Resident burnout has been a topic of growing interest, with few people doubting the impact of stress and burnout on job performance (1). Studies have found burnout to be associated with decreased job performance (2), low career satisfaction (3), and stress-related health problems (4). Bellini et al. (5) found that empathic concern among interns decreased in proportion to an increase in personal stress. Recently, Lockley et al. (6) linked extended work hours to increased attentional failures. Our previous work (7) found higher burnout prevalence among those working more than 80 hours per week. The research to date seems to implicate extended work hours as the reason behind burnout, stress, or attentional failures, which, in turn, likely translate into decreased job performance and learning.
An opportunity to test the effects of work hours on the prevalence of burnout was presented in July 2003 when Wayne State University School of Medicine implemented a nationally required measure by the Accreditation Council for Graduate Medical Education (ACGME) to reduce resident work hours. It was a unique opportunity to replicate our survey on burnout prior to the work hour limits and explore whether a change occurred.
We hypothesized that: 1) the rate of burnout among residents post-work hour limitations would be lower compared to pre-work hour limitations; and 2) first-year residents would have higher burnout rates than other residents (due to the presence of other factors that may not be corrected with work hour limits). In addition, we explored potential causes of burnout: number of hours worked, number of nights on call, lack of support/other stresses, satisfaction with faculty, demographic characteristics of gender, and country of training (U.S. versus international medical graduates).
Dr. Christina Maslach, author of the Maslach Burnout Inventory (MBI)(8), describes burnout as a combination of three factors or dimensions: depersonalization, emotional exhaustion, and the lack of personal accomplishment. The MBI, a validated and reliable tool for measuring burnout (9) that has been used in several studies on burnout, uses 22 statements that encompass three subscales based on these dimensions of burnout. Each statement requires participants to rate the frequency of such feelings ranging from never to every day. Standard validated cutoffs (or categorical representations of burnout) (8, 9) were used to determine if burnout was present.
All residents and interns in general surgery, internal medicine, family medicine, obstetrics/gynecology, pediatrics, and psychiatry in all years of training at Wayne State University School of Medicine were eligible for participation in the study, with the exception of the first author (a PGY-4 psychiatry resident at the time of conducting the study).
A package containing the MBI questionnaire, a consent form for participation in the study, and a postage paid return envelope was mailed to the residents after obtaining verbal consent from each residency program director. Also included in the package was a supplemental data sheet which requested the following information: year in training, program affiliation, gender, country of training, marital status, number of children at home, satisfaction with attending faculty, number of working hours per week, number of nights on call this month, number of nights on call last month, family stressors, financial stressors, and mood rating.
The packets were sent in three consecutive mailings between February and May of 2004 corresponding to time of the prior survey (7) to avoid confounding the results with seasonal trends. The total number of responders was 118 out of 384 (31%). Of those responding, 34% indicated they had participated in the prior year’s survey. The study was approved by the Wayne State University Institutional Review Board.
Burnout was present in 41% (95% confidence interval 33%, 50%) of the responding residents. The confidence intervals overlap the prevalence found in the survey (7) conducted prior to work hour limitations (49%). Only 11.1% of the current residents reported working over 80 hours per week. However, the burnout prevalence increased as the number of hours worked across four categories (<40; 40–59; 60–80; and >80 hours) increased (linear by linear χ2=5.12, df=1, p=0.024). Residents who reported working more than 80 hours per week had higher prevalence of burnout (69.2%) compared to those working fewer hours (38.5%: Fisher’s exact test, p=0.034).
First-year residents reported 43% (95% confidence interval 25%, 61%) burnout prevalence, significantly lower than the 77% burnout prevalence among first-year residents prior to implementation of work hour limitations. This lower prevalence brought their burnout prevalence down to those seen among other residents.
Satisfaction with clinical faculty was inversely related to burnout (linear by linear χ2=8.99, df=1, p=0.003). Mood was also negatively associated with burnout as only 24% of those “happy,” 48% “neutral,” and 73% among those “sad” or “depressed” reported burnout (linear by linear χ2=10.76, df=1, p=0.001).
There was no rank or linear association between number of nights on call and prevalence of burnout. Marital status, gender, presence of children, recent family stress and country of training similarly were not associated with prevalence of burnout.
Two key findings are significant in our survey: 1) The prevalence rates of burnout among first-year residents were significantly lower in this survey compared to our previous survey conducted prior to the work hour reform; 2) the overall burnout prevalence rates among residents were somewhat lower after the implementation of work hour limits compared to our survey results prior to such implementation. The overall difference was probably mostly accounted for by the difference in burnout in first-year residents.
A study by Gelfand et al. (10) reported no difference in surgery residents’ burnout before and after work hour limitations. This difference may be explained by perceived lack of anonymity (and other possible design specifications, which were not scrutinized due to lack of IRB review and approval) inherent in the design of the Gelfand et al. study (10), possibly leading to reporting bias. As a matter of interest, surgery residents in our study were slightly, yet not significantly, less burned out than during the previous year (this year 40%, previous year 42%). However, the small number of responses in various disciplines makes meaningful statistical comparison separately for each discipline impossible.
The significant lower prevalence of burnout among first-year residents following the implementation of work hour limits supports the hypothesis that work hours reform was needed most among interns new to the system who had not yet developed ways to cope with the demands of a busy schedule.
Our findings are limited by a low response rate, which, unfortunately, is not uncommon in medical education (11). Low response rates in medical education may be due to distrust of surveys, fear of loss of confidentiality, perceived possibility of retribution, perceived negative effect of reporting, or competing demands of work. We attempted to address this issue within the design of a traditional mail-survey study. We designed our study (three consecutive mailings approximately 1 month apart) to maximize the yield of responses within a limited time frame. We also used statistical comparisons, which may better allow addressing the issue of smaller numbers.
Because of the low response rate, our study, similar to others, does not allow control for another limitation—we do not know if those with high, or conversely, those with low level of burnout were more likely to respond. The response rate, however, is similar to that of the previous survey. There is no evidence to suggest any change in who would be more likely to respond over time. We can assume that, by chance, residents with both high and low burnout responded to our survey.
Finally, the residents’ probable knowledge that the purpose of the study was to measure improvement in job satisfaction (i.e., decreased burnout) following changes in work hour limits is another possible limitation. Residents have a vested interest in maintaining work hour limits and may adjust the rating accordingly. However, there was a significant difference in response among first-year residents compared to the previous year (this year 42.9% of them met criteria for burnout; in the previous year 77.3% did), while there was no significant difference between this and the previous year among residents in other years of training (this year 44.1%, previous year 41.8%). Thus, it seems logical to conclude that the residents’ responses were not affected by assumptions about the purpose of this study that would be self-serving and might have biased the results. Furthermore, first-year residents were not exposed to the previous year’s survey and would not be able to compare their responses to the previous year’s responses.
Our most important finding is the reported lower prevalence of burnout among first-year residents after the implementation of work hour limits. This is clearly in accord with the vision behind the ACGME’s implemented changes that were expected to have a positive impact on medical education. Another interesting finding was the statistically significant inverse relationship between burnout and satisfaction with clinical faculty. This may speak to the importance of strong clinical models in maintaining morale during training. Unfortunately, our survey did not provide any space for additional comments from residents regarding this and other issues.