Trends and Factors Associated With Physician Burnout at a Multispecialty Academic Faculty Practice Organization

IMPORTANCE Physician burnout is common, and prevalence may differ throughout a clinician’s career. Burnout has negative consequences for physician wellness, patient care, and the health care system. Identifying factors associated with burnout is critical in designing and implementing initiatives to reduce burnout. OBJECTIVE To measure trends and identify factors associated with physician burnout. DESIGN,SETTING,ANDPARTICIPANTS SurveystudyconductedfromMay16toJune15,2014,and againfromMay16toJune15,2017,measuringratesofphysicianburnoutinalargeacademicmedical practice. Factors associated with burnout out were evaluated. In 2014, 1774 of 1850 eligible physicians (95.9%) completed the survey. In 2017, 1882 of 2031 (92.7%) completed the survey. EXPOSURES Medical specialty, demographic characteristics, years in practice, and reported rates of burnout. RESULTS Respondents included 1027 men (57.9%) and 747 women (42.1%) in 2014 and 962 men (51.1%) and 759 women (40.3%) in 2017. The mean (SD) number of years since training completion was 15.3 (11.3) in the 2014 survey data and 15.1 (11.3) in the 2017 data. Burnout increased from 40.6% to 45.6% between the 2 points. The increased rate was associated with an increase in exhaustion (from 52.9% in 2014 to 57.7% in 2017; difference, 4.8%; 95% CI, 1.6%-8.0%; P = .004) and cynicism (from 44.8% in 2014 to 51.1% in 2017; difference, 6.3%; Maslach Burnout Inventory and a 5-point Likert scale); (3) administrative workload on physicians; and (4) leadership and diversity content. 2 The Maslach Burnout Inventory evaluates physician burnout by grouping items into 3 scales: exhaustion, cynicism, and professional efficacy. 2 As described in the Maslach Burnout Inventory manual, a score of 3.0 or greater on the exhaustion subscale, 2.0 or greater on the cynicism subscale, or 4.0 or greater on the professional efficacy subscale denotes a high level of burnout for the respondent in that subscale. 2 Many studies measuring physician burnout using the Maslach Burnout Inventory use the convention that burnout is determined by the percentage of physicians scoring high on either emotional exhaustion or cynicism. We elected to use the 3 domains because this was the original descriptionofthetool.Giventheconcernthatusingtheoriginalversionoftheinventorywouldresult in underreporting of our faculty group’s burnout rate, we calculated the data using both the 3 domains originally described in the Maslach Burnout Inventory and the revised version, using 2 domains and dropping professional efficacy. The results were similar, with a 5% increase noted in both calculations between 2014 and 2017. Career satisfaction is measured using a 5-point Likert scale (verysatisfied,satisfied,neutral,dissatisfied,orverydissatisfied). 21 Allmethodsforbothsurveysare in compliance with the American Association for Public Opinion Research (AAPOR) reporting guideline


Introduction
Physician burnout has been described extensively in the peer-reviewed literature. [1][2][3][4] Freudenberger 1 first described the concept of staff burnout in terms of physical signs and behavioral indicators.
Maslach et al 2 defined burnout as a syndrome characterized by depersonalization, emotional exhaustion, and sense of low personal accomplishment.
Burnout rates among physicians vary by geographic location, experience level, specialty, age, and sex. 4 It is estimated that at least 35% of physicians in the developing world and 50% in the United States experience burnout, now considered by many experts to be an epidemic. 4,5 Burnout rates are higher for physicians engaged in the front lines of care, including family medicine, emergency medicine, and general neurology. [6][7][8] Burnout carries significant professional and personal consequences. [6][7][8][9] Physician burnout is negatively associated with altruism, professionalism, and quality and safety of care. [9][10][11] At the individual level, physician burnout has been associated with cardiovascular disease, alcohol use, depression, suicide, and shorter life expectancy. [12][13][14][15][16][17][18][19] Burnout has also been implicated as having an adverse financial impact on practices and organizations. 6 Here, we present data on administrative burden from a hospital-wide physician practice survey conducted in 2017. When appropriate, we compare these data with those collected in 2014 using the same instrument and sampling the same population.

Survey Instrument
The Massachusetts General Physicians Organization conducts a biennial survey of active members of this multispecialty academic faculty practice. Qualifying physicians earn a financial incentive for completing the survey, which is part of the organization's Quality Incentive Program. 20,21 Eligibility for the incentive program is based on percentage of time physicians spend on clinical activity.
Incentive amount for completing the survey ranges from $166.67 to $833.34 depending on the percentage of time the participant spends on clinical activity. Physicians are asked to consent to survey participation on the cover letter inviting them to participate. The survey, approved by the Partners human research committee, was most recently administered in 2017. It assesses physician perceptions of the functioning of the clinical enterprise within and across departments, reflects progress made on organizational priorities, and evaluates hospital leadership. 20,21 The survey assesses 4 domains: (1) physician career and compensation satisfaction; (2) physician well-being (assessed using the Maslach Burnout Inventory and a 5-point Likert scale); (3) administrative workload on physicians; and (4) leadership and diversity content. 2

The Maslach
Burnout Inventory evaluates physician burnout by grouping items into 3 scales: exhaustion, cynicism, and professional efficacy. 2 As described in the Maslach Burnout Inventory manual, a score of 3.0 or greater on the exhaustion subscale, 2.0 or greater on the cynicism subscale, or 4.0 or greater on the professional efficacy subscale denotes a high level of burnout for the respondent in that subscale. 2 Many studies measuring physician burnout using the Maslach Burnout Inventory use the convention that burnout is determined by the percentage of physicians scoring high on either emotional exhaustion or cynicism. We elected to use the 3 domains because this was the original description of the tool. Given the concern that using the original version of the inventory would result in underreporting of our faculty group's burnout rate, we calculated the data using both the 3 domains originally described in the Maslach Burnout Inventory and the revised version, using 2 domains and dropping professional efficacy. The results were similar, with a 5% increase noted in both calculations between 2014 and 2017. Career satisfaction is measured using a 5-point Likert scale (very satisfied, satisfied, neutral, dissatisfied, or very dissatisfied). 21  for survey studies. 22 We report and compare data from the 2017 and 2014 Massachusetts General Physicians Organization surveys.

Statistical Analysis
To compare demographic distribution, level of burnout, and level of satisfaction between the 2014 and 2017 surveys, we used χ 2 tests and t tests, as appropriate. We then used multivariable regression to evaluate the association of factors with selected outcomes. All statistical tests were 2-sided and P values less than .05 were considered statistically significant. Regression results are reported as odds ratios (ORs). All statistical analyses were performed using SAS statistical software version 9.4 (SAS Institute, Inc).  (difference, 4.2%; 95% CI, 3.0%-5.3%; P < .001) ( Table 2). The top 3 duties requiring more faculty engagement between the 2 survey periods included time spent on prior authorizations, time spent on specialty recertification, and time spent on compliance regulations related to opioid prescribing.

Survey
Demographic factors associated with burnout, including sex, ethnicity/race, years of experience, and specialty, were evaluated using multivariate analyses. Work-related factors were also included in the assessment of variables deemed to be associated with burnout.  time and resources for continuing medical education, opportunity to affect decision making, workload, and having a trusted advisor were associated with low likelihood of burnout.
In both surveys, physicians were asked to indicate their plan for the next 3 years. Table 4 presents the association of emotional exhaustion and physicians' plan. We first sorted physicians into quartiles of emotional exhaustion (from lowest to highest) and then calculated the percentage of physicians who planned to continue in their current role as well as those who planned to make changes. In the 2017 survey, relative to physicians in the first quartile, physicians in the fourth quartile were less likely to continue in their current role (80.8% vs 44.4%), more likely to cut their work hours (9.0% vs 31.5%), more likely to decrease the number of patients in their practice (5.2% vs 20.2%), and more likely to relocate to another practice (3.2% vs 17.4%). Aspirations to attain an administrative role in the organization and to retire from practice were less associated with level of exhaustion. In multivariate analysis and after controlling for basic demographic characteristics, exhaustion was associated with several outcomes, but not with intent to seek an administrative position within the organization (Table 4). A 1-unit increase in exhaustion (scale of 0-6) was associated with 38% decrease in likelihood of continuing in current role (95% CI, 42%-33%), 58% increase in likelihood of reducing work hours (95% CI, 44%-73%), 48% increase in likelihood of reducing number of patients in practice (95% CI, 32%-65%), 61% increase in likelihood of relocating to other practice (95% CI, 42%-83%), and 38% increase in likelihood of retiring (95% CI, 15%-66%).
The results were similar for the 2014 survey. high score on 2 of the 3 burnout scales is defined as 1 and 0 otherwise. An alternative analysis using a continuous specification of exhaustion score and cynicism score resulted in similar results. We used the binary definition here to reduce the number of tables. b All the satisfaction and agreement factors were specified as binary variable where very satisfied and satisfied (strongly agree and agree) are coded as 1 and everything else as 0. In our study, we found physicians were more vulnerable to emotional exhaustion than any of the other subscales of burnout. Physicians reporting high levels of exhaustion were more likely to reduce their clinical schedules, reduce the number of patients in their practice, leave the practice, or retire.

JAMA Network Open | Health Policy
The association between burnout and physician clinical work effort, as well as turnover, has been described extensively in the literature. [25][26][27][28][29][30] In a longitudinal prospective study 30 evaluating the association between burnout and physician work effort, the authors reported that for every 1-point Physician turnover is associated with patient and clinician distress, as well as financial cost to the system. Studies show that the cost of replacing a physician, irrespective of practice, is 2 to 3 times the individual's annual salary. [25][26][27][28][29] Primary care physicians in our survey were more likely to report high levels of exhaustion when compared with medical specialists, suggesting they represent a population requiring targeted interventions to address exhaustion. These findings may be associated with the amount of time primary care physicians spend documenting on the EHR and serving as the clinicians responsible for the management of patients' multiple complex medical and social problems. 23 After adjusting for potential relevant factors, we found physicians with more than 30 32 These findings point to potential opportunities in this vulnerable group to mitigate burnout, such as initiatives that promote community building and networking and harnessing effective leadership. 8 In a study 33 including 2813 physicians, after adjusting for relevant variables, each 1-point increase in leadership score was associated with a concomitant 3.3% decrease in burnout and 9% increase in career satisfaction (P < .001).
In multivariate analysis, satisfaction with workload and relationship with colleagues were associated with lower burnout rates, while more time spent on administrative duties and lack of control over one's schedule had inverse associations. Other studies have shown that these factors are associated with burnout. 30,34 Continued decline in reimbursements coupled with increased expenditure in health care systems, such as investments in EHR platforms, have resulted in significant capital expenditures. These demands have been addressed by health care organizations, at least in part, by placing increasing productivity and higher efficiency demands on physicians and reducing health care delivery cost. 25 Furthermore, the health care industry is faced with larger demands for implementation, measurement, and reporting of quality metrics, for example, and growing assessment centered on reporting patient satisfaction and experience. 25 The end result has been increasing administrative tasks as part of physicians' daily responsibilities, often accompanied by allocation of fewer resources. Each of the burnout domains identified in this study is influenced by the individual physician, practice, organization, and national health care system. Addressing factors associated with burnout identified in our study and extensively documented in the literature will necessitate a shared commitment from both health care organizations and individual physicians. 25,34 Furthermore, efforts to enhance and maintain connectivity among physicians are also critical.

Limitations
Our study has several limitations. Our findings may not be generalizable to the physician population at large because the study was conducted in a single academic medical center. Although all respondents spend some time delivering clinical care, many of them also spend significant portions of their time engaged in research, education, and other administrative roles. Another study limitation is our classification of burnout into a dichotomous variable rather than a continuum of low to high with reported cutoff values for each classification. The result of our applied methodology may be an overestimate of burnout rates. 35 When we evaluate our data using the continuum methodology, we observed a proportional increase in burnout between the 2 points comparable to the results using a dichotomous variable. The anonymous nature of the study is another limitation. For example, we do not have data on specific actions physicians intending to address their burnout undertook, or the