The Limits of Using Medicare Data to Evaluate U.S. Health Care Spending
Six years ago, as a new president was taking office and the prospect of significant health care reform was moving from pipe dream to possibility, the city of McAllen, Texas, was thrust into the national spotlight. McAllen was the focus of a now-famous New Yorker article written by Atul Gawande, M.D., who used the Dartmouth Atlas of Health Care to highlight critical problems with health care in America. Health care spending in McAllen was exceptionally high—annual payments made by Medicare to providers in the area were $7,000 more per person than the national average—and patient health outcomes, compared with lower-spending communities like Rochester, Minnesota, were terrible. So compelling was Gawande’s article that President Obama reportedly had copies of it sent to several members of his Cabinet and advisors.
The findings Gawande reported were based on Medicare claims data, which have historically been the best source of information about U.S. health care spending. Medicare accounts for 20 percent of health care spending, and the federal government has a record of every payment that is made through the program. Payments under the other major public health care program, Medicaid, by contrast, are made by individual states, so there is no equivalent central database. The remaining source of information about spending is from private insurance claims, which comprise 33 percent of U.S. spending. But private insurance data have historically been the most challenging to access, because insurers consider the information proprietary and, like Medicaid data, they aren’t centrally collected.
As a result, we’ve gotten in the habit of using Medicare data to comment about spending across the entire U.S. health care system. But as private health insurance data have become available to researchers, we’re finding that using Medicare data this way is likely a mistake. Case in point: When the data became available in Texas, we learned that McAllen was only an expensive market when the Medicare program was footing the bill. Research by a team at the University of Texas, conducted in 2010, showed that even though Medicare spending in McAllen was 86 percent higher per person than in neighboring and demographically similar El Paso, spending by the largest private insurer in Texas, Blue Cross Blue Shield (BCBS), was 7 percent lower per person in McAllen. Recent research by the same team shows no correlation between BCBS of Texas and Medicare spending in the state. Other recent studies (Franzini 2015, IOM 2013) have made similar observations about other areas of the country.
New research led by Yale economist Zack Cooper, supported by The Commonwealth Fund and released this week, further illustrates just how different Medicare and private health care spending really are. Cooper and his team, which also includes Stuart Craig of the University of Pennsylvania, Martin Gaynor of Carnegie Mellon, and John Van Reenen of the London School of Economics and Political Science, analyzed a new database of private insurance claims that is administered by the Health Care Cost Institute and comprises data from Aetna, Humana, Kaiser Permanente, and UnitedHealthcare. The team’s work supports earlier research showing that Medicare and private spending are not highly correlated, and that spending drivers are very different in the two markets.
The team’s work begins to quantify those differences. Moreover, because the data are national, the researchers can compare Medicare and private spending across the United States, not just in isolated markets. Some of what they find is remarkable. In his 2009 article, Gawande cited Grand Junction, Colorado, as a low-spending foil to McAllen, Texas. Grand Junction has even been cited as a health care model for the nation. However, while Grand Junction had the third-lowest spending per Medicare beneficiary in the nation in 2011, that same year it was the 43rd highest-spending market for the privately insured (out of more than 300) and had the ninth highest hospital prices in the nation.
What accounts for these differences in spending? According to Cooper and his team, private insurance spending is driven by health care providers’ prices, rather than the volume of health care services provided, which is the primary driver of Medicare spending. And price, Cooper finds, varies dramatically both within and across geographic areas. Hospital-negotiated prices often vary by a factor of four or more within a single geographic area, and by a factor of more than 10 across the U.S. As Cooper notes, this variation far exceeds the variation in Medicare reimbursements, which differ by a factor of about three across the U.S. The large variation in private prices is also present for relatively homogenous services, like MRI scans, which vary by a factor of 12, and ninefold within a single U.S. city (Miami). Cooper also finds that a hospital’s market power is strongly associated with its prices. When a hospital is in a monopoly, for example, its prices are 15.3 percent higher than hospitals facing competition from three or more other providers. All of this suggests that price variation is driven by factors other than clinical quality.
Cooper’s work puts to rest the notion that we can generalize findings from analyses of Medicare data to the commercially insured population. Because the research is based on national data, it also allows us to compare markets and start to examine why we see such dramatic variation in private spending. The work also gives us a sobering picture of how inflated private-payer prices really are. His team estimates that if private-payer prices were replaced by administered prices set at 120 percent of current Medicare reimbursement rates, inpatient spending on the privately insured patients they studied would decrease by 17 percent. As affordability becomes the issue for the new insurance marketplaces, and as employers shift more cost to employees, it is going to be critical to learn much more about the actual spending drivers for private insurance. Cooper’s work represents a critical step in that direction. To read it, visit HealthCarePricingProject.org.