Medicine has a unique problem when it comes to its trainees. Although all fields must allow trainees some opportunity to “practice” their craft before being granted a credential allowing them to work without supervision, legal, accounting, or architectural errors made by trainees generally have fewer consequences than medical errors do.

Moreover, the demands of medical practice (particularly the need for around-the-clock and weekend coverage; Chapter 16) have led to the use of trainees as cheap labor, placing them in situations in which they have too little supervision for their skill level and experience. Although this early independence has been justified pedagogically as the need to allow “trainees to learn from their mistakes” and hone their clinical instincts, in truth much of it flowed from economic imperatives.

Yet the solution is not obvious. One can envision a training environment in which patients are protected from trainees—after all, who would not want the senior surgeon, rather than the second-year resident, performing his or her cholecystectomy? While such an environment might be safer initially, the downstream result would be more poorly trained physicians who lack the real-world, supervised experience needed to transform them from novices into experienced professionals. The problem would be similar for nurses and other caregivers.

These two fundamental tensions form the backdrop of any discussion of training issues in the context of patient safety. First, what is the appropriate balance between autonomy and supervision? Second, are there ways for trainees to traverse their learning curves more quickly without necessarily “learning from their mistakes” on real patients? This chapter will address these issues, closing with a short discussion about teaching patient safety. Other important training-related issues, such as teamwork training and duty-hour restrictions for residents, are covered elsewhere (Chapters 15 and 16, respectively).

The third-year medical student was sent in to “preround” on a patient, a 71-year-old man who had had a hip replacement a few days earlier. The patient complained of new shortness of breath, and on exam was anxious and perspiring, with rapid, shallow respirations. The student, on his first clinical rotation, listened to the man’s lungs, expecting to hear the crackles of pulmonary edema or pneumonia or perhaps the wheezes of asthma, yet they were clear as a bell.

The student was confused, and asked the patient what he thought was going on. “It’s really hot in here, doc,” said the patient, and, in fact, it was. The student reassured himself that the patient was just overheated, and resolved to discuss the case later that morning with his supervising resident. In his mind, calling the resident now would be both embarrassing and unnecessary—he had a good explanation for the patient’s condition. An hour later, the patient was dead of a massive pulmonary embolism. The student never told anyone of his observations that morning, and felt shame about the case for decades afterwards.

In his terrific book, Complications, Harvard surgeon Atul Gawande describes the fundamental paradox of medical training:

In medicine, we have long faced a conflict between the imperative to give patients the best possible care and the need to provide novices with experience. Residencies attempt to mitigate potential harm through supervision and graduated responsibility…. But there is still no getting around those first few unsteady times a young physician tries to put in a central line, remove a breast cancer, or sew together two segments of colon. No matter how many protections we put in place, on average, these cases go less well with the novice than with someone experienced.

This is the uncomfortable truth about teaching. By traditional ethics and public insistence (not to mention court rulings), a patient’s right to the best care possible must trump the objective of training novices. We want perfection without practice. Yet everyone is harmed if no one is trained for the future. So learning is hidden behind drapes and anesthesia and the elisions of language.1

Traditionally, supervisors in medicine erred on the side of autonomy, in the belief that trainees needed to learn by doing—giving rise to the iconic mantra of medical training, “see one, do one, teach one.” We now recognize this paradigm as being both ethically troubling and one more slice of the proverbial Swiss cheese, a constant threat to patient safety (Chapter 2).

Supervising physicians (the issues also play out with other health professionals such as nurses, but the autonomy that trainees in other fields can exercise and the danger they can cause seem less than those of physicians) are terribly conflicted about all this. Supervisors know that they could do many things better and more safely, but also recognize that trainees truly do need to learn by doing. Moreover, providing the degree of supervision necessary to ensure that trainees never get into trouble would create job descriptions for supervising attendings that might not be compatible with career longevity.

On a scale that has, at one extreme, supervising physicians doing everything while trainees watch, and, at the other, trainees doing everything and calling their supervisors only when they are in trouble, until fairly recently most medical training systems were far too tilted toward autonomy. Prodded by some widely publicized cases of medical error that were due, at least in part, to inadequate supervision (the death of Libby Zion at New York Hospital in 1986 was the most vivid example2; Table 1-1), the traditional model of medical education—dominated by unfettered resident autonomy—is giving way to something safer. (As a side note, while the Libby Zion case is popularly attributed to long resident hours, the chair of the commission that investigated the case [Dr. Bertrand Bell] clearly saw the root cause more as inadequate supervision than sleepy residents.3) We now recognize that “learning from mistakes” is fundamentally unethical when it is built into the system, and that it is unreasonable to assume trainees will even know when they need help, particularly if they are thrust into the clinical arena with little or no practice and supervision.4

Our new appreciation of these issues has led not only to some system reforms (including the increased supervision requirements enacted by the Accreditation Council on Graduate Medical Education in 20115; Table 16-2) but also to a crack in the dike of academic medical culture. For example, many attendings now stay late with their teams on admitting nights, a practice that would have been judged pathologically obsessive only 20 years ago (during my medical school days, one such attending picked up a nickname among the house staff of “the world’s oldest intern”).

In addition, the old culture of the “strong resident” or “strong student” (translated: one who never bothers his supervisors) is changing. Growing numbers of programs are building in expectations of oversight and creating structures to support it. For instance, around-the-clock attending presence (often with hospitalists and intensivists) to help supervise trainees at night is increasingly common. At UCSF Medical Center, for example, we now have hospitalists stay overnight with our residents. In addition to managing some patients independently, these faculty members supervise residents’ care of critically ill patients and are available to help with more stable patients. I routinely tell faculty who take on these “nocturnist” roles that they should always ask the resident, “What would you be doing if I wasn’t here?” before offering recommendations. I also ask these supervising physicians to emphasize to the residents that a call for help is a sign of strength and professionalism, not weakness.6

Ensuring appropriate communication is partly a cultural problem, but it also needs a scaffolding of thoughtfully developed guidelines. After analyzing a series of malpractice cases, the chiefs of surgical services at Harvard, working with the Harvard Risk Management Foundation, identified a series of “triggers” that should prompt direct communication between residents and attendings (Table 17-1).7 A study at four of Harvard’s teaching hospitals showed that, in one-third of cases, attendings were not notified after such trigger events.8 The guidelines were subsequently codified, with promising results.9

Oversight is not a monolithic construct. Supervisors must modulate their degree of oversight based on their comfort with the trainee’s level and demonstrated competency with the task at hand. After over 200 hours of ethnographic observation, as well as interviews with emergency department and general medicine teams, Kennedy et al. documented four types of oversight, each of which might be appropriate in different situations (Table 17-2).10

The challenge going forward will be to find the very narrow sweet spot between unfettered trainee autonomy and stifling attending oversight, varying the level of supervision to ensure patient safety while also permitting the graded independence that trainees need to become independent practitioners.11,12 Thankfully, we are now much closer to finding the correct balance than we were even a decade ago.

I appreciate the importance of this issue in a very personal way. You see, in the case that began this section, I was the third-year medical student who missed the fatal pulmonary embolism.12