Burnout Rate and Risk Factors Among Anesthesiologists in the United States

Anoushka M. Afonso, M.D.; Joshua B. Cadwell, M.B.A., M.S.; Steven J. Staffa, M.S.; David Zurakowski, Ph.D.; Amy E. Vinson, M.D.

Disclosures

Anesthesiology. 2021;134(5):683-696. 

In This Article

Materials and Methods

This study was determined to be exempt by the institutional review boards of Memorial Sloan Kettering Cancer Center (New York, New York) and Boston Children's Hospital (Boston, Massachusetts) in July 2019 and January 2020, respectively.

Participants

Invitations to participate with objectives of the study and a survey link were distributed by the ASA via email to all ASA member attending anesthesiologists in the United States. Participation was not mandatory, there were no incentives to participate, and all responses were anonymous. The initial email was sent on March 6, 2020, and was followed by two weekly reminders (March 14 and 21). A third reminder was planned for March 28, but this was canceled out of concern for the escalating COVID-19 pandemic and after ASA member feedback to the investigators raising concern for the overall email burden physicians were experiencing. The principal investigators did not believe it was ethical to continue to send reminder emails given the cognitive overload many individuals were beginning to experience during the initial phases of the pandemic.

Survey Questionnaire

Burnout was assessed using the validated Maslach Burnout Inventory Human Services Survey.[5] Our additional survey questions focused on both personal and occupational risk factors for burnout. The study survey was designed on the basis of guidance from the American Association for Public Opinion Research (Lenexa, Kansas).[19–21] The demographic and practice information questions were developed primarily by the authors after performing a literature review and receiving input from the ASA Committee on Physician Well-being and approval from the ASA Executive Committee for distribution to the membership. Survey development and pretesting, which included online interface, usability, and technical functionality of the electronic questionnaire, were tested on a targeted group of approximately 15 professional colleagues. Intended survey logic was retained throughout with all responses obtained during this testing phase erased before survey distribution to the ASA membership and not included in statistical analysis. Participants were first asked via email to participate in the voluntary study, with explicit assurance of confidentiality; this email contained a link to the 35-question survey tool (Appendix). Explicit consent was not required by either institution's institutional review board for participation in this survey study. Questions were presented in a forced-response format, with optional demographic responses, without any incentives offered.

Participants were asked to provide information on demographic characteristics (including age, gender identity, inclusion in a vulnerable or underrepresented group, and English as a second language), primary practice environment, availability of a professional mentor at work, length of time since completion of training, recent staffing shortages at work, level of support in their professional and personal lives, and magnitude of caregiving responsibility at home.

The survey also included the full version of the Maslach Burnout Inventory Human Services Survey, which is the predominant metric for assessment of burnout among physicians.[1,22,23] The Maslach Burnout Inventory Human Services Survey uses 22 items to assess levels of three dimensions: emotional exhaustion (nine items), depersonalization (five items), and feelings of personal accomplishment (eight items). Each item is scored using a 7-level scale ranging from 0 to 6 (from never to every day), allowing for subscale assessment in all three dimensions.[5] To facilitate potential comparisons to large previous studies of burnout in U.S. physicians, in a manner similar to these studies,[3,4,24] we considered a high score on emotional exhaustion (greater than or equal to 27) and/or depersonalization (greater than or equal to 10) to indicate a high risk of burnout.[1] Whereas the Maslach Burnout Inventory Human Services Survey assesses burnout over a continuum, to identify those with more significant degrees of burnout, we also classified the combination of a high score on emotional exhaustion and depersonalization and a low score on personal accomplishment (less than or equal to 33; i.e., all three dimensions present, using the same scoring thresholds as previously described for "high risk for burnout") as burnout syndrome, consistent with definitions by the World Health Organization[25] and Maslach et al.[5] Responses were automatically captured into SurveyMonkey without any participant identifiers for further analyses.

Statistical Analysis

Descriptive statistics of responses are presented as frequencies and percentages (for categorical variables) and medians and interquartile ranges (for continuous variables). Missing data in the final analysis sample were negligible; denominators are presented to indicate instances of missing data. Burnout rates are presented as frequencies and percentages, and means and SDs are presented for each continuous subscale (emotional exhaustion, depersonalization, and personal accomplishment). Assessment of the generalizability of the study respondents was performed by comparing age, gender identity, and time since completion of training between the analysis sample and the overall ASA population. Age and time since training completion were modeled as outcome variables in two separate median regression models with group indicator (survey respondents vs. ASA population) as a covariate, and the coefficient and 95% CIs are reported to estimate the difference in medians with corresponding 95% CI.[20] Differences between proportions were calculated for gender identity using exact 95% CIs.

For statistical analysis, work support questions were considered in three categories (Not at all/A little, A moderate amount, A lot/A great deal), and other Likert scale questions were dichotomized as (Not at all/A little/A moderate amount vs. A lot/A great deal). Practice environment, gender identity, and caregiving responsibilities were coded as a categorical variable, age was dichotomized as younger than 50 yr, and all other variables were considered as dichotomous predictors.

Univariate comparisons were performed by comparing respondents with and without one manifestation of burnout (high score on the scales for emotional exhaustion and/or depersonalization) and by comparing respondents with and without burnout syndrome. Demographic and practice characteristics and support perceptions were analyzed using the Wilcoxon rank-sum test for continuous variables and the chi-square test for categorical variables. After univariate associations were determined for screening, all variables with P < 0.05 on univariate testing were included in the multivariable logistic regression modeling. A final multivariable model was fit after backwards elimination model building to obtain the adjusted associations between each potential risk factor and burnout, with the purpose of identifying independent risk factors associated with burnout. Results from multivariable modeling are presented as adjusted odds ratios with corresponding 95% CIs and P values. A post hoc supplemental analysis was performed to determine the significant risk factors for emotional exhaustion, depersonalization, and personal accomplishment, using univariate and multivariable linear regression modeling, with results presented as adjusted coefficients with 95% CIs and P values.

No statistical power calculation was conducted before the study, because the sample size was based on the number of complete survey responses. For all statistical analyses, a two-tailed P < 0.05 was considered to be statistically significant. All statistical analyses were performed using Stata (version 16.0, StataCorp, USA).

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....