The Robot Will See You Now

IBM's Watson—the same machine that beat Ken Jennings at Jeopardy—is now churning through case histories at Memorial Sloan-Kettering, learning to make diagnoses and treatment recommendations. This is one in a series of developments suggesting that technology may be about to disrupt health care in the same way it has disrupted so many other industries. Are doctors necessary? Just how far might the automation of medicine go?

Watson "fills in for some human limitations," says IBM's Marty Kohn, a physician, who emphasizes that Watson is being developed to support doctors, not replace them. (Kareem Black)

So how would all these innovations fit together? How would the health-care system be different—and how, from a patient’s standpoint, would it feel different—from the one we have today? Imagine you’re an adult with a chronic condition like high blood pressure. Today, your contact with the health-care system would be largely episodic: You’d have regular checkups, at which a doctor or maybe a nurse-practitioner would check your blood pressure and ask about recent behavior—diet, exercise, and whatnot. Maybe you’d give an accurate account, maybe you wouldn’t. If you started experiencing pain or had some other sign of trouble, you’d make an appointment and come in—but by then, the symptom might well have subsided, making it hard to figure out what was going on.

In the future as the innovators imagine it—“Health 2.0,” as some people have started calling it—you would be in constant contact with the health-care system, although you’d hardly be aware of it. The goal would be to keep you healthy—and any time you were in danger of becoming unhealthy, to ensure you received attention right away. You might wear a bracelet that monitors your blood pressure, or a pedometer that logs movement and exercise. You could opt for a monitoring system that makes sure you take your prescribed medication, at the prescribed intervals. All of these devices would transmit information back to your provider of basic medical care, dumping data directly into an electronic medical record.

And the provider wouldn’t be one doctor, but rather a team of professionals, available at all hours and heavily armed with technology to guide and assist them as they made decisions. If, say, your blood pressure suddenly spiked, data-processing tools would warn them that you might be in trouble, and some sort of clinician—a nurse, perhaps—would reach out to you immediately, to check on your condition and arrange treatment as necessary. You could reach the team just as easily, with something as simple as a text message or an e-mail. You’d be in touch with them more frequently, most likely, but for much shorter durations—and, for the most part, with less urgency.

Sometimes, of course, office or hospital visits would be necessary, but that experience would be different, too—starting with the hassle of dealing with insurance companies. Watson has a button for submitting treatment proposals to managed-care companies, for near-instant approval, reducing the time and hassle involved in gaining payment authorization. The transformation of the clinical experience could be more profound, although you might not detect it: someone in a white coat or blue scrubs would still examine you, perform tests, prescribe treatment. But that person might have a different background than he’d have today. And as the two of you talked, your exam information would be uploaded and cross-referenced against your medical record (including the data from all those wireless monitors you’ve been toting around), your DNA, and untold pages of clinical literature.

The evolution toward a more connected system of care has already begun at some large organizations that use team models of care. One such institution is the Group Health Cooperative of Puget Sound, a nonprofit, multi-specialty group practice. Matt Handley, the medical director for quality and informatics, says that about two-thirds of Group Health’s patients now use some form of electronic communication, and that these methods account for about half of all “touches” between patients and the group’s doctors or nurses. “They set up their own appointments … They don’t need to call somebody and ask when I’m free,” Handley says. “They send messages to doctors; look up lab tests and radiology results; and order refills … The fascinating thing is that people of all ages are using it … I have people in their 90s who secure-message me.”

It’s a long way from Group Health to Health 2.0, and Handley is among those who are wary of the hype. Sure, the demos for products like Watson look great. They always do. But can such tools really winnow down information in a way that physicians will find useful? Can they effectively scour new medical literature—some 30,000 articles a month, by Handley’s reckoning—and make appropriate use of new evidence? Will they actually improve medicine? “While Watson could sometimes be helpful, it may actually drive up the cost of care,” Handley says, by introducing more possible diagnoses for each patient—diagnoses that clinicians will inevitably want to investigate with a bevy of expensive tests. A study in the journal Health Affairs, published in March 2012, found that physicians with instant electronic access to test results tended to order more tests—perhaps because they knew they could see and use the results quickly. It’s the same basic principle Handley has identified: if new tools allow providers to process far more information than they do now, providers might respond by trying to gather even more information.

Another reason for skepticism is the widespread lack of good electronic medical records, or EMRs, the foundation on which so many promising innovations rest. Creating EMRs has been a frustratingly slow process, spanning at least the past two decades. And even today the project is a mess: more than 400 separate vendors offer EMRs, and the government is still trying to establish a common language so that they can all “speak” to one another. “Our doctors have state-of-the-art electronic health-record systems,” says Brian Ahier, the health‑IT evangelist (yes, that is his real title) at the Mid-Columbia Medical Center, in northern Oregon, and a widely read writer on medical innovation. “But for clinical communication” outside the medical center, “they have to print it out, fax it, and then scan” what they get back.

But despite these risks and stumbling blocks, there are reasons to think the next wave of innovations might really stick. One is legislation enacted by the Obama administration. The 2009 Recovery Act—the $800 billion stimulus designed to end the economic crisis—set aside funds for the creation of a uniform standard for electronic medical records. It also made changes to Medicare, so reimbursement to doctors and hospitals now depends partly on whether they adopt EMRs and put them to “meaningful use.” The incentives seem to be working: according to a September 2012 survey by the consulting firm CapSite, nearly seven in 10 doctors now use EMRs. The trade publication InformationWeek called this tally a “tipping point.”

Under the Affordable Care Act, a k a “Obamacare,” Medicare will also begin rewarding providers who form integrated organizations, like Group Health Cooperative of Puget Sound, and groups that accept “bundled” payments, so that they are paid based on the number of patients in their care rather than for each service rendered. In theory, this financing scheme should encourage medical practices and hospitals to keep patients healthier over the long term, even if that means spending money up front on technology in order to reduce the frequency of patient visits or procedures. In other words, the new, digital model for health care should eventually become more economically viable.

One sign that medical care is in the midst of a massive transformation, or at least on the cusp of one, is the extraordinary rise in demand for information-technology workers within the health-care sector. All over the country, hospitals are on a hiring binge, desperate for people who can develop and install new information systems—and then manage them or train existing workers to do so. According to one government survey, online advertisements for health-IT jobs tripled from 2009 to 2010. And the growth is likely to continue. The Bureau of Labor Statistics estimates that in this decade, the health-IT workforce will grow by 20 percent. Most experts believe that such growth still won’t be nearly enough to fill the demand. But it’s the data revolution’s ability to change jobs within health care—to alter the daily workflow of medical assistants, nurses, doctors, and care managers—that might have the most far-reaching effects not just on medicine, but also on the economy.

Economists like to say that health care suffers from a phenomenon called “Baumol’s disease,” or “the Baumol effect,” first described half a century ago by the economist William Baumol, in collaboration with a fellow economist named William Bowen. In most occupations, wages rise only when productivity improves. If factory workers get an extra dollar an hour, it’s because they can produce extra value, thanks to better training or equipment. Baumol and Bowen observed that certain labor-intensive occupations don’t operate by the same principle: job productivity doesn’t rise much, but wages go up anyway, because employers need to keep paying workers more in order to stop them from pursuing other lines of work, in other sectors where productivity is rising quickly. That forces the employers to keep raising prices, just to provide the same level of service.

Over time, industries afflicted with Baumol’s disease tend to consume a larger and larger proportion of a nation’s income, because their cost, relative to everything else, climbs ever upward. The health-care industry has a textbook case of Baumol’s disease, and so far, technology hasn’t made much of an impact. Just as it still takes five string players to play a Mozart quintet (Baumol’s famous example), so it still takes a highly trained surgeon to operate on somebody. “We do now have robots performing surgery, but the robot is under constant supervision of the surgeon during the process,” Baumol told a reporter from The New York Times two years ago. “You haven’t saved labor. You have done other good things, but it isn’t a way of cheapening the process.” Likewise, a doctor in a clinic still sees patients individually, listens to their problems, orders tests, makes diagnoses—in the classic economic sense, the process of an office visit is no more efficient than it was 10, 30, or 50 years ago.

Now technology could actually change that process, not by making the exam faster but by enabling somebody else to conduct it—or to perform the test, or carry out the procedure. The idea of robots performing surgery or more-routine medical tasks with less supervision is something many experts take seriously—in part because, in the developing world, burgeoning demand for care is already pushing medicine in this direction. As part of an experimental program in Tanzania, rural health workers, many of whom have relatively little medical training, have access to a “decision-support tool” that can help them diagnose and treat illness based on symptoms. And thanks to an initiative called the Maternal Health Reporter, similar caregivers in India can submit patient information to a central data bank, then receive regular reminders about care for pregnant women.

“In Brazil and India, machines are already starting to do primary care, because there’s no labor to do it,” says Robert Kocher, an internist, a veteran of McKinsey consulting, and a former adviser to the Obama administration. He’s now a partner at Venrock, a New York venture-capital firm that invests in emerging technologies, including health-care technology. “They may be better than doctors. Mathematically, they will follow evidence—and they’re much more likely to be right.” In the United States, Kocher believes, advanced decision-support tools could quickly find a home in so-called minute clinics—the storefront medical offices that drugstores and other companies are setting up in pharmacies and malls. There, the machines could help nonphysician clinicians take care of routine medical needs, like diagnosing strep throat—and could potentially dispense the diagnoses to patients more or less autonomously. Years from now, he says, other machines could end up doing “vascular surgery, fistulas, eye surgery, microsurgery. Machines can actually be more precise than human hands.”

Nobody (including Kocher) expects American physicians to turn the keys of their practices over to robots. And nobody would expect American patients accustomed to treatment from live human beings to tolerate such a sudden shift for much of their care, mall-based minute clinics notwithstanding. But because of a unique set of circumstances, the health-care workforce could nonetheless undergo enormous change, without threatening the people already working in it.

Between the aging of the population and the expansion of health-insurance coverage under Obamacare, many more people will seek medical attention in the coming years—whether it’s basic primary care or ongoing care for chronic conditions. But we don’t have nearly enough primary-care doctors—in practice today or in training—to provide this care. And even if we trained more, we wouldn’t have enough money to pay them. With the help of decision-support tools and robotics, health-care professionals at every level would be able to handle more-complicated and more-challenging tasks, helping to shoulder part of the load. And finding enough nurses or technicians or assistants would be a lot easier than finding enough doctors. They don’t need as many years to train, and they don’t cost as much to pay once their training is finished. According to the Bureau of Labor Statistics, doctors’ median annual salary is $166,400, while nurses’ is $64,690 and medical assistants’ is $28,860. 

Health professionals at all levels tend to guard their turf ferociously, lobbying state officials to prevent encroachments from other providers. But the severe shortage of professionals to provide primary care means there should be plenty of work to go around. Already today, there’s a push within health-care-policy circles to more consistently allow providers to “practice at the top of their licenses”—that is, to let the people at each level of training do as much as their training could possibly allow them to do. That would enable higher-wage, more-highly-skilled professionals to focus on work that’s truly commensurate with their education. It would also reduce the cost of care. Watson and its ilk could help us take this concept further, by augmenting the capabilities of workers at every skill level. Physicians could lead large teams of mid- and low-level providers, delegating less complicated and more routine tasks. “Having nurses, with the assistance of these artificial-intelligence tools, [do more] frees physicians to perform the higher-level interventions, allowing everyone to practice at the top of their license,” Brian Ahier says.

That model actually isn’t so different from the collaborative approach that institutions like Group Health have been deploying with such success. “We focus on developing teams—teams of several doctors, physician assistants, nurse-practitioners, and/or nurses,” Matt Handley says. “Every day starts with a huddle: the team talks about the day and reviews a couple of topics and cases, figures out who is going to need what, from which provider, and so on.”

The providers with less medical training can be more technologically adept, anyway. “The doctor or clinician of course has the high analytic skills, makes the judgment calls, the diagnoses, prescribes medications,” says Catherine Dower, an associate director at UC San Francisco’s Center for the Health Professions. “But the medical assistants are frequently the ones who can actually use this new technology really well, including tele-health—they can get bio-feeds from patients sitting at home, they can tap insulin and cardiac rates. And then, as this information is fed into a central site, the medical assistant can read and make a decision on which patient should come in and be seen by the doctor, and which one needs some minor modification”—whether that means adjusting medication or scheduling a visit to discuss more-significant changes in treatment or therapy.

For the health-care system as a whole, the efficiencies from the data revolution could amount to substantial savings. One estimate, from the McKinsey Global Institute, suggested that the data revolution could yield onetime heath-care savings of up to $220 billion, followed by a slower rate of growth in health-care costs. Total health-care spending in the U.S. last year was $2.7 trillion, so that would be roughly the equivalent of reducing health-care spending by 7.5 percent up front. That’s the best reason to believe that the data revolution will make a difference, even if it never lives up to the hype of its most enthusiastic proponents. The health-care system is so massive, so full of waste, so full of failure, that even a marginal change for the better could save billions of dollars, not to mention quite a few lives.

And, in a small way, it could help us begin to fill the hole that’s developed in the middle class. David Autor, an economist at MIT, has noted that for the past generation, technological change in the U.S. has tended to favor highly skilled workers at the expense of those with mid-level skills. Routine clerical functions, for instance, have been automated, contributing to the hollowing-out of the middle class. But in the coming years, health care may prove a large and important exception to that general rule—effectively turning the rule on its head. “Look at physician-assistant positions,” Dower says. “After completing a rigorous academic program that's modeled on medical school but takes less time, PAs can provide many primary- and specialty-care services at starting salaries of about $77,000.”* If technological aids allow us to push more care down to people with less training and fewer skills, more middle-class jobs will be created along the way.

“I don’t think physicians will be seeing patients as much in the future,” says David Lee Scher, a former cardiac electrophysiologist and the president of DLS Healthcare Consulting, which advises health-care organizations and developers of digital health-care technologies. “I think they are transitioning into what I see as super-quality-control officers, overseeing physician assistants, nurses, nurse-practitioners, etc., who are really going to be the ones who see the patients.” Scher recognizes the economic logic of this transition, but he’s also deeply ambivalent about it, noting that something may be lost—because there are still some things that technology cannot do, and cannot enable humans to do. “Patients appreciate nonphysician providers because they tend to spend more time with them and get more humanistic hand-holding care. However, while I personally have dealt with some excellent mid-level providers, they generally do not manage complex diseases as well as physicians. Technology-assisted algorithms might contribute to narrowing this divide.”

“I don’t think physicians will be seeing patients as much in the future. I think they are transitioning into what I see as super-quality-control officers.”

Even Watson, which has generated so much positive buzz in medicine and engineering, has its doubters. “Watson would be a potent and clever companion as we made our rounds,” wrote Abraham Verghese, a Stanford physician and an author, in The New York Times. “But the complaints I hear from patients, family and friends are never about the dearth of technology but about its excesses.”

Marty Kohn, from the Watson team, understands such skepticism, and frequently warns enthusiasts not to overpromise what the machine can do. “When people say IT can be transformative, I get a little anxious,” he told me. Partly that’s because he thinks technology can’t change an industry, or a culture, if the professionals themselves aren’t committed to such a transformation—Watson won’t change medicine, in other words, if the people who practice medicine don’t want it to change. As a physician, Kohn is careful to describe Watson as a “clinical support” tool rather than a “decision making” tool—to emphasize that it’s a machine that can help health-care professionals, rather than replace them. “Some technologies are truly transforming health care, providing therapies that never existed before. I don’t view IT that way. I view IT as an enabler.”

Still, Kohn has reconciled himself to hearing people talk about Watson as if it were a person—he says he’s now used to answering the question “Who is Watson?” rather than “What is Watson?” He also likes to tell a story about a speech he gave in Canada, one that, like the Las Vegas presentation, attracted more people than the room could hold. That evening he called his wife, to tell her about the enthusiasm. “That’s really great, Marty,” he recalls her saying. “Just remember, they were there to meet Watson, not you.”

* Note: This quote has been modified from the original to correct an error about the amount of schooling required for physician assistants.

Presented by

Why Principals Matter

Nadia Lopez didn't think anybody cared about her middle school. Then Humans of New York told her story to the Internet—and everything changed.

Join the Discussion

After you comment, click Post. If you’re not already logged in you will be asked to log in or register with Disqus.

Please note that The Atlantic's account system is separate from our commenting system. To log in or register with The Atlantic, use the Sign In button at the top of every page.

blog comments powered by Disqus


A History of Contraception

In the 16th century, men used linen condoms laced shut with ribbons.


'A Music That Has No End'

In Spain, a flamenco guitarist hustles to make a modest living.


What Fifty Shades Left Out

A straightforward guide to BDSM

More in Health

More back issues, Sept 1995 to present.

Just In