I was caught off guard yesterday by the publication of a study calling into question the methods and the main conclusions of Dartmouth Atlas Project research, which, as many of you know, found wide geographic variations in the quality and cost of health care.
The Dartmouth findings have gotten a lot of attention and were at the core of an influential New Yorker story by Atul Gwande, M.D. And while they've inspired some spirited conversation, the findings are essentially looked upon as true by policymakers, journalists and others.
Until now. Authors of a new study published online in Medical Care Research and Review claim that the Dartmouth research is at its core flawed. If their analysis is to be believed, it throws a giant monkey wrench into the grand debate over geographic variation.
James Reschovsky, a senior fellow for the Center for Studying Health System Change, took some time yesterday to explain what he and his two co-authors did, and he made a convincing argument that the Dartmouth research rests on an incorrect hypothesis: "They made the assumption that people in their last months of life are of roughly equal health because they're about to die," Reschovsky says. "We found that it could not be supported. People near the end of their life vary in terms of the number and type of conditions as much as the entire Medicare population."
The researchers examined the health of a set of patients who died in 2006 and compared the health of those in areas deemed by Dartmouth to be high-cost areas with those in low-cost areas. "We found that amongst people about to die, people in higher cost areas have a greater number of medical conditions," Reschovsky says.
"The bottom line: It goes to show that a lot of the prior research on geographic variations in Medicare is deeply flawed," Reschovsky says.
Yet, Jonathan Skinner, one of the researchers for the Dartmouth Institute for Health Policy and Clinical Practice who worked on the geographic variation studies, says they don't find the same conclusions, and he sounded just as convincing.
"Their result depends on people in high-cost areas being sicker — and using our national Medicare data, with the same diagnoses, we just don't even come close to matching their results," Skinner writes via email. He says he's not questioning the researchers' methods, but is saying that Dartmouth's chosen data sets don't produce the same results.
He also describes some other reasons why the new study's approach is not appropriate, and notes that a panel convened by the Institute of Medicine concluded that unexplained geographic variation is a big problem. The panel issued an interim report in March and expects to issue a final report on the issue this summer.
Ideally, the final IOM report will lay the debate to rest.
Too much time and money is being spent on the issue of geographic variation in Medicare spending for the country to forge ahead without knowing for sure the magnitude of the problem.