Methodology for ‘Medicare Advantage Money Grab’

Introduction

For this series, the Center for Public Integrity analyzed Medicare Advantage (MA) data obtained from the Centers for Medicare and Medicaid Services (CMS). The data contained enrollment and payment information for MA contract-plans and Fee-For-Service (FFS or traditional Medicare) payment data by county, downloaded from three CMS web pages: CMS Plan Payment Data, CMS Monthly Enrollment and Fee-For-Service Data.

Brian Biles, an MD and professor in the Department of Health Policy at George Washington University, and Giselle Casillas, an MPP and senior research assistant at the GWU, offered methodology advice and helped guide parts of the analysis.

The Center used monthly enrollment data from March as enrollment remains relatively stable that month. Because of regulations in the Health Insurance Portability and Accountability Act (HIPAA), small Medicare Advantage plans with 10 or less people are omitted from the data so that patient identities remain private. These CMS public data contain no patient-level information other than aggregated enrollment figures.

We analyzed data covering 2007 to 2011. After the Center completed its analysis, CMS released 2012 data.

A key to understanding cost differences between MA and traditional Medicare lies in comparing the risk score in each plan and each county. To accomplish the comparison, MA data were weighted by insurance plan county enrollment as Fee-For-Service (FFS) risk score data was only available on the county level. Weighted risk scores were calculated by taking county total enrollment, further broken down by the MA providers’ total enrollment in that county, and weighting the risk score by that number. We weighted the payments by county enrollments, similar to methodology used by the Medicare Payment Advisory Commission (MedPAC).

To compare annual payments instead of monthly costs between MA and traditional Medicare, the weighted base payment for each county for MA plans and the FFS base payments were multiplied by twelve.

For the annual payments at the national level, the Center excluded payments in Puerto Rico, as do most researchers working with MA data, due to significantly different Medicare billing there.

Other limitations include CMS calculations of the traditional FFS risk scores, which are done on a rolling basis in 5-year groups rather than the single year calculation for MA risk scores. Small risk scores changes and differences, though, can significantly add to costs to the taxpayer. The analysis only highlights plans with a wide gap between their risk scores and comparable FFS risk scores, or uses other descriptive measures, such as percentage change in MA risk scores. Also, many of the aggregated calculations are rounded down to eliminate any over estimations.

The data behind the interactive graphic comes from the same CMS data. The Center used the raw data or simple descriptive percentage change in MA risk score to highlight MA trends from 2007-2011.

Calculating a national trend in risk scores is difficult at best. Limitations include the nature of the aggregated data provided by CMS and lack of more specific attributes related to patients who remain in the program through the years as compared with new enrollees. From the CMS data, calculations suggest the average risk score grew slightly under 2 percent. But a report from MedPAC suggest a projected 3 percent growth overall between just two of the years in the data. Finally, a Government Accountability Office Report suggests it could be higher. A 3 percent average is a conservative estimate for comparison purposes only.