Late last week, the Center for Medicare and Medicaid Services released the 2018 risk adjustment reports. Risk adjustment in the ACA context moves money on a net neutral basis from insurers with lower than statewide average coded health risk to insurers with higher than average coded health risk. The CMS model uses diagnosis on claims and some prescription data to estimate the incremental extra cost an enrollee “should” have after the model adds in some demographic (age/gender) factors. Each disease category in the individual market has four distinct co-coefficients, one for each metal band. Usually, Bronze plans have a lower risk score than Silver and Gold plans.
Risk adjustment is tough to get right. It involves actuaries projecting first who they think their insurer will be covering for the next benefit year and then projecting what the rest of the state’s market will look like and which carriers get which people at which price point. It is easy to miss on risk adjustment.
And there is a massive miss.
In 2018, Oscar expanded to Tennessee. They entered only the Nashville market where they had a dirt cheap Bronze plan and then priced competitively for Silver and Gold. This mostly worked as they got 20,000+ enrollees. HIOS ID 23552 is now being charged $31,691,661 in risk adjustment outflows. This means that Oscar’s actual experience was that they covered a healthy to very healthy population compared to the rest of the state. Being a net risk adjustment payer is not indicative of profitability or non-profitability; Centene for instance is a constant net risk adjustment payer but they are consistently profitable. I am not surprised that Oscar is paying $103 Per Member Per Month (PMPM) in risk adjustment as they optimized their plan offerings to sell a lot of bronze plans which are attractive to healthier and younger than average individuals. Their marketing is aimed at tech savvy, convenience shoppers who value a seamless experience.
Selling a lot of bronze plans and aiming at younger individuals will produce significant risk adjustment outflows in most “normal” situations. However 2018 was not normal due to CSR termination and the policy responses adopted by many insurers, including Blue Cross and Blue Shield of Tennessee. BCBS-TN both Silverloaded CSR costs and more importantly, they have consistently (since 2017) Silver-gapped the counties where they held a monopoly. In 2017, BCBS priced their benchmark silver plans way above the least expensive Silver plan so that the least expensive silver plan is free for families up to 300% or more of the federal poverty level. BCBS-TN also publicly declared early in the 2018 rate setting cycle that they would be embracing this strategy to the hilt. They also silver loaded CSR cuts so Gold would be a vastly better value proposition in 2018 than it was in 2017 in the BCBS-TN monopoly counties. The rest of the state would be priced to be heavy on silver and gold plans which means that even if there was no difference in inherent health characteristics of the covered population, the rest of the state would have higher risk scores and coded risk merely due to the metal plan selections.
All of that is fine if it is anticipated and priced correctly.
That was not the case, at least it is not the case in the official rate filing documents.
Looking at the 2018 Universal Rate Review Template (URRT) Worksheet 2 for HIOS ID 23552, and looking at column BT PRJ_RISK_ADJ and we see Oscar projected in 2018 that they would have no risk adjustment payments beyond the program operation fee of $0.14 PMPM.
Oscar’s projections for risk adjustment for 2018 were offer by a factor of 700.
That is a modest oopsie.
I’ve never understood Oscar’s path to profitability.
March 2018:
I just don’t understand what Oscar does differently than my former co-workers, besides lose a tremendous amount of money every year. They think they can get their MLR in the high 80s for next year, which is where a well run, lean insurer can profit but they are anything but a skinny insurer. Call me an old fuddy duddy but I am confused as to how they get to profitability with their admin costs as they are even assuming everything else goes right.
I am still stuck there as missing on risk adjustment is understandable, it is a tough thing to get right, but missing by over $100 PMPM takes a special type of disruptive effort especially when the general outline of the state’s markets were being publicly declared during the premium setting process.
On the Road and In Your Backyard
Amir Khalid
I don’t think anyone at Oscar does, either. They’re not the first startup ever to base their profit expectations on fantasy maths.
Mary G
Jared Kushner’s brother Josh is one of the founders and losing money is in their genes?
Barbara
IIRC, you posted on a United affiliate in Georgia that offered what sounded to me like an Oscar model exchange plan in the Atlanta area, and left after either one or two years. In my view, if United sees no short to medium turn path to profitability from a business model, Oscar is definitely going to be challenged, because in addition to their technical ability to accurately forecast risk adjustment into their premium setting, they don’t have the scale to command lower per unit prices on administrative and medical services. Does Oscar really think that other insurers aren’t doing many of the same things they are already? Yeah, I don’t understand it.
David Anderson
@Barbara: Harken was the UHC experiment. And they hit the same problem, providing a seamless experience at a high premium is not particularly attractive to price sensitive buyers who barely touch the system.
I was scratching my head on this problem in 2016 and I’ve worn away the hair above my left temple without a good solution since then.
https://balloon-juice.com/2016/05/19/lifestyle-insurers-in-mco-markets-how-do-they-work/
Barbara
@David Anderson: I wonder if the VC ubermenschen who are funding Oscar think that eventually someone will buy them out. But I can’t think who that would be. Unlike employer provided insurance, an individual open enrollment model allows existing competitors to take your market share without paying for it (other than through the marketing costs they are already spending to attract new enrollment). If Oscar gains traction in specific geographic and business markets, they might be a target for those lines, in those areas, for companies not already there. They would have to be a lot more successful than they currently are, but the analogy is to the many new companies that arose in Medicare Advantage between 2006 and 2010, when Congress began unspiking the punch bowl. Many were built to be sold, and were in fact sold if they had managed to gain a foothold in certain places.
David Anderson
@Barbara: But that only works if and when Oscar is competent enough to price appropriately for as soon as Oscar prices at a level that supports actual costs, their marketshare dives. I could see this strategy make sense in Medicare Advantage where we know that the population is sticky but the ACA individual market is Teflon for the low risk/low cost individuals and the OSCAR price levels don’t see full compensation on risk adjustment if there is a MCO like insurer in the state driving down state average premium.
I am scratching my head.
Barbara
@David Anderson: Oscar would likely have better luck in Medicare Advantage, agreed, especially in certain types of markets, and the duration of its efforts is truly kind of astounding. When the exchanges began in 2014, others (like United) made the mistake of approaching plan design and rate setting in a similar way to MA, and most of those left because as you point out, the enrollment is not similar to Medicare beneficiaries. It is similar to Medicaid, which is why Centene has succeeded where others have not. Other big players have Medicaid Managed Care plans as well, but for some reason they failed to make the connection that Centene did. I hope some of them come back in because I think competition would be good for the exchanges.
David Anderson
@Barbara: Competition on the exchanges is the manuscript my co-author is submitting this afternoon… Competition does weird things will be the abstract.
Barbara
@David Anderson: There needs to be enough of it but it’s never clear what that means. If you followed the hearings on drug pricing, the hearing that included actual individuals who needed expensive drugs eventually included the detail that one woman’s inability to get her drug more cheaply was tied to the fact that nearly every MAPD plan in the part of Alabama where she lived was offered by a single carrier (you already know who it is) and that MA beneficiaries in most other places, especially the Northeast, didn’t have nearly the same challenges she did in getting the drug in question, which she could afford only because she had been approved for a patient assistance plan. I once did a spreadsheet showing size of state population versus number of exchange plans and in most cases, number of plans follows population (overall, of course, not necessarily evenly across the state), but the standouts for bucking this trend were Alabama and North Carolina. Largish states with almost no insurance competition in many parts of the state.
Dan Munro
Uber lost $1.8B in 2018 and they lost $2.2B in 2017.
Losing money is THE business model at this stage of Oscar’s evolution. The ONLY thing they have to prove is high-growth — which is what they ARE trying to do. Only the investors can determine if it’s high enough growth for their level of risk.
This strategy is definitely high risk. Amazon is the preferred model that everyone likes to cite, of course, but at one point (for investor purposes) Uber had a market capitalization of $120 billion. Now that it’s publicly traded, we see Uber’s market capitalization today at $75 billion. Poof. There went $45 billion.
It’s also possible that everyone is tracking Oscar to the fundamentals of a traditional payer – when the real value (and model Oscar is pursuing) is health data – at scale. With a long enough horizon – and enough scale – the data they are collecting will absolutely support a premium acquisition strategy.
The idea that Oscar is somehow “disruptive” to the payer industry is pretty absurd because it’s all dependent on well known (and very well traveled) actuarial math. The “seamless” experience Oscar markets isn’t proprietary or protectable – and I doubt it would capture a high premium if sold as is.
Barbara
@Dan Munro: Every insurer currently in existence is tracking data at scale and has been for a long time. Really, seriously, it is not clear to us how Oscar is different even using that as the end point of its business model, rather than the usual metrics one applies to insurers. United, Anthem, Aetna, Humana, CIGNA and nearly every other entity that processes claims for a living is mining massive amounts of data, using tools they develop on their own and with the assistance of technology companies.
Barbara
@Dan Munro: And just p.s., having an enrollment base that is disproportionately younger (lives don’t revolve around medical issues yet) and willing to fly at the first sign of a price hike (hard to monitor over time) makes Oscar less likely to have the kind of data that yields real information.
sukabi
@Mary G: if that’s the case, the profitability may not be in the actual business, but in the venture capital flowin thru the company and into upper management’s pockets.
Another Scott
BeckersHospitalReview from August 2018:
Maybe they see that they’re doomed unless they do something beyond what they’re doing now?
Dunno.
Cheers,
Scott.
Dan Munro
@Barbara: Yes, but it’s also very different data in a post #ACA / #MA world. The individual (direct) market was almost non-existent before #ACA – and #MA has also mushroomed quickly.
Today, #MA is over 20 million and #ACA is about 12 million (2018). Those numbers (and growth rate) will attract new investors – and new investment thesis.
I’m not a fan of “market” oriented healthcare (because I think healthcare should be a basic human right), but until we end ESI (and other, strictly market-oriented approaches), we will absolutely see risk capital placing lots of different bets.
Barbara
@Dan Munro: Well, MA has been in existence as such since 2006. Prior to that, there were Part C plans and prior to that there was a program broadly referred to as Medicare risk, going all the way back to around 1972. There were a lot of new investors in MA when the program started in 2006, but Congress began slowly deflating the balloon of supercharged funding in around 2010, and it was completed by 2017. There are still new plans starting, and they are more focused on technology.
I am not arguing that Oscar is doomed, just that if its proposition is advancement of data mining and technology, it is the tail wagging the dog, that most insurers are already doing what it claims to be doing, and that its highest value is most likely to technology entities that have realized over time that their lack of access to data is going to hamper their effort to create AI in the health care space. That is, Oscar isn’t building a better mousetrap, but it might be able to provide a missing component for companies that are. Which is no doubt why Google made the investment. Nonetheless, you have to have enrollment in order to generate data. Being run as a gigantic experiment to generate data for Google is not what I deem to be a long term business strategy, but who really knows.
Also, the 12 million people in an ACA exchange plan is dwarfed by close to 66 million people enrolled in Medicaid managed care plans (nearly 11 million in California alone). Just for perspective on the scale of the enterprise.
Mario Schlosser
Mario here, Oscar co-founder and CEO. Thanks for looking at our numbers. David, let me help with your head-scratching. When we go into a new market, we assume we attract average risk, hence the numbers in the 2017 URRT (average risk = $0 RA). Whether we then end up getting more or less risk should not matter, unless you either (a) think the risk adjustment formula doesn’t work and/or (b) we’re not managing the process well. In TN’s case, our $100 RA PMPM was pretty much exactly offset by $100 lower claims than originally priced, and our MLR was in the 80s (just where we priced it!). It doesn’t make sense to just look at one of those things, as you probably know. Don’t just take my word for it, look at Health Supp 1 of our annual NAIC TN filing (we file that before this final report), or cut that against quarterly filed net premiums (which have our RA projections throughout 2018 in them; when we have plan mix and members onboard, we can start accurately projecting RA early in the year). In other states we entered in 2018 it was the opposite (higher risk than average, claims higher, RA receiver – similar MLR). So, the comparison you’re making is fun but pointless. Unless you want to debate me on the assumption of average market risk in a new market, which would be a great discussion for another day. Doesn’t mean risk adjustment works as well as it should btw, but that’s also a different discussion.
David Anderson
@Mario Schlosser: Thanks for the comment:
I have a couple of things to lay out on my thinking on OSCAR:
1) I hope you and/or your competitors succeed with success defined as getting customer friendly insurance that offers great value to people at a premium that is reasonable to the customer and reasonably profitable to you — I hope that is true for all of the start-ups out there that are trying some variant of a tech enabled health insurance play that is not a pure risk skimming play. If you guys can crack that, it unleashes great social value.
2) Regarding #1, insurance is tough as you well know and the ACA market is probably the toughest market to be in given risk adjustment and policy risk. So have fun storming the castle.
3) Now onto the technical question of should a new entry price for average state wide risk?
In my opinion, that would have been a reasonable thing to do in 2014-2015, a flip a coin decision in 2016 and a bad decision by 2017 going forward. There are a few strands of data/evidence that, in my opinion suggest that pricing for state wide risk by the 2017 market year, and especially for the 2018 policy year, that this is a sub-optimal decision.
First, Oscar’s own pricing for the 2018 market in Nashville strongly suggested that Oscar would dominate the over 201% FPL price sensitive/low morbidity Bronze market as you priced your least expensive Bronze to be $233 below benchmark for a single 40 year old non-smoker. Your pricing also indicates that OSCAR would be at a price disadvantage for the most price sensitive Silver CSR 94 buyers as CIGNA would be able to offer near zero premium plans to folks at 137% FPL. Oscar’s gold plans were priced fairly high so gold would be rare. Oscar would be predictably heavy on Bronze.
The silver spreads (cheapest silver to benchmark) were better for subsidized buyers in 2018 than 2017 but compared to the rest of the state, subsidized silver buyers were still seeing more expensive minimum silver premiums in the Nashville region. The rating region was priced to be heavy Bronze and that was being driven by Oscar’s good Bronze pricing for subsidized buyers and mediocre pricing on Silver. And this is what happened: Tennessee had a 25% Bronze uptake for OEP 2018 On-Exchange. Of the counties OSCAR was in for 2018, Trousdale County had 30% uptake in Bronze and it went to 50% Bronze in Williamson County. The Nashville Rating Region was heavy Bronze and light Silver.
We further know that BCBS-TN in the summer of 2017 loudly declared that they would be hiking premiums and doubling down on their silver spread strategies and a full silver load in the monopoly areas of the state. These strategies push effective net of subsidy premiums down significantly. Significant parts of the state were seeing BCBS-TN spreads between the cheapest Silver to Benchmark getting withing $40 PMPM of the OSCAR Bronze-Benchmark spread in Nashville. Zero or near zero premium Silver plans are more attractive than zero or near zero Bronze plans. The other competitive markets saw wider cheapest Silver to benchmark spreads than greater Nashville and smaller cheapest gold to benchmark spreads as well. If we are to assume perfectly uniform health risk distribution in the state, we should also expect that the insurer that is heavy Bronze compared to the rest of the state to pay into risk adjustment. That is just a fact of the formula interacting with strategy. And there is nothing wrong with that strategy.
There is a problem with the assumption of uniform health distribution. We know there is a significant urban/rural morbidity divide with urban areas being less morbid than rural areas without even taking into consideration that Appalachia is its own unique health risk case.
Here is a relevant January 2017 report from the CDC: https://www.cdc.gov/media/releases/2017/p0112-rural-death-risk.html
Being in the greater Nashville area means that the rest of the state compared to the Oscar market region is heavy on the rural and light on urban populations. All else being equal, that would indicate the rest of the state will have a higher disease burden.
Finally, OSCAR in 2018 was operating under an information disadvantage for risk adjustment optimization as you were a new entry. Even assuming your HCC optimization team is best in class, they were working on comparatively thin data sets to identify missing HCC coding opportunities. BCBS-TN has massive data sets that were very broad and deep in 2018. CIGNA should also have had a similar data advantage. I would expect that insurers with deeper and broader data sets should be able to find and legitimately score more HCC points from an identical population than teams working with thin data universes.
So the combination of Oscar pricing to be bronze heavy relative to the rest of the state, selling only in a single rating area that is again is heavily urban and working with thinner and narrower data sets for 2018 risk adjustment optimization purposes are the three reasons why I think assuming state average risk is at best a questionable decision for any new insurer entering a new region post 1/2/16 with a Bronze heavy pricing strategy.
Jay
@David Anderson:
David, good article and insights but isn’t the purpose of risk adjustment to normalize costs among payers with different morbidity at the rating area level? Essentially, a payer heavy on healthy bronze in the Nashville rating area would be compared against other payers of same rating area, so it doesn’t matter if Nashville is healthier than rural rating areas.
Furthermore, can you elaborate on why pricing for average market risk isn’t optimal after 2017? Isn’t $300 in claims and $100 risk adj payment essentially the same as $400 in claims and no risk adj, if you believe that the market average is $400?
David Anderson
@Jay: The risk adjustment system for the ACA is normalized against state average premium. That is why Nashville being heavy on urban and light on rural relative to the rest of the state matters. Different risk adjustment systems would change that salience.