Secondary Logo

Journal Logo

Well-Being of Students and Residents

Predictors of Persistent Burnout in Internal Medicine Residents: A Prospective Cohort Study

Campbell, Jessica MD, MA; Prochazka, Allan V. MD, MSc; Yamashita, Traci MA; Gopal, Ravi MD

Author Information
doi: 10.1097/ACM.0b013e3181f0c4e7
  • Free


Residency is a critical time in a physician's development. Residents are expected to gain skills in their specialty while simultaneously providing top-quality patient care. The internship year tends to be very time-intensive and stressful; as residents rise through the training program, the numbers of hours they work decrease while their levels of responsibility and competency increase.

The stress of residency places residents in internal medicine and other specialties at risk for burnout. According to Maslach and Jackson,1 burnout is a syndrome characterized by emotional exhaustion (EE), depersonalization (DP), and reduced sense of personal accomplishment (PA). Burned-out residents may feel emotionally depleted and unable to give of themselves such that they develop a cynical attitude, dehumanize their patients, and devalue their own clinical performance. They are also more likely to screen positive for depression.2 Among internal medicine residents, burnout has been linked to self-reported suboptimal patient care, deferred clinical decision making, and increased perceived medical errors.3–5

Despite the implementation of Accreditation Council for Graduate Medical Education (ACGME) work hours restrictions in July 2003, the proportion of residents considered burned out has remained significant,3,4,6 and burnout rates have continued to climb.7 Resident burnout remains a condition with significant side effects and no easy treatment.

In a review of the burnout literature, Thomas8 found that most studies of resident burnout are cross-sectional in design. These studies present the prevalence of burnout at a single point in time and then determine the demographic, personality, and work environment factors that correlate with burnout. The longitudinal studies that exist do not analyze residents individually, limiting the ability to control for personality constructs and follow individuals over time.

To date, there have been three previous significant longitudinal studies on resident burnout. Using the Maslach Burnout Inventory (MBI),1 Michels and colleagues9 prospectively studied all family medicine residents in South Carolina for three years, starting in 1993. Their study, which had a 40% response rate, found that burnout increased during residency. Specifically, residents' DP scores increased during their three years of training (first year = 8.22 versus second year = 8.60 versus third year = 11.43, P < .001), and EE and PA scores did not significantly change throughout training. They also found an association between gender and burnout: Men had significantly higher DP scores than did women.

The second longitudinal burnout study, from New South Wales, Australia, surveyed medical students using the MBI every three months, from May 2000 in their final year of medical school to the end of their internship in November 2001.10 That relatively large study of 117 students had a response rate between 75% and 98% for each of the six times the survey was administered. Burnout increased from 28% of students at baseline to 39% of students at graduation and continued to climb to 75% during the third quarter of internship. At the end of internship, only 61% of residents met criteria for burnout. The survey stopped at completion of internship, so it is unknown whether the burnout rates continued to decline, stabilized, or increased as the respondents progressed through residency.

The third longitudinal study, conducted in Israel by Tzischinsky et al,11 surveyed 78 multispecialty residents at the beginning of their internship, after their first postgraduate year, and after their second postgraduate year, using the MBI. The researchers found that mean burnout scores increased from baseline to the end of the first postgraduate year but then decreased after the second postgraduate year, even though perceived stress levels were higher at that time.

Just two of the aforementioned studies followed residents through all three years of training, and their contradictory results provide little insight into the incidence and prevalence of burnout as residents progress through training programs: Tzischinsky et al11 found decreasing burnout, whereas Michels et al9 found increasing burnout. Researchers do not know whether burnout is a transient condition that resolves over time or a chronic condition. This important distinction needs to be addressed. If the natural course of burnout is resolution, then there may not be a pressing need for interventions. However, if residents burn out and remain burned out throughout their training, and possibly their careers, the need to develop interventions and identify solutions becomes more urgent.

We conducted a longitudinal study of internal medicine residents to test the hypothesis that burnout would lessen as residents progressed through their training. We also examined predictors of persistent burnout. In this article, we attempt to answer the following questions: To what extent does burnout become a persistent phenomenon? What are the chances of recovery, and what are the risk factors associated with being burned out during the entire residency?



University of Colorado Denver Health Science Center (UCDHSC) internal medicine residents who began their residencies in 2003, 2004, 2005, and 2006 (n = 179) were eligible to participate in the study. We excluded preliminary interns from the study because they matriculate to other training programs. The three-year UCDHSC internal medicine residency training program is a large program at a multihospital academic health center and offers two training tracks: categorical and primary care. As residents progress through the program, the number of on-call months, including ward and ICU rotations, decreases (interns and second-year residents: categorical track = eight months, primary care track = seven months; all third-year residents = four months). ACGME-mandated work hours restrictions led UCDHSC in June 2003 to adopt a float system to decrease the number of hours worked by postcall residents.

Data collection

We mailed surveys with an introductory cover letter signed by the investigators, $5 reimbursement, and a postage-paid return envelope to each resident's home each May, from 2003 through 2008. The Colorado Multiple Institutional Review Board approved the research protocol.


Each survey questionnaire incorporated the instruments described below. We asked residents to answer survey questions in reference to their previous month's rotation. In addition, we asked them about their educational and work experience, the rotation they just completed, and the average number of hours worked per week that month.

The self-administered MBI, consisting of 22 items evaluated on a seven-point Likert scale, investigates the three domains of burnout: EE (average = 19–26, high ≥27), DP (average = 6–9, high ≥10), and PA (average = 39–34, low ≤33).1 In the Results, we present the raw scores, as recommended by Maslach and Jackson,1 but to aid in comparison with other studies we converted the three-dimensional definition of burnout into a one-dimensional construct by defining a burned-out resident as having a high EE or high DP score. We defined persistent burnout as meeting the criteria for burnout each of the entire three years of the resident's training.

The Primary Care Evaluation of Mental Disorders Patient Health Questionnaire screens for five mental disorders.12 In this survey, we used the depression subset, which consists of two self-administered questions scored on a Likert scale.

Finally, we assessed confidence with clinical decision-making and bedside procedures by asking residents to respond to a series of statements about their skills, using a five-point Likert scale anchored by “strongly agree” and “strongly disagree.” For procedural skills, in 2003–2005 we asked residents about central lines, arterial lines, and arthrocentesis. In 2006–2007, we added thoracentesis, lumbar puncture, paracentesis, and oral intubation to the procedures.

Statistical analysis

We compared residents who had persistent burnout with those who did not, using chi-square tests. We tested changes over time in categorical variables using the Cochran Q test, which is a test for binary outcomes (e.g., burned out or not burned out), with multiple related samples. In our study, the samples are related because the responses came from the same individuals as they progressed through the residency program. We tested changes over time in continuous variables with a repeated-measures linear model specified with an unstructured within-student variance–covariance matrix that allows the intercept to vary randomly for each student. We used repeated-measures analysis of variance to analyze continuous variables (e.g., EE score) to account for the correlation of the resident's responses over time.

We performed logistic regression modeling to calculate odds ratios (ORs) for persistent burnout. We used variables that were statistically significant in the bivariate analysis (gender and positive depression screen) and placed them into a stepwise regression to obtain ORs. Because we were interested in predictors of persistent burnout, we used intern-year positive depression screen as a predictor.

We conducted statistical analysis using SAS version 9.2 (SAS Institute Inc., Cary, North Carolina). A P value < .05 was considered statistically significant.


Of the 179 eligible residents, 86 (48%) responded to the survey every year that they were in the residency program (Table 1). As the residents gained seniority, the number of residents whose previous rotation was an on-call month decreased from 76 (91%) interns to 54 (63%) third-year residents (Cochran Q = 19.45, P < .001) (Table 2). The average number of hours they reported working per week during the previous month also decreased: Interns reported an average of 75.9 hours, second-year residents reported 66.8 hours, and third-year residents reported 57.7 hours (P < .001). All participating residents completed their training in the expected time frame.

Table 1
Table 1:
Cohort Characteristics of the 86 Internal Medicine Residents Who Completed the Survey for All Three Years of Their Residency, University of Colorado Denver Health Science Center, May 2003 to May 2008
Table 2
Table 2:
Cohort Changes Over the Three Years for the 86 Internal Medicine Residents Who Completed the Survey for All Three Years of Residency, University of Colorado Denver Health Science Center, May 2003 to May 2008

Sixty-seven residents (78%) were burned out at least once during their residency: 58 residents (67%) were burned out during their internship, 58 (67%) during their second year, and 50 (58%) during their third year (P < .08) (Table 2). There was no change in the number of burned-out residents from internship to the second year of residency; however, there was a downward trend from 58 burned-out interns to 50 burned-out third-year graduating residents (Cochran Q = 5.12, P < .08). The mean EE score did decrease from 25.8 among interns to the low end of average (21.0) among graduating residents. The number of respondents with high EE also decreased from 33 interns (38%) to 20 graduating residents (23%) (Cochran Q = 14.00, P < .001). However, levels of DP and PA (mean and percentage with high values) did not change significantly between the first and third years of residency. Fewer residents screened positive for depression as they gained seniority (45 [52%] in the first year versus 23 [27%] in the third year; Cochran Q = 17.79, P < .0001).

As the residents gained experience, their confidence increased. The number of residents who agreed or strongly agreed with feeling “confident in my ability to do bedside procedures” increased with postgraduate year, as did the number who agreed or strongly agreed with feeling “confident in making important clinical decisions” (Table 2).

As the 58 interns with burnout progressed through the residency, 7 (12%) recovered in their second year and 9 more (16%) recovered in their third year (Figure 1). Of the 28 interns who were not burned out, 7 (25%) developed burnout in the second year and 2 (7%) in the third year. A high number of residents, 42 (49%), were burned out during all three years of residency; 25 (29%) were burned out for one or two years, and 19 (22%) were never burned out. Those who were persistently burned out had statistically significant higher rates of a positive depression screen than did all other residents (Table 3). Twenty-nine of the 42 residents with persistent burnout (69%) screened positive for depression as interns, whereas 16 of the 44 residents who were burned out for zero, one, or two years (36%) screened positive (P < .002). Gender was also associated with persistent burnout: 27 (64%) of the 42 persistently burned-out residents were men, whereas 27 (61%) of the 44 occasionally and never burned-out residents were women.

Figure 1
Figure 1:
Residents (n = 86) with and without burnout as they progressed through three years of the internal medicine residency training program at the University of Colorado Denver Health Science Center, May 2003 to May 2008. PGY indicates postgraduate year.
Table 3
Table 3:
Characteristics of the 42 Residents Who Were Persistently Burned Out Versus the 44 Residents Who Were Burned Out for Zero, One, or Two Years, University of Colorado Denver Health Science Center, May 2003 to May 2008

Logistic regression modeling for persistent burnout revealed an OR of 3.31 for male sex (P < .01, CI 1.29–8.49). Screening positive for depression at internship confers an OR of 4.40 (P < .002, CI 1.72–11.28) for persistent burnout. The C statistic for goodness of fit for the logistic regression model was 0.783.


To the best of our knowledge, this is the first longitudinal study concerning burnout across the career of internal medicine residents at a single, large academic center since the implementation of work hours restrictions. In this cohort of residents, we found that most residents were burned out at least once during their residency. The majority of residents were burned out during their internship, and most of the burned-out interns remained burned out for the duration of their residency. The risk of developing burnout after internship was lower.

As the residents progressed through their training, they reported greater confidence in their clinical and procedural skills, fewer hours worked, and less EE. Additionally, positive depression screen rates declined. There was a trend toward decreased burnout rates from the second to the third year of residency that was nearly significant (P < .08) and seemed to be driven by a decrease in EE scores (P < .001). Scores on the DP and PA subscales did not significantly change throughout training. This indicates that DP may drive the persistence of burnout and that EE might be more easily modified.

Our findings stand in contrast to those described by Michels and colleagues,9 who found that the rate of burnout increased as residents progressed through training, driven by increasing DP scores. We found a trend toward decreasing burnout with decreasing EE scores and unchanged DP scores. In part, this difference may be because work hours restrictions have remarkably changed residents' work schedules since 2003, and Michels and colleagues' study was conducted before the restrictions. Our findings are more consistent with Tzischinsky and colleagues'11 study, which noted decreased burnout rates after the second year of residency.

This study has several limitations. We surveyed residents at only one large, multihospital academic health center in the United States. We do not know whether these data can be extrapolated to smaller, nonacademic programs or to specialties other than internal medicine. Also, during the second year of the study, the ACGME work restrictions began to be enforced.6 Although many aspects of our residency program had already been restructured to prepare for the requirements, we do not know whether the further evolution of the program to comply with work hours restrictions altered burnout rates in this cohort, as restructuring the program made the senior residents work more hours than they had in years past. Yet previous studies at this and other internal medicine programs do not show much effect of work hours on resident burnout.7,13,14

Our response rate was low and may bias the results: 48% of eligible residents returned the survey for three consecutive years. Further, as with all surveys, self-reported data may be inaccurate. We attempted to limit this source of error by instructing residents to answer survey questions in reference to their previous month's rotation.

We do not have data on nonresponders and are therefore unable to completely control for nonresponse bias. Because we do not have data on depression for all residents in the program, we do not know whether responders are more likely than nonresponders to be depressed. It is reassuring, however, that the sex and age of responders approximate those of residents in the program as a whole. Finally, there may be other confounders affecting burnout and depression that we did not measure, such as financial distress, personality traits, or relationship strain.


This study offers medical educators and researchers a better understanding of the progression of burnout during internal medicine residency. There was a trend toward decreased burnout as residents progressed through training, but we had hoped that the prevalence of burnout would decrease more substantially as residents gained experience and confidence. We found that after internship, the incidence of new burnout is lower than during internship. Unfortunately, residents who burn out as interns are likely to remain burned out during the entirety of their residency. Men and residents who screened positive for depression as interns are at the highest risk for persistent burnout. Although previous studies have shown that men and women have similar burnout rates,8 we found that women generally score higher on EE scales, whereas men are more likely to experience DP. It may be that lower rates of persistent burnout are found in women because their EE scores decrease as training progresses, whereas men's DP scores stay constant and burnout persists.

We are not aware of any randomized controlled interventional studies concerning resident burnout. Existing interventional studies are limited by small sample sizes and poor design.15 As such, few resources are available to guide residency programs on how to prevent burnout among residents or how to help residents once they are burned out. Our study identifies the intern year as a potential target for intervention. Because residents have a lower risk of becoming burned out in their second and third years, preventing burnout among interns may significantly decrease burnout throughout all of residency. Screening interns for depression and referring them for appropriate care may be a critical element given that depressed interns are at increased risk for persistent burnout. Interventions geared toward male residents, perhaps with an attempt to address the DP component of burnout, may also help. Likewise, female interns may benefit from interventions that target EE.

The problem of resident burnout is not going to solve itself. We found high rates of persistent burnout throughout residency despite work hours restrictions. More research on interventions and treatment of depression in residency, as well as additional longitudinal studies during and after residency, are needed to understand how to make trainees happier and healthier.


This study was funded in part by the University of Colorado Denver Division of General Internal Medicine Small Grant Program.

Other disclosures:


Ethical approval:

This study was approved by the Colorado Multiple Institutional Review Board 03-286.


1 Maslach C, Jackson SE. Maslach Burnout Inventory Manual. 3rd ed. Palo Alto, Calif: Consulting Psychology Press; 1986.
2 Purdy RR, Lemkau JP, Rafferty JP, Rudisill JR. Resident physicians in family practice: Who's burned out and who knows? Fam Med. 1987;19:203–208.
3 Shanafelt TD, Bradley KA, Wipf JE, et al. Burnout and self-reported patient care in internal medicine residency program. Ann Intern Med. 2002;136:358–367.
4 Gopal R, Carrera F, Baker W, et al. Does internal medicine burnout affect self-reported quality of care? J Gen Intern Med. 2005;20(suppl 1):153–154.
5 West C, Huschka M, Novotny P, et al. Association of percieved medical errors with resident distress and empathy: A prospective longitudinal study. JAMA. 2006;296:1071–1078.
6 Accreditation Council for Graduate Medical Education. Common Program Requirements. Available at: Accessed August 22, 2009.
7 Gopal RK, Yamashita TE, Campbell JE, Prochazka AV. Lack of decision latitude and high psychological demands are associated with internal medicine resident burnout. J Gen Intern Med. 2009;24(suppl 1):115.
8 Thomas NK. Resident burnout. JAMA. 2004;292:2880–2889.
9 Michels PJ, Probst JC, Godenick MT, Palsch Y. Anxiety and anger among family practice residents: A South Carolina family practice research consortium study. Acad Med. 2003;78:69–79.
10 Willcock AM, Daly MG, Tennant CC, Allard BJ. Burnout and psychiatric morbidity in new medical graduates. Med J Aust. 2004;181:357–360.
11 Tzischinsky O, Zohar D, Epstein R, Chillag N, Lavie P. Daily and yearly burnout symptoms in Israeli shift work residents. J Hum Ergol (Tokyo). 2001;30:357–362.
12 Spitzer RL, Kroenke K, Williams JB. Validation and utility of a self-report version of PRIME-MD. JAMA. 1999;282:1737–1744.
13 Goiten L, Shanafelt TD, Wipf JE, et 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.
14 Gopal R, Glasheen JJ, Miyoshi TJ, et al. Burnout and internal medicine resident work-hour restrictions. Arch Intern Med. 2005;165:2595–2600.
15 McCray LW, Cronholm PF, Bogner HR et al. Resident physician burnout: Is there hope? Fam Med. 2008;40:626–632.
© 2010 Association of American Medical Colleges