Managed Care Medicare Medicare Advantage

Medicare Advantage Plan’s Response to Risk Adjustment

Medicare beneficiaries have the option to enroll in private plans to have them operate their benefits rather than use the tradiational Medicare Fee-for-services (FFS) program.  Medicare pays these private plans, known as Medicare Advantage (MA) plans, premiums based on the health status of their enrollees.   Medicare uses a risk score to measure beneficiary health status.

Can MA plans strategically select enrollees who will fall below expected cost?  Does the MA program save Medicare money?  Today’s post answers these questions drawing largely from a 2011 study by Brown, Duggan, Kuziemko and Woolston.


During the 1980s and 1990s, CMS used a “demographic model” to generate individual level risk scores, so-called because it included only demographic variables (gender, age, and disability and Medicaid status) as opposed to disease or health conditions. CMS found that the model explained only one percent of the variation in payments among the FFS population.

“In 2004, CMS introduced a more comprehensive risk-adjustment regime that is based on the hierarchical condition categories (HCC) model…the HCC model uses claims data from the FFS population to calibrate a model that predicts FFS costs in the following year, though the HCC model accounts for not just inpatient claims, but physician and outpatient claims as well. The model distills the roughly 15,000 possible ICD-9 codes providers list on claims into seventy disease categories.”

Studies have found that the HCC model explains 11 percent of FFS cost for the subsequent year; CMS found that when FFS data are used to calculate HCC scores, the HCC score explains eleven percent of FFS costs the following year (Pope et al., 2004). A study by Newhouse et al. (1997) found that plans can predict about 20 percent of cost in the subsequent year; this finding implies that MA plans may still be able to engage in significant selection.


  • Risk scores should increase when patients switch from FFS to MA.
  • Conditional on risk scores, total Medical expenditure for those switching to MA relative to those staying in FFS should fall after risk adjustment.
  • MA recipients are positively selected with respect to over cost both before and after risk adjustment.


To test these hypotheses, the authors use individual-level data from the Medicare Current Bene ficiary Survey (MCBS) Cost and Use series from 1994 to 2006. The authors describe the data set as follows: “The MCBS links CMS administrative data to surveys from a nationally representative sample of roughly 11,000 Medicare enrollees each year. It also provides complete claims data from hospital admissions, physician visits, and all other Medicare-covered provider contact for all FFS enrollees in the sample, totaling about half a million claim-level observations each year.”


Below I describe the evidence the study uses to confirm their three proposed hypotheses listed above.

  • Risk scores risk for these switching to Medicare Advantage.  Individuals switching into MA before risk adjustment had risk scores roughly 0.31 points lower than those remaining in FFS, risk scores of those switching into MA rise after risk adjustment is introduced, making up about a third of the difference.
  • Medicare Advantage beneficiaries cost less than predicted by risk adjustment.  After risk adjustment, individuals switching into MA are over $1,200 “cheaper” than their risk-scores predict them to be.
  • When a county’s Medicare population shifts from FFS to MA its average per capita FFS costs increase.  Those switching between FFS and MA have costs roughly $575 lower than the remaining FFS beneficiaries.  This phenomenon occurs due to the fact that when low-cost enrollees leave the FFS population, the average cost of the beneficiaries remaining in FFS increases.

The observed cost relate to Medicare premium payments.  However, they may understate the true cost as individuals are unlikely to postpone expensive procedures until they join an MA plan because plans tend to have less generous cost-sharing arrangements for serious medical procedures than does FFS.

The authors find that individuals are still profitable if they fall in the first four quintiles of the risk score distribution.  Even after risk adjustment, however, the least healthy individuals (i.e., those with risk scores in the top 5 percentiles) are still unprofitable for MA plans.

I pose that the main reason this result occurs is “upcoding” by MA plans.  Medicare bases MA beneficiary health status on the diagnosis codes submitted by MA plans.  FFS providers typically have less of an incentive to extensively document all of a beneficiary’s health conditions.  For instance, physicians are paid based on a fee schedule that pays them based on the procedures they do rather than the health of the patient.  MA plans, on the other hand, receive more money the more diagnosis codes they document.  Thus, when FFS beneficiaries switch to MA plans, their risk scores increase even if their true health status changes little or not at all.

The reason that the most severely ill patients are still unprofitable is that medical spending is highly skewed with a long right tail.  Even upcoding will not sufficiently reimburse MA plans for their costs to treat high cost outliers because these beneficiaries–by definition–are much sicker and cost much more than the typical person with a given set of conditions.  Further, while the HCC model is (for the most part) additive in nature, health spending rises in a nonlinear fashion as the number of comorbidities a patient suffers from increases.