”).  However, the ACA regulatory framework – which includes guaranteed issue, adjustment community rating, mandatory coverage of pre-existing conditions, and no annual or lifetime limits – changes things completely.

Risk Adjustment in Medicare, Medicaid, and the ACA:

In recent years, policy makers have created risk adjustment programs for Medicare Advantage plans, Medicare Part D prescription drug plans, and Medicaid managed care plans.  Medicare and state Medicaid agencies programs give health plans risk-based payments to offset the cost of enrolling more expensive people, reducing the incentive health plans have to limit access to care for unhealthy people. The more risky and potentially expensive a health plan’s population is, the larger payments a health plan receives. Such risk-adjustment mechanisms, in essence, transfer money from plans with relatively healthy enrollees to those with relatively unhealthy and expensive enrollees.

Section 1343 of the Affordable Care Act (ACA) allows states to establish risk adjustment programs for health plans in the individual and group markets, both inside and out of the Health Insurance Exchanges (HIX). As with the exchanges, the federal government will run risk adjustment programs for states that choose not to. Read the Centers for Medicare and Medicaid Services (CMS) final rules on risk adjustment here.

Risk adjustment is particularly important now because the health reform law placed new restrictions on premiums health plans could charge to high-cost groups. For example, a consumer’s health status and sex may not be used to set premiums, even though unhealthy consumers and women tend to have higher per capita medical costs. Health plans will also be limited in their ability to vary premiums based on a person’s age.

For more information on risk adjustment and the related topics of risk corridors and reinsurance, see my previous blog posts:

Assessing, Improving Risk Adjustment Models:

Risk adjustment programs use models to project a person’s expected health costs and determine appropriate payments to health plans. Data used in risk models include a person’s demographic characteristics, medical conditions, and sometimes health usage data from the current or the preceding year. The model sets a risk score based on those factors to account for how much more a particular person is expected to cost compared to the average cost for eligible enrollees.

But how well do the models work? And how successful are risk adjustment programs at encouraging health plans to cover the least healthy people?

Researchers at The Synthesis Project, an initiative of the Robert Wood Johnson Foundation (RWJF), recently produced an excellent report and brief to help answer these questions.

Here is an overview of the findings:

  • Various models that use data from the previous year all underestimate costs for high-cost patients and overestimate costs for low-cost patients.
  • Nonetheless, using data from the previous year tends to make models more accurate, though health plans receiving risk adjustment payments then have an incentive to inflate costs to receive larger payments in subsequent years.
  • Models using current-year data are more accurate than those using prior year data because they take into account new or worsened medical conditions.
  • Studies show Medicare Advantage risk adjustment payments – in existence for more than a decade – have increased the spread of expected costs among patients – meaning health plans are willing to cover more expensive enrollees thanks to risk adjustment.
  • However, high-cost enrollees still made up a larger share of people who dropped out of Medicare Advantage plans than of people who stayed in the plans.

Authors Eric Schone and Randall Brown of Mathematica Policy Research and Sarah Goodell of The Synthesis Project make several recommendations to improve risk adjustment models. Here are a few of the recommendations I found particularly interesting:

1) Make upcoding reduction for risk adjustment payments plan specific.

Providers seeking payments sometimes label, or code, a treatment case in the most severe category the treatment will allow. Providers might also treat conditions such as heart disease or arthritis that are not related to the reason a person sought care.  Traditional fee-for-service payments create a strong incentive for providers to maximize their revenues through coding patients and services are high as possible. Meanwhile, risk adjustment gives health plans a strong incentive to gather as much data as possible to justify high risk adjustment scores.  Bottom line: “Upcoding,” as those practices are called, in addition to increasing an individual provider’s payments from the health plan, makes enrollees seem more expensive than they are, increasing risk adjustment payments to the health plan.

Risk adjustment payments in Medicare Advantage are summarily reduced by an average expected level of upcodingThe Synthesis Project brief points out that practice gives Medicare Advantage plans an incentive to upcode even more, to make up for the payment reduction. It would be better to set payment reductions from upcoding for each plan individually, based on whether the plan’s risk profile increased more than average from one year to the next.

2) Set maximum limits for per capita cost data to include in risk adjustment models.

As with any model, outliers will skew the results. The brief suggests setting a dollar limit for per-person costs included in the models. Doing so would make the model more accurate in calculating expected costs for most people. Health plans could then rely on reinsurance programs, such as those in the ACA, to recoup costs for outliers.

3) Improve incentives for insurers to encourage enrollees to improve their health.

Risk adjustment payments increase as enrollees become less healthy and therefore more expensive. In theory, that might create an unintended incentive for health plans to allow their enrollees’ health to deteriorate, though that conclusion assumes risk adjustment payments fully cover the costs of insuring expensive patients, which they don’t. Risk adjustment programs could encourage health plans to promote health improvements by including some form of quality or outcomes-based measure.

Read the full report here and the brief here.