And how would we know why, and what could we do about it?
An economist at the Federal Reserve Board, Louise Sheiner, recently released a paper on the subject that questions the premise on which much of Medicare policy, including several provisions in the Patient Protection and Affordable Care Act (PPACA) (P.L. 111-148), is based.In particular, Sheiner challenges the work of the Dartmouth Institute, which President Obama and PPACA proponents cited in support of the legislation. The Dartmouth Institute has studied health care spending and policy in detail since 1988. Sheiner’s main objection is to their methodology.
The Dartmouth Institute uses data on the services furnished to individual beneficiaries and payments made for those services . In several publications, the authors found wide regional disparities in spending. The per capita payments in the most expensive regions, Miami-Dade County, Florida and McAllen, Texas, were more than twice the payments in the lowest quintile. Most recently, they have disregarded the payments that are not attributable to individual patients, such as disproportionate share hospital adjustments, and have adjusted to compensate for the differences resulting from the wage index.
Still, they have found higher spending in some areas that is not attributable to price, but to more intensive use of discretionary services than in areas with lower per capita spending. For example, the increased use of test requiring equipment, such as CT scanners, is associated with the increased availability of the equipment, a phenomenon they refer to as “supply- sensitive care.” And increased intensity of service use does not necessarily correspond with improved health outcomes.
But the variation is reflected more clearly when hospital referral regions are compared rather than states. Dartmouth also has found that variations in per capita spending are reduced when adjusted for the intensity of service use. Most areas did not change significantly, although New York City moved from the highest quintile to slightly below average.
Sheiner posits that the best way to analyze health care spending is at the state level. She finds that most of the apparent geographical variation in Medicare spending arises from differences in the statewide rates of insurance status, exercise, and obesity, and race—that is, higher Medicare spending correlates with a higher percentage of black residents. Although Sheiner addresses spending, she does not relate it to the intensity of service use. Sheiner compared Medicare spending to Medicaid and private insurance in each state, and she found that Medicare spending was not a reliable predictor of healthcare spending in general, though there is interaction among Medicare, Medicaid and private spending. The Dartmouth experts agree.
There are some obvious differences between the Medicare population and the general population: elderly people are more likely to have one or more serious chronic illnesses, while young people usually need preventative care, with maternity and obstetric or accidental injuries as the typical reason for a hospital stay. So it makes sense that there would be differences in spending. That’s why Medicaid expenditures for the elderly account for more than half of Medicaid spending even though the elderly are about 25 percent of Medicaid beneficiaries.
The Medicare Payment Advisory Commission (MedPAC) also has examined regional disparities in payments the intensity of use of services at the county level. MedPAC found that the variation in spending was reduced when the analysis controlled for service use, but substantial variation remained, and, generally, the same regions were at the top and bottom of the scales. MedPAC notes that there is a great deal of variation in payments within states, and sometimes, within the hospital referral region. In fact, MedPAC’s data show that in 2006, the average spending per beneficiary on durable medical equipment in South Florida ranged from $220 in Collier County to $430 in Broward County (Fort Lauderdale) to $2,200 in Miami-Dade.
MedPAC distinguishes clearly between Medicare spending and utilization of Medicare services. Medicare spending includes the wage index, supplemental payments to hospitals that serve a disproportionate share of low-income patients, academic medical centers, and rural referral centers. These payments are not based on the intensity of service utilization. Service usage in the communities in the 90th percentile, the top 10 percent of utilization, is about 30 percent greater than the service use in the bottom 10 percent. And fraud is suspected in areas like Miami where usage is drastically out of proportion relative to neighboring counties.
Jonathan Skinner, PhD and Dr. Elliot Fisher, senior researchers at Dartmouth, released a brief response to Sheiner’s paper. They “confess” to using aggregated data in the 1990s before richer individual data on health risks and more powerful computers were available. However, they abandoned that approach because of the “ecological fallacy,” the mistaken idea that the behavior of individuals is based on group averages. And they gave the following example: In 2004, Republican George Bush won the ten states with the lowest incomes, and John Kerry won nine of the ten states with the highest incomes. From that statewide data, one might infer that low-income people tend to vote Republican, and high-income people, Democratic. But research shows the opposite is true—Skinner and Fisher cited research showing that high-income voters are at least 10 percent more likely to vote Republican than low-income voters are.