Changes in Pediatric Faculty Burnout During the COVID-19 Pandemic

We conducted serial cross-sectional surveys of pediatric faculty physicians at an academic, tertiary-care children’s hospital located in New York City at 3 distinct time periods. Our institution is part of a large, integrated academic health system and is the referral center for children in the health system and in the area. The children’s hospital is located on the main campus of our health system and housed in a separate building, although is connected to the adult hospital and therefore shares the same administrative structure, laboratory, pathology, and radiology services. Our faculty work with 86 residents, 4 chief residents, 66 fellows, and 94 advanced practice providers. This study was approved by our institutional review board.

The initial survey invitation (February 20, 2020, to March 20, 2020) was sent to examine baseline prevalence of factors associated with physician burnout. As part of our institutional departmental commitment to wellbeing, we planned to monitor these factors every 6 months. However, immediately after completion of the initial survey period, our institution coincidentally experienced a COVID-19 patient surge, causing significant changes in faculty work environment, workflow, and worklife factors (Fig 1).

To gauge the impact of these changes and stressors on pediatric faculty burnout, we administered a second survey (April 20, 2020, to May 20, 2020) during the peak of COVID-19 cases at our institution, with >75 COVID-19 cases at our children’s hospital (Fig 1) and >1000 cases within our hospital system. At the peak of the COVID-19 surge, pediatric faculty were caring for pediatric COVID-19 patients, deployed to care for adult COVID-19 patients in a unit in our children’s hospital, and deployed to care for adult COVID-19 patients in temporary units for convalescent patients.⁵  Other faculty joined an expanded palliative-care team and called COVID-19 patients and family members daily.⁶  Outpatient-based pediatricians shifted toward the use of telemedicine in place of face-to-face clinic visits. Institutional strategies were developed to address these challenges (Supplemental Table 2).

The third survey period (September 1, 2020, to October 1, 2020) was part of the initially planned semiannual schedule to monitor burnout prevalence. At this time, pediatric faculty no longer cared for adult patients. On average, <5 COVID-19 patients were admitted to the children’s hospital daily (Fig 1), and <30 COVID-19 patients were admitted to the entire health system daily. However, we theorized that the start of the school year, notable for uncertainty related to school openings, would exacerbate challenges in balancing worklife commitments, particularly for pediatric faculty who were caregivers of young children.

All primary and affiliated faculty of the Department of Pediatrics at our institution were eligible for this study. Starting from Survey 2, this included affiliated faculty who held appointments in 2 different departments (eg, pediatric surgeons with appointments in pediatrics and surgery). Eligible participants were defined from a department list of currently employed faculty. Because of new faculty hires, Survey 3 may have included faculty who were not employed at our institution during the COVID-19 surge. We defined pediatric faculty as participants who self-indicated they held a rank of “Instructor” or above because instructors are considered faculty at our institution.

Survey invitations were sent to a faculty list via e-mail with a cover letter in the body of the e-mail. The cover letter explained the purpose of the survey (to monitor and improve physician wellbeing), clarified that the survey participants would be anonymous, and was signed by the department chair and vice-chair for faculty development.

The survey was administered with an Internet-based survey hosting system (Mind Garden, Inc, Menlo Park, CA). The survey was accessed via an e-mail link, allowed for unique identities for each participant, and included automated scoring with no way to track individual participants. All survey questions had to be answered for participants to submit the survey.

During all 3 survey periods, faculty had one month to complete the survey. E-mail reminders were sent weekly in every survey period (3 in total) to all eligible participants. No incentives were offered for completion of the survey and participation was voluntary. Participants were able to review and change their answers before submission. Completion of the survey was considered implied consent by the participant.

This 68-question survey included 3 sections over 3 pages, including a section on demographic information, including age, race and ethnicity, sex, marital status, and responsibility for dependents. Professional information, such as faculty rank, years of experience after medical school, years at current organization, subspecialty, and employment status (full- versus part-time) were also collected. The next section included a commonly used measure for burnout evaluation in medical literature,^8,20,21  the validated Maslach Burnout Inventory-Human Services Survey (MBI-HSS).²²  This is a 22-question survey used to measure burnout in individuals working human services jobs. Finally, we examined the Areas of Worklife Survey (AWS), a 28-question validated survey, which is used to assess 6 key organizational areas of the workplace environment (worklife domains) strongly associated with burnout.¹⁷ 

The AWS includes 6 domains (workload, control, reward, community, fairness, and values). Each domain score is calculated by using level of agreement with statements evaluating the participant’s worklife and measured on a 1 (strongly disagree) to 5 (strongly agree) Likert scale. Lower scores (generally, <2.75 out of 5) suggested problems in a worklife domain. Consistent with other studies,²³  mean domain scores were used to compare and evaluate worklife domains between survey periods.

The Maslach Burnout Inventory-Human Services Survey is used to measure burnout by using 3 subscales. Emotional exhaustion (EE) measures feelings of stress or exhaustion, depersonalization (DEP) measures low enthusiasm and feelings of detachment, and personal accomplishment (PA) measures feelings of competency. Scores are calculated by using self-reported frequency of symptoms, on a 0 (never) to 6 (every day) Likert scale. Based on other studies, we focused on subscales commonly used to define burnout:⁸  high scores on EE (>3.0) and/or a high score on depersonalization (≥2.73). “High” scores were calculated by using population norms.²⁴ 

We grouped pediatric faculty as either “inpatient” or “outpatient” providers on the basis of their self-reported subspecialty. Participants who identified as critical care physicians, emergency department physicians, hospitalists, or neonatologists were considered “inpatient” providers. Because these physicians only worked in an acute hospital setting, they were identified as “frontline” workers during and after the COVID-19 surge in the survey 2 period. During the Survey 3 period, participants were also asked whether they had directly cared for COVID-19 patients between March and June 2020 as a dichotomous “yes/no” question to evaluate if exposure to COVID-19 patients contributed to burnout.

Our primary outcome was the prevalence of pediatric faculty burnout and its associated subscales (EE, DEP, and PA), comparing all 3 survey periods. Secondary outcomes included mean worklife domain scores, comparing 3 survey periods, and associations between demographics (such as frontline worker status) and burnout.

Categorical variables were analyzed by using χ² or Fisher’s exact tests to compare demographic and professional information between all 3 survey periods and between faculty demographics and survey demographics. We also used χ² analysis or Fisher’s exact test, as appropriate, to analyze burnout prevalence between survey periods, and to measure associations between certain demographics (eg, frontline worker status) and burnout. Continuous variables, such as burnout subscale scores and worklife domain scores. were analyzed by using a Student’s t test or Mann–Whitney test, as appropriate, comparing time periods as follows: Survey 1 to Survey 2, Survey 2 to Survey 3, and Survey 1 to Survey 3. Burnout subscales (EE, DEP, and PA) and worklife domain scores were reported as mean scores on the basis of previous literature. All statistical analyses were conducted by using Stata Corp 15.1 (Stata Corp, College Station, TX).

The response rate for each of the 3 surveys was 89 of 223 (40%) in Survey 1, 100 of 267 (37%) in Survey 2, and 113 of 275 (41%) in Survey 3. There were no differences in the demographics of the participants comparing survey periods (Table 1). When comparing overall faculty to survey participants, there were no differences in terms of age range or sex in Survey 1 (P = .34; P = .89; respectively), Survey 2 (P = .37; P = .89; respectively), or Survey 3 (P = .35; P = .28; respectively). In all 3 time periods, associate and full professors were more likely to respond to the survey (Survey 1: P = .01; Survey 2: P < .001; Survey 3: P = .007; data not shown).

In all 3 survey periods, the frequency of burnout ranged from 20% to 26% of participants (P > .05, all comparisons by χ² test), and inpatient (frontline) workers were not more likely to report burnout (P > .05 in Survey 2 and Survey 3; Fig 2). Fig 3 reveals changes in PA, DEP, and EE over the 3 survey time periods. Although faculty deployment occurred between survey 1 to survey 2, there was a decrease in EE, with the mean EE score decreasing from 2.25 (1.1) to 1.9 (1.2) from Survey 1 to Survey 2 (P = .04; Fig 3). The EE score remained lower (2.07) in Survey 3 compared with that of Survey 1 (P = .22), although not different compared with that of the previous time periods. There was no change in DEP or PA when comparing survey periods. Burnout frequency and changes in burnout subscales were unchanged when reanalyzing our data with primary pediatric faculty (ie, without affiliated faculty such as pediatric surgeons) in surveys 2 and 3. Participants self-reporting direct care for COVID-19 patients during the surge also did not have an increased likelihood of burnout (P = .48; data not shown).

Fig 4 reveals changes in the 6 worklife domains over the 3 time periods. There was an improvement in workload, when comparing Survey 1 to Survey 2 (mean: 2.5 [0.72] to 2.95 [0.78]) and survey 1 to Survey 3 (mean: 2.5 [0.72] to 2.8 [0.77]; both P < .01), suggesting participants had less problems with workload in later survey periods. There was also an improvement in the fairness domain comparing Survey 1 to Survey 2 (mean: 3.1 [0.71] to 3.3 [0.69]; Fig 4) and Survey 1 to Survey 3 (mean: 3.1 [0.71] to 3.3 [0.72]; both P = .03; Fig 4). These findings remained unchanged when reanalyzing our data with primary pediatric faculty. When separating faculty into inpatient (frontline) and outpatient providers, we still found improvements in the workload domain in inpatient faculty when comparing Survey 1 to Survey 2 (mean: 2.7 [0.75] to 3.1 [0.78]; P = .03). Outpatient faculty had improvements in the workload domain comparing Survey 1 to Survey 2 (mean: 2.5 [0.71] to 2.9 [0.77]; P = .006) and survey 1 to survey 3 (mean: 2.5 [0.71] to 2.80 [0.78]; P = .015).

We examined pediatric faculty burnout, and factors associated with burnout, using serial cross-sectional surveys in an early US epicenter of the COVID-19 pandemic over 3 time periods. There were no changes in burnout prevalence during the first 7 months of the pandemic. Contrary to our hypothesis, we found that EE decreased during the COVID-19 surge and did not increase in the postsurge period. Faculty perceptions of workload and fairness improved in the surge and post-surge survey periods. To our knowledge, this is the first study to explore the short-term impact of the COVID-19 pandemic on burnout and worklife in pediatric faculty deployed to care for COVID-19 patients. These findings may be helpful to other institutions who need to deploy pediatric faculty and require strategies to support pediatric faculty during another wave of the pandemic or during other future critical events. Strengths of this study include the ability to directly compare pediatric faculty burnout data during the height of the COVID-19 surge with recent pre- and postsurge data in an academic institution located in a COVID-19 epicenter, and the use of validated survey instruments.

We found burnout rates ranging from 20% to 26% of our pediatric faculty, which is lower than previously reported baseline national rates of pediatrician burnout (30% to 49%)^25–27  and is comparable with that of other studies exploring health care worker burnout during the COVID-19 pandemic.^20,28–37  Our findings of improved EE (the subscale most related to burnout⁸ ) during the COVID-19 surge may reflect a standard psychological response to a disaster. The “phases of disaster” theory is used to describe 6 stages (predisaster, impact, heroic, honeymoon, disillusionment, and reconstruction) characterized by predictable sequences of individual and community emotional responses to a disaster over time, which may occur over months to years.³⁸  The “heroic” stage is characterized by feelings of adrenaline and altruism and occurs soon after the initial impact of a disaster. A subsequent “honeymoon” or “community cohesion” stage then occurs when survivors (in this case, physicians) settle into the new routine created by the disaster and have a stronger sense of connection to each other because of shared experiences.

The timing of Survey 2 (immediately after the peak of the COVID-19 surge; Fig 1) may have captured this “honeymoon” stage, whereas Survey 3 may have captured the beginning of a “disillusionment phase.” In previous studies, researchers indicate EE and burnout in health care workers may continue years after an inciting event, such as a pandemic or earthquake.^15,16,39  In one study by Mattei et al,¹⁵  direct exposure to an earthquake in L’Aquila, Italy, was a significant predictor of burnout in health care workers, as long as 6 years after the event. With our findings, we support theories that long-term monitoring of physician wellbeing may be critical to ensure a healthy physician workforce in areas severely affected by the pandemic⁴⁰  and that studying individual and organizational responses to traumatic events can create opportunities for posttraumatic growth.⁴¹ 

We did not find changes in the majority of AWS domains (control, reward, values, and community) comparing 3 different survey periods, suggesting there may be a ceiling effect with these measures. However, faculty perceptions of workload improved during the surge and postsurge, despite faculty deployment. Our institution employed strategies on the basis of a theoretical framework to mitigate burnout development,⁴²  which may have improved faculty perceptions of workload. Deliberate scheduling ensured no faculty worked over 12 hours in a single shift and had enough time off between shifts.^38,43  The creation of an adult unit within our children’s hospital, rather than deploying pediatricians to unfamiliar adult units, staff, and workflows may have eased the burden of unfamiliarity. After the surge, diminished inpatient and outpatient patient volumes may have also contributed to reported improvements. Our findings of improved perceptions of workload is consistent with that of other studies in which researchers measure AWS sequentially. Gregory et al⁴⁴  examined an intervention in primary care clinics that was associated with improvements in physician workload, EE, and DEP. Associations between burnout subscales and AWS, especially in different settings, may also be particularly important to explore in further studies.

We also noted an improvement in the worklife domain of fairness. The domain of “fairness” is used to examine participants’ perceptions on appropriate allocation of resources and fairness of decision-making by leadership. We theorized frequent and clear communication from administrative leadership may have helped faculty understand how and why decisions were made. We ensured visible and engaged leadership, with the department leadership meeting with each pediatric division between the Survey 2 and Survey 3 periods to understand unique challenges faculty faced.⁴⁵  Prioritizing frequent and timely communication with all faculty may have helped the perception that resources and responsibilities were distributed as fairly as possible.

Although we did not find significant improvements in the worklife domain of community, this may have been due to a ceiling effect because it was consistently the highest mean worklife score in all time periods (4.0–4.2 of 5). Continued community-building was an important strategy during and after the surge because peer support may mitigate burnout.²⁷  Spaces were created for frontline staff to take a break together during shifts. Virtual faculty support calls were offered on a regular basis.⁴⁶  Mental health resources were shared, and psychiatry department colleagues offered their services to anyone requesting assistance. In July 2020 (before Survey 3 administration), “parent load” virtual support groups were created to help faculty with dependents <18 years of age cope with the challenges of child care and the uncertainties of school reopening. Celebrations of annual events, such as trainee graduations and faculty retirements, occurred virtually and focused on our community’s strengths. Strengthening feelings of community and support both virtually and in-person (for frontline workers) may have helped create feelings of cohesion and connection needed to mitigate stress from the pandemic surge.

There were several limitations to this study. The response rate was suboptimal, although similar to other physician burnout surveys,⁴⁷  and may not fully capture the true extent of burnout, although demographics of the survey population and our faculty were generally similar. Senior faculty were more likely to respond to the survey, which may have impacted our findings because older participants have been shown to be less likely to be burned out.⁸  However, our survey demographics remained similar in all survey periods, and we did not note any changes in burnout between survey periods. We used a cross-sectional survey, so we can only infer associations, not causal relationships. We did not track individual responses over time because the survey was anonymous, and study participants may have only responded in some time periods and not in others. Some participants may have been misclassified as “outpatient” or “nonfrontline” respondents in some cases because we were unable to specifically identify and analyze those outpatient providers who became “frontline” workers during the pandemic surge, which may impact our findings. This study may be subject to response bias, with busier frontline faculty choosing not to respond. Social desirability bias may also have played a role, despite the confidential nature of this study. We did not measure for physician stress or anxiety, which may impact the development of physician burnout and have negative effects on wellbeing. This study was conducted at a single institution, limiting the generalizability to our findings. We were unable to perform a multivariate analysis because of our sample size. Finally, although there was no difference in burnout rates between the 3 survey periods, the results may be subject to a type II error. On the basis of our limited sample size, we were only powered to detect a difference in burnout of 16%. However, despite the limited sample size, we were still able to detect changes in areas of worklife (eg, workload and fairness) that may prevent burnout.

Physician burnout remains a critical public health issue, particularly in the face of the COVID-19 pandemic. Our institution, located in an early epicenter of the pandemic and requiring faculty deployment to deal with a surge in COVID-19 patients, customized strategies to mitigate physician burnout. The implementation of these strategies may have initially improved pediatric faculty EE and worklife. These approaches may be helpful for other institutions facing surges of COVID-19 patients and pediatric faculty deployment. However, the full impact of the COVID-19 surge on pediatric faculty burnout may not yet be apparent only 4 months after the initial surge event. Future studies are needed to examine continued strategies to mitigate burnout and examine the long-term consequences of the COVID-19 pandemic and deployment on physician wellbeing.

We thank the pediatric housestaff, nurses, staff, and faculty at the Children’s Hospital at Montefiore for their hard work and dedication during the COVID-19 surge and beyond.