Each June, health insurers send the government detailed cost forecasts for providing prescription-drug benefits to more than 40 million people on Medicare.
No one expects the estimates to be spot on. After all, it is a tall order to predict the exact drug spending for the following year of the thousands of members in each plan.
However, year after year, most of those estimates have turned out to be wrong in the particular way that, thanks to Medicare’s arcane payment rules, results in more revenue for the health insurers, a Wall Street Journal investigation has found. As a consequence, the insurers kept $9.1 billion more in taxpayer funds than they would have had their estimates been accurate from 2006 to 2015, according to Medicare data obtained by the Journal.
Those payments have largely been hidden from view since Medicare’s prescription-drug program was launched more than a decade ago, and are an example of how the secrecy of the $3.5 trillion U.S. health-care system promotes and obscures higher spending.
Overdoing It
Health insurers reaped $9 billion in additional revenue from 2006 to 2015 by overestimating costs to Medicare and keeping a share of the extra money.
Medicare’s prescription-drug benefit, called Part D, was designed to help hold down drug costs by having insurers manage the coverage efficiently. Instead, Part D spending has accelerated faster than all other components of Medicare in recent years, rising 49% from $62.9 billion in 2010 to $93.8 billion in 2017. Medicare experts say the program’s design is contributing to that increase. Total spending for Part D from 2006 to 2015 was about $652 billion.
The cornerstone of Part D is a system in which private insurers such as CVS Health Corp., UnitedHealth Group Inc. and Humana Inc. submit “bids” estimating how much it will cost them to provide the benefit. The bids include their own profits and administrative costs for each plan. Then Medicare uses the estimates to make monthly payments to the plans.
After the year ends, Medicare compares the plans’ bids to the actual spending. If the insurer overestimated its costs, it pockets a chunk of the extra money it received from Medicare—sometimes all of it—and this can often translate into more profit for the insurer, in addition to the profit built into the approved bid. If the extra money is greater than 5% of the insurer’s original bid, it has to pay some of it back to Medicare.
For instance, in 2015, insurers overestimated costs by about $2.2 billion, and kept about $1.06 billion of it after paying back $1.1 billion to the government, according to the data reviewed by the Journal.
The companies say the pattern of overestimates the Journal identified is the result of unpredictable drug pricing by pharmaceutical manufacturers. Insurers say they strive to make accurate bids, which are certified by professional actuaries. Medicare reviews and approves all bids, and sometimes audits them, insurers also point out.
Some insurers also say some members purchase plans with so-called supplemental benefits, richer coverage than what Medicare will pay for. Under Medicare’s rules, the impact of that extra spending can lead to bigger discrepancies in the estimates, they say.
A CVS official said it is to be expected that companies would be “biased” toward overestimating costs “because we can’t have years where we lose money.”
In a statement, the company stressed it strives for accuracy when reporting to the Centers for Medicare and Medicaid Services, or CMS, the agency overseeing Part D. “We take great care to develop and submit the most accurate bid possible to CMS in order to have the lowest possible premium,” a CVS spokeswoman said in an email.
A Humana spokesman said in an email that the company “has not engineered its bids in such a way as to produce additional profit.”
“The Part D program provides a stable and sustainable way for seniors to obtain their medications at an affordable price, and the private sector innovation and competition it fosters deliver great value to the government and taxpayers,” a UnitedHealth spokesman said in an email.
CMS said in a statement that risk sharing—the part of the program that can reward overestimates—is “based on a statutory formula,” and that it has “taken action to bolster Part D plans’ negotiating power so that they can get the best deal for patients from prescription-drug manufacturers.”
CMS declined to release data to the Journal that would show individual insurers’ bidding patterns, saying it is confidential industry information. Separately, the Journal obtained the insurers’ data for a five-year period from 2009 to 2013 that shows the patterns of the more than 6,000 Part D plans sold during those years.
The Journal also analyzed aggregate payment data from 2006 to 2015 that has either been published by CMS or obtained via public-records requests.
A detailed analysis of the confidential industry data obtained by the Journal shows that 69% of Part D members from 2009 to 2013 were in plans that overestimated costs by at least 5% over that time. Among many of the largest insurers, the rates were higher. UnitedHealth overestimated costs by at least 5% for plans covering 93% of its members in the analysis; Anthem Inc. did so for plans covering 76% of members; and Humana did so for plans covering 74% of members.
Anthem said in a statement, “we are confident in our compliance track record.”
The Journal analysis of 2009-13 data excluded some plans that aren’t subject to normal Part D rules, including Medicare-subsidized plans backed by employers and some so-called demonstration plans, which are pilot projects that test out new ways of delivering or paying for care. Those plans are included in the 2006-15 aggregate records, however.
If those big insurers were aiming to submit accurate bids, the probability that they would have overestimated costs so frequently and by such a large amount is less than one in one million, according to a statistical analysis done for the Journal by researchers at Memorial Sloan Kettering Cancer Center, who study pharmaceutical pricing and reimbursement.
Insurance companies use heaps of data to predict future spending. If truly unpredictable events were blowing up their statistical models, the proportion of overestimates to underestimates would be closer to 50/50, says Peter Bach, director of Sloan Kettering’s Center for Health Policy and Outcomes, which conducted the statistical analysis.
“Even expert dart throwers don’t hit the bull’s-eye every time. But their misses are spread around in every direction,” says Dr. Bach. “If they start missing in one particular direction over and over they are doing it on purpose.”
The Medicare Payment Advisory Commission, an independent congressional agency, wrote in 2015 that insurers’ inaccurate estimates “show consistent patterns rather than the randomness one might expect from projection errors in the actuarial assumptions behind bids.”
The Part D drug benefit was passed into law in 2003 and launched in 2006. Congress designed the program so that it would be paid for by taxpayers and the seniors who enroll in it, but operated entirely by private-sector companies. The private health insurers that run Part D negotiate drug prices with pharmaceutical companies and compete to sell coverage to seniors. For standard drug coverage under the program, Medicare pays three-quarters of the costs and seniors who join pay monthly premiums covering the remaining roughly 25%.
That market-based design was supposed to reward insurers that hold drug spending low with greater market share, and penalize those whose costs come in higher than budgeted. The program’s designers expected insurers to keep their total bids as low as possible in order to keep premiums low. Big overestimates, they thought, would deter cost-conscious customers by leading to higher premiums.
But, a wrinkle in the program’s design allows plans to reap higher revenue from overbidding—while also keeping premiums low.
Insurers’ bids are split into two separate cost estimates. The two-guess system was designed by legislators who worried private companies wouldn’t participate in their new insurance market. To entice them, Congress sought to stem the risk of big losses if they made miscalculations. For one chunk of costs, called the “direct subsidy,” insurers and Medicare share the risk based on a complicated formula. For the other chunk, called the “reinsurance subsidy,” Medicare bears it all.
It is with the direct subsidy, which mostly covers routine costs, that overestimates can pay off. For starters, insurers that overestimate costs keep any extra money up to 5% of their guess, and some of money beyond that 5%, based on a Medicare formula. A high guess here leaves insurers with extra money. Those that underestimate can face losses of up to 5%; Medicare steps in to partially offset losses over that initial 5%.
With the reinsurance subsidy, which covers the government’s share of spending on a subset of patients with extremely high drug costs, Medicare bears all the risk for a wrong guess. Plans that overestimate pay all the extra money back, and ones that underestimate get fully reimbursed by Medicare. Since this guess is also part of the total cost estimate that member premiums are based on, a low guess here can help hold premiums down with no risk to the insurer.
The confidential data the Journal reviewed shows plans across the board guess wrong in the way that most benefits them in both cases. Around 46% of all Part D members in the Journal’s analysis were in plans that overestimated the first guess and underestimated the second one from 2009 to 2013. By contrast, only about 1% of members were in plans that did the opposite.
Across all insurers, direct subsidy estimates were high by $17.6 billion, and plans kept an extra $9.1 billion from 2006 to 2015, according to the Journal’s analysis of CMS data published online and obtained via a public-records request.
Reinsurance subsidy estimates across all insurers were low by $27.8 billion, and Medicare paid the plans back all of it, according to public CMS data.
Companies said they frequently underestimated reinsurance costs because of a steady rise in the use of expensive medicines for chronic conditions like rheumatoid arthritis. They also said the aggregate public data includes some employer-backed plans that are subject to different rules and inflate the total reinsurance pay-back figure.
“[P]lan sponsors have been able to keep part of catastrophic benefit spending out of enrollee premiums and receive the full reinsurance amounts due to them,” according to MedPac, the independent congressional agency. “Even though the plan must return some of its [direct subsidy] profit to Medicare…it still nets a portion of profits,” MedPac wrote in the same 2015 report to Congress.
Some of the largest Part D insurers had higher-than-average rates of estimating wrongly in both areas, despite ostensibly having the most data and resources to accurately forecast spending. Companies with the most market information should, in theory, have an advantage in predicting costs and constructing bids, according to economists and actuaries.
UnitedHealth simultaneously overestimated its direct subsidy costs and underestimated the government’s reinsurance subsidy, triggering extra revenue for the company and extra costs for the government, in plans covering more than half of its membership in the analysis, which averaged 6.3 million annually from 2009 to 2013. The proportion of members in plans where UnitedHealth got it wrong the other way, by underestimating the direct subsidy and overestimating the reinsurance subsidy, was less than a thousandth of 1%.
Humana bid inaccurately in the first way in plans covering 47% of its membership in the analysis, while bidding the opposite way in plans with just 0.1% of its members.
CVS—which covered about 3.4 million members in its Part D plans included in the analysis at the end of this five-year period—overestimated its direct subsidy costs and underestimated the government’s reinsurance subsidy across 63% of its membership, the Journal’s analysis found. Just 0.5% of its members were in plans where CVS did the opposite.
Across all of CVS’s plans from a longer time period, the company kept extra money from the direct subsidy payments in five of seven years from 2009 to 2015, and received an average of $513.9 million annually in extra reinsurance payments, according to data provided by CVS to the Journal for years that haven’t yet been made public by CMS, and the data separately obtained by the Journal.
The data show that CVS kept extra money from the direct subsidy again in 2016, but that its 2017 bid underestimated the direct subsidy costs. CVS overestimated its reinsurance costs in both of these years and as a result paid back the government money in each year.
CMS, the Medicare agency, said recent data “suggest that, on average, plans’ estimates of future costs in their bids are closer to their actual costs, resulting in a significant decline in revenue retained by Part D plans.”
UnitedHealth, Humana and CVS combined covered nearly half of all Part D members during the 2009-13 period.
The Obama and Trump administrations have both proposed slashing the reinsurance subsidy and increasing plans’ risk for patients with very high drug costs. The proposal, also advocated by MedPac, has some industry support. CVS said it backs the change.
The change can be made only by an act of Congress, and legislators aren’t actively considering it.