The art of medicine and whether computers can replace doctors

My medical school at the University of Utah developed a clever computer program in the late 80s that was meant to both educate medical students and assist in the treatment of patients.

For several years the tech geeks at the school had collected an immense database of information from all the patient admissions at the hospital—presenting symptoms, exam findings, tests, and final diagnoses. They took all this data and crunched it into a program that provided a statistical snapshot of all the clinical syndromes seen by our university hospital over several years and then created a user-friendly interface that allowed us new trainees to learn from all this experience.

With this program I was able to enter the term “chest pain” and quickly learn what the ultimate diagnosis was for all the thousands of patients who had reported chest pain as part of their initial symptom constellation (e.g. 12% heart attack, 6% pulmonary embolus, 46% esophageal reflux, 18% chest wall pain, etc.).

Even cooler was the function that allowed us to enter several symptoms and findings and the computer would spit out the most likely diagnosis. Mind you, this was not based on some sort of theoretical textbook list of syndromes, but rather on the real-world experience of our patients and our doctors at our hospital. I’d type in “abdominal pain, nausea, and elevated blood lipase” and the computer would tell me that this patient was 91% likely to have acute pancreatitis.

As far as a computerized training tool this program was unsurpassed. It could tap its database to create a fictionalized patient and problem list, then quiz us on what testing and therapy we’d recommend. Patient X with foot pain and history of diabetes, heart disease, and schizophrenia shows up with a fever. What’s your next test? What’s the most likely diagnosis?

As medical students, we couldn’t help but be a little intimidated by a computer program that seemed a million times smarter than we were. It seemed to us that the rapidly advancing technology, if left to evolve down this Orwellian path, would soon be able to diagnose and treat patients with precision and reliability that we humans could only dream of. It was a little deflating to think that we would suffer through a decade of schooling and countless thousands in loans only to find that we are all being replaced by computers.

Well, that was 1989 and in the 21 intervening years I’ve managed to find employment as a doctor and have not yet been replaced by a robot. Despite the fact that we have more evidence-based models of care (algorithms based on large population studies that help take the variability out of medical care) than we’ve ever had, human doctors—despite all their flaws and imperfections—are still an integral part of modern healthcare.

You wouldn’t think this should be the case. In a world where complex software at sites like Amazon and Pandora can predict your tastes better than you can, where Google can access trillion pieces of information in fractions of a second, where all details of everyone’s medical history will potentially be available for data mining, and where computers can trounce the grandmasters in chess, someone could come up with a computer program that would out-doctor even the best of doctors (is there an app for that?). How is it that our system continues to rely on the imperfect judgment of doctors to treat us?

The answer, I would suggest, is what we commonly know as the “art of medicine.” Most of you have heard this phrase before, but do you know what it actually means?

I don’t either. Look it up in a dictionary and you’ll come up empty. Wikipedia, one of my favorite resources, has no separate entry for the art of medicine. One of the few on-line essays I found dealing with the subject employed vague platitudes to define this concept: “Mastery. Individuality. Humanity. Morality.”

I would attempt to define “the art of medicine” a little more specifically.

In much of modern medicine there are pretty clearly established practice patterns that lead to reliable results. A good example of this is our treatment of acute appendicitis. Ever wonder why there aren’t a half dozen competing treatment strategies for appendicitis? Come down with leukemia and you’ll be presented with at least a couple of options for therapy (different types of chemotherapy, bone marrow transplant, alternative medicines, etc.), but get an infected appendix and you pretty much get an immediate trip to the operating room. The simple reason you don’t get many options with appendicitis is that the established therapy works so well—it’s curative in all but rare cases. The rate of cure for most leukemias is low, thus there are many different ways to produce a relatively poor outcome. If we had a pill that cured leukemia in nearly all cases, all other alternatives would quickly vanish.

The treatment of appendicitis is so straightforward (the diagnosis, by the way, is murkier) that even the simplest computer program would churn out the correct recommendation every time. One plus one equals two. At that point all you need is a pair of hands to dig the offending tissue out of the abdomen.

But what happens when one plus one doesn’t equal two? What happens when there is no cure, or when the intended cure doesn’t produce the results you hope for, or when a cure is available but the patient’s condition precludes its application? I maintain that the art of medicine is very simply defined as what you do when things don’t go how they’re supposed to—when one plus one adds up to something else.

The art of medicine is what you lean on when you have to explain to family members why the surgery you just performed on their loved one didn’t produce the beneficial results you had hoped for, or, even worse, resulted in an adverse outcome or complication.

The art of medicine guides you in dealing with an emotionally fragile patient who needs a procedure or treatment that you know will be beyond their ability to cope.

The art of medicine dictates how you break the news to a worried waiting room that their ill family member has just passed away; or how to cautiously guide families to the acceptance of impending demise in a patient about whom they clearly have denial; or how to comfort a surviving spouse when he or she asks you if they should have done more to save the person they’d been married to for 50 years.

The art of medicine is knowing when to stop asking questions during an office visit and just let the patient speak what’s on his or her mind. It’s knowing when the agenda you have for your interaction has to take a back seat to something that may be very unimportant to you but critical in the mind of your patient.

The art of medicine is how you handle the patient who returns faithfully to your office seeking relief of her symptoms but who just can’t bring herself to remain compliant with the medications you prescribe to treat her condition. Or how you handle the patient with an easily treatable illness who simply refuses therapy.

The art of medicine is choosing a course of therapy based as much on an understanding of the character and personality of the patient as on knowledge the disease process itself.

I could go on for page after page but I think you get the picture. To a computer program—even the most clever ones—one plus one will always be two and the treatment of medical problems will remain nothing more than a function of odds, statistics, and search engines. This may be satisfactory for appendicitis and a handful of other problems, but for most of what we do as doctors the math is never quite so simple.

The art of medicine—the art of dealing with the unanticipated, unwanted, and less-than-optimal—can’t be programmed into a computer and, for that matter, can’t be taught in medical school. It’s what clinicians develop after years of experience, application and sometimes failure of science, mistakes, introspection, and learned humility.

Those sorts of skills will be in high demand forever. And until they can program robots to do all that I’ll probably be able to keep my job.

Eric Van De Graaff is a cardiologist at Alegent Health who blogs at the Alegent Health Cardiology Blog.

Submit a guest post and be heard.

Comments are moderated before they are published. Please read the comment policy.

  • http://33charts.com DrV

    Great post, Eric. These subtle elements of human interaction are hard to quantify. As we become progressively more disconnected from patients via technology, patient independence and centralization of decision making, this connection is going to evolve as a real demand.

  • http://fertilityfile.com IVF-MD

    Nice post. :) You did a good job of asking and answering your own question. It will be a long time before doctors can be fully replaced by machines.

    The computer program you described seeks to tackle only one component of doctoring – making the diagnosis. As you pointed out, doctors are not diagnosis machines. In order for machines to replace us, there would have to be a module for gathering data from the history, physical exam and lab tests and spitting out a ranked differential diagnosis list. Granted, this is the subsection of medical practice that is most amenable to computerization at present. Then, there would also have to be a module that does counseling and bedside manner stuff. And another robotics module that can perform surgery. We are nowhere near the capability to do either of the latter two, although it’s my understanding that vision correction surgery is pretty much automated now.

  • anon

    reminds me of a short story i read. it was set in the future and in space. the people aboard the ship had a computer for a doctor. one of the people develops appendicitis, only the computer keeping diagnosing food poisoning. if the computer had been told to ignore food poisoning as a possible diagnosis, it would have come up with appendicitis and the computer could have take out the appendix. but alas, no one on the ship had any medical knowledge and the person dies. the flaw in the computer program was that there wasn’t an risk factor analysis. and as flawed as human doctors are, we are looking another person in the eye and can see that it might not be the #1 diagnosis all the time.

  • http://secondbasedispatch.com Jackie Fox

    A lot of what you’re describing is compassion, which is not programmable except by whoever you believe may have programmed it into human beings. The art of medicine in large part comes from that compassion and that human spark. On behalf of patients everywhere, I’d like to answer the question of whether computers can replace doctors. The answer is an unequivocal “NO.”

    Dr. Van De Graaff, I would be honored if you would consider submitting this for Grand Rounds when I host on August 10th. My theme is doctor-patient communication.

    • http://www.alegent.com/heartblog Dr. Van De Graaff

      Jackie,

      You’re welcome to use any of this for your grand rounds presentation and I’m flattered by your request. The topic of doctor-patient communication is so voluminous that I’m sure you’ll have plenty to discuss.

      Eric VDG

  • Doc99

    “I’m a doctor, not an algorithm..”
    Leonard “Bones” McCoy, MD

    • http://onhealthtech.blogspot.com/ Margalit Gur-Arie

      “Please state the nature of the medical emergency;”
      Emergency Medical Hologram, U.S.S. Voyager (a century later)

  • HJ

    If a well programmed computer can make a diagnosis, information gathering and compassion can then be supplied by other medical personal. Nurses, lab techs, radiology techs can collect most of the objective information. I find very few doctors have time for counseling and bedside manner stuff.

    I often see comments that nurse practitioners are better at the human side of medicine, but lack a depth of education of a doctor. A computerized diagnostic tool would close that knowledge gap, making a NP an attractive, cost effective choice.

    • jsmith

      NPs are great for the simple stuff or stuff that they see over and over, but even computerized diagnostic tools required skilled operators, and, crucially, judgment, which is linked to experience and intelligence. Most docs simply have a lot more experience with serious disease than NPs do. And , yes, they usually come from a higher cognitive class. These issues are critical if somewhat nebulous.
      Uncomplicated ear infection, give me an NP if there’s no doc available. Heart failure or lymphoma, I want a doctor, at least I can get diagnosed and stabilized. Then, fine, turn me over to the NP for routine follow-up.

  • jsmith

    Medicine is a personal service profession. Everyone in medicine knows this. All normal people know this. Even a lot of computer-heads know this. Why the long post?

  • http://www.emrconsultant.com EHR Scope

    I don’t think any doctor has anything to worry about until robots start passing the turing test. Until then (if it ever happens), human intuition, reasoning, and experience will continue to be invaluable. At the very least, humans will be needed to monitor robots.

    Anyways, don’t worry about anything until you start hearing that robots are passing the turing test. At that point everyone’s jobs are in trouble though, so you won’t be lonely!