Job burnout is a psychological syndrome characterized by emotional exhaustion, depersonalization, and feelings of decreased personal accomplishment.1 Burnout among resident physicians is common,2–5 affecting between 55%6 and 76%2 of internal medicine (IM) residents late in the academic year. The timing of burnout onset is variable. Some learners will develop burnout during medical school,7–9 with as many as 34%10 to 45%11 of medical students meeting criteria. One study showed a dramatic rise in prevalence from 4% to 55% over the course of the internship year.6
Previous studies have documented numerous correlates of increased burnout for residents, including excessive work hours,4,12 loan debt,11 work–home conflict,11 a variety of personality traits,10,13,14 limited emotional support,3 fatigue,15 and confidence in one's skills, though findings have been inconsistent.10 Having children was found to be protective in one study.10
Job burnout can affect both physician well-being and patient care. Consequences of burnout may include depression, with higher rates observed among burnt-out residents,2,4,6 and possibly suicidal ideation.16 A number of studies report that residents with burnout are more likely to self-report suboptimal patient care2,4 and medical errors.17 It is unclear whether burnout is associated with actual medical outcomes in IM residents.
Though several studies have examined potential risk factors associated with the development of burnout, most were based at a single institution and few have explored factors associated with new cases of burnout, which might better identify at-risk residents and inform potential interventions. We hypothesized that incident cases of burnout would be considerable during the internship year and associated with a number of previously reported measurable risk factors such as residents' psychiatric history, sleep debt, work intensity, and support network. We also hypothesized that other important risk factors might include personality type and frequency of performance feedback. We thus conducted a prospective, multiinstitution, paired-response survey to examine which factors were associated with incident cases of burnout.
Eligible individuals were all incoming preliminary and categorical first-year residents beginning IM or combined medicine–pediatrics internships in 2008 at the following training programs: New York Presbyterian Hospital–Weill Cornell Medical Center (New York, New York), Mount Sinai School of Medicine (New York, New York), Yale University School of Medicine (New Haven, Connecticut), University of Pennsylvania School of Medicine (Philadelphia, Pennsylvania), and Massachusetts General Hospital (Boston, Massachusetts). We offered no incentives to encourage participation in the study.
Our questionnaire has been described previously.10 The surveys included questions pertaining to demographics and resident characteristics, as well as validated instruments used to measure job burnout, sleep deprivation, and personality type. We based questions related to resident characteristics on previously observed associations such as work hours,4,12 loan debt,11 work–home conflict,11 personality traits,10,13,14 support network,15 and confidence.10 We chose also to examine the impact of self-perceived frequency of performance feedback and mentorship14 on the development of burnout and hypothesized that this might be a protective factor. Additionally, we explored possible associations of exercise,14 foreign medical graduate status,14 the availability of backup support and ancillary services,14 and self-reported history of depression or anxiety with the development of burnout. We chose not to use a validated screening instrument to measure depression. We felt that respondents who screened “positive” on such an instrument would need referral for psychiatric evaluation, which would require breaking the anonymity of the study and would identify a mental health condition in an individual who may have wanted to keep such a diagnosis confidential. Further, we postulated that history of depression is a risk factor for the development of burnout, focusing on this historical feature rather than defining new cases using a validated instrument. The pre- and post- surveys differed slightly in that the second survey included factors which could not be measured before the start of training (e.g., average number of patients admitted per call). The validated survey instruments used were as follows.
Maslach Burnout Inventory.
The Maslach Burnout Inventory (MBI) contains 22 questions which measure the participant's level of emotional exhaustion, depersonalization, and sense of personal accomplishment. Respondents use a seven-point scale to describe the frequency of job-related emotional experiences. This questionnaire was previously validated in a large population of physicians and other medical personnel18 and is used commonly in studies assessing burnout in resident physicians.
Some debate exists regarding the scoring of MBI responses to define burnout. The MBI generates a subscore for each of the three dimensions of burnout. The most commonly used definition for clinically significant burnout is a high subscore on either the emotional exhaustion or the depersonalization subscale13,18; however, some argue that a better definition would be high scores on both subscales.6,19 The personal accomplishment domain is typically not used to define a case of burnout because it is felt to measure a quality distinct from the other two.1,2 Others still argue that the best way to measure burnout is by comparing mean subscale scores as continuous data.18,20 In keeping with previous studies and to promote the greatest comparability, we primarily defined burnout as a high score on either the depersonalization or the emotional exhaustion domain (liberal definition) and used the alternative definition of high scores in both domains as a secondary outcome (strict definition). The use of two definitions allows for comparison with a more sizable body of the existing literature.
Epworth Sleepiness Scale.
Using the Epworth Sleepiness Scale,21 participants are asked to rate on a four-point scale their likelihood of falling asleep in eight different scenarios. Scoring greater than 10 is considered significant for excessive sleepiness. This instrument is a validated measure of sleep deprivation.21
Ten-item personality inventory.
This validated survey instrument asks participants to self-report the extent to which they possess five different personality dimensions: extroversion, agreeableness, conscientiousness, emotional stability, and openness to experience.22 These personality traits are believed to be stable in the adult population.23 We correlated scores on each inventory with presence of burnout.
A single unblinded coordinator from the administrative staff of each institution assigned all participating residents a unique identifier and had access to a code key linking each resident's name to his/her unique identifier. The unblinded coordinator had no access to survey questionnaire data, and members of the research team had no access to the code key, thereby ensuring survey response anonymity while maintaining the ability to pair individual responses over time. The authors administered each survey with a cover letter explaining the purpose, potential risks and benefits, and voluntary nature of the study. We gave respondents 10 to 15 minutes to complete the survey on each of two occasions. We administered the survey for the first time just before the start of internship in late June and early July 2008. The great majority of participants completed the first survey before day 1 of training; however, a small number (approximately five) completed the survey in the first week of training, likely because of misunderstanding the instructions. The authors administered the second survey between early April and mid-June 2009, varying by institution. We completed data analysis between December 2009 and April 2010. Participating institutions deidentified all data before interinstitutional transfer for data entry and analysis. Each institution obtained approval from its own institutional human subject research review board before the start of the study.
We reviewed respondents' demographic data using univariate distributions to compare pre and post profiles. We conducted multiple-comparison chi-square tests to identify statistically significant differences in proportions (P < .05) of burnout across survey items. The data were dichotomized with the desired responses (agree or disagree) versus all other responses combined. We used the Fisher exact test when cell sizes were small (n < 5). We used SAS for Windows statistical software, version 9.1 (SAS Institute, Inc., Cary, North Carolina) for data analysis. We analyzed all data with the intention to identify correlations between survey items, burnout incidence, and late-internship burnout prevalence.
A total of 263 first-year residents were eligible for participation in our study. Two hundred fifty-three (96%) completed the first survey, 193 (73%) completed the second survey, and 185 (70%) completed both surveys. Some participants chose not to respond to all survey items.
Demographic information for the cohort at the start of training is listed in Table 1. Categorical residents were slightly though significantly more likely to complete the second survey compared with other residents.
We found no other demographic differences between the group that completed the first survey and those that completed the second.
There were no statistically significant differences in burnout prevalence among participating residency training programs. Burnout prevalence overall was 36% (92/253) at the start of residency training and 81% (154/191) late in the internship year using the liberal definition, which was our primary outcome, with burnout prevalence among participating institutions varying from 28% to 43% (18/65 to 21/49) at the start of internship and from 75% to 84% (36/48 to 32/38) near the end of internship. Using the strict definition of burnout, we found a prevalence of 14% (35/253) at the start of residency training and 50% (95/191) late in the internship year, with burnout prevalence among participating institutions varying from 11% to 20% (7/65 to 10/49) at the start of internship and from 47% to 61% (18/38 to 17/28) near its end.
Of the 185 residents who completed both surveys, 114 were free of burnout at the start of internship. Of these, 86 developed burnout over the course of the year, yielding an incidence rate of 75%. Using the strict definition of burnout, 156 residents began their training free of burnout, of whom 78 developed burnout, yielding an incidence rate of 50%. There were no differences in basic demographic features between incident cases of resident burnout and those who remained free of burnout (see Table 2).
Residents developing burnout over the course of the year (incident cases) were significantly more likely to self-identify as having a disorganized personality style (9/85 versus 0/28; 11% versus 0%; P = .019) and trended toward being less likely to report receiving feedback on a monthly basis (34/54 versus 13/15; 63% versus 87%; P = .057) (see Table 2). Table 3 shows factors associated with the development of burnout using the strict definition. We found no significant associations between burnout incidence and duty hours, current rotation, demographic characteristics, social support network, educational loan debt burden, career plan, or psychiatric history using either definition of burnout.
Discussion and Conclusions
To the best of our knowledge, this is the largest multiinstitution study examining job burnout among first-year IM residents and the first to examine burnout incidence at multiple centers. Many studies looking at change in burnout prevalence of a cohort over time are limited by their inability to discriminate changes that occur on an individual level because some individuals who start burnt out may lose their burnout, whereas others who are free of burnout may develop it later. Previous studies have shown a dramatically high prevalence of burnout in first-year residents and have identified factors associated with burnout; however, none have examined burnout incidence. We chose to focus on burnout incidence as a way to more accurately identify factors associated specifically with the future development of burnout as a means to potentially identify at-risk trainees. Our hope is that characterizing at-risk trainees will take us one step closer to identifying an intervention to minimize burnout.
High rates of burnout among trainees persist despite changes in duty hours aimed at reducing fatigue. Studies examining the impact of reduced work hours show only a modest reduction in burnout with a questionable impact on medical errors and education time.4,12 Newer duty hours innovations such as mandated naps are currently under investigation and may ultimately impact burnout.24 However, it is not likely that scheduling innovations alone will substantially reduce burnout among trainees.
Several previously reported associations with burnout prevalence were unrelated to burnout incidence in our findings, including work hours,4,12 loan debt,11 depression,2,4,6 and social supports.15 The lack of association may be due to differences in study design. We chose to examine residents' self-reported history of depression as opposed to employing a depression screen such as was used in other studies.2 We also explored work hours through participants' self-report, in contrast to prior studies which measured burnout prevalence before and after imposed work hours limitations.4,12 We found that other previously unassociated factors remained as such in our study, including exercise and self-reported sleep needs.10 However, we did confirm that a self-reported disorganized personality style, previously shown to be associated with burnout prevalence at the start of training,10,14 was associated with burnout incidence as well. We also found subspecialty career choice to be related to burnout, though this association could be confounded by a connection with personality traits.25 Of interest, not a single resident who was free of burnout late in internship self-reported a disorganized personality style at the start of training. Though the prevalence of this trait was quite low overall, a lack of this trait may be protective. Other studies have demonstrated associations between personality characteristics and having greater empathy26 or being considered a positive role model27 in IM.
It therefore seems plausible that screening for personality type at the start of training might be one way to identify either at-risk residents or those who will be resilient. It may also be important to identify personality traits early on because some have been correlated with clinical skills,28 stress, learning style, work ethic, and career satisfaction.13 Of note, these traits are self-reported perceptions at a given point in time and may therefore correlate poorly with true personality, representing a limitation of the instrument we used. Though our chosen instrument measures personality imperfectly, it measures a perception that is associated with the development of burnout and is practically useful for identifying potentially at-risk trainees. Once identified, these individuals could be provided with greater guidance and supervision throughout their training.
We found gender to have no influence on burnout prevalence or incidence. Studies examining the role of gender in the development of burnout have found variable results,5 making it difficult to draw any conclusions about this relationship. One recent study did find that male gender was associated with the persistence of burnout throughout the three training years, suggesting the need for further research to explore this correlation.29
Whereas one previous study did examine the relationship between resident burnout and talking with faculty mentors or working in a supportive environment,14 we chose specifically to focus on the impact of receiving regular performance feedback. Formative and summative feedback are increasingly emphasized in resident education.30 Our finding of a significant association between lack of recent feedback and incident burnout suggests that receiving regular feedback might somehow protect against burnout, though our data cannot be used to establish causality. More objective measures of feedback delivery would help clarify whether this association reflects protection through receipt of feedback or variability in the perception of having received feedback based on the degree of burnout. Nonetheless, our finding that feedback may protect against burnout is actionable, and future studies should measure whether more feedback or particular types of feedback would reduce incident burnout.
Some might argue that the very high rate of burnout among residents indicates that it is a natural consequence of training and does not represent an appropriate measure of study. Indeed, we believe that all three features of burnout—emotional exhaustion, depersonalization, and feelings of decreased accomplishment—may be viewed as normal adaptations to the life and death stressors of residency training. Furthermore, limited autonomy and lack of control, which are both commonly experienced in residency, may also lead to burnout.14,31 Given the demands of residency, which include extended work hours, intense and potentially traumatic work conditions,32 work–life imbalance,33 and emotional stress,34 it seems that the emotional exhaustion that defines burnout would be a natural consequence of training. Little has been described about the normal psychological transitions of becoming a resident physician, but the development of pragmatic and efficient medical-decision-making skills may come at the expense of personal patient connection and may yield disillusionment with the profession. However, it is clear that not all residents become burnt out. In nearly every study examining this issue, despite great variability in study design, there remains a group that does not develop burnout. Perhaps some develop burnout only temporarily as they accept a short-term imbalance between personal and professional needs.33 However, given the known association between burnout and self-reported suboptimal patient care2,4 as well as mental health disorders in residents, it seems that the development of true burnout is maladaptive in at least a subset of residents. The current measures we have for burnout may not reliably identify the most at-risk residents; but, given the consequences, they remain important to measure. Future studies might seek to augment the burnout instrument to select those burnt-out residents who are at higher risk for the professional and mental health complications and yield a refined formula for predicting who might become an at-risk resident.
Our study has a number of important limitations. Our response rate is lower than anticipated, and a very small number of residents completed the survey immediately after the start of training as opposed to just prior. Response bias may be present, though the demographic characteristics of those who completed the first and second surveys are similar. The second survey was completed over a longer period of time than the first because of constraints of administering to residents during the academic year, which may lead to different rates of burnout among the institutions, though we found no statistically significant differences in prevalence or incidence across institutions. All of the participating programs are highly competitive academic IM residency programs, so it is difficult to generalize our findings to community-based programs or to specialties other than IM.
The literature to date suggests that there are a number of predictors or possibly mediators that lead residents at the start of training along the path to job burnout. Our study adds a new dimension to this body of literature. Work stress in the form of work hours and intensity, personality characteristics such as pessimism and lack of confidence, and poor social supports have all previously been associated with burnout. Our study adds the notion that self-identified disorganized personality style and a reported lack of appropriate regular performance feedback may predispose residents to developing burnout. Innovations to optimize duty hours to enhance learning and minimize fatigue and work stress are clearly important for protecting residents and patients. Our work suggests that other fundamental aspects of the training process are important as well, including regular constructive feedback for trainees to understand and improve their performance, and career counseling to help trainees confidently plan their professional lives. We believe that a multipronged intervention that includes reduced work hours, training in mindful practice, improved performance feedback measures, and greater guidance for those at highest risk would have the most impact on reducing burnout and its negative consequences. Novel training strategies and interventions that explore the optimal design of these aspects of residency training have the potential to mitigate job burnout and provide a training ground that promotes high-quality patient care without sacrificing physician well-being.
Administrative staff from each participating institution assisted in replication and distribution of survey materials. No independent funding or support was secured for this function.
Each institution obtained approval from its own institutional human subject research review board before the start of the study.
1 Maslach C, Schaufeli W, Leiter M. Job burnout. Annu Rev Psychol. 2001;52:397–422.
2 Shanafelt T, Bradley KA, Wipf JE, Back AL. Burnout and self-reported patient care in an internal medicine residency program. Ann Intern Med. 2002;136:358–367.
3 West CP, Tan AD, Habermann TM, Sloan JA, Shanafelt TD. Association of resident fatigue and distress with perceived medical errors. JAMA. 2009;302:1294–1300.
4 Gopal R, Glasheen JJ, Miyoshi TJ, Prochazka AV. Burnout and internal medicine resident work-hour restrictions. Arch Intern Med. 2005;165:2595–2600.
5 Thomas N. Resident burnout. JAMA. 2004;292:2880–2889.
6 Rosen IM, Gimotty PA, Shea JA, Bellini LM. Evolution of sleep quantity, sleep deprivation, mood disturbances, empathy, and burnout among interns. Acad Med. 2006;81:82–85. http://journals.lww.com/academicmedicine/Fulltext/2006/01000/Evolution_of_Sleep_Quantity,_Sleep_Deprivation,.20.aspx
. Accessed May 31, 2011.
8 Dahlin ME, Runeson B. Burnout and psychiatric morbidity among medical students entering clinical training: A three year prospective questionnaire and interview-based study. BMC Med Educ. 2007;7:6.
9 Guthrie E, Black D, Bagalkote H, Shaw C, Campbell M, Creed F. Psychological stress and burnout in medical students: A five-year prospective longitudinal study. J R Soc Med. 1998;91:237–243.
10 Ripp J, Fallar R, Babyatsky M, David R, Reich L, Korenstein D. Prevalence of resident burnout at the start of training. Teach Learn Med. 2010;22:172–175.
11 Prins JT, Gazendam-Donofrio SM, Tubben BJ, van der Heijden FM, van de Wiel HB, Hoekstra-Weebers JE. Burnout in medical residents: A review. Med Educ. 2007;41:788–800.
12 Goitein L, Shanafelt TD, Wipf JE, Slatore CG, Back AL. The effects of work-hour limitations on resident well-being, patient care, and education in an internal medicine residency program. Arch Intern Med. 2005;165:2601–2606.
13 McManus IC, Keeling A, Paice E. Stress, burnout and doctors' attitudes to work are determined by personality and learning style: a twelve year longitudinal study of UK medical graduates. BMC Med. 2004;2:29.
15 Prins JT, Hoekstra-Weebers JE, Gazendam-Donofrio SM, et al. The role of social support in burnout among Dutch medical residents. Psychol Health Med. 2007;12:1–6.
16 van der Heijden F, Dillingh G, Bakker A, Prins J. Suicidal thoughts among medical residents with burnout. Arch Suicide Res. 2008;12:344–346.
17 West CP, Huschka MM, Novotny PJ, et al. Association of perceived medical errors with resident distress and empathy: A prospective longitudinal study. JAMA. 2006;296:1071–1078.
18 Maslach C, Jackson SE, Leiter MP. Maslach Burnout Inventory Manual. 3rd ed. Palo Alto, Calif: Consulting Psychologists Press; 1996.
19 Brenninkmeijer V, VanYperen N. How to conduct research on burnout: Advantages and disadvantages of a unidimensional approach in burnout research. Occup Environ Med. 2003;60:i16–i20.
20 Dyrbye LN, West CP, Shanafelt TD. Defining burnout as a dichotomous variable. J Gen Intern Med. 2009;24:440.
21 Johns MW. A new method for measuring daytime sleepiness: The Epworth sleepiness scale. Sleep. 1991;14:540–545.
22 Gosling SD, Rentfrow PJ, Swann WB Jr. A very brief measure of the big-five personality domains. J Res Pers. 2003;37:504–528.
23 Terracciano A, McCrae RR, Costa PT Jr. Intra-individual change in personality stability and age. J Res Pers. February 1, 2010;44:31–37.
25 Markert RJ, Rodenhauser P, El-Baghdadi MM, Juskaite K, Hillel AT, Maron BA. Personality as a prognostic factor for specialty choice: A prospective study of 4 medical school classes. Medscape J Med. February 27, 2008;10:49.
26 Shanafelt TD, West C, Zhao X, et al. Relationship between increased personal well-being and enhanced empathy among internal medicine residents. J Gen Intern Med. 2005;20:559–564.
31 Wallace JE, Lemaire JB, Ghali WA. Physician wellness: A missing quality indicator. Lancet. 2009;374:1714–1721.
33 Ratanawongsa N, Wright SM, Carrese JA. Well-being in residency: A time for temporary imbalance? Med Educ. 2007;41:273–280.