For decades before the Affordable Care Act (ACA), American health insurers were very good at avoiding enrolling people who were likely to be high-cost in the individual (or “non-group”) health insurance market. When individuals applied for coverage, they were asked questions such as: “What illnesses do you have?” and “What medical treatment have you had in the last (choose one) three, five, or 10 years?” Any person that answered with a serious (or not so serious) condition was denied coverage. Others with less serious conditions may have been charged additional premiums for a pre-existing condition or had exclusions of certain services or conditions in their policies.

That all changed on January 1, 2014 with the ACA. All insurers offering new policies in the individual market now are required to provide coverage to anyone who can pay his or her premium. This is known as a “guaranteed-issue” requirement. But even with this requirement, we can’t assume insurers truly offer coverage to everyone in the individual market.

There are still ways for insurers to avoid the sickest people — the ones who might have future costs in the tens of thousands (or more) of dollars. Some methods appear innocuous — just aim advertising at healthy people who use gyms. Others are more controversial — develop networks of physicians or hospitals that discourage access to care with, for example, fewer oncologists. What could Congress do to minimize this avoidance of people who are more likely to use care? One good answer: risk adjustment.

The ‘Three R’s’

To increase the likelihood of the ACA’s success, three premium stabilization programs were embedded in the law and are known as the “Three Rs,” for reinsurance, risk corridors, and risk adjustment. The first two are temporary through 2016. Insurers were concerned that with the addition of guaranteed issue, many new enrollees in 2014 and 2015 would be those who had been denied coverage or who had had pre-existing condition limits in the past.

As a result, these new enrollees might be very high-cost and could drive up premiums. Since unpredictable high claims would be a temporary issue at the start-up of new ACA coverage, it was thought that reinsurance and risk corridors would not be needed after the first three years.

Reinsurance gathers funds from all types of commercial health coverage (including both employer-based and individual) on a per-head basis and distributes it to the new individual market to reduce any “spikes” from unknown health risk in 2014 and pent-up demand for services. This worked well in 2014, and the Department of Health and Human Services (HHS) was actually able to increase the amount of reinsurance payments 25 percent above the expected level.

Risk corridors are modeled after the approach used by the George W. Bush Administration for Medicare Part D to help stabilize financial results of insurers when that new law was implemented. While risk corridors are controversial among some in Congress and inside the Beltway, they worked well during the first few years of the Medicare Part D program.

Risk corridors reduce costs to insurers by using Treasury funds if actual claims turn out to be more than 3 percent higher than the “target” set by the insurer. These funds are offset by any gains from claim amounts more than 3 percent below this target, which insurers share with the Treasury. But Congress changed the rules for the ACA risk corridors, and so far, Treasury has collected insufficient funds to fulfill the original promise of the ACA to assist insurers that were adapting their business to the new guaranteed-issue world.

Risk adjustment is the permanent element of the Three R’s that works to “level the playing field” among insurers. It is structured so that any insurer that welcomes all new enrollees and gets sicker-than-average people (whether by chance or design) will receive compensation from the Centers for Medicare and Medicaid Services (CMS) to help pay for “extra services.” The agency collects funds from insurers who enroll more low-cost, healthy people and distributes them to insurers who enroll more high-cost people. In some cases, these additional risk-adjustment funds may be “cascaded” down from insurers to capitated physician groups, either increasing or reducing their revenue based on how sick or healthy their patients are.

A Short History Of Risk Adjustment

Starting with Medicare private policies (now called Medicare Advantage), CMS began using inpatient diagnostic information for risk adjustment in 2000, and then improved the precision of the risk-adjustment system with the addition of professional encounter data for chronically ill people in 2004. The ACA has now carried this approach forward to the individual and small-employer markets starting in 2014.

Has it worked? Some observers have said that when you saw an insurer’s billboard advertising to people with chronic conditions, we’d be there. Guess what? In late 2015, Aetna, one of the biggest U.S. insurers, began advertising to diabetics — because they believed they could help manage care better and control costs for diabetics, and risk adjustment would pay them enough to be profitable with these enrollees.

How Does Risk Adjustment Work?

Insurance is expected to pay for unexpected, random “bad things,” like accidents. Those continue to be the “risk” that insurers get paid to manage. But, for chronic conditions like diabetes (relatively low but regular costs) or cystic fibrosis (very high and regular costs), actuaries know that there may be a lifetime of extra expenses.

Risk adjustment systems choose a limited number of discrete, ongoing, costly conditions and pay insurers extra for them, in addition to regular premiums from individuals or employers. As used under the ACA, risk adjustment is a so-called “zero-sum game,” where insurers who have a large share of chronically ill people receive payments, and other insurers who have fewer than average ill people pay into a “risk adjustment pot” to make the payments. The U.S. Treasury does not have a funding role (contrary to the statements of some critics).

As shown in Exhibit 1, comparing Medicare risk adjustment with the similar method used under the ACA, there are close to 100 Hierarchical Condition Categories (HCCs) that are serious chronic conditions accounted for in the models. There is a “hierarchy” in the method, which means that a more serious type of one condition (for example, diabetes with complications) is not double-counted with the less-costly version of that condition that might be observed earlier in the year. Seniors (in the Medicare model) have many chronic conditions — 61 percent have one or more HCCs. Among people under age 65 (in the ACA model), 23 percent have one or more HCCs.

Exhibit 1: Medicare Versus ACA Risk Adjustment Characteristics

Bertko_Exhibit_1

Source: CMS

Another difference is that the Medicare risk adjustment is prospective which means it is used to predict costs for next year. A prospective model is thought to improve incentives to manage care efficiently, if prior data are available. In contrast, under the ACA model, CMS had no prior knowledge of conditions for 2014, so is operating a concurrent model, which uses 2014 diagnostic data to run the 2014 risk-adjustment algorithm (this is especially important for high-cost neonates — they are not chronic, just very expensive, and are included in the ACA risk adjustor).

The concurrent model also works well in the context of the considerable “churn” in the ACA-enrolled population during a year. Consumers who find better jobs move to employer-sponsored insurance (and out of Exchanges), and some who become very sick lose employment and become eligible for Medicaid (and also move out of Exchanges). Thus, the concurrent risk-adjustment method appropriately directs payments to insurers based on who was actually enrolled during the year or part of the year.

Data: Fuel For Risk Adjustment

What is one of the key requirements for risk adjustment to work? Data on enrollees’ health conditions. While this may seem obvious, it is fiendishly difficult in a brand new program to get insurers to collect the right diagnostic data from providers and submit it under strict privacy guidelines to CMS!

To keep the agency from obtaining “too much data,” CMS used a distributed data approach. In simple terms, this means that each insurer is given standard, regulated software code to use internally, on its own computers, to analyze its claims and encounter database. Then, when data quality checks are successfully completed, CMS sends a digital message that essentially says “run risk-adjustment program,” and the insurers’ aggregate answers (but not an individual person’s data) are made available to CMS.

And it worked! On June 30, 2015, after a monumental effort, CMS published the results of the 2014 ACA risk adjustment for all plans in all states for the individual and small-employer markets. It was on time and correctly done. Now, all that remains is to collect the risk-adjustment assessments and make the payments in each state.

The Picture So Far

Risk adjustment for the first full year of the ACA in 2014 is off to a solid start, albeit with some complaints from the “losing” insurers who must pay an assessment. Because collection of encounter data is complex, some insurers are likely to have done a mediocre job in the first year, especially if they had no prior experience or used capitation contracts and didn’t receive diagnostic data for every claim.

Some small insurers complained that they should be exempt from risk adjustment or have very limited assessments because they are small and new to this business. Some of the largest insurers have acknowledged “stumbling” out of the gate with financial losses in 2014 and 2015, but these appear to be more from not understanding the new ACA Exchange markets and mis-pricing within the risk corridor than from anything to do with risk adjustment data collection. In Covered California (the state’s Exchange), nearly all of our plans were profitable in 2014; good outreach and enrollment have contributed to the Exchange’s stability, and risk adjustment has minimized insurers’ concerns about what kinds of consumers they might enroll.

There is no argument that risk-adjustment data collection is hard! But, the consequences of giving only some insurers a “free pass” from this task are huge. The industry is known for finding loopholes in laws and using them to avoid paying a fair share. It would be wrong to allow some insurers to possibly avoid insuring sick people, just because the insurer is new to the business or on the small side.

The U.S. has now joined the European countries (Germany, the Netherlands, and Switzerland) who have successfully operated insurance markets for more than 20 years using risk adjustment. We have not reached “Nirvana” yet — just as the European countries have evolved their systems, we need to strive for continuing improvement (such as revising the payment weights for conditions as new treatments are used).

Andy Slavitt, CMS Acting Administrator, has announced a meeting on March 31 to receive feedback about the current method and to discuss changes that might be made over the next few years. Some of the changes under consideration are discussed in a white paper issued by CMS on March 24. Major questions include: whether to add prescription drugs to the model, how to account for partial year enrollment (due to enrollee churn), whether to pool very high risk enrollees separately, and what kind of recalibration of risk weights to use in 2018 and beyond.

But, keep in mind that we’ve made a great start. At Covered California, we have had success with balancing the risk of our participating plans and kept average statewide premium increases low at 4.2 percent in 2015 and 4.0 percent in 2016.