Big loss for CMS today: healthleadersmedia.com/finance/medica…
[Initiate long but important MA upcoding thread]
[Initiate long but important MA upcoding thread]
Background: CMS pays MA plans fixed per person monthly payments. Payments are set using a risk adjustment model that bases an individual's payment on their expected cost in FFS. The model includes demographic categories as well as a large set of indicators for chronic cond'ns.
These condition indicators are based on diagnoses appearing in claims. Model is estimated using data and people from FFS Medicare. Model effectively assumes a person who gets a diabetes diag in FFS Medicare has same expected cost in FFS as a person who gets a diabetes diag in MA.
This assumption is dubious because MA plans are highly incentivized to code every cond'n a person has any evidence of having, while no such incentive exists in FFS.
CMS audits diagnoses submitted by MA plans to determine whether they were valid (i.e. from a face-to-face encounter with a doc and substantiated based on clinical records). If diags are determined invalid, MA plans must pay back any overpayment caused by the invalid diag to CMS.
United contended that these audits violated the "actuarial equivalence" requirement of the MA program. Basically, they say that the FFS diags on which they original payments are based are not audited, so they should not be required to pay back overpayments appearing in an audit.
Their argument is effectively that in FFS there is a lot of statistical noise in the condition indicators due to inaccurate coding. This noise drives cond'n coefficients toward zero.
If MA is subject to audits that clean the noise, but the original FFS data that payments are based on is not subject to those audits, then this violates "actuarial equivalence" and hurts MA plans.
Basically, if the FFS data were cleaned of the noise, coefficients would go up and MA plans would get paid more for their enrollees who are coded with these conditions.
This is how the court ruled: That CMS can't force MA plans to pay back payments caused by diagnoses found to be invalid in audits unless CMS bases the payments to MA plans on audited FFS data.
Here's the problem: These audits only go in one direction. They only determine whether REPORTED diagnoses were valid. The biggest coding difference between MA and FFS is that there are many more UNREPORTED diagnoses in FFS.
So doing audits in FFS will remove even more diagnoses from a setting where "UNDER-diagnosis" is the problem.
MA plans on the other hand spend billions of dollars coding every possible condition for all of their enrollees, driving up their payments.
So here's the question (and why this case was a ballsy move by United): Does this ruling open up CMS to audit FFS records for UNREPORTED diagnoses? If they did so, the payments for chronic diseases would surely go down. A lot.
They'd go down because many relatively healthy diabetics or folks with cancer in remission would now get coded with these diseases in FFS. Since these folks are lower cost, they'll drive down the average, and thus the MA payments.
The big problem in coding in MA is that the MA plans are much more complete coders than FFS, yet they are paid as though they coded they same. That is surely a worse violation of actuarial equivalence than the noise problem in this case.
So, now we wait to see what happens. Will CMS step up and find the myriad unreported diagnoses in FFS, leveling the playing fiend with the MA plans? Or will they roll over and just keep the payments flowing for every insurer's favorite cash cow?
(quick note: I think it would be totally reasonable for CMS to go seek out these unreported diags only in FFS and not in MA and still be complying with "actuarial equivalence" given that the MA plans are already doing this. Indeed, by not doing this CMS is violating act equiv)
For more info on this topic, see my paper with Mike Geruso here: nber.org/papers/w21222
