Recently, I sat down to talk with Arlene Ash, PhD about risk adjustment. Dr. Ash is Professor and Chief of the Division of Biostatistics and Health Services Research, Department of Quantitative Health Sciences at the University of Massachusetts Medical School. As a methods expert on risk adjustment in health services research, she has pioneered tools for using administrative data to monitor and manage health care delivery systems, including those now relied upon by various federal programs, including CMS’ Medicare Advantage. About one-third of Medicare’s beneficiaries are currently enrolled in Medicare Advantage, a privately administered managed care program run by dozens of insurance companies in different regions.
We started talking about a recent Medicare Advantage fraud class-action law suit in which the government and others allege that UnitedHealth systematically gamed the risk adjustment system and fraudulently collected millions of dollars through “upcoding” — essentially claiming that their enrollees were sicker than they were. Two Florida Medicare Advantage plans, Freedom Health and Optimum Healthcare, have recently settled a similar lawsuit by paying $32 million. By one estimate, unjustified coding intensity in Medicare Advantage could cost Medicare $200 billion over the next 10 years.
I asked Dr. Ash about the problem of upcoding, as well as what the future of risk adjustment might be. What follows is an edited transcript of Arlene’s comments.
Arlene Ash: The goal of Medicare Advantage’s risk adjusted payment method is to give more money to plans that enroll sicker people, and we only know how sick someone is from the medical diagnoses that plans submit. That is, we give plans a huge incentive to code aggressively. We’ve always known that we need checks against fraud.
Since the beginning, CMS has studied trends in upcoding, watching continuously enrolled people in both Medicare Advantage plans and traditional (fee-for-service) Medicare, and comparing trends in risk scores. They’ve seen a greater year-over-year increase in risk scores for Advantage enrollees, leading them to apply a coding intensity adjustment that reduces payments to the Advantage program as a whole.
However, CMS has also seen big differences in risk score growth across Medicare Advantage plans, and — whatever adjustors are applied across the board — the more that an individual plan upcodes, the more money it gets.
When Washington State adopted diagnosis-based risk adjustors for its Medicaid program in the mid-1990s, it decided to try to level the playing field by teaching everyone to code as aggressively as could be legally justified. But how do we distinguish legitimate upcoding from fraud? Broadly, it is only legal to submit codes that relate to conditions that are addressed during at least one face-to-face encounter with a clinician during a year. If the code for diabetes, say, does not appear during the next year, the diabetes-related payment will go away, since we don’t want to pay more for a person with diabetes who isn’t actually being treated for the disease.
When payment became linked to these coding rules, some plans started looking at their data and found that many people with a diabetes code in one year had none the next. Digging deeper, they saw that this wasn’t careless coding — many of these people had not had their diabetes addressed for over 12 months! A plan could respond by improving care — seeing to it that people with a serious chronic disease were seen more regularly — and thus justifying the disease-associated payment bump. This is a win for quality. However, if it simply carries forward the diagnosis — whether or not there was a diabetes-related visit — that would be fraud.
Aside from fraud, a problem with the current methodology is that if really good care made a health problem disappear, the plan’s payment would be reduced. On the one hand, if the person is less sick, less money it may take fewer resources to care for them. However, such payment rules obviously don’t encourage the best possible care; we should separately reward exemplary management of chronic illness, especially when a serious medical problem, such as Type II diabetes, can be entirely eliminated.
The Future of Risk Adjustment
When an entire health sector (here, Medicare Advantage) documents an exceptionally rapid accrual of serious illnesses (compared to traditionally-insured Medicare enrollees), that is not reassuring. Of course we should pay more when a plan “takes on” caring for sick people, and we should pay for finding and treating previously unrecognized underlying illness. But at some point, increasing risk scores start to suggest less benign causes: inadequate care or fraudulent upcoding. We could address these concerns by monitoring changes in risk measures, and introducing both incentives that reward decreases in risk and fraud surveillance systems that look for suspicious “risk accrual” patterns.
These concerns arise for any program that contracts with managed care entities, such as MassHealth (Massachusetts’ Medicaid program), with which our team has been working. Most managed care organizations in our State have been anchored in hospitals that don’t have strong relationships with the non-medical supportive services community. The new Delivery System Reform Incentive Payment (DSRIP) program (part of our State’s 1115 Waiver) will, among other things, include funding to strength relationships with “Community Partners” to coordinate crucial “wraparound” services that can help avert medical emergencies. Will Accountable Care Organizations (ACOs) and Managed Care Organizations (MCOs) be able to provide good care for many of those sickest Medicaid beneficiaries who have previously been cared for through fee-for-service arrangements? Risk-adjusted payments should encourage MCOs to enroll these people, and the State’s DSRIP money will be used to help set up the infrastructure to enable providing all needed services.
Along with other State reforms, the risk-based payment model for MCOs and ACOs that we built for MassHealth now adjusts not only for age, sex and medical diagnoses, but also for social risks (including disability, housing problems and neighborhood indicators of economic stress). We’re proud of that refinement. However, while really good risk adjustment is necessary, it is not sufficient. Will paying for social risk result in managed care organizations engaging in activities that address social risks? Our payment system allocates Medicaid funds in a way that should help, but can MassHealth use CMS’ DSRIP investment to encourage, support, and even push MC organizations to successfully transform? We certainly hope so – and we’ll be watching carefully to see what happens.
To read more from Arlene on delivery system reform, check out this series on risk adjustment in Medical Care.