Every health economist remembers their first Dartmouth Atlas.
Louise Sheiner got hers in 1998, when she worked at the Federal Reserve. She still has that same copy today — even though it was actually the Fed's property, and she left that job earlier this year for a new post at the Brookings Institution.
"Nobody has come after me for it, at least not yet," Sheiner says.
It's hard to overstate the importance of the Dartmouth Atlas, a research project begun in the mid-1990s by health-care researchers at (unsurprisingly) Dartmouth College. The 18-year study has shown the incredible variation in American health spending. What Medicare spends on a single patient's hospital care ranges from $5,371 in Utah to $8,937 in Maryland.
Dartmouth Atlas researchers used this variation as the base of an influential argument — one that, among other things, was an inspiration for Obamacare: the people in Maryland aren't sick enough to justify $3,600 more in health care than the people in Utah. Maryland doctors are providing inefficient, wasteful care.
This raised the rare possibility of a free lunch in health care. By cutting the wasteful spending, America could save money and improve care at the same time.
Atul Gawande used research from the Dartmouth Atlas to write his seminal piece on McAllen, Tex., the area of the country with the highest health care spending. The Atlas came up repeatedly during the Obamacare debate as evidence that America could cut costs without hurting care.
Sheiner has had her copy of the Atlas for more than a decade now. But her newest research argues that big, clunky book she's held onto for years now is wrong.
She thinks that differences in health do explain the variation in spending — a more pessimistic conclusion that suggests doctors can't simply reduce spending by copying high-performers elsewhere.
"It's nice for Dartmouth to say here's a good measure of how much we can save, and nobody has to sacrifice," Sheiner told me in an interview. "But I don't know how much space for those type of savings actually exists."
There is definitely huge variation in health spending
Here's where Sheiner and the Atlas researchers agree: There is lots of variation in American health care spending. You can see that in the Atlas's many maps, like this one, which shows average Medicare spending in the last two years of a patient's life.
Where Sheiner and the Atlas disagree is why spending varies so much.
The Dartmouth Atlas argues it's largely about how doctors practice medicine. Researchers there have spent lots of time looking at individuals' health (how old they are, for example, or whether they have diabetes) and how much Medicare spends on them. They find that differences in health status explain only about 30 percent of the variation in health spending.
The other 70 percent, they argue, has nothing to do with patients — and everything to do with how doctors provide medicine. Some doctors decide to provide lots of medicine that isn't supported by research. They try things, hoping they'll work. Dartmouth describes these physicians as "cowboy doctors."
"Cowboy doctors are more likely to do things where there's no clinical basis," says Jon Skinner, a health economist at Dartmouth and researcher on the Atlas. "People are sick everywhere, but in these particular areas, there seems to be a lot more spending. And a lot of it comes down to providers' decisions."
Sheiner's new working paper for the Brookings Institute argues this is wrong. While Dartmouth uses individual level data, Sheiner does something different. She compares average Medicare spending across an entire state to average health. Here's one of her charts that plots Medicare spending against states' diabetes rates.
The diabetes averages, Sheiner finds, are really good at predicting the average Medicare spending. She finds that a state's average diabetes rate can explain 70 percent of the variation in spending. And if health problems explain most of the spending variation, that suggests much less room for doctors to cut costs without hurting care.
Whose data is better?
Sheiner argues that her averages do a better job at picking up on the health of a population than individual-level characteristics.
An example: it's definitely helpful to know whether an individual has diabetes to predict how much they spend on health care. But it's even more helpful to know whether that person has had diabetes for years or just acquired it. State-level data, Sheiner argues, does a better job at capturing some of those longer-term trends than a single data point.
"States where the incidence of diabetes is high," Sheiner argues in her paper, "are likely to be in worse health, on average, than in states where the incidence of diabetes is low."
Armed with this data, Sheiner argues that the Dartmouth Atlas has made the task of cutting health-care spending seem too easy. She says it explains why, even 16 years after the Atlas launched, the high-spending states haven't become more like the low-spending states.
"Cross state variation isn't telling us much," Sheiner says. "I don't say we can't cut spending, but I don't think its going to be as easy as Dartmouth has suggested."
Most research comes to a different conclusion
Sheiner's argument is at odds with a lengthy body of research.
The Atlas authors, to start, reject Sheiner's analysis. They argue that looking at state level data can lead to all kinds of weird conclusions that don't make sense. Skinner likes to draw an analogy to state voting patterns.
"If you look at voting, we know that people who earn less tend to vote Democratic and those who earn more vote Republican," Skinner says. "If you ran state-level regressions on the states that supported Bush, you'd find a correlation with lower-income states. And you'd say that, well this must mean Republicans are poor. But if you look at the actual data within the state, you quickly learn this isn't the case."
Instead, Skinner points to a strong body of research that studies variation among individuals — and makes a compelling case that wide swaths of the country have hugely inefficient health care systems.
One such paper published this summer looked at what happens to Medicare patients' spending when they move from a high-cost region to a low-cost region. And it shows that, all of a sudden, these patients start to have higher medical spending just by virtue of switching to a higher-spending city.
Separate Dartmouth research has looked at what percent of patients with chronic heart failure receive beta blockers. In a perfect world, this number would be 100 percent: guidelines recommend these drugs as the standard of care for this particular condition. Research finds that's not the case. There is lots of variation, even among these ideal candidates, of who gets beta blockers. In Connecticut, it's more than 80 percent; in Mississippi, it's less than 40 percent.
"This is an area where you're looking at ideal candidates for a treatment who are in the same state of health," Skinner says. "This tells you that its not the patients' health that's varying. It's the care they receive."
Another angle the Dartmouth research has taken is looking at doctors' beliefs about medicine. There, researchers have found that doctors who think that ineffective treatments work are (surprise!) more likely to provide that unnecessary care. That suggests that doctors' views about medicine do matter in terms of what care patients receive.
The big question: is it possible to make high spenders more efficient?
There's one place that both Sheiner and Dartmouth's supporters agree: even if the Dartmouth Atlas is right, that doesn't make health-care reform easy.
When we spoke Wednesday, Sheiner made the point that, for a long time, we've known that Kaiser Permanente, an insurer that owns its own hospitals and employs its own doctors, is really efficient. When economists want to point to a the type of system that is the model for improvement, they often point to Kaiser.
But here's the thing with Kaiser: we've known for decades they do better, but most health-care systems haven't adopted their methods. They remain fragmented and inefficient. "Economists don't understand why Kaisers haven't taken over the world," Sheiner says. "But they haven't, and you have to wonder why that is."
Amitabh Chandra, who has been a pointed critic of Sheiner's work, thinks this a valid place to critique the Dartmouth Atlas. It can point out the differences between efficient and inefficient regions of the country. But that doesn't mean we know how to transform the latter areas into high-performing health care systems.
"Kaiser hasn't taken over the world, and one reason might be we've hit a ceiling of doctors who want to work under that model," Chandra says. "Or maybe some patients don't want Kaiser. That's a constraint that we have to deal with."
One of the things that's so difficult about tackling wasteful care, is actually identifying the waste. At many appointments, doctors have difficulty knowing when treatment is needed - and when it won't provide help at all.
The medical community tries to address these issues with comparative effectiveness research, studies that compare how different treatments stack up against each other. Obamacare is trying to figure out some of this with its Patient-Centered Outcomes and Research Institute, or PCORI. But that effort is small and hamstrung: it can only compare the effectiveness of treatments and is strictly barred from looking at costs.
Dartmouth, in either words, might tell us accurately what's broken about American health care. That doesn't necessarily mean that its given us the tools to fix it.