The new findings drew skepticism from an economist at the Dartmouth Institute for Health Policy and Clinical Practice, which has suggested for years that the way providers practice medicine drives the differences. The Dartmouth Institute produces the Dartmouth Atlas of Healthcare, a geographic analysis of health spending.
The report, published in the journal Medical Care Research and Review, looks at Medicare spending; diagnoses during the final months of life; and patients with unambiguous conditions, such as heart attacks or hip fractures, that are not influenced by physicians' discretion.
James Reschovsky, a senior fellow at the Center for Studying Health System Change, and co-authors Jack Hadley of George Mason University and Patrick Romano of the University of California Davis, said the results undercut the Dartmouth Atlas findings, which rely on end-of-life spending to conclude that intensity of care in the last six months of life is to blame for much of the deviation.
Reschovsky and the other researches said most of the gap separating the highest- and lowest-cost areas disappears after adjusting for patients' diagnoses—the distance from the top and bottom quintiles narrows 84%. They then more broadly analyzed spending variation to see whether physicians' possible influence over diagnoses could have skewed the results.
Diagnoses were categorized by whether they were more or less likely to be vulnerable to physicians' discretion and controlled for market forces that could influence physician behavior, such as competition.
Jonathan Skinner, a Dartmouth Institute economist and researcher, countered that recent research by the Institute of Medicine's committee on geography and health spending found “substantial” variation cannot be explained by demographic or health differences.
The IOM found in a recent interim report that adjusting for health status “does considerably decrease” the variation in spending. However, the IOM could not identify the source—wasteful or otherwise—for other variation, the report said.
Skinner said Dartmouth has been unable to replicate findings by Reschovsky and his colleagues using a data source that's different but comparable enough that it should yield similar results.
Dartmouth's analysis relies on a sample of patients, he said, whereas Reschovsky's team sampled physicians, which he said could skew results depending on the mix of physicians and the likelihood that the physician sample includes a greater number of acutely ill patients.
Reschovsky declined to comment on Skinner's assessment of methodology but said he trusts the results. “Our sample was benchmarked against administrative data from CMS, so I am confident that our results are not the result of how our beneficiaries were sampled,” he said in an e-mail.
Follow Melanie Evans on Twitter: @MHmevans