Every four days, for long stretches of the year, I try to take some time out of my morning to pick my wife up from work at the hospital. These are her “Q4” rotation months, when medical residents are assigned to 28-hour shifts every four days, usually with regular 12-16 hour workdays in between. Earlier in her residency, she used to drive herself the short distance home after these marathon shifts. But, especially when she was running late, whether to finish up some paperwork or to address a late-emerging crisis, I would worry myself sick with visions of her falling asleep while driving over a narrow bridge, or of a darting bicyclist combining tragically with her dulled reaction time. Better to have the peace of mind of my own relatively well-rested hands on the wheel.

These fears are all too real. Medical residents are at a vastly increased risk of being involved in traffic accidents following long shifts. It is a serious concern for my family. But in the grand scheme of things, permitting sleep-deprived residents to drive is a modest public safety concern. The more jarring question is this: How does it make sense to entrust doctors with control of operating rooms and intensive care units when they cannot be trusted on the road?

Last May, researchers at Johns Hopkins University School of Medicine announced a sensational finding: Medical mistakes cause more than 250,000 deaths in the United States annually. This would make medical error the nation’s third leading cause of death—trailing only heart disease and cancer. Many have criticized these numbers as overstated and misleading, noting, for example, that many preventable deaths involve very sick patients who probably would have died within six months anyway. Other studies have found the number of annual deaths attributable to medical error to be as low as 25,000. And, even broadly defined, the overall error rate is still reassuringly low—with fatal errors occurring in less than 1 percent of the 35 million hospital admissions Americans undergo each year. Yet, however you spin it, the prospect of even 25,000 preventable deaths demands serious attention.

It is natural to wonder whether the punishing nature of medical training compromises patient safety. Physicians, after all, are only human, and sleep-deprived humans make mistakes. After just 17 hours without sleep, our ability to perform many basic mental tasks declines to levels at or below our functioning when we are drunk. Chronic sleep deprivation degrades many measures of cognitive performance relevant to the work of being a doctor, such as basic reasoning, motor skills, concentration, memory, and even moral judgment. Not surprisingly, residents have been documented making anywhere from 36 percent to 700 percent more serious medical errors when working recurrent 24-hour shifts, including more than five times as many serious diagnostic errors, and experiencing twice as many attentional failures, such as (in extreme cases) falling asleep during surgery. In one survey, over 40 percent of residents reported fatigue as the cause of their most serious mistake—which in one-third of instances led to a patient’s death. Even more troubling, many doctors report feeling fine when their performance has actually dipped due to chronic sleep deprivation, indicating false faith in their relative immunity to sleep-related cognitive impairments.

Following the Challenger Space Shuttle’s explosion in 1986, a Presidential Commission determined that overwork-induced fatigue among key personnel played a substantial role in causing the disaster. In the decades that followed, most professions with responsibility for public safety—from pilots to nuclear power plant operators—were subject to detailed government requirements mandating rest and forbidding excessive shift lengths. In the early aughts, momentum built to regulate residency work hours in a similar manner. In early 2003, Representative John Conyers of Michigan and then-Senator Jon Corzine of New Jersey introduced in Congress the Patient and Physician Safety and Protection Act, which would have imposed strict limits on resident work hours, delegated broad enforcement authority to the Department of Health and Human Services, and, critically, established a mechanism to provide additional funding to hospitals to make up for the cost of lost resident hours.

Just a few months later, in a successful bid to stave off this sort of federal oversight, the Accreditation Council for Graduate Medical Education (ACGME), the nonprofit licensing body that accredits medical residency programs, voluntarily introduced resident duty hour restrictions. Under the rules, which were expanded in 2011, residents are capped at a monthly average of 80 hospital hours each week, single shifts are limited to 16 hours for first-year residents (or “interns”) and to 28 hours for more experienced residents, and there must be an 8-hour break between shifts. Before the voluntary adoption of duty hour reforms, most residents in the United States were not subject to any meaningful limits on their work hours—36 and 48-hour shifts were routine, as were 100-120-hour workweeks. To most, it seemed self-evident that the new rules would reduce medical errors and save lives.

The data tell a different story. Study after study has concluded that duty hour reform has had little appreciable effect on patient outcomes. This prompted the ACGME to sponsor large-scale, national experiments that turned most of the nation’s teaching hospitals into laboratories for examining the effect of sleep deprivation on medical care. Last February, the New England Journal of Medicine published the results of the first round of this research, focusing on surgeons. For purposes of the study, the ACGME waived the most important duty hour restrictions for surgical residents, including on maximum shift length and minimum time off between shifts, for roughly half of accredited surgery residency programs across the country for one year. Consistent with earlier findings, researchers could identify no statistically significant differences in patient outcomes at programs with and without the restrictions.

How is it possible that doctors are able to practice medicine about as effectively when they are extremely sleep deprived as when they are well rested? Doctors are not superhuman; the data are clear that longer hours lead to many mistakes. But longer shifts also reduce the number of “handoffs,” or the transfer of care from one physician to another. Among the “patient safety community,” there is no question “handoffs are the most dangerous time” in terms of the risk of medical error, says Christopher Peabody, an emergency physician and director of the Acute Care Innovation Center at the University of California, San Francisco. Following a handoff, patient-specific information and expertise is invariably lost. An estimated 80 percent of serious medical errors involve a miscommunication between caregivers during the transition of care. What the data reveal, therefore, is that the risks associated with extreme work hours are offset, at least in part, by improved continuity of care. Better for a fatigued resident to finish the surgery rather than hand the scalpel off to a fresh-faced colleague who is insufficiently familiar with the case.

Moreover, because the ACGME’s duty hour rules were an unfunded mandate, hospitals were placed in the difficult position of attempting to do the same amount of work with fewer resident hours. As a result, most teaching hospitals instituted “night float” rotations—where a single resident covers overnight all the patients for several “day” residents who, before the restrictions, would have been at the hospital themselves. While night float is often popular among residents because it reduces the overall number of nights they must work, its introduction undermined the anticipated safety benefits of the reforms. It is no surprise that multiple doctors, even if leaden with fatigue, are often better than one.

In light of these problems, the ACGME has begun rolling back reform. Starting next academic year, the 16-hour limit on intern shifts will be eliminated. A study of similar scale to that conducted on surgeons is currently underway for internal medicine residents (i.e., generalist doctors). If the results of that study are consistent with past research, the ACGME is expected to scale back duty hour reform even further.

So is duty hour reform simply a well-intentioned but failed experiment? At first glance, this might seem a reasonable interpretation of the data. If increasing the number of handoffs and instituting shortcuts like night float offset the benefits of better-rested residents, then reform did not accomplish its self-described goals. And fewer resident hours means less resident training. It is perfectly reasonable to believe any policy that is neutral with respect to patient care should weigh in favor of more training opportunities for the nation’s future doctors—their families and sanity notwithstanding.

But this perspective also reflects a path-dependent decision to avoid addressing the true nature of the underlying problem. That low-quality handoffs and overwhelmed residents also cause medical errors does not mean we need more cognitively drunk doctors. It means we need better handoffs and more resources devoted to supporting doctors-in-training. Considered through this lens, the ACGME’s research is best understood as demonstrating that longer shifts—and thereby fewer handoffs—are not a viable solution to the handoff problem.

Handoffs are often disorganized and informal, conducted only verbally, and lacking in systematic oversight or direction. One survey found the vast majority of emergency department academic programs did not even have standard, written policies regarding handoffs, and offered no formal didactic handoff training to residents. Handoffs are routinely infected by basic communication errors. For instance, one study found that nearly 30 percent of handoffs inaccurately convey patient medication lists; another found that nearly 15 percent of handoffs in emergency departments omit critical information regarding patient vital signs.

Yet basic, cost-effective handoff interventions have been shown to succeed in reducing medical errors: One study found that instituting a simple, standardized routine for conducting handoffs reduced medical errors by 23 percent, and reduced preventable harm to patients by 30 percent. Similar studies have found that deliberate and thoughtful procedures to improve handoffs can have an even greater impact on patient outcomes. While there is ample room for more research and innovation, “we basically understand how to do a good handoff,” explains Dr. Peabody. But “the majority of hospitals around the country have not implemented these things that we know work” in reducing medical errors. “Why doesn’t every hospital have a system it uses to hand off patients?”

There is a growing awareness in the medical community of the need to take handoffs seriously. In 2011, the ACGME instituted general requirements for residency programs to provide handoff training and to monitor handoff quality. But there remains vast room for improvement. Best practices are under-researched and underutilized, and transfer of care is still too often seen as an ancillary concern in both medical education and practice. If the medical profession is serious about reducing medical errors, the ACGME and other organizations should support large-scale research into improving handoffs, and residency programs should be required to prioritize the implementation of best practices. This is far more likely to improve patient care than returning to the days of expecting residents to operate effectively at the tail end of 36 or 48 hours without sleep, in willful blindness to fundamental human limitations. Instead of a return to the discredited past, the medical community would do well to explore creative, forward-thinking solutions to the resource constraint and continuity-of-care problems that have, thus far, limited the success of duty hour reform.