In an old joke, a man sprinkles a magic powder to ward off elephants. “But there are no elephants around here!” protests a skeptic.
“You see?” says the man. “It works!”
In one sense, the funding of Medicare Advantage (MA) clearly works. Insurers are flocking to offer these privately managed plans, through which a growing share of Medicare enrollees—in 2018, it was 34%—choose to get their benefits. When companies contract with CMS to offer an MA plan, they receive monthly capitated payments for enrollees, which must cover all Part A and Part B benefits except hospice, and, in most cases, a separate payment for Part D prescription drug benefit as well. But the calculation of these payments raises several questions, and one big one echoes singer Avril Lavigne in a 2002 pop hit.
She didn’t inquire—although one might—whether MA payment creates a true “level playing field” with government-run, fee-for-service Medicare. And she didn’t ask if the way MA plans are paid is the best way to incentivize quality and the “value-over-volume” transition that U.S. health care is widely believed to need. Lavigne sang simply—and plaintively: “Why do you have to go and make things so complicated?”
Of course, it’s complicated partly because it’s tough. “Figuring out how to pay the plans accurately and fairly has posed a persistent challenge,” wrote Richard Kronick, former director of the Agency for Healthcare Research and Quality (AHRQ), in a February 2017 Health Affairs blog. He went on to warn that because the MA payment system over-rewards plans for the medical risks their enrollees face, “unless there is a further policy response, Medicare will substantially overpay MA plans over the coming decade—likely to the tune of hundreds of billions of dollars.”
Source: Kaiser Family Foundation analysis of CMS state/county market penetration files, 2018
A different assessment came on the same blog early last year from Robert E. Moffit of the conservative Heritage Foundation and coauthors Rita E. Numerof and Christen M. Buseman. Noting that plans are paid a capitated rate tied to CMS quality measures (and conceding that the system isn’t perfect), they stated approvingly that MA plans “leverage the business-rich experience and care management capabilities of commercial payers and have demonstrated, to a large degree, the special advantages of the private-sector market response to patient wants and needs.”
“Business-rich,” maybe. But truly competitive? Last May, Steven Lieberman, a nonresident fellow at the Brookings Institution, weighs in with five coauthors, once again on the Health Affairs blog, with a negative on that one. “Research suggests that the current MA bidding structure does not strongly promote competition, allowing plans to bid above their true costs and to retain rebate dollars instead of fully pass[ing] savings through to beneficiaries,” they wrote. The rub, they argued, is that complicated payment system.
So let’s take a look at how plans are paid.
The short answer—and it is not all that short—is that CMS establishes benchmark monthly amounts for each county based on fee-for-service Medicare spending, and those benchmarks guide insurers in submitting “bids,” which CMS then customarily accepts. “Bids” is in quotation marks because, as MA critics note, the system is not one of true competitive bidding with some plan (or plans) getting the contract and others not. In the MA world, a bid includes the benchmark amount and the plans’ administrative expenses and profit margin, which can’t exceed 15% of total plan revenues. When a bid exceeds the plan’s risk-adjusted benchmark, an extra premium is charged to enrollees. When it’s lower than the benchmark, as is more often the case, the plan is paid not only the amount of its bid, but also a share (from 50% to 70%, depending on quality ratings) of the difference between the benchmark and the bid. That amount is known as a rebate, and the plan is required to use it—after deducting small percentages for administration and profit—either to provide additional services to enrollees or to lower their premiums.
But almost every word in our paragraph of MA 101 is a metaphorical rock which, if upended, reveals a swarm of wriggling complications. In fact, it speaks volumes—and not very reassuring ones—that the way this country pays for its best stab at cost-effective care for seniors is an ever-evolving and almost impenetrable agglomeration of legislation, regulation, market behavior, and accumulated custom. With benchmarks, bids, and rebates, the three basic units in this payment process, however, there’s also a toolbox of modifying factors that, examined one by one, shed further light on the details, which—as usual—are where the devils reside.
Quality star ratings
It makes sense in this value-focused era that MA plan payment should be influenced by the quality of care and service, and it is—with some limitations, as we’ll see. Plans are assessed with a quality rating of from one to five stars (five is tops) based on nearly 50 measures that include how well they provide screenings, immunizations, and checkups; help members manage chronic conditions; and satisfy customers and respond to their calls and questions. Data come from member surveys, reports by plans, and reviews of billing, among other sources. For plans’ Part D services, a separate rating is based on drug safety, accuracy of drug pricing, and member satisfaction, and the two ratings are combined to create an overall plan rating.
Plans with four- or five-star ratings are awarded quality bonuses that nominally are supposed to boost benchmarks by 5%. In actual practice, for 2018 they added 4% to the average plan’s benchmark and 3% to plan payments, according to the Medicare Payment Advisory Commission (MedPAC), a nonpartisan agency that advises Congress on Medicare. Two important details about quality bonuses: First, in roughly half of U.S. counties, benchmarks are capped at pre-ACA levels, which may block some or all of a quality bonus from applying, and second, bonuses may be doubled in certain urban counties with low fee-for-service spending and historically high MA penetration. MedPAC has called for an end to that practice.
By the time a quality rating goes into the baker’s oven of MA plan payment, it’s pretty stale dough. As Catherine M. Murphy-Barron, principal and consulting actuary for the consulting firm Milliman, explains, the star ratings that will trigger quality bonuses and inform consumers’ choices for 2020 coverage will be based on ratings issued in October 2018—“and most of the data for that year’s star rating was collected in 2016.”
Another problem with the use of quality star ratings, noted by MedPAC in its annual report to Congress, is “contract cross-walking.” In this practice, a company operating in geographically disparate areas combines far-flung units under a single contract so they can all enjoy bonus payments earned by one unit’s high quality rating—even when the other units thereby included haven’t merited such a grade. The commission cites one company that used such contracting to piggyback its lower-rated 300,000-member plan in the South onto a four-star, 20,000-member regional plan in the Northeast so that the big Southern plan could garner bonus payments it had not earned. MedPAC has long recommended that CMS fix this problem, but it hasn’t yet.
Benchmarks don’t just happen. You might think they’d simply be last year’s Medicare FFS spending in the county, right? Wrong! In the lullaby prose of the Congressional Research Service (CRS), “to project per capita FFS spending in each county, the Secretary [of HHS] calculates historic spending data from original Medicare claims and estimates a trend to determine the growth” in FFS Medicare. Then a national FFS per capita monthly spending estimate (for 2018, this was $848.21) is multiplied by a county-level index called the “average geographic adjustment” to “determine the relative difference in per capita spending in each county.”
At least in recent years, several other adjustments have then been made “which are either specified in the statutes or made at the Secretary’s discretion,” as CRS reports. They’ve included tweaks to reflect price inflation, changes in growth-estimate methodology mandated by the ACA, a subtraction representing reduced Medicare spending for care veterans get from the VA instead of their plan, and a phaseout of what amounted to a double compensation for hospitals’ indirect medical education.
The quartile system
The MA game can’t be played without players, and the ACA’s drafters wanted to be sure insurers would suit up and come onto the MA field—even in remote rural areas. They also wanted to push most medical costs in the direction consensus says they need to go: downward. So they devised a system in which the nation’s 3,142 counties (or county equivalents, where parishes, boroughs, etc., predominate instead of counties) are ranked according to total Medicare FFS spending. That ranked list is divided into quartiles, and benchmarks are adjusted by a factor that corrects somewhat for geographic disparities.
For the lowest-spending counties, benchmarks are multiplied by 115% “to help attract plans,” as MedPAC writes. In the next-lowest quartile the factor is 107.5%, then it’s 100%. The biggest-spending quartile’s benchmarks are multiplied by 95%, again in MedPAC’s words, “to generate Medicare savings.”
“In remote areas, there might not be any Medicare Advantage plans available if we didn’t overpay them,” comments Laura Skopec of the Urban Institute, who says that the fact that the percentage downside is smaller than the upside is neatly balanced by the distribution of enrollees. “The vast majority of people live in 95% counties,” she says. “And this 95%-to-115% spread has the magical property of getting payments on average nationally to about 100% of traditional Medicare.”
Local MA plans that operate in more than one county use a benchmark that’s averaged across the counties it served, weighted by enrollment.
Did we say “local” plans? There’s another kind too. Unlike local plans—which may be HMOs (as most are), PPOs, special-needs plans, or private FFS plans—regional plans must be PPOs, and each must serve all of one of 26 regions established by CMS. Calculating benchmarks for regional plans, confesses the advocacy group Better Medicare Alliance in deadpan printed prose, “is a bit more complex.” Basically, these regional benchmarks are a weighted average of two things: (1) the average county-level benchmark, weighted by the number of Medicare beneficiaries in each county and the national share of beneficiaries in FFS Medicare, and (2) the average plan bid, weighted by each plan’s projected enrollment and the national percentage of beneficiaries in MA.
Enrollees in bad health, of course, cost more than those in good health. And, as Kronick, the former head of AHRQ notes, plans are rewarded (over-rewarded, he believes) for higher-risk patients. Based on ICD-10 diagnostic codes from patient encounters submitted by MA plans and using a model called Hierarchic Condition Categories, CMS pools data in diagnostic buckets it terms DxGroups and combines these in categories, announcing annually the coefficients by which the various condition categories are expected to increase costs. This becomes the basis for each enrollee’s risk score. A risk score of 1.0 means the member is expected to generate roughly the average level of spending in the coming year, while a 2.0 risk score anticipates twice the average cost, doubling the payment.
MedPAC’s March 2018 report to Congress gives the example of an 84-year-old male in a certain county who’s not eligible for Medicaid. An annual capitated payment for his care would be $5,707, it reports. For an otherwise equivalent man with diabetes the heightened risk score would turn that into $6,765. And what about one with vascular disease as well? The plan would get $9,796 for that gentleman’s care.
Coding intensity adjustments
But FFS Medicare will always be an imperfect template for MA plan payment. That’s because MA plans have an incentive FFS lacks to identify risk aggressively—after all, it affects how much they’re paid. FFS providers mostly code for procedures performed. An enrollee’s diabetes that triggered a visit last year may not be reflected in coding the treatment of this year’s broken leg. But MA plans go to great lengths to identify every possible diagnosis. Through medical record reviews, home-visit health risk assessments, widely promoted annual checkups, and shared risk programs with providers (or even what MedPAC calls “pay-for-coding” programs), plans collect diagnoses the way New York City’s homeless population collects recyclable cans in the street.
Of course, that might entitle them to a halo rather than horns—such identification of risk can, theoretically, enable the follow-up and robust condition management that proactive care requires. But coding is also a system that is allegedly systemically gamed. Says MedPAC’s 2018 report: “Some plans appear to have modified their approach to coding diagnoses to maximize revenue to the detriment of accurate reporting of diagnosis codes or consideration of patient needs. In recently unsealed lawsuits, whistleblowers alleged that plans ignored evidence of improper coding; used software that is incapable of deleting invalid diagnoses, or ignored the status of a diagnosis as valid or invalid; and focused clinical programs on patients with a potential for coding a higher level of severity (e.g., diabetes without complications) but not on patients already coded with the highest level of severity for a condition (e.g., diabetes with complications) who might benefit the most from disease management.”
Since pre-ACA days, going back at least to the Deficit Reduction Act of 2005, Congress has recognized that this disparity of incentives between MA and FFS in risk adjustment has justified a numerical fix. It called for a compensatory downward adjustment in MA risk scores, and for the 2014 payment year, according to Kronick, that adjustment factor was 3.41%. Following further direction in the ACA and the American Taxpayers Relief Act of 2012, the downward fix was increased by 0.25% yearly until it reached 5.91% for 2018. Many experts believe that for a truly level playing field it should be larger—more like 8% to 11%. CMS has statutory authority to increase it more, but hasn’t.
“They never redid the analysis,” says Shawn Bishop, the Commonwealth Fund’s vice president for controlling health care costs and advancing Medicare. Indeed, a new risk-adjustment model could be created based on MA’s own claims data, she says, and CMS is working on that—but at an apparent snail’s pace. Meanwhile, coding intensity adjustment is a political hot potato the agency would rather not juggle further.
“CMS just isn’t motivated to take on this issue,” says Bishop. “It’s a political decision, not a data decision. The analysis is difficult, and the plans don’t like the adjustment to begin with. They’re making a ton of money doing it this way.” In any case, it’s widely agreed that average MA risk scores have been rising relative to FFS—by 1.5% annually over the past decade, says Kronick.
Quality star ratings (remember them?) play a role once again when it’s time for MA plans to receive rebates. The rebate, of course, is a fixed percentage of the difference between an MA plan’s risk-adjusted benchmark and the bid it has submitted as an offer to provide capitated care. Plans with three quality stars or fewer receive 50% of the benchmark-bid difference as a rebate; those with three-and-a-half to four stars get 65%, and those with four-and-a-half or five stars are awarded 70%. And the additional services that can be offered with rebate money—sweeteners to attract new enrollees at the next open enrollment period—include at-home care and telemedicine.
Pulling back out of the weeds a bit, concerns about MA plan payment aren’t limited to the growing risk-adjustment gap between MA and FFS and all the other details. They also go to the whole benchmark-bid process and the way it works. As health economist Austin Frakt has noted, benchmarks really have nothing to do with plan costs. They’re an open invitation to plans to “bid” up to a clearly stated permissible level.
Agrees the Urban Institute’s Skopec: “When you give plans a benchmark to bet against—when you tell them, basically, the maximum you’re willing to pay—you may get higher bids than if they were bidding blind.” She concedes that fewer plans might bid if true price competition ruled instead. But she believes the present Rube Goldberg-style MA payment system suffers from an unresolved tension between two goals: maximum choice for seniors and maximum savings for taxpayers.
Indeed, Harvard health economist Vilsa Curto and his colleagues studied Medicare costs from the year 2010 and found that basing MA benchmarks on FFS Medicare costs caused $21 billion in overpayments that year.
But are you ready for a little policy whiplash? Curto and company also found that MA still saved money over FFS, with spending reduced by 9% to 30% at no sacrifice in quality.
We live in an age of irreconcilable politics, when red won’t give blue an inch, nor vice versa. But even a conservative can’t avoid noticing some element of cross-subsidy here from the bad ol’ government to the knight-on-a-white-horse private sector. No more than a liberal/progressive can deny that the preventive medicine and care coordination MA seeks to encourage just may be where hope lies.