Health economists and policy analysts have long known that Medicare spends much more, per patient, in some parts of the country than in others. In fact, the primary project of a large research group at Dartmouth is devoted to analyzing the geographic patters in Medicare spending.
Last year, Atul Gawande brought this phenomenon greater public attention with an article in The New Yorker on medical care in the areas of McAllen and El Paso, two regions in Texas that have superficially similar demographics but vastly different levels of per-patient Medicare spending. However, several recent studies suggest that the regional variation might be the result of Medicare’s payment systems and that privately insured patients experience less variation in treatment levels, perhaps sacrificing quality of care.
Many policy wonks tend to assume that the lower levels of spending are the correct levels, and the higher levels are due to high levels of wasteful health care. The Dartmouth group attributes the variation to differences in “practice patterns”—essentially, regional variations in physician culture that lead to more unnecessary care in the more expensive locations. Dr. Gawande, who interviewed doctors and hospital administrators in McAllen (but not in El Paso), attributed the difference to “the culture of money,” which led doctors to recommend more intensive procedures to more patients than absolutely necessary.
Underlying both of those explanations is the assumption that physician behavior is to blame, not anything about the Medicare program. Indeed, the researchers use Medicare data not because they are interested in Medicare as such, but because Medicare datasets are easily available, and they might be reasonable if there are no systematic differences between how physician treat Medicare patients and how they treat other patients.
But is that assumption valid? Can we really assume that Medicare data is a valid sample for analyzing treatment of all patients? Or is this just a case of “looking where the light is better”? Might the pattern look different for non-Medicare patients? If so, the problem might not be doctors in certain areas but in the Medicare program itself and the way it adjusts payments for different areas.
An article published this week in the journal Health Affairs by Luisa Franzini, Osama I. Mikhail, and Jonathan S. Skinner suggests that the difference between McAllen and El Paso investigated by Dr. Gawande might really be a feature of Medicare, not of local physician culture. They obtained data on treatment of patients in both locations covered by Medicare and also data on patients in both locations covered by a private, non-profit insurance company—Blue Cross and Blue Shield of Texas.
What they found calls into question the assumptions that health policy wonks have been making for years: While Medicare indeed spends almost twice as much more per patient in McAllen than in El Paso, Blue Cross spends about the same in both places. In fact, Blue Cross’s per-patient spending was actually slightly lower in McAllen. These findings persisted for overall spending, as well as for spending on specific types of services and several specific diseases.
This confirms results found by other researchers looking at the same question on a nationwide basis. Andrew Rettenmaier and Thomas Saving (a former Medicare trustee) found that state-by-state variation in per-patient Medicare spending was not strongly correlated with total per-capita health care spending. In addition, Rong Yi found that regional variation in spending on privately insured patients was closely correlated with measures of disease burden. And Tomas Philipson and colleagues found that geographic variation in health care utilization is substantially larger for patients in government programs compared to those insured in the private sector.
Why the difference? Philipson and colleagues point out that economic theory suggests that private insurers are less able to control prices but have stronger incentives to restrain utilization while keeping patients (their customers) satisfied. Franzini and colleagues point out that Blue Cross and Blue Shield of Texas, like most large private insurers, implements disease management programs for patients with chronic disease—programs that both improve patient health and reduce hospital visits, emergencies, and other sources of excess cost. Medicare does not provide, or reimburse for, disease management services; these services are available to Medicare patients only if they enroll in private-sector Medicare Advantage plans that provide them.
Furthermore, private insurers have substantial incentives to root out fraudulent bills. Fraudulent billing, if undetected, not only hits their bottom line but would force them to raise premiums and lose business to competitors that can better prevent fraud. Medicare, on the other hand, has little incentive to prevent fraud and devotes only a small amount of resources to that end. Indeed, the areas with the highest per-patient spending figures are often those with the highest levels of known fraud.
It appears from the evidence that the large variation in Medicare is not indicative of something deeply wrong with the way health care is delivered throughout the system even outside of Medicare but rather an indication that something is deeply wrong with Medicare. Not the doctors and not the patients—but with Medicare’s payment system, service mix, and incentives.