2024 Risk Adjustment Report
Last month, CMS issued the 2024 final risk adjustment report, which shows how much each carrier paid or received from the Affordable Care Act’s risk adjustment program. In our blog series last year, we went in depth on why the risk adjustment transfer numbers in isolation don’t tell you very much, and we also summarized the data at the parent level as well as made estimates as to the total member months involved with carrier to try to contextualize all of this data. We’ve put together a similar summary this year which we’ve posted below.
We’ve done our best to validate this data, but of course, mistakes are possible and we’d love any feedback or issues you may notice.
The comparison of estimates to actuals showed a not-surprising pattern: carriers tended to over-estimate their payable (or underestimate their receivable) at year end. This makes sense: carriers want to be conservative about their estimates – regulators looking at their financials want to be sure carriers have enough assets to continue to pay the claims they have promised to pay, and having a big nasty surprise in June from risk adjustment undermines regulators confidence in a carrier and may lead to some unwelcome scrutiny. In aggregate, carriers over-estimated their accruals by about $1.8B, almost 20% of the total amount of the $10.4B aggregate payments in 2024.
There weren’t many huge surprises this year, and in many ways, it may be “the calm before the storm.” Centene’s market share was the largest by far, with nearly 20% of the reported individual market. Their population was also higher risk relative to the market than in prior years, with a PMPM receivable of around $10, versus roughly breakeven last year. They conservatively accrued their receivable at year end, and in fact, all of the top ten largest carriers were either over-accrued or only under-accrued by a few dollars PMPM – the largest carrier with a significant miss looked to be BCBS of North Carolina at around a $21 PMPM miss. All of this to say, it seems like carriers have gotten the hang of estimating risk adjustment, which made some news that dropped last week all the more surprising:
When the risk adjustment report first dropped on June 30, Nona Tepper from Modern Healthcare called and talked with me about the report. She also asked me about rumors of some impending 2025 bad news about risk adjustment, which I hadn’t heard of and shared my skepticism. But the next day, Centene issued a press release withdrawing its 2025 earnings guidance. Nona has been working the ACA beat exhaustively over the years and I should have known better than to doubt her! But the press release was quite specific about why they were withdrawing their guidance, and I thought it was worth a deep dive on what might be going on. To do so, we’ll take a brief tour through a fascinating and brilliant piece of ACA history from the consulting firm Wakely and their product: WNRAR (pronounced Win-rar)
WNRAR
If you’re like me and were a nerd in the 1990s and early 2000s, when you hear WNRAR, you think of the file compression program that was all the rage back in the day. But there’s another WNRAR that’s critical to how many ACA plan estimate their risk adjustment receivables and payables: the Wakely National Risk Adjustment Reporting (WNRAR) study.
Wakely is an actuarial firm with roots going back to 1967. At the time when the ACA was being implemented, Wakely was a relatively small firm with around 25 actuaries. They had a few clients planning to enter the ACA market, and all the carriers trying to enter that market had the same problem: projecting risk adjustment payables and receivables is a bigchallenge.
Calculating the carrier’s own risk score was hard enough – but even if they successfully interpreted the emerging guidance on calculating risk scores, a carrier’s own risk score doesn’t tell them anything about whether they’ll be a payer or receiver of risk adjustment.
So in 2012, Wakely began talking to carriers across the country and asked if they’d send in anonymized data about their members’ risk. If they could get enough carriers in a market to give them data, they would simulate the risk adjustment program in that state and provide an estimate for how much the carrier should expect to pay or receive. They were quite successful and today they run the WNRAR study in 35 states. In states without WNRAR, carriers have to make educated guesses about what changes may be happening in their market so they can estimate their risk adjustment transfers – because the only indication they’ll get from CMS is an interim report in March of the next year, and then the final report in June drops. A lot can change in a market from one year to the next, and a big shock in June could mean that you’re halfway through rate filing season and suddenly learn that last year you lost money when you thought you were profitable. So, you can see why it’s so valuable (and why it’s helped propel Wakely through a lot of growth). If you really want to dig into the weeds, the participation guide and other information is on a Wakely web site here:
https://wramodel.com/wakelyraf/#overview
Wakely releases results several times a year – the first round of results for a plan year tends to come out in mid-June and is based on claims paid and incurred through April of that year. This Wakely report is the first insight health plans get about the current year statewide average risk scores. This is where there seems to have been a surprise for some healthplans in 2025. Even with all the headwinds in the ACA market, it still grew year over year in 2025. There’s some question about whether some of this enrollment is fraudulent, but in general, membership growth has meant that average risk has decreased. This makes sense generally, because people joining the market in the last several years have been enticed by richer subsidies, and you’d generally expect that a marginal buyer only purchasing coverage because it’s more highly subsidized is less likely to have a lot of health conditions.
What Centene reported, though, is that risk scores in states where Wakely reports results was “significantly higher than, and materially inconsistent with” their expectations. In other words, they expected that overall health of the members enrolling in the ACA market would be fairly stable, but in fact, it got worse. So, if Centene had been seeing their risk scores get worse, they assumed that the market wasn’t also getting worse at the same rate – so if they had a higher risk population, they’d get compensated for it by increased risk adjustment transfers. If this isn’t happening, then everyone’s risk is going up which means if a health plan didn’t raise its rates much in 2025, they’ll have a problem. This sent the market into a panic, and their stock has lost 45% of its value in the last month. Molina also modified their earnings guidance downward.
There’s a risk of a perfect storm: the fraudulent enrollment problem would mean that potentially millions of “fake” members are disenrolling as they report the fraud, and someone who was enrolled without their consent by definition had no claims. On top of that, there’s a risk of induced demand from the people who remain: if they think their health plan is going to cost a lot more next year due to enhanced subsidies expiring, then they may try to get a surgery or procedure they’ve been putting off done now. On top of that, with the Medicaid PHE unwinding more or less complete, the people who lost Medicaid and switched to the Exchanges may have been the higher risk members. If everyone underpriced for this scenario, it could be catastrophic for the market in 2025. So if you lose a huge chunk of the population that didn’t have a lot of claims, but the population that remains is costing a lot more, then the market will have dramatically underpriced, and risk adjustment can’t fix market underpricing.
However, I remain a bit skeptical that this portends doom for the entire market, and think that other health plans shouldn’t panic too much (yet). A few reasons:
Based on our analysis of the Unified Rate Review filings for 2025, Centene had the smallest weighted average increase in their rates in 2025. On average, carriers raised rates by about 6% in 2025, but Centene only raised theirs by 3%, as did Molina. By contrast, Oscar, United, Elevance, and Cigna all raised rates by 6-7%. Part of the story may be that other carriers priced for some higher market risk while Centene didn’t, and the WNRAR report made it clear that risk adjustment wouldn’t save them from underpricing.
The 2024 risk adjustment report showed an interesting trend: there were actually more members with HCCs (risk adjustable conditions) in 2024 compared to prior years. It wasn’t a huge increase, but given that the market grew meaningfully in 2024 versus 2023, so you’d expect if a lot of the 2024 enrollment growth was fraud, then the number of people with risk adjustable conditions would go down. Yesterday, CMS added a few interesting appendixes to the risk adjustment report, including one that breaks this data down to the state level. Without prior year data, it’s hard to know what state level variations are persistent, and the last time CMS posted an appendix like this was in 2020 for plan years 2017-2019, which was before enhanced APTCs and the market was less than half of the reported size from 2024.
However, if millions were enrolled fraudulently, you’d expect the states where fraud is occurring to be substantially different than those where it’s less prominent. If we compare the percentage of members without HCCs in 2019 to those reported in 2024, this number overall is down from 22.9% to 21.9%. If we assume the overall decrease was due to the market growth due to APTC enhancements attracting lower risk individuals, then the state level variations might be explained by fraud patterns. Using this assumption, we end up with potentially 1.5M fraudulent members across FFM states which did not expand Medicaid, with half coming from Texas. This is meaningful to be sure, but not the 15-20% of the market some are claiming. This will have an outsized effect on Centene who has a lot of membership in those markets, but it doesn’t necessarily portend a market in a death spiral.
We’ll share this analysis in full in the coming weeks, but the story around fraud is more nuanced than one might think.The Wakely report Centene references only uses 4 months of claims data, with no runout. And there’s reason to believe that this year’s claims completion patterns may be different than last year’s: the number of people passively enrolling or actively renewing but not switching plans (and thus remaining on the same carrier) is up meaningfully from 2025 – from 45% in 2024 to 60% in 2025. This means that people aren’t having to get new prior auths for drugs and procedures, find new doctors, etc. before accessing care. It also means health plans know a lot about their members already compared to a brand new member. So, risk scores may be higher, but it might be because the patterns around claims submission and completion are different this year.
Of course, all this is speculative… and I’m anxious to see the earnings of Oscar for Q2. Health plans should be paying attention – but they shouldn’t be panicking unless their own data is also showing some alarming patterns. And of course, you can reach out to us if you want help analyzing your circumstances – there’s tons of uncertainty going into next year and even if 2025 shouldn’t alarm you, 2026 is just around the corner and the decisions you make in the next month will be ones you have to live with through the end of 2026.
2024 Risk Adjustment Transfer Data
About this data: We took the data from CMS’s 2024 risk adjustment report and added additional data as it was available. This is similar to our report from last year, but we added some new fields: namely premium data and data about what carriers had accrued at year end. Premium and enrollment data was sourced from the Unified Rate Review Public Use File where available (using the “current enrollment” and “current premium” fields and using an annualization factor of 12), with the Supplemental Health Care Exhibit (SHCE) state level data used when URR data could not be found or when the URR data appeared non-credible (e.g. was wildly divergent from SHCE data and led to state level totals well outside the amount reported by CMS). The totals don’t match 100% to what was reported from CMS state level data; this is expected as we won’t have issuer level enrollment data at the HIOS ID level until the URR PUF for 2026 comes out, and that won’t include issuers exiting the market – the approximations we use will have some noise such as the inclusion of transitional and grandfathered membership in SHCE data that is not subject to risk pool requirements.
We used the amount receivable (payable) booked at year end for each carrier from the SHCE as well, and then compared that to the final receivable (payable) shown in the CMS report to highlight any carriers who had payments materially divergent from what they had estimated on their year-end financials. We included the High Cost Risk Pool (HCRP) receivable in the comparison, since this should be reported as an offset to the payable or an addition to the receivable. Because California’s financial reporting system for carriers doesn’t use standard NAIC filings, they aren’t included here (and anything relying on the SHCE had to be omitted since that is not a required CA filing).
For data sourced from the SHCE, we mapped carrier HIOS IDs to NAIC company codes and states using a data set we’ve developed over the years which has relied on various sources (rate filings, MLR PUFs, etc.) to create as comprehensive a mapping as possible between carrier subsidiaries and their parents. We also made some assumptions when something seemed way off – for example, on their SHCE, Baylor Scott & White reported a $125M payable for their small group block and a $3.8M receivable for their individual block, while their final risk adjustment results were essentially the reverse of this. We assumed this to be an error on the SHCE and adjusted the data accordingly.