Medicine Has Forgotten How to Touch the Patient

When was the last time your doctor actually examined you? I do not mean when someone checked your blood pressure, copied your medication list, looked at your laboratory results, or listened briefly through a shirt before returning to the computer. I mean examined you. Looked at you. Watched you breathe. Felt your pulse. Looked at your eyes. Examined your neck. Listened carefully to your heart and lungs. Palpated your abdomen. Watched you walk. Noticed the color of your skin, the movement of your hands, the expression on your face, and the thousand other small pieces of information that a human body offers to someone who has learned how to observe it.

For most of the history of medicine, this was not an optional ritual performed before ordering the “real tests.” This was medicine. Today, increasingly, it is treated as something quaint, inefficient, or unnecessary.

I have spent most of my professional life taking care of very sick people. Critical care is one of the most technologically intensive areas of medicine ever created. I love technology. I have watched mechanical ventilators breathe for patients who could no longer breathe for themselves. I have used ultrasound, CT imaging, sophisticated hemodynamic monitoring, extracorporeal technologies, molecular diagnostics, and laboratory measurements that physicians a generation before me could only have dreamed about.

I would never want to return to a world without these tools. But after decades at the bedside, I have also learned something that becomes more obvious, not less, with experience: technology is most powerful when it extends the physician’s senses. It becomes dangerous when it persuades the physician that those senses are no longer necessary.

The Patient Before the Data

Experienced clinicians understand something that is remarkably difficult to teach in a lecture hall or classroom. Sometimes you walk into a room and know immediately that something is wrong. You may not yet know what it is, but you know it before the laboratory does. The patient is breathing differently. His skin looks different. She answers a question a fraction of a second more slowly than she did several hours earlier. Posture, speech, facial expression, respiratory effort, or level of attention changes. None of these observations is magical. They are data, collected by a biological instrument trained through thousands and thousands of encounters: the physician.

When I round in an intensive care unit (ICU), I can learn an enormous amount before touching the computer. I can see whether a patient is fighting the ventilator. I can hear secretions from the doorway. I can see accessory muscles contracting. I can recognize when someone who looked reasonably comfortable earlier suddenly looks exhausted. I can see peripheral perfusion, diaphoresis, agitation, abdominal distension, or a subtle change in mental status. None of this means I should ignore the arterial blood gas, the chest radiograph, the CT scan, or the monitor. Of course I need them. The mistake is believing that because those measurements are more technologically sophisticated, they are necessarily more real than what is happening directly in front of me.

Medicine once began with the patient and proceeded toward testing. Increasingly, it begins with testing and eventually works its way back toward the patient. Abraham Verghese and colleagues have described this inversion of the diagnostic process, in which clinicians may encounter an image or electronic representation of a patient before encountering the patient himself. That change sounds subtle. It is not. Once the CT scan, laboratory panel, consultant note, medication list, and previous diagnoses have entered our minds, we no longer approach the patient without preconceptions. The data have already told us what we are supposed to see.

This is one reason the physical examination should never be reduced to nostalgia. Its value is not that physicians once carried black bags and made house calls. Its value is epistemological. It provides an independent stream of information. The history and physical examination can support what the tests tell us, but they can also contradict them. That contradiction is often where important medicine begins. A normal number should make us reconsider an abnormal observation, but an abnormal observation should also make us reconsider a normal number.

The Screen between Us

Something extraordinary happened to the architecture of the medical encounter. The physician and patient used to face each other. Today, they frequently face a computer.

A widely cited time-and-motion study of ambulatory physicians found that during the clinic day physicians spent approximately 27 percent of their time in direct clinical face time with patients and nearly half their time on electronic health records and desk work. For every hour of direct clinical face time, physicians spent almost two additional hours on electronic documentation and related desk activities during the workday. The electronic medical record was introduced as a tool for storing information. Somewhere along the way, the physician became one of the tools used to feed the electronic medical record.

This has consequences that go well beyond physician irritation or burnout. Attention is a finite resource. When I am looking at a screen, I am not fully looking at you. When I am clicking boxes, reconciling lists, responding to alerts, and documenting everything necessary to satisfy billing, regulatory, legal, and institutional requirements, I am performing work. But that work is not the same thing as observing a patient. We have confused documentation of care with care itself.

The modern medical record can create a remarkable illusion of precision. A note may contain a beautifully populated physical examination with normal findings listed across multiple organ systems. Anyone who has practiced modern medicine knows the uncomfortable truth: sometimes the completeness of the electronic examination exceeds the completeness of the actual one. Elder and colleagues warned years ago about examinations transformed by templates and drop-down boxes into records that may look more impressive than what truly occurred at the bedside.[2] The computer makes it possible to document fifteen normal observations in seconds. Unfortunately, it cannot guarantee that anyone actually made those observations.

The irony is striking. Never in history have physicians possessed more information about their patients, yet we may sometimes know them less. We know their hemoglobin from three years ago. We can review every medication prescribed by another physician. We can scroll through imaging reports, genetic data, vaccination histories, risk scores, and dozens of laboratory trends. We may possess thousands of data points about a human being who has received only a few minutes of our undivided attention.

The Physical Examination Is Not Primitive Technology

There is a tendency to speak about the physical examination as though it were the horse and buggy of diagnosis, something admirable in its time but sensibly replaced by superior machinery. That misunderstands what examination is. A physical finding is simply another diagnostic test. It has sensitivity, specificity, likelihood ratios, false positives, and false negatives, just like laboratory measurements and imaging studies. The Rational Clinical Examination movement made precisely this point decades ago, arguing that clinical findings should be subjected to the same rigorous evaluation applied to other diagnostic tools.

Some physical findings are poor. Some are excellent. Some things we were taught generations ago probably deserve to disappear. There is no virtue in preserving an examination maneuver merely because an old professor taught it to us. Evidence-based medicine should apply to the stethoscope as rigorously as it applies to an MRI scanner. But there is a tremendous difference between saying that certain bedside findings are unreliable and concluding that bedside observation itself has become obsolete.

The evidence also suggests that inadequate examination can cause harm. In one collection of cases involving physical-examination failures, missed or delayed diagnoses were common, as were unnecessary testing and treatment. Reviews of diagnostic error continue to identify deficiencies in history taking and physical examination as important contributors. These are not arguments against technology. They are arguments against voluntarily throwing away information before making decisions.

Consider something as simple as breathing. A pulse oximeter gives me a number, and that number can be enormously useful. But it does not tell me everything I learn by watching someone breathe. Two patients can have identical oxygen saturations while one is resting comfortably and the other is approaching respiratory failure. Respiratory rate, depth, pattern, accessory muscle use, paradoxical movement, ability to speak, mental status, and fatigue matter. Anyone who has managed respiratory failure has seen a patient whose monitor still looked reassuring while the patient absolutely did not.

The same is true throughout medicine. An echocardiogram can characterize cardiac anatomy better than my stethoscope ever will. Still, jugular venous distension, cool extremities, edema, diaphoresis, or altered mentation may tell me instantly that circulation is failing. A CT scan can show me an abdomen in exquisite detail, but it cannot substitute for knowing whether that abdomen has suddenly become rigid and exquisitely tender. Technology provides resolution. Examination provides context. Good medicine requires both.

We Are Losing the Teachers Too

Skills disappear surprisingly quickly when institutions stop rewarding them. Medical students enter training fascinated by physical diagnosis. They buy stethoscopes, learn cardiac sounds, practice neurologic examinations, and struggle to distinguish normal from abnormal findings. Then they enter the clinical environment and observe what the system actually values. They watch residents review laboratory values outside the room. They watch teams conduct portions of rounds around computers. They learn that ordering a test is often faster than finding someone who can teach them how to elicit a finding.

They discover that no billing system awards extra productivity points because a physician spent ten additional minutes teaching a student how to examine jugular venous pressure. Eventually they absorb the real curriculum of modern medicine: what the institution measures matters, and what it does not measure becomes optional.

Research has documented deficiencies in physical examination skills among medical trainees and has shown that bedside teaching itself has declined. Particularly troubling is evidence that greater time in training does not automatically produce better examination technique. One direct-observation study found physical examination technique correlated with recognition of clinical signs, differential diagnosis, and clinical judgment, yet performance actually deteriorated during internship. That should concern us. We may be creating a system in which physicians accumulate more medical knowledge while simultaneously losing some of the mechanisms through which they apply that knowledge to real human beings.

Clinical medicine has always depended on apprenticeship. You can learn the physiology of aortic stenosis from a book. Still, there is something different about standing beside an experienced cardiologist who places your stethoscope exactly where the murmur is clearest and tells you what to listen for. You can memorize the manifestations of cirrhosis, but you remember them differently after someone shows you spider angiomas, ascites, muscle wasting, and asterixis in a real patient. You can watch a thousand videos about respiratory distress, but watching an experienced intensivist recognize impending fatigue at the bedside teaches something that no multiple-choice examination can measure.

Once one generation loses those skills, the next generation has nobody to teach them. That is how knowledge disappears from a profession. Not through a decree announcing that the skill is forbidden, but through gradual neglect until eventually nobody remembers that it mattered.

Touch Is Also Communication

Another element of the physical examination is harder to quantify, and perhaps for that reason easier for modern medicine to discount. Examination is a form of human communication. When a frightened patient comes to see a physician, he does not simply want information. He wants to know that someone competent has assessed him. There is a difference between saying, “Your CT scan looks fine,” and sitting beside the patient, listening to his story, examining the area that hurts, and then explaining why the findings are reassuring. The diagnostic conclusion may be identical. The human experience is not.

Studies of patients’ experiences with physical examination suggest that touch has relational and emotional significance and can contribute to trust. A recent review addressing the physical examination in the age of artificial intelligence similarly emphasized that examination may serve therapeutic and relationship-building functions beyond diagnosis. Experienced physicians do not need a randomized trial to recognize that difference, although it is reassuring when research confirms what patients have been telling us for generations.

Of course, touch must be appropriate, explained, and respectful. Medicine’s history contains abuses of authority, and no physician is entitled to invade a patient’s physical boundaries under the banner of tradition. But an appropriate examination performed with consent is fundamentally different from unwanted touch. The very act of asking permission, examining carefully, and explaining what one is doing can reinforce something that is increasingly fragile in health care: trust.

There are moments when a hand on a patient’s shoulder communicates something that a laboratory result cannot. There are moments when sitting beside the bed rather than standing over it changes a conversation. Patients notice whether we are looking at them or at the computer. They notice whether we seem hurried. They notice whether we touch the painful area they came to have evaluated. They know, often instinctively, whether we have actually examined them.

This is particularly important when people are seriously ill. Critical illness strips patients of control. Clothes disappear. Privacy disappears. Strange machines surround them. People discuss laboratory values outside the room. Alarms sound. Families become frightened. In that environment, the physician’s examination can become more than information gathering. It reminds the patient that beneath all those numbers, tubes, diagnoses, and billing codes, there remains a human being whom another human being is trying to help.

Covid Accelerated the Distance

The Covid era accelerated trends that were already underway. Telemedicine expanded with extraordinary speed and, in many circumstances, provided a valuable service. Patients who could not travel could see physicians. Follow-up visits became easier. People living far from specialists gained access that they previously lacked. Virtual medicine is not inherently inferior medicine, and it would be foolish to discard its advantages.

But every technology has limits. A 2024 review of virtual physical examination acknowledged both the potential usefulness of remote examination techniques and the limitations involved in reproducing the physical encounter. A camera can show me a great deal. Still, it cannot yet allow me to palpate an abdomen, feel the temperature of a limb, detect subtle crepitus, assess the character of a pulse with my fingers, or perform many other elements of examination ordinarily.

The problem is not telemedicine itself. The problem arises when convenience quietly changes our definition of adequate medicine. A tool originally designed for encounters that do not require physical presence can slowly create the assumption that physical presence rarely matters. Once that assumption becomes culturally accepted, the burden of proof reverses. Instead of asking, “Can this visit safely be done remotely?” we begin asking, “Why does this patient need to be seen at all?”

That is how distance becomes normalized.

And Now Comes Artificial Intelligence

Artificial intelligence (AI) may become one of the greatest tools physicians have ever received. I use AI. It can help physicians process information, identify patterns, summarize records, improve communication, explore differential diagnoses, and reduce some of the enormous clerical burden that modern medicine has created. The correct response to AI is not fear. It is to decide which parts of medicine we want machines to amplify and which parts we should never surrender.

The conversation about AI and medicine often asks the wrong question: Will AI replace doctors? That makes for good headlines but misses the more immediate danger. The real question may be whether physicians will gradually stop performing the parts of medicine that AI cannot perform well because we have convinced ourselves those parts are inefficient.

Imagine the medical encounter several years from now. Before seeing the patient, an AI system has summarized 20 years of medical records. It has analyzed laboratory trends, imaging, genomic risks, medications, wearable-device data, and previous encounters. It has generated a differential diagnosis and recommended tests. During the conversation, ambient software records everything and creates the note. An algorithm suggests treatment. Much of this could be wonderful. It might even return time to physicians that computers previously stole from us.

But what will we do with that returned time? If we use it to look at the patient again, AI could paradoxically help restore medicine. If we use it merely to increase the number of encounters per hour, we will have created an even more efficient system for keeping doctors and patients apart. Technology does not determine which future arrives. Incentives do.

There is also a deeper paradox. Artificial intelligence learns from information that humans have collected. Clinical medicine, however, contains information that never enters the dataset if nobody observes it. If future physicians stop examining patients because algorithms appear sufficiently accurate without examination findings, the clinical record itself becomes poorer. We will not merely stop using bedside data. We will stop generating it.

A physician who cannot recognize respiratory distress without an oxygen saturation, volume overload without an echocardiogram, delirium without a score, or circulatory failure without a monitor has not become more technologically advanced. He has become technologically dependent. There is a difference.

The greatest danger of technological medicine is not that machines will become better doctors. It is that doctors may gradually stop doing the things machines cannot do.

The Patient Is Not a Dataset

Modern medicine has accomplished miracles precisely because we learned to measure things that previous generations could only guess about. I do not want less science. I want more of it. I want better imaging, better biomarkers, better artificial intelligence, better genomics, better remote monitoring, and better therapies. But the triumph of measurement has produced an intellectual temptation: if something cannot easily be converted into a number, perhaps it does not matter. Human beings do not cooperate with that philosophy.

Patients tell stories badly. Symptoms evolve. People forget details. Diseases fail to read textbooks. Laboratory tests return false positives and false negatives. Imaging shows abnormalities that have nothing to do with why someone is sick. Algorithms inherit assumptions from the data used to build them. The clinical encounter exists because somebody must integrate all of this uncertainty and decide what matters now, in this patient, sitting in this room.

That act of integration is not separate from science. It is the practice of science under conditions of incomplete information. We should therefore be careful about what we call efficiency. A five-minute encounter followed by five laboratory tests, an imaging study, and two referrals may look efficient from the standpoint of physician throughput. It may be extraordinarily inefficient for the patient and the health care system. Sometimes five more minutes of history and examination can eliminate an entire cascade of unnecessary testing. Sometimes it cannot. The point is that we increasingly work in systems that reward the cascade more reliably than they reward the five minutes.

We have spent years discussing the cost of health care while paying remarkably little attention to the cost of not knowing our patients.

Bring Medicine Back into the Room

The solution is not to smash computers, abandon telemedicine, throw away ultrasound machines, or teach medical students 19th-century rituals whose diagnostic value has never been demonstrated. That would be absurd. The solution is far simpler: technology should return the physician to the patient rather than remove the physician from the patient. Artificial intelligence should eliminate clerical work so I can spend more time examining and talking with patients. Electronic records should present information rather than demand endless data entry.

Medical education should treat bedside examination as a skill that develops throughout training, not something completed during the first years of medical school. Senior physicians should round with trainees at the bedside and show them what decades of accumulated experience actually look like when applied to a human being.

Most importantly, physicians need to remember that observation is not an inferior form of data acquisition. The eyes, ears, hands, and mind of an experienced clinician remain remarkable instruments. They are imperfect instruments, certainly. That is why we invented laboratory medicine, radiology, ultrasound, and every other technology we now possess. But imperfect does not mean useless, and old does not mean obsolete.

After decades in medicine, I trust technology more than I did when I began because I understand how extraordinary it has become. I also trust it less unquestioningly because I have seen its limitations. Numbers can reassure us when we should be worried. Images can distract us with findings that do not matter. Beautifully written electronic notes can describe patients whom nobody has really examined. The longer I practice medicine, the more convinced I become that the best physicians are not those who choose between technology and bedside medicine. They insist on both.

Perhaps that is what we are in danger of forgetting. Medicine is not ultimately practiced on images, laboratory values, algorithms, or electronic records. It is practiced on human beings. The data matter because the person matters, not the other way around.

We may be asking the wrong question when we debate whether artificial intelligence will someday learn to think like a physician. A more urgent question is whether physicians will continue to see, hear, touch, and know their patients well enough to recognize what the machine cannot. And if someday a patient can pass through an entire medical encounter while being scanned, measured, categorized, predicted, documented, and treated without ever being truly examined, we should ask ourselves something even more uncomfortable: have we created the most technologically advanced medicine in history, while simultaneously forgetting how to practice medicine?