Public Integrity has a good report on Medicare Advantage over-coding. The Center for Medicare and Medicaid Services (CMS) conducts annual audits of Medicare Advantage carriers to see if the chart data supports the diagnosis codes that build a case that a member is more expensive and needs a bump payment from CMS to make the insurer whole. CMS conducted 37 audits in 2007 and thirty four of the plans had at least 10% of their claimed risk adjustment cases judged to be unsupported by the medical records. One carrier had, in my mind, de minimis variation (2%) and two carriers were spot on. On the other hand seven carriers or about 20% of the sample had at least 60% of their risk adjustment claims judged to be unsupported by documentation in the chart.
Why is this important?
It is important for a couple of reasons. First, this is a rip-off of public money. Medicare Advantage has a history of upcoding the risk adjustment model. The same patient in traditional Fee for Service Medicare looks a lot healthier based on claims data than if they are in an aggressively upcoded Medicare Advantage plan. That means the Medicare Advantage plan gets extra bump payments that are either sent back to members as added benefits, sent to provider as higher payments or consumed by company insiders and owners as hookers and blow.
Secondly, there has been tentative industry discussion/advocacy to move the Exchange/QHP risk adjustment model of revenue neutral flows towards the Medicare HCC bump payment system. That would change the Red Queen race incentive that rewards technically sophisticated coding operations at the expense of naive or data poor coding operations towards a Cash Dash against the Federal government. I’m not too thrilled with this idea.
So how does CMS stop it? There are three ways that limit overpayments and one that could reduce legitimate discrepancies.
The first way has been a floating discussion to not allow diagnosis codes that show up on only a Health Risk Assessment to be counted for risk adjustment. This was talked about for the 2017 plan year rules but it did not happen.
Health-risk assessments are wellness exams usually conducted in a senior’s home. Insurers send nurse practitioners or other clinicians to check a beneficiary’s vital signs, verify drug prescriptions and identify physical-safety risks. The goal is to get a sense of a patient’s health status, which insurers code to get a risk-adjusted score….
MedPAC commissioners voted last month to deny payment for diagnoses found during home visits unless a doctor confirmed the diagnoses in a separate office visit. Insurers have vigorously opposed that idea, arguing it restricts their ability to care for seniors at home. Advantage plans could also see a large financial hit. Gorman estimated that eliminating health-risk assessments for payment could ding revenue by 2%.Yet payments from home visits avoided the chopping block in the rate notice posted Feb. 19. The CMS said that although it “appreciates the work of MedPAC,” not all of the recommendations are addressed in rate policies.
The second way for CMS to limit over-coding is to hire more full time auditors. We’re probably still at the point where an auditor will pay for themselves in cost savings by the middle of the year. CMS is auditing more aggressively now than they have in the past, but it is still fairly light from my observations given how much money is at risk.
An extension of the auditing would be to bump up penalties. If carriers are showing a multi-year pattern of extensive over-coding, triple penalties to the company as well as significant fines to the executive who signs off on the claims submission would change the incentive structure from Medicare risk adjustment as a pure revenue center towards a clinical data center with some revenue implications. CMS should make it expensive to be significantly wrong.
A technical adjustment to make things better and reduce legitimate discrepancies would be to allow the HCC model to incorporate some prescription drug data like the Exchange HCC model will. Some prescriptions are obvious indicators. The Hep-C drugs are tied directly to Hep-C while there might not be a current year diagnosis of Hep-C, the costs of the disease are still being borne by a carrier.
I'mNotSureWhoIWantToBeYet
Looking for unwarranted up-coding would seem to be a natural application for AI like IBM’s Watson. It presumably would pay for itself in short-order, and might be useful for quickly catching things that need to be addressed by policy-makers (“Hey, we’re seeing an upsurge in W61.61XDs – maybe there’s something going on!”)
Do you know if anyone, besides the AI vendors looking for funding, are looking into such things?
Cheers,
Scott.
I'mNotSureWhoIWantToBeYet
Looking for unwarranted up-coding would seem to be a natural application for AI like IBM’s Watson [there’s that bad word again]. It presumably would pay for itself in short-order, and might be useful for quickly catching things that need to be addressed by policy-makers (“Hey, we’re seeing an upsurge in W61.61XDs – maybe there’s something going on!”)
Do you know if anyone, besides the AI vendors looking for funding, are looking into such things?
Cheers,
Scott.
Soprano2
This explains why my husband’s Medicare Advantage provider constantly pesters him with phone calls about wanting to conduct a home visit. My husband always turns them down because he doesn’t think he needs it. He’s got diabetes and high blood pressure, so I’m sure they’re anxious to find reasons to use him in their efforts to adjust upward.
liberal
Those are a good start. How about making it a felony for the executive, with mandatory minimum jail time?
AFAICT there’s just obscene amounts of white collar crime in the US that goes unpunished. (Not that I think it’s necessarily better anywhere else, or that it’s a new problem.) Stuff like wage theft should be accompanied by nasty, draconian sanctions, as should be ripping off Uncle Sam.
burntoutdoc
I have lots of disdain for the players in healthcare who exploit issues (such as coding) in a way meant to inappropriately inflate payments for themselves. But in a system as complex as ours, it is to be expected that the border between correct and excessive behavior will become increasingly difficult to police.
I think we need to re-examine the whole premise that one can reliably describe a complex ongoing disease process by a set of code numbers, and then use those codes to reliably assign the right number of health care dollars for the care of that particular illness in that particular patient.
For example, ten years ago two of my diabetic patients who started out with subjective tingling in the feet were assigned the code for diabetic neuropathy. Patient #1 had no progression of the symptoms, was never treated for it despite continued symptoms that the patient ignored, and never developed lack of sensation, neuropathic injury, or charcot joints. Zero health care dollars spent on this code. Patient #2 has progressed rapidly (despite foot checks, specialty consultations, and multiple medications) to insensate feet and amputations. LOTS of dollars spent here.
The point I’m trying to make is that the underlying assumption that patients with similar coding can be used to reliably distribute health care dollars in the future is not an accurate way to allocate resources to the health plans that are paying for their care. I understand that from an insurer point of view, some of this billing data is prospectfully useful to set premiums for groups of patients. But when you turn it around and try to assign payment to one patient based on diagnosis codes, this can really be as arbitrary as paying per patient visit. You just change who can game the system, the nuts and bolts of how it is done, and who wins and looses.
burntoutdoc
@burntoutdoc: Oops. “prospectively”
chris murphy
“That means the Medicare Advantage plan gets extra bump payments that are either sent back to members as added benefits, …”
Really? Not much chance of that.