In the mid-1960s, the Johnson administration launched a federal initiative to more effectively deliver health care to regions across the country. Jack Wennberg was appointed director of the Regional Medical Program based in Vermont. Though Wennberg had only just finished his residency, he had developed a deep interest in how social systems worked while earning a master’s degree in Public Health. In Vermont, he became intrigued by a question: Why was it that physicians in some parts of the state seemed to prescribe more treatments than in other parts, even though the population of patients was essentially the same? Near his home town of Morrisville, to pick one early example, tonsillectomies were performed 10 times more frequently than they were 60 miles away in Middlebury. And in time, as those patterns persisted, an equally compelling question arose: Why was there no correlation in outcomes?

Wennberg published his findings in a 1973 issue of Science, and came up with a startling conclusion. “Given the magnitude of the variations,” he wrote, “the possibility of too much medical care and the attendant likelihood of iatrogenic illness is as strong as the possibility of not enough service and unintended morbidity and mortality.”

"The Vermont work made clear that there was a new problem for health care policy makers: too much medical care."
– Gil Welch, Professor of Medicine, The Dartmouth Institute

He extended his inquiry into the 13 hospital referral areas in the state of Maine, and found exactly the same kind of “unwarranted variations.” He formed a center at Dartmouth to deepen his research, investigated not only treatments and diagnoses, but hospital admissions, insurance policies, costs of service — discovered the same lack of correlation with positive outcomes — and became convinced that a host of factors beyond a patient’s illness and needs were driving decision making. It turned out that, to a large degree, where someone lived determined the kind of care someone got.

In the 1990s, hoping to add science-based neutrality to the Clinton health care reform debate, Wennberg and his colleagues at Dartmouth turned their research to the entire country. Relying heavily on Medicare records, they established the first national database of geographic variations in health care. Over time they discovered that the number of hospital admissions per capita was driven not by science or patient need, but by the number of hospital beds in a given region. They determined that physicians’ values and preferences greatly influenced the way they interpreted their science. They estimated that 30 percent of Medicare spending was unnecessary.

Since its initial publication in 1996, the increasingly detailed Dartmouth Atlas has become a powerful tool among health care policy- and decision-makers. Two of its emerging legacies: a greater emphasis on outcomes-based research, and the growth of shared decision-making between patients and their care givers.Back to top