Prevalence of burnout and its relation to the neuroendocrine system among pediatric residents during the early Covid-19 pandemic: A pilot feasibility study

Abstract

Background

Measuring burnout relies on infrequent and subjective surveys, which often do not reflect the underlying factors or biological mechanisms that promote or prevent it. Burnout correlates with cortisol levels and dysregulation of the hypothalamic-pituitary-adrenal axis, but the chronology and strength of this relationship are unknown.

Objective

To determine the prevalence and feasibility of studying burnout in pediatric residents using hair cortisol and hair oxytocin concentrations.

Design

/Methods: Longitudinal observational cohort study of pediatric residents. We assessed burnout using the Stanford Professional Fulfillment Index and hair cortisol (HCC), and hair oxytocin concentrations (HOC) at four 3-month intervals from January 2020–January 2021. We evaluated test-retest reliability, sensitivity to change using Pearson product-moment correlations, and relationships between burnout and hair biomarkers using hierarchical mixed-effects linear regression.

Results

17 Pediatrics residents provided 78 wellness surveys and 54 hair samples. Burnout symptoms were present in 39 (50%) of the surveys, with 14 (82%) residents reporting burnout in at least one time point. The lowest (41%) and highest (60%) burnout prevalence occurred in 04/2020 and 01/2021, respectively. No significant associations between burnout scores and HCC (β −0.01, 95%CI: 0.14–0.13), HOC (β 0.06, 95%CI: 0.06–0.19), or the HCC:HOC ratio (β −0.04, 95%CI: 0.09–0.02) were noted in separate analyses. Intra-individual changes in hair cortisol concentration were not associated with changes in burnout score.

Conclusions

Burnout was prevalent among Pediatrics residents, with highest prevalence noted in January 2021. This pilot longitudinal study demonstrates the feasibility of evaluating burnout with stress and resilience biomarkers in Pediatrics residents.

Highlights

• •Burnout is prevalent among pediatric resident physicians.
• •Longitudinal evaluation of burnout in combination with hair samples is feasible.
• •Hair cortisol and oxytocin concentrations have complex relationships with burnout.

1. Introduction

Professional burnout affects up to 40% of the 100,000 US pediatricians at any given time, with peak prevalence during residency training, [1] with adverse consequences for quality of care and patient outcomes [2,3]. However, measuring burnout relies on infrequent and subjective surveys, resulting in an inability to identify the underlying factors or biological mechanisms that promote or prevent professional burnout.

Cortisol is a key stress hormone that regulates the autonomic system to alter physiological stress responses [4]. Cortisol levels fluctuate in saliva or serum with acute stress, but cortisol also binds to growing hair reflecting a non-specific cumulative or chronic stress. Hair cortisol concentrations (HCC) are thus widely accepted as a measure of chronic stress over 3–6 months [5]. Stress-induced cortisol release can be buffered by positive social experiences and at least partially mediated by oxytocin release [6]. Similarly to cortisol, oxytocin levels fluctuate acutely in serum and saliva, but hair oxytocin concentrations (HOC) can index the supportive and positive social experiences of an individual. Diurnal variations in serum cortisol (peaks after waking) and oxytocin (peaks during sleep) may also be affected by sleep deprivation among healthcare providers, further underscoring the utility of HCC and HOC as markers of chronic stress, particularly when used to evaluate other chronic markers of dysregulated stress such as burnout [7].

Burnout status correlates with cortisol levels and dysregulation of the hypothalamic-pituitary-adrenal axis, but the chronology and strength of this relationship are unknown [8,9]. This study sought to establish the prevalence of burnout among pediatric residents during the early pandemic and the feasibility of assessing HCC and HOC as longitudinal and objective measures of burnout.

2. Materials and methods

We report a longitudinal observational cohort study of pediatric residents, following approval from the Stanford Institutional Review Board and informed written consent. We followed the CONSORT reporting guidelines for pilot feasibility trials [10].

2.1. Eligibility and recruitment

We recruited resident physicians in Pediatrics from a large quaternary residency program. Residents were eligible if they were not anticipated to graduate during the study period. Residents with natural hair of all types were eligible to enroll, but were excluded if hair had been bleached, dyed, or permanently straightened, was shorter than 3 cm at the posterior vertex, or if they had used systemic glucocorticoids in the prior month.

2.2. Hair sample collection and analysis

We assessed hair cortisol and hair oxytocin concentrations at four intervals, including January 2020, July 2020, October 2020, and January 2021. Hair samples were not collected in April 2020 due to restrictions related to the COVID-19 pandemic, but distal hair collected in July 2020 was used a surrogate for the April 2020 hair sample when available.

Approximately 50 strands of hair were clipped from the posterior vertex region, and retained in a −20 °C freezer until analyzed. A 10 mg sample of hair from the proximal 3 cm was cut to powder consistency, followed by four extraction phases alternating 1 mL of methanol (52 °C, 15 h, 200 rpm) with 1 mL of acetone extraction (25 °C, 5 min, 200 rpm); each repeated twice. Supernatants from each sample were pooled for air evaporation at 4 °C, reconstituted in phosphate-buffered saline (PBS, pH 7.6) according to hair weight (70μL/10 mg hair). Total protein content was determined using the Take 3, Gen 5 spectrophotometer (BioTeK plate reader, Winooski, VT). HCC values were determined via Cortisol ELISA assay kit (ALPCO Diagnostics, Salem, NH) and HOC values via Oxytocin ELISA assay kit (Enzo Life Sciences, Farmingdale, NY), using the Epoch BioTek plate reader for all spectrophotometer readings.

In our lab, HCC and HOC assays are reliable and reproducible, with intra-assay coefficients of variation <4% and <6% for HCC and HOC, respectively. Inter-assay coefficients of variation are <8% for both HCC and HOC [11]. Detection thresholds are 4 ng/ml for HCC and 15 pg/ml for HOC. Our detection of HOC has also been validated using both mass spectroscopy and a highly specific aptamer for oxytocin [11]. We evaluated HCC as nanograms per milligram of hair and HOC as picograms per milligram of hair. For cortisol:oxytocin ratios, we first converted HCC and HOC to millimoles per milligram of hair to account for differing molecular weights.

2.3. Burnout assessment

In line with prior longitudinal burnout research among trainees, we assessed burnout symptoms at five 3-month intervals from January 2020 to January 2021, coincident with the hair sample collections [12]. We used the Stanford Professional Fulfillment Index, which includes 10 prompts assessing burnout symptoms, 4 related to work exhaustion and 6 related to interpersonal disengagement. In line with prior research [13], we calculated the mean score for these 10 burnout items, which are scored 0–4 on a Likert scale. A score of ≥1.33 was considered indicative of burnout symptoms, though we retained burnout as a continuous variable in all analyses. In exploratory analyses, we also considered the outcomes of emotional exhaustion (EE) score and depersonalization (DP) score separately.

2.4. Statistical approach

To test reliability, we calculated Pearson product-moment correlations of HCC, HOC, and cortisol:oxytocin ratios with their values for adjacent assessments among participants with <0.4-point (10%) change in burnout score during the same time period. To test sensitivity to change, we evaluated HCC, HOC, and HCC:HOC molar ratios using hierarchical mixed-effects linear regression models with participant as random intercept and burnout score as independent variable, leveraging the temporal variations in stress states inherent to monthly residency rotations. For example, inpatient rotations often feature long hours and emotionally charged situations which may exacerbate chronic stress, while vacation may provide an opportunity to relieve stress. In secondary exploratory analyses, we also (1) included gender and most recent rotation (categorized as ICU, acute inpatient, outpatient, or vacation) as independent variables, and (2) evaluated the outcomes of EE and DP scores separately.

Categorical reference values were set at the modal value. Due to the hypothesis-generating nature of these analyses, we made no corrections for multiple testing. All statistical analyses were performed in Stata version 17.0.

3. Results

3.1. Recruitment and study population

We advertised the study to 63 eligible pediatrics residents, screened 19 residents for eligibility, and enrolled 17 residents (27% enrollment rate) in this pilot study during a 9-day enrollment period, as outlined in Supplemental Table 1. Residents were mostly female (11/17, 65%); most identified as White (10/17, 59%), with others identifying as Asian (4/17, 24%), multiple races (1/17, 6%), or not reported (2/17, 12%). Each participant received 5 wellness surveys, with a total of 78/85 (92%) surveys returned. Fourteen residents completed all 5 surveys, one participant completed 4/5 (80%) surveys, and two participants completed 2/5 (40%) surveys. Burnout symptoms were present in 39 (50%) of the wellness surveys, with 14 (82%) residents reporting symptoms of burnout at least once. April 2020 had the lowest burnout prevalence (41%), while January 2021 had the highest burnout prevalence (60%), as shown in Fig. 1.

Fig. 1.

Average burnout scores (0–4 scale) during 5 study timepoints. N = 17 pediatrics residents.

Each participant was also invited to give 4 hair samples, with a total of 49/68 samples (72%) collected. Because hair grows at approximately 1 cm/month and measurements from 3 cm of sequential hair segments have been validated to represent HCC values at 3-month intervals [14], 13 of the missed hair sample collections were extrapolated from distal segments of hair (6 from missed sample collections, and 7 from the April 2020 timepoint during which COVID-19 restrictions prevented sample collection), for a total of 62 useable hair samples out of a potential 85 (73%). Most common reasons for not obtaining useable hair samples were COVID-19 restrictions (10/23, 43%), unable to schedule due to away rotation or vacation (9/23, 39%), or unable to schedule for other reasons (4/23, 17%). All hair samples were adequate for measurement of HCC, but 4/62 (6%) had insufficient volume for accurate HOC measurement. Individual EE, DP, HCC, and HOC trajectories are shown in Supplemental Fig. 1.

3.2. Reliability

For adjacent time periods with a <10% absolute change in burnout score (28/45, 62%), HCC correlation was 0.52 (P = .005), HOC correlation was 0.30 (P = .15), and HCC:HOC molar ratio correlation was 0.26 (P = .25).

3.3. Sensitivity to change

In mixed effect regressions with participant as random intercept, burnout score did not associate with HCC (β 0.03, 95%CI -0.11 to 0.17), HOC (β 0.06, 95%CI -0.06 to 0.19), or HCC:HOC ratio (β −0.01, 95%CI -0.08 to 0.06).

3.4. Secondary analyses

In 3 separate multivariable analyses adjusted for most recent rotation and gender, burnout score did not associate with HCC (β 0.02, 95% CI -0.12 to 0.16), HOC (β 0.06, 95% CI -0.06 to 0.19), or HCC:HOC molar ratio (β −0.01, 95% CI -0.08 to 0.06), as shown in Table 1.

Table 1.

Hair cortisol, oxytocin, and cortisol:oxytocin ratio in relation to burnout score. Mixed effects linear regressions: dependent variable is burnout score (0–4 scale). N = 62 samples from 17 pediatrics residents (resident as random effect).

	β (95% CI)	P
Model 1: Cortisol
Intercept	1.33 (1.05–1.61)
Hair cortisol concentration (ln)	0.02 (−0.12–0.16)	.78
Rotation (vs. outpatient)
ICU	0.13 (−0.23–0.48)	.49
Inpatient	−0.01 (−0.27–0.26)	.97
Vacation	−0.26 (−0.70–0.19)	.26
Male sex (vs. female)	−0.04 (−0.46–0.39)	.87
Model 2: Oxytocin
Intercept	1.33 (1.06–1.60)
Hair oxytocin concentration (ln)	0.06 (−0.06–0.19)	.34
Rotation (vs. outpatient)
ICU	0.14 (−0.30–0.59)	.52
Inpatient	−0.01 (−0.27–0.25)	.94
Vacation	−0.30 (−0.74–0.14)	.18
Male sex (vs. female)	−0.06 (−0.45–0.34)	.79
Model 3: Cortisol to oxytocin ratio
Intercept	1.35 (1.02–1.68)
Hair cortisol:oxytocin ratio	−0.01 (−0.08–0.06)	.76
Rotation (vs. outpatient)
ICU	0.16 (−0.30–0.61)	.50
Inpatient	−0.00 (−0.29–0.28)	.98
Vacation	−0.29 (−0.75v0.17)	.21
Male sex (vs. female)	−0.07 (−0.49–0.35)	.75

Burnout score had a≥10% absolute change in 17/45 (38%) of adjacent timepoints. Hair cortisol reactivity was not associated with burnout reactivity, with change in HCC between adjacent time points associated with no significant difference in burnout score between adjacent time points (β 0.06, 95% CI -0.09 to 0.21), as illustrated in Supplemental Fig. 2.

HCC, HOC, and HCC:HOC ratios did not associate with EE scores or DP scores, although HCC showed positive point estimates for EE (β 0.09, 95% CI -0.05 to 0.23), but negative point estimates for DP (β −0.03, 95% CI -0.19 to 0.13).

4. Discussion

We found high prevalence of burnout among pediatric resident physicians throughout the early phase of the COVID-19 pandemic comparable to rates reported previously [1]. We also found that recruiting and retaining pediatrics residents for longitudinal evaluation of burnout symptoms and hair biomarkers is feasible, with a 27% enrollment rate within a 9-day enrollment period, and 82% retention throughout the study. In this pilot cohort, burnout scores were not associated with HCC, HOC, or HCC:HOC molar ratios.

HCC exhibited strong autocorrelation across time periods in which participants had minimal change in burnout symptoms, suggesting its reliability as a biomarker of static burnout state. In contrast, HOC and the HCC:HOC ratio only exhibited moderate autocorrelation, which may suggest more complex relationships between oxytocin and burnout symptoms. Further, sensitivity to change was low for both HCC and HOC, with point estimates near zero for burnout score associations with each log increase in HCC or HOC. This low sensitivity to change may reflect the complexity of the relationships among stress, dysregulated stress, and burnout,¹² or may suggest that the chronology of burnout symptoms differs from the 3-month lead time of HCC which was evaluated.

Begin et al. [15] recently suggested a possible positive association between HCC and EE, and a negative association between HCC and DP, with similarly non-significant findings when accounting for clustering within subjects. Our study replicates that pattern, further suggesting that cortisol release may be more closely tied to emotionally stressful situations, but that disengagement and depersonalization may correspond to a blunted physiologic response. Staples et al. reported PGY1 residents with burnout had lower training milestones scores in patient care, systems-based practice, practice-based learning/improvement, professionalism, and interpersonal and communication skills, but not medical knowledge. [16] Interestingly, lower patient care scores were associated with higher DP and low personal accomplishment scores in PGY1 residents [16]. Although correlates of burnout in pediatric residents have shown differences by race and gender in previous studies, [17]) we did not observe such associations in our pilot study—a larger prospective trial would be needed to more fully evaluate these relationships.

This study must be interpreted in light of its design. As a pilot feasibility study, it was not adequately powered to detect small associations among HCC, HOC, and burnout symptoms. Participants were all exposed to similar work environments, were self-selected, and 3 of them did not complete all timepoints, which may result in some response bias. Although the baseline assessment was prior to the local pandemic response, the remainder of this longitudinal study spanned the first year of the US COVID-19 pandemic response. Contextual stressors related to this exogenous shock may have introduced larger than typical variations in stress states among trainee physicians. Although we used the well-validated Stanford Professional Fulfillment Index to assess burnout symptoms in pediatric residents, our findings may not be directly comparable to those of other studies that may use a variety of other survey instruments. [1,3,13].

5. Conclusion

Burnout is prevalent among pediatrics residents, with highest burnout apparent by January 2021. Longitudinal assessment of burnout and hair biomarkers is feasible among a cohort of resident physicians. Studies with larger sample sizes will be required to ascertain whether hair cortisol and hair oxytocin are valid physiological measures relevant for the study of burnout in trainee physicians.

Funding

This work was supported by the Agency for Healthcare Research and Quality [K08 HS027837].

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: