We found that extended-duration work shifts were associated with adverse effects on patient safety (i.e., fatigue-related significant medical errors and preventable adverse events, including those resulting in fatalities). The risk of self-reported attentional failures during patient-care and educational activities was also significantly affected by extended-duration work shifts. That is, during months with frequent extended-duration work shifts (i.e., five or more extended-duration shifts in the month), interns were significantly more likely to fall asleep during surgery, while talking to or examining patients, during rounds, and during lectures or seminars, potentially affecting their ability to deliver patient care or to learn.
The hours of reported sleep per month decreased as the number of extended-duration work shifts, the number of reported medical errors, and the number of reported attentional failures increased—and this is consistent with laboratory studies that have unequivocally demonstrated the profound effects of sleep deprivation on alertness and performance [9
]. Indeed, the adverse effects after 24 h of continuous wakefulness on neurobehavioral performance are comparable to that of a blood alcohol concentration of 0.10% [16
]. Our results confirm the recently published findings by Landrigan et al. [3
] and by Lockley et al. [2
]. During a traditional schedule of extended-duration work shifts every other shift, as compared to a timetable where the scheduled length of work shifts was limited to less than 16 h, Landrigan et al. reported that serious medical errors were 36% more frequent [3
], while Lockley et al. reported that polysomnographically defined attentional failures were approximately twice as frequent [2
]. Moreover, our findings extend these results by confirming the associations between extended-duration work shifts, medical errors, preventable adverse events (including those resulting in fatalities), and attentional failures in interns in multiple specialties and hospitals across the United States.
In addition to the impact of these fatigue-related serious medical errors on patient safety, the perception of having made a medical error that causes an adverse patient outcome creates significant emotional distress for physicians that can last days or years and may include feelings of fear, guilt, anger, embarrassment, and humiliation [18
]. The personal distress and reduced empathy associated with perceived medical errors increase the odds of future medical errors [19
]. Without coping mechanisms, such as accepting responsibility, discussion with colleagues, and disclosure to patients, physicians may use dysfunctional methods of dealing with errors, such as alcohol and drug abuse [20
We also demonstrated that extended-duration work shifts adversely affected medical education, as the odds of falling asleep in lectures and while on rounds with attending physicians increased significantly when extended-duration shifts were worked. Not only is adequate sleep crucial for maintaining vigilance [22
], but sleep also plays an important role in memory consolidation and learning [23
]. The amount of sleep reported by most interns is less than that shown to be necessary to perform satisfactorily on cognitive performance tasks that are implicitly necessary for learning [25
]. Additionally, the high level of chronic partial sleep deprivation, consistent with prior reports [26
], may account for the high base rate of attentional failures in these participants.
We acknowledge that, despite the large amount of data on intern-months collected from interns across specialties from programs throughout the United States, there are a number of limitations to our study. First, the individuals who volunteered for our study represent a small proportion of interns in the United States. Therefore, it is possible that the participants in our study may not be representative of the entire population of interns. For instance, our sample could be biased towards interns with a specific interest in the issue of resident work hours, thus raising the potential of a reporting bias. Additionally, those individuals who responded to our survey may have been more responsible and therefore more likely to report medical errors (although the responsibility of the participant could not account for the relationship between the numbers of extended-duration shifts and reported medical errors that were identified by the within-person analysis). However, potential participants were not informed of the specific study hypotheses, and the questions addressing our primary exposure and outcome variables were distributed among questions regarding secondary outcome measures such as caffeine usage, mood, and stress. Moreover, because each intern serves as his/her own control in our case-crossover analysis, it would have taken considerable time over a number of months—and a conscious effort to mislead—for intentionally biased respondents to affect our results deliberately. However, it remains a possibility that those who were more attentive to the topic of medical errors may have volunteered to participate in the study.
Second, the data for medical errors relied on self-reports and were not independently validated. Additionally, because “fatigue or sleep deprivation” was part of the question regarding medical-error outcomes, it may have inflated the association between this suggested “cause” and the medical error. Third, the definition of what incidents interns believe constituted a ”significant error” or “adverse patient outcome” may have varied from person to person. However, the consistency of the results when the outcome of the error resulted in a fatality provides a common metric of severity. In addition, because each intern was compared only to him- or herself, there is no reason to believe that he or she would have changed the criteria for what was significant by the nature of the rotation schedule.
Fourth, in our case-crossover analysis, it is possible that during months without extended-duration work shifts, interns may be on rotations with fewer opportunities to make errors (e.g., radiology rotation, dermatology, ambulatory elective), than on more clinically-oriented rotations when extended-duration work-shift frequency was greater (e.g., intensive-care unit, hospital ward) resulting in more opportunities to make errors. Additionally, there may be other additional organizational features (e.g., staffing, supervision) that differ between rotations. To address partially this issue of monthly activity, we conducted a secondary analysis in which we limited the data to only those months on hospital wards. When we did so, we found that the association between the frequency of extended-duration work shifts and medical errors and adverse events was comparable to the results of the overall analysis. While the number of hours at work increases as the number of shifts increases and may partially account for the increased rate of events, it should be noted that the percentage increase in weekly work hours (a 16.7% increase in the category of between one and four extended-duration shifts per month and a 42.9% increase in the number of five or more extended-duration shifts per month, compared to months with no extended-duration shifts) cannot account for the observed increase in the rate of events. For months with between one and four extended-duration shifts and for months with five or more extended-duration shifts, fatigue-related adverse events increased by ~450% and ~700%, respectively. For months with five or more extended-duration shifts, fatigue-related adverse events resulting in a fatality increased by ~300%.
We also recorded the frequency of both fatigue-related and non-fatigue-related errors. We found that even though the rate of non-fatigue-related errors increased as extended-duration work-shift frequency increased (presumably for the two reasons outlined above), the magnitude of the increase was much less than the increased rate of fatigue-related errors. This finding suggests that the increased rate of fatigue-related errors was not predominately due to a confounding effect of monthly activity. However, it remains possible that interns were more tired when working more frequent extended-duration shifts and that they may have been more likely to under-report or to over-report medical errors, misperceive errors in their work, or to attribute errors to fatigue. Landrigan and colleagues found that physicians perceived fewer medical errors than those assessed by an independent observer [3
Fifth, for many of our analyses, we collected information about the events (i.e., errors) and exposures (i.e., number of extended-duration shifts worked per month) simultaneously from the monthly surveys. This could lead to reporting bias as interns may be more inclined to send back surveys during months in which incidents occurred. However, when analyses were limited to participants returning all surveys throughout the year, the relationships between errors and the frequency of extended-duration shifts did not change substantially. Sixth, there may be a carry-over effect from one month to the next. For example, an intern may have worked five extended-duration shifts in the first month and no extended-duration shifts in the second month. If that intern reported a medical error at the beginning of the second month, our data analysis would not have been able to detect the potential carry-over effect from the high number of extended-duration shifts in the first month.
Finally, although self-reporting is a well-established method of eliciting information regarding medical errors, prior work has demonstrated that even very robust stimulated self-reporting systems detect only a fraction of all medical errors [28
]. Consequently, it is likely that the interns committed a greater number of serious medical errors, including those resulting in preventable adverse events, than we have reported here. Thus, although there are a number of potential biases and confounders in our study, given the consistency and magnitude of our findings, it is unlikely that they could account for the striking association between extended-duration work shifts and fatigue-related significant medical errors, preventable adverse events, and attentional failures.
According to the 1999 report from the Institute of Medicine, between 48,000 and 98,000 deaths each year occur due to a medical error [1
]. Recently, the ACGME has placed limitations on resident work hours in an attempt to reduce fatigue-related medical errors. Historically, when such attempts have been made to reduce resident work hours in the United States, the frequency of extended-duration work shifts (>24 h) has been reduced (e.g., from a frequency of Q2 to Q3 to Q4 [where Q2 is an extended duration shift occurring every other night, Q3 is an extended duration shift occurring every third night, and Q4 is an extended duration shift occurring every fourth night]). However, the practice of working for more than 24 h consecutively has remained the cornerstone of American postgraduate medical education. In fact, the recent (2003) ACGME work-hour guidelines for postgraduate medical education programs effectively continue to sanction up to nine extended-duration shifts (of up to 30 h consecutively) per month, since every other shift can be an extended-duration work shift under the new ACGME guidelines [29
]. Still, interns working extended-duration shifts within these ACGME guidelines reported significant numbers of medical errors, including those that resulted in adverse patient outcomes and fatalities. Furthermore, 83.6% of interns reported working more hours than allowed by ACGME standards in the year following their introduction [30
These data, collected from interns in all specialties across the United States, are not consistent with the recent suggestions by a member of the ACGME Residency Review Committee in Surgery [31
] that safety hazards associated with resident fatigue are limited to a small subset of trainees. Even interns who worked well below the current 80-h ACGME weekly work-hour limits (averaging 64.8 h of work per week), but who continued to work up to one extended-duration shift per week (half the weekly frequency allowed under current ACGME standards), had 8-fold greater odds of reporting an adverse event than those who did not work extended-duration work shifts. This finding is consistent with data from numerous studies documenting that 24 h consecutively of wakefulness impairs short-term memory, degrades neurobehavioral performance, and greatly increases the risk of both errors of commission and omission and attentional failures [23
]. Additionally, Ayas and colleagues recently reported that the odds of an intern having a percutaneous injury increased by 61% after ≥ 20 h at work [34
]. These findings are also consistent with the recent demonstration that elimination of extended-duration work shifts reduces attentional failures and serious medical errors among interns working in intensive-care units [2
Our results thus reveal that the practice of scheduling 24-h or greater extended-duration work shifts, as currently sanctioned by the ACGME, may pose a significant increased risk of safety hazards to patients, contribute to the occurrence of medical errors that are attributable to fatigue or sleep deprivation and to consequent preventable fatal and nonfatal adverse events, and may also interfere with the primary educational purpose of residency training. These results have important public policy implications in terms of postgraduate medical education and suggest that directors of training programs should consider alternative coverage schedules for trainees with the objective of eliminating extended-duration shifts. In Europe, where the tradition of extended-duration “on call” shifts originated more than a century ago, work shifts of all physicians (including those in training) have recently been limited to 13 h consecutively [35
], thereby eliminating extended-duration work shifts altogether. Fletcher et al. recently published a review of interventions aimed at reducing US resident work hours, including strategies such as day and night float teams and the use of physician extenders [36
]. Future studies should explore the applicability of our findings regarding the association between medical errors and extended-duration work shifts to all practicing physicians in the United States.