Methods
This is a cross-sectional, survey-based, national study conducted during the COVID-19 pandemic from April 14, 2020 to April 25, 2020. The self-administered, anonymous online survey collected demographic data and mental health measurements from otolaryngology physicians from academic institutions throughout the United States. Participation was voluntary, and participants were allowed to terminate the survey at any time. A REDCap (Research Electronic Data Capture) database was developed specifically for this project and used to capture survey data. It was accessible only to study personnel. This project was reviewed and determined to qualify as quality improvement by the University of Pennsylvania’s Institutional Review Board.
We contacted otolaryngology program directors via e-mail from all 109 allopathic academic programs in the US to disperse the survey to their residents, fellows, and attendings. Demographic data were self-reported by the participants, including sex (male or female), age, occupation (attending physicians, fellows, resident physicians), and geographic location. Date of projected peak resource utilization for each state was obtained from the Institute for Health Metrics and Evaluation’sCOVID-19 Projections in order to categorize participants based on the “Surge status” of their state.24 States reaching their date of projected peak resource use during our study period were in the “Surge”, while states that had not reached that date were “Pre Surge,” and states that were already past that date were “Post Surge.” Numbers of positive COVID-19 cases and numbers of COVID-19 deaths per state were obtained from the COVID Tracking Projectfrom date April 19, 2020, the midpoint of our study period.25
We focused on symptoms of burnout, anxiety, distress, and depression for all participants, using validated measurement tools.26-30 The single-item Mini-Z burnout assessment (range, 1-5) was used to assess burnout, with burnout defined as> 3.27,28 The 7-item Generalized Anxiety Disorder (GAD-7) scale (range, 0-21) was used to assess symptoms of anxiety over the past two weeks, with a scale of normal (0-4), mild (5-9), moderate (10-14), and severe (15-21) anxiety.26A score of 10 has been reported to be a cut-off point for identifying cases of GAD. The GAD-7 included a final question assessing the “difficulty [these problems] made it for you to do your work, take care of things at home, or get along with other people” (range, 0-3). The 15-item Impact of Event Scale (IES; range, 0-75) was used to assess symptoms of distress over the past seven days, with a scale of subclinical (0-8), mild (9-25), moderate (26-43), and severe (44-75) distress.29 A score of 27 has been reported as a cut-off for risk of post-traumatic stress disorder (PTSD).31 The IES total score was also divided into two sub-scores: intrusion (range, 0-35) and avoidance (range, 0-40). Per Horowitz et al., the intrusion sub-scores assessed symptoms of “unbidden thoughts and images, troubled dreams, strong pangs or waves of feelings, and repetitive behavior.”29 The avoidance sub-score measured “ideational constriction, behavioral inhibition and counterphobic activity, and awareness of emotional numbness.”29 The 2-item Patient Health Questionnaire (PHQ-2; range, 0-6), was used to assess symptoms of depression over the past two weeks, with a score of 3 as the cut-off for a positive depression screening requiring further evaluation with the more in-depth PHQ-9.30 These categories were based on values established in the literature.26-30
Data analysis was performed using R software version 3.6.3. The difference in distribution of symptoms across multiple groups is tested by the chi-square independence test (Table 2 ) and by the nonparametric Wilcoxon rank sum test and Kruskal-Wallis test (Table 3 ). To determine risk factors for severity of burnout, anxiety, distress, and depression, multiple logistic regression models were used (Table 4 ). The binary outcome variables were created for anxiety (normal vs other categories) and for distress (subclinical vs other categories). Type of physician, sex, age, surge status, and number of positive cases were included in the model, while location and number of deaths were found to be highly correlated with the number of positive cases and therefore excluded to alleviate the issue of collinearity. All tests were two-sided and the significance level α=0.05 was applied. 95% confidence intervals were constructed, where applicable.