One evening during my residency, I found a friend, another doctor in training, standing at the foot of his patient’s bed in the intensive care unit, shaking his head. The middle-aged patient was well known to many doctors and nurses in the hospital. Unemployed and living in one of the city’s rougher neighborhoods, he appeared regularly in the emergency room with his blood sugars way out of control.

This time, however, he was in with a different diagnosis: Fournier’s gangrene, a rare but life-threatening complication of his poorly controlled diabetes. The devastatingly aggressive infection of his groin and pelvic area was so severe that even hours after the surgeons had finished the emergency operation to cut away the infected tissue, people passing through the operating suites could still smell the overwhelming odor of pus and rotting flesh.

That night in the I.C.U., I saw the mounds of saturated gauze that covered the patient’s gaping wounds and the medley of machines and monitors that now supported his life. “We knew how to treat his diabetes whenever he came in,” my friend said, a tone of sad resignation in his voice. “But the problem was when he went home.” He sighed, shook his head again and added: “Every time he went home, everything went right out the window.”

My friend and I had grown up professionally with an unshakable belief in the science of clinical care, the idea that chronic illnesses like diabetes, high blood pressure and asthma could be controlled and complications averted if the treatments were based on scientifically sound research data. But in the case of this patient — and those of many others I would care for through the years — the most effective treatment recommendations that research could offer were insufficient. All our best efforts were thwarted once the patient went home.

For over 20 years, the ideal of practicing evidence-based medicine has enjoyed more than the clinical benefit of the doubt; it has become the unquestioned sine qua non of quality patient care. And in many cases, it has resulted in the widespread acceptance of several significant, and beneficial, changes in clinical practice. Most patients who have had a heart attack now fare better because of evidence-based recommendations to take a medication known as a beta-blocker. Diabetic patients who follow the proven guidelines for self-monitoring and blood sugar control now tend to experience fewer complications. And because of the findings of a recent large-scale clinical trial, some women with early-stage breast cancer may no longer have to suffer the complications of lymph node removal to prevent tumor spread.

But for all the attention and effort directed toward disseminating and adhering to proven treatments in patient care, one uncomfortable fact has become increasingly clear: for many patients, evidence-based medicine isn’t working. Two-thirds of patients with diabetes, a disease with some of the strongest evidence-based guidelines available, continue to have trouble controlling their blood sugar levels; and only half of all patients with hypertension, another well studied disease, ever get their blood pressures under control.

Now, in a perspectives article from The Journal of General Internal Medicine, two primary care doctors attempt to explain why. They assert that evidence-based medicine ignores the impact of the patient’s life at home and results in fractured and desultory care. To remedy the current system, they call for a fundamental shift in the way that primary care is practiced. They advocate an approach that blurs the traditional division between doctors’ offices and communities, a new paradigm they call “evidence-based health.”

“Right now we provide all this clinical care in our offices,” said Dr. David Moskowitz, lead author and a researcher and primary care physician at the University of California, San Francisco. “But then we send the patient back home to a community that can pose significant obstacles to what we are recommending they do.”

A doctor, for example, might advise diabetic patients in the office to eat more whole grains and fewer sugary foods. But many patients, particularly in low-income communities, have limited access to farmers’ markets or can’t afford such wholesome foods. “Patients certainly have a role in their health care,” Dr. Moskowitz said, “but their choices are frequently constrained.”

And while some programs have aimed at helping patients better manage their own care, creating parks and neighborhood walking paths, for example, teaching patients how to take medications correctly, or instituting healthy changes on school lunch menus and implementing communitywide smoking bans, most have operated independently of primary care practices.

But an evidence-based health model would change all of that. Primary care providers and community and public health workers would no longer have to work alone but together in coordinated efforts that would extend from the exam room to the home.

Such integration, at least in early pilot programs, appears to work. Since 2008, Vermont’s evidence-based health initiative, the Blueprint for Health, has successfully coordinated the care of roughly 60,000 patients. Community health teams that include nurses, social workers and behavioral health counselors spend time both within doctors’ practices and out in the community, tackling care-related obstacles like transportation, insurance applications and even housing and unemployment. While the program is still relatively young, hospital admissions and emergency department visits have dropped, resulting in lower monthly costs per person.

“If you can’t buy heating fuel, you aren’t going to buy insulin,” said Dr. Lisa Dulsky Watkins, associate director of the program. “And if you can’t afford transportation, you’re going to wait until the last possible moment, until you’re really sick, to take a taxi to go to the emergency room. And that costs a lot.”

With bipartisan support, Vermont officials are planning to expand the program statewide. But even the most enthusiastic supporters of the Blueprint plan — and of evidence-based health — concede that the critical factor for success has little to do with physician or community engagement, patient care or even, perhaps most surprising, the increasing evidence supporting such an approach. Rather, the crucial first step is a change in the fee-for-service reimbursement system. Vermont Blueprint community health teams and physician practices are paid for their work coordinating patient care, and the state’s health plans and insurers have been voluntary supporters of the program from the start.

Obtaining that level of support in other states may prove more difficult.

“Evidence-based health isn’t an easy endeavor,” Dr. Moskowitz acknowledged. “Health care providers on the ground know the kinds of lives patients lead and the constraints that exist. They understand that perfect evidence-based recommendations for chronic diseases won’t really affect things.

“What matters is filtering that information up to the payment system, to where the change needs to start.”