“The patient is having an adverse drug reaction,” I announced as I walked out of Mrs. J’s E.R. room holding a bottle of antibiotics that had recently been prescribed to her. The medical student assigned to the patient looked sheepish and the senior resident looked surprised. Along with the emergency-department registration staff, the triage nurse, and the nursing student, they had already asked Mrs. J. if she was taking any new medications. Yet the patient waited to tell me—the attending E.R. doc and the final and most senior questioner—about her new antibiotics.
My line of questioning was guided by other clues to Mrs. J.’s diagnosis, like the peeling skin on her fingers and her fever. Perhaps when I asked her whether she was on medication, the probing tone of my voice finally triggered her memory and led her to dig for the pills in her purse. But the frequency with which attending doctors elicit different and more useful information from patients belies the explanation that the most skilled physician is simply the best detective. Medical students and resident trainees get over their embarrassment and often start referring to the lead doctor’s seemingly magical effect on a patient by a nickname—the Attending Sign. The idea is that the presence of an attending physician, like a reflex hammer that hits the right spot on the knee, causes the patient to disclose a cardinal sign of her diagnosis.
The quirks of medical information-gathering have provoked concern that doctors and nurses are collectively doing something wrong—and the realization that while medical history-taking is one of the most important things we do, we still don’t know how best to do it. Medical educators have tried to codify the techniques for taking a patient’s history since medical education in the United States evolved from apprenticeships to formal schooling. Sir William Osler, the Johns Hopkins physician-in-chief who is considered the forefather of modern medical education in the United States, instructed his residents, “In taking histories, follow each line of thought; ask no leading questions; never suggest.” Osler’s century-old guidance is still standard teaching in medical school classes that teach clinical skills, yet recent studies have demonstrated that his rules are routinely violated. In one study, investigators taped 93 patient encounters with E.R. residents and nurses. While the questioners tended to begin the interviews with appropriate open-ended questions, 80 percent of them interrupted the patient’s answer with a closed or leading question within 12 seconds on average.
In light of findings like these, some medical educators argue that instead of “taking” a history, by extracting facts through questioning, we should “build” a history, by eliciting a biomedical narrative that a patient develops herself. Organized medicine has aggressively supported such efforts, as in this 1999 report by the Association of American Medical Colleges. A 2003 editorial in the AMA journal Archives of Internal Medicine argues that interviews that leave little room for a patient’s ideas and feelings anger patients and threaten the quality of their care, concluding, “We simply cannot afford the time to do our job so poorly.”
Alongside the movement to humanize the medical interview is a competing push to mechanize it. In its highly publicized report on patient safety, To Err Is Human, the Institute of Medicine pointed to computerized information-gathering as a way to decrease mistakes and improve communication. Computers are also a way to save time as understaffed emergency departments, expanded patient loads, and the model of medicine as a business squeeze the minutes that providers have to gather medical histories. Then there’s the increasing value that patients and their families attribute to what they see as good customer service, which translates into complaints when they have to repeat information they think the doctor should already know. Small wonder, then, that computer programs for taking medical histories, like Instant Medical History, RelayHealth and Kryptig, are becoming increasingly popular. Patients can use the software to enter their histories in waiting-room kiosks or at home via the Web.
The problem with these systems is that in all likelihood they often don’t work. Let’s say a patient enters a medication dosage. If the dosage is later changed by a physician who is not linked to the computer network, the patient may not remember to update the info. One 2004 study at a Veterans Administration hospital in Iowa City found significant discrepancies between the medications a patient was actually taking and what was reported in the database. Slate’s Tim Noah is right to point out that the kinks get worked out of many software applications that cause trouble when they initially go on line. But the problem of human error persists: When the computer displays a medication list, most of which is usually correct, the nurse, doctor, and even patient may be less likely to double-check it.
Automated systems may be even worse at collecting subjective and nuanced patient histories. A European study from last year compared medical histories obtained by self-reported questionnaires and by in-person interviews. The investigators found poor agreement between the methods, with patients omitting important facts from the questionnaires, especially about less severe or short-lived problems. When they answer a computer rather than a doctor or nurse, patients cloud the results based on their own interpretation of what is important about their health.
Despite all the research, no one has successfully demonstrated the ideal approach for gathering a patient’s history. From my perspective as an E.R. doc, the narrative approach seems valiant but unrealistic—too cumbersome for an emergency setting. The automated approach, on the other hand, is potentially dangerous because it narrows the scope of patients’ stories. As doctors swing along this pendulum, what may be getting lost is the organic nature of a patient’s story—the natural unfolding that led Mrs. J. to identify her new medication not the first time she was asked, but the fifth.
What explains this process? A patient’s goals vary, even over the course of a 30-minute E.R. visit, in ways that shape how he conveys his condition. When a patient arrives in the E.R. with chest pain, for example, all he may want at first is to be seen right away. Once he is “triaged” into a room, his next goal is often to get pain relief. Only after he is more comfortable and less anxious may the patient be able to tell the story that includes the most useful diagnostic information. His doctors and nurses’ agendas also vary along the way. The triage nurse must determine if the patient’s complaint of chest pain could indicate an immediately life-threatening condition; her questions will be directed toward the goal of choosing where in the E.R. to move the patient and how quickly to do so. The medical student’s questions are designed to build a diagnosis that could explain the patient’s symptoms—what the attending doc will inevitably ask for. The E.R. nurse focuses on determining if the pain medication is helping. Only when the attending steps into the room, armed with a reassuring EKG and a set of normal vital signs (OK, he is not dying on me, let’s try to figure out what is going on) do the goals of the patient and his caretakers coalesce around building a complete history.
The benefits of multiple medical interviews haven’t really been quantified. But the Attending Sign captures the way in which patients’ histories self-evolve as multiple interviewers take multiple approaches. I am pretty sure that Mrs. J.’s antibiotic pills were buried so deeply in her brain (and her purse) that it was the iterative experience of the E.R. itself, of talking to many people about her condition, that finally prompted her memory and her diagnosis.