Following our September 2021 Health Affairs Forefront articles (parts one and two) on the “Medicare Advantage Money Machine,” the most common reaction of which we are aware has been that this was an “Emperor Has No Clothes” moment. Many knew the Medicare Advantage (MA) game, but few had called it what it was. Two responses to our articles, by individuals we have long admired, have been published in Forefront that find fault with our analyses. One is by George Halvorson, for 30 years one of the most accomplished executives in US health care and a passionate advocate for MA as a solution to our nation’s health care cost and quality conundrum. The other comes from Don Crane, another highly respected leader for many years as the CEO of America’s Physician Groups; he has also been a strong advocate for physician practices that accept accountability for care. We have deep respect for both, but, for the most part, their arguments fail.
We count more than a dozen assertions in these critiques, some matters of opinion and others matter of fact. These can be grouped into the following six categories, which we address here:
- The accuracy of our characterization of the MA business model, the MA Money Machine and its effects;
- The profitability of MA firms;
- Effects of MA on care and underinsurance for low-income MA beneficiaries;
- Financial savings for MA beneficiaries;
- The quality of care in MA; and
- The Direct Contracting Model now redesigned as the accountable care organization (ACO) Reach Model.
The Accuracy Of Our Characterization Of The MA Business Model
Halvorson cites our claim that “Risk score gaming creates a major transfer of wealth from taxpayers and Medicare beneficiaries to MA plans, and it lies at the heart of the business model for most MA plans” as well as our claim that the Centers for Medicare and Medicaid Services (CMS) “consistently overpays MA plans with no demonstrable clinical benefit to patients.” He writes, “These two statements are entirely incorrect.” He asserts that the Risk Score Game is only played for “a short burst of possible productivity, as all diagnoses tend to be found quickly and then it’s done” and “…after all there is only a small number of finite diagnoses…. Finding more codes is not a business plan; it’s a care improvement plan. The truth is that we want every diagnosis code to be found. We very much want our care teams to know every diagnosis for every person because that information and knowledge is extremely useful, indeed necessary, for optimal care delivery.”
As we point out, the Medicare Payment Advisory Commission (MedPAC) and many others have documented that MA costs more than fee-for-service. Every extra dollar paid to MA plans is, indeed, a transfer of taxpayers’ money to MA plans, which then benefit in terms of growth and higher profits. Those profits represent direct transfers of wealth from taxpayers to plans.
Medicare beneficiaries, as well, pay for this subsidy. Of every extra dollar paid to MA plans, CMS actuaries allocate approximately 58 percent to Part B costs. Part B premium is then calculated to cover 25 percent of total Part B costs. Every MA beneficiary and most traditional Medicare beneficiaries pay this premium directly. MA plans, in collecting these overpayments, are taking dollars directly from 57 million Part B beneficiaries (likely including, by the way, Halvorson and the two of us). We will address the clinical benefit to patients in the quality-of-care section below.
Halvorson’s claim that the Risk Score Game is limited (that is, done) is not correct. The “small number of finite diagnoses” involves more than 10,000 International Statistical Classification of Diseases and Related Health Problems, Tenth Revision diagnosis codes that feed into 80-plus Hierarchical Condition Categories (HCC) that drive payment, each of which must have evidence of clinical attention (usually in the electronic health record) and be recoded and documented every year. Hence, the growth of an entire coding industry of which Halvorson is either unaware or turning a blind eye toward. Many MA firms, such as Signify, Inovalon, Change Healthcare, and Optum, to name but a few, highlight their MA risk coding products as major drivers of revenue. Furthermore, research that we cite by Richard Kronick and F. Michael Chua, MedPAC, and others demonstrates that MA risk scores have for decades risen continually, and apparently without limit, at 1 percent and now almost 2 percent per year faster than those in traditional Medicare. That is not a mark of continually increasing relative severity of actual illness in MA compared with the fee-for-service population; it is the Risk Coding Game, and, despite Halvorson’s claims, apparently it is never “done.”
Among the well-documented methods MA plans use for the Risk Coding Game is to send nurses into homes to do health risk assessments to collect codes. Halvorson disputes that coding is a primary purpose of plan home visits saying, “… in fact, the plan had direct, in-home support in place and was using nurses and processes that were already set up to provide care. They did not send nurses into new homes; they sent them back into homes where they had already been to support patient care.”
The link Halvorson provides goes directly to United Healthcare’s Optum HouseCalls website, a program that was the subject of the following exchange on United Health Group’s Q1 2021 Earnings Call:
Questions from Justin Lake, Wolfe research analyst: “And then on Medicare Advantage, it sounded like your health assessments are going better in 2021. Should we expect that you should get a significant amount of that risk score headwind for 2021 to reverse next year and get those revenues back?”
Reply from Andrew Witty, CEO: “Hey, Justin, thanks so much for the questions. … Let me just touch on the health assessment. You heard in the prepared comments the very strong performance of the house calls program (600,000 visits) …, record quarter, … in terms of performance of house calls. And you’re quite right. That makes us feel pretty optimistic that that headwind that we saw is going to dissipate pretty rapidly … gives us kind of a rising optimism … that much of that kind of the negative headwind that characterized the emergence of the pandemic starts to mitigate, at least on that dimension.”
Halvorson makes no mention of recent reports from the HHS Office of the Inspector General (OIG) documenting that diagnosis codes not reported in service records but only through Chart Reviews and Health Risk Assessments (HRAs), a majority of which are done in the home, resulted in $9.2 billion in additional payments. The OIG writes:
“Diagnoses that MAOs [Medicare Advantage organizations] reported only on HRAs, and on no other encounter records, resulted in an estimated $2.6 [billion] in risk-adjusted payments for 2017. In-home HRAs generated 80 percent of these estimated payments. … We found that twenty of the 162 MA companies drove a disproportionate share of the $9.2 billion in payments from diagnoses that were reported only on chart reviews and HRAs ….” United Healthcare “further stood out in its use of chart reviews and HRAs … it had 40 percent of the risk-adjusted payments from both mechanisms ($3.7 [billion]) yet enrolled only 22 percent of MA beneficiaries.”
Almost every MA expert, observer, or regulator with whom we have spoken agrees that the primary purpose of home visits is to drive risk scores and premium. We agree that it can be important to document properly the diagnoses and care requirements of individuals with chronic disease. But, as Halvorson points out, that needs to be done only once. That goal would not be sufficient rationale to drive the widespread use in MA of metrics such as gap closures and code recapture rates on a routine and recurrent annual basis. It would not support performing 2.4 million registered nurse (RN) home visits a year or ordering scientifically non-indicated tests such as screening carotid ultrasounds, as we described. These efforts and costs make sense only in the Risk Score Gaming hunt for revenue in which codes must be resubmitted every year, not in an authentic quest for better care. Halvorson’s recent statements that “CMS has now officially cancelled and retired the CMS Hierarchical Conditions Categories Risk Adjustment Model that has been used for almost two decades to calculate risk for plans. It is dead and completely gone for 2022” appear to show that he does not understand how the HCC model and the Money Machine work. CMS has moved to using “encounter data” rather than data from the previous Risk Adjustment Processing System (RAPS). But Home Visits and Chart Reviews are explicitly included in encounter data and Plans continue to incent providers to document every possible code through the use of pay for coding metrics cited above. The increasing trend of MA Risk Scores described by Kronick demonstrates that MA Plans have effectively adapted to the change to encounter data.
Crane offers no comments regarding our description of the risk score gaming or the Money Machine Model. His primary critique is that “Drs. Gilfillan and Berwick’s attacks on MA and HCC coding seem to reflect a lack of appreciation for the indispensable role that risk adjustment plays in MA compensation.” He then connects risk adjustment to the greater percentages of vulnerable populations in MA. We understand that risk adjustment is necessary. That is why we recommended that CMS create an alternative approach that considers social determinant measures directly, which the HCC system does not.
Rather than confront the issue of plan overpayment, Crane argues that MA costs CMS less than fee-for-service. He cites a Milliman study commissioned by the Better Medicare Alliance, an MA lobbying firm. A close reading of the study in fact proves the reverse. Milliman, one of the largest actuarial consultants to MA plans, provided two analyses. The first used their own “proprietary tools” to analyze bid data, not actual costs, or payments, to estimate what CMS pays MA plans to provide Medicare Parts A and B services. They compared this bid data to their computed national average cost in traditional Medicare and conclude that MA costs CMS 0.7 percent ($7 per member per month [PMPM]) less.
But as Milliman notes on page 8, “The estimates reflect coding intensity differences captured by the 5.9 percent MA coding pattern adjustment set by CMS … but do not reflect any differences in coding intensity that may not be captured by this adjustment.” As acknowledged in the Milliman report, with the exception of one cited AHIP report that was debunked by MedPAC, most observers believe there is ample uncorrected extra risk coding intensity above the 5.9 percent. In fact, in figure 8, which includes results of a 5 percent sensitivity analysis, attempting to adjust for county distribution and additional risk score differences, the Milliman report shows that MA cost $41 PMPM more than traditional Medicare, or 4.5 percent more. Given the current projection by Richard Kronick and F. Michael Chua of a 14 percent uncorrected risk score difference, the total overpayment approaches 15 percent.
The Profitability Of MA Firms
Halvorson claims that our essays include: “A long string of inaccurate information, starting with the characterization of MA profits as ‘extraordinary’…. We can argue about whether 4.5 percent profit is a good number; but it’s clearly inaccurate to say that it is an ‘extraordinary number,’ as Gilfillan and Berwick contend. Most businesses in most industries would see their stock prices dropping with only 4.5 percent profits.”
This is a simple misreading by Halvorson. We did not say that MA profits are extraordinary (although the increases in plans’ stock prices and valuations surely are). We said the following regarding the provider incentive payments generated by the MA Money Machine: “While all can agree that we should improve compensation for primary care, these extraordinary profits are more likely to be captured by the for-profit parent entities rather than passed through to physicians delivering care.” We base this on the Deal 2 Money Machine delivering up to $200 per member per month in provider incentive payments, well above an average of $40 in traditional Medicare.
Halvorson says: “The health plans could use every technique and every approach known to the industry and described by Gilfillan and Berwick to inflate the risk scores for MA patients today and the full impact of that risk-pool enhancement, even risk pool distortion if they could achieve it, would still be to end up with pretty much the same amount of profit that we have now—because the 10 percent administrative costs are real, and because the ACA is not going to allow any health plan to exceed that 15 percent limit on costs and profits.”
Profits are the product of profit margin multiplied by revenue. Risk Score Gaming positions plans to increase profits from both. Margins get larger because of the difference between artificially high revenue from coding and static medical costs for a population that is in reality no sicker. MedPAC reports that for-profit margins now average 6.9 percent; no doubt some are much higher than the average. Rebates increase with risk scores providing more revenue as well as subsidies for improved benefits and lower premium that attract more members. Total revenue increases as membership grows dramatically due to the subsidized products. Administrative costs as a percentage of revenue are not fixed, they achieve economies of scale with growth, becoming lower as plans get larger. Large national plans typically have administrative costs well below smaller plans, often below 10 percent. This provides more room for profits even within the 85 percent medical loss ratio (MLR).
But as we showed, MA plans have ways to avoid the 85 percent MLR constraint. Under what we termed “Deal 2,” plans reward providers for generating higher risk scores by setting a medical expense target based on a percentage of the plan premium. As risk scores go up, premium goes up, and the difference between premium and medical costs become provider incentive payments that are recorded by the plan as medical costs. If the target is set at 85 percent, the MLR will always be reported as 85 percent or higher, regardless of the actual cost to provide care. The 85 percent constraint is rendered meaningless. Under “Deal 3” (in which MA plans employ physicians), including these payments in medical expenses may hide an actual MLR approaching 70 percent. Each of these deals drive Medicare costs up significantly. Halvorson and Crane addressed none of these realities.
The Effects Of MA On Care And Underinsurance For Low-Income MA Beneficiaries
Halvorson reports, “Lower-income individuals are more likely to choose MA, and more than 50 percent of African American and 60 percent of Latino beneficiaries choose MA.” And, furthermore, “Gilfillan and Berwick are also incorrect … when they write that, with MA, low-income beneficiaries remain underinsured.”
Crane cites two studies by ATI Advisory showing that low-income, minority, and vulnerable populations make up a greater percentage of MA as opposed to traditional Medicare.
Of course, lower-income people choose MA. Here is why: The median income for individuals older than age 65 was $27,000 in 2018, approximately $20,000 and $17,000 for African Americans and Latinos respectively. Those in the lower end of the income distribution have little discretionary income. As we described, the Money Machine Model drives MA growth because plans use some of the subsidies from CMS to offer products with improved benefits and lower premiums for members. MA plans consciously create different mixes of benefits and premium for different customer segments. Zero-premium products, with less-rich benefits and more out-of-pocket costs, are targeted at cost conscious buyers, particularly those with limited disposable income. Lower-income individuals have little choice but to trade a known zero premium cost for an unknown likelihood of greater out-of-pocket costs. But that tradeoff leaves many low-income beneficiaries underinsured and possibly experiencing significant negative impacts on their health.
The Commonwealth Fund definition of “underinsured” includes individuals whose “out-of-pocket costs, excluding premiums, over the prior 12 months are equal to 10 percent or more of household income; or whose out-of-pocket costs, excluding premiums, over the prior 12 months are equal to 5 percent or more of household income for individuals living under 200 percent of the federal poverty level” In 2019, they estimated that 25 percent of Medicare beneficiaries are effectively underinsured.
Does MA eliminate being “underinsured” for low-income beneficiaries? Milliman estimated MA enrollees on average would have out-of-pocket costs of $1,645 to cover copayments, deductibles, and coinsurance in 2019. Out-of-pocket expenses for traditional Medicare beneficiaries with supplemental coverage are much lower while for those without such coverage are about $600 higher. MA decreases but does not eliminate out-of-pocket expenses. Two hundred percent of federal poverty level in 2019 for an individual was $25,000, so the 5 percent line for being underinsured would be $1,250. This suggests that almost half of the MA population, and a likely larger percentage of African American and Hispanic MA beneficiaries could face the prospect of being underinsured. However, in states that expanded Medicaid under the Affordable Care Act (ACA), individuals with incomes up to 138 percent of poverty level (or higher in some states), about $17,250, would be eligible for Medicaid. This still leaves many MA beneficiaries facing out-of-pocket expenses greater than 5 percent of household incomes. Our point stands: Many lower-income individuals remain underinsured despite MA coverage.
Does being underinsured affect health? Out-of-pocket expenses present significant obstacles to care for many individuals and families. According to the Henry J. Kaiser Family Foundation (KFF), “Half of US adults report skipping or delaying medical care due to cost, including about six in ten lower-income, Black, and Hispanic adults.” More than 20 percent of African American and Latino adults report that they have postposed or delayed hospital care and physician care in the past year. This included 40 percent of Medicare beneficiaries earning less than $40,000.
In another report, KFF researchers found that the “cost-related problems” among MA members were reported by 17 percent of beneficiaries overall versus 12 percent for traditional Medicare with supplemental coverage and 30 percent for traditional Medicare alone. Additionally, 32 percent of African American and 24 percent of Latino MA members reported cost-related problems. The researchers conclude: “our findings suggest that enrollees in Medicare Advantage do not generally receive greater protection against cost-related problems than beneficiaries in traditional Medicare with supplemental coverage, particularly for some enrollees, such as Black beneficiaries in relatively poor health, despite having an out-of-pocket cap and additional benefits.”
Studies of low-income Medicaid populations have demonstrated that “cost sharing is associated with reduced utilization of services, including vaccinations, prescription drugs, mental health visits, preventive and primary care, and inpatient and outpatient care, and decreased adherence to medications.” It has also been shown to increase mortality rates.
Financial Savings For MA Beneficiaries:
Halvorson writes: “Gilfillan and Berwick do not mention … that MA members actually spend much less money each year on care. The combination of better benefits and better care for MA members means that the average MA member saves more than $1,600 a year on personal health care costs, as compared to traditional Medicare enrollees.”
Crane says: “the Milliman study found that that each dollar spent by the federal government on MA provides beneficiaries with additional benefits and lower cost sharing than beneficiaries enrolled in traditional Medicare receive. MA beneficiaries also spend less in out-of-pocket costs for Medicare-covered services.”
As already noted, we fully recognize that MA subsidies result in lower premiums and better benefits for MA members. Thus, of course, they spend less. How much less is not clear because the methodology in studies on this assumes that the populations have the same underlying burden of illness. Researchers have shown that the MA population is healthier than those in traditional Medicare. They should use fewer services and have lower out-of-pocket costs. We cannot project how much of the out-of-pocket difference is related to population differences, but we readily agree that out-of-pocket spending is lower because of MA subsidies. However, Halvorson’s cited study also makes the point that 12 percent of MA beneficiaries spend more than 20 percent of their income on out-of-pocket expenses including premium, confirming, again, that underinsurance persists in Medicare Advantage.
Champions of Medicare Advantage, such as Halvorson, repeatedly celebrate the rapid enrollment growth as evidence of better care and better service than in traditional Medicare. A far better explanation of MA’s growth is the cheaper prices and savings experienced by MA beneficiaries that result from the subsidies and wealth transfers we described above. If an ice cream firm can offer ice cream for free because of a subsidy, its market share will of course grow against firms that charge for ice cream. The cause is not better ice cream; it is free ice cream. The questions are: Where do the subsidies come from? And is it fair to the people who do not have subsidies? In the MA world, the subsidies come from taxpayers, Medicare beneficiaries, and the federal treasury. A free lunch for MA is not free; it is being paid for by others.
The Quality Of Care In MA
Halvorson writes: “Standard [fee-for-service] FFS Medicare has no quality agenda at any level. The care teams that are created by MA plans using the revenue flow from capitation do not exist in FFS Medicare. The patient chart information is extremely incomplete in FFS settings; information is siloed and not shared between caregivers. These major shortcomings are not addressed by Gilfillan and Berwick.”
Crane writes: “Drs. Gilfillan and Berwick pay almost no attention to quality.”
We are not unbridled supporters of traditional fee-for-service Medicare. We are quite supportive of CMS efforts to contract with providers through alternative payment models such as ACOs to deliver more coordinated care. But Halvorson’s comments on quality of care are in important respects misleading, and they deserve direct response.
It is seriously incorrect to say that traditional Medicare does not have a quality improvement (QI) agenda. The traditional Medicare QI program dwarfs the MA Star program in scope and results. The MA program is largely focused on the important but limited areas of office-based primary care and patient experience. CMS has been pursuing quality improvement in traditional Medicare for many years through numerous avenues, including competitive bidding, extensive survey and certification activities, and the Quality Improvement Organization (QIO) program. Over the past 13 years, the CMS Value-Based Purchasing Program has been developed to drive improved quality across hospitals, primary care and specialists physicians and advanced practitioners, nursing homes, and home health agencies.
While not all programs have demonstrated significant positive outcomes, we point to the work done by QIOs and the Partnership for Patients initiative, which resulted in significant improvements in hospital-acquired conditions and substantial cost savings. A literature search to identify MA plans, or any insurance company initiatives, to improve the outcomes of care in hospitals, nursing homes, or home care yields sparse results. In fact, recent studies have documented that MA plans are actively ignoring these system-improvement efforts as they use lower-quality skilled nursing facilities and home health providers more, and high-quality hospitals less, than traditional Medicare. Halvorson’s unwarranted claim disrespects the thousands of individuals working at CMS and their vendors as well as provider teams driving broad, systemwide quality improvement efforts.
Regarding care teams and integrated information systems, we agree that chart information in most physician practices is incomplete, siloed, and not shared. In fact, most MA members are cared for in the same practices as traditional Medicare patients, with the same charts for all patients. Large comprehensive and integrated care team practices using a common medical record such as Kaiser Permanente are the exception, not the rule, in Medicare Advantage. Some MA plans attempt to overcome this by providing care coordination for high-risk patients. ACOs operating in traditional Medicare are providing similar care coordination, producing almost $4 billion in net savings to CMS over the past eight years, as opposed to the $75 billion in excess costs that MA plans have laid on CMS in the same interval. And they are driving improvement in the same primary care provider office metrics used in Medicare Advantage for the 10 million patients they serve, almost one-third of all traditional Medicare patients.
Curiously, Halvorson’s sole citation of examples of team-based care, which he refers to as “care teams that are created by MA plans,” links to a 2008 Institute for Healthcare Improvement report that describes the implementation of effective team-based care in four organizations, only one of which, a Kaiser Permanente care site, had any relationship to MA. In fact, MA has neither a claim on authorship rights nor any unique corner on team-based care, which has progressively found its way into much of US health care, no matter how it is financed.
Crane and Halvorson cite the Star Quality ratings system: “MA has an extensive, well-structured, and strongly administered five-star system that measures plan performance on more than 40 categories of quality and service and rates every plan based on their most recent performance levels. … Gilfillan and Berwick suggest that we should ignore MA’s progress on quality; they call the improvement in all those plans and care settings “The Lake Wobegon Effect” … although they are not entirely silent on the overall quality issue—they misdescribe the MA reality by, as mentioned, falsely claiming that MA provides “no discernable (sic) clinical benefit to members.”
We did write that “MA harbors an arbitrage game in which CMS consistently overpays MA plans with no demonstratable clinical benefit to patients.” We chose the word “demonstratable” (not “discernable”) purposely to capture what MedPAC’s 2021 report to Congress said, as follows:
“Over the years, the Commission has discussed the flaws in the 5-star system and the QBP [quality bonus payment] and the continuing erosion of the reliability of data on the quality of MA plans …. The current state of quality reporting is such that the Commission’s yearly updates can no longer provide an accurate description of the quality of care in MA. The Commission’s March 2019 report to the Congress contains a detailed discussion of the difficulty of evaluating the quality of care within the MA sector and changes in MA quality from one year to the next.”
MedPAC reiterated this at its January 14, 2022, meeting. Halvorson neither cites nor rebuts MedPAC’s reports or conclusions. Instead, he urges celebration of the Star Rating achievements of MA plans, 80 percent of which are now rated at four stars or above. Regarding the “Lake Wobegon Effect,” CMS 2022 Star Ratings Documentation shows that from 2017 to 2022 the number of plan contracts above 3.5 increased from 49 percent to 68 percent and membership in the “above average plans” went from 68 percent to 90 percent. MedPAC described the longstanding MA plans’ approach to manipulating contracts to push more of their members into four star and above plans in their 2019 Report to the Congress as follows:
“For 2019, plan sponsors have used this strategy (contract consolidation) to move about 550,000 enrollees from non-bonus contracts to bonus-level contracts, resulting in unwarranted bonus payments in the range of $200 million in 2019. In the preceding five years, over four million enrollees were moved from non-bonus plans to bonus plans, including situations in which surviving contracts that fell below 4 stars underwent subsequent consolidations and were consumed by bonus-level contracts.”
Would that this reflected the success of the program to drive “continuous improvement” in outcomes for patients. However, a recent review of the Five-Star program found no evidence of that concluding:
“Together these results do not support the conclusion that the quality bonus program succeeded in improving overall quality of care for MA beneficiaries.”
Unfortunately, the Five-Star program, while well intended, primarily creates a “performing to the test” result rather than solid and important quality improvements in outcomes. Halvorson does not acknowledge, let alone dispute, the wide scientific controversy about the nature and value of current quality rating schemes, as highlighted by MedPAC.
Understanding the effects of MA on quality is extremely difficult in part because MA operates in an information environment in which it is all but impossible to demonstrate better clinical outcomes for patients. The reasons for this include:
- Lack of comparability across MA and traditional Medicare of HEDIS metrics, most of which rely on administrative data and assess processes, not outcomes. There remains serious uncertainty about the link between HEDIS metric improvement, as measured in incentive programs, and improved clinical outcomes;
- Selection bias among those who chose Medicare Advantage versus traditional Medicare;
- Lack of comparability between MA data sets, laden with the extra diagnosis codes driven by the HCC Risk Adjustment model, and traditional Medicare data sets; and
- Failure of risk adjustment to produce populations with the same illness burden when used to normalize populations for comparisons.
For all these reasons, as well as MA claims payment practices that result in frequent denial or downgrading of medical services and incomplete data, use and cost statistics are not comparable. With no acknowledgment of these confounders and gaps, nor of the conflicting evaluation literature, Halvorson simply asserts that:
“MA plans get better results. For example, compared to FFS Medicare, on a risk-adjusted basis, a 2018 study found that MA members have 33 percent fewer emergency department admissions and almost 23 percent fewer standard hospital admissions because of the better care that plans deliver in a wide range of areas.”
As illustrated by the Milliman study Halvorson cites, the literature comparing MA and traditional Medicare is replete with studies commissioned, sponsored, or conducted by MA plans and related entities using their proprietary claims data sets. The 2018 study cited by Halvorson is one of several non-peer reviewed analyses performed by a subsidiary of Inovalon Health, one of the first and largest coding optimization firms, at the request of the Better Medicare Alliance, an MA lobbying firm. The studies were built using the Inovalon MORE2 database, built upon data from the MA firms they support.
It is not entirely clear how the comparator populations were selected from the much larger MA and traditional Medicare populations. While some characteristics of the groups were very different (for example, only 64 percent of the MA group first entered Medicare based on their age versus 78 percent for traditional Medicare), the risk scores of the two groups were identical. Furthermore, table 2 (in the 2018 study) illustrates that the traditional Medicare population had a significantly higher prevalence of many chronic diseases. Since we know that MA data sets have many more diagnosis codes than fee-for-service, it seems obvious that these were not comparable populations. The traditional Medicare population was sicker, had a greater illness burden, and had higher costs as expected.
In a 2020 study by the same firm using the same data, this data bias was accentuated by using a “propensity matching” approach at the diagnosis-code level to create three different MA and traditional Medicare high-risk population subcategories for comparison. The inevitable outcome in both studies was that the up-coded MA population likely had a lower severity burden when compared to a matched traditional Medicare population with the same codes and was therefore likely to have lower costs and use.
The second study also illustrates one key MA business practice—downgrading inpatient stays to outpatient observations stays. These “downgrade” decisions are usually made after the patient has left the hospital. While provider reviewers often differ on the appropriateness of these downgrades, appeals are infrequent and mostly unsuccessful. The patient’s care and clinical outcome are unaffected by a retrospective change in categorization. The reported count of inpatient stays and total medical costs in claims data are lower as a result.
As shown in exhibit 1, the overall MA population seems to have a lower inpatient hospitalization rate. (It is surprising that the difference is only 10 percent given the claims of the prior study of a 23 percent reduction for the chronic disease population.) However, combining the observation stays and inpatient stays shows that inpatient use is slightly higher in Medicare Advantage than traditional Medicare. The relevant point for evaluating outcomes is that lower admission rates in MA cannot be taken at face value. Interestingly, in these studies emergency department visits were also higher for the MA population compared to the 33 percent reduction in the 2018 study.
We do not use these analyses to claim that care is worse in MA but rather to demonstrate the knottiness of comparing traditional Medicare and MA quality using what one would think would be reliable metrics and data. We suspect these results are just demonstrations of the four confounding issues raised above. Even MA advocates report confusing outcomes when they use their own data to compare MA and traditional Medicare quality.
Exhibit 1: Hospitalizations and emergency department visits, Medicare Advantage versus Traditional Medicare
Source: Data analysis by Avalere Health: Better Medicare Alliance, Positive outcomes for high-need, high-cost beneficiaries in Medicare Advantage compared to traditional fee-for-service Medicare. 2020 Dec.
Two recent reviews of the Medicare Advantage versus traditional Medicare literature, Eva DuGoff and colleagues and Amol Navathe and colleagues, showed mixed and at times contradictory results. MA tended to do better on Five-Star program quality process and intermediate outcome metrics such as HEDIS. The data on clinical outcomes, defined in various ways, was contradictory. DuGoff concluded there was some evidence of better health outcomes in MA, based primarily on readmission and mortality studies. Researchers have generally seen MA’s lower mortality rates as indicative of favorable risk selection, but one study suggests some of this may be due to MA care or extra benefits. Most recently, Navathe and colleagues summarize their review of the literature as follows:
Medicare Advantage was associated with higher use of preventive care visits, fewer hospital admissions, fewer emergency department visits, shorter hospital and skilled nursing facility lengths-of-stay, and lower health care spending relative to traditional Medicare. Although MA plans outperformed traditional Medicare in most studies comparing quality-of-care metrics, their enrollees were more likely to receive care in average-quality hospitals and lower-quality nursing facilities and home health care agencies. The evidence on readmission rates, mortality, experience of care, and racial/ethnic disparities did not show a trend of better performance in MA plans than traditional Medicare, despite the higher payments to MA plans. Challenges with the comparability of data (due to selection bias, challenges with risk adjustment, and unobserved differences related to social determinants of health) complicate comparisons between the performance of Medicare Advantage and traditional Medicare.
We acknowledged that clinical care in some provider-based MA plans has been different from MA plans at large. We wrote, “Aside from some notable exceptions with group model and staff model health maintenance organizations, capitation of privatized Medicare plans has simply allowed insurance companies to collect from CMS a toll of 15 percent or more on the total cost of care….” We believe, as Halvorson does, that capitation can help reshape the delivery of care. Kaiser Permanente has been a pioneer in care innovation and lowered the cost of care, quite possibly improving outcomes for their patients. But that success really demonstrates the impact of capitating providers, as the Permanente Medical Group, not the Kaiser insurance business arm, is responsible for those outcomes. Kaiser Permanente, sadly, is the exception rather than the rule; many of its innovations occurred outside of MA, and historically well before MA was on the scene. And, we note, in the perverse Marketplace of MA, successful efforts to redesign care may still not obviate the business-driven pursuit of Risk Score Gaming by nonprofit plans.
MA plans do have lower costs of care and use of services. They impact hospital use through care preauthorization, retrospective denials, and downgrading inpatient stays to observation status as well as care coordination. Total denials average about 6 percent according to an HHS OIG 2018 analysis. In one of our experiences (RG), the rate of inpatient denials and downgrades to observations status varied from 10 percent to 20 percent for the large national MA plans. Plans’ focus on postacute care does lead to fewer and shorter skilled nursing facility and acute rehabilitation facility stays. Home health care use is also lower. While plans generally use Medicare contracting approaches and prices, there may also be opportunities to pay less than Medicare for some professional and ancillary services. The most rigorous comparison of MA versus fee-for-service costs found it difficult to estimate the net cost savings of MA concluding:
We find that MA insurer revenues are 30 percent higher than their health care spending. Health care spending per enrollee-month in MA is 30 percent lower than in traditional Medicare; holding enrollee county and risk score fixed, this spending difference shrinks to 25 percent, and adjusting for mortality differences (used as a proxy for differential risk) further reduces it to 9 percent.
These differences, as well as MA plans’ avoidance of high-quality and specialty hospitals, reduce costs, but the overall impact of MA network approaches on clinical outcomes is not documented and could be unfavorable. With MA now covering almost 50 percent of the Medicare population, if all the reported better metric performance compared with traditional Medicare is leading to significantly better clinical outcomes for patients, that should be unequivocally demonstrable. It is not.
The Direct Contracting Model/ACO Reach Model
Crane says: “Curiously, the Forefront authors seem to take the view that Direct Contracting entities (DCEs) are not accountable care organizations (ACOs). And … DCEs build upon the framework of earlier ACO models, from 2012 Pioneer ACOs to 2016 Next Generation ACOs, and represent the evolutionary path of permitting physicians to manage populations and accept accountability for quality and outcomes.”
Our point was that direct contracting was a hybrid of the MA and ACO approaches that purposely targeted the very MA insurer and investor firms that drive MA overpayments. This is a path that positions financially driven intermediaries, not physicians, to manage populations as they do in MA. Our concern was that these firms will use direct contracting either as another Money Machine or as an opportunity to move patients into MA.
CMS’s recently announced changes addressed some of the issues we raised including limits to risk score gaming and monitoring movement into MA. However, we remain concerned that it brings MA investor and insurer-controlled firms directly into the fee-for-service alternative payment model (APM) world. The 52 of 100 Reach ACOs that are investor-controlled operate in 49 states (including the District of Columbia and Puerto Rico) containing 99 percent of the 35 million fee-for-service beneficiaries. Sixteen of 100 are insurers and approved to operate in 43 states with 90 percent of the fee-for-service population. They include all but one of the large national MA plans. The extensive existing primary care provider networks of MA insurers position them to disrupt existing ACOs and potentially expand their reach to millions of traditional Medicare beneficiaries.
While CMS has said they are particularly interested in entities with “experience providing direct patient care,” they have said it is not a requirement. Furthermore, most insurers have now created care delivery arms that could meet that criterion. We believe that insurers should be limited to operating in the MA space, but if CMS cannot eliminate them from ACO Reach they should at least limit their growth. At almost two million beneficiaries in 2022, ACO Reach is much larger than the Next Gen population of about 1.3 million beneficiaries before the expected expansion in 2023. We believe this should be a limited test of the model without allowing network and service area expansion.
A Path Forward
Medicare is at a crossroads with years of MA subsidies and overpayment driving ever faster privatization of the program. The ACO Reach programs threatens to drive this even faster. The concentration of health care dollars in the hands of private insurers creates a massive lobbying force on Capitol Hill, virtually disabling the will for change as we saw in CMS’s announcement of a projected 8.5 percent increase in MA revenue for 2023. As we said in our original articles, CMS should replace the HCC system with a new system that would include consideration of Social Determinants of Health such as the Area Deprivation Index. This could be done in a gradual process that avoids sudden changes to benefits and premiums.
The impact of MA on the delivery of care goes well beyond the Risk Score Game. Traditional Medicare presents a relatively straightforward, highly standardized approach to coverage and the financing of health care. MA creates a highly fragmented coverage and financing system that brings complexity and additional cost for all segments of the industry, particularly the providers. The result is an expensive MA administrative superstructure that we have layered on top of, and that now permeates and distorts, the actual delivery of care. Estimates of administrative costs and profits for all health system parties vary significantly but the broadest definitions are in the range of 25 percent. MA is a major driver of those costs. We must ask what we are getting for that if, after 35 years of privatized Medicare costing more than fee-for-service, we cannot demonstrate real clinical outcomes improvement. The answer is, “We are not getting much.” The opacity that comes from the privatization of public payment confounds objective, scientific analysis.
We are very supportive of the continuation of CMS alternative payment model (APM) efforts through the Center for Medicare and Medicaid Innovation and in conjunction with CMS programs such as the Medicare Shared Savings Program. We simply believe that the better pathway is for CMS to contract directly with providers focused on patient care, not with financial parties focused on financial results. Hence, our recommendations to develop an advanced ACO track that could involve total cost of care capitation with providers. Traditional Medicare needs a better benefit package, some basic managed care capabilities to eliminate areas of overuse, and a network of providers focused on delivering better outcomes.
Large employers have many of the same needs as CMS. But, they do not make the mistake of giving insurers their entire health care spend because they know that would only increase total costs. They simply pay plans an administrative fee to help them design and administer an effective benefit plan. A simple alternative to MA is to do the same in TM. Many third-party administrators (TPAs) can provide managed care services that MA Plans use. ACOs—which can put into the hands of clinicians both the opportunity to spend resources where they can best help patients and some extra economic motivation to do so—can be the basis for the network. We believe it is time for CMS to create a level playing field for a competition between ACOs and MA. One step to that end would be for CMS to establish a TPA for ACOs to use to administer a capitated model.
Given Halvorson’s personal experience and longstanding belief in the merits of capitated payment, as well as his success with MA and other programs at Kaiser Permanente, it is not surprising to find him such a vocal champion of MA. Indeed, if the only approach on the table were exactly Kaiser Permanente’s and the only leader were Halvorson, we would have far less reason to raise the cautions we have. Unfortunately, that is not the case. Building care delivery systems that improve clinical outcomes for patients is hard work and takes time. Most MA plans have neither the capabilities nor the mission to do that. It is much easier to excel at the arbitrage opportunities MA presents such as Risk Score Gaming and differentials in county benchmarks.
In his enthusiasm, Halvorson ignores important facts, selects his citations, and appears frankly naïve in his understanding of the MA industry outside of Kaiser. Crane, a tireless, effective, and highly respected advocate for accountable care, offers no response to our description of the MA Money Machine that is a central feature of the business model driving the financial success of many of his member organizations.
Notwithstanding their critiques, as we have heard from many others, the MA emperor remains exposed for what it is: the largest threat to Medicare solvency and a driver of continued waste of precious social resources. MA leaders talk about the importance of addressing the social determinants of health. We have proposed specific concrete steps they could take to move excessive spending from health care to social services, and still maintain current benefits for their members. Action on their part would speak much louder than their words.
Dr. Gilfillan was the CEO of Trinity Health System from 2013 until 2019. He is a trustee for US Pharmacopeia; a director for the Health Care Transformation Task Force; a member of the Advisory Committees for the Institute for Exceptional Care and several Robert Wood Johnson Foundation programs addressing health equity and social determinants of health (all uncompensated). Dr. Gilfillan also recently consulted for an integrated health system (compensated). Dr. Gilfillan was the director of the Center for Medicare and Medicaid Innovation during the roll-out of several ACO Models and was involved in the development of CMS ACO regulations. He has also been a leader and member of teams that managed multiple ACOs and Medicare Advantage plans. Dr. Berwick served as administrator of the Centers for Medicare and Medicaid Services from July 2010 to December 2011, during which he oversaw the issuing of the initial CMS regulations for accountable care organizations, as well as numerous other regulations devolving from the Affordable Care Act. He serves on the boards of LumiraDx (stock option compensation); Virta Health (stock option compensation); NRC Health (stipend and stock option compensation); Institute for Exceptional Care (uncompensated); CareVisor (stipend); Partners in Health (uncompensated); Results for Development (uncompensated). He also serves on advisory boards for the National Institute for Health Care Management Foundation, Datavant, and the Institute for Accountable Care, and on the American Medical Association Journal Oversight Committee. Dr. Berwick occupies multiple committee positions with the National Academies of Sciences, Engineering, and Medicine.