Doctor

What your doctor doesn't know could kill you

A computer program that provides vast amounts of information for diagnosing and treating patients could revolutionize the practice of medicine. So why won't physicians use it?

By Chris Gaither, Globe Staff, 7/14/2002

Dr. Lawrence L. Weed, developer of the Problem Knowledge Coupler software, at his office in Burlington, Vermont.
(Photo by Lane Turner / The Boston Globe)

r. Harold D. Cross, an occupational health and emergency physician in Beaufort, South Carolina, recalls few cases during his five decades of practicing medicine as bizarre as that of the flailing plumber. The 47-year-old man had begun to act strangely at work, waving his arms involuntarily for a minute at a time and acting dazed. A few times each night for three years, the plumber would begin thrashing about in bed, then get up and pace for a few moments before climbing back under the covers. The man had been seen by a neurologist, who diagnosed him with a sleep disorder and prescribed a drug. It didn't help, so he stopped taking it and turned to Cross.

Cross has a trick, an unusual tool he sometimes pulls out of his virtual black bag. The tool is not a stethoscope, which amplifies his ability to hear a heartbeat, or an MRI, which makes up for his inability to see through flesh. Instead, it is a piece of computer software that makes up for the limits of the human brain. The software, called the Problem Knowledge Coupler (PKC), was conceived by an old Vermont friend of his, Dr. Lawrence L. Weed. Instead of listing the symptoms of a disease and asking a doctor to choose the closest fit, as some medical Web sites do, Weed's program asks a doctor to first answer a long list of questions about the patient's troubles. Then, up comes the most likely diagnoses and ways to test them out. The program helps doctors match (or "couple") the patterns of a patient's problems with the relevant knowledge that exists, perhaps buried deep in a textbook or journal article, to recognize and treat those problems.

Firing up Weed's software, Cross plugged the plumber's problems into the "sleep disorder" function. In the end, Cross says, the man's symptoms satisfied all three findings for nocturnal paroxysmal dystonia, a rare disorder that Cross had never heard of. The plumber and his wife printed out the findings and brought them to their primary care doctor, who had referred them to Cross. As Cross tells it, the doctor accepted the results and ordered the prescription recommended by the cp10PKCcp11 program: a low dosage of anti-seizure medication, taken every night before bed. The plumber and his wife have slept better ever since.

But, according to Cross, the neurologist who originally diagnosed the case as an REM sleep disorder had a very different reaction to the use of the software. When the plumber and his wife handed that doctor the PKC printouts, he shuffled them, left the room, and, Cross says, "returned with a very hostile, angry disposition." Viewing the results as computer-generated quackery, he refused to back down from his original diagnosis.

"The issue is not who is right but how the patient can be helped," Cross wrote me in an e-mail. "Physician response to new information, especially when they have not done their homework, sometimes results in pretty violent behavior. That is embarrassing to report, but I have been at this quite a while, and it is a fact of life."

One of the great myths of modern medicine, often perpetuated by the doctors themselves, is that the physician knows best. As health care organizations seek ways to make medical care more reliable, some are installing systems to automate billing and prescriptions. Doctors are going online to answer e-mail from patients. And a few providers have sought for years to create databases of patient records that can be easily accessed by physicians across a health care organization. But Weed, the creator of Knowledge Couplers, argues that none have truly explored the most basic limitation confronting doctors: They are trying to do something they simply cannot do.

Humans, Weed argues, cannot consistently process all of the information needed to diagnose and treat a complicated problem. The more information the physician gets about a patient, the more complex the task becomes. A doctor working without software to augment the mind, he argues, is like a scientist working without a microscope to augment the eye.

Some accomplished physicians and scientists who have explored ways to use artificial intelligence to diagnose patients say that it is impossible with today's technology. Many other doctors strongly oppose the mere concept, calling software incapable of matching their expertise; computers merely get in the way, they argue. But a small band of physicians and Weed's company's biggest customer, the US Department of Defense, have begun to use the Knowledge Couplers, and an early study suggests that their patients are healthier for it. If the software catches on, Weed's ideas may forever change the way doctors make decisions, removing much of the mystery and leaving us, the patients, with more control over our care. Weed's supporters say the medical industry will one day recognize the genius behind the software, much as it recognized the promise of Weed's first major innovation, which changed medicine four decades ago.

arry weed looks like an aging eagle. His nose is sharp, and his head shaved close, leaving white stubble. He rolls the sleeves on his dress shirts well past mid-biceps and wears his pants cinched low, which makes his torso look abnormally long. When he walks for exercise, he walks backward, because, he says, it strengthens muscles that are unused by regular strides. When I visited Weed at his office in Burlington, Vermont, I had to run to keep up with him as he zipped (forward) down the hallways. Weed is 78; I am in my late 20s.

Weed takes great pleasure in arguing. To extend his range, he devours literature of all kinds. The first time we spoke about his views on health care, he referred to Karl Marx, Winston Churchill, the space shuttle Challenger, George Eliot, Aristotle, Pope John Paul II, Tolstoy, and Wilt Chamberlain. When making an important point about what he perceives as the ills of modern health care and how computers can cure them, he reaches out and grabs my arm with strong fingers, like talons.

In 1956, Weed swooped into Eastern Maine General Hospital like a raptor into a henhouse. The Bangor clinic had hired him, at the age of 32, to become its medical director. The medical residents never knew what hit them. Weed believed, and eventually persuaded most of them to believe, that during their six or seven years of medical training at fine universities, they had never learned to keep a proper record of what they saw and how they treated it. And in that lapse, Weed contended, originated many of the problems of modern medicine, as well as their solution.

Physicians first started keeping patient records in the 19th century. Most patient records looked like spaghetti tossed against a wall, indecipherable to all but the doctor himself, and sometimes to him as well. Physicians would often write down things that supported their diagnosis, omitting symptoms that failed to fit the hypothesis. Anyone who came along later and read the chart would see only what the first doctor concluded, not the symptoms the doctor may have dismissed to reach his conclusions.

In Bangor, Weed began his crusade to turn doctoring into hard science. He came up with a new system, which he called the problem-oriented medical record. It places relevant information into four categories, known as SOAP notes: subjective (the patient's history), objective (the observed physical findings and test results), assessment (the diagnosis, with an explanation of the thought process), and plan (more testing and/or the treatment). The rigid structure of the record keeping encouraged doctors to better organize their examinations and ask questions they might normally have omitted.

"Everybody thought that it was a heck of a thing that someone was addressing the issue of how we keep records," says Dr. Donald A. B. Lindberg, director of the National Library of Medicine. "In truth, he was probably the first to seriously reexamine it in a hundred years. You see his ideas across medicine now."

Weed's system for record keeping was slow to catch on. It was not until the late 1960s, after he had left Maine for a professorship at Case Western Reserve University in Cleveland, that several prominent physicians, including Dr. J. Willis Hurst, then a professor at Emory University in Atlanta and Lyndon B. Johnson's personal cardiologist for 18 years, recognized its benefits and began trumpeting the cause. The problem-oriented record earned Weed fame within medical circles. Medical schools across the country began teaching Weed's method. The National Academy of Sciences elected him to its ranks and, many years later, in 1995, awarded him the Gustav O. Lienhard Award for distinguished contribution to health care in the United States.

But even Weed realized there was a drawback to the problem-oriented record: It took too long to fill out. In 1968, Weed accepted multimillion-dollar grants from the National Institutes of Health to create a massive, computerized extension of the record-keeping system, one that would require less time. The project was known as the Problem-Oriented Medical Information System, or PROMIS. The next year, Weed moved it from Case Western to the University of Vermont, where, within five years, he had an entire patient ward of the university's medical center equipped with minicomputers running the computerized medical records. The university's doctors could document each step of the path they took in diagnosing and treating each patient. And for the first time, a medical director - Weed - could sit at his desk and review each of his doctor's moves, each possible diagnosis explored, each lab test given, each drug prescribed.

But the more he could see, the more depressed he grew. Doctors skipped steps to save time, sometimes failed to ask the right questions, failed to recognize patterns, and overlooked obvious possibilities under pressure. "Why are we so thrilled that drug orders are going at the speed of light to the pharmacy and coming back, if they shouldn't have been ordered in the first place?" Weed asked. "What are we computerizing this for? We're computerizing chaos." Instead of improving medicine, Weed believed, PROMIS was simply exposing the limitations of the human brain. Asking doctors to make decisions using their memories alone, he says, is like asking a travel agent to book a client's journey from memorized flight schedules.

Disheartened after running PROMIS for more than a decade, in 1982 Weed retired from the University of Vermont and, with his wife, Laura, a son, Chris, and a friend named Richard Hertzberg, created the PKC Corp. He secluded himself in Burlington with a small team of computer programmers to develop new software. For two decades, the father of the problem-oriented medical record, as he is known, almost disappeared from the medical world's radar. But now Weed believes his newest innovation, the Problem Knowledge Coupler, is finally ready for mainstream use.

mong doctors, stories circulate about crack physicians who make brilliant diagnoses on cases that had puzzled their colleagues. These hotshots are able to perform superbly the physician's most fundamental task: Observe the patient, make a mental list of the symptoms, then match those symptoms with the patterns of a diagnosis. But what happens when you get sick and aren't seen by such a doctor? As Dr. Donald M. Berwick, a clinical professor of pediatrics and health care policy at the Harvard Medical School, says, genius diagnosticians "make great stories, but they don't make great health care. The idea is to make accuracy reliable, not heroic."

Weed reached that conclusion three decades ago. In a 1969 book, Medical Records, Medical Education, and Patient Care, he wrote that the medical profession must fix itself before it sinks "deeper into the quagmire of raw information upon which our footing is already insecure." That is true today, more than ever. Medical research has progressed in leaps and bounds, but it far outpaces physicians' abilities to put the new information into daily practice. If a doctor is a hungry man and information is the food, Weed might say, the medical profession is dumping two tons of wheat on his plate and telling him to make a sandwich.

Researchers have found that many medical errors are caused by the system - a pharmacist misreads a doctor's handwriting on a prescription and dispenses the wrong drug, or the doctor prescribes a drug that interacts badly with some other medication. Hospitals have begun automating the prescription process. But with the exception of some software that can recognize patterns in blood tests or electrocardiograms, little attention has been paid to automating the physician's diagnostic process. It's not incredible that so many medical errors are made, Weed says; it's incredible that doctors are able to achieve as much as they do within the existing system.

Medical schools teach a kind of "best guess" or "average man" method: Ask a patient about his personal and family history, perform a physical examination, then combine the data into a list of possible reasons for the problem (this list is called a differential diagnosis). Treat the patient for the most statistically probable of those reasons and, in case you are wrong, also treat for the most serious possibility. For most patients, the most likely guess is right. But for a significant minority, it turns out to be wrong.

Moreover, what if a doctor doesn't ask the right questions in the first place? Doctors are human, after all, and they begin drawing conclusions very early. "Does the public realize that doctors can't possibly do what the public thinks they can do?" Weed asks. "In the field of medicine, for many, many years, we have all been pretending we do not know that everybody has been pretending."

So how do we remove the memory limitations, the processing limitations, and the human bias from medicine? Look at "Upon This Age," a poem by Edna St. Vincent Millay, which Weed recites from memory to anyone willing to listen:

Upon this gifted age, in its dark hour,
Rains from the sky a meteoric shower
Of facts . . . they lie unquestioned, uncombined.
Wisdom enough to leech us of our ill
Is daily spun; but there exists no loom
To weave it into fabric.

With his software, Weed is trying to build that loom for physicians, to weave the information into a fabric of good medical care. In the November 1999 edition of the British Medical Journal, Weed published the distillation of 50 years of his experience, "Opening the Black Box of Clinical Judgment." In the five-part article, co-written by his son Lincoln, an attorney, Weed said, "Medicine requires a new division of intellectual labor, a division between tools that retrieve and process information and users who apply judgment and values to arrive at decisions."

The concept behind the PKC program is simple: Don't start drawing conclusions until you have all your facts straight. Say, for example, that someone complaining of a headache visits a doctor. Instead of asking a few simple questions and guessing which cause is most likely, the doctor or nurse sits the patient down in front of a computer to complete the "headache coupler" (there are about 80 couplers, covering topics from depression to general health). The program takes the patient through a long list of questions. (Does your vision or speech change before the headaches come? Does your eyelid droop when the headache hits?) It then instructs the doctor to enter laboratory or examination results.

The more information that a physician gets about a patient, says Weed, the more complex the task becomes. A doctor working without software to augment the mind, he argues, is like a scientist working without a microscope to augment the eye.

Once the questionnaire is completed, the doctor clicks on "primary options." The computer goes into its vast database of medical knowledge - originally compiled by PKC Corp. librarians and updated every six months on CDs or over the Internet - and matches up the patient's answers with all the known symptoms for various causes of headaches. For example, a patient may match 10 of 14 signs of classical migraines, or two of two for sinusitis, or one of six for a cerebellar hemorrhage. But the coupler may also warn the doctor to check for other life-threatening conditions that, because at least one sign is present, need further investigation before they can be ruled out. The doctor can click on each condition to learn more about it or get references to medical journals in which the condition is described.

The physician's new job, Weed says, is to perform any follow-up tests that the software recommends, then discuss the findings with the patient, and, again using the software, devise an appropriate treatment plan. Weed believes doctors should use the diagnostic software with every patient who comes into the office, not just with those who present challenging diagnoses.

This has prompted one of the strongest criticisms of Knowledge Couplers: Many doctors would probably complain that sitting in front of a computer to answer a hundred questions for a flu case takes far too much time. Weed says this reluctance stems from the false confidence that doctors entertain about their abilities. Patients and nurse practitioners can answer most of the questions before the doctor's visit begins. Weed argues that regular use of couplers can catch problems early and, when tied to a database, help chart a patient's health over time. Taking 20 minutes to get the facts straight from the start, he says, saves time and money down the road.

His vision isn't that of a science fiction writer - he doesn't picture a scenario in which computers replace doctors. Computers can't perform a physical examination or interpret a chest X-ray film. Willis Hurst, the Emory University professor, calls Weed a genius and says the software should allow doctors to focus on their clinical skills. "Knowledge coupling is the wave of the future," Hurst says, "but the information that is entered into the computer must be accurate."

The idea is to let computers do what they do best (process information) and let humans do what they do best (perform examinations and navigate the uncertainties of the computer findings). Then, Weed says, once all the facts have been collected, doctors and patients can make a shared decision. Howard Pierce, PKC Corp.'s chief executive officer, says he does not worry about possible malpractice suits against the company, because the software, like an enormous textbook, shows doctors their options instead of recommending actions. "The implications of the tool are very large and challenge the notion of what a doctor is," Pierce says. "But once the doctor rolls over and accepts it, he can go right back to being a doctor. Just a better doctor."

he doctors were bewildered. For almost seven months, a 15-year-old girl had struggled to catch her breath when she exercised. Scrawny and pale, she had lost 15 pounds and her menstrual periods had stopped. Too tired to drag her weary body through a school day, she missed three months of classes.

A physical examination at a hospital yielded no answers, so she was sent home. But over the next few weeks, she visited the hospital time and again, complaining of troubles such as nausea, stomach pain, and dehydration. Chest X-rays, brain scans, blood tests, abdominal ultrasounds, internal scopes - nothing explained her condition. And she was getting worse.

Finally, after several false leads, another blood test revealed abnormal levels of serum electrolytes, a sign of a rare hormone deficiency called Addison's disease (it counts John F. Kennedy among its sufferers). The girl's doctors had missed it, because the symptoms were atypical. Once they identified Addison's, doctors began hormone replacement therapy, and the patient improved.

The tale of this girl was described in the January 1996 edition of The New England Journal of Medicine. The article, "Just in Time," was an exercise in clinical problem solving. Two doctors, David J. Keljo and Robert H. Squires Jr., revealed each step of the patient's illness to an anonymous "expert clinician," who recorded his suspicions. By the end, the doctor's list of possible causes included inflammatory bowel disease, poisoning, Munchausen's syndrome, an abdominal tumor, and pancreatic cancer. Addison's disease was never considered in the exercise; in real life, according to the article, Addison's disease was not considered until the girl almost died.

"In retrospect, the diagnosis seems obvious," the researchers wrote. "A rare diagnosis that is obvious in retrospect, however, is often not so obvious prospectively."

From the moment the article crossed his desk, Weed began holding it up as an example of the medical system's failings. Both the clinician's and the girl's physicians' ignorance of the medical research and their eagerness to start guessing without all the facts (two traits common among most doctors, Weed would argue) prevented them from combining the individual symptoms into the correct diagnosis. With computers, Weed says, the doctors would have spotted Addison's disease almost from the start.

I am no doctor. But just for fun, I decided to try the couplers myself. Seated on a leather couch in a coffee shop with my notebook computer, I went through the Addison's disease exercise that the New England Journal put its physician through. By the time I had finished, at the top of the screen, under the heading "Medical Conditions That May Escape Routine Work-up," was: "3 of 6, adrenocortical insufficiency (e.g., Addison's disease)." I clicked on it. As evidence, the program listed decreased appetite, hyper-pigmentation, and significant weight loss. Under evidence not present for Addison's, the program listed low serum sodium levels, low blood pressure, and high serum potassium levels. Though I didn't know what the girl's blood pressure should have been, a doctor almost certainly would have; if it was low, her symptoms would satisfy four of six criteria for Addison's.

A doctor could then have followed the program's recommendation and tested the girl's serum sodium and potassium to see whether her symptoms matched the textbook cases. The program also listed other possible syndromes and disorders. Only pancreatic cancer was close, with three of six, but the remaining symptoms didn't fit. If I had seen this girl and run her facts through the coupler, I would have said she probably had Addison's disease and ordered tests to confirm it.

Weed will tell you that he could teach a 15-year-old to use his software. He argues that medical schools, or "diploma mills," as he calls them, teach students to remember the answers to questions they never asked and memorize observations that they never made. That, he says, is the antithesis of scientific training.

Weed shared his views on medical training and diagnosis during my visit to his offices, a former wool mill on the Winooski River in Burlington. Darkness crept early into his office that day, but he refused to turn on the overhead lights. To me, sitting in darkness on the other side of his desk, Weed looked like a hysterical shadow puppet, waving his arms about as he complained for three hours about what the medical profession fails to understand about computers and doctoring. He finished with a flourish.

"Medical students will say, `You mean the human mind can't process as well as a computer?' Well, the human mind can make every calculation that you need to get a rocket from here to the moon. That's true. But you can't do it when the rocket's in motion. You can't do it at that speed. When a quantitative difference becomes big enough, it has qualitative implications. Do you see what I mean?"

I was beginning to. But I was paying the price. When I arrived at the PKC Corp. offices that day, Pierce, its CEO, had shaken my hand and said: "It's your first time meeting with Larry, huh? I have aspirin to help with the headache you're going to get." Sure enough, as I sat there talking with Weed, my head began to throb. By the end of the evening, spots danced in front of my eyes and my temples felt about to burst. Maybe it was the lack of light, maybe it was Weed's ideas. Then again, maybe I was just hurrying to a diagnosis on limited evidence.

ot surprisingly, Weed's views have been received coldly by physicians in most medical schools and teaching hospitals, the places in which new technologies make their way into popular use. Dr. Jerome P. Kassirer, a distinguished professor of medicine at Tufts University and the editor of The New England Journal of Medicine from 1991 to 1999, worked on computer-aided diagnosis for two decades before giving up in the 1980s. He has not seen Weed's software but knows doctors who have, and he still believes that no computer software can properly diagnose a patient. Such programs sometimes remind doctors to check for things they might otherwise have forgotten, he says, but they can also prompt doctors to chase unlikely diagnoses and perform unnecessary tests. "I'm skeptical that you could do as well as a doctor with these programs," he says.

Weed's critics and supporters alike admit that his unrelenting attack on the medical establishment's values and his willingness to call out offending physicians by name (as in the Addison's case) have driven away many potential supporters of his software. "It has to be said that, in the medical community, he was his own worst enemy," says Donald Lindberg, the national medical library director.

Other doctors wither at Weed's views without even meeting him. Shortly after Weed's "Black Box" paper was published, a radiation oncologist in Johannesburg named Leopold Reinecke posted a response on the British Medical Journal's Web site, calling it "a black day indeed" if Weed's words were to be accepted. "We (the creators of the computer) must now believe that the computer is able to function as our master. Our brain is `inferior' to this instrument of technology." He concluded dramatically: "Caring is our business, love is our method, and science is our tool - how can a computer do that?"

Despite the resistance, the PKC Corp. has won over some individual followers in the medical community. While it does so, the company has kept itself in business through contracts totaling about $8 million a year with the Department of Defense. The department has commissioned screening couplers for a wide range of things, from general health assessment to diagnostic tools for Persian Gulf syndrome. After September 11, a PKC program surveyed Pentagon employees about their mental health following the terrorist attacks.

A research survey in March found that 73 million people in the United States have sought health information online.

For PKC employees, the most exciting use of Knowledge Couplers is in the medical clinics run by the military's department of health affairs. The department is looking for ways to give consistent medical care to US troops, military families, and veterans scattered across the world, and officials think couplers may help. "We know for a fact they will never replace a physician's judgment," says Lieutenant Colonel Scott Goodrich, "but just like having a deskful of textbooks and reference journals is a help to us, the couplers are going to be just that same type of physician-support tool." Goodrich is a family physician and manager of an Army program that is evaluating the couplers.

Yale University is running clinical trials for the Department of Defense on Knowledge Couplers at an Army hospital in Fort Knox, Kentucky, and at a Navy hospital in Mayport, Florida. If the trials show that the software improves patient care without taking too long to complete, Pentagon officials plan to begin tying it in to a second-generation, computerized database of patient records scheduled for widespread rollout starting later this year. The goal, officials say, is to have patient information automatically flow between the couplers and the enormous patient-record database, making it easier to both fill out a coupler session and update the medical record. In addition to possibly improving care for regular patients, the resulting system could make it easier for medics and nurses to follow the proper course of diagnosis and treatment on patients in remote locations, then electronically send the cases for review by specialists anywhere in the world.

ighteen months ago, at the recommendation of the Department of Defense, the James A. Haley Veterans Affairs Hospital in Tampa completed a seven-month trial of the diabetes-management Knowledge Coupler. Most of the 38 test patients were in their 70s and 80s and had to be taught how to use a computer mouse. The patients filled out their part of the questionnaire with help from a secretary or pharmacist; a physician examined the patients and filled in their history. "I'm astounded at some of the knowledge that comes out" of the software, said Dr. Willard S. Harris, the former chief medical officer, who ran the test. "Much of it is knowledge that is repeatedly shown in the literature not to be used in actual practice," he said shortly after the testing. The trial found that regular use of the couplers improved the ability of the clinic's doctors to control the disease. Patients improved in seven of eight standards for diabetes management, compared with only one improvement for the patients who did not use the PKC software.

Despite the encouraging results, the Tampa VA decided not to incorporate the use of Knowledge Couplers into its clinics - an indication of the administrative resistance that PKC Corp. expects to face.

To reach wide audiences, the couplers would need acceptance by general hospitals, private practices, and health care organizations. Pierce says his company has begun trials with several large companies, which he would not name, to license an Internet version of the software. So, for example, a worker with a sore back could fill out an "acute lower back" questionnaire before visiting the company physician. That might result in a shorter visit.

The average patient may show less tolerance for the software than VA patients. VA patients tend to be less affluent, less busy (many are retired), and more tolerant of bureaucracies. Other visitors to a physician's office may resist the idea of sitting down before a computer each time. But with the rise of health-related Web sites, patients may be more willing to try. A March survey by the Pew Internet Project found that 73 million people in the United States have sought health information online. Other researchers have discovered that, as patients learn more, they are growing more and more skeptical of the "paternalist model" of medicine. Even if patients eventually decide to defer to their doctor's judgment, most want to know as much as possible about their conditions and their options for treatment.

Robert R. Weaver, an associate professor of sociology at Youngstown State University in Ohio, has taken an interest in Weed's work. He surveyed patients in a primary care clinic where Weed's software is used. Some grumbled that the couplers took too long to fill out. Others said they missed the interaction with the doctor. "It is just another way that the patient is alienated from his health care provider," one wrote. But Weaver found that most enjoyed being able to read all of the findings about their cases and print out copies to take home to review later or share with their spouses. "I like to be that much more a part of my own health care," one patient wrote.

Last year, the Institute of Medicine at the National Academy of Sciences, in Washington, D.C., released a report called "Crossing the Quality Chasm." The highly regarded report outlined dozens of flaws in the health care system in the United States. Among them was a tremendous gap between scientific knowledge and practice. That gap will only grow wider, the report said, as new technologies like genetic mapping and cell restoration make medicine more potent but exponentially more complex.

In an article in the Harvard Business Review the previous fall, three Harvard professors wondered in an article of the same title, "Will Disruptive Innovations Cure Health Care?" A disruptive innovation, as the authors described it, is a technology that shatters existing structures and changes everything, like the personal computer did to the mainframe computer. The authors, Drs. Richard Bohmer and John Kenagy and business professor Clayton M. Christensen, author of The Innovator's Dilemma, a bestseller among business books, wrote that the US health care industry is in crisis. "We believe that a whole host of disruptive innovations, small and large, could end the crisis," they wrote, "but only if the entrenched powers get out of the way and let market forces play out."

Both pieces made Weed angry: the first because he felt the Institute of Medicine team didn't go far enough in its criticisms or recommendations; the second because Christensen wouldn't reply to the correspondence in which Weed labeled the Knowledge Coupler as the disruptive technology that was prepared to tear up the health care industry. (A Harvard Business School spokesman says people in Christensen's office, which is swamped with mail, do not recall receiving the correspondence, but Harvard professors generally avoid endorsing any products.)

To disrupt the medical industry, Weed doesn't need every doctor to use his program on every patient, as he himself argues; he needs more hospitals to test the program, alleviate fears that it is impersonal and takes too much time to use, and begin incorporating it into existing medical systems, as the military is considering.

Some physicians and patients may be uncomfortable with the direction in which Weed would have us go. Ten years ago, if we wanted to trade a stock, we had to go to a broker. Today, we have E*Trade. Could Weed's program be disrupting medicine in the same way, bringing us closer to the days of E-Diagnose? In some ways, that will never happen. Patients will always need trained medical professionals to perform physical examinations, order lab tests, write prescriptions, and perform surgeries. Some patients will be too sick (or uninterested) to fill out the couplers. And few patients will be willing to throw their health completely into the virtual hands of a computer program. When we are sick, we need to feel as though our suffering is understood, and computers are poorly designed to show such compassion.

But when the PKC software makes it onto the Web, ordinary people who never labored through medical school will be able to record their own ailments and link those problems with a vast store of medical knowledge that is, in many ways, superior to the physician's memory. Although the software has the potential to become a hypochondriac's dream (or worst nightmare), it could cause the doctor-patient relationship to shift dramatically: The doctor will have to prove to patients that he followed the procedures called for by science.

As patients know more, they can collaborate better with the doctor in both diagnosis and treatment planning. Of course, patients who come into the doctor's office with reams of computer printouts and a newfound sense of understanding might be terribly annoying to physicians who are used to making unilateral decisions. But once their bruised egos heal, doctors may discover that they can leave work each day having cured more patients.

Berwick, the Harvard professor, has considered Weed a hero and "a giant of American medicine" ever since he devised the problem-oriented medical record. Berwick sat as chairman of the Institute of Medicine committee that produced the report that Weed criticized. He has seen the Knowledge Couplers demonstrated several times and has heard Weed's claims that they are ready for widespread use, but Berwick is not yet convinced.

Nevertheless, he contends that Weed's basic concept, that physicians need computers to make up for their brains' limits, is right on. "He's got to be right that it's computer-based," Berwick says of Weed. "Whether it's his [program] or not, I don't know. But my guess is, 30 years from now, we're going to look back on this and say, `How could we ever have thought we should do it the other way?' "

Chris Gaither is a Globe staff member who covers California's Silicon Valley.