Digital Diagnosis

The art of medicine takes on the science of technology.

As American troops roared north through the Iraqi desert into Baghdad in the spring of 2003, their vehicles joined the wind in kicking up clouds of dust. Soldiers inhaled fine particles of sand and dirt. Between March and August 2003, 10 came down with a rare strain of pneumonia. It killed two of them.

Doctors never established the exact cause of the outbreak, but they did notice that nine of its victims had significant exposure to fine particulate matter and all of them smoked, the majority having started smoking not long before they became ill. A team from the Walter Reed Army Medical Center warned physicians to look out for symptoms of acute eosinophilic pneumonia among soldiers, and to be aware that the best treatment is steroids, not antibiotics.

Army Col. Bart Harmon, chief medical information officer of the Military Health System, describes the pneumonia outbreak to illustrate the benefits of health information technology. In an office building in Falls Church, Va., that serves as the headquarters of military health IT, he stands in front of a projected slide depicting a bar chart of a typical disease outbreak. This, he says, pointing to the rising bars where the number of cases begins to climb, is where the military wants to head off outbreaks. Not here, he says, where the numbers spike-a full-blown epidemic. Close monitoring gave the military an edge in spotting and identifying the rare pneumonia in Iraq. In the civilian world, a universal electronic medical system could cut the time between outbreak and detection by two weeks, Harmon says.

How to do it? Swap paper for computer records, put laptops, personal digital assistants or tablet computers in clinicians' hands. Get them to stop writing medically significant notes in idiosyncratic shorthand. Record patient symptoms in standard fashion, Harmon says. Have a private and secure means for aggregating that data so trends become apparent.

Of course, setting up an electronic health IT system hasn't been easy even within the confines of the military. Only now, after working for 10 years, spending hundreds of millions of dollars and often forcing cooperation between doctors and technology specialists, is the Defense Department almost ready to fully deploy its health IT system.

In April 2004, President Bush ordered the Health and Human Services Department to actively promote adoption of health IT. The plan: Within 10 years, a majority of Americans should have their electronic health records accessible nationwide, probably through a federation of health databases. The benefits: lowered medical costs (now 16 percent of the gross national product), better health care through instantaneous communication of medical histories and improved detection of disease outbreaks. All this is supposed to be accomplished without massive federal expenditures.

Savings vs. Cost

Doctors already using electronic health systems can't help but gush about them. Dr. Richard Baron is an internist at a small practice in Philadelphia that switched to electronic medical records in mid-2004. Transitioning from paper was not easy, he admits. Previously unnoticed variations in clinical style became the subject of tense office discussions. Members of the practice didn't disagree about medical practices, but rather on how to prioritize them. "It's a challenge to get everybody on the same page for that, because we're talking about the actual work people do all day," Baron says. The practice reduced its patient load during the adjustment period, losing some clients for good. But Baron wouldn't go back to paper for anything. "We were able to run a report of all the women between 50 and 65 who have not had a mammogram in the past year," he enthuses. On their next visits, the women automatically were reminded. "If they came in for pinkeye, they would get a sheet that said, 'By the way, you're overdue for a mammogram.' "

Financially, however, adoption of the new system was a wash, Baron says. Tech support costs ballooned. True, the practice no longer pays for a transcribing service, but it now has piles of fussy electronic equipment that needs constant attention and replacement.

A key assumption underlying the presidential vision of nationwide health IT is that it will be adopted without federal subsidy. Bush's executive order instructed HHS not to rely on additional federal resources. Instead, the department will encourage adoption by promulgating national standards for health IT. An executive order signed in August brings the government's purchasing power to bear by mandating that should physicians who contract with the government (including those receiving payment via Medicare) buy health IT equipment, it must meet HHS standards. But the executive order doesn't require clinicians to buy the equipment.

Going electronic will save money, proponents of the market-based approach say: A paper record medical system generates waste. Administrative costs are high. Too often, misplaced test results cause repeats. Preventive care can be improved by electronically searching records-have all the elderly patients gotten their flu shots yet? But will such savings reach the bottom line?

A 2005 RAND Corp. study found that if 90 percent of hospitals and medical practices adopted health IT and produced efficiency gains at a rate similar to other industries' adoption of technology, annual savings could amount to $346 billion after the adjustment period. Without the IT-fueled productivity gains, the study predicts savings of $81 billion a year. Critics blasted the study as a public relations exercise paid for by the medical IT industry, which subsidized it. The Wellesley, Mass.-based Center for Information Technology Leadership, whose mission is to "communicate the financial and clinical value delivered by specific technologies and strategies," predicted in 2005 that a fully interoperable nationwide health IT system would generate $77.8 billion in savings each year.

Not true, says Clifford Goodman, senior scientist at the Falls Church, Va., health care consulting firm The Lewin Group. "I concur that it will lead to efficiencies," he says. The savings estimates may or may not turn out to be accurate, depending on the optimism of those studying them, he says. He agrees that savings eventually will outweigh implementation costs. But as improved medical technology pushes down per-unit costs, Goodman notes, more people buy health care, increasing overall spending. So IT efficiencies probably will add to the national health bill, not reduce it. "The health care system always finds way to spend money [saved]," Goodman says.

Vocabulary Test

Standing in a hallway of the Veterans Affairs Medical Center in Washington, Dr. Divya Shroff gets excited about what's on the screen of a laptop perched on a wheeled cart. With a few mouse clicks, she can see the results of a patient's lab tests, X-rays and colonic endoscopy-from 1993. VA, like De-fense, is on the forefront of health IT. Computerized since the mid-1980s, the improvements in VA's system are easy to see, users say. "Other hospitals, it's like running on sand . . . because you're spending so much time going to medical records, pulling the last admission chart," says Shroff. "Here, all of a sudden you're running on pavement again."

But if the Defense and VA departmental systems show the promise of health IT, they also illustrate why projections of cost savings might be unrealistic. In developing its savings estimates, the Center for Information Technology Leadership assumed various levels of system interoperability. The highest level (for which it predicted the $77.8 billion number) is for a system using interoperable machine-readable data in which all medical terms are structured using a restricted vocabulary. The largest benefits of health IT accrue when the computer can read all the data entered and therefore plot it graphically, automatically search it for pre-existing conditions and scan for trends. Instead of paging through charts, whether paper or computer images of paper originals, machine-readable data lets doctors automatically put medical information in context. "You have much more awareness of a whole variety of clinical information," Baron says.

Both the Defense and VA health programs are vertically integrated-the departments pay the doctors, own the hospital facilities and control the software. They have unparalleled power to shape their technology environment. And yet, their health IT systems are encased in sturdy data stovepipes, and the agencies struggle to communicate with each other.

The struggle isn't because the systems lack data standards. In fact, there are many of them. "We're at the stage where industry was before interchangeable parts," says Dr. Howard Bleich, a Harvard Medical School professor of medicine specializing in health IT. "You can transmit the data from one [system] to another, but the data has another meaning from one to another." Two hospital campuses running the same commercial software might not be interoperable. The cause isn't deliberate resistance to interoperability or a Luddite aversion to technology. The reasons go to the heart of medicine.

Laboratory test results, for example, can mean different things in different clinics and parts of the country, depending on demographics, cost or past practice. A serum creatinine test is a simple and common method of evaluating kidney function. But some tests isolate creatinine levels and other don't. Measuring a basket of similarly reacting chemicals is cheaper, but less precise. Thyroid hormone levels can be measured by a combination of thyroid-stimulating hormone and thyroxine, or just triiodothyronine. Which is the right one? "Different hospitals, different doctors, different administrators make different decisions," says Bleich. The problem, he adds, is that the differences can be medically significant.

"Is the answer that we shouldn't have electronic medical records, or is the answer that the laboratory people should agree what units they're going to report in?" asks Baron. But complete standardization of tests might not be beneficial. Certain diseases are more common in certain parts of the country, requiring even simple lab procedures to be more sensitive to certain indicators.

The federal response has been to promulgate standards for data exchange under a formal process led by the federally funded Healthcare Information Technology Standards Panel. "It's like saying I wrote a song, and whether I put it on cassette tape or CD or an 8-track tape, it's the same song," says panel chairman Dr. John Halamka, who also is chief information officer at Harvard Medical School. Electronic transmission between clinics and labs already can and does occur, but it requires tens of thousands of dollars and probably months of tweaking in order craft an interface-the opposite of interoperability.

Halamka's panel in October initially approved interoperable lab result standards that account for variation in lab test components, he adds. Health care organizations can record data elements however they want, but for transmission, the data must be translated into a common form to qualify for a certification the government says it will enforce to the extent it can.

But data translation adds complexity and cost. It requires standards to be constantly upgraded, and some say it could stifle innovation because standard-setting bodies rarely are at the cutting edge of an industry or profession.

No Translation

If the prospect of standardizing lab results data is daunting, then consider the difficulty of translating doctors' notes. When writing down patient symptoms, most doctors don't use a standard terminology. They make notes to themselves in shorthand or use phrases their clinics have invented. There are detailed, standardized vocabularies intended to reduce ambiguity in doctors' notes; creating interoperable machine-readable health records requires such a restricted terminology if the technology is to live up to its full promise.

But of course, there is more than one vocabulary. Defense uses Medcin, which was created by Medicomp Systems in Chantilly, Va. In 2003, Health and Human Services licensed SNOMED Clinical Terms, produced by the College of American Pathologists. Do they communicate? "You can't get chicken salad from chicken manure," says Defense's Harmon. For example, data translation from Medcin to SNOMED back to Medcin would be extremely imperfect.

Not every doctor even believes in taking notes using a controlled vocabulary. "Some physicians believe doing so communicates a standard, but not what's wrong with a specific patient," says Dr. Ross Fletcher, the Washington VA Medical Center chief of staff. Though imposing a note-taking vocabulary is conducive to data mining for outbreak detection, VA's electronic medical records contain space for doctors to type in their own notes, in their own vocabulary. VA eventually does assign a machine-readable diagnostic code to describe patient encounters-International Statistical Classification of Diseases and Related Health Problems (commonly known as ICD)-but ICD codes operate at a higher level of abstraction than Medcin.

It all adds up to policy and market failure, says Dr. David Himmelstein, a Harvard Medical School associate professor of medicine. The marketplace has created more standards than can be dealt with through the hands-off federal approach. "We don't tell airplanes to use whatever radio frequencies that they want and have different standards and tell them we're going to try to make those standards work together," he says. "We actually set some standards, and we adopt them."

Himmelstein's solution is subsidization of private sector adoption of health IT. The Bush health IT effort "is partly just a smoke screen to distract from the fact that they're doing nothing about health care more generally," he says. Without more aggressive federal intervention, nationwide health IT adoption won't happen, charges J.D. Kleinke, executive director of Omnimedix Institute, a Portland, Ore.-based nonprofit health care IT research and development organization. There is a growing market for health IT systems, but unless implementation policy changes, it will be centered in wealthy areas "with affluent doctors who serve affluent patients on the right side of the digital divide," Kleinke says.

Rating Doctors

Even those who praise the Bush administration's health policy agree that it might not be sufficient to spur nationwide health IT adoption. Their solution is to change the way the federal government pays physicians. "Payment should be tied to the quality outcomes that they deliver," says David Merritt, a project director at the Center for Health Transformation, a Washington consulting firm founded in 1999 by former House Speaker Newt Gingrich. The 2004 executive order mandating IT adoption alludes to using quality measures, saying technology will foster "a more effective marketplace, greater competition and increased choice through the wider availability of accurate information on health care costs, quality and outcomes."

Combining health IT and pay for performance creates a two-way feedback loop, according to Merritt: Pay for performance decouples compensation from mere numbers-patient visits, tests or procedures performed-prodding doctors to take a broader approach to health care, he says. But successful performance-based pay must be predicated on a health IT system that aggregates medical data, detects whether appropriate practices are used consistently and checks for medical outcomes.

Performance-based compensation is a touchy subject in medical circles, however. Hardly anyone defends the current payment system, but setting up a new one that rates doctors partly on whether they perform certain procedures ignites opposition. "There exists a set of standards that if I apply uncritically across my patients, bad things will happen," says Baron, whose support for performance pay would depend on incorporating risk adjustment, so doctors are not penalized for seeing very sick patients or from deviating, based on patient need, from protocols determined by general demographics.

Adoption of health IT need not imply performance pay, of course. "Health IT is being adopted around the world much faster than here, in places that have nothing to do with consumer-driven health care," Himmelstein says. But "the main thing [the Bush administration] seems to want computers to do is to keep track of even more complexity that they want to introduce in health financing," he adds.

Merritt of the Center for Health Transformation says that's not the case at all. "Americans have a right to know that information. We're paying the health care bills in this country."

The medical industry is not the first to grapple with interoperability, or the first to do so in a field where achieving it could save lives. First responders, for example, have tried for decades to find radios that can communicate across levels of government and jurisdictions.

Limited interoperability does exist-but only on a restricted scale, despite the time, money and widespread support for it. And this in emergency communications, a setting where no one is questioning the motivations of interoperability boosters. Says Kleinke: Getting the medical industry to adopt health IT "requires a president who actually understands the problem," rather than simply mandating that it be solved in 10 years.