Resident work hours continue to be an issue of importance.1 Report after report in the literature describes levels of fatigue and sleep deprivation far in excess of those experienced by any other group in the U.S. workforce.2-6 Although concern for public safety has led to specific work-hour limits for airline pilots, bus drivers, and railroad engineers, medical residents still average close to 80 hours a week on duty, with many exceeding that figure.2-7
Over a decade ago, similar concerns, arising in part from the unfortunate Libby Zion case, prompted the New York State Department of Health to establish regulations governing resident work hours and supervision in that state.8,9 These essentially provided for an 80-hour workweek averaged over a four-week period, at least one scheduled 24-hour free period per week, and a maximum shift of 12 hours for emergency medicine residents, together with around-the-clock supervision by attending physicians. Surgical programs were exempted from the 24-hour maximum provided they met certain other requirements.
In 1998, surprise visits by the New York State Department of Health to 12 hospitals that had agreed to follow these guidelines in exchange for sizable financial subsidies from the state found them all to be in violation, with 94% of residents in New York City and 37% in the rest of the state shown to be working more than 85 hours per week. At the same time, 77% of surgical residents in New York City were working in excess of 95 hours per week.9 In April 2001, a group of petitioners, including Public Citizen, the American Medical Student Association, the Committee of Interns and Residents, and Drs. Bertrand Bell and Kingman Strohl, urged the Occupational Safety and Health Administration (OSHA) to set national regulations regarding hours of service for all residency and fellowship programs.10 On November 6, 2001, Representative John Conyers (D-Mich.) introduced the Patient and Physician Safety and Protection Act (H.R. 3236), which would essentially mandate by law the regulations requested in the OSHA Petition.10,11 On June 12, 2002, Senator Jon Corzine (D-N.J.) introduced similar legislation in the Senate (S. 2614).12 On October 4, 2002, OSHA denied the petition, stating “other knowledgeable groups are taking action on this problem.”10,13
With governmental regulation imminent, the key organizations responsible for graduate medical education responded. In October 2001, the Association of American Medical Colleges recommended, “in no case should residents be scheduled to be on duty more than 80 hours a week,” and published their complete policy on graduate medical education in August 2002.14 This was followed by action of the House of Delegates of the American Medical Association (AMA) in June 2002 that recommended a limit of 80 hours per week averaged over two weeks and the study of possible increases of 5% for some programs.15 Finally, in February 2003, the Accreditation Council for Graduate Medical Education (ACGME), whose responsibility it is to set standards for, monitor, and accredit all residency programs, approved the final version of the much-discussed recommendations of its Work Group on Resident Duty Hours and the Learning Environment.16 These standards, scheduled to take effect on July 1, 2003, will apply to all programs and specialties. Averaged over a four-week period, they call for a limit of 80 hours per week, every third night on-call, and one day out of seven free of patient care. Programs are also required to give residents a minimum ten-hour rest period between times on duty. Finally, there is a controversial “24/6” provision that limits on-call duty to 24 hours, plus “an added period of up to six hours for continuity and transfer of care, educational briefing and didactic activities,” during which no new patients may be admitted. It is possible to request up to a 10% exception to these standards for sound educational reasons if approved by a specialty’s residency review committee and the institution’s graduate medical education committee. However, no specialty-wide exceptions will be permitted during the first year of implementation. Of note is the ACGME’s insistence on placing a higher value on education and patient safety than on meeting service demands.
The 80-hour workweek has emerged as the consensual standard without clear empirical support.1 To date, no broad-based account of residents’ actual work experiences across specialties is available to anchor these discussions. In this article, we report the responses of a large sample of postgraduate year 1 (PGY1) and year 2 (PGY2) residents in a variety of specialties and hospitals to a national random survey covering a number of aspects of their work experience during the first two years of training. The focus of this report is on their self-reported work hours and their correlates.
The study sample was identified in early 1999 using the Graduate Medical Education database, which is secured as part of the AMA’s annual survey of graduate medical education programs.7,17 Using the random selection feature of a standard statistical software package (SPSSPC), a 15% sample was drawn from residents who were in PGY1 and PGY2 positions, had no prior training, and were scheduled to complete their residency year in the summer of 1999. There were 21,607 PGY1 and 19,748 PGY2 residents in the 1998-1999 period (n = 41,355).17 Certain small specialties such as nuclear medicine and medical genetics were excluded, as were residents in Puerto Rico. There was no attempt at oversampling or stratification. The final sample consisted of 6,106 residents (14.8%).
The questionnaire was based on one used for an earlier national survey conducted by the lead author in 1990, as well as on those used for surveys conducted by the AMA in 1979, 1983, and 1987.5,6,18 Questions focused on the broad residency experience including work, sleep, learning, behavioral change, stress, supervision, and impairment, as well as on incidents of humiliation, sexual and racial harassment or discrimination, and observations of unethical or unprofessional behavior. The final questionnaire was five pages long and contained 44 questions covering 144 separate data elements. It was pilot tested for clarity, and cognitive interviews were conducted with residents in different specialties. Revisions were based on this feedback.
The questionnaire was mailed in late April 1999 to all identified residents, along with a cover letter, a prestamped return envelope, and a postcard with the resident’s name and address. Residents were asked to return the postcard and the completed questionnaire separately to maintain respondents’ anonymity, while enabling the investigators to track who had responded to the survey. Residents were also instructed to return the postcard if they declined to participate or were not enrolled in either their first or second postgraduate year. Those who did not return the postcard were mailed a follow-up questionnaire package on three subsequent occasions. In addition, efforts were made to encourage nonresponders through their program directors. Returned questionnaires were accepted until September 1999 and were numbered and coded in the order in which they were received. Correlations were calculated between these identification numbers and reported variables. None were outside the range of r = ± .03.
The index question relating to work hours was “During your current year of residency, on average, how many hours per week did you spend in each of the following activities? (A) Providing direct patient care (excluding moonlighting). (B) All other activities related to your residency.” An additional question asked, “During your current year of residency, did you work outside your program to earn money? If yes, how many hours, on average, did you work (per month)?” On the index question, the reported figures for “direct patient care” and for “all other activities” were combined to derive total work hours for the purposes of this paper.
Of the 6,106 surveys mailed, 295 were found to be undeliverable, and 145 residents responded that they were not in PGY1 or PGY2. An additional 39 were identified as having left their programs before the date of the survey. Two residents had died, and there was no record of nine others ever having been in the identified program. Overall, 3,604 of 5,616 residents responded (64.2%): 1,665 (59%) PGY1 residents and 1,912 (68.7%) PGY2 residents. Twenty-seven respondents did not indicate their graduate year. Response rates were consistent across specialties, although two surgical specialties registered below 50%. Distribution across specialties paralleled national figures.17 There was no significant correlation between response rate by specialty and work hours.
Men constituted 56.5% (n = 2,016) of the respondents. Of the respondents, 40.4% were single (never married), 56.6% were married, 2.9% were divorced or separated, and 0.1% were widowed. A total of 72.3% (n = 2,590) were graduates of U.S. medical schools (USMGs), while 23.7% (n = 849) were international medical graduates (IMGs). Smaller numbers (n = 132; 3.4%) were graduates of U.S. osteopathic schools. Ethnic representations included: Native American/Alaskan Native 0.4%, Asian/Pacific Islander 13.9%, black (not Hispanic) 4.9%, Hispanic 5.5%, Indian Subcontinent 9%, Middle Eastern 2.3%, and white (not Hispanic) 60%. Slightly more than 2% listed themselves as “Other,” or failed to respond. With the exception of slightly fewer men (2.1%) and blacks (0.9%), and slightly more women (2.2%), USMGs (1.3%), and Hispanics (0.7%), all our demographic figures were close to those recorded for the entire group of PGY1 and PGY2 residents reported for 1998-1999.17
Reported Work Hours
Of the PGY1 residents who answered the question, 1,625 reported that they had worked an average of 83 hours per week (see Table 1), and 1,868 of the PGY2 residents reported averaging 76.2 hours per week; a 9.7% difference (p < .0001). There were significant differences across specialties, with general surgery residents reporting nearly twice the average work hours per week as pathology residents in both years. In PGY2, the average work hours in emergency medicine, family practice (FP), and psychiatry declined by as much as ten hours per week, while those in general surgery went up. There was a noticeable consistency in the percentage of time spent in direct patient care by nearly all the specialties; 81.7% in PGY1 and 77.8% in PGY2. A negative correlation (r = -.54) was found between patient care and nonpatient care work hours.
Overall, 49.7% of PGY1 residents reported working an average of more than 80 hours per week, 28.1% over 90 hours per week, and 12.4% over 100 hours per week. For PGY2 residents, the figures were 35.1%, 20.8%, and 10.9%, respectively. Sixteen residents claimed to have averaged 130 hours a week or longer. There was also a marked difference in the distributions of work hours reported within specialties. General surgery and some of the surgical subspecialties were skewed to the right (longer hours), while others, such as psychiatry, FP, dermatology, and emergency medicine, were skewed to the left (shorter hours).
Only 39 PGY1 residents (2.3%) reported working outside their programs, averaging 3.7 hours per week. Approximately 16% of PGY2 residents who responded (n = 304) reported moonlighting, averaging 5.7 hours per week. Few of these residents were in general surgery (2.8%), while 33.2% were from family practice and 28.7% from emergency medicine. Those who moonlighted worked significantly (p < .001) fewer program-related hours (70.4 versus 77.4 hours) than did nonmoonlighters. There was no statistically significant relationship between moonlighting and total indebtedness for USMGs.19
Using the discriminant function analysis feature of SPSS, we derived a model to predict moonlighting. In this model, moonlighting was more common for PGY2 residents, those who worked fewer hours, were USMGs, had more debt, had children, or were men. Using the assumption of equal assignment between classes, this multivariate model correctly identified 79% of the moonlighters and 66% of the nonmoonlighters.
Sleep and other hours
PGY1 residents reported sleeping an average of 40.3 hours per week, or 5.76 hours per night, while PGY2 residents averaged 41.9 hours per week, or 5.98 hours a night, a significant difference (p < .0001). PGY1 residents reported an average of 37.2 hours for their longest single period without sleep, while their PGY2 colleagues averaged 35.7 hours, again a significant difference (p < .0001). There were also significant differences across specialties, with residents in neurosurgery and general surgery averaging 33.5 and 34.7 hours of sleep per week (4.78 and 4.95 hours per night) respectively, while residents in pathology averaged 48 hours per week (6.9 hours per night).
When weekly hours for work and sleep were combined, what remains were the “other” or so-called “free” hours, time residents had to commute to work, shop, study, and take care of personal and business affairs.3Figure 1 displays the distribution of work, sleep, and “other” hours across specialties. There was an average of 47.5 such “other” hours per week for the total sample. However, there was a significant difference (p < .0001), between the figures for PGY1 residents (44.9 hours) and PGY2 residents (50 hours). There were also significant differences across the specialties, ranging from psychiatry (60 hours), to neurosurgery (23.9 hours) and general surgery (29.7 hours).
Reported work hours differed by level of training, gender, and marital status. When data from PGY1and PGY2 residents were combined, men reported longer work hours (80.1 hours) than did women (78.3 hours, p < .05), as did residents without children (80.4 hours) compared with those with children (76.1 hours, p < .0001). Single residents worked 81 hours, compared with 78.3 hours for married residents and 76.4 hours for divorced residents (p < .001). Overall, there was a significant difference (p < .0001) between USMGs (80.4 hours) and IMGs (76.2 hours). These differences were smaller in internal medicine (IM) (80.9 hours for USMGs versus 79.4 hours for IMGs), suggesting that the differences above may be related to the varying representation of IMGs and women in different specialties.
When these results are considered together, residents who were in PGY1, single, had no children, or were men tended to report longer hours, while those who were in PGY2, married, had children, or were women reported shorter hours. Each of these variables produced significant main effects. The interaction terms did not reach statistical significance, suggesting that each of these variables functions independently. The combined impact of these four variables is demonstrated in Figure 2, which shows how much they add to or subtract from the grand mean of working hours reported by the whole sample. At the extreme, these four variables produce a potential difference of 17 hours per week.
Work hour correlates
Table 2 shows that total work hours significantly correlated (Pearson product moment correlation) with residents’ reported levels of stress (r = +.33), and feelings that their current year of residency was more stressful than other programs in their specialty (r = +.21). Significant correlations were also found between total hours and reported hours of sleep per week (r = -.39), the longest period without sleep (r = +.31), and the frequency of periods of prolonged sleep deprivation (r = +.42). Some reflection of the residents’ views of their workload was found in the high correlation between total hours worked and their level of agreement with the statement that the number of hours they were required to work “was too long” (r = +.35).
Standard odds ratios were calculated for work hours and a number of variables (see Table 3). Residents were divided into those averaging 80 or fewer hours per week and those who reported averaging more than 80 hours. The resulting odds ratios compare the rates of events reported within these two groups and provide an index of the risk associated with being over or under the 80-hour figure. Residents who worked longer hours were 1.58 times more likely to have experienced a serious accident or injury, 1.90 times more likely to have had a serious conflict with a fellow resident, and 1.27 and 1.85 times more likely to have experienced conflict with attending staff and with nurses. Longer work hours also meant that residents were 1.62 times more likely to report increased use of alcohol, 1.63 times more likely to report a noticeable weight change, 2.50 times more likely to state that they took medications to stay awake, and 1.36 times more likely to take medications to cope with residency. They were also 1.54 times more likely to report having made a “significant medical error.” With minor variations, these relationships held within medical specialties as well as for the sample as a whole.
Types of Residency Experience
We believe that the average figures reported above obscure what is a considerable variation in work hours across specialties, within specialties, and even within the same residency program. To illuminate this variation, we employed cluster analysis in an effort to delineate different types of residency work experience. Using the SPSS K-cluster program, solutions for two through eight clusters were derived. We settled on a four-cluster solution because it offered the best balance between differentiation while avoiding fragmentation. These four clusters are useful, not as a test of hypothesis, but for their heuristic value in identifying different types of reported residency experience. Three types of variables were selected for inclusion in these analyses: demographic variables, indicators of satisfaction, and residents’ reports of work and sleep time. Variables were retained in the model if mean values showed differences across the four clusters and dropped if mean values showed little differences. This final model was based on nine variables: reported weekly working hours, PGY, weekly sleep hours, USMG versus IMG, general satisfaction, reported stress, gender, having children, and moonlighting hours.
Based on our examination of the mean values for each of the variables used to derive the cluster solution, we labeled these four clusters: High Intensity, Moderate Intensity, Low Intensity, and Moonlighters (see Table 4). The High Intensity group (n = 695) averaged 106.5 hours per week, reported the highest stress, and the lowest levels of satisfaction and hours of sleep per week among the groups. The Low Intensity group (n = 1,186) averaged 60 work hours per week, and reported lower stress and the highest levels of satisfaction and sleep. An offshoot of this Low Intensity group was a small group of Moonlighters (n = 88), who also worked shorter hours (65 hours per week) and were largely PGY2 residents, men, and married with children. They reported high satisfaction, the lowest stress, and did most of the reported moonlighting. The Moderate Intensity group (n = 1,320) had work hours (83.2) and other characteristics similar to the overall figures for the entire sample.
Although these types of residency experiences show noticeable correspondence across specialties, there is also considerable variation within each specialty (see Table 4). For example, although surgery residents tended to be in the High Intensity group, there are residents in some surgical programs clustered in the Moderate Intensity and Low Intensity groups. IM residents were found chiefly in the Moderate and Low Intensity groups. FP residents tend to cluster in Low Intensity experiences, although a few reported High Intensity experiences. These findings suggest that intensity of residency experience offers an alternative way to frame discussions of residency working conditions.
The data reported by the PGY1 and PGY2 residents in our survey indicate that, far from being an outer limit, working close to 80 hours per week represents an average, with approximately half of PGY1 residents and a third of PGY2 residents exceeding this figure. Nearly one quarter of our respondents claimed to be working 90 hours or more per week, and over 11% reported regularly working at least 100 hours per week. Moonlighting was of no consequence among PGY1 residents, and it was limited to a few specialties and residents in PGY2. Moonlighting had little effect on total work hours. Wide variations in work hours were reported across specialties, with residents in the surgical specialties reporting averages of 90 to 100 hours per week, while those in the primary care specialties reported averages close to 80 hours. During PGY2, between 90% and 100% of residents in neurosurgery, general surgery, orthopedic surgery, and obstetrics/gynecology reported working more than 80 hours a week.
These data are averages. Substantial variation exists within every specialty. Demographic variables accounted for some of this variation in working hours. Residents who were women, or had children, or were married, or were in their PGY2 year reported shorter work hours than their counterparts. Selecting from the variables at our disposal, cluster analysis identified four distinct types of residency experience both within and across specialties.
Statistically significant associations were found between longer work hours and higher levels of perceived stress, overall hours of sleep and perceived sleep deprivation, satisfaction with learning, inadequate supervision, working while ill, observations of both others and themselves working in an “impaired condition,” and perceptions of humiliation and belittlement. Residents who worked longer hours were significantly more likely to have reported conflict with others in the workplace, made a significant medical error, and sustained a serious personal accident or injury. They were also more likely to have experienced a variety of behavioral changes, such as significant weight change, increased alcohol use, and greater use of medications to cope with the residency.
The idea of limiting work hours has achieved acceptance elsewhere in the western world. Denmark leads the way, prescribing a limit of 37 1/2 hours per week for all physicians, including those in training.10,20 The European Union, Germany, and the Netherlands, have recently established maximum work hours from 48 to 56 hours per week, while Great Britain and Australia have mandated a 72-hour range.10,20
In the United States, average resident work hours have decreased over the past two decades. Average total work hours per week for PGY1 residents reported on national random surveys conducted in 1983 and 1987 were 84.7 and 87.5 hours per week respectively.5,6 New ACGME regulations limiting resident work hours for all specialties began in July 2003, although many programs had already complied.16
Simple hourly comparisons, however, ignore the differences in specialty and program requirements as well as the enormous changes in hospital care and residents’ responsibilities over this period. Patients today are sicker, spend less time in the hospital, and receive more complicated treatment than earlier.21 Yet, surprisingly, we found that residents reported spending close to the same proportion of their time in direct patient care as was reported in earlier surveys.5,6
The real issue, as we have argued elsewhere, is not merely the time that residents put in, but what they gain from that time.18,21 The mandated 80-hour per week limit on work hours tends to equate all residents and all residency experiences. Unfortunately, this universal rule attempts to resolve the varied problems of residency training by imposing a “procrustean bed,” or “one size fits all” limit on work hours. Our data suggest that the issue is far more complex. Attention needs to be focused on other aspects of the experience including, especially, sleep deprivation and fatigue, the learning environment, and the personal needs and goals of residents.
Other data from our survey suggest that free time resulting from lower work hours may not necessarily be used for sleep. For example, the small percentage of PGY1 and PGY2 residents who moonlighted were also among those who reported the fewest work hours.19 The issue of work hours goes far beyond the simple numbers reported here. Extended work hours impact the residents’ personal life and professional development, with inevitable consequences for learning, health, and well-being, as well as for clinical performance, patient safety, and quality of care.2-4,10,18,21 Their impact must be viewed in the context of the total life of the resident. For example, how do residents utilize the remaining hours of the week? If granted fewer work hours, would they sleep more, or use the time for learning, income augmentation, or personal matters? PGY2 residents in our study used only 1.6 of the 6.8 weekly work-hour difference between the PGY1 and PGY2 years for sleep, a mere 13.8 extra minutes of nightly sleep! We also found that those residents who were married with children tended to work fewer hours and, therefore, had more hours available to meet personal, family, and financial obligations. Data from other studies are consistent with this notion. Residents in obstetrics and gynecology recently gave the “need for more personal time” (76.3%) as their second-highest reason after fatigue (77.6%) for wanting limits on their work hours.2
We believe that the work-hours debate needs to shift from a focus on the number of hours to what is achieved in those hours and from a focus on averages to the reasons for variations within and across residencies.
A Matter of Choice?
Residents are not simply passive reactors in this issue. If they choose to exercise it, residents do have some control over their learning and work experience through the process of specialty and program selection. Our data suggest that demographic factors such as gender, marriage, and children are strongly associated with the intensity of the residency experience. Senior medical students who have greater sleep requirements and/or personal obligations have the option of applying to programs that are known to have less intense environments. Applicants can, and do, take into account program characteristics and their own personal needs and expectations in constructing their rank-order preferences. What applicants need is more complete and reliable information concerning the educational opportunities and the work-hour expectations of individual programs. Better information should lead to a better “fit” between program requirements and the personal and professional needs of applicants.
As with all studies, our report has limitations. The most obvious is that the data are based on self-reports from residents and, thus, the figures may be overreported or underreported owing to differing perceptions of the difficulties of the training years.22 However, the findings are clearly in line with those reported by the New York State Department of Health, as well as most studies.2,5-6,8-10. Furthermore, the relative differences between the specialties are consistent with our prior work.18 A possible nonrespondent bias also seems unlikely because of the large sample, an acceptable response rate, and the close match between the demographic characteristics of our respondents and those of the concurrent nationwide sample.17
Second, in an effort to compare our findings with those of previous national surveys, we asked the residents to provide summary information for their current training year. Our data reflect averages, and it is clear that further research is needed regarding the important variations across time and service commitments, as well as between individual programs.
Finally, our data reflect responses from residents during a particular year. It is possible that the recent expressions of concern about the residency experience from residents, medical students, educators, and the public have had some positive impact since the time of our study. At least, that is our hope.
The authors wish to express their appreciation to Rebecca Miller and Tom Richter, currently at the ACGME, for their assistance in drawing the sample and setting up the survey procedure, and to Beverley Rowley for her help in the design of the survey. This work was supported in part by a grant from the AMA Education and Research Foundation.
The opinions expressed in this article represent those of the authors and should not be ascribed to the organizations with which they are associated.